Cpi Manual Concepts and Methods

CONSUMER PRICE
INDEX MANUAL
Concepts and Methods | 2020
International Monetary Fund | International Labour Organization

Statistical Office of the European Union (Eurostat)
United Nations Economic Commission for Europe
Organisation for Economic Co-operation and Development | The World Bank
CONSUMER PRICE
INDEX MANUAL
© 2020 International Labour Organization/International Monetary Fund/Organisation for Economic
Co-operation and Development/European Union/United Nations/The World Bank
Cataloging-in-Publication Data
IMF Library
Names: International Labour Office. | International Monetary Fund. | Organisation for Economic Co-operation
and Development. | European Union. | United Nations. | World Bank.
Title: Consumer price index manual : concepts and methods.
Description: Washington, DC : International Monetary Fund, [2020]. | Includes bibliographical references.
Identifiers: ISBN 978-1-48435-484-1 (Paper)
978-1-51354-299-7 (ePub)
978-1-51354-298-0 (PDF)
Subjects: LCSH: Consumer price indexes—Methodology. | Price indexes.
Classification: LCC HB225.C66 2020
Eurostat Catalogue number: OA-04-20-634-EN-C (Paper)

Eurostat Catalogue number: OA-04-20-634-EN-N (PDF)
Disclaimer: The views expressed in this book belong solely to the authors. Nothing contained in this book
should be reported as representing the views of the organizations, member governments, or any other entity
mentioned herein.
Please send orders to:

International Monetary Fund, Publication Services
P.O. Box 92780, Washington, D.C. 20090, U.S.A.
Tel.: (202) 623-7430 Fax: (202) 623-7201
E-mail: [email protected]
Internet: www.elibrary.imf.org
www.bookstore.imf.org
CONTENTS
Foreword ........................................................................................................................................................................ ix
Preface ........................................................................................................................................................................... xi
Acknowledgments ........................................................................................................................................................... xv
Acronyms and Abbreviations ......................................................................................................................................... xvii
1 Introduction, Overview, and Basic Steps for the Consumer Price Index Development ................................. 1
Introduction ............................................................................................................................................................. 1
Developing the Consumer Price Index .................................................................................................................... 1
Overview of the Consumer Price Index Uses and Needs ........................................................................................ 1
Overview of the Consumer Price Index Concepts . ................................................................................................. 2
Deciding on the Index Coverage and Classification Structure ................................................................................ 6
Deriving the Weighting Pattern ............................................................................................................................... 7
Designing the Sample .............................................................................................................................................. 10
Collecting and Editing the Prices ............................................................................................................................ 12
Maintaining and Updating the Sample .................................................................................................................... 14
Missing Products and Adjusting for Changes in Quality ........................................................................................ 14
New Products . ......................................................................................................................................................... 19
Calculating the Consumer Price Index .................................................................................................................... 21
Updating the Weights and Linking of Series ........................................................................................................... 24
Organization and Management . .............................................................................................................................. 25
Publication and Dissemination ................................................................................................................................ 25
Special Cases ........................................................................................................................................................... 26
Errors and Biases ..................................................................................................................................................... 26
Annex 1.1 Formula Notations ................................................................................................................................. 27
2 Uses, Concepts, Scope, and Classifications of Consumer Price Indices . .......................................................... 29

Introduction ............................................................................................................................................................. 29
Consumer Price Index Uses . ................................................................................................................................... 29
Concepts and Scope of the Consumer Price Index . ................................................................................................ 34
Consumer Price Index Classifications ..................................................................................................................... 48
Key Recommendations . .......................................................................................................................................... 50
Annex 2.1 Use of Price Statistics in the National Accounts—Supply and Use ...................................................... 51
3 Expenditure Weights and Their Sources ............................................................................................................. 53

Introduction ............................................................................................................................................................. 53
Conceptual Basis of the Weights ............................................................................................................................. 53
The Weighting Structure of the Consumer Price Index ........................................................................................... 55
Data Sources ............................................................................................................................................................ 57
Deriving the Weights in Practice ............................................................................................................................. 59
Weight Reference Period ......................................................................................................................................... 61
Frequency of Weight Updates ................................................................................................................................. 61
Items Requiring Special Treatment ......................................................................................................................... 62
4 Sampling ................................................................................................................................................................. 65
Introduction ............................................................................................................................................................. 65
Sampling Techniques . ............................................................................................................................................. 65
Sampling Stages in a Consumer Price Index . ......................................................................................................... 69
Central Price Collection . ......................................................................................................................................... 72
Sample Maintenance . .............................................................................................................................................. 73
Variance Estimation and Optimal Allocation .......................................................................................................... 74
iii
Contents
5 Price Collection and Validation ............................................................................................................................ 77

Introduction ............................................................................................................................................................. 77
Organization Options . ............................................................................................................................................. 77
The Principles of Price Collection . ......................................................................................................................... 78
The Practical Aspects of Managing Price Collection .............................................................................................. 86
Data Validation and Editing . ................................................................................................................................... 93
Price Collector Training .......................................................................................................................................... 98
Documentation: Work Instructions .......................................................................................................................... 99
Disaster Recovery . .................................................................................................................................................. 99
Other Methods of Price Collection .......................................................................................................................... 100
Annex 5.1 Consumer Price Index Price Collection Procedures .............................................................................. 105
Annex 5.2 Consumer Price Index—Example of a Price Collection Form .............................................................. 107
Annex 5.3 Consumer Price Index—Automated Data Checking ............................................................................. 109
Annex 5.4 Documentation Control Template .......................................................................................................... 113
Annex 5.5 The Calculation of Average Product Price When Combining Prices from Different Price
Collection Methods and for Different Price Collection Frequencies ................................................................ 115
Annex 5.6 Web Scraping ......................................................................................................................................... 117
6 Temporarily and Permanently Missing Prices and Quality Change ................................................................ 121
Introduction ............................................................................................................................................................. 121
Background . ............................................................................................................................................................ 121
Potential Errors in the Matched-Model Method . .................................................................................................... 122
Useful Concepts for the Treatment of Missing Prices ............................................................................................. 123
The Treatment of Temporarily and Permanently Missing Variety Prices . .............................................................. 125
The Treatment of Temporarily Missing Price Observations . .................................................................................. 126
The Nature of Quality Change . ............................................................................................................................... 130
Permanently Missing Price Observations . .............................................................................................................. 131
Implicit Methods of Quality Adjustment . ............................................................................................................... 132
Explicit Methods of Quality Adjustment . ............................................................................................................... 136
High Technology and Other Sectors with a Rapid Turnover of Models ................................................................. 148
Annex 6.1 Overall Mean (or Targeted) Imputation ................................................................................................. 155
Annex 6.2 Quality Adjustment Using a Replacement and Price Overlap ............................................................... 157
Annex 6.3 The Nature and Extent of the Index Number Bias If Only Matched Varieties Are Used ...................... 159
7 Maintaining the Sample ........................................................................................................................................ 161

Introduction ............................................................................................................................................................. 161
Sample Maintenance and Matching ........................................................................................................................ 161
Incorporation of New Products . .............................................................................................................................. 163
Information Requirements Maintaining the Sample ............................................................................................... 172
8 Calculating Consumer Price Indices in Practice ................................................................................................ 175

Introduction ............................................................................................................................................................. 175
The Calculation of Price Indices for Elementary Aggregates ................................................................................. 175
The Calculation of Higher-Level Indices ................................................................................................................ 190
9 Updating CPI Weights and Linking New to Previous CPI Series . ................................................................... 199
Introduction ............................................................................................................................................................. 199
Calculating a Chain Index ....................................................................................................................................... 199
Updating Weights for Price Change: Pros and Cons ............................................................................................... 200
Detailed Methods for Updating Weights ................................................................................................................. 202
Linking the Previous CPI to the New Price Index Reference Period ...................................................................... 208
Frequency of Weight Updates ................................................................................................................................. 214
Annual Updating and Linking ................................................................................................................................. 215
Introducing New Classification Systems ................................................................................................................. 218
Expanding CPI Geographic Coverage . ................................................................................................................... 219
iv
Contents
10 Scanner Data .......................................................................................................................................................... 221

Introduction ............................................................................................................................................................. 221
Practical Considerations .......................................................................................................................................... 221
Multilateral Price Index Methods ............................................................................................................................ 226
11 Selected Special Cases ........................................................................................................................................... 235

Introduction ............................................................................................................................................................. 235
Seasonal Products .................................................................................................................................................... 235
Internet Purchases . .................................................................................................................................................. 241
Housing . .................................................................................................................................................................. 245
Second-Hand Goods ................................................................................................................................................ 256
Own-Account Production ........................................................................................................................................ 259
Tariffs . ..................................................................................................................................................................... 263
Telecommunications ................................................................................................................................................ 269
Transport Services ................................................................................................................................................... 272
Health, Education, and Social Protection Services . ................................................................................................ 274
Financial Services . .................................................................................................................................................. 279
Annex 11.1 Example of Price Collection Checklist for Second-Hand Clothing . ................................................... 289
Annex 11.2 Example Price Collection Letter to Retailer ........................................................................................ 291
Annex 11.3 Calculation of a Price Index for a Deposit Product ............................................................................. 293
12 Errors and Bias ...................................................................................................................................................... 297

Introduction ............................................................................................................................................................. 297
Types of Errors ........................................................................................................................................................ 297
Measuring Error . ..................................................................................................................................................... 298
Types of Bias ........................................................................................................................................................... 301
13 Data Quality Management and Reporting .......................................................................................................... 307

Introduction ............................................................................................................................................................. 307
Data Quality Assessment Framework (DQAF) for a Consumer Price Index . ........................................................ 307
Quality Management ............................................................................................................................................... 308
Quality Management Systems ................................................................................................................................. 308
Prototype of a Quality Management System ........................................................................................................... 310
Documentation ........................................................................................................................................................ 311
Internal and External Audits of Production Processes . ........................................................................................... 312
Quality Reporting and Improving the Consumer Price Index: Frameworks . ......................................................... 314
Work Programs: Programming, Planning, and Reporting ....................................................................................... 316
Annex 13.1 Data Quality Assessment Framework (DQAF) for the Consumer Price Index ................................... 317
Annex 13.2 Documentation Control Template ........................................................................................................ 319
Annex 13.3 Pro Forma for an Audit Schedule ........................................................................................................ 321
Annex 13.4 Audit Report Template ......................................................................................................................... 323
Annex 13.5 Model Quality Report Document for the Consumer Price Index ........................................................ 325
14 Publication, Dissemination, and User Relations ................................................................................................. 327

Introduction ............................................................................................................................................................. 327
Time-Series Presentation of Level and Change . ..................................................................................................... 327
Seasonal Adjustment and Smoothing of the Index .................................................................................................. 328
Analysis of Contributions to Change ...................................................................................................................... 328
Economic Commentary and Interpretation of the Index ......................................................................................... 328
Presentation of Related or Alternative Measures .................................................................................................... 329
Press Release, Bulletin, and Methodological Statement ......................................................................................... 330
International Standards Concerning the Dissemination of Consumer Price Indices . ............................................. 331
Timing of Dissemination of the Consumer Price Index .......................................................................................... 331
Timeliness of Release versus Data Accuracy .......................................................................................................... 332
Access to Data ......................................................................................................................................................... 332
Confidentiality ......................................................................................................................................................... 332
Presentation of Methodology .................................................................................................................................. 332
Explaining Index Quality . ....................................................................................................................................... 333
User Consultation .................................................................................................................................................... 333
v
Contents
A Glossary of Main Terms .................................................................................................................................... 335

Appendix 1 The Harmonised Index of Consumer Prices (European Union) . ................................................. 343
Appendix 2 Classification of Individual Consumption According to Purpose 1999 (COICOP 1999) ........... 347
Appendix 3 Classification of Individual Consumption According to Purpose 2018 (COICOP 2018) ........... 361
Appendix 4 R
esolution Concerning Consumer Price Indices Adopted by the Seventeenth International
Conference of Labour Statisticians, 2003 ........................................................................................ 427
Appendix 5 S
patial Comparisons of Consumer Prices, Purchasing Power Parities, and the International
Comparison Program ....................................................................................................................... 443
Appendix 6 Some Basic Index Number Formulas and Terminology . .............................................................. 453
Appendix 7 Consumer Price Index Research Agenda . ...................................................................................... 455
Bibliography . ......................................................................................................................................................... 461
Index ....................................................................................................................................................................... 477
Box
1.1 Classification of Individual Consumption According to Purpose (COICOP)................................................... 7
Figures
5.1 Price Changes in Plot Chart during Sales Season ........................................................................................... 96
A5.1 Planning and Organizing Price Collection ...................................................................................................... 105
A5.2 Price Collection Form ..................................................................................................................................... 107
A5.3 Documentation Control Template ................................................................................................................... 113
6.1 Quality Adjustments for Different Sized Varieties . ........................................................................................ 137
6.2 Scatter Diagram of Price against Capacity: Washing Machine Data .............................................................. 140
6.3 Guide to Treatment of Missing Prices . ........................................................................................................... 146
A6.2 Matched-Model Price Index Bias and Pricing Strategies . .............................................................................. 159
8.1 Illustrative Aggregation Structure of a CPI . ................................................................................................... 176
9.1 The CPI Life Cycle ......................................................................................................................................... 200
11.1 The Acquisitions Approach for Owner-Occupied Housing ............................................................................ 251
11.2 Loose Specification Sampling and Tight Specification Pricing ...................................................................... 276
11A.1 Price Collection Checklist for Second-Hand Clothing ................................................................................... 289
11A.2 Price Collection Letter to Retailer..................................................................................................................... 291
13.1 An Example of a Quality Management System for CPI Data Collection ....................................................... 310
13.2 Documentation ................................................................................................................................................ 311
13A.1 Documentation Control Template ................................................................................................................... 319
13A.2 Pro Forma for an Audit Schedule .................................................................................................................... 321
13A.3 Audit Report Template .................................................................................................................................... 323
13A.4 Model Quality Report Document for the CPI ................................................................................................. 325
14.1 Example of an Illustrative CPI Press Release ................................................................................................. 330
14.2 Model Note on Methodology—To Be Included in Press Releases on CPI or on the Official Website . ......... 331
Tables
3.1 Example of Plutocratic versus Democratic Weights ....................................................................................... 55
3.2 Deriving Expenditure Weights by Region . ..................................................................................................... 56
3.3 Deriving Expenditure Weights by Region and by Outlet Type ....................................................................... 57
3.4 Treatment of Products for Which No Prices Are Collected ............................................................................ 61
3.5 Estimation of Net Expenditure Weights .......................................................................................................... 63
4.1 Systematic Sample of 3 out of 10 Outlets, Based on PPS Sampling .............................................................. 66
4.2 Cutoff Sample of 3 out of 10 Outlets .............................................................................................................. 67
4.3 Stratification by Region, Outlet Type, and Product Type . .............................................................................. 68
4.4 Different Allocation Strategies ........................................................................................................................ 75
5.1 Determining Purchaser Price When Bargaining Takes Place . ........................................................................ 83
5.2 Selected Values of C and the Proportion of Observations Flagged . ............................................................... 94
A5.1 Price Relatives Showing Movement from Previous Period (Example 1) ....................................................... 109
A5.2 Parameters and Derived Limits (Example 1) .................................................................................................. 109
vi
Contents
A5.7 Combining Prices from Different Price Collection Methods and for Different Price Collection
Frequencies . .................................................................................................................................................... 115
A5.8 Web Scraping—Typical Data Structure .......................................................................................................... 118
6.1 Example of a Replacement Variety with Overlap ........................................................................................... 123
6.2 Illustrative Variety Codes for Price Collector for Missing Values .................................................................. 125
6.3 Temporarily Missing Price Observations and Imputed Prices ........................................................................ 127
6.4 Overall Mean and Targeted Mean Imputations ............................................................................................... 128
6.5A Illustration of Treatment of Comparable Replacements ................................................................................. 131
6.5B Illustration of Treatment Using the Overlap Method, Noncomparable Replacements: Actual Preceding
Period Price ..................................................................................................................................................... 132
6.5C Illustration of Treatment Using the Overlap Method, Noncomparable Replacements: Imputed
Succeeding Period Price .................................................................................................................................. 133
6.6 Introducing a Noncomparable Replacement via an Overlap .......................................................................... 135
6.6A Introducing a Noncomparable Replacement to Illustrate Link-to-Show-No-Change .................................... 135
6.7 Example of Size, Price, and Unit Price of Bags of Flour . .............................................................................. 137
6.8 Estimated (Linear) Equation of Price against Capacity: Washing Machine Data . ......................................... 141
6.9 Illustrative Hedonic Regression Estimates for Washing Machines . ............................................................... 141
6.10A Hedonic Regression Imputation of New Variety’s Price . ............................................................................... 143
6.10B Hedonic Regression Imputation of Old Variety’s Price .................................................................................. 144
6.10C Hedonic Regression Imputation of New Variety’s Price . ............................................................................... 144
6.11 Illustration of Rapid Model Turnover ............................................................................................................. 149
6.12 Equivalences of Hedonic Approaches............................................................................................................... 152
6.13 Difference between Hedonic and Matched Indices . ....................................................................................... 153
A6.1 Example—Bias from Implicit Quality Adjustment When the (Mean) Price Change of
Quality-Adjusted New Varieties Compared with the Varieties They Are Replacing Is Assumed Not
to Change (r2 = 1.00)......................................................................................................................................... 156
7.1 Example of Sample Augmentation . ................................................................................................................ 166
7.2 Example of Introducing a New Elementary Aggregate .................................................................................. 168
7.3 Example of Introducing New Weights for Higher-Level Aggregates ............................................................. 169
8.1 Jevons and Dutot Price Indices Using Averages of Prices .............................................................................. 178
8.2 Jevons and Carli Price Indices Using Averages of Long-Term Price Relatives . ............................................ 178
8.3 Jevons and Carli Price Indices Using Chained Short-Term Price Relatives ................................................... 178
8.4A Jevons and Dutot Elementary Price Indices Using Averages with Missing Prices ......................................... 184
8.4B Jevons and Carli Elementary Price Indices Using Relatives with Missing Prices . ........................................ 184
8.5A Jevons and Dutot Elementary Price Indices Using Averages with Imputed Prices ........................................ 185
8.5B Jevons and Dutot Elementary Price Indices Using Relatives with Imputed Prices ........................................ 185
8.6 Replacing Varieties with No Overlapping Prices: Jevons and Carli Price Indices ......................................... 187
8.7 Replacing Varieties with No Overlapping Prices: Dutot Index . ..................................................................... 188
8.8 Disappearing and Replacement Varieties with Overlapping Prices ................................................................ 188
8.9 Calculation of a Weighted Elementary Index . ................................................................................................ 189
8.10 Aggregation of Elementary Price Indices ....................................................................................................... 197
8.11 Aggregation of Elementary Price Indices (Arithmetic) across Locations . ..................................................... 197
9.1 Calculation of a Chain Index . ......................................................................................................................... 199
9.2 Updating Weights for Price Change from Weight Reference Period .............................................................. 204
9.3 Updated CPI with the Same Weight and Price Reference Periods . ................................................................ 206
9.4 Updated CPI with a New Price Reference Period . ......................................................................................... 207
9.5 Linking CPI Series Using a Single Period Overlap on a New Index Reference Period ................................. 209
9.6 Linking Old and New Index Series to a Previous Annual Average ................................................................ 210
9.7 Linking New Series to an Old Index Reference Period .................................................................................. 212
9.8 Aggregating New CPI Series Using an Annual Period Overlap ..................................................................... 213
9.9 Partial Weight Updates at the COICOP Class Level ....................................................................................... 214
9.10 Linking Annual Indices for Multiple Periods with Chained Linking Factors . ............................................... 216
9.11 Decomposition of Index Changes ................................................................................................................... 218
10.1 A Numerical Example of Chain Drift ............................................................................................................. 228
10.2 Movement Splice Linking with Rolling Window of 13 Months . ................................................................... 231
11.1 Short-Term Price Relatives Based on Monthly Indices for Four Selected Elementary Aggregates ............... 238
11.2 Monthly Price Indices Using Imputed Prices for Missing Values of Winter and Summer Nightwear ................. 238
11.3 Calculation of a Mortgage Interest Charges Series ......................................................................................... 250
11.4 Relationship between the Choice of Owner-Occupied Housing (OOH) Approach and CPI Purposes . ............... 254
11.5 Matched Models: Landline Telephones .......................................................................................................... 266
11.6 Consumer Profiles: Mobile Telephones .......................................................................................................... 267
vii
Contents
11.7 Bus Fares: Old and New Tariffs ...................................................................................................................... 267

11.8 Internet Services .............................................................................................................................................. 268
11.9 Changes in the Tariff for Internet Prices ......................................................................................................... 268
11.10 An Illustrative Index Structure for Telecommunication Services (representative item approach) ................. 270
11.11 Examples of Specifications of Telecommunication Services . ........................................................................ 270
11.12 Example of a User Profile for Mobile Phone Services . .................................................................................. 272
11.13 CPI Weights: Medical Insurance and Medical Care Elementary Aggregates ................................................. 277
11.14 The Effects of a Subsidy and Tax Credit on a CPI . ........................................................................................ 279
11.15 Illustration of the Impact of Taxes on Measures of Insurance Services ($) .................................................... 287
11A.1 Calculation of a Price Index for a Deposit Product: Base Period Sample Account ........................................ 293
11A.2 Calculation of a Price Index for a Deposit Product: Fee Schedule ................................................................. 293
11A.3 Calculation of a Price Index for a Deposit Product: Bank Accounts Debit Tax ............................................. 294
11A.4 Calculation of a Price Index for a Deposit Product: Financial Institutions Duty (percent) ............................ 294
11A.5 Calculation of a Price Index for a Deposit Product: Interest Data .................................................................. 294
11A.6 Calculation of a Price Index for a Deposit Product: CPI Data . ...................................................................... 294
11A.7 Calculation of a Price Index for a Deposit Product: Projected Current Period Sample Account ................... 295
11A.8 Calculation of a Price Index for a Deposit Product: Indices for Current Accounts ........................................ 295
12.1 A Taxonomy of Errors in a CPI . ..................................................................................................................... 297
viii
FOREWORD
The Consumer Price Index Manual: Concepts and Methods contains comprehensive information and explana-
tions on compiling a consumer price index (CPI). The Manual provides an overview of the methods and practices
national statistical offices (NSOs) should consider when making decisions on how to deal with the various prob-
lems in the compilation of a CPI.
The chapters cover many topics. They elaborate on the different practices currently in use, propose alternatives
whenever possible, and discuss the advantages and disadvantages of each alternative. Given the comprehensive
nature of this Manual, it will satisfy the needs of many users.
This publication on the practice of compiling CPIs is an update of Consumer Price Index Manual: Theory and
Practice, published in 2004. Through the mechanism of the Intersecretariat Working Group on Price Statistics
(IWGPS), the update has been managed by the International Monetary Fund (IMF) and jointly published by the
organizations of the IWGPS: the Statistical Office of the European Union (Eurostat), the International Labour
Organization (ILO), the IMF, the Organisation for Economic Co-operation and Development (OECD), the United
Nations Economic Commission for Europe (UNECE), and the World Bank.
The primary purpose of the Manual is to assist countries in producing CPIs that reflect internationally recom-
mended methods and practices. It draws upon a wide range of experience and expertise to describe practical and
suitable methods to guide countries in their efforts to improve the quality and international comparability of their
CPIs and to meet user needs. The Manual is intended for statistical offices (or other compiling agencies) as a ref-
erence to compile the CPI and for training purposes. CPI users, such as employers, workers, policymakers, and
researchers, are also targeted. The Manual will not only inform them about the different methods that are employed
in collecting data and compiling such indices but will also provide them with information on the limitations of
CPIs, so that the results may be interpreted correctly.
The 2004 Manual included extensive theoretical chapters. The theoretical chapters are omitted in the updated
version of the Manual, which focuses on providing guidance on best practices for CPI compilation concepts and
methods. A companion publication will be released separately that focuses on the theoretical foundations of CPIs.
This publication, labeled Consumer Price Index Theory, provides an overview of the conceptual and theoretical
issues that drive the methods and practices.
Mariana Kotzeva
Director General
Eurostat
Rafael Diez de Medina

Chief Statistician/Director, Department of Statistics
International Labour Office
Louis Marc Ducharme

Chief Statistician and Data Officer, Director, Statistics Department
International Monetary Fund
Paul Schreyer
Chief Statistician, Director, Statistics and Data Directorate
Organisation for Economic Co-operation and Development
Lidia Bratanova
Director, Statistical Division
United Nations Economic Commission for Europe
Haishan Fu
Director, Development Data Group
World Bank Group
ix
PREFACE
The Consumer Price Index Manual: Concepts and Methods, herewith referred to simply as the Manual, is an
update of the 2004 publication Consumer Price Index Manual: Theory and Practice. Since 2004, methods and
best practices, as well as user needs, have continued to evolve. Countries have expressed the need for a manual
that better reflects current best practices and includes more practical compilation advice. The Manual was prepared
under the auspices of the Intersecretariat Working Group on Price Statistics (IWGPS), which consists of six orga-
nizations: the Statistical Office of the European Union (Eurostat), the International Labour Organization (ILO),
the International Monetary Fund (IMF), the Organisation for Economic Co-operation and Development (OECD),
the United Nations Economic Commission for Europe (UNECE), and the World Bank. The Manual is published
jointly by the six organizations.
The IWGPS, together with experts from a number of national statistical offices (NSOs) and academia, has
collaborated since 2015 on updating this Manual. The sponsoring organizations endorse the principles and rec-
ommendations contained in it as good practice for statistical agencies in compiling their consumer price indices
(CPIs). Because of practical and resource constraints, some of the current recommendations may not be immedi-
ately attainable by all NSOs, and they should therefore serve as guidelines or targets for agencies as they revise
their CPIs and improve their CPI programs. There are not always clear-cut solutions to specific conceptual and
practical problems such as sample design, choice of index formula, adjustment of prices for quality changes, and
the treatment of new products. NSOs must therefore rely on the underlying economic and statistical principles laid
out in this Manual to arrive at practical solutions.
The Consumer Price Index

The CPI is an index that measures the rate at which the prices of consumption goods and services are chang-
ing from one period to another. The prices are collected from shops or other retail outlets. The usual method of
calculation is to take an average of the period-to-period price changes for the different products, using as weights
the average amounts that households spend on them. CPIs are official statistics that are usually produced by NSOs,
ministries of labor, or central banks.1 They are published as quickly as possible, generally within four weeks after
the reference period.
The Manual is intended for the benefit of agencies that compile CPIs, as well as users of CPI data. It explains
in some detail the methods that are recommended for use to calculate a CPI. A separate companion publication,
Consumer Price Index Theory, explains the underlying economic and statistical theory on which the methods are
based.
A CPI is a measure of price changes of the goods and services purchased by households in their role as con-
sumers. It is also widely used as a proxy measure of inflation for the economy as a whole, partly because of the
frequency and timeliness with which it is produced. It has become a key statistic for purposes of economic policy
making, especially monetary policy. It is often specified in legislation and in a wide variety of contracts as the
appropriate measure for adjusting payments (such as wages, rents, interest, social security, other benefits, and
pensions) for the effects of inflation. It can therefore have substantial and wide-ranging financial implications for
governments and businesses, as well as for households.
This Manual provides guidelines for NSOs or other agencies responsible for constructing a CPI, bearing in mind
that the resources available for this purpose are limited. Calculating a CPI cannot be reduced to a simple set of rules
or standard set of procedures that can be mechanically followed in all circumstances. While there are certain gen-
eral principles that may be universally applicable, the procedures followed in practice, whether they concern the
collection or processing of the prices or the methods of aggregation, must take account of particular circumstances.
These include the main use of the index, the nature of the markets and pricing practices within the country, and
the resources available to the national statistical office (NSO). NSOs have to make choices. The Manual explains
the underlying economic and statistical concepts and principles needed to enable NSOs to make their choices in
efficient and cost-effective ways and to be aware of the full implications of their choices.
The Manual draws upon the experience of many NSOs throughout the world. The procedures they use are not
static but continue to evolve and improve in response to several factors. First, research continually refines and
strengthens the economic and statistical theory underpinning CPIs. For example, clearer insights have recently
been obtained on the relative strengths and weaknesses of the various formulas and methods used to process the
1
For simplicity, the Manual refers in general to NSOs as the statistical agencies responsible for compiling the CPI.
xi
Preface
basic price data collected for CPI purposes. Second, recent advances in information and communications technol-
ogy, such as the availability and the technical capabilities to make effective use of large-scale administrative data
sets, have affected CPI methods. Both of these theoretical and data developments can impinge on all the stages in
compiling a CPI. New technology can affect the methods used to collect prices and transmit them to the NSO. It
can also improve the processing and checking, including the methods used to adjust prices for changes in the qual-
ity of the goods and services covered. Finally, improved formulae help in calculating more accurate and reliable
higher-level indices, including the overall CPI itself.
International Standards for Consumer Price Indices

The objectives of the international standards for CPI compilation are to provide guidelines on best practices that
can be used by countries when developing or revising a CPI and to promote the quality and international compa-
rability of national CPIs.
In many countries, CPIs were first compiled mainly to be able to adjust wages to compensate for the loss of
purchasing power caused by inflation. Consequently, the responsibility for compiling CPIs was often entrusted to
ministries, or departments, of labor. The International Conference of Labour Statisticians (ICLS), convened by
the Governing Body of the ILO, therefore provided the natural forum in which to discuss CPI methodology and
develop guidelines.
The first international standards for CPIs were promulgated in 1925 by the Second ICLS. The first set of stan-
dards referred to “cost of living” indices rather than CPIs. A distinction is now drawn between two different types
of indices. A CPI can be defined simply as measuring the change in the cost of purchasing a given “basket” of
consumption goods and services, whereas a cost of living index is defined as measuring the change in the cost of
maintaining a given standard of living, or level of utility. For this reason, the Tenth ICLS in 1962 decided to adopt
the more general term “consumer price index,” which should be understood to embrace both concepts. There need
not be a conflict between the two. As explained in the Manual, the best practice methods are likely to be very
similar, whichever approach is adopted.
The international standards for calculating CPIs have been revised four times, in 1947, 1962, 1987, and 2003 in
the form of resolutions adopted by the ICLS. The 1987 standards on CPI were followed by a manual on methods
(Turvey and others 1989), which provided guidance to countries on the practical application of the 1987 standards.
The 1989 manual on methods was revised, expanded, and published in 2004.
The 51st Session of the United Nations Statistical Commission endorsed this Manual as an international statisti-
cal standard on March 4, 2020 and urged all countries to use this Manual in the compilation of their national CPIs.
The Background to the Present Update

Since 2004, substantial progress has been made in developing new data sources, price collection methods, and
related index calculation methods. This update incorporates these developments and reflects experience gained
improving CPI compilation methods. Finally, evolving user needs and the need for greater international compara-
bility contributed to the necessity for updating the 2004 Manual.
In response to the various developments in CPI compilation methods and the emergence of new data sources,
the need to update the 2004 Manual was recognized and agreed in 2014. A formal recommendation to revise the
manual was made at the meeting of the UNECE Expert Group on Consumer Price Indices, Geneva, May 2014,
jointly organized with the ILO. The participants of this meeting noted a need for clearer, more prescriptive recom-
mendations where research, methodological development, and practical experience support such recommenda-
tions and guidelines.
Following the 2014 meeting in Geneva, the IWGPS endorsed the need to update the Manual and selected the
IMF as lead agency to manage the update. The overall objective of this update was to develop a more concise man-
ual that provided more practical advice wherever possible. The updated material takes into account experiences
gained on the applicability and usefulness of the 2004 Manual; incorporates relevant developments in methods
and practices as well as theory and research over the last decade; updates material on data sources, data collection
methods, and related calculation methods to reflect developments since 2004; reflects recent developments in user
needs; and harmonizes the CPI concepts in line with the System of National Accounts 2008.
The Organization of the Update

The six international organizations listed at the beginning of this preface, concerned with the measurement of
inflation, have collaborated on the update of this Manual. They have provided, and continue to provide, technical
assistance on CPIs to countries at all levels of development. They joined forces to establish the IWGPS to develop
international standards and recommendations on price statistics, document best practice guidelines, and support
their implementation.
xii
Preface
The responsibilities of the IMF, as lead agency within the IWGPS for this update, were to:
• Appoint the various experts on price indices who participated in the updating process, as members of the Techni-
cal Expert Group (TEG/CPI), providing substantive advice on the content of the Manual and serving as authors
• Provide the financial and other resources needed
• Arrange meetings of the TEG/CPI, prepare the agendas, and write the reports of the meetings
• Arrange for the publication and dissemination of the Manual
The drafting and updating have entailed meetings and discussions over a five-year period, during which CPI
experts from NSOs, international and regional organizations, and academia have participated. The updated Manual
owes much to their collective advice and expertise.
The experts participating in the TEG/CPI were invited in their personal capacity as experts and not as represen-
tatives, or delegates, of the NSOs or other agencies in which they might be employed. Participants were able to
give their expert opinions without in any way committing the offices from which they came.
The update of the Manual involved multiple activities:
• The development of the Manual outline and the recruitment of experts to draft the various chapters
• The review of the draft chapters by the members of the TEG/CPI, the IWGPS, other experts, and CPI compilers
• The posting of the draft chapters on a special website for comment by CPI compilers and data users
• Discussions by the TEG-CPI to finalize each of the chapters
• Agreement by the IWGPS on the adequacy of the content and quality of the chapters for having a global consul-
tation of countries’ views
• Formal global consultation by the United Nations Statistics Division
• Inclusion of comments and suggestions from the global consultation
• Final copyediting of the whole Manual
• Endorsement by the 51st Session of the United Nations Statistical Commission
Electronic versions of the Manual are available on the IMF website (www.imf.org/cpi) and the IMF eLibrary
(r.imfe.li/cpi-manual). The IWGPS will issue guidance notes that will amend and update specific chapters to
address and clarify particular issues as needed. This is especially true for emerging discussions and recommenda-
tions to be made by international groups reviewing the CPI, such as the ICLS, the United Nations City Group on
Price Indices (the “Ottawa Group”), and the UNECE Expert Group on Consumer Price Indices.
Comments and suggestions on the Manual are welcomed by the IWGPS and should be sent to the International
Monetary Fund (email: [email protected]). They will be considered for any future revisions.
xiii
ACKNOWLEDGMENTS
The organizations of the Intersecretariat Working Group on Price Statistics (IWGPS) wish to thank all those
involved in the drafting and production of the Consumer Price Index Manual: Concepts and Methods. Particular
thanks go to Brian Graf, the editor, and Margarida Martins, who coordinated work on the Manual. Their efforts
greatly enhanced the quality of the Manual. Nada Hamadeh, IWGPS Chair, also deserves special thanks for her
efforts to ensure a timely completion of this update.
The current Manual represents an update of the 2004 Manual published by the International Labour Organiza-
tion (ILO). Individual authors were recruited to review and update each chapter. Some chapters required extensive
updating and rewriting, while others needed only minimal updating from the 2004 version of the manual. Two new
chapters have been added on scanner data and updating Consumer Price Index (CPI) weights.
The IWGPS established the Technical Expert Group on the CPI (TEG-CPI) for updating the Manual. Members
of the TEG-CPI were as follows: Maria Mantcheva, IMF (Retired), Co-chair of the TEG-CPI; Brian Graf, IMF,
Co-chair of the TEG-CPI and editor; Margarida Martins, IMF, Secretariat; Badria Al-Aadi, NCSI (Oman); Paul
Armknecht, IMF (Retired) and U.S. Bureau of Labor Statistics (Retired); W. Erwin Diewert, University of British
Columbia (Canada); David Fenwick, UK ONS (Retired); Claude Lamboray, Eurostat and STATEC (Luxembourg);
Yunita Rusanti, BPS-Statistics Indonesia; Raphael Posse, INEGI (Mexico); Mick Silver, IMF (Retired); Valentina
Stoevska, ILO; and Jan Walschots, Statistics Netherlands (Retired).
The Manual benefited from the experience of several experts responsible for updating the individual chapters.
The authors included: Paul Armknecht, IMF (Retired); Corinne Becker, Swiss FSO; David Fenwick, UK ONS
(Retired); Jan de Haan, Statistics Netherlands; Brian Graf, IMF, editor of the Manual; Claude Lamboray, Eurostat;
Maria Mantcheva, IMF (Retired); Valentina Stoevska, ILO; Marcel van Kints, ABS; Mick Silver, IMF (Retired);
and Jan Walschots, Statistics Netherlands (Retired).
The Manual has also benefited from valuable contributions by many other experts who served as primary
reviewers for individual chapters, including: Badria Al-Aadi, NCSI (Oman); Carsten Boldsen, UNECE; Rob
Cage, Bureau of Labor Statistics (United States); Barra Casey, CSO (Ireland); Ronald Johnson, Expert (external
reviewer); Patrick Kelly, Statistics South Africa; Brent Moulton, Expert (external reviewer); Ragnhild Nygaard,
Statistics Norway; Niall O’Hanlon, IMF; Federico Polidoro, ISTAT (Italy); Rafael Posse, INEGI (Mexico); Yunita
Rusanti (BPS-Statistics Indonesia); and V. Thuy, GSO (Vietnam).
Finally, the Manual benefited from a comprehensive review and feedback from members of the IWGPS. These
include: Carsten Boldsen, UNECE; Yuri Dikhanov, World Bank; Robert Dippelsman, IMF; Louis Marc Ducharme,
IMF; Claudia Dziobek, IMF; Anne-Sophie Fraisse, OECD; Nada Hamadeh, IWGPS Chair, World Bank; Francette
Koechlin, OECD; Paul Konijn, Eurostat; Claude Lamboray, Eurostat; Jarko Pasanen, Eurostat; Pierre-Alain Pion-
nier, OECD; Valentina Stoevska, ILO; Gabriel Quiros-Romero, IMF; James Tebrake, IMF; and Peter van de Ven,
OECD.
The IMF acted as the Secretariat of the TEG-CPI.
The TEG/CPI officially met twice: March 7–9, 2016 (Washington, DC) and January 18–20, 2017 (Vienna).
Informal meetings were held on the sidelines of the meetings of the UNECE Expert Group on CPIs in 2016 and
2018. The Manual also benefited from detailed chapter-by-chapter discussions of the IWGPS members during
October 29–30, 2018 (Paris) and January 9–10, 2020.
Gemma Diaz from the IMF Communications Department provided extensive editorial and production support
for the final version of this Manual.
xv
ACRONYMS AND ABBREVIATIONS
2008 SNA System of National Accounts 2008

API Application Programming Interface
CADC Computer-Assisted Data Collection
COGI Cost of Goods Index
COICOP Classification of Individual Consumption According to Purpose
COLI Cost of Living Index
CPI Consumer Price Index
DQAF Data Quality Assessment Framework
DQRS Data Quality Reference Sites
DSBB Dissemination Standards Bulletin Board
EAN European Article Number
EFQM European Foundation for Quality Management
e-GDDS Enhanced General Data Dissemination System
EU European Union
EUROSTAT Statistical Office of the European Union
FISIM Financial Intermediation Services Indirectly Measured
G20 Group of Twenty
GDP Gross Domestic Product
GEKS Gini, Eltetö, Köves, and Szulc
GK Geary–Khamis Method
GSBPM Generic Statistical Business Process Model
GST Goods and Services Tax
GTIN Global Trade Item Number
HBS Household Budget Survey
HFCE Household Final Consumption Expenditure
HFMCE Household Final Monetary Consumption Expenditure
HGMC Hedonic Geometric Mean Characteristics
HGMI Hedonic Geometric Mean Imputation
HICP Harmonised Index of Consumer Prices
ICP International Comparison Program
ILO International Labour Organization
IMF International Monetary Fund
IWGPS Intersecretariat Working Group on Price Statistics
L-T Long Term
MMM Matched-Model Method
MSE Mean Square Error
N.A. Not Available
NPISH Nonprofit Institutions Serving Household
NSDP National Summary Data Page
NSO National Statistical Office
OECD Organisation for Economic Co-operation and Development
OLS Ordinary Least Squares
OOH Owner-Occupied Housing
PPI Production Price Index
PPP Purchasing Power Parities
PPS Probability Proportional to Size
R&D Research and Development
SDDS Special Data Dissemination Standard
SKU Stock Keeping Unit
SPD Structured Product Descriptions
SRS Simple Random Sampling
S-T Short Term
TDH Time Dummy Hedonic
TPD Time Product Dummy
xvii
Acronyms and Abbreviations
TQM Total Quality Management

UN United Nations
UNECE United Nations Economic Commission for Europe
UPC Universal Product Code
VAT Value-Added Tax
VIF Variance Inflation Factors
xviii
INTRODUCTION, OVERVIEW, AND BASIC
STEPS FOR THE CONSUMER PRICE
INDEX DEVELOPMENT 1
Introduction or low-income households. Their product coverage can be
adapted to show what the rate of inflation is in particular
1.1 Chapter 1 provides a self-contained overview of the sectors, such as energy or food, or excluding particular
uses and the basic steps for compiling the consumer price products, such as alcohol and tobacco. They can shed light
index (CPI). More than just a summary of the chapters to on the effect of tax changes or government-regulated price
follow, Chapter 1 guides the reader through the compilation changes on the rate of inflation. They can be compiled on
process and highlights best practices that are explained in a regional basis, showing different inflation rates within
greater detail in subsequent chapters. The flow of the chap- different parts of a country or between urban and rural
ter follows the different steps needed to develop and main- areas.
tain a CPI program that better reflects the standards and best 1.6 CPIs are now considered as one of the most impor-
practices set out in the Manual. tant economic and social indicators produced by national
1.2 Both compilers and data users will benefit from read- statistical offices (NSOs)1 throughout the world. Against
ing Chapter 1. This chapter provides a primer on the methods this background, the challenge of NSOs is fourfold: to
used to compile the CPI and explains why one method is identify user needs; to conceptualize user needs with
preferred over another. The chapter aims to provide a clear, regard to economic concepts; to translate the underlying
easily understood summary of best practices and compilation concept into statistical measurement terms following the
methods without being overly technical. fundamental principles of price index measurement; and to
construct the indices so defined and evaluate them against
Developing the Consumer Price purpose.
Index
Overview of the Consumer Price
1.3 CPIs measure changes over time in the general
level of prices of goods and services that households Index Uses and Needs
acquire (use or pay for) for the purpose of consumption. In 1.7 A CPI can be used for a variety of purposes, the
many countries, they were originally introduced to provide most common ones being: the indexation of wages, rents,
a measure of the changes in the living costs faced by work- contracts, and social security payments; the deflation of
ers, so that wage increases could be related to changing lev- household final consumption expenditure in the national
els of prices. However, over the years, CPIs have widened accounts; and the use as a general macroeconomic indica-
their scope and now are widely used as a macroeconomic tor, especially for inflation targeting and for setting inter-
indicator of inflation, as a tool by governments and central est rates. Elements of a CPI are also often used in the
banks for monetary policy and for monitoring price sta- calculation of purchasing power parities and extrapolat-
bility, and as deflators in the national accounts. With the ing purchasing power parities between benchmark years
globalization of trade and production and the liberalization as required in the International Comparison Program
of the markets, national governments, central banks, and (ICP).2
international organizations place great importance on the 1.8 Given the many uses of CPIs, it is unlikely that one
quality and accuracy of national CPIs, and their interna- index can perform equally well in all applications. Some
tional comparability. countries therefore construct several CPI variants for spe-
1.4 Different conceptual frameworks can be used to cific purposes. Each index should be properly defined and
address fundamental issues relating to the nature of the named to avoid confusion, and a “headline” CPI measure
index. For example, different concepts are used if the CPI should be explicitly identified. Where only one CPI is pub-
intends to measure the change in the cost of a fixed-weight lished, it is the main use that should determine its type and
basket of goods and services or whether the target is to mea- scope. If there are several major uses, compromises may
sure the change in the cost of living, that is, the cost of main- have to be made with regard to how it is constructed. The
taining a given standard of living, taking into account the purpose of a CPI should guide all aspects of its construction.
fact that when prices change consumers change their expen- CPI producers need to know how their index is being used
diture patterns. The use and conceptual basis of the index
will determine the method of construction.
1.5 The method of construction also allows (or should 1
NSO is used to describe the compiling agency in this Manual regardless of
allow) CPIs to be adapted for a wide range of specific the institutional setting in a country.
uses. For example, they can be adapted to calculate spe- 2
See Chapter 4, Consumer Price Index Theory, and Appendix 5,
cific inflation rates for social groups such as pensioners International Comparison Program.
1
CONSUMER PRICE INDEX MANUAL
to ensure that it fulfills its purpose. In this connection, user Overview of the Consumer Price
consultation is important.
1.9 This section reviews the various uses of a CPI before Index Concepts
examining some of the issues confronted by the index com-
piler relating to the scope of the index and the practical mea-
Type of Index Number Formula
surement and compilation decisions which must be made. 1.11 Experts generally agree that the ideal type of index
for a CPI would be a “superlative” index such as the Fisher
index, which will be discussed in the following text and in
The Different Uses of a Consumer Price Chapter 8. Superlative indices make equal use of the prices
Index and quantities (that is, the expenditure weights) in both peri-
1.10 CPIs have three main uses: ods being compared (the reference period and the current
period). Current period expenditure weights are usually
• Indexation not known, so in practice nearly all CPIs rely on weights
A CPI may be used for wage or contract indexation of any relating to a weight reference period some time earlier. An
specific group, whether of a population acquiring products, exception to this are actual transactions that can be captured
or of a subset of products themselves. In either case, it at the points of purchase through the use of scanner data (as
should represent the coverage of the group concerned. For discussed in Chapter 10).
instance, it can be argued that the weights of a CPI used 1.12 Some countries aim to produce a COLI. But such
for indexation of pensions should cover only the expendi- an index is in fact a type of superlative index and suffers
ture of the pensioner population. The product and outlet list from the same practical defect as mentioned previously, and
could also be more appropriately targeted, if the data exist. it is not possible to compile in real time.
This means, for example, that a CPI used for indexing pen- 1.13 Many countries state in their published metadata
sions may use weights relating to pensioner households and that they use a Laspeyres index or a “Laspeyres-type” index
may exclude products which may be thought largely irrele- for their national CPI which, in practice, is not the case.
vant to this group of households such as educational items. It is important, nevertheless, for NSOs to state publicly
Similarly, for domestic indexation, the CPI should cover what type of index is being calculated in their CPI. A true
only expenditure of the resident population (see Section on Laspeyres index uses quantity data which relate to exactly
Geographical Coverage and Chapter 2 for more informa- the same period as the price reference period. In practice,
tion). More generally, it must be decided whether the CPI however, this is difficult to obtain and rarely the case. Most
should be, in principle, a cost of living index (COLI) or a NSOs have a price reference period which is later than the
cost of goods index—these two very different concepts are period to which the quantity data or weights relate. Also, the
discussed in the following text. weights usually will span, say, a year rather than a month (or
For certain specific types of indexation, such as for quarter). This is because one of the main sources of weights
rents, users may prefer to use just the subindex for rents. data is a household budget survey (HBS), as discussed in
In such cases, the subindex should be of a statistical qual- Chapter 3, which, ideally, run for 12 consecutive months.
ity sufficient for that purpose. The HBS generally produces usable results a year or more
• National accounts deflation after the end of a survey period.
This use requires consistency between the price data used Index Formula at Lower (Elementary Aggregate)
for the CPI and the expenditure data used in the national Level
accounts. Both data sets should cover the same set of 1.14 The first stage in the calculation of CPIs is the
goods and services and use the same concepts and same calculation of elementary price indices, which are then
classification, in principle the Classification of Individ- aggregated to obtain higher-level price indices. Expendi-
ual Consumption According to Purpose (COICOP). For ture weights are not usually available below the elementary
example, the national accounts require the valuation of aggregate level. The three most widely known elementary
goods produced for own consumption, whereas this is index formulas are the Carli, the Dutot, and the Jevons. These
sometimes excluded from the CPI either as a matter of are all based on unweighted averages of prices or price rela-
principle or for pragmatic reasons. This applies mainly to tives, and each is associated with a number of assumptions
the valuation of the services of owner-occupied housing which will have an impact on measured inflation. The Carli
and the consumption of own-produced food. (a simple arithmetic average of price relatives) and Dutot
• Inflation measurement (the ratio of simple arithmetic averages of prices) formu-
It can be argued that central banks ideally need a timely las have a number of problems associated with their use—
index relating to total inflation, not just consumer infla- particularly the chained Carli, which is discouraged as it is
tion. But NSOs generally are unable to construct such particularly associated with having a known, and potentially
indices, in part because of the measurement issues relat- substantial, upward bias. The Jevons formula (the ratio of
ing to government consumption. In the absence of such simple geometric averages or the geometric average of price
an index, most central banks rely on a CPI, using the relatives) is increasingly used as it avoids many of the prob-
domestic concept (as described in the next section and in lems associated with the arithmetic versions. It should be
Chapter 2), but measured on as wide a basis as possible, noted that an arithmetic average is always greater than or
with regard to both products and geographical coverage. equal to a geometric average and that the difference will be
The same applies to the use of the CPI as a general mac- greater, the greater the variance in the price relatives. The
roeconomic indicator. choice of formula becomes more important the greater the
2
INTRODUCTION, OVERVIEW, AND BASIC STEPS FOR THE CONSUMER PRICE INDEX DEVELOPMENT
diversity of price movements which is one argument for days of purchase. Consumption is likely to take place in the
ensuring that elementary aggregates are as homogeneous as month for which the CPI is calculated. But a semidurable
possible. This topic is discussed in more detail in Chapter 8. good such as a shirt will be worn many times over a period
of probably several years. Durable goods such as televisions
Index Formula at Higher Level and refrigerators may be used for a decade or more. The
1.15 The higher-level indices are calculated simply question arises as to which CPI month (or months) should
as weighted averages of the elementary price indices. The the purchase be allocated.
weights typically remain fixed for a sequence of at least 12 1.20 With services, these questions can be even more
months. Some NSOs revise the CPI weights at the begin- complex. Take, for example, the purchase of a season ticket
ning of each year to try to approximate as closely as pos- for a bus service. This may be a single payment for a pass
sible the current consumption patterns, but many countries which gives (“free”) unlimited bus transport for a year. It
continue to use the same weights for several years. At a can be seen that although this example is clearly a service
minimum, weights should be updated every five years. The (the use of bus transport over a period of time), it has much
use of fixed weights has a considerable practical advantage in common with the purchase of a durable or semidurable
that the index can make repeated use of the same weights. good such as a television or shirt which provides a type of
This saves time and resources. Revising the weights can be service over a long period. A service such as a medical oper-
both time-consuming and costly if it requires a new HBS ation can also be regarded as durable, since it is likely to
to be carried out. However, the longer the period between give long-term health benefits to the patient.
weight updates, the less relevant and representative the CPI 1.21 CPI theory devotes much thought to these issues,
becomes. Many NSOs are moving toward annual or bian- which can have important implications not only for how
nual weight updates. a CPI is compiled but also for the results. Three different
1.16 In Chapters 2–4 of Consumer Price Index Theory, approaches can be identified:
the superlative indices Walsh, Fisher, and Törnqvist show up
• The acquisitions approach relates to when the good or
as being “best” in all the approaches to index number theory.
service is acquired, irrespective of when it is actually used
These three indices give very similar results so that for any
or consumed. The time of acquisition of a good is the
practical reason it will not make any difference which one
moment at which the legal ownership of the good passes
is chosen as the preferred theoretical target index because
to the consumer. This is usually the point at which the
they most closely approximate a COLI. The theoretical tar-
purchaser incurs a liability to pay. On the other hand, with
get index is a matter of choice and affects the choice of for-
a service there is no change in ownership; it is “acquired”
mula used to calculate the index. In practice, an NSO may
at the time the producer provides it (for example, a sin-
prefer to designate a basket index that uses the actual bas-
gle bus journey or airline flights). A CPI based on this
ket in the earlier of the two periods as its target index on
approach measures the change in the cost of acquiring a
grounds of simplicity and practicality. As noted previously,
product. The timing of the recorded prices should be con-
it is not possible to compile a superlative index in real time.
sistent with the way in which the value would be recorded
In other words, the Laspeyres index may be the theoretical
in the expenditure data used for the CPI weights.
target index because NSOs produce a CPI that lies between
a COLI and a cost of goods index. • The use approach relates to the period over which
the product is consumed or used. A CPI based on this
approach measures the change in the cost of using the
Acquisition, Use, or Payment Approach product over time; in other words, the cost of the good is
1.17 A CPI is based on the measurement of the change distributed over its useful life. Expenditures on durable
in prices of the goods and services included in the basket. goods and services are liable to fluctuate depending on
The vast majority of goods (but not necessarily of total val- the expected duration of their useful life.
ues) are priced in the retail outlets selling them. It should • The payments approach relates to the period of time when
be noted that most often the prices recorded are the labeled the actual period-to-period payments for the product are
prices, which are assumed to be the prices actually paid by made. This can differ from the period when it is acquired
consumers. It is also generally assumed that payment for and when it is used. When payments are not made in cash,
the goods is made at the time of purchase—indeed the con- there may be a long period before the purchase is paid
sumer would regard the two events as identical. However, for, whether by credit card or other methods. The time at
payment can be in cash or on credit, including credit cards which these debits are made is irrelevant for the recording
for which the due date of payment may be several weeks of the price. The price to be recorded is the price payable
after the actual purchase. at the time of acquisition (though sometimes the method
1.18 The time factor is important also in other ways. of payment may itself affect the price).
A consumer may decide to buy a larger than normal quan-
tity of a particular good if there is a special price reduction. 1.22 NSOs need to have a clear policy on which of
The product may then be stored at home and “consumed” these approaches is used in their CPI. In practice, the choice
(that is, used) over a relatively long period. Cans of food, for between the three approaches is an issue relating to durable
example, offered cheaply for a limited period, may be stored goods and its impact is likely to affect the weight given
at home without deterioration for months, and consumed at to owner-occupier housing costs. Each of these approaches
the usual frequency. is discussed in Chapter 10 of Consumer Price Index
1.19 Another issue concerns the definition of “usage.” Theory. In countries where food expenditures and other
A bottle of milk will typically be consumed within a few expenditures on nondurables, semidurables, and services
3
account for a significant share of the CPI basket and where countries on a continuous basis, although surveys of prices
credit financing is rarely used, the acquisition, use, and pay- done in order to compute purchasing power parity may pro-
ment approaches will give very similar results and hence the vide an occasional benchmark. Where the regular collection
CPI can satisfy many uses equally well. This is the principal of the relevant prices is not practical, it may be possible to
reason why most countries use, either implicitly or explic- obtain a reasonable proxy for price movements using pub-
itly, the acquisitions approach to define what constitutes lished subindices of the other countries’ CPIs.
consumption expenditure. 1.26 Under the domestic concept, the treatment of inter-
net purchases requires a broader approach, especially given
Geographical Coverage its growing importance. Most NSOs which have exam-
ined the issue have concluded that internet purchases from
1.23 There are two distinct aspects to the question of home whether from domestic or foreign websites, should
the geographical coverage of a CPI. The first relates to the be included in the CPI. Care must be taken to convert for-
country as a whole (domestic versus national coverage), the eign purchases to the national currency as differences in
second to its regions. exchange rates will affect the price paid (for more informa-
tion on this topic, see Chapters 2 and 11). Chapter 2 provides
The National versus Domestic Concept more information on the domestic versus national concepts,
1.24 A CPI can have “national” or “domestic” coverage. while Chapter 3 provides more detail on the application of
• National coverage means that the CPI should cover the these concepts with regard to developing weights.
expenditure made by resident households (at purchasers’
prices)3 of the country, regardless of where the expendi- Regional Coverage
ture takes place. The national concept is appropriate when 1.27 Concerning the regional coverage of the CPI, the
the CPI is being used for indexation of incomes and cost general rule is that a national CPI should cover expenditures
of living measures. The weights for expenditure abroad and prices throughout the country. However, comprehensive
can be included in the HBS, but measuring prices paid coverage is not always necessary, especially if regional CPIs
abroad poses problems. The national concept thus poses a are not published and the sampling program ensures that the
measurement problem for collecting prices abroad. index is representative of the whole country. In such situa-
• Domestic coverage means that the CPI would cover all tions, CPI compilers should collect evidence from time to
the expenditure made within the economic territory of time on the trends in prices in different regions over periods
the country, including the household final consumption which cover differences in seasonal variations, to ensure that
expenditure made by foreign visitors. It is appropriate the sample remains representative. Any region which shows
where the CPI is used for national inflation analysis and price trends significantly different from the others should be
monetary policy. Many countries carry out surveys of the covered by the CPI if its inclusion is likely to have a signifi-
expenditures of foreign visitors, for example, via Inter- cant effect on the national CPI and will improve representa-
national Passenger Surveys conducted at major border tiveness. But there is little point in spending scarce resources
crossings and airports by NSO staff. This is particularly collecting prices in sparsely populated regions if that would
important for those countries which have a large number have little or no impact on the national CPI. When carrying
of foreign tourists, or a high level of cross-border shop- out these sensitivity tests, regional weights can often be an
pers. Foreign visitors will generally have very different issue. In this case, population may sometimes be used as a
expenditure patterns from those of resident households proxy for regional consumer expenditure. However, where
(for example, they will spend more on hotels and restau- regional CPIs are aggregated to compute the national CPI,
rants) and to omit them could introduce serious distor- weights should be based on regional expenditure rather than
tions into a CPI aiming to follow the domestic concept, population data.
especially if the main purpose of the index is to measure 1.28 Another difficulty regarding regional CPIs is
the inflation trends in the economy. related to the national versus domestic concept. It can some-
times be the case that a household lives in one region but
1.25 For the national concept, internet purchases from does most of its shopping in an adjacent region, particu-
foreign websites or websites of retailers based abroad should larly when a household lives close to a regional “border.”
be included in the CPI. So should purchases made abroad The question of whether the expenditure weights and the
more generally, including such items as fees for foreign prices should be allocated to the region of expenditure or the
boarding schools, even if the item, in this case education, region of residence is usually dictated by practical issues. As
is consumed outside the country. Where such purchases are with the national concept discussed previously, if the region
made in the foreign currency, they should be converted to of expenditure is used (equivalent to expenditure abroad by
the domestic currency at the relevant exchange rate. Clearly, a domestic resident) some method must be found for esti-
it would be impracticable to collect prices directly in foreign mating the proportions of expenditure made by “visiting”
consumers in the various regions so that this can be reflected
in the price indices.
3
The System of National Accounts 2008 (2008 SNA, paragraph 6.69) for- 1.29 Finally, the question often arises as to whether a
mally defines purchasers’ prices. According to the 2008 SNA, the purchas- CPI can be limited to urban areas or if rural areas should also
er’s price is the amount paid by the purchaser, excluding any value-added be covered, as in many countries a significant part of popu-
tax or similar tax deductible by the purchaser, in order to take delivery of
a unit of a good or service at the time and place required by the purchaser.
lation resides in rural areas. Having a rural CPI is important
The purchaser’s price of a good includes any transport charges paid sepa- for poverty analysis. Again, in principle, the whole territory
rately by the purchaser to take delivery at the required time and place. should be covered, but clearly, the impact on the national
4
CPI of including rural areas where relatively few monetary 1.33 In considering the practical issues relating to the
transactions take place will often argue against their inclu- inclusion of institutional households in a CPI, two questions
sion on grounds of cost. The view taken will depend, at least need to be asked. First, is the expenditure pattern of insti-
in part, on the size and treatment of own-account produc- tutional residents likely to be significantly different from
tion. If own-account production is included in the CPI, the household residents? Second, even if the answer is yes,
weights should include a valuation of the physical quantities would their exclusion from the CPI be likely to significantly
of such products, often derived from the HBS. The prices affect the national (or regional) CPI? To answer these ques-
will normally be the same as those used for actual transactions, some research should be carried out on a sample basis.
tions for the same goods sold in the same locality. 1.34 Some countries exclude certain household types
1.30 Where the weights derived from an HBS are avail- from the CPI, such as the very wealthy or the very poor.
able for rural as well as urban households it is generally Such exclusions may be on theoretical grounds (for exam-
better to use the weights for urban and rural households ple, using the argument that the expenditure of the wealthy,
combined, even if price collection is limited to urban areas, who are relatively few in number, should not be allowed
as this will normally improve the representativeness of the to affect a CPI which may be used for indexation of wages
index. But, where feasible, price statisticians should under- of ordinary workers) or on practical grounds (for example,
take pilot calculations to test whether this is the case. using the argument that wealthy households tend to have
low response rates to HBSs, and their inclusion can lower
the quality of expenditure weights). Such exclusions make
Reference Population for the Consumer the CPI inconsistent with the national accounts coverage.
Price Index For a CPI which is used for indexation of wages, the exclu-
1.31 The group of households included in the scope of a sion of pensioners and wealthy households may be justified
CPI is referred to as the “reference population.” According to on conceptual grounds. For example, it may be considered
the 2008 SNA, households are made up of private households that such households are likely to spend their money on
and institutional households. Private households are defined atypical products and including them would distort the rel-
as groups of persons who share the same living accommoda- evant overall average. It is also argued by some that pen-
tion, who pool some or all of their resources and consume sioner households should be excluded in principle from an
select goods and services collectively. Members of the same index used for the escalation or indexation of state pensions
household do not necessarily have to belong to the same fam- because of the circularity involved (the level of state pen-
ily, as long as there is some sharing of resources and con- sion influences expenditure patterns which are then used in
sumption. Institutional households consist of persons living the up-rating calculation) while others would argue that it is
permanently, or for a very long period of time, in institutions logical that indexation should be based on an index reflect-
other than private households. These include religious insti- ing the expenditure of pensioner households and their spe-
tutions, hospitals, the military, prisons, or retirement homes. cific inflation experience. Note that if wealthy households
Persons who enter such institutions only for short periods of are excluded, the CPI basket should not include products
time should be treated as members of the individual house- likely to be bought only by the excluded group, nor should
holds to which they normally belong (the 2008 SNA defines outlets specializing in such products be included in the sam-
household, private, and institutional household in paragraphs ple. Conversely, if the wealthy are included, some “luxury”
4.149–4.154). Temporary foreign workers may live together products and outlets should also be included in the sample.
in special housing blocks, which may also be treated as insti- For the analysis of national inflation, it is considered that the
tutional households in the population census. more comprehensive the CPI, the better.
1.32 Expenditure on accommodation and living costs,
such as lodging fees and charges for meals, imposed by the Democratic versus Plutocratic Weights
institution, as well as personal expenditure by the individual 1.35 A “democratic” CPI uses weights based on the aver-
on, for example, clothes and toiletries, should be included age proportionate share that households in the whole popula-
in the CPI. However, care should be taken to ensure there tion have spent on the item. Hence, the share of a specific item
is no double counting where, for instance, a family rather in the basket is calculated for each household. The weight for
than an individual pays the accommodation costs charged the item is the sum of the household shares divided by the
to a patient in a hospital. A data source like the HBS should number of households. Many households may purchase the
be designed to pick up the amount spent on such charges item, but also, a number of households will not make such
just once and the standard convention is to record it against purchases. The average share is thus based on the experience
the household which incurs the costs. In the previous exam- of each individual household, whether they have purchased
ple, the accommodation costs should be included under the the item or not. Each household share is equally weighted in
expenditure of the family, and not the individual. If individ- deriving the average. The alternative “plutocratic” CPI cal-
uals spend their own money on clothes and other incidental culates the item shares as the total expenditure on the item by
expenditure, then the HBS should record this expenditure as all households divided by the total expenditures of all items
being incurred by the individuals. In practice, many HBSs purchased by households. This latter method gives more
do not cover institutional households and, where this expen- weight to the high-spending households. It is argued that a
diture is considered to be significant, estimates will need to democratic index is more suitable for showing the impact of
be made from, for example, special surveys of people living inflation on the average household but is very rarely com-
in institutions or by reference to the expenditure patterns of piled by NSOs. There is a consensus that a plutocratic index
similar people, say, the same sex, age, and socioeconomic is the appropriate index to use for national accounts deflation
group, living in noninstitutional households. and for a general measure of inflation.
5
1.36 Many countries publish a range of CPIs relating coverage this category of nonmonetary transaction, so no
to subsectors of the population such as all households, low- price should be imputed.
income households, or pensioner households, but a CPI 1.42 A CPI should measure the prices of final consumption
based on democratic weights is very rare. by a household. In principle, the first category, imputed trans-
actions where households do not incur a liability but bear the
Product Exclusions costs of acquiring the good or service in another way, should
1.37 In its role as an indicator of total consumer infla- be included in a CPI used for deflation in the national accounts
tion, the CPI should in principle cover all types of goods compilation and, in principle, can be included in a CPI com-
and services which are consumed in the national retail mar- piled for other purposes. Perhaps the most important example
ket. In practice, some types of products may be excluded is the consumption of own-account production products such
for policy reasons while other exclusions are unavoidable as food and owner-occupied housing services. Here, there is
in practice. These may include: goods sold illegally, such no actual transaction at all, and thus no price. If the transaction
as narcotics; black market sales of mobile phones and other is to be valued, a price has to be imputed. This would usually
goods; gambling; and prostitution. In most of these cases, be done by reference to actual purchases of the same product
except perhaps for legal lotteries, there will be no expendi- in, say, nearby markets. But even when this is done for the
ture data from the HBS and prices are difficult if not impos- purpose of national accounts compilation, it is not necessarily
sible to collect. Thus, in principle, estimates of weights and appropriate to include production for own consumption in a
prices will need to be made for the purpose of producing general-purpose CPI or in a CPI used for indexation. Indices
deflators, even if the expenditure is not covered in the CPI. used as a general measure of inflation or for indexation should
Solutions will need to be found to the practical measure- be based on a narrow definition of consumption that includes
ment issues. For instance, if a CPI covers gambling, it is only monetary expenditure. From the point of view of measur-
not the gross stakes which should be included in the weight, ing inflation and also for the purposes of income indexation,
but the net stakes, which is broadly equivalent to the margin the most common view is that it is best to omit production for
taken by the gambling operator. As it is not likely that the net own consumption on pragmatic grounds. Furthermore, goods
stakes can be measured, one solution may be to distribute and services purchased by households which are then used as
the weight for gambling across other subclasses in COICOP inputs into own-account production are normally treated as if
class 09.4.6 (recreational and sporting services). they themselves were consumption goods and services, and
1.38 The treatment of second-hand goods is often found therefore are included in the CPI. Some countries may find it
to be problematic. As far as transactions within the house- useful to produce two versions of the CPI: one including and
hold sector are concerned, sales will balance purchases, so the other excluding own-account production. This topic is fur-
the weights will be zero, and they may be excluded from ther discussed in Chapter 11.
the CPI. But in many countries, there are significant sales
of imported second-hand goods from dealers and other third Deciding on the Index Coverage
parties, such as cars and clothing. Where sales of imported
second-hand goods sold by dealers or other third parties
and Classification Structure
are significant relative to sales of new goods of the same 1.43 Classification is a central theme in the compila-
product, such sales should be included, in both weights and tion of the CPI. Choosing a classification system is the first
prices (as described in Chapter 11). step in compiling the CPI because its subaggregates must
1.39 CPI compilers sometimes face proposals from gov- be defined in such a way that the expenditure weights and
ernments or pressure groups to exclude certain categories prices will relate precisely to the coverage of the subaggre-
of products for nonstatistical reasons. Common examples gates. The classification is also important because it estab-
are alcohol and tobacco in some countries where their con- lishes the framework to define and draw the boundaries for
sumption is socially discouraged or they can only be pur- the inclusion of the representative items in the index (and
chased illegally. While it is acceptable to produce a variant sometimes the outlets). Finally, the classification system
of the CPI excluding such products, the all-items CPI should helps in defining which level of the hierarchy will be suit-
include them, where practical, to ensure that the index pres- able for publication.
ents a true and accurate picture of national inflation. 1.44 In years past, countries used their own distinct sys-
1.40 Chapter 2 covers in some depth the treatment of tems for classifying the range of products covered by their
other excluded or partly excluded products, including taxes CPI. Most countries have now, however, moved to the inter-
and licenses, subscriptions, insurance, gambling, financial national standard classification COICOP.
transactions, hire purchase, and interest payments. 1.45 COICOP was first developed for the 1968 SNA to
provide the structure for classifying household final con-
Imputed Transactions and Imputed Prices sumption expenditure. The various components of house-
1.41 A distinction can usefully be made between hold final consumption expenditure are often used as the
imputed transactions and actual transactions where a price is basis for the weights in the CPI. The 2003 International
imputed. In the category of actual transactions, a prescribed Labour Organization (ILO) Resolution on CPIs requires
medicine would be provided free as part of a national health that national CPI classifications should be reconcilable
service. There is a “transaction” in the sense that a product with COICOP at least at its higher aggregation levels. Most
changes hands but at “zero” price so that it does not consti- countries have adopted COICOP in their economic statis-
tute a monetary transaction. The conventions for a CPI con- tics (for example, in the CPI, national accounts, ICP, and
structed for the purposes of indexation or the measurement HBSs), with a clear advantage for integration of data sets
of inflation as a macroeconomic indicator exclude from and enhanced analytical capabilities.
6
Classification Systems: The General Case purpose. Households will select various goods and ser-
1.46 In its broadest sense a classification is a proce- vices in order to satisfy their consumption objectives (that
dure in which individual items are organized into categories is, renting an apartment for the provision of shelter or eat-
(classes) and subcategories (subclasses) based on informa- ing an apple for the purpose of nourishment). These goods
tion on one or more characteristics inherent to the items. A and services are disaggregated further into various groups
classification structure will usually have these same items and may not be based solely on the principle of purpose
(or elements) arranged in a hierarchically ordered system but also according to product type. For example, oranges
based on category–subcategory relationships where the and apples are included in the “Fruit” category. The more
subcategory has the same description as the associated cat- detailed breakdown is often a product-type classification
egory in addition to one or more descriptions. For example, because these items share a similar production process and
an apple is a subclass of fruit. So any apple is a fruit, but are certainly sold at fruit stands or the same location in the
not every fruit is an apple. A product needs to have a more supermarket.
detailed description to be an “apple” and not just a “fruit.”
1.47 In principle, a classification system can be based Deriving the Weighting Pattern
on any attribute of the objects being classified. Normally,
1.50 A CPI measures changes in the cost of a represen-
organizing a population of items into categories must leave
tative basket of goods and services. This involves weight-
no two categories with any item in common; in other words,
ing aggregated prices for different categories of goods and
the categories must be mutually exclusive. Also, the cate-
services so that each takes an appropriate share to reflect
gories must collectively include all of the items which are
the budgets of the households covered by the index. For
in the population—the categories must be exhaustive. For
instance, if most people spend far more on fresh vegetables
example, in the case of the CPI, its classification should
than on electricity, then a price rise for fresh vegetables must
include the entire universe of goods and services that are
have more effect on overall price rises than a similar-sized
covered by the index (for example, fresh food purchased
increase for electricity. Therefore, at the lowest level, each
in a store by a consumer is part of the CPI, while heavy
elementary aggregate should receive a weight equal to the
machinery such as a tractor is not) and no product should be
ratio of expenditure by the covered households on items rep-
included in two different categories in the structure.
resented by that aggregate to total expenditure by covered
households on all items within the scope of the CPI. Chapter
The Consumer Price Index 3 discusses the derivation and sources of the expenditure
Classification System weights and provides detailed guidance on specific issues.
1.48 COICOP, as its name implies, is founded on the 1.51 The 2003 ILO Resolution on CPIs makes the obvi-
principle of “purpose” (see Box 1.1). It is a purpose-type ous but important point that the weights follow directly from
classification because throughout the aggregation program the scope of the index as well as from the choice between the
the products are grouped according to the purpose (or func- “acquisition,” “use,” or “payment” approach, as described in
tion) they usually fulfill such as transport, nourishment, paragraph 1.22. It also states that there are two main sources
shelter, and so on. Most national CPIs aim at measuring the of information: HBSs and national accounts, and that the
change of the cost of a basket of goods and services, which weights should be reviewed and updated at least every five
is consumed for the purpose of satisfying certain needs. A years. Additionally, new sources of weight information are
purpose-based classification would therefore appear to be being developed such as actual expenditures on various types
the logical classification system for a CPI. of transactions based on scanner data and other electronic
1.49 The official COICOP is a five-digit classification. formats. Such sources are being evaluated and exploited for
NSOs will expand the COICOP to six and seven digits to use in development of weights now and in the future.
obtain more detail for their use. At the higher level of the 1.52 The use of expenditure weights is consistent in
classification, the products are grouped according to the concept with a CPI based on the acquisition, payment, and
user cost approaches although the treatment of major dura-
ble goods and housing can present a problem, particularly
the costs of owner-occupied housing. The use of weights
Box 1.1 Classification of Individual Consumption in a CPI based on total consumption expenditure is often
According to Purpose (COICOP) referred to as plutocratic weights because this concept gives
more weight to the expenditure patterns of high-spending
The COICOP classification structure has been updated and
the new version, COICOP 2018, was endorsed by the United households (which will also tend to be those with higher
Nations Statistical Commission in March 2018. COICOP 2018 incomes). Paragraph 1.36 provides more details on pluto-
replaces COICOP 1999 as the official standard classification for cratic weights.
the CPI. 1.53 The goods and services consumed by the house-
While a number of countries have begun to implement
COICOP 2018, many others will implement the updated classifi-
holds can in principle be acquired in four ways: (1) purchase
cation in the coming years during a future CPI update. Because in monetary transactions; (2) from own production; (3) as
many countries continue to use COICOP 1999, the Manual payment in kind; and (4) as transfers or gifts from other eco-
presents some examples using COICOP 2018 and others using nomic units, including social transfers in kind provided by
COICOP 1999. government and nonprofit institutions serving households.
Chapter 2 summarizes the key changes made to COICOP.
The detailed classification structures for both COICOP 2018 and 1.54 The weights are determined by the scope of the
COICOP 1999 are included as appendices to the Manual. CPI and should be derived on the basis of the relevant cov-
erage and types of consumption and with reference to SNA
7
concepts. But it should be noted that there are several delin- where the retail market is heterogeneous (see Chapter 5).
eations of consumption used, as described in Chapter 2. The Some levels are the following:
broadest possible scope for goods and services would cover
all four of the previous categories. It would include all social • Central shops weights to represent a small number of
transfers in kind in the form of education, health and hous- large supermarket chains or chain stores which have uni-
ing, and other goods or services provided free of charge or at form prices across branches and prices are provided by
nominal prices. The total acquisition of goods and services the shops’ headquarters.
thus described is equivalent to household actual final con- • Stratum weights. For some types of expenditure, pur-
sumption in the SNA. For the CPI as a general measure of chasing patterns may differ markedly by region or type
inflation, the more relevant would be to include only goods of outlet, and in these cases, stratification will improve
and services purchased in monetary transactions by the the accuracy of item indices. For example, each locally
households. Only monetary transactions generate prices that collected item in the index could be allocated to one of
can be observed for the CPI, but this then leaves outstand- the different stratum types to allow the best available
ing the issue of owner-occupier housing services which is information about purchasing patterns to be incorporated
considered in detail in Chapter 11. in the index calculation. Depending on the structure of
1.55 It is against this background that a CPI often fol- the retail market, the stratum types could be: region and
lows the concept of household final consumption expendi- outlet type; region only; outlet-type only; and no strati-
ture, as laid down in the 2008 SNA. This approach is often fication. The assignment of stratum type will depend on
recommended for a CPI being used as a macroeconomic the information available for constructing the weights for
indicator, restricted to the appropriate reference population, each item and the number of prices collected per item.
or “Index Households,” where the CPI is being used as a In principle, all locally collected items might be stratified
compensation index. This compensation index might, for by both region and outlet type, but if the weights data are
example, exclude the very rich. unreliable or nonexistent at this level of detail, then the
1.56 Household final consumption expenditure includes item may be allocated to another stratum type. Allocation
nonmonetary transactions (such as for owner-occupier also partly depends on which outlet types are specified for
housing or consumption of own production of food). The the collection of prices and the number of prices collected.
concept of “household final monetary consumption expen- If the rules for the choice of outlets did not specify that
diture,” used in the European Union Harmonised Index of both a chain and an independent outlet should be chosen
Consumer Prices defines both the goods and services to be for an item, there may be too few prices collected in one
covered, and the price concept to be used, and refers only of these outlet types to make stratification by outlet type
to monetary transactions. Household final monetary con- meaningful. In some instances, there may be no stratifi-
sumption expenditure is a useful concept but many coun- cation because research has shown that stratification has
tries prefer to also include some nonmonetary transactions little effect.
(in particular owner-occupier housing) in their CPI, some- The weight of an elementary aggregate (that is, the
times using imputed costs thus moving the coverage of the stratum weight) should reflect the expenditure on the
index closer to household final consumption expenditure. entire elementary aggregate and not the weights of the
Chapter 2 provides more details on the different types of outlets and items that have been chosen to represent it.
consumption. For instance, if spaghetti is chosen as the representative
1.57 The conceptual issues relating to the construction product under the elementary aggregate with the heading
of weights are discussed in detail in Chapter 3. of pasta products, then the weight of this category should
reflect expenditures for all pasta products and not solely
Weighting Structure the “lower” weight of spaghetti; that is, the weight of the
1.58 The weighting structure should follow the aggre- pasta category will be represented entirely by spaghetti.
gation structure of the CPI. For instance, if this structure is Similarly, if an expenditure category is divided into two
based on COICOP, then this is the structure which should be elementary aggregates according to outlet type, say, open
used for the weights. markets and supermarkets, with corresponding market
1.59 Additional subdivisions can be introduced where shares of food sales, 70 percent and 30 percent, respec-
there is further stratification of the sample to include geo- tively, then the same rule as mentioned previously should
graphical location, outlet type, or a more detailed prod- apply. For instance, suppose a single store is selected as
uct level classification. Thus, the weighting structure will the representative outlet for a particular food item sold in
depend on the sample design for price collection and com- supermarkets in a country where two supermarket chains
pilation and in particular the need for more detailed weights have equal sales, then the sales from the sampled store
which may be generated by additional sample stratification. will account for the total value of the expenditure weight
In general, NSOs will collect some prices centrally and of 30 percent; the weight of the elementary aggregate for
adopt up to four levels of sampling stratification for local this food item sold in supermarkets should not be 15 per-
price collection: locations; outlets within locations; items cent (0.30 × 0.50), that is, a weight based only on the sales
within different sections of expenditure; and varieties. The of the selected supermarket.
sampling of varieties is normally conducted in the field • Product or item weights (in the current context the terms
by price collectors and does not normally involve explicit can be interchangeable). Some products or items may be
weights (see Chapter 5). Stratification is frequently used intended only to represent themselves; others represent
to increase sampling and operational efficiency, especially a subclass of expenditure within a section. For instance,
8
within electrical appliances, an electric cooker may repre- expenditure weights. There are two ways of doing this:
sent only itself and not any other kinds of electrical appli- (1) including the weight in a related elementary aggregate
ances. However, other products or items will represent (this may involve the creation of a “miscellaneous” cate-
price changes for a set of products or items, which are gory), or (2) having the weight of the product for which no
not all priced, so for these the weight reflects total expen- representative prices exist equal to zero, which essentially
diture on all products or items in the set. For example, a redistributes the weight to other elementary aggregates.
screwdriver may be one of several items representing all 1.63 In general, the prices for the product for which
spending on small tools within home improvement and prices are not collected will be expected to exhibit a similar
maintenance materials, and there are other items within movement to the other products in the elementary aggregate
the section representing all spending on paint, timber, fit- and the first of the previous two methods should be used.
tings, and so on. The second method may be used where the elementary
• Upper-level or section weights. It is common practice to aggregate is heterogeneous, or the associated price index is
give each section an integer weight in parts per thousand not considered very reliable. Because of the negligible size
or per hundred so that the sum of the section weights is of the weight value involved, the consequence on the overall
1,000 or 100. It is likely that most of these weights will index will also be negligible whichever method is used.
be based on the HBS results. The main exceptions will
be some housing sections, including (where applicable)
mortgage interest payments and depreciation, where Data Sources
weights are estimated from other sources (see Chapter 1.64 Depending on the population coverage, weights
11), and for certain other sections (for example, tobacco, for a CPI are derived either from expenditure data based on
confectionary, soft drinks, and alcoholic drinks) where estimates drawn from a sample covered in the HBS or from
the HBS may be thought to under-record expenditure national accounts estimates of household final consumption
and better data are available elsewhere. Many countries expenditure. It should be emphasized, however, that expen-
also use national accounts household final consumption diture estimates in the national accounts are usually partially
expenditure estimates where available at the COICOP based on HBS information, although they may differ with
group or class level and the HBS expenditure distribution regard to coverage, and that in some countries these are
at the lower levels. This is also an area where scanner not available or not compiled at a detailed level. The two
data can be used to more accurately reflect the expendi- sources are not entirely independent. Note also that national
ture distribution. accounts data may also be used when the HBS is conducted
too infrequently to ensure the reliability of the CPI or when
the expenditure weights need to be updated more often than
Implicit Weights Within Elementary Aggregates the periodicity of the HBS. Nevertheless, an HBS will still
1.60 An unweighted formula (for example, Jevons or
have to be conducted eventually because it is an important
Dutot—see paragraphs 1.145–1.151 on elementary price
source for benchmarking the components of household final
indices) is usually used when aggregating the elementary
consumption expenditure of the national accounts. Other
aggregate price relatives of the sample of products at the
sources for the weights are also available and are usually
elementary aggregate level. This practice is usually justi-
complementary to these two main sources. These include
fied on the grounds that the required information such as
administrative data sources or retail trade statistics data.
market shares is simply not available to a sufficient level
These are discussed in Chapter 3.
of precision. However, if broad-based estimates of mar-
1.65 When various sources of information are used for
ket shares are available, then these can be used as implicit
generating the weights in a CPI, the compiler should take
weights for determining the sample of price observations to
the time to check the data to ensure that the results are con-
enhance the accuracy of the elementary price index. Some
sistent and plausible with what is expected or investigate
possible sources for this information are transaction shares
further if necessary.
from scanner data, trade publications, market reports, and
consultation with industry experts.
Household Budget Surveys
1.61 The sample of price observations based on implicit
1.66 When using the HBS as the basis for developing
weights can be updated independently and more frequently
CPI weights, the sample size (number of households) should
than the weights of the elementary aggregate; however, the
be sufficient to ensure that the expenditure data yielded
price statistician will need to ensure that the weights are both
produce statistically reliable and representative weights
coherent and consistent within the elementary aggregate. It
at the elementary aggregate level. In some countries, the
is best to review them at the time of updating the basket.
acceptable statistical quality is based on the coefficient of
variation (the ratio of the standard deviation to the mean).
Weights for Products for Which Prices For those expenditure weights that are unable to meet the
Are Not Collected minimum requirement of reliability, three options should be
1.62 As it is not feasible to collect a full set of prices
considered:
from every outlet, including market stalls and street ven-
dors, and from every provider of a service, all prices have • If reliable expenditure data are not available for an ele-
to be sampled. This means that in practice there will be mentary aggregate, they can be combined with another
some products which consumers spend money on for which related elementary aggregate to form a new broader ele-
prices will not be collected (see Chapter 4). However, the mentary aggregate (for example, “wheat bread” could be
expenditures for these products need to be included in the combined with “rye bread” to form a new category called
9
“bread”). This approach will often lead to a more reliable accounts definitions and classification systems for house-
elementary aggregate but may require an adjustment to hold consumption, which is an advantage when compiling a
the existing structure of the CPI. CPI as a macroeconomic indicator and for use as a deflator.
• If annual HBS data are available, then expenditures could 1.70 National accounts have two inherent advantages:
be averaged over more than one year thus improving the sta- • The household final consumption expenditure aggregate
tistical reliability of the data, with regard to standard errors, of national accounts may be derived mainly from the
but to the detriment of timeliness. It should also be noted HBS, but national accountants will often use other sources
that averaging may not improve the statistical quality of the of information before finalizing their results, especially
expenditure estimates if a particular category of household in cases where the accuracy of the HBS is in doubt such
expenditure is rapidly growing or declining. Averaging is as where underreporting is present. National accounts go
useful if the particular expenditure category under consid- through an additional quality assurance process and rees-
eration shows a lot of variability over several HBSs but no timation should increase the reliability of the weights.4
clear trend. This is an area where statisticians will have to
• Even if the HBS is updated infrequently, CPI weights can
use their judgment. The basket reference period should not
still be updated at regular intervals from national accounts
be arbitrarily chosen and periods of less than a year should
data for higher-level aggregates at the division or group
be avoided because of seasonal influences on consumption
level.
patterns. Furthermore, some countries exclude, from multi-
year averages, years which are exceptional, for example, as 1.71 However, there are two inherent disadvantages
a result of particularly poor harvests leading to high prices with national accounts data:
and distorted expenditure weights.
• They are generally only available at the national level, so
• Leave the CPI structure unchanged and simply accept deconstruction of the national accounts data to provide
that the weight for the particular elementary aggregate a finer level of detail or to produce regional expenditure
concerned is less than ideal. Whether this is an accept- weights may be necessary using other available sources
able position to take will depend on the weight of the of information. Other data sources include HBSs, retail
elementary aggregate and on its importance, particularly surveys, aggregated transaction from scanner data, and
to analysts. For example, it would be a difficult position administrative data such as statistics on excise duty.
to sustain if the elementary aggregate has a large weight National accounts data can be used to derive weights
and is presented as a published subindex. at the more aggregate level, and HBS data can then be
1.67 It should be noted that in normal circumstances applied to derive weights at the lower levels of the aggre-
weights can tolerate a certain degree of imprecision before gation program. If the expenditure data from the HBS are
having a significant effect on the overall CPI, particularly not viewed as sufficiently detailed to ensure a minimum
at the higher levels of aggregation, or main divisions of the of acceptable accuracy, or if a demand exists for indices
CPI. But this is less so at lower levels. For instance, an index of a finer breakdown (for example, there is a need for a
described as “fruits and vegetables” where the true weight price index for apples but only expenditures for all fruit
for fruit is 40 percent and the weight of vegetables is 60 per- can be derived from the HBS), then other potential data
cent, but with a biased estimate of the expenditure weights, sources can be used for disaggregating the expenditures,
fruits account for 60 percent of the index and vegetables the including surveys of retail sales from establishments,
remaining 40 percent. The biased weights affect the relative point-of-sales surveys and aggregated scanner data, sur-
importance of both fruits and vegetables in the basket, giving veys of production, export and import data, and adminis-
too much weight to fruit and too little weight to vegetables. trative data. Note that some of these sources may also be
Consequently, the price index for “fruits and vegetables” used for stratifying expenditures according to sales vol-
will be also biased. To minimize the potential for such occur- umes by retail outlet type and by region.
rences, it is recommended that the compiler always strives to • National accountants apply an element of discretion and
get the best possible estimates for the expenditure weights. judgment when making operational decisions relating to
1.68 An annual HBS is optimal for a CPI because as the construction of national accounts. Some of the details
well as avoiding one-off setup costs, it permits the annual of these decisions are not always readily available to
updating of the weights, hence reducing the substitution bias users. Consequently, compilers of the CPI should consult
associated with out-of-date weights in a fixed-basket index with their national accounts counterparts regularly before
like the CPI. Furthermore, it provides the opportunity for using their data for weights in order to ensure that they are
using multiyear weights to reduce the sampling error and, consistent with the objectives of the CPI.
where considered appropriate, minimize the sampling vari-
ance associated with unusual expenditure patterns in a par-
ticular year (for instance, abnormal circumstances affecting Designing the Sample
consumer behavior such as political events, natural disas-
1.72 Chapter 4 gives advice both on sample selection,
ters, or oil shocks). But obtaining reliable consumption
that is, how to construct a sample, and on estimation pro-
estimates is challenging and there is a persistent trade-off
cedures, that is, how to estimate the CPI from the sample
between data quality and survey cost.
National Accounts 4
This assumes the national accounts have independent expenditure side
1.69 The use of national accounts weights ensures con- estimates, and household final consumption expenditure is not estimated
sistency and comparability between the CPI and national as a residual.
10
of prices collected. This Manual recognizes that in practice decided upon, such as regional indices or separate subindices
nonprobability sampling sometimes needs to be carried out. for urban and rural areas. As well as cost, a limiting factor in
Similarly, although the 2003 ILO Resolution states that sample design is the time taken in collecting prices. Practi-
probability sampling techniques are to be preferred, it goes cal considerations that will need to be considered include
on to say that “where appropriate sampling frames are lack- the availability of price collectors and transportation issues.
ing and it is too costly to obtain them, samples of outlets and 1.77 In general, NSOs adopt four levels of sampling for
products have to be obtained from non-probability meth- local price collection: locations; outlets within locations;
ods” and that “statisticians should use available information items within different sections of the expenditure classifica-
and apply their best judgement to ensure that representative tion; and product varieties. Stratification is also frequently
samples are selected.” used to increase sampling efficiency, especially where the
1.73 To construct a perfectly accurate CPI, the price retail market is heterogeneous. Often a combination of prob-
statistician would need to record the price of every variety ability sampling and nonprobability sampling is used.
of every good and service purchased by the consumer. This 1.78 When using probability sampling, the units in the
would mean collecting a full set of prices from every outlet, sample are selected so that each has a known nonzero proba-
including market stalls and street vendors, and from every bility of selection. For instance, locations could be randomly
provider of a service, including public utilities such as water selected from local administrative areas with probability
and electricity, private transport including shared minibuses according to total population,5 the population representing a
and the hire of rickshaws, modern forms of communication, proxy for the retail turnover in the area. Within a location,
such as mobile telephones, and the provision of domestic outlets could be randomly selected from a business register,
service. As this is not feasible in practice, most prices have with probability according to their individual turnover or
to be sampled and this involves local price collection in a sales or by floor area measured by an enumerator listing and
selection of outlets in a sample of locations chosen to be rep- visiting each shop. Sample selection based on probability
resentative of the country as a whole and at selected times according to size increases sampling efficiency. Also, as the
on selected days. aim is to have a sample which is representative of retail turn-
1.74 The exceptions to selecting a sample of prices as over, the prices subsequently collected on the previous basis
described in the previous one are prices which can be col- would then not need to be rebalanced by reweighting if the
lected from a central source, such as a public utility pro- assumption holds that those population and floor areas are
vider, national retail chain headquarters, or a government good proxies for turnover. Alternatively, each location and
department. For many of these items, all prices will be taken outlet could be given an equal chance of selection in the sam-
and no sampling will be involved. For example, the service ple, regardless of the total proportion of the retail market that
provider may give the NSO a full price list or tariff from they account for, but then reweighting would be necessary.
which all the prices can easily be extracted. This may be 1.79 In practice, sample selection is never straightfor-
the case where sampling would not make sense or would ward, and compromises must be made for good practical
be unreliable because the number of prices is very small. reasons even when a sampling frame exists. Administrative
For instance, no sampling would be involved if an electric- boundaries may not coincide well with statistical targets. For
ity tariff consisted of a standard standing charge for service instance, choosing administrative areas using probability
provision and a standard charge per kilowatt of electricity according to population could ignore the inconvenience of
used, which was the same for all customers regardless of administrative boundaries that straddle the border between
location and only varied with total usage (with heavy users a commercial district and a residential area so that, contrary
getting a discount after a certain threshold). In this case, to the intention, the commercial district has no chance of
the tariff prices would be collected and applied to a typical being selected as it contains no houses. Also, a visit to the
cross-section of users and varying quantities of electricity. location may indicate that it is impractical for the collection
Sampling would be used to choose a cross-section of users. of prices, for example, because of a physical barrier such as
a railway or river bisecting the area and causing difficulties
of access. Similarly, very rarely do NSOs have readily avail-
Approaches to Drawing Samples able sampling frames which reliably list all retail outlets,
1.75 The focus of this section is on sampling proce- particularly recent openings, and even fewer will have lists
dures for local price collection in outlets, including options that cover all market stalls in all types of markets, or mobile
relating to probability and nonprobability sampling. There street vendors. The relative advantages and disadvantages of
is a section in Chapter 5 which specifically addresses the random and purposive sampling should be examined at each
special challenges of sampling prices in markets and prices stage of the sample selection. It is recommended that the
charged by street traders. Chapter 5 also addresses the asso- NSO should first decide on the ideal sampling solution and
ciated issue of price bargaining and discusses the issue of then modify this to take into account practical constraints.
volatile prices. 1.80 The ultimate goal should be:
1.76 As only a sample of prices will be recorded in the
• An overall sample which is representative of the total
course of local price collection, there is inevitably some
population of goods and services being offered for sale
sampling error in measuring the CPI. The sampling proce-
dures should aim to minimize this sampling error, maximize
sampling efficiency (that is, obtain the maximum sampling 5
A more appropriate alternative might be number of workers employed or
precision for minimum fieldwork and processing costs) and aggregated retail turnover in the location if this is known. This may be the
reduce bias as much as possible. The sample design should case where most shopping is done in town centers but with most people
allow publication of subindices at all levels which have been living in residential areas in the suburbs.
11
and purchased. The sample chosen should be represen- 1.83 Goods and services that normally are paid accord-
tative of price levels and, most particularly, price move- ing to a tariff can pose problems when their structures
ments. All variations of items and outlet types should be are modified over time, compromising the principle of
considered for each product and chosen to reflect typical unchanged consumption. Examples include public transport
consumer purchasing habits. fares, electricity, main water supply, physicians, hospital
• A variance or mean square error which is as low as pos- services, and telecommunications. For utilities, the pay-
sible. Samples should be reasonably optimized. At the ment may consist of a standard rate per unit of consumption
very least, a basic analysis of sampling variance should sometimes in combination with a fixed charge. A solution to
be conducted, even if an overall estimate of the precision this problem is to define representative services or bundles
of the CPI cannot be made. of services (for example, categories of consumers and spe-
cific services consumed). For these, it is important that the
• Optimization. The entire set of sample prices should be
prices experienced by a representative range of customers
optimized to meet the publication needs of the CPI, taking
and tariffs are observed and that customer profiles are kept
into account user requirements, practical data collection
constant over time. More detailed advice is given in Chapter 11
considerations, and cost.
on selected special cases.
1.84 The focus in the next paragraphs is on locally col-
Collecting and Editing the Prices lected prices. It begins by reviewing the principles behind
collecting prices for a CPI and then considers the practical
1.81 Chapter 5 provides an overview of the most appro-
collection issues and how these should be managed. A work-
priate survey methods for the collection of prices while
ing assumption has been made that the index being compiled
Chapter 10 discusses the detailed approaches that can be
is an acquisition index (see Chapter 4). It is also assumed
used for scanner data. In large part, the considerations are
that prices are being collected for a monthly price index with
the same as for sample design and will depend on local cir-
prices therefore being collected, in general, every month.
cumstance. For instance, the methods should consider: the
Some countries produce only a quarterly CPI, while others
purchasing habits of consumers and the extent to which they
produce a weekly index, especially for fresh food. The con-
use licensed and unlicensed markets and internet purchases;
cepts and procedures discussed will apply to price collection
the structure of the retail market including the balance
practices, whatever the frequency of index publication.
between markets, small independent shops, and large retail
chains; the extent of public ownership and price control; the
diversity of goods and services being sold; the pricing struc- The Principles of Price Collection
tures used, including tariffs; and whether bargaining is com-
mon. The availability of central records of prices charged 1.85 Except in a small number of cases, such as the
also has an important bearing. The 2003 ILO Resolution treatment of owner-occupied housing costs, a CPI is usu-
emphasizes the importance of well-trained price collectors ally designed to measure the change in the actual transaction
who adhere to the standard procedures. prices of goods and services bought by consumers. How-
1.82 There are two basic price collection methods: ever, collectors cannot normally observe individual transac-
tions as they occur, so they must usually observe the price
• Local price collection where prices are obtained from marked on or assigned to the product and assume that these
outlets located around the country. This will include are the transaction prices. Many countries are also improv-
licensed and unlicensed markets and street vendors as ing the collection of actual transaction prices through the
well as shops. Normally the price collector will need to use of electronic technologies such as scanner data and
visit the outlet to observe the prices although the prices internet purchases. An exception to observation and elec-
for some items may be collected by other means, includ- tronic capture is bargaining, where a price might not even
ing telephone and price lists. be displayed, and this case is discussed in detail in Chapter 5.
• Central price collection is often used where prices can Special procedures apply when outlets in which items are
be collected by the head office without the need for being priced close down, or the items which the price col-
fieldwork. This may also include centrally regulated or lector was pricing at a particular outlet are no longer sold
centrally fixed prices which can be obtained from the by that outlet. These procedures are covered in Chapters 4
regulatory authorities, although in these cases checks will and 7.
need to be made to ensure that the goods and services
in question are actually available and sold at the stated The Principle of a Fixed Basket
price. It is not unusual to find goods subject to price con- 1.86 Underlying all of what follows in this section is
trol being sold at a different “unofficial” price. Central an important principle: the necessity to compare prices on a
price collection can be further broken down into: like-to-like basis from one period to the next. This has two
• Prices which are combined with prices collected consequences:
locally. For instance, this may occur when a supermar- • Where the price collector has the role of selecting what
ket chain provides centrally determined price lists or variety of a product to price in a particular outlet, a con-
actual transaction data, from scanners, eliminating the sideration should be whether that variety will be available
need for the price collector to visit a shop in person. to price over a reasonably long period (tight specifications
• Prices which are used on their own to compute cen- are of no use if the described variety cannot be found in
trally constructed indices. Most tariffs fall into this the outlets). This is in addition to being typical of what is
category. sold to customers.
12
• The price collector should record additional information The detection of errors in the collection and recording of
needed to ensure the unique identification of the variety price information must occur as soon as possible after the
priced so that: information is collected. Detection is usually achieved by
• The same variety continues to be priced in the case of examining price movements and checking those that exceed
price collection being carried out by a different person. some predefined limits or appear to be unrealistic based on
an analysis of all available information. Statistical outputs
• The identification of quality change can be evaluated
should sufficiently portray the reality of the economy, and
when the variety disappears and is replaced by a differ-
with an output review, the compiler ensures the indices
ent one allowing an adjustment for quality change to be
reflect reality.
made.
1.91 It should be noted that while price collectors should
examine every price they collect, subjecting every collected
Variety Specifications price to the same level of examination by collection super-
1.87 There are no firm rules, especially regarding the visors and index compilers is not considered necessary and
use of looser or tighter variety specifications: each country generally is not feasible. It is recommended that, to improve
may choose its preferred methods—and stick to them. How- cost-effectiveness, some form of significance rating should
ever, there are a number of considerations in deciding on be applied to determine how much time and effort should be
variety specifications: expended on editing individual prices.
• Tight specifications leave less discretion to the price col- • In general, prices from elementary aggregates with rela-
lector, so the reliability and training of collectors are fac- tively small price samples should receive more attention
tors to consider when deciding whether to use loose or from the index compiler. This is because the weights of
tight specifications. the elementary aggregates are broadly equal. Each indi-
• Particular care should be taken to ensure that the specifica- vidual price movement from these elementary aggregates
tion is very detailed for heterogeneous items where there will have potentially a much more significant influence
is scope for significant difference between one variety and on the index calculations than any individual price move-
another, and for items which by nature are highly speci- ment from an elementary aggregate with a large number
fied. Cars and hi-tech goods fall into the latter category. of price quotes.
• Tight specifications also allow for the calculation of • Price samples from elementary aggregates with high
meaningful average prices: expenditure weights should be examined critically as the
high expenditure weight will make all price movements
• Average prices are useful in identifying outliers and
within the sample significant to the index calculation.
assessing the accuracy of the reported prices.
• The highest risk is associated with elementary aggregates
• Average prices allow comparisons of price levels, includ-
with relatively large weights but few price quotes and
ing between, for example, regions or urban and rural areas. with complex index construction. This situation is associ-
1.88 Responsibility for specifying the items to be priced ated with utilities and other services which account for
should normally rest with the head office. Specifications relatively large expenditures and where there may be only
should be reviewed on a regular basis to determine whether one or a handful of suppliers and prices are based on com-
they continue to be relevant. A revision of specifications plex tariffs. Petrol prices could be another example.
could be implied by: (1) a large number of missing price 1.92 There are two main categories of checking and
quotations; (2) a large number of substitutions; or (3) a wide identifying possible data errors and outliers, as further
variation in the distribution of collected price levels. described in Chapter 5:
1.89 Some countries find structured product descrip-
tions (SPDs) from the ICP helpful for specifying items to be • Nonstatistical checking
priced in a CPI. As well as providing a ready-made frame- • Statistical checking
work for detailed item specifications the use of SPDs has
the additional advantage of facilitating greater integration 1.93 Some of the editing techniques discussed in Chap-
between the two price collection exercises leading, among ter 5 work best when applied to large quantities of data. The
other things, to savings in collection costs from using the best results will therefore be obtained if conducted at the
same price quotes in both the CPI and ICP. location where many prices are available to any analyst.
This will generally be regional offices or, more probably, in
the head office. However, the techniques can be adapted and
Data Editing still be applied to prices held by a small collection center as
1.90 Once the price information has been collected and a way of quickly and efficiently detecting extreme prices.
recorded it has to be edited. Data editing is the process of Abnormal prices such as sale prices, or price movements,
ensuring correct and usable data for calculation of price such as sale recovery prices, may be excluded from auto-
indices for elementary aggregates. Data editing is some- mated procedures for the detection of outliers, in particular
times referred to as input editing. There are three steps in the setting of upper and lower bounds, but should neverthe-
this process: less be checked, for instance by reference to previous price
history. For seasonal items, such as fresh food and cloth-
• The detection of possible errors and outliers
ing, abnormal price movements will be normal. These price
• The verification and correction of data movements should not be excluded from the outlier detec-
• Output review tion procedures. It is important that an appropriate method
13
is used to validate these prices. Chapter 5 provides more detailed guidance both on implicit methods of quality adjust-
details on the different data validation methods. ment, such as the overlap method and class mean imputa-
tion, and on explicit methods, including expert judgment,
and the hedonic approach. Chapter 7 delves more deeply
Maintaining and Updating into the issue of item substitution, particularly on methods
the Sample of incorporating new products into the index.
1.94 One strategy to deal with the changing universe of 1.100 The 2003 ILO Resolution advises that when a
products would be to resample, or reselect, at regular inter- product disappears “clear and precise rules should be devel-
vals the complete set of items to be priced. For example, oped for selecting the replacement product.” It lists three
with a monthly index, a new set of items could be selected selection strategies: the most similar; the most popular;
each January. Each set of items would be priced until the and the most likely to be available in the future. On qual-
following January. Each January, both sets would be priced ity adjustment, this resolution states that “when a quality
in order to establish a link between each set of 12 monthly change is detected, an adjustment must be made to the price,
changes. Resampling each year would be consistent with a so that the index reflects as near as possible the pure price
strategy of updating the expenditure weights each year. For change.” It guards against the automatic assumption that
example, the Harmonised Index of Consumer Prices and “all price change is a reflection of the change in quality.” It
many national CPIs in the European Union Member States does not recommend particular explicit or implicit methods
resample items annually. of quality adjustment but does state that “the methods used
1.95 Although resampling may be preferable to main- should as far as possible be based on objective criteria.”
taining an unchanged sample or selection, it may not be prac-
tical for those countries that update weights infrequently. Missing Products
When the CPI is updated every five years, for example, 1.101 In order to measure price change from one period
systematically resampling the entire set of products each to the next, the price statistician tracks, for each elementary
year would be difficult to manage and costly to implement. aggregate, the prices of a fixed sample of items. The detailed
Moreover, it does not provide a complete solution to the characteristics of the products, that is, the varieties of goods
problem of the changing universe of products, as it does and services selected for pricing, are recorded to assist the
not capture price changes that occur at the moment when price collector in fulfilling the aim of pricing exactly the
new products or new varieties or models are first introduced. same product in the same outlet in the same location so that
Many producers deliberately use the time when products are the CPI compares “like-to-like” in subsequent periods. Also,
first marketed to make significant price changes. the recording of detailed characteristics, especially price-
1.96 A more practical way in which to keep the sample determining features, can help when needing to make adjust-
up to date is to rotate it gradually by dropping certain items ments to the recorded price due to changes in specification
and introducing new ones. Items may be dropped for two and hence quality. In practice, the particular product being
reasons: priced in a specific outlet may become unavailable—for
• The product is believed by the price collector or head example, the product is discontinued, may be in temporary
office to be no longer representative. It appears to account short supply, or may be a seasonal product which disappears
for a steadily diminishing share of the total expenditures when out of season. These situations are discussed in Chap-
within the basic categories in question. ters 6 and 7 of this Manual, and in Chapter 9 of Consumer
• The product may simply disappear from the market alto- Price Index Theory. In all other cases, the price statisticians
gether. For example, it may have become obsolete as a need to estimate the price of any missing product that they
result of changing technology or unfashionable because believe will return to the market within a reasonable time,
of changing tastes, although it could disappear for other or, if they believe it will not return, find a suitable replace-
reasons. ment. If the price statisticians believe that the product will
not return, the replacement should be either (1) as similar as
1.97 At the same time, new products or new qualities possible to the previous one, or (2) the most popular “simi-
of existing products appear on the market. At some point, lar” product in the shop, or (3) the similar product that most
it becomes necessary to include them in the list of items likely will be available for future pricing. Unlike approach
priced. This raises the general question of the treatment of (1) which leaves the sample “static” with the risk that it will
quality change and the treatment of new products. be increasingly out of date and difficult to collect prices for,
approaches (2) and (3) have the advantage of introducing
Missing Products and Adjusting an element of sample replenishment. This is where quality
adjustment becomes an issue. The price index should reflect
for Changes in Quality only pure price changes—the price index should not reflect
1.98 A CPI should reflect the change in the cost of buy- any portion of the price difference that is due to increases
ing a fixed basket of goods and services of constant quality. or decreases in quality between the missing item and its
In practice, this represents a challenge as products can per- replacement. A value needs to be placed on any change in
manently disappear or be replaced with new versions of a specification between the old and replacement item and a
different quality or specification, and new products can also quality adjustment applied accordingly. This approach to
become available. quality adjustment applies to any replacement strategy, but
1.99 Chapter 6 discusses the nature of “quality” and is particularly relevant where sample replenishment takes
methods for adjusting prices for quality change. It provides place.
14
1.102 Three situations that regularly occur are the has stayed in the same stratum. Similarly, a visit to the
following: new location may indicate that it is impractical for the
collection of prices, for example, because of a physical
• Substitution procedures where an item, product, or outlet
barrier such as a railway or river bisecting the area and
disappears, including the introduction of new items
causing difficulties of access.
• The imputation of a price when a product is temporarily
out of stock (excluding seasonal goods) Products
• Quality adjustment where a change of product involves 1.106 If a chosen product is temporarily missing and no
changes in its price-determining characteristics price is recorded, a note to this effect should be made by the
price collector. For a product temporarily missing, a price
Substitution Procedures has to be imputed. Nonseasonal items and varieties should
1.103 In a dynamic retailing environment, there is a be replaced if missing more than a predefined period of
continuous turnover both in outlets and in products. time. For example, if it is out of stock for three consecutive
months, then the collector should be instructed to choose a
Outlets replacement which matches as closely as possible the prod-
1.104 If an outlet goes out of business or refuses to par- uct description unless it is decided to take the opportunity of
ticipate in the price collection survey, it should be replaced a disappearing good to update the sample. Where a product
with the same sort of outlet (for example, a market stall is permanently unavailable for pricing, procedures need to
should be replaced with a market stall, or a single shop with be in place for determining a replacement and then impute
a single shop) in the same location and conducting the same a new base price if the replacement is of a different quality.
type of business (in other words selling the same types of Methods for imputing a missing price are discussed in the
goods). For example, if the previous shop was a butcher sell- following text.
ing refrigerated meat, then another butcher selling refriger- 1.107 As the issues relating to temporarily and per-
ated meat should replace it. If probability sampling was used manently missing products differ—and their treatment is
to select the original outlet, the sampling frame should be different—it is important for the price collector to establish
revisited and a replacement outlet selected from the same whether the unavailability of a product is likely to be tempo-
stratum. Regardless of how the replacement is found, the rary or permanent. A price may be considered as temporarily
original outlet’s sampled items should be assigned to the missing if the same product is likely to return to the mar-
replacement outlet for price observation. ket within a reasonable time period. This includes seasonal
1.105 If an outlet changes location, a decision on whether items for which special procedures apply (see Chapter 11).
the price collector should follow the outlet to its new location Permanent unavailability, on the other hand, occurs when a
needs to consider both sampling and operational issues: variety is withdrawn from the market with no prospect of
returning in the same form. Products may be temporarily
• Sampling issues. The principle of maintaining a like-for-
missing, for example, due to supply shortages caused by
like comparison holds. In practice, the balance of the factors such as the seller underestimating demand, strikes
sample needs to be maintained to ensure that it continues by factory or transportation workers, or supply problems
to be representative. Stratification is frequently used to with imported products. In these cases, the price collector,
increase sampling efficiency and ensure that the sample although not able to observe a price in the current period,
is representative. This means that when an outlet changes may have obtained information (for example, from the shop-
its location, reference needs to be made to the stratifica- keeper) to suggest that the same variety will become avail-
tion and selection procedures used in the initial sample able again at some, perhaps unknown, time in the future.
selection. For instance, assume that shopping locations
1.108 The previous discussion does not cover seasonal
were initially selected from local administrative areas, products, that is, where a product or item may disappear,
stratified by an urban/rural split and outlet type, and then because it is a seasonal product and it may be expected to
outlets randomly selected from those outlets within the reappear when it is next in season. The case of seasonal
chosen local administrative areas. Then the relocating products is covered in detail in Chapter 11. Imputation pro-
outlet can be followed to its new location if it continues cedures are fairly similar for both temporarily missing and
to fall in precisely the same stratum. However, if the out- seasonal products.
let moves away from the original shopping location, for
example, from an urban shopping district to a rural area
outside the city or to another urban district within the city, Temporarily (Nonseasonal) Missing Products
or if it relocates within the same shopping location but 1.109 If it is believed that a missing product will be
becomes part of a multiple chain, then it has moved to available again in a reasonable time, then the price statisti-
another stratum and a suitable replacement for it should cian has three options:
be found from within the original stratum, in order to • Omit the variety for which the price is missing so that
maintain the sample balance. a like-for-like comparison is maintained using matched
• Operational issues. As mentioned in Chapter 5 there may pairs. The elementary index uses only those observations
sometimes be operational reasons for departing from the for which the price collector obtained prices of exactly
sample generated by the standard selection procedures. the same variety in the current and previous periods. In
For instance, efficient scheduling of price collection and this approach, the price change for the deleted product,
the availability of price collectors may make following which was recorded up to the point immediately before
the outlet to its new location impractical, even though it its disappearance will be disregarded from that point on.
15
This may cause problems, for example, if it unbalances different specification, normally two actions are required by
the sample. the price statistician:
• Carryforward the last observed price. Carrying forward • Selecting a replacement product for pricing
the last observed price is only recommended in the case • Quality adjusting the price if there are differences in qual-
of fixed or regulated prices. Although this provides price ity between the outgoing product and its replacement
continuity in the periods when observations cannot be
made, it is likely that short-term movements in the index Each is considered in turn.
are biased, since the subindices in question will show no
change when prices are not available. If prices in general The Selection of a Replacement Product for Pricing
are rising, the bias will be downward, whereas if prices 1.112 In practically every period for which a price index
are falling, the bias will be upward. Carryforward is not is compiled, some varieties of a product become permanently
recommended, particularly when there is high inflation or missing, not just in particular outlets but also because they
when period-to-period movements (as opposed to annual are no longer produced or imported. If no action is taken, the
movements) in the price index are important. The carry- sample of prices will diminish. Permanently missing variet-
forward method is appropriate only if there is reason to ies are problematic not just because of the potential impact
believe that the price has not changed. Typically, it will on how representative the sample is, but also because it will
lead to estimation of price change with samples that do not
be difficult for the price statistician to validate the belief
match from period to period; that is, the composition of the
that the price has not changed, unless the price is fixed or
matched pairs changes. In addition, the index number for the
regulated.
latest month will be less reliable than that for the previous
• Imputation. The best solution by far is to impute a price. month because of the smaller sample size.
Imputation makes use of the best available information 1.113 The price statistician’s task is to maintain the
to provide an unbiased estimate of price movement. sample size by finding replacements for the specific variet-
The principal methods for imputing prices are shown ies when they are no longer available and are not expected
in more detail in Chapter 6. There are essentially two to return within a reasonable time. One of two alternative
choices: strategies can be adopted (replace with the most similar
• Impute the missing price by reference to the average product and sample replenishment). Under both options it is
price change for the prices that are available in the important to identify any differences in quality between the
elementary aggregate (overall mean imputation). This original and replacement varieties as it is crucial to ascertain
assumes that the price change of the missing product, whether there is a quality difference, and, if there is, to esti-
if it had been available in the shop, would have been mate its value and calculate a quality-adjusted price.
equal to the average change in prices in the elemen- 1.114 Replace with the most similar product. This
tary aggregate. This may be a reasonable assumption reduces the role of quality adjustment, as the more similar
if the elementary aggregate is fairly homogeneous. a product is, the less is the required quality adjustment (see
This method of imputation is equivalent to the “omit” paragraphs 1.118–1.131) but contributes to the deteriora-
method (see first bullet point), no matter whether a tion in the representativeness of the sample where a prod-
Jevons, Carli, or Dutot method of aggregation is used uct starts disappearing from the shops because it is being
replaced by something new and sales are declining. Finding
at the elementary aggregate level. In a given month
the replacement that is most similar to the original variety
this approach provides similar results to the “omit”
requires knowledge about characteristics of the previous
method described in the first bullet previously; how-
variety. Good practice in price collection involves maintain-
ever, across time, the two approaches will not produce ing up-to-date descriptions of the variety’s characteristics.
the same results if the index is compiled using the The ICP developed SPDs for most item categories: these
short-term formulation. This is because the imputed provide a framework to list the various characteristics—pri-
prices are used to compile the index from month to oritized in order of their importance—of the varieties for
month. each category.6 Such descriptions of characteristics can be
• Impute the missing price by reference to the average used to match the characteristics between the old variety
price change for the prices of “comparable” varieties and various replacements so that critical characteristics are
from another similar outlet (class mean imputation). matched and less important characteristics can be noted.
This represents a more precise match between the miss- Critical characteristics are those that impact or contribute
ing product and the products supplying an imputed to determining the price, such as type of product (canned
price. It is normally preferable to impute using the aver- tuna fish), brand (StarKist), size (150 milliliter), and packag-
age price change in the elementary aggregate unless ing (tin, in water). Less important characteristics are those
the imputations are unreliable because of small sample that do not affect the price, such as color of the label on
sizes. the package or the location in the shop where the product
is displayed. The salient considerations can be incorporated
1.110 A detailed discussion on imputation methods is
presented in Chapter 6.
Permanently Missing Products 6

See the ICP 2003–2006 Handbook on the World Bank ICP website
1.111 When the situation arises where a product per- at https://www.worldbank.org/en/programs/icp/brief/handbooks-and-
manently disappears or is replaced by a new version with a operational-guides.
16
into a decision-making framework for identifying a similar collector should determine if the outlet is likely to continue
product, as follows: selling the replacement. If it is also expected to be discon-
tinued in the near future, then a different variety should be
• There is a basic match of the main characteristics, particu-
selected—either another that is similar to the first, or the
larly those which determine price.
most popular variety within the product line. As will be
• Consumers perceive them as similar even though some char- seen, there are several different methods for both the explicit
acteristics may be different. This may be the variety in the and the implicit types of quality adjustment, but there are
shop that buyers are most likely to buy in place of the original. some common themes in the methods of each of these two
• They are used for the same purpose and in similar situ- main types. Explicit methods estimate the impact on price of
ations. For food this may include a consideration of changes in characteristics or features of the product. Implicit
whether the brand is one for everyday consumption or methods estimate the impact on price indirectly by reference
only for special occasions. to price differences between different varieties.
1.115 Under the “replace with a similar product” strat-
egy, an updated version of a product (that is, the one that the Comparable Replacements
supplier lists as the replacement), is the logical starting point 1.119 If the selected replacement product is regarded as
for the replacement for its predecessor. In most instances comparable, then the observed price change is treated as a
this would be the one that is the most similar to the origi- pure price change. But the price statistician should gather
nal variety, so the price statistician can compare “like-to- and examine all the available evidence, if possible taking
almost-like.” In instances where the most similar variety is advice from market experts where necessary, before coming
also one that is likely to soon disappear, the price collector to such a conclusion. Even in cases where a replacement
should select the variety that is most popular within the out- product is believed to be of comparable quality, care should
let for the product class. Although this strategy is less likely be taken, since experience suggests that most goods tend to
to yield a replacement that is sufficiently similar to permit undergo steady improvements, especially hi-tech and elec-
direct price comparison, it will reduce the likelihood of the trical goods.
replacement disappearing in the near future and will keep
the need for quality adjustment to a minimum. Direct or Explicit Quality-Adjustment Methods
1.116 Sample replenishment—replace the missing 1.120 There are a number of direct or explicit methods
product or variety with the currently most representative for determining the price associated with a change in quality.
one by going through a process of resampling. The extent to But quality adjustment is not an exact science, and different
which a sample remains representative is highly dependent evaluations can yield different results. The point is that even
on the rules used for item replenishment when a particular if the evaluation methods used are somewhat imprecise in
variety or product disappears from the shelf of a particular the measurement of the quality difference, it is important to
outlet. Compared with the option of replacement with the make an adjustment. If quality differences are not removed,
most similar product, sample replenishment has the benefit the price index will reflect the quality change in addition
of maintaining the current representativeness of the sample. to pure price change and is likely to have an upward bias
If disappearing products are always replaced with similar because quality usually increases.
products, the sample will gradually become less relevant 1.121 Chapter 6 provides a flowchart of the decisions
to market reality. Sample replenishment also increases the needed for making quality adjustments. The most common
chances of the replacement being available on the shop shelf ways of making an explicit quality adjustment in practice
for pricing. The problem of a deteriorating sample increases are as follows:
with the rate of turnover in varieties and products and with • Package size adjustment. The value of a change in pack-
the rate of product development. age size is assumed to be proportional to the relative
change in the package size.
• Expert judgment. The price collector determines the value
Adjusting the Price for Differences
either through direct knowledge or in consultation with
in Quality personnel in the shop where the product is sold. Alterna-
1.117 When a product disappears or is replaced with tively, NSO staff, who may be knowledgeable about the
a new version of a different quality or specification, then product, can estimate a value. Reliance on an individual's
one of the following methods of introducing the price of the knowledge concerning the products depends, however,
replacement is adopted: on the individual obtaining sufficient market information
• Comparable replacement and is also liable to subjectivity. The judgment needs to be
properly informed.
• Direct (explicit) quality adjustment
• Reference to the producer cost. The production cost from
• Implicit quality adjustment (imputation)
the producer can be used in the case of an improvement
In all cases, a nominal price in the base month is needed for to an existing product, although a judgment then needs
the new or replacement product—this nominal base price is to be made on whether to apply an adjustment factor.
used until the next rebasing. For example, the adjustment might consist of the normal
1.118 The price collector should record the specifica- retail markup to reflect wholesalers’ and retailers’ costs
tions (the price-determining characteristics) of the new and profits. In the context of a COLI, a downward adjust-
variety so that the head office can determine if the replacement may be appropriate to account for the fact that the
ment is of similar quality to the original variety. The price new “standard” feature will not increase the utility to all
17
purchasers, for example, some may not welcome air con- observed price for the “old” item. This will yield an estimate
ditioning in a car because of the extra running costs. Com- of what the replacement item would have cost in the previ-
pilers of the producer price index often attempt to gather ous period.
information on production costs from manufacturers for 1.123 Assume that the NSO was able to determine that,
quality-adjustment purposes. However, producers may based on the differences in characteristics, the value of the
be unwilling to provide information on their marginal quality difference between Brand C and Replacement 1 was
costs for reasons of confidentiality. There are a number of $25 in period 1. The price statistician can add this amount to
potential problems with using the producer cost method. the price of Brand C in period 1 to obtain an adjusted price.
In particular, it is not necessarily the case that the cost
of production, with an adjustment factor along the lines Implicit Quality Adjustment
described previously, gives a good indication of the sell- 1.124 If the replacement product is of a different qual-
ing price.7 ity or specification and no information is available to quan-
• The former “option” price. In the case where an optional tify the difference, then assumptions have to be made about
feature has become standard, the former price charged what proportion of the price difference is accounted for by
for the optional feature can be used as the explicit differences in quality. Implicit quality adjustment creates
quality-adjustment value but again, consideration will an imputed “quality-adjusted” price based on price changes
need to be given to whether a scaling-down factor should from similar varieties of the product. The precise nature of
be applied—in this case possibly a downward adjustment the imputation depends on the index formula that is used.
to reflect the reduced production costs of making a feature A basic assumption underlying the most commonly used
“standard” and also, in the cost of living context, that the implicit quality-adjustment methods is that the difference
utility gain is less than the increase in price. Some of the in quality between varieties simultaneously available in the
concerns relating to using producers’ costs (see the previ- market is equal to the difference in price between the variet-
ous bullet point) apply, the main difficulty being that it is ies or models.9 Thus, when a product disappears from the
likely that the market valuation of the options will have shelf, an underlying assumption is made when imputing a
changed once they become standard, indeed, it is often price that a price differential continues to reflect a difference
because of changing market circumstances that producers in quality. The different methods of imputation available to
make former options standard. the index compiler are described in paragraphs 1.127–1.131.
1.125 Most countries construct some form of a fixed-
• Hedonic regression. Another way to obtain a value of
base price index. If price movements are estimated using
the quality difference is to use hedonic regression to esti-
long-term price relatives from the base period, then the
mate the value of changes in a product’s characteristics.
base price may be adjusted proportionately for the esti-
See Chapter 6 and also the Handbook on Hedonic Price
mated quality difference. If price movements are estimated
Indexes and Quality Adjustments in Price Indexes pub-
using short-term relatives from the previous period, then an
lished by the Organisation for Economic Co-operation
imputation adjustment can be implicitly made by estimating
and Development for more information.8 Hedonic meth-
the missing variety’s price in the current month from the
ods require large databases with a wide range of product
average price change in its elementary aggregate. The price
characteristics. Such databases are seldom available in
change of the omitted observation is equal to the change in
NSOs (though for some products scanner data from large
its elementary aggregate.
stores may be a viable source in some countries) and can
involve substantial development and maintenance costs.
Overlap Pricing
In addition, hedonic models need to be reestimated peri-
1.126 This method requires knowledge of the prices
odically. Hedonic methods should be applied only where
of the two varieties in the same time period—the overlap
they add significantly to the statistical integrity of the
period. If the old and new varieties are available at the same
index. This is most likely to be the case with hi-tech high-
time because the price collector either knows in advance that
turnover goods.
the old variety will disappear soon and selects and prices a
1.122 In explicit methods, the monetary value of the replacement, or the outlet is able to accurately give the price
quality difference is determined directly using one of the of the replacement in a previous period when the old vari-
previous methods and then applied to a previous period’s ety was priced, then the price difference between the two is
taken to be the value of the quality difference. The rationale
is that it cannot be due to price change because price change
7
In constructing a CPI, consumer valuations of products should be decisive,
occurs only over time. The price index uses the old variety
and the cost of production is irrelevant in this context. In “equilibrium” in the overlap period and the new variety in the next period
situations, the cost of production (using a “normal” cost of capital in the and the price differential between the old and new varieties
producer’s user cost formula) should give a close approximation to the sell-
ing price (and hence the consumer’s valuation of the product). But it is in
disequilibrium situations, involving new products, that research and devel-
9
opment has to be amortized using monopolistic prices where new products For the argument to be valid it has to be assumed that the consumers are
and obsolescence of old products is likely to occur that the assumptions of rational and well informed about available alternative choices. It also has
equilibrium break down. to be assumed that the market is “in a state of equilibrium,” that is in a state
8
Jack Triplett. “Handbook on Hedonic Indexes and Quality Adjustments in that would persist indefinitely if all given conditions remained unchanged.
Price Indexes: Special Application to Information Technology Products.” In practice, these assumptions may be adequate or inadequate to varying
STI Working Paper 2004/9, Statistical Analysis of Science, Technology and degrees in different product areas. Equilibrium conditions may be disrupted
Industry; OECD Publications, Rue André-Pascal, 75775 Paris, Cedex 16, by temporary imbalances between supply and demand, such as stock clear-
France. http://www.oecd.org/dataoecd/37/31/33789552.pdf. ances, bad harvests, or scarce, much-demanded new models.
18
never affects the index. In this case, the market has deter- comparable with explicit quality adjustments, but the
mined the value of the quality adjustment. remaining ones will need imputed prices. Class mean impu-
tation calculates imputed price relatives using only the prices
Overall Mean Imputation of comparable and, where appropriate, explicitly quality-
1.127 Overall mean or nonclass mean imputation adjusted varieties or models. In general, it does not use the
(also referred to as “linking,” “splicing,” or bridge overlap prices for the varieties or models that were not replaced,
method in the context of the European Union’s Harmonised because these are likely to be different from those of new
Index of Consumer Prices) imputes an overlap price for the models. The prices of old models tend to fall as they become
old variety in the current period by reference to the price obsolete, while the new models (represented by the replace-
changes between the previous and current periods of simi- ments) tend to have a higher price before falling. This may
lar varieties or items. The latter are used to impute a price not be relevant in developing countries where new products
change for the old variety, which can then be used to obtain appear in the retail market relatively late at a “mature” price.
the imputed price. The ratio of the imputed price for the old Using class mean imputation adds complexity but reduces
variety and the price of the new variety in the current period two types of bias referred to earlier: bias from ignoring qual-
is the estimated quality adjustment. ity change altogether and treating all price movements as
1.128 An estimate of the price of the missing variety can price change, and bias from overadjusting for quality change
still be made even though the price of the replacement vari- by treating some pure price change as quality change.10
ety may not be known for the previous pricing period using
the overall mean price change for the elementary aggregate.
1.129 The overall mean procedure assumes that the
New Products
pure price change from the replaced item to the replace- 1.131 An entirely new product, in contrast to a new
ment item is the same as that for the composite of all other variety of an existing product, is essentially a replacement
goods in the same group. It is used frequently because of of a previously popular product and represents a good or
its simplicity, but there can be an inherent bias built into service that:
the methodology, particularly when major model changes
• Was not included and could not be included in the price
are occurring. The direction and extent of this bias is a mat-
index during the initial selection of the current market
ter for debate but depends on whether the actual quality-
basket and which is now available for possible inclusion
adjusted change in price is bigger or smaller than the
in the index
measured price changes of the items used in the imputation.
Major price changes can frequently occur at the time new • Cannot be easily linked to the service flow or produc-
varieties or models of a product are introduced. This is quite tion technology of existing goods and services; that is, it
common, for example, with new vehicles, household appli- represents a distinct departure from previously available
ances, electronic equipment, and clothing items. As the new products insofar as it is a step change with regard to tech-
varieties are introduced there may still be a substantial sup- nology or utility to the consumer
ply of the old varieties which are showing little price change • Has a recognizable and generally accepted new benefit to
or may actually be declining in price. In consequence, using consumers as a result of becoming available
the old varieties’ price changes to impute the price changes 1.132 The last two cases help to distinguish an entirely
for new ones will underestimate the actual price change for new product—referred to as a revolutionary product—from
the new varieties and cause a downward bias in the price an existing product whose features and, in consequence,
index. The use of the overall mean imputation procedure, “quality” has changed—an evolutionary product. A revolu-
in which all observations in the elementary aggregate are tionary product is an entirely new good or service that is not
used, is not recommended for such cases. It is also possible, closely tied to a previously available product. A revolution-
and can be argued (but less convincingly), that the use of ary product tends to be a good or service that is expected to
the overall mean leads to an upward bias because the price satisfy some need in a new way and is unlikely to fit neatly
changes associated with models that are unchanged in qual- into an existing CPI item category. As an example, when
ity will be further along the evolutionary cycle and therefore mobile telephones were first introduced, they provided a sig-
will be rising less rapidly. An alternative imputation proce- nificantly new service. While on the one hand, they provided
dure, called “class mean” imputation, avoids some of the an extension of an existing flow of service (telecommunica-
problems associated with bias. tion), on the other hand, they provided a dimension of ser-
vice that was new (the opportunity to make “mobile” calls
Class Mean Imputation
away from a fixed telephone) and a distinct product from
1.130 Class mean imputation is similar in procedure to existing landline telephone services (it was a step change in
the overall mean imputation, but uses only the price changes technology). It is therefore an example of a revolutionary
of “comparable” replacements to impute the overlap price, product. More recent examples of revolutionary products are
the replacements being limited to those that have exactly broadcasting (streaming) services for television and smart-
the same price-determining characteristics, or those items phones, downloads of games, and electronic storage of data
with replacements that have been declared comparable after (the “cloud”). Examples of evolutionary products would be
review or have already been quality-adjusted through one
of the “explicit” methods. For example, when the arrival of
a new model of a particular brand of motor vehicle forces 10
Clothing items usually require additional procedures to control substi-
price collectors to find replacements, some of the replace- tutions depending on the season of the year (for example, autumn/winter
ments will be of comparable quality, others can be made versus spring/summer clothing).
19
new models of household appliances such as refrigerators the sample, as manufacturers usually introduce more up-to-
and washing machines where improvements in quality are date features into new models. However, the representative-
introduced from time to time. Evolutionary products can ness of the new model, as measured by its popularity, will
also be newly added brands of currently available products only be determined over time.
such as a new type of canned fish or electronic appliance 1.136 Further discussion of the methods for introducing
which differs from those currently available. For example, into the CPI basket replacements for products which no lon-
a current brand of canned fish may consist of certain types ger exist and disappear from the shop shelves is presented
of fish (mackerel, salmon, or tuna) and then a new variety in Chapter 7.
for one of the canned fish is introduced which is packed in
• As a supplement to the sample by adding a new variety
water rather than oil.
or making a targeted replacement to drop an old vari-
1.133 The focus is on keeping the basket of goods and
ety and add a new one. This represents a more proactive
services that are priced up to date and relevant. It covers
approach to product substitution. Again, it is normally
both truly and completely new products, that is, those which
associated with evolutionary products. The CPI collec-
are revolutionary, as well as those which are evolutionary
tion procedures instruct price collectors to replace the
and also goods or services previously provided free and thus
old, less popular variety of canned tuna fish in oil, with
previously excluded from the CPI. It does not deal directly
the new, more frequently bought tuna packed in water
with substitution and quality change when a good or service
even though the former remains available. This is differ-
unexpectedly disappears. This is the subject of Chapter 7.
ent from the standard reactive approach of replacing the
old disappearing variety with the new one because the old
variety still exists and may not be discontinued although
Planning for the Introduction may be increasingly difficult to find (and less popular/
of New Products lower turnover shares). The new variety is supplementary
1.134 There are three sets of circumstances in which to the old variety and begins to gain market share while
new goods and services are included in the CPI: the old variety declines in market share. This more proac-
tive approach requires the price statistician to monitor the
• As “replacements” for products which no longer exist.
market for the entry of new varieties and to get a sense of
This is normally associated with evolutionary products.
their popularity with consumers, for example, by notic-
Producers often discontinue old versions of their prod-
ing the changing proportions of shelf space occupied by
ucts and introduce new versions that are quite similar but
the different varieties or by talking to the shopkeeper. The
may be of a different quality at a different price. Note
head office can also help by gathering sales information
that “quality change” includes changes in technical speci-
from other sources.
fications as well as more clearly visible outward changes
in design. This can happen frequently and is usually • As a planned introduction of a revolutionary product
unplanned for in a CPI, although not necessarily unex- which consumers begin to buy so that the product has an
pected. It is often associated with forced replacements increasing share of the market. The appearance of revo-
when collectors go to price a product only to find that it is lutionary products in the marketplace and consumer reac-
no longer sold. In the CPI, collection procedures usually tion to them, as measured by sales, are less predictable
instruct price collectors to replace the old versions (mod- than for evolutionary products. Revolutionary products
els) with: also tend to have different price trends from other prod-
ucts in the sample and can therefore exert an influence on
• The most similar model. For example, when the old
the CPI disproportionate to actual sales. For these reasons,
model of washing machine is discontinued, the price
revolutionary products are important, represent a signifi-
collector is instructed to replace it with another model
cant challenge, and warrant special attention, requiring the
which has similar (though probably not identical) spec-
price statistician to be particularly attentive and proactive.
ifications and to record any changes in characteristics
(specifications) to aid the evaluation of potential qual- 1.137 The previous circumstances can be managed
ity differences. either in a planned way, as part of a regular process of
• Alternatively, replacements can be products that are updating the CPI basket, including chain linking (covered in
currently the most popular with consumers. This rep- Chapter 9) or in an ad hoc way when the need arises, or the
resents a deliberate attempt to refresh the CPI basket circumstances warrant action to be taken.
when a replacement has to be made. For example, the
current varieties of canned fish may include tuna. Pro- Timing of the Introduction of
ducers may have introduced a new variety that contains New Products
tuna packed in water rather than oil, and consumers are
1.138 The timing of introduction of new products can
now shifting their buying patterns to purchase more of
vary by the type of product and may be dictated by the
the new variety. There is no external factor forcing the
method of incorporation into the index. For revolutionary
consumer to change to the new product.
products, it can be particularly critical to the accuracy of the
1.135 In some instances, when a model ceases to be index as there is a greater potential for introducing bias if
produced, the manufacturer will indicate which model is the these products are ignored. This is less likely to be the case
replacement and the CPI collection procedures instruct the for evolutionary goods.
price collector to start collecting the price of this replace- 1.139 Chapter 7 presents two alternative strategies that
ment. This also normally contributes to the replenishment of can be adopted when choosing a forced replacement, already
20
mentioned in the previous paragraphs—replacing with the does not include the new product class is still being used.
most similar product or replacing the missing item or variety Because it is not known exactly which other expenditures
with the currently most representative one. The strategy of are being reduced relatively as the new product is purchased,
replacement with the most popular is more likely to main- this rescaling of weights is somewhat arbitrary and may lead
tain the relevance of the sample by going through a process to credibility issues.
of resampling and identifying an appropriate replacement
product. In order to make a properly judged decision on
which strategy to follow and to inform the choice of replace- Calculating the Consumer
ment product and timing of introduction, the NSO needs to Price Index
be aware of current consumer market trends, including what 1.143 The calculation of CPIs usually proceeds in two
new products are becoming popular and what supplemen- stages. First, price indices are estimated for the elementary
tary products are being introduced. It also needs to monitor expenditure aggregates, or simply elementary aggregates.
product turnover, which can be an indication of the rate of Then these elementary price indices are averaged to obtain
product development associated with evolutionary products, higher-level indices using the relative expenditure values of
and can vary between different categories of products. This the elementary aggregates as weights.
information can be obtained from data gathered by price
collectors and their supervisors, commodity experts in the
NSO (for instance, working on the CPI or the producer price Elementary Price Indices
index) or from trade journals, products identified through 1.144 The weights used in the CPI are generally derived
scanner data, and consumer reports. from the HBS at levels that are typically for an item group-
1.140 Evolutionary products should be included in the ing such as cheese, butter, or milk. There is no identifica-
sample as soon as it is clear that consumers are shifting to tion of the specific variety of the product and an associated
these new products from the old versions. A frequent updat- weight. NSOs select a sample of individual varieties to rep-
ing of the basket reduces the need for the ad hoc introduc- resent each item, but there are often no weights available
tion of evolutionary products. at the variety level. The NSOs then use some method of
1.141 Revolutionary products usually first appear in averaging to produce an average price or an average price
the marketplace at a high initial price to cover development change to use in deriving the item or elementary index. This
costs and to exploit the novelty value to the consumer. The level of computation is usually referred to as an elementary
prices generally start to decline as they become more estab- aggregate because it is the first level at which an index is
lished and competing varieties enter the market resulting in compiled for aggregation to higher levels of the CPI.
increased supply. The timing of introduction into the CPI 1.145 When weights are not available, the choice of the
basket is a critical issue—if introduced too late it will not averaging method can be very important. Chapter 6 of Con-
only reduce the representativity of the CPI basket but could sumer Price Index Theory shows that the larger the variation
also give too much weight to any price decline associated in the individual prices, the larger the difference among the
with obsolescence of the product it is replacing when it nears standard averaging methods. Both arithmetic and geometric
the end of its life cycle but is still in the CPI basket. For rev- averaging can be used, but as demonstrated in Chapter 8 of
olutionary products the timing of their introduction into the this Manual and Chapter 6 of Consumer Price Index Theory,
CPI is important. In practice, often they are not introduced geometric average formulas are recommended.
until they can be included in a new basket at the time of a
CPI revision. This can lead to out-of-date and unrepresenta- Arithmetic Average
tive baskets if the revisions are carried out infrequently or 1.146 The two methods used historically by NSOs to cal-
with a long time lag, for instance as a result of delays in culate the elementary indices are the ratio of average prices,
processing HBS data. But the price statistician is also con- known as the Dutot index, or the average of price relatives,
fronted with uncertainty; it is not always clear how the retail known as the Carli index. Each of these formulas can be
market will react in the longer term to the introduction of calculated using either the long-term price relative formula
a revolutionary product—some will be highly successful, (comparing current to base period prices) or the short-term
achieving significant sales volumes and market stability in a price relative formula (comparing current to previous period
relatively short time, while others may achieve high sales at prices). The short-term versions of these formulas calculate
an early stage which are not maintained. a long-term relative change by chaining together the short-
1.142 Methods of introduction which overcome the term price relatives. Appendix 6 illustrates the mathematical
problem of lack of timeliness include sample supplemen- formulas for the Dutot and Carli indices. Chapter 8 provides
tation, targeted replacement procedures, and reinitiating details on the use of these formulas.
(or rotating) the sample for the elementary aggregate or 1.147 As shown in Chapter 8, it should be noted that the
COICOP class. These methods are discussed in detail in chained Carli produces different results than those for the
Chapter 7 and, sample reinitiation apart, are generally appli- fixed-base Carli using the average of long-term price relatives.
cable to evolutionary products. For revolutionary products, The chained Carli price index has a definite upward bias. This
a new elementary aggregate must often be created. Frequent chained version of the Carli index should not be used by NSOs
updating of the CPI basket reduces the potential problems for calculating elementary-level indices in the CPI.
and the introduction of revolutionary new products at the
time of a basket update has a number of operational advan- Geometric Averages
tages: namely, the old weights do not need to be rescaled if 1.148 With the introduction of the CPI Manual in 2004,
a new product class is introduced when an old basket which a major emphasis was placed on using geometric averaging
21
when weights are not available for the individual prices in make a rational choice between the various possibilities
the CPI elementary indices. The geometric price index is means having a clear idea of the target index that would be
known as the Jevons price index and is calculated either as preferred in principle. The target index can influence practi-
the ratio of the geometric average prices or as the geometric cal matters such as whether the weights used in the index
average of the price relatives. The Jevons formula can be should be price updated.
found in Appendix 6, and Chapter 8 provides more detail on 1.153 A comprehensive, detailed, rigorous, and up-to-
using the Jevons. date discussion of the relevant index number theory is pro-
1.149 The Jevons index generally provides different vided in Chapters 2–10 of Consumer Price Index Theory.
estimates than both the Dutot or the Carli. Whether using Most of the standard indices used to compile CPIs (Laspey-
the long-term price relative method or the short-term rela- res, Lowe, and Young) have known biases. The next para-
tive method Jevons indices yield the same index numbers graphs include summaries of the different formulas that can
as shown in Chapter 8. The same property holds true for the be used. Chapter 8 provides more details on these formulas
long-term and short-term Dutot indices. This property does that can be used to calculate the CPI.
not hold true for the Carli index. The chained short-term
Carli index is always equal to or greater than the long-term Price Indices Based on Baskets of Goods and
Carli index. Services
1.150 Chapter 6 strongly encourages the use of the 1.154 The purpose of an index number may be explained
Jevons price index for calculating elementary indices where as comparing the values of households’ expenditures on
weights are unavailable. It notes that the Dutot price index consumer goods and services in two time periods. Knowing
should only be used in cases where the sample of transac- that expenditures have increased by 5 percent is not very
tions is homogeneous with respect to base price levels or informative if one does not know how much of this change
price trends. It strongly discourages the use of the short-term is attributable to changes in the prices of the goods and ser-
Carli price index because of its known upward bias. The vices, and how much to changes in the quantities purchased.
short-term method for the Jevons index will easily accom- The purpose of an index number is to decompose propor-
modate replacement varieties or adjustments for quality tionate or percentage changes in value aggregates into their
changes. As mentioned earlier, the NSO will only need to overall components of price and quantity change. A CPI is
collect prices for the current and previous periods to enter in intended to measure the price component of the change in
the system. If the long-term method is used, quality adjust- households’ consumption expenditures. One way to do this
ments will involve changing the base price of the transaction is to measure the change in the value of an aggregate, hold-
for the value of the quality change as shown in Chapter 8. ing the quantities constant.
Lowe Indices
Choice of Higher-Level Index Number 1.155 One very wide, and popular, class of price indi-
Formula ces is obtained by defining the index as the relative change,
1.151 The NSO has to decide on the kind of index between the periods compared, in the total cost of purchas-
number to use. The extensive references dealing with index ing a given set of quantities, generally described as a “bas-
theory in the bibliography of this Manual reflect the fact that ket.” The meaning of such an index is easy to grasp and to
there is a very large literature on this subject. Many kinds explain to users. This class of index is called a Lowe index,
of mathematical formulas have been proposed over the past after the index number pioneer who first proposed it in 1823
two centuries. While there may be no single formula that (see Chapter 2 of Consumer Price Index Theory). Most
would be preferred in all circumstances, most economists NSOs make use of some kind of Lowe index in practice.
and compilers of CPIs seem to agree that, in principle, The Lowe index formula can be found in Appendix 6 and
the index formula should belong to a small class of indi- is described in more detail in Chapter 8. Lowe indices are
ces called superlative indices. A superlative index may be widely used for CPI purposes.
expected to provide an approximation to a COLI. A char- 1.156 In principle, any set of quantities could serve as
acteristic feature of a superlative index is that it treats the the basket. The basket does not have to be restricted to the
prices and quantities in both periods being compared sym- quantities purchased in one or other of the two periods com-
metrically. Different superlative indices tend to have similar pared, or indeed any actual period of time. The quantities
properties, yield similar results, and behave in very similar could, for example, be arithmetic or geometric averages of
ways. Because of their properties of symmetry, a superlative the quantities in the two periods. For practical reasons, the
index is also likely to be seen as desirable even when the basket of quantities used for CPI purposes usually has to
CPI is not meant to be a COLI. be based on a survey of household consumption expendi-
1.152 When a monthly or quarterly CPI is first pub- tures conducted in an earlier period than either of the two
lished, however, it is invariably the case that there is not suf- periods whose prices are compared. For example, a monthly
ficient information on the quantities and expenditures in the CPI may run from January 2018 onward, with January 2018
current period to make it possible to calculate a symmetric, = 100, but the quantities may be derived from an annual
or superlative, index. In cases where the prices and quanti- expenditure survey made in 2015 or 2016, or even spanning
ties are available, care must be taken to use index methods both years. As it takes a long time to collect and process
that do not result in biased estimates (for example, chain expenditure data, there is usually a considerable time lag
drift). Such methods are discussed in detail in Chapters 2–4 before such data can be introduced into the calculation of
of Consumer Price Index Theory. While it is necessary to CPIs. The basket may also refer to a year, whereas the index
resort to second-best alternatives in practice, being able to may be compiled monthly or quarterly.
22
1.157 The index can be written, and calculated, in two for different approaches for the modified versions of the
ways: either as the ratio of two value aggregates, or as an Lowe and Young indices.
arithmetic weighted average of the price ratios, or price
relatives, pit/pi0 (where pit refers to the price of the item in Geometric Young, Laspeyres, and Paasche Indices
the current period and pi0 refers to the price of the item in 1.162 In the geometric version of the Young index, a
the reference period) for the individual products using the weighted geometric average is taken of the elementary
hybrid expenditure shares si0b as weights. The expenditures aggregate price relatives using the expenditure shares of
are described as hybrid because the prices and quantities period b as weights. The geometric Laspeyres is the spe-
belong to two different time periods, 0 and b, respectively. cial case in which b = 0; that is, the expenditure shares are
The hybrid weight may be obtained by updating the actual those of the price reference period 0. Similarly, the geomet-
expenditure shares in period b, namely, pibqib/Σ pibqib, for the ric Paasche uses the expenditure shares of period t. It should
price changes occurring between periods b and 0 by multi- be noted that these geometric indices cannot be expressed
plying them by the relative prices, pi0/pib. as the ratios of value aggregates in which the quantities are
fixed. They are not basket indices, and there are no counter-
Laspeyres and Paasche Indices part Lowe indices.
1.158 Any set of quantities could be used in a Lowe 1.163 The geometric Young and Laspeyres indices have
index, but there are two special cases which figure very the same information requirements as their arithmetic coun-
prominently in the literature and are of considerable impor- terparts. They can be produced on a timely basis. Thus, these
tance from a theoretical point of view. When the quanti- geometric indices must be treated as serious practical pos-
ties are those of the price reference period, the Laspeyres sibilities for purposes of CPI calculations. As explained in
index is obtained. When quantities are those of the other Chapter 8, the geometric indices are less likely to be subject
period, the Paasche index is obtained. Appendix 6 provides to different types of index number biases than their arith-
the mathematical formulas for the Laspeyres and Paasche metic counterparts. Their main disadvantage may be that,
indices. because they are not fixed-basket indices, they are not so
easy to explain or justify to users.
Young Index
1.159 Instead of holding constant the quantities of Symmetric Indices
period b, an NSO may calculate a CPI as an arithme- 1.164 The standard price index methods used in most
tic weighted average of the individual price relatives, countries today, that is, the Lowe and Young indices,
holding constant the expenditure shares of period b. date back 90 years to those proposed by W. C. Mitchell
The resulting index is called a Young index, again after (1927) and G. H. Knibbs (1924). Index number theory has
another index number pioneer. The Young index formula advanced substantially, particularly in the past 30 years, to
can be found in Appendix 6 and is described in more provide better information on what the target index number
detail in Chapter 8. formula should be. Various approaches have been used to
1.160 Whether to price update or not, and the resulting evaluate index number formulas and derive those best suited
choice of index, is discussed in more detail in Chapter 9 (see for inflation measures. The research presented in Consumer
also Chapter 3 of Consumer Price Index Theory). Price Index Theory has resulted in improvements for fixed-
basket formulas and identified target indices that are sym-
Short-Term Price Index Formulas metric averages of standard formulas. The target indices are
1.161 Many countries use a modified version of the the Fisher, Törnqvist, and Walsh price indices, discussed in
Lowe or Young index that compiles the index based on detail in Chapter 8. However, these usually cannot be pro-
short-term price changes rather than the long-term price duced in final form except with a lag because they require
changes. This modified method can be compiled in two both current and past weight information. Thus, most NSOs
ways. Using the first approach, index compilation involves use the fixed-basket indices where the weight data are
a two-step estimation process that breaks down the price derived from some past period. An exception can occur in
movements into short-term, period-to-period changes that countries where scanner data are available and symmetric
are used to bring forward the index from the previous indices can be produced in real time using the methods pre-
period. In the second approach, elementary-level indices sented in Chapter 10.
are compiled based on chained short-term price changes and 1.165 A symmetric index is one that makes equal use of
the calculation of upper-level indices uses the base period the prices and quantities in both the periods compared and
weights. There is no preference as to which approach is treats them in a symmetric manner. There are three sym-
preferred. Countries can decide which approach should be metric indices that are widely used in economic statistics;
applied. The use of the short-term formula makes it easier these three indices are also superlative indices, the Fisher,
for NSOs to introduce replacement items in the sample if the Walsh, and Törnqvist. The formulas for each are provided in
ones they have been tracking are no longer available. The Appendix 6, and Chapter 8 provides details on using each of
short-term approach also enables the NSOs to make quality these index formulas.
adjustments as improvement (or deterioration) is made to 1.166 Different formulas are used at different stages of
the sampled varieties. The NSO only needs to collect the aggregation. At the elementary or first stage, where prices
current and previous prices for the item in order to introduce are first combined to form an index, many countries will not
it into the index. In using the long-term method, the base use weights. At the second and higher levels, weights are
price will need to be adjusted for the changes in the quality applied, but these weights generally relate to some period in
of the items in the sample. Chapter 8 provides the formulas the past that becomes less representative with the passage of
23
time. When compared to the target indices (Fisher, Walsh, 1.171 For infrequently updated CPIs, a single year
or Törnqvist), it becomes apparent that the indices produced is preferred as the price reference period. Where a single
in practice are of substantially lower quality than the target month (or quarter) is used, the prices of some seasonal
indices. This Manual discusses these issues thoroughly and products will be unavailable or unusually high or low and
provides approaches that countries can implement over time many unusual or imputed prices may have to be used for the
to move closer to the target measures. price reference period. For countries with infrequent weight
updates, it is preferable that the price reference period is a
whole year in which seasonal prices would be appropriately
Updating the Weights and Linking represented. In some months there will be no sales of, for
of Series example, seasonal fruit, but an average price for the whole
1.167 Chapter 4 notes that over time the CPI weights year would still be available for the price reference period.
and basket become less representative of the consumer mar- 1.172 The index reference period should be one year.
ket. The weights and basket should be updated at least once Using a single month or quarter to serve as the index refer-
every five years to maintain their relevance. Many countries ence period (= 100), can result in distorted index changes
strive to conduct an HBS on a five-year cycle to use for because of the unusual or imputed prices in that period.
updating the CPI basket and provide detailed information 1.173 Chained CPIs weights are updated on an annual
on household expenditures for use in the national accounts. basis. Because annual updating allows for a relatively small
From the HBS consumption estimates, NSOs will identify lag between the weight and price reference periods, a single
the most important items to use in the new CPI basket as month is used as the price reference period. There is a con-
discussed in Chapter 3. Deriving the new basket involves tinuing flow of price data that may include imputations, and
adding items that have gained importance since the previous a relatively small number of changes in specifications or
HBS and deleting those that are no longer important based products; the major exercise is to introduce the new weights
on their shares in the HBS. into the flow of price data and link this to the existing chain.
1.174 For infrequently rebased CPIs, the use of a single
month as the price reference period is not advised; however,
Introducing New Weights and the it is often the case that the country’s resources only allow for
Consumer Price Index Basket a price reference period of less than a full year. NSOs should
1.168 The new basket is then used to review the sample strive to maximize the number of months used as the price
of items and outlets to ensure that the samples are repre- reference period, with the goal of using a full year.
sentative of products being purchased by consumers and 1.175 A primary shortcoming of using a short reference
the places in which they are purchased. The new weights period for infrequently rebased CPIs is that out-of-season
and sample are used to start a new CPI with updated weight items in the price reference period will have no observed or
and price reference periods. Typically, the weight reference economically meaningful price. The decision as to which
period precedes the price reference period. The NSO must month to use for the price reference month should consider
decide how the new weights and sample will be introduced when seasonal items with relatively high weights are in
in the CPI. It also must decide on the price reference period, season. If these items are not in season, an imputed price
if it will differ from the weight reference period. will have to be used and consideration should be given to
1.169 If the price reference period and weight refer- the validity of imputation methods for out-of-season items
ence period are the same (Laspeyres), then the new sample in this context. For example, if the carryforward method
is used to compile the CPI directly. If the two periods differ, (which is discouraged) is used and the reference month
most likely with the weight period preceding the price refuses an imputed price for the out-of-season item, the index
erence period, there are two main options for updating the may be unduly low. As mentioned previously, the two-stage
CPI weights. One approach is to price update the weights to Laspeyres aggregation is preferred since it avoids the need
the price reference period (Lowe index) to keep the implied for long-term price comparisons with this one-month price
quantities fixed at the weight reference period levels. The reference period.
second approach is to keep the expenditure shares fixed 1.176 A primary shortcoming of using a longer price
(Young index). reference period are the resources needed to collect a full
1.170 Chapter 9 examines the different approaches to year of prices for all items included in the CPI basket. Some
price updating the weights (paragraphs 9.6–9.10) and dis- countries struggle with the resources to collect prices for
cusses the recent research on price updating (paragraphs two baskets (old and new) simultaneously. To minimize the
9.11–9.18). A key issue in the decision is whether there has burden, some countries have begun looking at preliminary
been substantial price change between the weight reference expenditure data on a quarterly basis during the HBS col-
period and the price reference period when the weights are lection period and to identify any new items that may be
introduced into the CPI. Generally, if there is a substantial introduced into the basket. Once new products have been
price change between the price and weight reference peri- identified, prices can be collected.
ods, the weights should not be price updated. As noted previ-
ously, price updating the weights assumes that the quantities
have remained fixed. If prices have changed substantially, Linking Previous Consumer Price Index
this assumption is less likely to be true. It would be more to the New Price Index Reference Period
realistic to assume the expenditure shares have remained 1.177 The NSO may choose to start the new series using
fixed, in which case the price change is offset by a compen- the new price reference period as the new index reference
sating quantity change. period. In Chapter 9 the recommendation is that when new
24
weights are introduced there should be an overlap period for Products may disappear and have to be replaced by others,
the two indices so that they can be linked together. The over- but it may also be appropriate to drop some products before
lap period is used to develop adjustment factors that may they disappear altogether if they have become unrepresenta-
be applied to the old series to bring it to the same level as tive. Price collectors need to be provided with appropriate
the new series. The linking of the old and new index series training and very clear instructions and documentation about
creates a continuous time series of data, which users need. how to proceed. Clear instructions are also needed to ensure
1.178 At minimum, a single common period is required that price collectors collect the right prices when there are
as an overlap period between the indices. While a single sales, special offers, or other exceptional circumstances.
overlap period can be used when the CPI is updated annually, 1.183 The price data collected must also be subjected
this is not the preferred method when the index is updated to careful checking and editing. Many checks can be car-
less frequently. In the case of infrequent updates, an annual ried out automatically, using standard statistical control
overlap is preferred. Some NSOs update the CPI weights methods. It may also be useful to send out auditors to
each year so that the time lapse between the weight refer- accompany price collectors and monitor their work. The
ence period and the link month is short. The single period various possible checks and controls are explained in detail
link could be used in these instances. Chapter 9 provides a in Chapter 13.
detailed discussion of annual weight updates. 1.184 The head office staff also need to be trained on
1.179 Most NSOs will establish a new index reference index methods and procedures for review of collected data,
period using an annual average from a previous year. The imputation of missing data, quality-adjustment procedures
simplest and easiest method for users is to link the series when needed, as well as index compilation and dissemina-
with data for the month preceding the introduction of the tion processes. Improvements in information technology
new series (link month). This involves re-referencing the old should obviously be fully exploited. New and more efficient
series at each level to the annual average index for the new computers and database applications are continuously being
price reference period. However, there will be a discontinu- developed. As resources permit, new technologies and orga-
ity between the index level for the new index and that for nization improvements should be implemented.
the re-referenced index level for the old series in the link 1.185 Staff training and process reviews are an essen-
month. This reflects the difference in price trends between tial part of continuous quality improvement. Staff should
the old and new series as the new weights are being intro- receive regular training in their discipline and should be
duced. There are three steps involved to link this difference: invited to operational reviews where all team members can
raise concerns and, where appropriate, tackle specific issues
• Re-reference the old index series to the new index refer-
through individual or group training.
ence period
• Compile the new index series in the link month using the
new weight structure Publication and Dissemination
• Link the new series to the old series by using forward 1.186 As noted previously and in Chapter 2, the CPI is
linking factors or, if using the short-term price relative an important statistic whose movements can influence the
method, start the new series indices at the level of the old central bank’s monetary policy, fiscal policy, and the national
series in the link month budget; affect stock markets; influence wage rates and social
security payments; and so on. There must be public confi-
1.180 If the NSO or users want to continue the old CPI
dence in its reliability, and in the competence and integrity
series for future periods in time, they can produce a set of
of those responsible for its compilation. The methods used
forward linking factors to use in future months as the new
to compile it must therefore be fully documented, transpar-
CPI is released. The forward linking factor raises the level of
ent, and open to public scrutiny. Many countries have an
the new CPI series to that of the old series thus keeping the
official CPI advisory group consisting of both experts and
series on the old reference period.
users. The role of such a group is not just to advise the NSO
on technical matters, but also to promote public confidence
in the index.
Organization and Management 1.187 Users of the index also attach great importance
1.181 The production of the CPI is a complex operation to having the index published as soon as possible after the
involving extensive fieldwork by data collectors; process- end of each month or quarter, preferably within two or three
ing, review, and editing of the collected data; compilation weeks. There are also many users who do not wish the index
of indices; and their dissemination to the public. The whole to be revised once it has been published. Thus, there is likely
process requires careful planning and management to ensure to be some trade-off between the timeliness and the quality
that the data products conform to good management and of the index.
statistical practice. Appropriate management procedures for 1.188 Publication should be understood to mean the dis-
the CPI are described in Chapter 13. semination of the results in any form. Most countries do not
1.182 Price collectors should be well trained to ensure release their CPI in print, or hard copy. NSOs now tend to
that they understand the importance of selecting the right release the CPI electronically and make it available through
products for pricing. Inevitably, price collectors are bound to the internet on their website.
use their own discretion to a considerable extent. As already 1.189 As explained in Chapter 14, good publication
explained, one issue of crucial importance to the quality and policy goes beyond timeliness, confidence, and transpar-
reliability of a CPI is how to deal with the slowly evolving ency. The results should be made available to all users, in
set of products with which a price collector is confronted. both the public and the private sectors, at the same time and
25
according to a publication schedule announced in advance. internet purchases, second-hand goods, own-account pro-
There should be no discrimination among users in the timing duction, tariffs, transport services, health, education, social
of the release of the results. The results should not be subject protection, and financial services.
to governmental scrutiny as a condition for their release and 1.192 Wherever possible, the chapter identifies the
should be seen to be free from political or other pressures. preferred approach for the treatment of each special case;
1.190 There are many decisions to be taken about the however, currently there is no preferred approach for the
degree of detail in the published data and the different ways treatment of owner-occupied housing services. The sec-
in which the results may be presented. Users need to be con- tion on housing in Chapter 11 identifies the different pos-
sulted about these questions. These issues are discussed in sible methods for the treatment of owner-occupied housing
more detail in Chapter 14; however, it is recommended to and describes the advantages and disadvantages of each
provide users with detailed data that are presented in a long method.
time series. Detailed indices should include detailed item
and area indices. Errors and Biases
1.193 The CPI, like all other statistics, may be subject
Special Cases to general error that may occur during any stage of the esti-
1.191 Certain products and issues have proven to be mation process but also errors that are unique to the CPI (for
challenging for compilers with regard to both develop- example, substitution bias and quality change bias). Chap-
ing weights and collecting prices. Chapter 11 focuses on ter 12 describes not only the different types and sources of
selected special cases and provides detailed advice for some potential errors, but also potential biases and their sources.
of the more problematic products and issues compilers face. Finally, Chapter 12 provides insight into how to address
These include the treatment of seasonal items, housing, these errors and biases.
26
Annex 1.1
Formula Notations
I 0:t Index
0:t
I EA Elementary index
IC
0:t
Carli index
I D0:t Dutot index
I J0:t Jevons index
I Jc0:t Chained Jevons index
0:t
I Dc Chained Dutot index
0:t
I Cc Chained Carli index
IL
0:t
Laspeyres index
0:t
I Lo Lowe index
IY
0:t
Young index
I ge0:t Geometric index
0:t
I GL Geometric Laspeyres index
0:t
I GLo Geometric Lowe index
I GY
0:t
Geometric Young index
0:t
I MY Modified Young index
0:t
I MLo Modified Lowe index
0:t
I HR Harmonic mean price relatives
0:t
I RH Harmonic mean prices
I LM Llyod-Moulton index
0:t
ln p Log of price
n Sample size
N Population size
p Price
p̂ Quality adjusted price
P Price relative
pi0 Base price observed for variety i
pit Price in current period for variety i
pit−1 Price in previous period for variety i
q Quantity
sit Percentage shares in period t
t Current period
t −1 Previous period
T End of index link
w Weight
wib Weight of the item in the weight reference period
wib ( t ) Weight price updated from the weight reference to period t
wagg b (t )
Aggregate weight price updated from the weight refer-
ence period to period t
27
USES, CONCEPTS, SCOPE, AND
CLASSIFICATIONS OF CONSUMER
PRICE INDICES 2
Introduction who may view consumer inflation as a pervasive phenom-
enon affecting all households similarly. The coexistence
2.1 Chapter 2 begins with an overview describing the of alternative measures could undermine their credibility
uses of the consumer price index (CPI). The primary uses for many users. Additionally, given budgetary constraints,
drive the decisions regarding the concepts and scope for the many national statistical offices (NSOs) cannot compile
index. As described in this chapter, countries face the real- more than one index. They face the reality of having to com-
ity of having to compile and disseminate a general-purpose pile a single general-purpose index.
index that tries to meet a variety of user needs. The System 2.5 This section describes the most important uses
of National Accounts 2008 (2008 SNA) provides the general for CPIs and indicates how the coverage of a CPI can be
framework for the concepts used to compile the CPI. The affected by the use for which it is intended. The question
chapter explores each of these concepts in detail and how of the appropriate coverage of a CPI must be addressed
they are applied to the CPI. The chapter concludes with a before defining the appropriate methodology to be used. As
discussion of the recommended classification for the CPI. described in the following text, whether the theoretical basis
The Classification of Individual Consumption According of a CPI is intended to be a cost of living index (COLI), a
to Purpose (COICOP) serves as the internationally recom- cost of goods index, or a general-purpose index, it is neces-
mended classification system for developing weights and sary to determine exactly what kinds of goods and services
compiling CPIs. The previous version of COICOP, adopted and what types of households are meant to be covered. This
in 1999, was updated and has been released as COICOP can only be decided based on the main uses of the index,
2018. The classification section highlights the key changes which will drive decisions on the concepts and methods
between COICOP 2018 and the previous version, and iden- used.
tifies some important issues for consideration as countries 2.6 Where only one CPI is published, it is the main use
begin implementing the new classification system. that should determine its type and scope. If there are sev-
eral major uses, compromises may have to be made about
Consumer Price Index Uses how it is constructed. The purpose of a CPI should influ-
ence all aspects of its construction. CPI producers need to
2.2 The CPI represents a key indicator of economic per- know how the index is being used if they are to ensure that
formance in most countries. This section will focus on why it is fit for purpose. In this connection user, consultation is
CPIs are compiled and their uses. important.
A Range of Possible Consumer Consumer Price Indices and General

Price Indices Inflation
2.3 As noted in Chapter 1, compilers have to consider 2.7 Measures of the general rate of inflation in the econ-
the needs of users in deciding on the group of households omy are needed for various purposes:
and the range of consumption goods and services covered
• Controlling inflation is usually one of the main objec-
by a CPI. As the prices of different goods and services do
not all change at the same rate, or even move in the same tives of government economic policy, although respon-
direction, changing the coverage of the index will change sibility for controlling inflation may be delegated to the
the results obtained. Thus, there can be no single, unique central bank. A measure of general inflation is needed to
CPI that meets all needs, and a range of possible CPIs could set targets and to assess the degree of success achieved by
be defined. the government or central bank in meeting anti-inflation
2.4 While there may be interest in a broadly defined targets.
CPI, covering all the goods and services consumed by all • A measure of general inflation is also needed for both
households, there are many other options for defining CPIs business and national accounting purposes, particularly
that cover specific sets of goods and services, which may for current purchasing power accounting.
be more useful for particular analytic or policy purposes. • The concept of a relative price change is important in
When only a single CPI is compiled and published, there economics. It is convenient therefore to be able to mea-
is a risk that it may be used for purposes for which it is not sure the actual changes in the prices of individual goods
appropriate. More than one CPI could be published to meet or services relative to some measure of general inflation.
different analytic or policy needs but each index must be There is also a need to measure neutral holding (or capi-
clearly defined and explained by metadata. Otherwise, the tal) gains and losses on assets, including monetary assets
publication of more than one CPI can be confusing to users and liabilities.
29
2.8 A CPI is not a measure of general inflation, as it only inflation, the use of indexation may become widespread
measures changes in the prices of consumer goods and ser- throughout the economy.
vices purchased by households. A CPI does not cover capital 2.13 The objective of indexation of money incomes
goods, such as houses, or the goods and services consumed may be either to maintain the purchasing power of those
by enterprises or the government. Any analysis of inflation incomes with respect to certain kinds of goods and services
pressures in the economy must also take into account other or to preserve the standard of living or economic well-being
price movements, such as changes in the prices of imports of the recipients of the incomes. These two objectives are not
and exports, the prices of industrial inputs and outputs, and quite the same, especially over the longer term. Maintaining
also asset prices. purchasing power may be interpreted as changing money
2.9 However, even though the CPI does not measure income in proportion to the change in the monetary value
general inflation, its movements may be expected to be of a fixed basket of goods and services purchased with that
highly correlated with those of a more general measure, income. Maintaining the purchasing power of income over a
because household consumption expenditure represents fixed set of goods and services does not imply that the stan-
a large proportion of total final expenditure. In particular, dard of living of the recipients of the income is unchanged.
the CPI should provide a reliable indicator of whether infla- 2.14 When the indexation applies to monetary assets or
tion is increasing or decreasing and also of possible turn- liabilities, it may be designed to preserve the real value of
ing points in the rate of inflation. This information is highly the asset or liability relative to other assets or relative to the
valuable even if the CPI may be systematically understating values of specified flows of goods and services.
or overstating the general rate of inflation.
The Type of Index Used for Indexation
Use of the Consumer Price Index for 2.15 When income flows such as wages or social secu-
Monetary and Economic Policy Purposes rity benefits are index-linked, it is necessary to consider the
2.10 The CPI is a key macroeconomic indicator and implications of choosing between a COLI and a price index
widely used for assessing economic policy, for monetary that measures the changes in the cost of purchasing a fixed
policy purposes, and for macroeconomic planning. It is also basket of goods and services, a type of index described here
commonly used by governments and central banks to set as a Lowe index. The widely used Laspeyres and Paasche
inflation targets. As part of this, some countries produce indices are examples of Lowe indices. The Laspeyres index
measures of “core” or “underlying” inflation, based on the uses a typical basket purchased in the earlier of the two peri-
CPI or selected subindices of the CPI. In a few countries, ods compared, while the Paasche uses a basket typical of the
the central bank produces measures of core inflation. Cen- later period. In contrast, a COLI compares the cost of two
tral banks require access to detailed weight, index, and baskets that may not be exactly the same, but which bring
price data for these calculations, and the NSO should pro- the same satisfaction or utility to the consumer.
vide these data. Central banks calculate core inflation using 2.16 Indexation using a Laspeyres price index will
a variety of different methods that require access to these tend to overcompensate the income recipients for changes
detailed data. For those NSOs that compile core inflation, in their cost of living. Increasing incomes in proportion
the most common method used is known as the exclusion to the change in the cost of a basket purchased in the past
method. Under this method, the weights and prices of those ensure that the income recipients have the opportunity to
items deemed volatile (that is, susceptible to short-term continue purchasing that same basket if they wish to do so.
shocks) are excluded from the index. The items excluded They would then be at least as well off as before. However,
can include items such as fresh fruit, vegetables, fish, meat, by adjusting their pattern of expenditures to take account
fuel, and so on. The NSO should work closely with the of changes in the relative prices of the goods and services
central bank to define the volatile items. The list of vola- they buy, they would be able to improve their standard of
tile items should be reviewed and updated as needed on a living or economic well-being because they can substitute
regular basis. When publishing the core inflation measure, goods and services that have become relatively cheaper
the NSO or central bank should fully explain the methods for ones that have become relatively dearer. In addition,
used and define the purpose and uses for these data to avoid they may be able to start to purchase completely new kinds
confusing users. of goods that provide new kinds of benefits that were not
2.11 For monetary policy purposes, flash estimates of available in the earlier period. Such new goods and ser-
the CPI may be produced, which are released before the offi- vices tend to lower a COLI when they first appear even
cial CPI and give early warnings or signals about the devel- though no price can actually be observed to fall, as there
opment of consumer price inflation. More information on was no previous price.
related or alternative measures can be found in Chapter 14.
Indexation of Wages and Pensions
2.17 As noted in Chapter 1, the indexation of wages was
Indexation the main objective behind the original compilation of CPIs,
2.12 With indexation, the monetary values of certain although there has always been a general interest in measur-
payments, or stocks, are increased or decreased in propor- ing inflation. If the indexation of wages and pensions is the
tion to the change in the value of a specified price index. main justification for a CPI, it has direct implications for the
Indexation, most commonly applied to monetary flows such coverage of the index. First, it suggests that the index should
as wages, pensions, social security benefits, rents, interest, be confined to expenditure made by households whose prin-
or taxes, may also be applied to the capital values of certain cipal source of income is either wages or pensions. Second,
monetary assets and liabilities. Under conditions of high it may suggest that expenditure on certain types of goods
30
USES, CONCEPTS, SCOPE, AND CLASSIFICATIONS OF CONSUMER PRICE INDICES
and services, such as luxury items, may be excluded from interest is not to maintain the standard of living of the credi-
the index. See also paragraph 1.35 in Chapter 1. tors but rather to maintain their real wealth by compensat-
ing them for the neutral holding, or capital losses on their
Indexation of Social Security Benefits loans incurred as a result of general inflation. A CPI may not
2.18 It has become common practice in many countries be the ideal index for this purpose but may be used in the
to index the rates at which social security benefits are pay- absence of another more suitable index.
able. There are many kinds of benefits, including retirement 2.25 Many other forms of contractual payments may
pensions, unemployment benefits, sickness benefits, or child be linked to the CPI. For example, legal obligations to pay
allowances. When indexing benefits of this kind is the main alimony or for the support of children may be linked to the
reason for compiling the CPI, it may suggest restricting the CPI. Payments of insurance premiums may be linked either
coverage of the index to certain types of households and to the all-items CPI or to a subindex relating to some spe-
goods and services. Some kinds of goods and services could cific types of expenditure, such as the costs of repairs.
then be excluded from the index based on political decisions
that certain goods and services are deemed undesirable—for Taxation
example, expenditure on alcoholic beverages and tobacco. 2.26 Movements in a CPI may be used to affect the
2.19 Alternatively, separate CPIs for different catego- amounts payable in taxation in several ways. For example,
ries of households could be compiled. For example, an index liability for income tax may be affected by linking per-
may be compiled covering the basket of goods and services sonal allowances that are deductible from taxable income
purchased by households whose principal source of income to changes in the CPI. Under a system of progressive taxa-
is a social security pension. tion, the various thresholds at which higher rates of personal
2.20 As already noted, publishing more than one CPI income tax become effective may be changed in proportion
may be confusing if inflation is viewed as affecting every- to changes in the CPI. Liability for capital gains tax may be
one in the same way. Such confusion can be avoided by pro- reduced by basing it on real, rather than nominal, capital
viding a detailed explanation of the purpose and use of each gains by reducing the percentage increase in the value of
alternative index (as explained in Chapter 14). It is not dif- the asset by the percentage change in the CPI over the same
ficult to explain that price changes are not the same for dif- period. In general, there are various ways in which some
ferent categories of expenditure. In practice, some countries form of indexation may be introduced into tax legislation.
publish more than one index.
2.21 The main reason why it may not be justifiable to Deflation: Changing Current Values
publish more than one index is that the movements in the to Volume Measures
different indices may be virtually the same, especially in the
short term. In such cases, the costs of compiling and pub- 2.27 Price indices, such as the CPI, can be used to
lishing separate indices may not be worthwhile. In practice, remove the effect of changes in prices from current value
differences in the patterns of expenditure between different data or to convert nominal data into data in volume terms
groups of households may not be large enough to yield sig- (2008 SNA, paragraph 2.146). In general, changes in current
nificantly different CPIs. value aggregates of output or expenditure can be decom-
2.22 Finally, it should be noted that the deliberate exclu- posed into changes in price and volume components. Defla-
sion of certain types of goods and services by political deci- tion separates the price effect from the volume changes.
sion on the grounds that the households toward whom the While aggregates of income may not, strictly speaking, be
index is targeted ought not to be purchasing such goods or broken into price and volume components, they may be
services or ought not to be compensated for increases in the calculated in constant purchasing power or to maintain a
prices of such goods and services, cannot be recommended constant standard of living, which is described as being in
because it exposes the index to political manipulation. For real terms. Converting nominal data into data in real terms
example, suppose it is decided that certain products such provides a consistent and meaningful measure across time.1
as tobacco or alcoholic beverages should be excluded from
a CPI. There is a possibility that, when taxes on products Measures of Real Consumption and Income
are increased, these products may be deliberately selected 2.28 Price indices can be used to deflate final consump-
for higher taxes with the knowledge that the resulting price tion expenditure at current prices or money income to derive
increases would not be reflected in the CPI. Such practices measures of consumption expenditure in volume terms (also
are not unknown. known as real consumption) and real income. Real measures
involve volume comparisons over time (or space). There
Indexation of Interest, Rents, and Other Contractual are two different approaches to such comparisons which
Payments are analogous to the distinction between a Lowe, or basket,
2.23 It is common for payments of both rents and inter- index and a COLI.
est, especially mortgage rates, to be indexed to the CPI.
Governments may issue bonds with an interest rate specifi- 1
The SNA distinguishes between volume and real measures. For example,
cally linked to the CPI. The interest payable in any given the 2008 SNA, in paragraph 15.181, notes that many flows in the SNA, such
period may be equal to the fixed rate of interest plus the as cash transfers, do not have price and quantity dimensions of their own
percentage change in the CPI. Payments of housing rents and cannot, therefore, be decomposed in the same way as flows related to
goods and services. While such flows cannot be measured in volume terms,
may also be linked to the CPI. they can nevertheless be measured “in real terms” by deflating their values
2.24 Creditors receiving interest payments do not con- with price indices in order to measure their real purchasing power over
sist only of households. In any case, the purpose of indexing some selected basket of goods and services that serves as the numeraire.
31
2.29 The first approach defines the change in the con- national accounts of a single country. Thus, while the pro-
sumption expenditure as the change in the total value of the cessing and aggregation of the basic data for CPI purposes
goods and services actually consumed, measured at the fixed should be determined by the needs of the CPI itself, it is
prices of some chosen period. This is equivalent to deflating appropriate to take into account the requirements of these
the change in the current value of the goods and services other kinds of price indices at the data collection stage.
consumed by an appropriately weighted Lowe price index. There may be important economies of scale obtained by
The change in real income can be measured by deflating the using a single collection process to meet the needs of several
change in total money income by the same price index. different types of indices.
2.30 The alternative approach defines the change in 2.36 The International Comparison Program (ICP) rep-
consumption expenditure in volume as the change in eco- resents one of the largest efforts to develop PPPs. Given lim-
nomic well-being derived from the goods and services ited staff and budgetary resources, there is a need for careful
actually consumed. This may be estimated by deflating the consideration to identify potential synergies between the ICP
change in the current value of consumption expenditure by and CPI. These possible synergies are the topic of an ongo-
using a COLI (see paragraphs 2.84–2.91). Real income may ing task force under the auspices of the ICP. The objective
be similarly obtained by deflating money income by the of this task force is to develop guidelines for countries to
same COLI. maximize limited resources for the collection of price data
2.31 The two approaches cannot lead to the same results needed for both programs, while ensuring that the reliabil-
if the pure price index and the COLI diverge. The choice ity and representativity of the CPI is not adversely affected.
between the two approaches to the measurement of con- Appendix 4 provides detail on the ICP which is managed by
sumption expenditure in volume and real income will not be the World Bank.
pursued further here, as the issues involved are essentially 2.37 Thus, operationally as well as conceptually, the
the same as those already considered in the parallel discus- CPI needs to be placed in the context of a wider set of inter-
sion of the choice between a Lowe, or basket, price index related price indices. The compilation of CPIs predates the
and a COLI (see paragraphs 2.82–2.91). compilation of national accounts by many years in some
countries, so the CPI may have originated as a free-standing
Consistency between Price Indices index. However, the CPI should no longer be treated as an
and Expenditure Series isolated index whose compilation and methodology can pro-
2.32 The data collected on prices and the data collected ceed quite independently of other statistics.
on household expenditure must be mutually consistent when
measuring the volume of consumption. This requires that
both sets of data cover the same set of goods and services Use of the Consumer Price Index
and use the same concepts and classifications. Problems for Accounting under Inflation
may arise in practice because the price indices and the 2.38 When there is inflation, both business and national
expenditure series are often compiled independently of each accounts have to introduce adjustments that are not needed
other by different departments of an NSO or even by differ- when the price level is stable. This is a complex subject
ent agencies. that cannot be pursued in any depth here. Two methods of
accounting are commonly used, and they are summarized
National Accounts in paragraphs 2.39 and 2.40. Both require price indices for
2.33 The CPI is mainly used in the national accounts their implementation.
to derive volume measures of household final consump-
tion expenditure (HFCE). HFCE at current prices should Current Purchasing Power Accounts
be deflated at the most detailed level as possible using the 2.39 Current purchasing power accounts are accounts
respective CPIs (2008 SNA, paragraph 15.140). In addition, in which the monetary values of the flows in earlier time
and in the absence of other indicators, the CPI is often used periods are scaled up in proportion to the increase in some
to derive volume measures of other national accounts aggre- general index of inflation between the earlier period and
gates, but this use of the CPI is neither ideal nor encouraged. the current period. In principle, the index used should be
a general price index covering other flows in addition to
Purchasing Power Parities HFCE, but in practice, the CPI is often used by default in
2.34 Many countries participate in regular international the absence of a more suitable general index.
programs enabling purchasing power parities (PPPs) to be
calculated for household actual consumption expenditure. Current Cost Accounting
The calculation of PPPs requires the prices of individual con- 2.40 Current cost accounting (2008 SNA, paragraphs
sumer goods and services to be compared directly between 1.65–1.67) is an accounting method in which the cost of
different countries. Real expenditures and real incomes can using goods and assets in production is calculated using
then be compared between countries in much the same way current prices of those goods and assets as distinct from
as between different time periods in the same country. the prices at which the assets were purchased or otherwise
2.35 It is not proposed to examine PPP methodology acquired in the past (historic costs). The current cost of
here but simply to note that PPPs create another demand using a good or asset takes account not only of changes in
for basic price data. When such data are being collected, it the general price level but also of changes in the relative
is important to recognize that they can be used for PPPs as price of that type of asset since it was acquired. In principle,
well as CPIs. PPPs are essentially international deflators that the price indices that are used to adjust the original prices
are analogous to the intertemporal deflators needed for the paid for the goods and assets should be specific price indices
32
relating to that particular type of good or asset, and such • Although price changes for certain kinds of consumer
indices are calculated and used in this way in some coun- products are difficult to measure because of quality
tries. However, when there are no such indices available, the changes, price changes for other kinds of goods and ser-
all-items CPI, or some subindex of the CPI, have been used vices, such as capital goods and government services,
for this purpose. especially public services, tend to be even more difficult
to measure.
Consumer Price Indices and International • Most countries have deliberately adopted a policy of not
Comparisons of Inflation revising the index once it has been published. This makes
2.41 CPIs are also commonly used to make interna- it more attractive for many purposes, especially those
tional comparisons of inflation rates. An important example with financial consequences such as indexation. The lack
of their use for this purpose is provided by the European of revisions may create a somewhat spurious impression
Union (EU). Since the mid-1990s, EU member countries of certainty, but it also seems to enhance the credibility
have compiled Harmonised Indices of Consumer Prices and acceptability of the index.
(HICP) that are used as an aggregate measure of inflation
for the euro area and to compare consumer price inflation 2.45 The widespread use of the CPI for purposes other
across member countries, and for monetary and economic than those for which it was designed can be explained not
policy purposes of the EU. only by the various factors listed in paragraph 2.44 but also
2.42 Another example would be the Group of Twenty with the fact that no satisfactory alternative or more compre-
(G20) CPI compiled by the Organisation for Economic Co- hensive measures of inflation are available monthly in most
operation and Development. One of the outcomes of the countries. For example, the CPI may be used as a proxy for
G20 Data Gaps Initiative was the compilation of aggregates a more general measure of inflation in business accounting,
for the G20 countries. Using the headline2 (all-items) indi- even though it may be clear that, conceptually, the CPI is
ces, the G20 aggregate provides a measure of inflation for not the ideal index for this purpose. Similarly, the fact that
the G20 countries. For more information on the G20 CPI, the CPI is not subject to revision, together with its frequency
see Methodology Notes: Compilation of a G-20 Consumer and timeliness, may explain its popularity for indexation
Price Index available on the Organisation for Economic Co- purposes in business or legal contracts in contexts where it
operation and Development website.3 also may not be very appropriate conceptually. These prac-
2.43 Finally, CPIs are used for purposes of international tices may be defended on the grounds that the alternative to
comparisons by a range of international organizations, using the CPI may be to make no adjustment for inflation.
including the Food and Agriculture Organization, the Inter- Although the CPI may not be the ideal measure, it is much
national Labor Organization, the International Monetary better to use it than to make no adjustment.
Fund, Organisation for Economic Co-operation and Devel- 2.46 Although the CPI is often used as a proxy for a
opment, World Bank, and others. general measure of inflation, this does not justify extending
its coverage to include elements that go beyond household
final consumption. If broader indices of inflation are needed,
Popularity of Consumer Price Indices they should be developed in addition to the CPI, leaving the
for Economic Analysis CPI intact. Some countries are in fact developing additional
2.44 The CPI is a high-profile statistic. CPIs seem to and more comprehensive measures of inflation within the
have acquired a unique status among economic statistics in SNA framework.
most countries. Changes in the CPI tend to receive signifi-
cant publicity and often make headline news. There are sev- Inflation Perceptions
eral factors which help to explain this: 2.47 While countries strive to compile and dissemi-
• All households experience the phenomenon the CPI nate a CPI that provides a reliable measure of price change,
intends to measure. The general public faces changes in many data users, especially individual households, perceive
the prices of consumer goods and services, and the direct the headline inflation rate to be incorrect. This perception
impact those changes have on their living standards. negatively affects the credibility of the index and user con-
Wages, pensions, and social security benefits may be fidence. NSOs should make every effort to improve these
adjusted according to changes in the CPI, which also will perceptions.
have a direct impact on households’ income. Interest in 2.48 Individual consumers may regard the headline
CPIs is not confined to the press and politicians. measure of inflation as wrong because they do not fully
• Countries disseminate the CPI frequently, usually each
appreciate CPI concepts and compilation methods. First,
month, so that the rate of consumer inflation can be closely the CPI represents a composite of all households, reflecting
monitored. The CPI is also a timely statistic that is released the average total expenditure of all households on the goods
very soon after the end of the period to which it refers. and services included in the basket, and not the expendi-
ture of any single household. Second, the index represents
• The CPI is a statistic with a long history, as noted in Chap- a weighted average change over time in the prices paid
ter 1. People have been familiar with it over this period of by consumers. This means that some prices are increasing
time. while others are decreasing. The impact of the price change
of each item included in the basket is determined by the
2
For the EU and Turkey, the HICP is used in the aggregation, while for all
weight of that item.
other countries, the national CPI is used. 2.49 To build user confidence and enhance credibil-
3
http://www.oecd.org/sdd/prices-ppp/CPI-G20-methodology.pdf. ity, NSOs should enhance transparency and provide more
33
details on the concepts and methods used to compile the CPI work done; and (4) they may receive them as free gifts,
(that is, metadata). Index compilation methods should be or transfers, from other economic units, including social
explained in detail and made available to all users on the transfers in kind provided by government. The treatment of
NSO website and in hardcopy as needed. In addition, NSOs household production activities is addressed in paragraphs
can provide better explanations to frequently asked ques- 2.154–2.168.
tions. Many NSO websites include a frequently asked ques- 2.55 The broadest concept of final consumption for
tions section to respond to questions regarding concepts and CPI purposes would be a price index embracing all four
methods used or to avoid common misperceptions. Finally, categories of consumption goods and services listed pre-
detailed data should be disseminated on the NSO website. viously. This set of consumption goods and services may
More detailed data provide more information to all users be described as total acquisitions. Total acquisitions are
to better understand the source(s) of price change. Chapter equivalent to the total household actual final consumption as
14 provides more information on the need to provide more defined in the SNA (2008 SNA, paragraph 9.81). The 2008
detailed data and metadata. SNA distinguishes between household actual final consump-
tion and HFCE, as explained in paragraphs 2.56–2.59.
The Need for Independence and Integrity 2.56 According to the SNA (2008 SNA, paragraph
of Consumer Price Indices 9.56), HFCE includes the expenditure incurred by resi-
dent households on consumption goods or services. These
2.50 Because of the widespread use of CPIs for vari- include purchases of consumer goods and services, as well
ous types of indexation, movements in the CPI can have as the estimated (imputed)5 value of barter transactions,
major financial ramifications throughout the economy. The goods and services received in kind, and goods and services
implications for the government alone can be considerable, produced and consumed by the same household (production
given that the CPI can affect interest payments and taxa- for own consumption).
tion receipts, as well as the government’s wage and social 2.57 Household actual final consumption, as defined in
security outlays. the 2008 SNA (2008 SNA, paragraph 9.81) consists of the
2.51 When financial interests are involved, there is consumption goods and services acquired by individual
always a risk that both political and nonpolitical pressure households. The total value of household actual final con-
groups may try to exert an influence on the methodology sumption includes the sum of these three components:
used to compile the CPI. The CPI, in common with other (1) the value of households’ expenditures on consump-
official statistics, must be protected from such pressures. tion goods or services, including expenditures on nonmar-
Partly for this reason, many countries establish an advisory ket goods or services sold at prices that are reduced or not
committee to ensure that the CPI is not subject to inappro- economically significant,6 (2) the value of the expenditures
priate outside influence or pressure. The advisory committee incurred by government units on individual consumption
may include representatives of a cross-section of interested goods or services provided to households as social transfers
parties, as well as independent experts able to offer profes- in kind, and (3) the value of the expenditures incurred by
sional advice. To enhance user confidence and transparency, nonprofit institution serving households (NPISHs) on indi-
details on the methodology used to compile CPIs should be vidual consumption goods or services provided to house-
made available to the general public. holds as social transfers in kind.
2.58 Household actual final consumption is a broader
Concepts and Scope of the concept than HFCE, and the difference between the two
Consumer Price Index concepts is linked to the concepts of acquisitions and expen-
diture, as explained in paragraphs 2.59–2.61.
2.52 As noted previously in paragraphs 2.38–2.43 and in
Chapter 1, the primary uses of the CPI will determine the Expenditure
concepts and scope used to compile the index. This section 2.59 Expenditure on goods and services is the value of
provides a more detailed look at the different concepts used the amount that buyers pay, or agree to pay, to sellers in
to compile the CPI, as well as scope and coverage issues. exchange for goods or services that sellers provide to them
or to other institutional units designated by the buyers (2008
Concepts SNA, paragraph 9.32). In other words, expenditure is made
by the economic units who pay for the goods and services
2.53 While this chapter does not provide a detailed that bear the costs. However, many of the goods and ser-
overview of SNA definitions and concepts, it should be vices consumed by households are financed and paid for by
noted that the SNA serves as the conceptual basis, or frame- government units or nonprofit institutions. They are mostly
work, for the CPI. Definitions for expenditure, consumption, services such as education, health, housing, and transport.
households, and other concepts are derived from the SNA.
2.54 Households may acquire goods and services4 for
purposes of final consumption in four main ways: (1) they 5
may purchase them in monetary transactions; (2) they may The term “imputation” is used in the SNA (2008 SNA, paragraph 3.75)
with a specific narrow meaning: it is preferable to reserve that term for the
produce them themselves for their own consumption; (3) kind of situation that involves not only estimating a value but also con-
they may receive them as payments in kind through bar- structing a transaction. In this manual, the term “imputation” has a broader
ter transactions, particularly as remuneration in kind for meaning.
6
Prices are said to be economically significant when they have a significant
influence on the amounts the producers are willing to supply and on the
4
The 2008 SNA defines goods and services in paragraphs 6.14–6.21. amounts purchasers wish to buy (2008 SNA, paragraph 6.95).
34
Household Actual Final Consumption

Value of Expenditures on Consumption Goods or Services, Including Payments for Nonmarket Value of Expenditures Incurred by
Goods or Services Sold at Nonsignificant Economic Prices (9.56) Government Units or NPISH on
Individual Consumption Goods or
Services Provided to Households as
Social Transfers in Kind (9.81)
Household Final Consumption Expenditure (HFCE)
Value of Expenditures on Consumption Goods or Services, Including Payments for Nonmarket
Goods or Services Sold at Nonsignificant Economic Prices (9.56)
Purchases of Consumer Purchases Goods and Services Consumer Goods and
Goods and Services (monetary of Consumer Provided by Employers Services Produced
transactions) Goods and to Their Employees in and Consumed by the
Services Remuneration for Work Same Household
(barter Done
transactions)
Household Final Monetary Consumption Expenditure (HFMCE)
Purchases of Consumer
Goods and Services (monetary
transactions)
Individual goods and services provided free of charge, or in the CPI. Alternatively, if the price decreases from some
at prices that are not economically significant, to individual positive amount to zero, this also must be reflected in the
households by governments or nonprofit institutions are CPI. The question remains how to deal with the weight for
described as social transfers in kind7 (social transfers in kind that item. One option may be to use a zero price and adjust
do not include collective services provided by governments the weights during the next update. Another option may be
to the community as a whole, such as public administration to redistribute the weight to the other items within the class.
and defense). Finally, it may be decided that the best approach would be to
2.60 The expenditure on social transfers in kind is redistribute the weight broadly over all items. The treatment
incurred by the governments or nonprofit institutions that of health, education, and social protection services in the
pay for them, and not by the households that consume them. CPI is further discussed in Chapter 11.
It could be decided that the CPI should be confined to final
consumption expenditure incurred by households, in which Acquisitions and Uses
case free social transfers in kind would be excluded from 2.62 It has been customary in the literature on CPIs to
the scope of the index. Even if they were to be included, draw a distinction between acquisitions of goods and ser-
they can be ignored in practice when they are considered vices by households and their subsequent use (consumption)
to be provided free, on the grounds that households incur to satisfy their households’ needs (the 2008 SNA defines
zero expenditure on them. Of course, their prices are not acquisitions in paragraphs 9.36–9.38 and consumption of
zero from the perspective of the units that finance the social goods and services in paragraphs 9.39–9.41). Consumption
transfers, but the relevant prices for a CPI are those payable goods are typically acquired at one point of time and used
by the households. at some other point of time, often much later, or they may
2.61 Social transfers cannot be ignored, however, when be used repeatedly, or even continuously, over an extended
governments and nonprofit institutions decide to introduce period of time. The times of acquisition and use nevertheless
(or eliminate) charges for them, a practice that has become coincide for many services, although there are other kinds
increasingly common in many countries. For example, if the of services that provide lasting benefits and are not used up
CPI is intended to measure the change in the total value of at the time they are provided. This is further explained in
a basket of consumption goods and services that includes paragraphs 2.63–2.67.
social transfers, increases in their prices to households from 2.63 Acquisition of a good refers to the moment at
zero to some positive amount increase the cost of the basket which ownership of a good transfers to the consumer. In
and should be captured by a CPI. The CPI should reflect a market setting, it is the moment at which the consumer
the full extent of the price change in the period the price incurs a liability to pay, either in cash, credit, or in kind.
increases from zero to some positive amount. With a zero Acquisition of a service cannot be determined as eas-
price, the good or service is generally excluded from the ily because the provision of a service does not involve an
CPI; however, when the price changes from zero to some exchange of ownership. Instead, it typically leads to some
positive amount, it becomes a newly significant good or ser- improvement in the condition of the consumer. Consumers
vice and can be treated in the same way as other new goods acquire a service at the same time that the producer pro-
or services are introduced. Chapter 7 provides an overview vides it and the consumer accepts a liability to pay. In a
of the different methods for including new goods or services market situation, therefore, the time of acquisition for both
goods and services refers to the time at which the liability to
pay is incurred. The distinction between time of acquisition
7
The 2008 SNA refers to social transfers in kind in paragraphs 6.234 and and time of use is particularly important for durable goods
6.235. and certain kinds of services.
35
2.64 A “nondurable” good would be better described as 2.70 Another important category of nonmonetary trans-
a single-use good (durable goods are defined in paragraph action occurs when households consume goods and services
9.42 of the 2008 SNA). For example, food and drink are used that they have produced themselves (2008 SNA, paragraphs
once only to satisfy hunger or thirst. Heating oil, coal, or 9.53–9.55). The households incur the costs, while the expen-
firewood can be burnt once only, but they are nevertheless diture is deemed to occur when the goods and services are
extremely durable physically and can be stored indefinitely. consumed. Own-account production of this kind includes
Households may hold substantial stocks of so-called nondu- the production of housing services by owner-occupiers. The
rables, such as many foodstuffs and fuel, especially in peri- treatment of goods and services produced for own consump-
ods of political or economic uncertainty. tion raises important conceptual issues that are discussed in
2.65 Conversely, the distinguishing feature of consumer more detail in paragraphs 2.154–2.157 and in Chapter 11.
durables, such as furniture, household equipment, or vehi- 2.71 The narrowest concept of consumption that could
cles, is that they are durable under use. They can be used be used for CPI purposes is one based on monetary transac-
repeatedly or continuously to satisfy consumers’ needs over tions only. As noted previously, HFCE includes barter trans-
a long period of time, possibly many years. For this reason, actions, goods and services received in kind, and production
a durable is often described as providing a flow of “services” for own consumption. A concept of consumption based only
to the consumer over the period it is used. Some durables on monetary transactions excludes many of the goods and
last much longer than others, the less durable ones being services actually acquired and consumed by households.
described as “semidurables” in COICOP, for example, cloth- On the other hand, only monetary transactions generate the
ing. Dwellings are not classified as consumer durables in prices needed for CPI purposes. The prices of the goods and
COICOP. They are treated as fixed assets and not consump- services acquired through nonmonetary transactions cannot
tion goods, and therefore fall outside the scope of COICOP be directly observed and have to be imputed or estimated on
and the CPI. However, the housing services produced and the basis of the prices observed in monetary transactions.
consumed by owner-occupiers are included in COICOP and Imputed or estimated prices do not generate more price
classified in the same way as the housing services consumed information. Instead, they affect the weighting, increasing
by tenants. More information on the treatment of housing is the weight of the prices used to value nonmonetary transac-
available in Chapter 11. tions for which prices have to be imputed or estimated.
2.66 Consumers may continue to benefit, and derive 2.72 If the main reason for compiling a CPI is the mea-
utility, from some services long after they were provided surement of inflation or as an input into monetary policy
because they bring about substantial, long-lasting, or even decisions, the scope of the index should be restricted to
permanent improvements in the condition of the consum- monetary transactions only, especially since nonmonetary
ers. The quality of life of individuals receiving medical transactions do not generate any demand for money. The
treatments such as hip replacements or cataract surgery, for HICP, used to measure inflation within the EU, is confined
example, is substantially and permanently improved. Simi- to household final monetary consumption expenditure
larly, consumers of educational services can benefit from (HFMCE) (see Appendix 1).
them over their entire lifetimes.
2.67 For some analytical purposes, it may be appropri- Domestic and National Concept of Expenditure
ate to treat certain kinds of services, such as education and 2.73 The concept of residence is based on the center of
health, as the service equivalents of durable goods. Expen- predominant economic interest of the institutional unit (that
diture on such services can be viewed as investments that is, household, in the case of the CPI), and not on nationality
augment the stock of human capital. Another characteristic or legal criteria. The SNA states that the most commonly
that education and health services share with durable goods used concept of economic territory is the geographic territory
is that they are often so expensive that their purchase has to administrated by a government, although it may be larger or
be financed by borrowing or by running down other assets. smaller than this (as in a currency or economic unit, or a
region in a country). The economic territory also includes
Monetary and Nonmonetary Transactions the territorial enclaves in the rest of the world (embassies,
2.68 A distinction may also be drawn between mone- consulates, military bases, scientific stations, and so on). As
tary and nonmonetary transactions depending if the payment a general rule, institutional units are considered resident in
made or liability incurred is stated in units of currency. A a certain economic territory if they are engaged in activities
monetary transaction occurs when a household pays in cash, and transactions on a significant scale for one year or more.
by check or credit card, or otherwise incurs a liability to pay, According to the SNA, the residence of individual persons is
in exchange for the acquisition of a good or service. Non- determined by that of the household of which they form part
monetary transactions occur when households do not incur a and not by their place of work, and all members of the same
liability but bear the costs of acquiring the goods or services household have the same residence as the household itself.
in some other way. If members of a household work and reside abroad so long
2.69 Households may incur nonmonetary transactions that they acquire a center of economic interest abroad, they
when household members receive goods and services from cease to be members of their original household (2008 SNA,
their employers as remuneration in kind (2008 SNA, para- paragraphs 4.10–4.15).
graphs 9.51 and 9.52). The employees pay for the goods and 2.74 Therefore, according to the 2008 SNA (2008 SNA,
services with their own labor rather than cash. Consumption paragraph 9.79), HFCE refers to the expenditure incurred
goods and services received as remuneration in kind can, in by resident households, whether that expenditure is incurred
principle, be included in a CPI using the estimated prices within the economic territory or abroad. Additionally,
that would be payable for them on the market. nonresident households may make expenditure inside the
36
economic territory of a country. Depending on the residence (for example, streaming or downloading music, movies, or
principle, there are two important concepts related to CPI television content) and other services (for example, soft-
coverage: the domestic and national concept. ware). If the service is consumed in the economic territory
2.75 The national concept, which aligns with the HFCE where the household is resident, it should be included in the
concept as defined in the 2008 SNA, includes expenditure CPI; however, if the service is consumed outside of the eco-
made by resident households, whether it takes place on the nomic territory of the country, it would be excluded. For
economic territory or elsewhere. This means that the expen- example, if a household reserves a hotel room that will be
diture of resident households at home should be adjusted used and paid for in another country, it would be consid-
by adding the expenditure of resident households abroad ered out of scope. For digital services, because the service is
(imports) and deducting the expenditure of nonresident being consumed within the economic territory of a country,
households in the home territory (exports). the expenditure and prices should be included in the country
2.76 The domestic concept is not based on the residence where the household resides. More details on the treatment
of households, rather on the territory where the consumption of internet purchases can be found in Chapter 11.
or expenditure occurs. It covers the consumption expendi-
ture made by all households on the economic territory of
a country, including the expenditure of nonresidents of the Basket Indices and Cost of Living Indices
country and excluding the expenditure of residents abroad. 2.82 A fundamental conceptual distinction may be
2.77 Decisions about the choice of the concept used drawn between a basket index and a COLI. In a CPI context,
depend on the main use of a CPI. For inflation analysis and a basket index is an index that measures the change between
monetary policy purposes, it is the price change within a two time periods in the total expenditure needed to purchase
country which is of interest. An index of inflation is needed a given set, or basket, of consumption goods and services. It
that covers all “domestic” consumption expenditure that is called a “Lowe index” in this Manual. A COLI is an index
takes place within the geographical boundaries of the coun- that measures the change in the minimum cost of maintain-
try, whether made by residents or nonresidents. ing a given standard of living. Both indices therefore have
2.78 When CPIs are used for escalating the incomes of very similar objectives in that they aim to measure the
residents, it may be appropriate to adopt the “national” con- change in the total expenditure needed to purchase either the
cept which covers all the expenditure of residents, whether same basket or two baskets whose composition may differ
inside or outside the country. Household budget surveys somewhat but between which the household is indifferent.
(HBSs) can cover all these types of expenditure. For exam-
ple, the prices paid for airline tickets and package holidays Basket Indices
bought within the domestic territory should be covered. It 2.83 CPIs are often calculated as either Lowe or Young
can be difficult, however, to obtain price data for the goods indices in practice. Their properties and behavior are
and services purchased by residents when abroad, although described in detail throughout this Manual. The operational
in some cases subindices of the partner countries’ CPIs target for most CPIs is to measure the change over time in
might be used. the total value of a specified basket of consumption goods
2.79 The treatment of purchases made online requires and services purchased, or acquired, by a specified group
special consideration. In principle, the domestic and of households in a specified period of time. The meaning
national concepts could provide guidance on how to treat of such an index is clear. It is necessary to ensure that the
the expenditure made on goods and services, including digi- selected basket is relevant to the needs of users and also kept
tal downloads, purchased online. In many cases, however, up to date. The basket may be changed at regular intervals
internet-based outlets may be based (registered) abroad and and does not have to remain fixed over long periods of time.
this expenditure would be considered cross-border shop- The determination of the basket is considered in more detail
ping. For those countries following the national concept, in Chapter 3.
the approach is clear. Strictly speaking, under the domes-
tic concept, this expenditure would not be included because Cost of Living Indices
it would be defined as expenditure abroad; in practice, this 2.84 The economic approach to index number the-
requires a broader interpretation. The nature of internet pur- ory treats the quantities consumed as being dependent on
chases, therefore, requires a different way of thinking and the prices. Households are treated as price takers who are
special consideration, especially with regard to the domestic assumed to react to changes in relative prices by adjusting
concept. Additionally, internet purchases continue to grow the relative quantities they consume. A basket index that
in importance. works with a fixed set of quantities fails to allow for the fact
2.80 Many countries have carefully considered how to that there is a systematic tendency for consumers to substi-
include the expenditure (and prices) made on goods and ser- tute items that have become relatively cheaper for those that
vices via the internet. For the purchase of goods, the expen- have become relatively more expensive. A COLI based on
diture and prices should be reflected in the country where the economic approach does take this substitution effect into
the goods are delivered. account. It measures the change in the minimum expenditure
2.81 Services purchased on the internet can be more needed to maintain a given standard of living when utility-
problematic for CPI compilation because there are both tan- maximizing households adjust their patterns of purchases in
gible and digital services. Tangible services would include response to changes in relative prices. In contrast to a basket
the traditional services such as transportation, hotels, index, the baskets in the two periods in a COLI will gener-
entrance to cultural/sporting events, or education. Digital ally not be quite the same in the two periods because of these
services would include telecommunications, broadcasting substitutions.
37
2.85 The properties and behavior of COLIs are 2.89 A conditional COLI should not be viewed as sec-
explained in some detail in Chapter 4 of the publication ond best. An unconditional COLI is a more comprehensive
Consumer Price Index Theory. A summary explanation COLI than a conditional COLI, but it is not a more compre-
has already been given in Chapter 1. The maximum scope hensive price index than a conditional index. An uncondi-
of a COLI would be the entire set of consumption goods tional index does not include more price information than
and services consumed by the designated households from a conditional index and it does not give more insight into
which they derive utility. It includes the goods and services the impact of price changes on households’ economic well-
received free as social transfers in kind from governments being. On the contrary, the impact of the price changes
or NPISH. Because COLIs measure the change in the cost is diluted and obscured as more variables impacting on
of maintaining a given standard of living or level of util- economic well-being are included within the scope of the
ity, they lend themselves to a uses rather than an acquisi- index.
tions approach (see also paragraphs 2.62–2.67), as utility is 2.90 Lowe indices, including Laspeyres and Paasche,
derived not by acquiring a consumer good or service but by are also conditional, being dependent on the choice of the
using it to satisfy personal needs and wants. basket. The fact that the value of a basket index varies in
2.86 Economic well-being may be interpreted to mean predictable ways according to the choice of basket has gen-
not only economic welfare, that is, the utility linked to eco- erated much of the large body of literature available on index
nomic activities such as production, consumption, and work- number theory. Conceptually, Lowe indices and conditional
ing, but also the general well-being associated with other COLIs have much in common. A Lowe index measures the
factors such as security from attack by others. It may not be change in the cost of a specified basket of goods and ser-
possible to draw a clear distinction between economic and vices, whereas a conditional COLI measures the change in
noneconomic factors, but it is clear that total economic well- the cost of maintaining the level of utility associated with
being is only partly dependent on the amount of goods and some specified basket of goods and services, other things
services consumed. being equal.
2.87 Conditional and unconditional COLIs. In prin- 2.91 In practice, it is not possible to compile a true
ciple, the scope of a COLI is influenced by whether it is COLI. Some countries construct the CPI for indexation
intended to be a conditional or unconditional COLI. The total purposes and employ methods to approximate a COLI.
welfare of a household depends on a string of noneconomic Similarly, other countries construct their CPI with the pri-
factors such as the climate, the state of the physical, social, mary purpose of measuring inflation, in which case a cost
and political environment, the risk of being attacked either of goods index would be the target index. The selection of
by criminals or from abroad, the incidence of diseases, and a cost of goods index or a COLI as target for the CPI has
so on, as well as by the quantities of goods and services con- consequences concerning the scope of the index and the way
sumed. An unconditional COLI measures the change in the different goods or services are included in the index. In prac-
cost to a household of maintaining a given level of total eco- tice, most countries compile a general-purpose index that
nomic well-being allowing the noneconomic factors to vary strives to meet both compensation and inflation measure-
as well as the prices of consumption goods and services. If ment needs in a single index.
changes in the noneconomic factors lower economic well-
being, then some compensating increase in the level of con-
sumption will be needed in order to maintain the same level Consumer Price Index Scope
of total economic well-being. For example, if there is an Households and Outlets Included in
adverse change in the weather, that requires more fuel to be the Scope of a Consumer Price Index
consumed to maintain the same level of comfort as before.
The cost of the increased quantities of fuel consumed drives Population Coverage
up the unconditional COLI, irrespective of what has hap- 2.92 The group of households included in the scope of
pened to prices. There are countless other events that can a CPI is referred to as the “reference households,” or the
impact on an unconditional COLI, from natural disasters “reference population.” In this context, scope refers to the
such as earthquakes to man-made disasters such as climate households intended to be represented in the CPI; on the
change or acts of terrorism. other hand, coverage refers to the actual households repre-
2.88 While there may be interest in an unconditional sented by the index.
COLI for certain analytical and policy purposes, it is delib- 2.93 According to the SNA, households are an institu-
erately defined to measure the effects of many other factors tional sector made up of private households and institutional
besides prices. If the objective is to measure the effects of households. Private households are defined as groups of per-
price changes only, the nonprice factors must be held constant. sons who share the same living accommodation, who pool
Given that a COLI is meant to serve as a CPI its scope must some or all of their resources and consume selected goods
be restricted to exclude the effects of events other than price and services collectively. Members of the same household do
changes. A conditional COLI is defined as the ratio of the min- not necessarily have to belong to the same family, as long as
imum expenditure needed to maintain a given level of utility, there is some sharing of resources and consumption. Institu-
or standard of living, in response to price changes, assuming tional households consist of persons living permanently, or
that all the other factors affecting economic well-being remain for a very long period of time, in institutions. These include
constant. It is conditional not only on a particular standard of religious institutions, hospitals, the military, prisons, or
living and set of preferences but also on a particular state of the retirement homes. Persons who enter such institutions only
nonprice factors affecting economic well-being. COLIs, in this for short periods of time should be treated as members of
Manual, are to be understood as conditional COLIs. the individual households to which they normally belong
38
(the 2008 SNA defines household, private, and institutional course, the borderline between urban and rural is inevitably
household in paragraphs 4.149–4.154). arbitrary and varies from country to country. For example, in
2.94 Following the SNA definition of HFCE (2008 one country, urban price collection is interpreted to include
SNA, paragraph 9.56) both private and institutional house- villages with as few as 2,000 residents, while in others this
holds are considered in the scope of the CPI, and in prin- threshold may be higher.
ciple should be covered. For example, the EU’s HICP 2.99 Decisions about geographical coverage with
coverage of households includes institutional households, regard to urban versus rural coverage depend on the popula-
consistent with the SNA definition (for more information, tion distribution and the extent to which expenditure pat-
see Appendix 1). Nevertheless, in many countries, the con- terns and the movements of prices tend to differ between
sumption expenditure of institutional households may be urban and rural areas.
of negligible importance, or it may be problematic to col- 2.100 When compiling regional indices, the concept
lect suitable expenditure data from institutional households. of residence (as described in 2.73) applies to the region in
For these reasons, it may be decided to exclude the con- which a household is resident. It is possible to draw a dis-
sumption expenditure of institutional households from the tinction between the expenditure within a region and the
actual coverage of the CPI. It may also be the case that the expenditure of the residents of that region, analogous to the
compensation (indexing wages and salaries) of only private distinction between the “domestic” and “national” concepts
households is one of the main purposes of the CPI, which of expenditure at the country level. The same issues arise for
may also justify disregarding the consumption expenditure regional indices as were discussed in the previous section
of institutional households. for national indices. The principles applying to cross-border
2.95 In almost all countries, the CPI scope includes as shopping between regions are the same as for international
many private households as possible and is not confined cross-border shopping (for a description of cross-border
to those belonging to a specific socioeconomic group. In internet purchases, see the section on internet purchases
some countries, however, extremely wealthy households are in Chapter 11), but data availability is generally different.
excluded for various reasons. Their expenditure may be con- if the scope of the regional index is defined to include the
sidered to be very atypical, while their expenditure data, as purchases by regional residents when in other regions (and
collected in an HBS, may be unreliable, as the response rates exclude purchases by consumers that are not residents of
for wealthy households in an HBS are usually quite low. In the region), although price data for the other regions should
addition, it may be too costly to collect prices for some of be readily available, it is unlikely that expenditure data will
the consumer goods and services purchased exclusively by be available with the necessary split between expenditure
the wealthy. Other countries may decide to exclude other within, and expenditure outside, the region of residence.
kinds of households. Such exclusions affect the expendi- 2.101 When compiling a regional index, care must be
ture weights to the extent that the patterns of expenditure taken to treat cross-border purchases in the same way in all
of the excluded groups differ from those of the rest of the regions. Otherwise double counting, or omission, of expen-
population. As far as possible, countries should define the diture may occur when regional data are aggregated. Where
scope to include all households regardless of size, location, regional indices are aggregated to give a national index, the
or income. weights should be based on regional expenditure data rather
2.96 In addition to a single wide-ranging official (head- than on population data, as discussed in Chapter 3.
line) CPI relevant to the country as a whole, more and more 2.102 Many countries try to satisfy the differing needs
countries publish a range of alternative indices relating to of their many CPI users by deriving a family of indices with
subsectors of the population (for example, low-income different coverage, headed by a single wide-ranging official
households or retirees). (headline) CPI which is relevant to the country as a whole.
In some large countries, regional indices are more widely
Geographical Coverage used than the national CPI, particularly where the indices
2.97 Geographical coverage refers either to the geo- are used for indexing wages and salaries. Thus, in addition
graphical coverage of expenditure or the coverage of price to the headline CPI, which has the widest coverage possible,
collection. Ideally, these two should coincide, whether the alternative indices are published which may relate to: (1)
CPI intends to be a national or a regional index. In general, subsectors of the population; (2) geographical regions; and
for the geographical coverage of expenditure, the index (3) specific product groups. Subindices of the overall (offi-
should include expenditure made by all households, urban cial all-items) CPI should be published at a level as detailed
and rural, throughout the country. as possible, since many users are interested in the price
2.98 In many countries, prices are collected in urban change of specific product groups.
areas only because their movements are considered to be
representative of the price movements in rural areas. In Outlet Coverage
these cases, if national weights are applied, the resulting 2.103 The coverage of outlets is dictated by the pur-
index can be considered a national CPI. If price move- chasing behavior of the reference households. In principle,
ments in urban and rural areas differ significantly and price the prices relevant to CPIs are the prices paid by households.
collection is restricted to urban areas because of resource In practice, however, it is necessary to rely mainly on the
constraints, then urban weights should be applied, and the prices at which products are offered for sale in retail shops
resulting index must be considered as an urban and not a or other outlets (including online outlets). All the outlets
national CPI. Most countries tend to use weights covering from which the reference population makes purchases are in
urban and rural households, although in the majority of the scope of the CPI and should be included in the sampling
cases, price collection takes place in urban areas only. Of frame from which the outlets are selected.
39
2.104 Examples of outlets are: (1) retail shops (from consumers actually pay for the goods and services included
very small permanent stalls to multinational chain stores); in the scope of the index. The expenditure and prices
(2) market stalls and street vendors; (3) online and web- recorded should be purchaser’s prices (2008 SNA, para-
based retail outlets; (4) establishments providing household graphs 6.64–6.68). Purchaser’s prices refer to those prices
services (for example, electricians, plumbers, or window paid by consumers to acquire ownership of goods or ser-
cleaners); (5) leisure and entertainment providers; (6) health vices. Purchaser’s prices include any taxes (not deductible
and education services providers; (7) communication and by the purchaser) and service charges on the products, and
transport providers; (8) public utilities; and (9) government taking account of all discounts, subsidies, and most rebates,
agencies and departments. even if discriminatory or conditional. It may be virtually
2.105 Given the growing importance of online and impossible, however, to take account of all discounts and
web-based retailers, there is a significant need for NSOs rebates in practice. Sensible practical compromises are
to carefully review and augment outlet sampling frames to needed, for which recommendations and examples are given
include online and web-based retailers. Collecting prices in Chapter 5.
from online and web-based retailers raises a number of dif- 2.111 When households pay the full market prices for
ferent issues that are addressed in more detail in Chapter 11. products and are then subsequently reimbursed by govern-
2.106 In practice, prices are collected from only a sam- ments, social security programs, or NPISH for some of the
ple of outlets and the samples may change, either because amounts paid, CPIs should record the market prices less the
outlets open and close or because there is a deliberate peri- amounts reimbursed. This kind of arrangement is common
odic rotation of the sample. When the prices in the outlets for educational and medical expenditure (health, education,
newly included in the sample are different from those in the and social protection are addressed in Chapter 11).
previous outlets, it is necessary to decide whether the price 2.112 Taxes and subsidies. All taxes on products, such
differences are apparent or genuine price changes. If they as sales taxes, excise taxes, and value-added tax (VAT),
are assumed to be apparent, the difference between the price are part of the purchasers’ prices paid by consumers which
recorded previously in an old outlet and the new price in should be used for CPI purposes. Similarly, subsidies8
the new outlet is not treated as a price change for CPI pur- should be taken into account, being treated as negative
poses, the difference being treated as attributable to quality taxes on products. For information on the treatment of taxes
difference. As explained in more detail in Chapter 6, if this and subsidies on the national accounts, see 2008 SNA, para-
assumption is correct, the price changes recorded in the new graphs 7.71–7.106.
outlets can simply be linked to those previously recorded in 2.113 For some analytical and policy purposes, it may
the old outlets without introducing any bias into the index. be useful to estimate a CPI that measures price movements
The switch from the old to the new outlets does not have any excluding the effects of changes in taxes and subsidies. For
impact on the CPI. monetary policymakers, the price increases resulting from
2.107 If the price differences between the old and the changes in taxes on products or subsidies are not part of an
new outlets are deemed to be genuine price changes, how- underlying inflation process but are attributable to their own
ever, the simple linking just described can lead to bias. When use of these economic levers. Similarly, when a CPI is used
households change the price they pay for a product by chang- for escalation purposes, any increase in a CPI resulting from
ing outlets, the price changes should be captured by the CPI. increases in taxes on products leads to an increase in wages
As explained in more detail in Chapter 6, it seems that most and benefits linked to the CPI, despite the fact that the aim of
NSOs assume that the price differences are not genuine and the tax increase might have been to reduce consumers’ pur-
simply link the new price series to the old. This procedure, chasing power. Alternatively, an increase in subsidies might
although widely used, assumes that markets are always per- be intended to stimulate consumption, but the resulting
fect, and that genuine price variation never occurs. This unre- lower prices could be offset by a smaller increase in indexed
alistic and questionable assumption may lead to an upward wages and benefits.
bias. Such bias is described as outlet rotation bias. 2.114 Net price indices. Net price indices in which taxes
2.108 The differences between old and new outlets on consumer goods or services are deducted and subsidies
becomes even more pronounced when the outlet sample are added to the purchasers’ prices may be compiled. Such
expands to include web-based outlets. In principle, web- indices do not, however, necessarily show how prices would
based outlets provide some greater utility to the consumer have moved if there were no taxes or changes in taxes. It is
with regard to choice, convenience, and service; however, in notoriously difficult to estimate the true incidence of taxes
practice, it is difficult to quantify the value of this added util- on products: that is, the extent to which taxes or subsidies,
ity. While concrete guidance cannot be provided at this time, or changes therein, are passed on to consumers. It is also
accounting for the differences when consumers switch from difficult to account for the secondary effects of changes in
old to new outlets will be considered as part of the research taxes. In order to estimate the secondary effects, input–out-
agenda included in Appendix 6 of this Manual. put analysis can be used to work out the cumulative impact
2.109 Outlet replacement and the treatment of price dif-
ferences between the old and new outlets are discussed in
Chapter 7. Methods for selecting a sample of outlets from 8
The SNA defines subsidies as current unrequited payments that govern-
which to collect prices are discussed in detail in Chapter 4. ment units make to enterprises on the basis of the level of their production
activities or the quantities or values of the goods or services that they pro-
duce, sell, or import. Subsidies are not payable to final consumers; current
Price Coverage transfers that governments make directly to households as consumers are
2.110 A CPI should reflect the experience of the con- treated as social benefits (2008 SNA, paragraphs 7.98–7.99). The use of the
sumers to whom it relates and should therefore record what term “subsidies” is broader than the use in the SNA.
40
of taxes and subsidies through all the various stages of pro- may not be aware of them, or they may have imperfect infor-
duction. For example, some of the taxes on vehicle fuel will mation because the costs of searching for the retail outlets
enter the price of transport services which in turn will enter selling at the lowest prices may be too high. Even when
the prices of transported products, some of which will enter the households are aware of the price differences, it may be too
prices paid for consumer products by retailers and hence inconvenient or costly to visit the outlets selling at the low-
the prices which they charge to consumers. A more practi- est prices. Another reason for the persistence of price differ-
cable alternative is therefore simply to confine the taxes and ences is that many service producers deliberately practice
subsidies for which correction is made to those levied at the price discrimination by charging different households dif-
final stage of sale at retail; that is, primarily to VAT, sales, ferent prices for identical services (for example, by charg-
and excise taxes. Estimating prices less these taxes only, or ing lower prices or fees to pensioners or people with low
corrected for changes in these taxes only, is more feasible. In incomes). As services cannot be retraded, price discrimina-
the case of a percentage sales tax or VAT, the calculation is tion is extremely common, or even prevalent, among service
simple, but in the case of excise taxes, it would be necessary producers. Household expenditure is nevertheless recorded
to ascertain the percentage markup by the retailer, since the at the prices actually paid, as this is the appropriate value of
excise tax will also be marked up by this percentage. How- the transaction.
ever, in practice, it is not possible to determine the percent- 2.120 Apparent price differences between the same
age markup by the retailer. goods or services are often not genuine price differences as
they may be due to differences in quality, including differ-
Price Variation ences in the terms or conditions of sale. For example, lower
2.115 Price variation occurs when exactly the same prices are often charged for bulk purchases of goods or off-
good or service sells at different prices at the same moment peak purchases of services. Such expenditure is recorded
of time. Different outlets may sell exactly the same product at the prices actually paid; that is, after deducting from the
at different prices, or the same product may be sold from a standard or list prices or charges any discounts for bulk or
single outlet to different categories of purchasers at different off-peak purchases.
prices. 2.121 Price discrimination can cause problems in the
2.116 If markets were “perfect” in an economic sense, compilation of price indices. Suppose, for example, that a
identical products would all sell at the same price. If more service supplier discriminates by age by charging senior
than one price was quoted, all purchases would be made at citizens, aged 60 years or more, price p2 and everyone else
the lowest price. This suggests that some products sold at price p1, where p1 > p2. Suppose, further, that the supplier
different prices may not be identical but must be qualita- then decides to redefine senior citizens as those aged 70
tively different in some way. When the price differences are, years or more, while otherwise keeping prices unchanged.
in fact, attributable to quality differences, the price differ- In this case, although neither p1 nor p2 changes, the price
ences are only apparent, not genuine. In such cases, a change paid by individuals aged 60–70 years changes and the aver-
in the average price resulting from a shift in the pattern of age price paid by all households increases.
quantities sold at different prices would reflect a change in 2.122 The example in paragraph 2.121 illustrates a
the average quality of the products sold. This would affect point of principle. Although neither of the stated prices, p1
the volume and not the price index. and p2, at which the services are on offer changes, the price
2.117 If NSOs do not have enough information about paid by certain households changes if they are obliged to
the characteristics of goods and services selling at different switch from p2 to p1. From the perspective of the house-
prices, they have to decide whether to assume the observed holds, price changes have occurred and a CPI should, in
price differences are genuine or only apparent. The default principle, record a change. When prices are collected from
procedure most commonly adopted in these circumstances sellers and not from households, such price changes can be
is to assume that the price differences are not genuine. This missed if price collectors do not pay attention to the condi-
assumption is typically made for both CPI and national tions of sale.
accounts purposes.
2.118 However, markets are seldom perfect. One reason Price Variation between Outlets
for the coexistence of different prices for identical products 2.123 The existence of different prices in different
may be that the sellers are able to practice price discrimi- outlets raises similar issues. Genuine price differences are
nation. Another reason may simply be that households lack almost bound to occur when there are market imperfec-
information and may buy at higher prices out of ignorance. tions, if only because households are not perfectly informed.
Also, outlets may provide different levels of service which When new outlets sell at lower prices than existing ones,
would be reflected in the prices paid by consumers. Finally, there may be a time lag during which exactly the same item
markets may be temporarily out of equilibrium as a result of sells at different prices in different outlets because of con-
shocks or the appearance of new products. It must be recog- sumer ignorance or inertia.
nized, therefore, that genuine price differences do occur for 2.124 Households may choose to switch their pur-
a number of different reasons. chases from one outlet to another or even be obliged to
switch because the universe of outlets is continually chang-
Price Discrimination ing, some outlets closing down while new outlets open up.
2.119 Economic theory shows that price discrimination When households switch, the effect on the CPI depends on
tends to increase profits. Different households may pay dif- whether the price differences are genuine. When the price
ferent prices for identical products because of market imper- differences are genuine, a switch between outlets changes
fections. Price differences may persist because households the average prices paid by households. Such price changes
41
ought to be captured by CPIs. On the other hand, if the price an unconditional COLI or within a more broadly defined
differences reflected quality differences, a switch would conditional COLI that allows for changes in some other fac-
change the average quality of the products purchased, and tors besides changes in the prices of consumption goods and
hence affect volume, not price. services.
2.125 Most of the prices collected for CPI purposes 2.130 Licenses. Households have to pay to obtain vari-
are offer prices. Offer prices may not in all cases corre- ous kinds of licenses and it is often not clear whether they
spond to the actual transaction prices paid by households. are simply taxes under another name or whether the gov-
In these circumstances, the effects of switches in the pat- ernment agency issuing the license provides some kind of
tern of households’ purchases between outlets may remain service in exchange, for example, by exercising some super-
unobserved in practice. When the price differences reflect visory, regulatory, or control function. In the latter case, they
quality differences, the failure to detect such switches does could be regarded as purchases of services (2008 SNA, para-
not introduce any bias into the CPI. In this case, buying at graphs 8.64c and 9.70).
a lower price means buying a lower-quality product, which 2.131 Payments by households for licenses to own or
does not affect the accuracy of the price index. However, use certain goods or facilities are, by convention, classi-
when the price differences are genuine, the failure to detect fied as consumption expenditure, not transfers, and are thus
switches will tend to introduce an upward bias in the index, included within the scope of a CPI. For example, license
assuming households tend to switch toward outlets selling at fees for radios, televisions, driving, firearms, and so on, as
lower prices. This potential bias is described as outlet sub- well as fees for passports are included. On the other hand,
stitution bias. licenses or fees for owning or using vehicles, boats, and air-
craft, and for hunting, shooting, and fishing are convention-
Expenditure and Other Payments Generally ally classified as current taxes and are therefore outside the
Outside the Scope of Consumer Price Indices scope of CPIs. Many countries, however, do include taxes
2.126 Given that, conceptually, most CPIs are designed for private vehicle use as they regard them as taxes on con-
to measure changes in the prices of consumption goods sumption for CPI purposes. As the actual circumstances
and services, it follows that purchases of items that are not under which licenses are issued, and the conditions attach-
goods and services fall outside the intended scope of a CPI: ing to them, can vary significantly from country to country,
for example, purchases of bonds, shares, or other financial NSOs may wish to deviate from the proposed conventions
assets. Similarly, payments that are not purchases because in some instance. However, in general, it seems appropriate
nothing is received in exchange fall outside the index: to make use of conventions internationally agreed by the rel-
for example, payments of income taxes or social security evant experts and to be consistent with the SNA.
contributions. 2.132 Gifts and subscriptions. Gifts are transfers, by
2.127 The implementation of these principles is not definition (2008 SNA, paragraph 3.82), and thus outside the
always straightforward, as the distinction between an expen- scope of a CPI. Payments of subscriptions or donations to
diture on a good or service and other payments may not charitable organizations for which no easily identifiable ser-
always be clear in practice. A number of conceptually dif- vices are received in return are also transfers (2008 SNA,
ficult cases, including some borderline cases of a possibly paragraph 8.132). On the other hand, payments of subscrip-
controversial nature, are examined in paragraphs 2.128 and tions to clubs and societies, including charities, which pro-
2.129. vide their members with some kind of service (for example,
regular meetings or magazines) can be regarded as HFCE
Transfers and included in a CPI.
2.128 Transfers. A transfer is a transaction in which one 2.133 Tips and gratuities. Noncompulsory tips or gratu-
unit provides a good, service, or asset to another without ities are considered gifts and are outside the scope of a CPI.
receiving any good, service, or asset in return (2008 SNA, There may be cases, however, where gratuities are compul-
paragraph 8.10), that is, transactions in which there is no sory, and in these cases, the payment should be included in
counterpart. Transfers are unrequited. As no good or service the expenditure on, and the price of, the good or service in
of any kind is acquired by the household when it makes a questions. For example, where restaurants include a com-
transfer, the transfer must be outside the scope of a CPI. The pulsory service charge on the bill, this would be included in
challenge is to determine whether or not certain kinds of a CPI.9 In cases where payment of a tip or gratuity is cus-
transactions are in fact transfers, an issue common to both tomary, although not compulsory, it could be argued that
CPIs and national accounts. because payment of a tip is expected, it should be included
2.129 Social security contributions and taxes on income in both the expenditure and the price; however, on practical
and wealth. As households do not receive any specific, indi- grounds, these tips and gratuities are often excluded because
vidual good or service in return for the payment of social the amount given is discretionary and it is not possible to
security contributions, they are treated as transfers (2008 define a single amount or percentage that should be added.
SNA, paragraph 8.16) that are outside the scope of CPIs.
Similarly, all payments of taxes based on income or wealth
(the ownership of assets) are outside the scope of a CPI
since they are unrequited compulsory transfers to govern- 9
The SNA states that commissions, gratuities, and tips received by employ-
ment (2008 SNA, paragraph 8.15). Property taxes on dwell- ees should be treated as payments for services rendered by the enterprise
employing the worker, and so should also be included in the output and
ings (commonly levied as local authority taxes or rates) gross value added of the employing enterprise when they are paid directly
are outside the scope. It may be noted, however, that unre- to the employee by a third party (2008 SNA, paragraph 7.44). Therefore, the
quited compulsory transfers could be incorporated within prices considered for final uses include the tips paid.
42
Insurance rendered by financial auxiliaries such as brokers, banks,

2.134 Insurance. There are two main types of insur- insurers (life and nonlife), pension fund managers, finan-
ance: life and nonlife insurance (2008 SNA, paragraph 17.6). cial advisors, accountants, and so on are within the scope
In both cases the premiums have two components: the net of a CPI. Payments of such fees are simply purchases of
premium, described as the payment for the insurance, and services. For more detailed information, see the section on
the service charge payable to the insurance enterprise for Financial Services in Chapter 11.
arranging the insurance (that is, a fee charged for calculat- 2.141 Purchases and sales of foreign currency. Foreign
ing the risks, determining the premiums, administering the currency is a financial asset. Purchases and sales of foreign
collection and investment of premiums, and the payment of currency are therefore outside the scope of CPIs. Changes
claims). in the prices payable, or receivable, for foreign currencies
2.135 The service charge is not directly observable by resulting from changes in exchange rates are not included
the policyholders or insurance companies. It is an integral in CPIs. In contrast, the service charges made by foreign-
part of the gross premium that is not separately identified in exchange dealers are included within the scope of CPIs
practice. As a payment for a service it falls within the scope when households acquire foreign currency for personal use.
of a CPI, but it is difficult to estimate. These charges include not only explicit commission charges
2.136 In the case of nonlife insurance, the net premium but also the differences between the buying or selling rates
is essentially a transfer that goes into a pool covering the offered by the dealers and the average of the two rates (2008
collective risks of policyholders as a whole. As a transfer, it SNA, paragraph 11.34).
falls outside the scope of a CPI. In the case of life insurance, 2.142 Transactions in financial assets do not change
the net premium is essentially a form of financial invest- wealth and there is no consumption involved. A transaction
ment. It constitutes the purchase of a financial asset, which of a financial asset merely rearranges the individual’s asset
is also outside the scope of a CPI. portfolio by exchanging one type of asset for another. For
2.137 Finally, it may be noted that when insurance is example, when a loan is made, the lender exchanges cash
arranged through a broker or agent separate from the insur- for a financial claim over the debtor. Similarly, the borrower
ance company, the fees charged by the brokers or agents for acquires cash counterbalanced by the creation of an equal
their services are included within the scope of the CPI, over liability. Such transactions are irrelevant for CPI purposes.
and above the service charges made by the insurers. 2.143 In general, when a household borrows from
financial institutions, including moneylenders, the borrowed
Gambling funds may be used for a variety of purposes including the
2.138 Gambling. The amounts paid for lottery tickets purchase of assets such as dwellings or financial assets (for
or placed in bets also consist of two elements that are usu- example, bonds or shares), as well as the purchase of expen-
ally not separately identified—the payment of an implicit sive goods and services. Similarly, the credit extended to the
service charge (part of consumption expenditure) and a cur- holder of a credit card can be used for a variety of purposes.
rent transfer that enters the pool out of which the winnings In itself, the creation of a financial asset and liability by new
are paid (2008 SNA, paragraph 8.136). Only the implicit borrowing has no impact on a CPI. There is no good or ser-
or explicit service charges payable to the organizers of the vice acquired, no expenditure, and no price.
gambling and taxes fall within the scope of a CPI. The ser- 2.144 Hire purchase (see 2.147) and mortgage loans
vice charges are usually calculated at an aggregate level as must be treated consistently with other loans. The fact that
the difference between payables (stakes) and receivables certain loans are conditional on the borrower using the funds
(winnings). for a particular purpose does not affect the treatment of the
loan itself. Moreover, conditional loans are by no means
Transactions in Financial Assets confined to the purchase of durable goods on “hire pur-
2.139 Transaction in financial assets. Financial assets chase.” Conditional personal loans may be made for other
are not consumption goods or services (2008 SNA, para- purposes, such as large expenditure on education or health.
graph 11.8). The creation of financial assets and liabilities, In each case, the contracting of the loan is a separate transac-
or their extinction (for example, by lending, borrowing, and tion from the expenditure on the good or service and must
repayments), are financial transactions that are quite differ- be distinguished from the latter. The two transactions may
ent from expenditure on goods and services and take place involve different parties and may take place at quite differ-
independently of them. The purchase of a financial asset is ent times.
not expenditure on consumption, being a form of financial 2.145 Although the provision of finance is a separate
investment. transaction from the purchase of a good or service for which
2.140 Some financial assets, notably securities in the it is used, it may affect the price paid. Each case needs to be
form of bills, bonds, and shares, are tradable and have mar- carefully considered. For example, suppose the seller agrees
ket prices (2008 SNA, paragraph 11.64). They have their to defer payment for one year. The seller appears to make an
own separate price indices, such as stock market price indi- interest-free loan for a year, but this is not the economic real-
ces. Many of the financial assets owned by households are ity. The seller makes a loan, but it is not interest-free. Nor is
acquired indirectly through the medium of pension pro- the amount lent equal to the “full” price. Implicitly, the pur-
grams and life insurance. Excluding the service charges, chaser issues a short-term bill to the seller to be redeemed
pension contributions by households are similar to payments one year later and uses the cash received from the seller to
for life insurance premiums. They are essentially forms of pay for the good. However, the present value of a bill at
investment and are thus excluded from CPIs. In contrast, the the time it is issued is its redemption value discounted by
explicit or implicit fees paid by households for the services one year’s interest. The amount payable by the purchaser at
43
the time the purchase of the good actually takes place is the Nominal interest is a composite payment covering four main
present discounted price of the bill and not the full redemp- elements whose mix may vary considerably. The first com-
tion price to be paid one year later. It is this discounted price ponent is the pure interest charge, which is the interest that
that should be recorded for CPI purposes. The difference would be charged if there were perfect capital markets and
between the discounted price and the redemption price is, perfect information. The second component is a risk pre-
of course, the interest that the purchaser implicitly pays on mium that depends on the creditworthiness of the individual
the bill over the course of the year. This way of recording borrower and can be regarded as a built-in insurance charge
corresponds to the way in which bills and bonds are actually against the risk of the debtor defaulting. The third compo-
valued on financial markets and also to the way in which nent is a service charge incurred when households borrow
they are recorded in both business and economic accounts from financial institutions that make a business of lending
(2008 SNA, paragraph 17.266). Deferring payments in the money. Finally, when there is inflation, the real value of a
manner just described is equivalent to a price reduction and loan fixed in monetary terms (that is, its purchasing power
should be recognized as such in CPIs. The implicit interest over goods and services) declines with the rate of inflation.
payment is not part of the price. Instead, it reduces the price. However, creditors offset the real holding, or capital, losses
This example shows that in certain circumstances the mar- they expect to incur by charging appropriately high rates of
ket rate of interest can affect the price payable but it depends nominal interest. For this reason, nominal interest rates vary
on the exact circumstances of the credit arrangement agreed directly with the expected rate of general inflation, a uni-
between the seller and the purchaser. Each individual case versally familiar phenomenon under inflation conditions. In
needs to be carefully considered on its merits. these circumstances, the main component of nominal inter-
2.146 This case needs to be clearly distinguished from est may therefore be the built-in payment of compensation
hire purchase, considered in 2.147, when the purchaser actu- from the debtor to the creditor to offset the latter’s real hold-
ally pays the full price and borrows an amount equal to the ing loss. When there is very high inflation it may account for
full price while contracting to make explicit interest pay- almost all of the nominal interest charged.
ments in addition to repaying the amount borrowed. 2.150 The treatment of the first component, the pure
2.147 Hire purchase. Under a hire purchase agreement, interest charge, is somewhat controversial but this com-
a buyer leases a good over a period of time and does not ponent may account for only a small part of the nominal
acquire ownership until the full amount of the contract is interest charged. The treatment of the second component,
paid (2008 SNA, paragraph 9.73). In the case of a durable insurance against the risk of default, is also somewhat
good bought on hire purchase, it is necessary to distinguish controversial because it can be difficult to measure in
the de facto, or economic, ownership of the good from the practice.
legal ownership. The time of acquisition is the time the hire 2.151 The third component, the payment of compensa-
purchase contract is signed and the purchaser takes posses- tion for the creditor’s real holding loss, is clearly outside
sion of the durable. From then onward, it is the purchaser the scope of a CPI. It is essentially a financial transaction.
who uses it and derives the benefit from its use. The pur- It may account for most of nominal interest under inflation
chasing household becomes the de facto owner at the time conditions.
the good is acquired, even though legal ownership may not 2.152 The fourth component constitutes the purchase
pass to the household until the loan is fully repaid. of a service from financial institutions whose business is to
2.148 By convention, the purchasing household is make funds available to borrowers. It is known as the finan-
treated as buying the good at the time possession is taken cial service indirectly measured and clearly falls within the
and paying the full amount in cash at that point. At the same scope of a CPI. It is included in COICOP (COICOP 2018,
time, the purchaser borrows, either from the seller or some class 12.2.1). The service charge is not confined to loans
financial institution specified by the seller, a sum sufficient made by “financial intermediaries,” institutions that bor-
to cover the purchase price and the subsequent interest pay- row funds in order to lend them to others. Financial institu-
ments. The difference between the cash price and the total tions that lend out of their own resources provide the same
sum of all the payments to be made is equal to the total inter- kind of services to borrowers as financial intermediaries.
est payable. The relevant price for CPI purposes is the cash When sellers lend out of their own funds, they are treated
price payable at the time the purchase takes place whether as implicitly setting up their own financial institution that
or not the purchase is facilitated by some form of borrow- operates separately from their principal activity. The rates
ing. The treatment of hire purchase is the same as that of of interest of financial institutions also include service
financial leases whereby fixed assets, such as aircraft, used charges.
for purposes of production are purchased by a financial 2.153 It is clear that interest payments should not be
institution and leased to the producer for most or all of the treated as if they were just pure interest or even pure interest
service life of the asset (2008 SNA, paragraph 17.304). This plus a risk premium. It is very difficult to disentangle the
is essentially a method of financing the acquisition of an various components of interest. It may be practically impos-
asset by means of a loan and needs to be distinguished from sible to make realistic and reliable estimates of the service
operational leasing, such as hiring out cars for periods of charges embodied in most interest payments. Moreover, for
time (2008 SNA, paragraph 17.301). The treatment of hire CPI purposes it is necessary to estimate not only the values
purchase and financial leasing outlined here is followed in of the service charges but changes in the prices of the ser-
both business and economic accounting. vices over time. Given the complexity of interest flows and
2.149 Interest payments. The treatment of interest pay- the fact that the different flows need to be treated differently,
ments on the various kinds of debt that households may have payments of nominal interest should not be included in a
incurred raises both conceptual and practical difficulties. CPI, especially in inflation conditions.
44
Household Production have with the market and through which they experience the
2.154 Household production. Households can engage effects of inflation. It therefore seems preferable to record
in various kinds of productive activities that may be either the actual prices of the inputs and not the imputed prices of
aimed at the market or intended to produce goods or services the outputs in CPIs.
for own consumption (2008 SNA, paragraphs 1.41 and 1.42). 2.160 Housing services produced for own consump-
2.155 Households may engage in business or commer- tion. There is no unique recommendation for the treatment
cial activities such as farming, retail trading, construction, of owner-occupied housing services in the CPI, as described
and the provision of professional or financial services. Goods in Chapter 11. There is no general consensus on what con-
and services that are used up in the process of producing stitutes best practice. Conceptually, the production of hous-
other goods and services for sale on the market constitute ing services for own consumption by owner-occupiers is no
intermediate consumption (2008 SNA, paragraph 6.213). different from other types of own-account production taking
They are not part of the final consumption of households. place within households. The distinctive feature of the pro-
The prices of intermediate goods and services purchased by duction of housing services for own consumption, as com-
households are not to be included in CPIs. In practice, it pared with other kinds of household production, is that it
is sometimes difficult to draw a clear distinction between requires the use of an extremely large fixed asset in the form
intermediate and final consumption, as the same goods and of the dwelling. In economics, and also national account-
services may be used for either purpose. ing (2008 SNA, paragraphs 6.34 and 6.117), a dwelling is
2.156 Households do not in fact consume directly all regarded as a fixed asset so that the purchase of a dwelling
the goods and services they acquire for purposes of con- is classified as gross fixed capital formation and not as the
sumption. Instead, they use them as inputs into the pro- acquisition of a durable consumer good.
duction of other goods or services which are then used to 2.161 It is important to note that there are two quite
satisfy their needs and wants. For example, basic food- distinct service flows involved in owner-occupied housing
stuffs such as flour, cooking oils, raw meat, and vegeta- services. One consists of the flow of capital services pro-
bles may be processed into bread, cakes, or meals with the vided by the dwelling which are consumed as inputs into
assistance of other inputs including fuels, the services pro- the production of housing services. The other consists of the
vided by consumer durables, such as fridges and cookers, flow of housing services produced as outputs which are con-
and labor of members of the household. Inputs of materi- sumed by members of the household. The two flows are not
als, equipment, and labor are used to clean, maintain, and the same. The value of the output will be greater than that
repair dwellings. Inputs of seeds, fertilizers, insecticides, of the input. The capital services are defined and measured
equipment, and labor are used to produce vegetables, in exactly the same way as the capital services provided by
flowers, and so on. other kinds of fixed assets, such as equipment or structures
2.157 The practical alternative, if the output is not other than dwellings. As explained in more detail in Chapter
intended for sale, is to treat the goods and services acquired 11, the value of the capital services is equal to the user cost
by households on the market for use as inputs into the vari- and consists primarily of two elements, depreciation and
ous kinds of household production activities as if they were interest, or capital costs. Capital costs are incurred whether
themselves final consumer goods and services. They provide or not the dwelling is purchased by borrowing on a mort-
utility indirectly assuming that they are used exclusively gage. When the dwelling is purchased out of own funds, the
to produce goods and services that are directly consumed interest costs represent the opportunity cost of the capital
by households. This is the practical solution that is gener- tied up in the dwelling, that is, the forgone interest that could
ally adopted not only in CPIs but also in national accounts have been earned by investing elsewhere.
where household expenditure on such items is classified as 2.162 There are three main options for the treatment
final consumption. Although this seems a simple and con- of own-account production and consumption of owner-
ceptually acceptable solution, exceptions may be made for occupied housing services in CPIs. One is to price the out-
specific kinds of household production that are particularly put of housing services consumed by owner-occupiers. The
important and whose outputs can readily be identified. second is to price the inputs, including the inputs of capital
2.158 Subsistence agriculture. In the national accounts, services. The third is to include the price of the dwelling.
an attempt is made to record the value of the agricultural If housing services are to be treated consistently with other
output produced for own consumption. In some countries, forms of production for own consumption by households,
subsistence agriculture may account for a large part of the the input approach should be adopted. The production and
production and consumption of agricultural produce. The consumption of housing services by owner-occupiers may,
national accounts require such outputs to be valued at their however, be considered to be so important as to merit spe-
market prices (2008 SNA, paragraphs 24.47–24.49). For an cial treatment.
index used for monetary policy purposes or as a general 2.163 If it is decided to price the outputs, the prices may
macroeconomic measure of inflation, it is not appropriate to be estimated using the market rents payable on rented accom-
follow this procedure for CPI purposes. modation of the same type. This is described as the rental
2.159 A CPI may record either the actual input prices equivalence approach. One practical problem is that there
or the imputed output prices, but not both. If the imputed may be no accommodation of the same type that is rented
output prices for subsistence agriculture are included in a on the market. For example, there may be no rental market
CPI, the prices of the purchased inputs should be excluded. for rural dwellings in developing countries where most of
This could remove from the index most of the market trans- the housing may actually be constructed by the households
actions made by such households. Expenditure on inputs themselves. Another problem is to ensure that the market
may constitute the principal contact that the households rents do not include other services, such as electricity or
45
heating, that are additional to the housing services. An addi- flows from durables are included in CPIs at present because
tional challenge is that market rents, like the rentals charged their acquisitions were included; similarly if the prices of
when durables are leased, have to cover the operating expenses dwellings are included in CPIs, the service flows would
of the renting agencies as well as the costs of the housing ser- have to be excluded. As explained in Chapter 11, the acqui-
vices themselves and also provide some profit to the owners. sitions approach may give insufficient weight to durables
Finally, rented accommodation is inherently different from and dwellings over the long term because it does not take
owner-occupied housing in that it may provide the tenants account of the capital costs incurred by the owners of the
with more flexibility and mobility. The transaction costs assets.
involved in moving from one house to another may be much
less for tenants than for owners. Treatment of Some Specific Household
2.164 In principle, if the output approach is adopted, Expenditure
then the prices of the inputs into the production of hous- 2.169 Some of the expenditure made by households
ing services for own consumption, such as expenditure on may not be on goods and services for household consump-
repairs, maintenance, and insurance, should not be included. tion and would therefore fall outside the scope of a CPI. One
Otherwise, there would be double counting. major category consists of the business expenditure made by
2.165 The alternative is to price the inputs into the households.
production of housing services for own consumption in
the same way that other forms of production for own con- Fees of Agents and Brokers
sumption within households are treated. In addition to 2.170 Fees of agents and brokers. The 2008 SNA clas-
intermediate consumption, such as expenditure on repairs, sifies expenses associated with the transfer of real estate
maintenance, and insurance, the costs of the capital services (including real estate agents’ commissions) as part of gross
must be estimated and their prices included in the CPI. The fixed capital formation (2008 SNA, paragraphs 10.48–
technicalities of estimating the values of the flow of capi- 10.52). Although harmonization is desirable, in some coun-
tal services are dealt with in Chapter 10 of the publication tries some CPI concepts do not precisely follow the national
Consumer Price Index Theory. As in the case of other types accounts concepts.
of production for own consumption by households, it is not 2.171 When a house is purchased for own use by an
appropriate to include the estimated costs of labor provided owner-occupier, it can be argued that the transfer costs asso-
by the owners. ciated with purchase (and sale) should be treated as con-
2.166 Whether the input or the output approach is sumption expenditure. The fees paid to an agent to buy or
adopted, it can be challenging to estimate the relevant prices. sell houses are included in many national CPIs, provided
The practical difficulties experienced may sometimes be so that the house is to be occupied by the owner and not rented
great as to lead compilers and users to query the reliability to a third party.
of the results. There is also some reluctance to use imputed
prices in CPIs, whether the prices refer to the inputs or the Undesirable, Informal, or Illegal Goods and Services
outputs. It has therefore been suggested that the attempt to 2.172 Nonobserved economy: informal or illegal pro-
measure the prices of housing service flows should be aban- duction of goods and services (2008 SNA, paragraphs 6.39–
doned. Instead, it may be preferred to include the prices of 6.48). As described in 2.56, HFCE includes the expenditure
the dwellings in the CPI. In most cases, these are observable incurred by resident households on consumption goods and
market prices, although many dwellings, especially in rural services, and therefore these fall within the scope of a CPI,
areas in developing countries, are also built by their owners, irrespective of whether their production, distribution, or
in which cases their prices still have to be estimated on the consumption is informal or illegal. Particular kinds of goods
basis of their costs of production. Ultimately, it may be use- or services must not be excluded from the CPI because their
ful for CPI compilers to rely on owner-occupied dwelling consumption is socially discouraged, or their production
services estimated in the national accounts. informal or illegal. Such exclusions could be quite arbitrary
2.167 Including the prices of dwellings in CPIs should and undermine the objectivity and credibility of the CPI. In
involve a significant change in the scope of the index. A the case where certain goods or services have been excluded
dwelling is clearly an asset and its acquisition is gross fixed from the index, these exclusions should be clearly docu-
capital formation and not HFCE. While the same argument mented and explained.
applies to durables, there is a substantial difference of degree 2.173 First, it should be noted that some goods and ser-
between a durable consumption goods and a dwelling as vices might be deemed to be undesirable at sometimes and
reflected by the considerable differences in their prices and desirable at others, or vice versa. Similarly, some goods or
their service lives. In principle, therefore, extending the services may be deemed to be undesirable in some countries
scope of a CPI to include dwellings implies extending the but not in others at the same point of time. The concept of
scope of the index to include gross fixed capital formation. an undesirable good or service is inherently subjective and
2.168 The advantage of including dwellings in the CPI somewhat arbitrary and volatile.
is that it does not require estimates of either the input or out- 2.174 Second, if it is accepted that some goods and ser-
put service flows, but conceptually it deviates significantly vices may be excluded on the grounds that they are unde-
from the concept of a CPI as traditionally understood. In sirable, the index is thereby exposed to actual or attempted
the case of both durable consumption goods and dwellings, manipulation by pressure groups.
the options are either to record the acquisitions of the assets 2.175 Third, attempts to exclude certain goods or ser-
in the CPIs at their market prices or to record the estimated vices by pressure groups may be based on a misunderstand-
prices of the service flows, but not both. Just as no service ing of the implications of so doing. For example, if the CPI
46
is used for escalating incomes, it may be felt that households expenditure (that is, expenditure less receipts) is zero, so
ought not be compensated for increases in the prices of cer- no weight is attached to purchases and sales from one
tain undesirable products. However, excluding them does household to another.
not imply lowering the index. A priori excluding some items • From another household via a dealer—in principle, house-
is just as likely to increase the CPI as reduce it, depend- hold expenditure on the services of the dealers is given by
ing on whether the price change for the item in question is the values of their margins (the difference between their
below or above the average for other goods and services. buying and selling prices). These intermediation services
For example, if it is decided to exclude tobacco from a CPI should be included in CPIs. They should be treated in the
and the price increase for tobacco products is below aver- same way as the fees charged by financial intermediar-
age, excluding tobacco actually increases the real income ies. The margins may be extremely difficult to estimate in
of smokers. practice. Care should be taken to include trade-ins either
2.176 While goods and services that households con- as purchases by the dealers or receipts of households.
sume should not, in principle, be excluded from a CPI • Directly from another sector (for example, from a corpo-
because they are acquired in the informal economy or even ration or from abroad)—the weight would be household
illegally, it may be impossible to obtain the requisite data on purchases of the second-hand goods from other sectors
the expenditure or the prices, especially on illegal goods and less sales to other sectors.
services. They may well be excluded in practice.
• From another sector (for example, corporation or from
abroad) via a dealer—the appropriate weight is given by
Luxury Goods and Services
household purchases from dealers less any household
2.177 Luxury goods and services. When a CPI is used
sales to dealers plus the aggregate of dealers’ margins on
as an index of general inflation, it ought to include all house-
the goods that they buy from and resell to households.
holds regardless of their socioeconomic group and also all
consumer goods and services regardless of how expensive 2.181 In some countries, many of the durables pur-
they are. Similarly, the scope of an index used for purposes chased by households, especially vehicles, may be imports
of escalating incomes should include all the goods and ser- of second-hand goods from other countries. The prices and
vices purchased by the reference households, irrespective of expenditure on these goods enter the CPI in the same way
whether any of these goods and services are considered to be as those for newly produced goods. Similarly, in some coun-
luxuries or otherwise unnecessary. tries, there may be significant net purchases of second-hand
2.178 If the reference households are confined to a selected vehicles by households from the corporations’ sector.
group of households, the index will effectively exclude all
those items that are purchased exclusively by households that Imputed Expenditure on Goods and Services
are not in the group. For example, excluding the wealthiest 2.182 Imputed expenditure on goods and services. As
5 percent of households will, in practice, exclude many lux- explained in paragraphs 2.59–2.67, many of the goods and
ury items from the scope of the index. As already noted, such services acquired and used by households for their final
households may be excluded for various reasons, including the consumption are not purchased in monetary transactions but
unreliability of their expenditure data and the fact that collect- are acquired through barter or as remuneration in kind, or
ing prices for some items purchased exclusively by a relatively are produced by households for their own consumption. It
small minority of households may not be cost-effective. How- is possible to estimate what households would have paid if
ever, once the group of reference households has been decided they had purchased these goods and services in monetary
and defined, judgments should not be made about whether to transactions or, alternatively, what it cost to produce them.
exclude certain part of their expenditure that is considered to In other words, values may be imputed for nonmonetary
be nonessential or on luxuries. transactions.
2.183 The extent to which it is desirable to include
imputed expenditure within the scope of a CPI depends
Second-Hand Goods
partly on the main purpose of the index. If the CPI is
2.179 Second-hand goods. Markets for used or second-
intended to be a measure of consumer inflation, it can be
hand goods exist for most durable goods. HFCE includes
argued that only monetary transactions should be included.
expenditure on second-hand goods and these are therefore
Inflation is a monetary phenomenon measured by changes
within the scope of a CPI. Household sales of durables con-
in prices recorded in monetary transactions. Transactions
stitute negative expenditure, and the weights for second-hand
include the buying and selling of a good or service. Mon-
goods are based on household net expenditure (that is, total
etary transactions occur when a seller exchanges ownership
purchases less sales). The total expenditure on a particular
of a good or service in exchange for some form of monetary
type of second-hand good is a function of the rate at which
payment. Even when the main use of a CPI is for indexation
it is bought and sold (that is, a higher turnover rate or num-
purposes, it can be argued that it should only reflect changes
ber of transactions gives a higher total expenditure). A higher
in the monetary prices actually paid by the reference popu-
turnover does not, however, increase the rate at which any
lation. For example, consistent with the objective of moni-
individual good can be used for purposes of consumption or
toring inflation in the EU, the aim of the HICP compiled
the flow of services that may be obtained from the good.
by Eurostat is to measure inflation faced by consumers.
2.180 Households may buy second-hand goods through
The concept of HFMCE used in the HICP defines both the
any of the following routes:
goods and services to be covered, and the price concept to
• Directly from another household—the selling house- be used (that is, prices net of reimbursements, subsidies, and
hold will record the proceeds of the sale as receipts. Net discounts). HFMCE refers only to monetary transactions
47
and includes neither consumption of own production (for 2.189 Given the practical difficulties in correctly recording
example, agricultural goods or owner-occupied housing ser- all these types of price decreases, it is usual to reflect discounts
vices), nor consumption of goods and services received as and rebates only if widely available. More and more countries
income in kind. include discounts associated with loyalty cards because most
2.184 Discounts, rebates, loyalty programs, and “free” shoppers obtain and use the loyalty card, effectively meeting
products. CPIs should take into account the effects of the criteria that the discount or rebate should be widely avail-
rebates, loyalty programs, and money-off vouchers. Given able. Discounts during seasonal sales may be recorded pro-
that a CPI is meant to cover all the reference households, vided that the quality of the goods does not change.
whether in the country as a whole or in a particular region, 2.190 As noted in Chapter 10, scanner data more effec-
discounts should be included even if they are available only tively reflect discounts, sales, and promotions.
to certain households or to consumers satisfying certain pay-
ment criteria.
2.185 It may be difficult to record discriminatory or Consumer Price Index
conditional discounts for practical reasons. When only one Classifications
selected group of households can enjoy a certain discount
on a specific product, the original stratum for that product is 2.191 The classification system upon which any CPI is
split into two new strata, each experiencing different price built provides the structure essential for many stages of CPI
changes and each requiring a weight. So, unless weight ref- compilation. Most obviously, it provides the weighting and
erence period expenditure for all possible strata are known, aggregation structure, but it also provides the basis for strati-
it is not possible to record discriminatory discounts cor- fication of products in the sampling frame, at least down to a
rectly. Similarly, with conditional discounts (for example, certain level of detail, and it dictates the range of subindices
discounts on utility bills for prompt payment), it can be dif- available for publication.
ficult to record the effect of the introduction of such offers 2.192 The international standard for classification of
unless data are available on the proportion of customers tak- individual final consumption expenditure is the Classifi-
ing advantage of the offer. These kinds of practical problems cation of Individual Consumption According to Purpose
also arise when there is price discrimination and the sellers (COICOP). COICOP is part of a set of classifications of
change the criteria that define the groups to whom different expenditure according to purpose, also known as func-
prices are charged, thereby obliging some households to pay tional classifications, and have formed an integrated part of
more or less than before without changing the prices. These the SNA since 1968 (2008 SNA, paragraphs 29.9–29.20).
cases are discussed further in Chapter 5. COICOP covers the individual final consumption expen-
2.186 Although it is desirable to record all price diture incurred by three institutional sectors: households,
changes, it is also important to ensure that the qualities of NPISHs, and general government. Individual final con-
the goods or services for which prices are collected do not sumption expenditure is that which benefits individual per-
change in the process. While discounted prices may be col- sons or households.
lected during general sales or discount seasons, care should 2.193 A few countries continue to use a country-specific
be taken to ensure that the quality of the products being classification system and have not adopted COICOP. To
priced has not deteriorated. enhance international comparison, these countries should
2.187 The borderline between discounts and rebates can provide bridge tables to map their national classification
be hazy and is perhaps best drawn according to timing. In systems to COICOP.
other words, a discount takes effect at the time of purchase, 2.194 While part of the SNA, COICOP is intended for
whereas a rebate takes effect some time later. Under this clas- use in several other statistical areas. In addition to CPIs,
sification, money-off vouchers are discounts and, as with the COICOP is also used for HBS, analysis of living standards,
conditional discounts mentioned in paragraph 2.185, can only and for compilation of PPPs.
be taken into account in a CPI if they relate to a single product 2.195 COICOP was revised in 2018 to reflect changes
and if the take-up rate is known at the time of CPI compi- in consumption patterns and the emergence of new goods
lation. Since this is highly unlikely, the effect of money-off and services since the previous version, introduced in 1999.
vouchers is usually excluded from a CPI. It should be noted The updated version, referred to as COICOP 2018, consists
that the discount is recorded only when the voucher is used, of 15 divisions:
not when the voucher is first made available to the consumer. • Divisions 01–13 covering the final consumption expendi-
2.188 Rebates may be made in respect of a single product ture of households
(for example, air miles), or may be more general (for exam-
• Division 14 covering the final consumption expenditure
ple, supermarket loyalty programs where a $10 voucher is
of NPISHs
awarded for every $200 spent). As with discounts discussed
previously, such rebates can only be recorded as price falls if • Division 15 covering the individual consumption expen-
they relate to single products and can be weighted according diture of general government
to take-up. Bonus products provided “free” to the consumer, 2.196 COICOP 2018 has four levels of detail organized
either by larger pack sizes or offers such as “two packs for in a hierarchical structure—divisions, groups, classes, and
the price of one,” should be treated as price reductions, subclasses:
although they may be ignored in practice when the offers
are only temporary and quickly reversed. When permanent • Division (two-digit level), for example, 03 Clothing and
changes to pack sizes occur, quality adjustments should be footwear
made (see Chapter 6). • Group (three-digit level), for example, 03.1 Clothing
48
• Class (four-digit level), for example, 03.1.2 Garments only once or whether they can be used repeatedly or con-
• Subclass (five-digit level), for example, 03.1.2.1 Gar- tinuously over a period of more than one year. Semidurable
ments for men or boys goods differ from durable goods in that their expected life-
time of use, though more than one year, is often significantly
2.197 Divisions 01–13, which cover households, shorter and their purchasers’ value is substantially less.
include 63 groups, 186 classes, and 338 subclasses. The full 2.202 Although a systematic separation between goods
COICOP 2018 structure can be found in Appendix 3. and services is applied, some classes and subclasses con-
2.198 Classifying according to purpose. COICOP tain both because it is difficult for practical reasons to break
groups HFCE on individual goods and services according to them down into goods and services. Such classes and sub-
the purpose they are deemed to fulfill, such as nourishing the classes are usually assigned an S, as the service component
body, preventing and curing illness, acquiring knowledge, is considered to be predominant. Similarly, there are classes
or traveling from one place to another. The principle of clas- that contain either both nondurable and semidurable goods
sifying according to purpose means that where similar or or both semidurable and durable goods. Such classes and
related products exist in either physical or virtual forms (for subclasses are assigned an ND, SD, or D according to which
example, books, music, videos, or games) the product should type of good is considered to be predominant.
be categorized in a unique class based on the predominant 2.203 Bundled goods and services. Single expenditure
purpose. For example, the purchase of electronic or virtual outlays (that is, where there is no itemized price informa-
books (for example, eBooks or audiobooks) should be clas- tion for the individual goods or services) may sometimes
sified in the same class or subclass as paper books because comprise a bundle of goods and services that serve differ-
they are used for the same purpose. Similarly, software and ent purposes. Examples include telecommunication (for
apps may provide the household with a specific service. If example, payment of one price for multiple services that
the payment is actually not for the software but for an asso- include mobile phone, internet, television, and landline
ciated service, which is provided with the help of the soft- telephone); package tours which include payment for trans-
ware or the app, the expenditure should be classified under port, accommodation, and catering services; education ser-
the corresponding service. As a general rule, expenditure vices that include payment for transport, accommodation,
on second-hand goods are classified together with the new and educational materials; inpatient hospital services that
goods since they are used for the same purpose. One excep- include payments for medical treatment, accommodation,
tion is the recording of motor cars, where the subclass level and catering; and transport services that include meals and
allows a separate recording of new motor cars and second- accommodation in the ticket price (for example, passenger air
hand motor cars (COICOP 2018 Subclass 07.1.1.2). transport). Single outlays covering two or more purposes and
2.199 Multipurpose goods and services. While most not separately invoiced should be classified according to the
goods and services can be assigned to a single purpose, predominant product or service of the bundle.
some goods and services could plausibly be assigned to
more than one purpose. Examples include motor fuel which
may be used to power vehicles classified as transport as well Key Changes from COICOP 1999 to
as recreational vehicles; bicycles which may be purchased COICOP 2018
for transport or recreational purposes; or sports footwear 2.204 COICOP 2018 reflects changes in consumption
which may be used for sports or for leisure wear. In cases patterns and the emergence of new goods and services.
where goods and services can be used for different purposes, The main changes from COICOP 1999 and COICOP 2018
they should be assigned to the division considered to repre- include the following.
sent the primary or predominant purpose.
2.200 Disaggregation of COICOP. The detail provided by Introduction of a New Subclass Level
COICOP, even at its most detailed level, may not be sufficient 2.205 COICOP 2018 introduces an additional fifth-digit
for the required analysis or to meet country-specific needs. In level denominated subclass that was not part of COICOP
such cases, classes or subclasses can be further subdivided as 1999. The introduction of these new subclasses facilitates
needed. There are clear advantages in maintaining the basic further harmonization of data collection and aggregation,
structure of COICOP to facilitate comparison between coun- improving comparability of the resulting statistics. It also
tries, over time and between different statistical domains improves the correspondence with the Central Product Clas-
such as CPIs, household expenditure statistics, and national sification to more easily reconcile with production data.
accounts aggregates. It is recommended that additional
detailed categories created to meet specific needs still can be Restructuring of Division 06 Health
aggregated into the existing COICOP Class or Subclass. 2.206 Division 06 is restructured to allow for a better
2.201 Type of product. COICOP classes and subclasses alignment of COICOP with the International Classification
are divided into services (S), nondurables (ND), semidurables for Health Accounts and its family of classifications.
(SD), and durables (D). This additional classification facili-
tates other analytical applications. For example, an estimate Restructuring between Division 08
may be required of the stock of consumer durables held Information and Communication, and
by households, in which case the goods in those COICOP Division 09 Recreation, Sport, and
classes that are identified as “durables” provide the basic ele- Culture, and Renaming of the Divisions
ments for such estimates. As explained earlier in paragraphs 2.207 To better reflect household use of information and
2.64 and 2.65 the distinction between nondurable goods and communication technology, a number of goods and services
durable goods is based on whether the goods can be used have been moved from Division 09 to Division 08. Division
49
08 has been renamed Information and Communication Key Recommendations

(formerly Communication) and Division 09 was renamed
Recreation, Sport, and Culture (formerly Recreation and • Consult with key data users to identify and define uses
Culture) to better reflect the coverage of the divisions. for CPI data. This ensures that the data compiled remain
relevant. It is important that NSOs consult with data users
Division 12 Insurance and Financial on a routine and regular basis.
Services, and Division 13 Personal Care,
• When the CPI is used for inflation analysis and monetary
Social Protection, and Miscellaneous
Goods and Services policy purposes, the domestic concept should be used.
2.208 Personal care, social protection, and miscel- • When the CPI is used only for escalating the incomes of res-
laneous goods were included in Division 12 of COICOP idents, it may be appropriate to adopt the national concept.
1999; Division 12 of COICOP 1999 has been divided into • The weights and prices in the CPI should be based on the
two divisions in COICOP 2018—Division 12 Insurance and purchaser’s price. Purchaser’s prices refer to those prices
Financial Services, and Division 13 Personal Care, Social paid by consumers to acquire ownership of goods or ser-
Protection, and Miscellaneous Goods and Services. This vices and include any taxes and service charges on the
change creates two, more homogeneous divisions. products, and taking account of all discounts, subsidies
2.209 In addition to the previously mentioned major and most rebates, even if discriminatory or conditional.
changes, a number of changes were introduced at the more • NSOs are encouraged to provide central banks with
detailed levels for most divisions. detailed weight, item, and price data in anonymized forms
so that they may calculate different measures of core
Implementing COICOP 2018 inflation or for analytical purposes.
• Geographic coverage of expenditure should include all
2.210 Implementation of COICOP 2018 must be done
with care to avoid confusing data users and to prevent any expenditure of households regardless of income, size, or
loss in user confidence that could result from this confusion. location (urban and rural).
NSOs should coordinate the implementation of COICOP • Geographic coverage of price collection should be as
2018 simultaneously across programs (for example, national broad as possible.
accounts and CPI). To further minimize any impact on data • It is critical to communicate and explain changes adopted
users, the introduction of COICOP 2018 should coincide when implementing COICOP 2018. The introduction of
with a routine update of the CPI. COICOP 2018 should coincide with a routine CPI update.
50
Annex 2.1
Use of Price Statistics in the National Accounts—Supply and Use
Uses
Domestic Intermediate Final Consumption Fixed Capital Formation

Production
Agriculture Mining and Construction Wholesale Other Households Government Construction Equipment Exports
Manufacturing and Retail Services
Trade
Agriculture APPI APPI APPI APPI APPI CPI PPI — — XPI/PPI

Mining and PPI PPI PPI PPI PPI CPI PPI — PPI XPI/PPI
Manufacturing
Construction CPPI CPPI CPPI CPPI CPPI CPI PPI CPPI — XPI/CPPI
Supply Wholesale and (a) (a) (a) (a) (a) (a) (a) (a) (a) (a)
Retail Trade
Other Services SPPI SPPI SPPI SPPI SPPI CPI SPPI SPPI XPI
Imports MPI MPI MPI MPI MPI CPI MPI MPI MPI
Labor LCI LCI LCI LCI LCI LCI — — — —
APPI, Agriculture producer price index (output); MPI, Import price index; CPI, Consumer price index; PPI, Producer price index; CPPI, Construction producer
price index (output); SPPI, Services producer price index; LCI, Labor cost index; XPI, Export price index.
(a) Margins—these are combined with prices and goods—not observed separately.
51
EXPENDITURE WEIGHTS AND THEIR SOURCES 3
Introduction by resident households, regardless of whether expenditure
is made within the country or abroad. Under the domestic
3.1 A consumer price index (CPI) is usually calcu- concept, the scope of the CPI (with regard to both prices
lated as a weighted average of the relative price changes of and weights) considers the economic territory and includes
the goods and services covered by the index. The weights expenditure of both residents and nonresidents.
attached to each good or service reflect their relative impor- 3.6 An HBS generally identifies all relevant expendi-
tance as measured by their shares in the total consumption tures made by resident households and may make a dis-
of all households. The weight determines the impact that its tinction between expenditure made within the economic
price change will have on the overall index. The weights territory and abroad. If the main purpose of the index is to
should be made publicly available for the information of measure price changes experienced by resident households,
data users, to ensure public confidence in the index and the weights should, in principle, include their expenditure
enhance transparency. The International Monetary Fund abroad. If, however, the aim is to include the expenditure
maintains a CPI database that includes detailed weight data.1 made by foreign visitors, thus reflecting all purchases of
3.2 Because the weights assigned to the different goods consumer goods and services made by resident or nonresi-
and services in the basket influence changes in the CPI, the dent households within the country, sources other than the
accuracy and reliability of CPI estimates depend upon the HBS must be used. An HBS does not cover expenditure
quality of the weights used. Therefore, it is necessary to made by nonresident households.
have weights that reflect, as closely as possible, current con- 3.7 For practical reasons, even if the weights cover the
sumer expenditure patterns. expenditure made both at home and abroad, prices may be
3.3 Chapter 3 discusses what the CPI weights should collected only for those goods and services acquired within
represent and how they can be derived. First, this chapter the economic territory of the country. Such an approach
provides an overview of the conceptual basis of the weights assumes that the price changes of the goods and services
to clarify what expenditure should be included when devel- acquired abroad are similar to the price changes for the same
oping weights. Next, the main data sources for the weights goods and services acquired at home. Alternatively, it may
are introduced and the practical steps that must be completed be possible to use CPI subindices compiled by the respec-
when deriving weights are discussed. Finally, some special tive countries to measure the price changes of the goods and
cases are discussed at the end of this chapter. services acquired abroad.
3.8 In principle, the weights should represent the whole
Conceptual Basis of the Weights country and all regions should be covered. Separate expen-
diture weights can also be derived for each region, in which
3.4 The expenditure weights used in a CPI have to be case the expenditures must be sufficiently representative at
consistent with the conceptual framework of the index as the regional level. This is especially important if the expen-
discussed in Chapter 2. Expenditure data can be obtained diture pattern differs between regions. For instance, it is
from different data sources with household budget surveys common that the consumption habits vary between urban
(HBSs) being the most common. When exploiting these data and rural areas. The same principle applies to harmonized
sources, the scope and the concept of the index will deter- CPIs, covering a group of countries, where each country can
mine which goods and services should be covered in the be regarded as a “region” with its own national weights.
CPI weights. If countries compile supplementary or alterna- 3.9 The discussion of national and domestic concepts
tive CPIs to meet multiple user needs, separate weighting applies to regional subdivisions. It can sometimes be the
structures would then be estimated, reflecting the different case that a household lives in one region but does most of
concepts or intended uses of the index. The main conceptual its purchases in an adjacent region, particularly if it lives
aspects that must be considered when deriving weights are close to a regional border. Practical issues dictate whether
discussed in paragraphs 3.5–3.18. the expenditure weights (and the prices) should be allocated
to the region of expenditure or the region of residence. In
Geographical Coverage: National versus any case, treatment should be consistent across all regions
Domestic Concept to avoid missing or double counting parts of household
expenditure.
3.5 The geographical coverage of a CPI may follow
either the “national” concept or the “domestic” concept.
The national concept measures price changes experienced Population Coverage
3.10 The target or reference populations will be defined
1
http://data.imf.org/. based on the main purpose and use of the index, as described
53
in Chapter 2. In principle, all types of households should be Expenditures That Are Out of Scope
covered, irrespective of their income or other socioeconomic 3.13 In its role as a measure of total consumer infla-
factors. If any income groups, types of households, or geo- tion, the CPI should, in principle, cover all types of goods
graphic areas are excluded, for example, for cost or practical and services that are consumed by the reference popula-
considerations, this should be explicitly stated in the index tion. Some types of products may be excluded for practi-
metadata. In some countries, the wealthiest households are cal reasons. These may include products that are illegal,
excluded because their expenditure may be atypical or the black market sales, gambling, or prostitution. Some prod-
HBS information may be less reliable. Other countries may ucts may also be excluded because of policy reasons. For
exclude the expenditures of the very poor for the same rea- instance, it can be decided to exclude certain categories
son. Another practice is to compile a CPI which excludes of goods or services for purposes of compiling alternative
both extremes of the income distribution. If the primary use aggregations or for analysis (for example, exclude tobacco
of the CPI is for adjusting incomes of a certain subgroup of from an index that is used for a specific purpose). When-
the population for increases in the cost of living, then such ever certain in-scope goods or services have been excluded
subgroup may be the appropriate target population. from the index, this should be clearly documented and
3.11 The weights may or may not include the expendi- explained to users.
ture made by people living in institutional households. Insti- 3.14 Only household consumption expenditure is rel-
tutional households refer to people living permanently in an evant for the construction of CPI weights. As explained
institution or who may be expected to reside in an institution in Chapter 2, expenditure on assets such as works of art,
for a very long time, as described in Chapter 2. Many countries financial investment (as distinct from financial services),
exclude expenditure of such households in their CPI because payments of social security contributions, fines or income
of the difficulty of obtaining reliable expenditure information, taxes, interest payments, or repayments of debts are not
or because the expenditure associated with such households considered to be household consumption expenditure and
is unlikely to be significant in comparison with private house- should be excluded from the coverage of the weights and
holds. The choice to include or exclude the expenditure made the index.
by institutional households may depend on the main source 3.15 Business-related expenditure is explicitly excluded
for the weights. While weights primarily based on national from the scope of a CPI. Consequently, this expenditure
accounts household final consumption expenditure (HFCE) must also be excluded from the CPI weights. Households
data may include consumption of institutional households, the may own unincorporated or informal enterprises, whereby
weights primarily based on an HBS would exclude consump- expenditure for some products is used partly for business
tion expenditures of institutional households. In considering purposes and partly for final consumption. In principle, only
the practical challenges relating to the inclusion of institu- the portion used for final consumption should be included in
tional households in a CPI, two questions need to be asked. the CPI weights.
First, is the expenditure pattern of institutional households
likely to be significantly different from private households?
Second, even if the answer is yes, would their exclusion from Democratic and Plutocratic Weights
the CPI be likely to significantly affect the national CPI? 3.16 The use of aggregated expenditure to derive CPI
weights reflects the principle that each household con-
tributes to the weights with an amount proportional to its
Monetary and Nonmonetary Transactions expenditure. This is referred to as plutocratic weighting
3.12 A portion of HFCE consists of nonmonetary trans- and means that the expenditure patterns of high-spending
actions, such as expenditure on goods and services produced households have more influence on the index. The use of
for own account and remuneration in kind. Furthermore, plutocratic weights is generally considered more appropri-
as described in Chapter 2, the broader concept of house- ate particularly for CPIs which have been constructed to
hold actual final consumption includes goods and services be a general measure of inflation, for national accounts
provided without charge or subsidized by governments and deflation, and as the basis for monetary policymaking. In
nonprofit institutions serving households. Depending on the principle, it is also possible to derive democratic weights,
main use of the CPI, a decision must be made whether to where each household is given equal weight. While demo-
include nonmonetary transactions in the weights or if it is cratic weights may be considered appropriate for an index
more appropriate to limit the scope of the index to mone- used to reflect the consumer inflation experience of the
tary transactions only. Production for own consumption is “typical” or “average” household, they are rarely used in
treated by some countries as within the scope of the CPI, practice. If all households have similar expenditure pat-
and in other countries as out of scope. In some countries, terns, the democratic and plutocratic approaches lead to
own-account production constitutes a significant portion of similar results. However, most often the expenditure pat-
HFCE. It can be argued that while it is part of gross domestic tern depends on the total level of expenditure of a house-
product and should therefore be included to improve consis- hold. There can be significant differences between both
tency with national accounts, and especially for producing approaches, especially if the distribution of household
deflators, it is not necessarily appropriate for a CPI used as expenditure is unequal. Table 3.1 presents an example. The
a general measure of inflation or for indexation where the lower-spending household (household 1) spends in rela-
narrowest concept of consumption, based on monetary trans- tive terms more on food than the higher-spending house-
actions, is used. If the CPI includes own-account production, hold (household 2). Consequently, the weight for food is
the weights should include a valuation of the physical quanti- higher in the democratic approach than in the plutocratic
ties of such products, the latter often derived from the HBS. approach.
54
Expenditure Weights and Their Sources
Table 3.1 Example of Plutocratic versus Democratic Weights

Expenditures for Household 1 Expenditures for Household 2 Expenditures for Plutocratic Weights Democratic Weights
Households 1 & 2
Value Share (percent) Value Share (percent) Value Weights (percent) Weights (percent)
(a) (b) (c) (d) (e) (f) (g)
Food 10 33 20 20 30 23 27
Other Goods and 20 67 80 80 100 77 73
Services
Total 30 100 100 100 130 100 100
3.17 In Table 3.1, plutocratic weights are calculated level. These product categories are not part of COICOP, but
first by summing the expenditure made by each household in many cases, more detailed breakdowns of COICOP sub-
in each group. Expenditure on food is summed (10 + 20) classes are needed for CPI purposes to better reflect country-
to derive the total expenditure on food (30). The same is specific circumstances and needs. For example, a country
done for expenditure on other goods and services (20 + may develop a more detailed weighting structure for “Rice”
80) to derive total expenditure on other goods and services to include specific types of rice (for example, basmati rice
(100). The total expenditure for each group is then summed or long-grain rice).
to derive total expenditure on all items (30 + 100 = 130). 3.21 The weights for the COICOP groups, classes, and
Plutocratic weights reflect the relative importance of each subclasses are their shares in the total consumption expen-
group (30/130 = 23 percent and 100/130 = 77 percent). For ditures of the reference population. The data sources used
the democratic weights, an average of the individual shares to derive these shares are discussed in the following text. In
for each household is calculated and used for the weight. addition, the weight for a subclass can be further stratified by
The democratic weight for food equals the average of the region, by outlet or outlet type, or by a combination of both
shares for each household ([33 + 20]/2 = 27 percent). The region and outlet. The elementary aggregate weights are the
same calculation is made for other goods and services ([67 + stratum weights according to expenditure class or subclass,
80]/2 = 73 percent). region, and type of outlet. If no breakdown by region or out-
3.18 For analytical purposes, additional weighting let is used, the subclass becomes the elementary aggregate.
structures may be derived to measure the inflation experi- 3.22 The weight of an elementary aggregate should
ence of different subgroups of households. Such an analysis reflect the expenditure on the entire elementary aggregate
is typically conducted by classifying households according and not the weights of the outlets and varieties that have been
to a sociodemographic variable such as income, age, or edu- chosen to represent it. For instance, the weight for the sub-
cational level. Depending on the objective of the analysis, class “Rice” should be based on the total expenditure made
either democratic or plutocratic weights could be derived for on rice, although the rice varieties selected for regular
the different household groups. price collection only represent a fraction of this expendi-
ture. Likewise, if an expenditure category is divided into
two elementary aggregates according to outlet type (for
The Weighting Structure of the example, open markets and supermarkets, with correspond-
Consumer Price Index ing market shares of food sales, 60 percent and 40 percent,
3.19 The calculation of a CPI usually proceeds in two respectively), these proportions would be used to estimate
stages. In the first stage, elementary indices are estimated the stratum weights, whatever the importance of the specific
for each of the elementary aggregates. In the second stage, outlets eventually sampled.
a weighted average is taken of these elementary indices 3.23 The methods used to calculate the elementary
using the expenditure shares of the elementary aggregates price indices from the individual price observations col-
as weights. Elementary aggregates are usually the smallest lected within each elementary aggregate are explained in Chap-
groups of goods and services for which expenditure data are ter 8. Working upward from the elementary price indices, all
available. They may cover the whole country or separate indices above the elementary aggregate level are described
regions within the country. Likewise, elementary aggregates as higher-level indices that can be calculated from the ele-
may be developed for different types of outlets. The nature mentary price indices using the weights of the elementary
of the elementary aggregates depends on data needs and the expenditure aggregates. The aggregation structure is consis-
availability of expenditure data. Elementary aggregates may tent, so that the weight at each level above the elementary
therefore be defined differently in different countries. Other aggregate is always equal to the sum of its components. The
types of weights (for example, nonexpenditure weights rep- price index at each higher level of aggregation can be calcu-
resenting market shares), if available, may be used within lated using the weights and price indices for its components,
the elementary aggregate. that is, the lower-level or elementary price indices.
3.20 The weights are usually classified according to
the Classification of Individual Consumption According to Regional Weights
Purpose (COICOP), also used in the national accounts and 3.24 Weights stratified by region may be used in the
the HBS. Some countries develop a more detailed product CPI, depending on the size and structure of the country,
classification that refines the products defined at the subclass data availability, resources, and the purpose of the index.
55
Introducing regional weights creates more homogeneous level. Alternatively, expenditures are first estimated at the
entities which are likely to experience similar price move- country level before being distributed across different regions.
ments and have similar consumption patterns. It may be 3.30 Consistency should be ensured between the prod-
necessary to distinguish different regions because CPIs for uct and the regional dimensions. In the example in Table 3.2,
individual provinces or states may be required for adminis- expenditure is estimated for three products in two regions
trative or political purposes. In addition, in federal countries, (Table 3.2A). The expenditure share of each stratum in the
indirect taxes and hence price developments may differ national index corresponds to the stratum expenditure divided
between the provinces or states. by the national total (Table 3.2B). According to this example,
3.25 In some countries, there may be large differences region 1 has a total weight of 38.8 percent. At the same time,
between urban and rural areas with regard to consumption pat- product 1 has a weight of 24.5 percent at the national level.
terns and price developments. A common practice is to intro- 3.31 If detailed product expenditures are only avail-
duce a stratification level that distinguishes between urban and able at the national level, the regional product expenditures
rural areas. This approach assumes that data sources are avail- must be estimated. This can be done by using shares derived
able to derive expenditure weights separately for urban and from the total expenditures that are made within a region.
rural areas. Moreover, prices must be collected in both rural Table 3.2C assumes that product expenditures at the coun-
and urban areas to compile the respective stratum indices. try level are available, and only total expenditure by region
3.26 If the weights derived from the HBS are available is known. In Table 3.2D the product expenditures for the
for rural and urban households and if price collection is lim- country are then disaggregated by region, using a weight
ited to urban areas, one approach is to combine the weights
for urban and rural households. This approach enhances the
representativity of the index because the weights represent Table 3.2 Deriving Expenditure Weights by Region
all households, both urban and rural. In this case, the prices Table 3.2A Expenditure by Product and Region
faced by rural households will implicitly be imputed by
Region 1 Region 2 Country
those collected in urban areas. This can be an acceptable
assumption if most of the monetary transactions made by Product 1 400 800 1,200
households living in the rural areas take place in urban areas Product 2 500 1,000 1,500
Product 3 1,000 1,200 2,200
and if the relative changes in prices in rural and urban areas Total 1,900 3,000 4,900
follow the same general trend. Alternatively, if prices are
only collected in urban areas, it could also be decided to Table 3.2B Regional Product Shares
restrict expenditure weights to urban households and com- Region 1 (percent) Region 2 (percent) Country
pile an accurate urban CPI. (percent)
3.27 Within a given COICOP subclass, the regional Product 1 400/4,900 = 8.2 16.3 24.5
weight represents the consumption expenditure in the region Product 2 10.2 20.4 30.6
in proportion to the expenditure in the whole country for Product 3 20.4 24.5 44.9
Total 38.8 61.2 100.0
that subclass. For example, if 60 percent of the total expen-
diture on fresh fruits occurs in the North region and 40 per- Table 3.2C Expenditure by Region (Regional Product
cent in the South region, then these proportions can be used Expenditures Not Available)
to derive the regional elementary aggregate weights. If at Region 1 Region 2 Country
the country level, the expenditure share for fresh fruits is 5 Product 1 N.A. N.A. 1,200
percent, this share can then be split between the regions so Product 2 N.A. N.A. 1,500
that 5 percent × 60 percent = 3 percent of the total national Product 3 N.A. N.A. 2,200
expenditure relates to fresh fruits in the North and 5 percent Total 1,900 3,000 4,900
× 40 percent = 2 percent to fresh fruits in the South. Table 3.2D Product Expenditures for the Country
3.28 Regional weights may typically be obtained from Disaggregated by Region
the HBS if the sample design of the HBS is representative at Region 1 Region 2 Country
the regional level and supports the development of reliable
Product 1 1,200 × 1,900/4,900 1,200 × 3,000/4,900 1,200
regional weights. When reliable estimates are not available = 465 = 735
from an HBS, population statistics are sometimes used to Product 2 1,500 × 1,900/4,900 1,500 × 3,000/4,900 1,500
split household expenditure across regions; however, this = 582 = 918
approach is not preferred as it assumes that expenditures Product 3 2,200 × 1,900/4,900 2,200 × 3,000/4,900 2,200
per capita or per household are the same in all regions. For = 853 = 1,347
Total 1,900 3,000 4,900
instance, there are usually large differences between urban
and rural populations in the level and pattern of items con- Table 3.2E Regional Product Shares (estimated)
sumed. Finally, national accounts data, if compiled by the Region 1 (percent) Region 2 (percent) Country
expenditure approach and available at the regional level, can (percent)
also be used to estimate regional weights. Product 1 465/4,900 = 9.5 15.0 24.5
3.29 In practice, there are different strategies to derive Product 2 11.9 18.7 30.6
regional weights depending on the availability and the qual- Product 3 17.4 27.5 44.9
ity of data sources. To develop a national weighting structure Total 38.8 61.2 100.0
based on regional weights, first estimate household expendi- N.A. = Not available
ture by region. Each of the regional expenditure values can
All totals are shown in bold.
then be summed up to obtain the expenditures at the national
56
of 1,900/4,900 = 38.8 percent for region 1 and a weight of accounts data can serve as a data source. While national
3,000/4,900 = 61.2 percent for region 2. The regional product accounts data rely to some extent on expenditure data from
shares in Table 3.2E obtained with this method differ slightly the HBS, other data sources can be used to supplement the
from those in Table 3.2B. Ideally, the sample size by region HBS data. Apart from the HBS, many other data sources
will support expenditure estimates by product within each can and should be used to improve the accuracy of the CPI
region, as shown in Tables 3.2A and 3.2B. weights. Such complementary data sources can be used in
a number of ways, including (1) validating and correcting
Outlet-Type Weights HBS estimates; (2) disaggregating the higher-level expen-
ditures to develop a more detailed weighting structure; and
3.32 Prices are collected from a variety of outlets and (3) estimating expenditures for certain product categories.
outlet types. In addition to the regional dimension, the sub-
class could also be stratified by outlet or by outlet type. This
can be especially useful if price levels and price changes dif- Household Budget Surveys
fer significantly across outlets. Information about the sale or 3.35 In most countries, the HBS serves as the primary
market share of the outlets may be used to form elementary data source for deriving expenditure shares for the goods
aggregate weights specific to a given outlet type. In some and services covered by the CPI. As the HBS may have been
countries, the HBS directly collects expenditure data by type designed to serve more than one purpose, the survey design
of outlet which can then be used to estimate these stratum should meet the requirements for the CPI. Ensuring that CPI
weights. The use of HBS data ensures consistency between specific needs are met requires close coordination between
the product weights and outlet-type weights. Other potential the CPI and HBS staff. The classifications used for HBS and
sources to derive outlet or outlet-type weights include point- CPI should be consistent and preferably based on COICOP.
of-purchase surveys, scanner data, business registers, or 3.36 Data for a normal year should be used as the basis
retail trade statistics. It may only be possible to develop out- for developing CPI weights. Given that many HBSs are con-
let or outlet-type weights for a broader category of products ducted infrequently, this can be challenging to manage in
so that the same proportions have to be used to disaggre- practice. If possible, an average of multiple years could be
gate the expenditures on the different items within product used, but this can only be done if the HBS is conducted con-
groups. One should bear in mind that these additional data tinuously. In the absence of a continuous HBS, alternative
sources may not be completely aligned with the scope and data sources would be needed to develop weights. This issue
the coverage of the CPI and may have their own limitations. of relying on a normal year to develop weights supports the
It may, therefore, be necessary to make further adjustments need for more frequent weight updates.
to the data extracted from such sources. 3.37 The main requirements for the CPI include ensuring
3.33 In the example in Table 3.3, outlet types 1 and 2 have that the survey broadly represents the CPI reference popula-
market shares of 60 percent and 40 percent, respectively, and tion, and that all types of final consumption expenditure made
this same breakdown holds for both regions and for all three by households are covered. The HBS sample size (number of
products. If a division is made according to the type of outlet households) should ensure that the expenditure data yielded
and region, then each item within a given region comprises are suitable to develop statistically reliable weights for the
two elementary aggregates: one for outlet type 1 and one for CPI at the elementary aggregate level. The HBS item list
outlet type 2. In this example, the weight for the elementary should be designed so that the information obtained maps
aggregate for product 1 in region 1 sold in outlet type 1 would directly into the CPI classification system. Ideally, the survey
then be 8.2 percent × 60 percent = 4.9 percent. This same cal- should also collect the data needed for deriving net weights
culation would be used to develop weights for outlet type 2. for second-hand goods (see paragraphs 3.89–3.92). The inter-
view and recording periods should be appropriately distrib-
Data Sources uted over time to ensure that annual estimates can be obtained
taking into account possible seasonal patterns.
3.34 The HBS serves as the primary data source for 3.38 The expenditure values used to derive weights
CPI weights for most countries. These are household-based should be consistent with the conceptual approach (acquisi-
surveys that collect data on households’ expenditure and tion, use, or payment) adopted for the CPI (for additional
consumption of goods and services. Alternatively, national information on this topic, see Chapters 2 and 11). For some
services, the moment when the service is consumed and
acquired can differ from the moment when the service is
Table 3.3 Deriving Expenditure Weights by Region paid for. The approach used in the HBS for determining
and by Outlet Type the timing of consumption should be the same as the one
Region 1 (percent) Region 2 (percent) Country used in the CPI. For most goods, the moments of acquisi-
Outlet Outlet Outlet Outlet
(percent) tion, payment, and consumption coincide. Sometimes, the
Type 1 Type 2 Type 1 Type 2 HBS focuses on the consumption of food products, which
includes food purchased by the household, but also own pro-
Product 1 8.2 16.3 24.5
4.9 3.3 9.8 6.5
duction of food or food received for free. For durable goods,
Product 2 10.2 20.4 30.6 there is no difference between the acquisition approach and
6.1 4.1 12.2 8.2 the payment approach if the purchase has not been financed
Product 3 20.4 24.5 44.9 with a loan. Once acquired, the use of such goods can span
12.2 8.2 14.7 9.8 over several years. The HBS may collect different types of
Total 38.8 61.2 100.0
information regarding durable goods.
57
3.39 Theoretically, the HBS provides information on the because of events such as droughts or floods, civil strife, oil
regional breakdown of expenditure used to develop reliable price shocks, or exceptionally mild or cold winters.
regional weights. For this purpose, the regional dimension
must be appropriately incorporated in the sampling design National Accounts Data
of the HBS. Depending on its design, the HBS may also pro-
vide information on the types of outlets and the brands pur- 3.44 National accounts can be an alternative source for
chased. In particular, it can be possible to identify internet deriving CPI expenditure weights, if reliable estimates for
purchases. Disaggregation of expenditure by type of outlet HFCE are available. The practical advantage is that HFCE is
can be useful not only to construct elementary aggregates at updated every year, whereas an HBS may only be conducted
a more detailed level but also to improve the sample design on a less frequent basis. However, national accounts may
for outlets and items for price collection. only be available at the national level and the use of other
3.40 Like the HBS, national food surveys are special sur- available data sources would be needed to develop more
veys with the primary emphasis on collecting information on detailed or regional expenditure weights.
family expenditure for food products. These surveys provide 3.45 The index compiler must understand the differ-
a very detailed breakdown of food expenditure that can be ences in scope and definition of consumption before using
used to derive the weights for detailed elementary aggregates. national accounts data for CPI weights. If the CPI is restricted
3.41 The detailed expenditure obtained from the HBS to monetary transactions, then only a subset of HFCE must
can be subject to measurement errors. There is often under be used, excluding its nonmonetary components. Addition-
or overreporting on luxury goods and services, as well as on ally, the CPI may adopt a different treatment for expendi-
certain types of products such as alcoholic beverages and ture on types of goods and services, such as owner-occupied
tobacco. Moreover, the respondent household may not cor- housing, or even a different scope, especially concerning
rectly remember the details of all the expenditure that took institutional households.
place during a given recall period. Therefore, HBS estimates 3.46 National accounts data may be used to improve
should be reviewed and evaluated for completeness of cov- HBS weights for products that are underreported in the
erage by comparing with secondary source data. HBS. Note that national accounts’ estimates for HFCE are
3.42 Given that the HBS in many countries is based on usually based on data from the HBS as well as a wide range
probability sampling methods, statistical quality indicators of other sources such as domestic production, retail sales,
(for example, standard error and coefficient of variation) for tax information, and import and export data.
the weights can be calculated. The analysis of these indi- 3.47 In practice, weights for the main product groups
cators can provide the index compiler with information on can be obtained from the national accounts down to a cer-
the structure of the elementary aggregates. If an analysis of tain level of disaggregation. Each of these weights can then
the HBS data shows, for example, that the expenditure data be disaggregated by applying the detailed HBS expenditure
for oranges are of poor statistical quality, then alternative groups to the national accounts consumption groups. The
data sources could be used to improve the reliability of the combination of national accounts and HBS data ensures
weights or a more broadly defined elementary aggregate consistency between the CPI and the national accounts data
such as citrus fruit could be considered instead. In a similar on HFCE at the level of the main consumption groups. The
way, suppose that the HBS provides an unreliable expendi- use of national accounts data also facilitates more frequent
ture estimate for an item at a detailed local level. It may then weight updates. For instance, CPI weights can be updated at
be preferable to derive an estimate based on the expenditure regular intervals using national accounts data for the higher-
share of that item within a broader geographical area. To find level aggregates. The updated expenditure is then distributed
the best compromise between lowering the variance of the using the shares obtained from the HBS or other sources that
item expenditure estimate and biasing it toward the spend- may only be updated less frequently.
ing pattern of the broader geographical area, a composite 3.48 CPI compilers should consult with their national
share can be computed that averages the initial direct share accounts counterparts regularly before using national
with the share obtained for the broader geographical area. accounts data for weights to ensure that they are consis-
3.43 The frequency of updating the CPI weights depends tent with the objectives and uses of the CPI. In some cases,
on how often an HBS is conducted. For the purposes of the national accountants need to apply an element of discretion
CPI, it is desirable for the HBS to be conducted at least and judgment when making operational decisions related
every five years. This will allow countries to revise and to the construction of some national accounts aggregates.
update their expenditure weights every five years (or more Moreover, the preliminary national accounts estimates are in
frequently). Some countries conduct continuous HBSs with general revised several times before the final estimates are
gradually rotating samples. However, a program of annual available. The most recent available data may not be suffi-
surveys with samples large enough to provide the type of ciently stable for CPI purposes. There is a trade-off between
estimates required for CPI weights can be very costly. For timeliness and quality.
this reason, some countries conduct large-scale surveys at
five-year intervals, perhaps supplemented with a smaller Other Data Sources
annual sample. Other countries distribute a large sample
over several years. The average of the results over several Administrative Data Sources
successive years of smaller-scale surveys may provide a set 3.49 For some product categories, one option is to esti-
of satisfactory annual estimates. The weights derived as the mate expenditures from administrative data sources instead
average of two or three years will also smooth any erratic of relying on survey data. For instance, expenditure on med-
consumer behavior over a shorter period, for example, ical products can often be obtained from relevant bodies in
58
charge of the national health systems. In some countries, and outlet stratification can be introduced by disaggregat-
tax data for certain products such as alcohol or tobacco lead ing the expenditure for a broader product category that was
to more accurate estimates of expenditure than HBS data obtained from the primary data source. The use of scanner
which suffers from underreporting. It can be also difficult data to construct weights is further discussed in Chapter 10.
for households to properly report explicit charges paid for
financial services in an HBS. Moreover, the derivation of Market Intelligence and Trade Associations
weights for insurance services requires special consider- 3.54 If the product categories are sufficiently important,
ation (see paragraphs 3.87 and 3.88). Therefore, it may be additional data sources might be consulted. A survey may be
preferable to use regulatory data sources to derive expendi- conducted with a small selection of outlets to obtain general
ture estimates for financial services and insurance. Admin- information on the breakdown of sales for a specific product
istrative data sources are not always perfectly comparable category. Existing market information can also be an option.
with CPI coverage and may suffer from their own errors. Associations of importers or distributors, other industry
For instance, tax revenue on tobacco also covers sales made groups, or marketing agencies and boards are likely to have
to nonresident households whereas the CPI may only be some general information on the breakdown of sales for spe-
limited to resident households. Most of the administrative cific products.
data sources are also used in the national accounts which,
as noted previously, can serve as a data source for weights. Tourism Expenditure Surveys
3.55 If the CPI follows the domestic concept, the
Retail Trade Statistics expenditure of nonresident households within the national
3.50 Statistics on retail sales by region and type of out- territory must be included in the weights. In countries where
let may be available for broad groups of products. One dis- tourism is important, tourism expenditure surveys can be
advantage of using these data is that some of the sales may conducted to estimate nonresident household expenditures
be to groups outside the reference population, including cor- that can be added to the expenditure made in the country by
porations or the government. The corresponding purchases resident households obtained through the HBS. Nonresident
do not form part of household final consumption. Some visitors will generally have very different expenditure pat-
sales may also be to nonresidents, who may be part of the terns from those of residents (for example, they will spend
reference population. Furthermore, for regional sales data, it more on hotels and restaurants).
needs to be kept in mind that sales may include purchases
made by households living in other regions. Point-of-Purchase Surveys
3.56 Point-of-purchase surveys provide data that can
Population Censuses be used to estimate weights for different types of out-
3.51 Population censuses provide data on the geograph- lets and can be used to develop a sample frame of outlets
ical distribution of the population and households, as well as where households make purchases. For products purchased,
on the regional differences in household size and composi- households are asked about the amounts spent in each out-
tion. Combined with estimates of regional levels of house- let where purchases have been made. Given that household
hold expenditure, these data can be used to estimate regional surveys are costly and that there is overlap between the HBS
expenditure weights, especially when such estimates are not and point-of-purchase surveys, it is possible to combine the
available from an HBS with a satisfactory degree of preci- two into an integrated survey that collects expenditure and
sion. In the absence of any expenditure statistics, population outlet data at detailed levels.
statistics might be used as the basis for regional weights. 3.57 A simpler version of this survey may be conducted
However, such an approach should be avoided because it to obtain weights for groups of products by outlet type. As
assumes that expenditure per capita or per household is the an alternative, in the absence of this type of survey, national
same in all regions and ignores the fact that there are usually retail sales statistics by outlet type from a survey of outlets
large differences between the urban and rural populations in could be used to estimate a breakdown of sales by outlet type.
the level and patterns of consumption.
Scanner Data Deriving the Weights in Practice

3.52 Scanner data can also be used to derive and update 3.58 Once the reference population and the coverage of
weights in a more frequent and timely manner. These data are goods and services have been decided, the weights need to
based on electronic data records that are stored in the data- be derived. The weights are calculated as the proportions
bases of sellers. Such scanner data sets include the quantities of the total consumption expenditure of all goods and ser-
sold and the corresponding value aggregates. The limitations vices included in the index basket for the reference popula-
of this information should, however, be borne in mind. The tion during the reference period. The reliability of the CPI
first one is that scanner data cannot be connected to a spe- weights will obviously depend, to a large extent, on the reli-
cific type of household, whereas the data from the HBS can. ability of the household expenditure data. In practice, the
Moreover, scanner data may only have limited outlet cover- derivation of weights involves a series of steps.
age. Finally, scanner data may not be fully consistent with
the scope of the CPI, as no distinction can be made between
sales to businesses, government, or households. Arrange the Data According to the
3.53 Scanner data can also be a good source for deriv- Classification and Coverage of the CPI
ing detailed weighting structures especially at the lower lev- 3.59 The detailed expenditure items identified in the
els of the index hierarchy. For instance, a detailed product HBS, or other data sources, must be mapped to the CPI
59
expenditure classes. If HBS classes do not match CPI sources to be reconciled and balanced against each other.
expenditure classes, the HBS results must be transformed The commodity-flow method may be used to improve esti-
to match the CPI classes. This can be done by aggregating mates of HFCE derived from the HBS by adjusting them
or disaggregating the relevant HBS headings over the rel- to account for additional information provided by statistics
evant CPI expenditure subclasses. Such transformation is on sales, production, and imports and exports of consumer
achieved much more easily and more reliably if the coding goods and services.
list for expenditure items in the HBS is coordinated with 3.65 The Supply and Use Tables form an integrated
the corresponding list of items used for collecting price framework where supplies of different kinds of goods and
observations for the CPI. As noted previously, both the HBS services originating from domestic industries and imports
and CPI should use the same classification system (ideally are allocated between various intermediate or final uses,
COICOP). including HFCE. The product balance for any product rec-
3.60 The HBS expenditure data may include payments ognizes that the sum of output at basic prices plus imports
that are outside the scope of the CPI. For example, pay- plus trade and transport margins plus taxes on products less
ments of income taxes or social security contributions, life subsidies on products is equal to the sum of intermediate
insurance premiums, remittances, gifts and other transfers, consumption, final consumption, and capital formation,
investments, savings, and debt repayments should not be all expressed at purchasers’ prices, plus exports (System
considered because they are not consumption expenditure. of National Accounts 2008 [2008 SNA], paragraph 14.5).
These should be excluded from the total used to calculate The HFCE estimates obtained can be compared with the
the expenditure shares that serve as the basis to estimate the corresponding estimates from the HBS to provide conver-
CPI weights. sion factors to adjust HBS expenditure data for under or
overreporting.
3.66 There are some practical limitations to applying
Correcting for Over- and Underreporting the commodity-flow method. Often the balance can only be
by Combining Different Data Sources established for a category of products broader than the prod-
3.61 The results from the HBS need to be carefully uct categories used in the CPI classification. Moreover, data
examined and adjusted to take account of under or over- are usually compiled at the national level and no detailed
reporting of consumption expenditure on different types regional breakdown is available. Finally, the HBS may not
of products. This is a relevant problem affecting HBS data. be the only inaccurate data source for a particular product
Evidence suggests that the understatement of expenditures and the estimates of other components of the product bal-
can be significant for certain goods and services in the HBS ance can also be unreliable.
if their consumption is socially discouraged, such as tobacco
and alcoholic beverages. Expenditures on durable goods also Adjusting Household Budget Surveys
tend to be underreported. Some products, such as vehicles 3.67 Even if expenditure data obtained from the HBS
or other major durable goods, are purchased infrequently. results are considered accurate, adjustments might still be
When purchased, the amount spent on such products can be needed to account for any significant changes in expen-
considerable. As the HBS is a sample survey, estimates are diture patterns between the period when the survey was
subject to sampling errors, which may be relatively large conducted and the period when the new weights are intro-
for such infrequent expenditures. For major durable goods, duced. Adjustments will typically be made for products
HBS estimates should be compared with other sources such belonging to fast-evolving markets and which are signifi-
as import statistics or administrative data. cantly losing or gaining importance during this period. It
3.62 Other expenditures are not reported because the is possible that expenditure on some products may not be
purchases were small, and easily forgotten by the respon- available from the HBS because the products appeared
dent. Therefore, to the extent possible, results from the HBS on the market after the survey had been completed. Addi-
should be compared and combined with other data sources tional data sources must then be accessed to estimate
when constructing CPI weights, especially when the HBS expenditure for new products. Expenditure should also be
sample is small. reviewed if there are known changes that have occurred
3.63 The usual strategy to correct for over or underre- following administrative decisions, such as changes in
porting is to use supplementary information from other rele- taxation, that entered into force only after the HBS was
vant data sources such as tax data, administrative data, other conducted.
independent surveys, or HFCE from the national accounts,
and to apply correction factors. For instance, assume that tax
revenue data indicate that annual sales for cigarettes may be Treatment of Expenditure for Unimportant
twice as high as the total annual expenditure estimated from or Difficult-to-Measure Products
the HBS. The CPI weight is thus obtained by increasing the 3.68 The HBS, which in most cases is the main source
HBS expenditure on cigarettes by two to better reflect the for deriving the detailed weights, usually includes obser-
sales data. vations on a much larger variety of goods and services
3.64 In countries with reliable and detailed national than it is practical to collect prices for in the CPI. Some
accounts data, the commodity-flow method can be used to products may have a weight which for all practical pur-
adjust data considered less reliable from the HBS. The use poses is negligible. The prices of products with very small
of the commodity-flow method within the Supply and Use weights may not be worth collecting if their contribution
Table framework, as described in the System of National to the CPI is very small. In practice, a cutoff threshold can
Accounts, enables data drawn from different primary be defined to select the products to be included in the CPI
60
basket (see Chapter 4 for a description of different sampling Weight Reference Period
techniques).
3.69 Among the consumption expenditure, there are 3.72 The weight reference period refers to the time
also likely to be a few products for which the prices, or price period to which the expenditure data used to estimate
changes, cannot be directly or satisfactorily measured, such weights were collected. For a fixed-basket CPI which keeps
as gambling. It may be difficult if not impossible to com- weights constant over several periods, the weight reference
pile a reliable price index for such products. A decision must period is typically a 12-month period, such as a calendar
then be made on how to treat the expenditure allocated to year. A month or quarter is too short to be used as a weight
such difficult-to-measure products. reference period, since any month or quarter is likely to be
3.70 Even if the product weight is small or if there are affected by accidental or seasonal influences. This is espe-
measurement problems, the product is still included in the cially important in countries where the composition and size
scope of the CPI. For those products with relatively low of expenditure can fluctuate significantly throughout the
weights, it is not efficient to collect prices; however, the CPI year. In some countries, data for a single year may not be
should cover these expenditures. Some price change should adequate because of unusual economic conditions or due to
be explicitly or implicitly assumed or imputed. In practice, insufficient sample size. In these cases, an average of more
there are different options: than one year of expenditure data may be used to calculate
the weights.
• The weight of the product is combined with one or several 3.73 As the CPI is sensitive to the selection of the
related products. For instance, if the weight for “cheese” weight reference period, it might be preferable to choose a
is small, it can be combined with the weight for “milk.” “normal” consumption period as the basis for weights and to
This means that the elementary aggregate is now weighted avoid periods in which there are special factors of a tempo-
using expenditures on milk and cheese although only rary nature at work. To achieve this, it may be necessary to
prices for milk will be collected. This option assumes that adjust some of the values to normalize them and to overcome
the price index for cheese would change in the same man- any irregularities in the data. One option might be to smooth
ner as the observed price index for milk. In practice, the particularly erratic observations, for example, by taking an
weights for the products could be kept separated but price average of HBS data over more than one year. All avail-
changes must be explicitly imputed. To continue using the able information concerning the nature of consumption in a
previous example, the price index attached to the weight weight reference period should be taken into consideration.
for cheese would simply be identical to the price index for 3.74 During periods of high inflation, multiple year
milk. This is the preferred approach. weights may be calculated by averaging value shares rather
• An alternative approach is to set the weight for which than averaging actual value levels. Averaging value levels
no representative prices exist equal to zero. This option will give too much weight to the data for the most recent
removes the item from the scope of the CPI and is equiva- year. Another option is to update the values for each year to
lent to the assumption that the price of the excluded prod- a common period and then to compute a simple arithmetic
uct would have moved in the same way as the all-items average of adjusted yearly data.
CPI for all the products actually included in the index. 3.75 As the weight reference period usually precedes
This approach is, in general, not recommended. the price reference period, the expenditure weights may be
3.71 Table 3.4 shows examples of different options for price updated to take account of the relative price changes
the treatment of unimportant expenditures. In the example, from the weight reference period to the price reference
the expenditure for product 3 is not relevant. One option period. The technique of price-updating expenditure data
is to add this expenditure to item 1. Alternatively, it may or CPI weights from a past period does not make weights
be added to products 1 and 2 proportionally to the weights reflect current expenditure patterns. New CPI weights can
of these two products. Finally, it can be removed, which only be derived if new expenditure data are available. Price
implies that the total expenditure for the three products is updating of weights is discussed in more detail in Chapter 9.
now reduced. In general, the last option should be avoided,
and expenditures should be allocated to products with Frequency of Weight Updates
similar price behavior. Because of the negligible size of
the weight value involved, the consequence on the overall 3.76 The expenditure weights should be updated at regu-
index will in general be negligible whichever method is lar intervals, as often as possible, but at least every five years.
used. The decision of when to update the weights often depends on
the availability of appropriate data sources. Rather than wait-
ing for the weights to become outdated before deciding to
update them, the recommended approach is to plan for regu-
Table 3.4 Treatment of Products for Which No Prices lar scheduled updates subject to the available data sources.
Are Collected In general, there is a lag between the weight reference period
Initial Allocate Allocate to Remove and the moment that the weights are introduced in the CPI.
Expenditures to Item 1 Items 1 and 2 Expenditure The overall principle is to minimize the implementation lag
Product 1 7 8 7.7 7 so that the weights used in the CPI are as up to date as pos-
Product 2 3 3 3.3 3 sible. New weights should be introduced in a timely man-
Product 3 1 0 0 0 ner, as soon as the required source data are available. For
(minor item) instance, expenditure weights referring to year y ‒ 2 could be
Total 11 11 11 10
introduced with the January index of year y.
61
3.77 If weights are kept fixed over longer time periods, the 3.81 An intermediate solution is to update weights only
index will be unresponsive to substitution effects or changes at the higher level, for example using national accounts data,
in consumer preferences. In the short term, consumers may which can then be disaggregated down to the lower levels
change consumption patterns in response to shifts in relative using shares from the HBS, which has not been updated.
prices, mostly between products included in the same class or Partial weight updates are further discussed and illustrated
subclass. In the medium or long term, consumption patterns in Chapter 9.
are also influenced by factors other than price changes. Most 3.82 Frequent updating of weights and chaining can lead
importantly, under rising incomes, changes in the level and to chain drift. For a Laspeyres or a Lowe index, the drift can
distribution of household income will cause a shift in demand be upward if there are systematic fluctuations in consumption
for goods and services toward goods and services with higher and prices. Suppose that the prices for petroleum products
income elasticities. Demographic factors such as aging of the went down between the old and the new weight reference
population and technological changes are examples of other period, causing an increase in the quantities consumed and
factors that affect spending behavior in the longer term. Fur- expenditure for these products. If petroleum prices sub-
thermore, new products will be introduced, and existing ones sequently rise after the new weights are introduced, the
may be modified or become obsolete. As a result of both rela- aggregate CPI will rise more rapidly than before since these
tive price changes and long-term effects, the weights may products now have more importance. At the level of broad
become out of date and less representative of current con- product categories, expenditure data analyzed over longer
sumption patterns. As shown in Chapter 2 of the publication periods often follow trends. Chaining can have a downward
Consumer Price Index Theory, the bias in a fixed-basket index impact if, for instance, there is a gradual shift from one prod-
is likely to increase with the age of the weights. At some point, uct to another product for which prices are rising slower. A
it therefore becomes desirable to use the weights of a more direct CPI compiled with unchanged weights is often found
recent period to ensure that the index is weighting appropri- to be higher than the chained index compiled, for instance,
ately the price changes currently faced by consumers. with annually updated weights. Chain drift is more problem-
3.78 If data sources such as the HBS are available only atic with high-frequency updating of weights and chaining
at irregular or infrequent intervals, the frequency of weight at the level of an individual variety. Multilateral methods
revision may necessarily be linked to the availability of that are discussed in Chapter 10 (and Chapter 7 of Consumer
results from the HBS. When the weights are to be fixed for Price Index Theory) are one solution to this problem.
several years, the objective should be to adopt weights that 3.83 The selection of the level in the index hierarchy at
are not likely to change much in the future, rather than pre- which the structure and weights are fixed for a period is par-
cisely reflect the activity of a particular period that may be ticularly important. The main advantage of setting the level
abnormal in some way. Even if weights are updated only relatively high is that the actual samples of products and
every five years, it is desirable to review the weights in their prices below this level can be adjusted and updated as
between to ensure that they remain sufficiently reliable and needed (see Chapter 7). New products can be introduced into
representative. The review, which may be limited to weights the samples, and the weights at the lower level updated using
at the elementary index level and their major components, more recent information. There is thus a greater opportunity
should examine whether there are indications that important to keep the index representative through an ongoing review
changes may have taken place in the consumption pattern of the sample of representative products. If the level is set
since the weight reference period. relatively low in the index structure, there is less freedom to
3.79 To reduce upper-level substitution bias, it would maintain the representativeness of the index on an ongoing
be preferable to update the weights more often such as every basis, and there will be a greater dependence on the periodic
three years, every two years, or even every year; however, this index review and reweighting process. In such circumstances,
depends on the availability of data to develop new weights. the arguments for frequent reweighting become stronger.
Any bias which may follow from using a Lowe index (see 3.84 Whenever the weighting pattern has been updated,
Chapters 1 and 8 for information on price index formulas) the new index using updated weights should be calculated for
with a fixed basket of goods and services will not have time to an overlapping period with the old one so that the two can be
accumulate to a significant magnitude. Also, by updating the linked. These techniques are discussed in Chapter 9. During
weights more frequently, there is the opportunity to introduce the year that follows the weight update, the year over year
newly significant goods and services in a timely manner. At rates of the higher-level aggregates compiled from the linked
the same moment, the sample of outlets and varieties to be series not only reflect changes in prices, but are also impacted
priced can be updated simultaneously, although the sample by the use of different weights and a different item structure.
should be maintained in between two weight updates. Coun-
tries that are experiencing significant economic changes, and
thus more rapid changes in the consumption pattern, should Items Requiring Special Treatment
consider updating their weights more frequently.
3.80 The benefits of updating weights more often should Seasonal Products
be assessed and compared to the additional costs of such an
3.85 In practice, there are two types of approaches for
exercise. If feasible, empirical studies can be conducted on
the treatment of seasonal products:
historical data to assess the impact of updating weights more
frequently, of choosing different weight reference periods, • A fixed-weight approach, which assigns the same weight
and of minimizing the time lag between the weight refer- for the seasonal product in all months, using an imputed
ence period and the price reference period. The resulting price in the out-of-season months. Seasonal products are
indices can then be compared to specific target indices. treated in the same way as other consumption products.
62
• A seasonal-weight approach, in which a zero weight is 3.89 If the weights are based on the gross premiums,
attached to a product which is out of season, and a positive expenditure weights should normally exclude goods and
weight is used for the in-season periods when the product is services provided for or reimbursed by the insurance com-
available for pricing. The in-season weights are kept fixed pany on the basis of claims. If the weights are based on the
as much as possible and only vary to the extent necessary to implicit service charge, expenditure weights include goods
reflect changes in the composition of the basket. Moreover, and services that households buy and that are reimbursed by
the principle of a fixed basket (that is, fixed weights) should insurance companies, and also goods and services that are
be maintained at least at some level of aggregation. One dis- paid for and provided by insurance companies on the basis
advantage of such an approach is that the monthly changes of claims. In general, it seems preferable to base the weights
in the index may be more difficult to interpret as they may for nonlife insurance on the service charges.
reflect not only price changes but also quantity changes.
The treatment of seasonal products is further discussed in Second-Hand Goods
Chapter 11. In addition, this topic is examined in Chapter 9 3.90 The prices of used, or second-hand, goods pur-
of the publication Consumer Price Index Theory. chased by households are included in the CPI in the same
way as the prices of new goods (Chapter 11 discusses this
Internet Purchases topic in more detail). However, households also sell used
goods, such as cars. If the price of a second-hand good rises,
3.86 As internet purchases become increasingly impor- a purchasing household is worse off, but a selling household
tant, these expenditures should be included in the CPI weights is better off. From a weighting perspective, sales constitute
according to the conceptual approach that has been adopted. negative expenditures, which implies that price changes for
Chapter 11 discusses how internet purchases should be used goods sold by households implicitly carry a negative
treated either under the national or the domestic concept. Ide- weight in the CPI. In effect, purchases and sales of second-
ally, data on the share of internet purchases for all relevant hand goods between households, whether directly or indi-
product categories could be available, so that detailed strata rectly through a dealer, cancel out (except for the dealers’
can be defined for this outlet type. The HBS should, in prin- margins, as explained in Chapter 2) and carry no weight
ciple, cover internet purchases made by resident households. in the CPI. However, households also buy from, and sell
The survey should be designed so that it is possible to sepa- to, other sectors. For the reference population, namely the
rate internet purchases from other purchases. The estimation entire set of households covered by the CPI, the weight to
of expenditure on internet purchases is of interest to both the be attached to a particular kind of second-hand good is given
HBS and the CPI programs and therefore it would be ben- by the households’ total expenditure on it less the value of
eficial to cooperate on this topic. Similarly, the treatment of the households’ revenue from sales to/from outside the
internet purchases in the national accounts should be carefully household sector, including the rest of the world. There is no
examined if this is the primary data source for CPI weights. reason why these should cancel out on aggregate. For exam-
3.87 Alternative data sources for measuring e-com- ple, many of the second-hand cars purchased by households
merce transactions include bank account data and credit may be imported from abroad. The difference between total
card data. However, excluding business expenditure from expenditures and total sales is usually described as house-
these data sets could be problematic and identifying specific holds’ net expenditure. This is the weight to be attached to
products and outlets may not be straightforward. Finally, the second-hand good in question. Table 3.5 illustrates how
reports prepared by external organizations that monitor to develop weights for used goods. In this example, the net
and describe e-commerce markets can also help to estimate weight would be obtained as (100 + 300) − (100 + 200) =
weights for internet purchases. The scope, coverage, and 400 − 300 = 100.
data collection methods underlying the statistics presented 3.91 Second-hand markets may exist for a whole range
in such reports must be well understood when combining of durable and semidurable goods. Except in the case of
such figures with those obtained from other sources. used cars, it is often very difficult to estimate the net expen-
diture because most HBSs do not collect the data that would
Insurance allow for a comparison between expenditure and revenue
3.88 As explained in Chapters 2 and 11, the weights from sales of individual kinds of second-hand goods. Usu-
for nonlife insurance could be based on either (1) the gross ally, only the total amount received from the sale of second-
premiums paid, consisting of the payment for the insur- hand goods is collected. However, this information gives an
ance, or (2) on the implicit service charges payable to idea of the volume and significance of these transactions in
the insurance enterprise for arranging the insurance. The the economy.
implicit service charges for administering the insurance 3.92 In countries where the volume of second-hand
and providing the insurance services are estimated by the purchases is small, second-hand goods (except used cars)
gross premiums plus the income from investment of the
insurance reserves less the amounts payable to policyhold-
ers in settlement of claims. This definition can potentially Table 3.5 Estimation of Net Expenditure Weights
lead to negative weights for instance if there are irregular
Seller Buyer Households Other Institutional Sectors
and unexpected large fluctuations in claims due to natu-
ral disasters or large-scale accidents. To avoid negative Households 100 300
weights, an average service charge covering several years Other Institutional 200 Out of Scope
Sectors
could be considered.
63
may be ignored when calculating the weights of the index. expenditure, it should be included in the index structure.
In countries where second-hand purchases are important, One option is to wait and introduce the new product at the
and their prices are believed to change at different rates time of a complete weight update. Alternatively, the current
from those of new goods, separate weights are needed. The weighting structure could be adjusted, using the techniques
information may be obtained, at least for some major dura- described in Chapter 7, to introduce the new products in the
bles, from the HBS, if the survey asks about expenditure on CPI in a timely manner. In any case, expenditure weights
second-hand and new goods. Because the amounts spent on must be estimated for the new products. Such data may not
purchasing second-hand cars are usually large, they should always be readily available from the HBS. If that is the case,
be included in the CPI basket if the data are available. the newly identified products should be incorporated in the
3.93 Even if countries include expenditure on second- HBS as soon as possible. In the meantime, alternative data
hand goods in the estimation of CPI weights, second-hand sources may be used to estimate the weights of the new
goods may not be covered in the price collection. By exclud- products so that their inclusion in the CPI is not delayed.
ing second-hand goods from the price collection, it is
implicitly assumed that the prices of new and second-hand
goods move in the same way. If they are included, the price- Key Recommendations
determining characteristics of the second-hand good must
be kept constant over time so the same good is priced each • CPI weights represent the expenditure shares for specific
month. If the goods are different, appropriate quality adjust- goods and services that define the basket.
ments must be made. • CPI weights should be consistent with the decisions made
regarding the geographic coverage, the household refer-
Owner-Occupied Housing ence population, the type of consumption expenditure
3.94 If the CPI includes owner-occupied housing under included in the CPI, and the product coverage.
the rental equivalence approach, weights represent the rents • The objective is to estimate expenditure consistent with
that owners would pay for the dwelling that they occupy. The the scope of the CPI at a sufficiently detailed product
owners could be directly asked what rent they would pay for level, possibly stratified by region or by outlet type, so
their dwelling; however, such answers are often found to be that reliable weights can be attached to the different ele-
unreliable. The preferred option is to rely on observed rent mentary aggregates that make up the index structure.
data to impute a rental value for the stock of owner-occupied • The primary data source for deriving expenditure weights
dwellings. The assumption is that the rent of an owner- are typically HBSs, but expenditure data may also be
occupied dwelling is expected to be equal to the observed directly obtained from the national accounts.
rent of a dwelling with similar characteristics. Such an • When using the HBS as the primary source for deriving
approach is also consistent with the 2008 SNA (paragraph the weights, the data have to be reviewed and corrected
6.117) which recommends valuing the output of the hous- for possible under or overreporting by comparing them
ing services produced by owner-occupiers at the estimated with other data sources in order to obtain a representative
rental that a tenant would pay for the same accommodation. weighting structure.
3.95 If the CPI includes owner-occupied housing under
• Secondary data sources can be useful to estimate expendi-
the acquisitions approach, the weights are based on the (net)
purchases of dwellings plus major repairs. The purchase of ture for certain product categories, to further disaggregate
a dwelling includes both a structure and a land and location the expenditure at the lower levels, and to correct for over
component. Only the value of the structure element should or underreporting.
be included in the weights of a CPI that aims at measur- • The weights should be updated as often as possible, but at
ing consumption expenditure, whereas the value of the land least once every five years, to ensure their relevance and
should in principle be excluded. Moreover, only dwellings the representativity of the index.
acquired for the purpose of own occupation should be cov- • Countries which are experiencing significant economic
ered. In addition, an owner-occupied housing index under changes, and thus more rapid changes in consump-
the acquisitions approach would include transaction costs tion patterns, should update their weights even more
and other costs related to the use of the dwelling, in which frequently.
case weights must also be estimated for these components. • The lowest level weights are likely to become out of date
Possible data sources for deriving owner-occupied housing more quickly than upper-level weights. Therefore, these
weights are statistics on building activity, business statistics lower-level weights, for at least some categories, need
in construction, or administrative records of property trans- to be reviewed and updated more frequently than upper-
actions. Chapter 11 describes in more detail the treatment of level weights to reflect changes in consumption patterns.
owner-occupied housing in the CPI. The best use of the available statistical information (for
example, market research or small annual surveys) should
Inclusion of Newly Significant Products be made for this purpose.
3.96 As discussed in Chapter 7, newly significant • A regular schedule of updating the weights (at least every
products that appear over time should be identified. If a five years or as often as possible) should be followed and
new product becomes important with regard to household made available to users.
64
SAMPLING 4
Introduction is not static but evolves over time. The samples that were
initially designed for the price reference period may not
4.1 To construct a perfectly accurate consumer price be fully representative of the current period. Therefore,
index (CPI), the price statistician would need to record the samples should be continuously monitored and updated as
price of every variety of all the goods and services that are needed. Chapter 7 describes in more detail the challenges of
in the scope of the CPI. Because it is too costly and, in a dynamic target universe.
practice, impossible to regularly record all the prices of the 4.6 The sample design should support the publication
universe in a timely manner, sampling techniques are used of the detailed subindices that have been agreed for dissemi-
to select a subset of prices that eventually enter the index nation, such as regional indices or separate subindices for
compilation. Consequently, a CPI compilation is based on urban and rural areas. Sample designs should be efficient to
samples. maximize sampling precision while minimizing the costs for
4.2 Sampling occurs on several different levels in the fieldwork and processing. Even without any formal variance
CPI: estimation (as described in 4.83–4.94), sample allocation
• Geographic (locations) and outlet samples: all places and should be optimized by taking into account the weight and
outlets where a product is sold the magnitude of the price change variance of the subindices.
• Product samples: all goods and services available for
purchase Sampling Techniques
• Time: the subperiods of the index 4.7 In survey sampling theory,1 there is a distinction
4.3 In practice, CPI sampling follows a multistage between the parameter and the estimator. In the context of
approach. The universe of products is structured by first a CPI, the parameter is the target price index number that is
selecting the items within the different categories of the based on prices and quantities of the products that belong to
expenditure classification. For each item of the CPI clas- the universe. The estimator is the price index that is actu-
sification, one or more representative varieties can then be ally compiled using the sampled data as input. The result
sampled. For the geographic and outlet samples, specific of the estimator depends on the price index formula that
locations for price collection are selected first. In a second may or may not use weights, and on the sampling scheme
step, outlets are selected within the sampled locations. The that has been adopted for selecting the varieties for which
specific varieties to be priced are ideally selected within the prices are collected. In practice, the parameter is unknown
sampled outlets. Finally, time can be considered as another although simulations can be conducted, for example, using
level of sampling as it must be decided at which moment scanner data to study the performance of different sampling
during the reference month prices are observed. strategies.
4.4 Within each of the different sampling stages, either 4.8 In assessing the quality of a sample estimator (that
probability or nonprobability sampling methods can be con- is, how well it estimates the parameter), two measures are
sidered, possibly in combination with stratification. If the often considered in the case of probability sampling. The
sampling frames required for probability sampling are not first measure is the bias of the estimator, which is the dif-
available, the price statistician either creates one or relies ference between the universe parameter and the average of
on nonprobability sampling techniques. For this reason, the estimator over all possible samples that could be drawn
nonprobability sampling techniques are commonly used to under the specified sample design. An estimator is unbiased
draw samples for price collection in the CPI. However, the if it has zero bias. The second measure is the variance of the
use of some form of probability sampling is generally the estimator with respect to the sampling distribution.
preferred option as it avoids the need for arbitrary decisions 4.9 An estimator is considered accurate if both its bias
and ensures unbiased results. In addition, there are practi- and variance are small; that is, the estimator is on average
cal considerations when organizing the price collection in very close to the parameter and does not vary significantly
the field. In practice, different sampling approaches may be from its mean. The mean square error, defined as the sum
adopted for different parts of the CPI basket. It is thus essen- of the variance and the squared bias, measures the accuracy
tial that sampling procedures are clearly defined and well of the estimator. In addition, samples should be efficient, so
documented. that the maximum sampling precision (and minimum vari-
4.5 When sampling designs are planned, the full uni- ance) can be obtained for the minimum cost with regard to
verse of locations, outlets and outlet types, items, and variet-
ies belonging to the scope of the CPI should be considered.
All significant parts of that universe should be appropriately 1
For a full treatment of the subject, please refer to one of the many text-
represented. An additional challenge is that representativity books available, for example, Särndal and others (1992) or Cochran (1977).
65
fieldwork and processing. If bias is a more relevant problem 4.16 In systematic PPS sampling, the list of units is first
than sampling error, the deterioration in precision caused by ordered randomly and the cumulative total of the auxiliary
small samples can be offset by spending sufficient resources variable xj is calculated. Selection of units then takes place
to select and maintain representative samples. using interval sampling with the interval value calculated
4.10 There are two approaches that can be used to sam- by dividing the cumulative total ΣiϵN xi by the sample size n.
ple units: probability sampling techniques and nonprobabil- This is done by generating a random starting point between
ity sampling techniques. Probability sampling techniques zero and the interval value to select the first unit. The second
require that the probability of a unit being selected as part random number is generated by adding the interval value
of the sample is known in advance and is strictly positive to the starting point and used to select the second unit. This
for each unit. Units can represent the locations and the out- process of adding the interval value to the previous random
lets where households shop or the items and varieties that number, and selecting the corresponding units, is repeated
households buy. Nonprobability sampling techniques do not until the requisite number of units has been sampled. If
rely on sampling probabilities. In practice, a combination of the size of a unit is larger than the interval value, then it is
probability sampling and nonprobability (or purposive) sam- selected with certainty. These units are then removed from
pling is used at the various sampling stages of a CPI. These the sampling frame and the process of cumulating the size
techniques are often applied together with stratification. variable and compiling an interval value is repeated with the
remaining units.
Probability Sampling Techniques 4.17 Systematic sampling is best explained using an
example. Table 4.1 shows how a sample of three outlets
4.11 In probability sampling, a sample of n units from can be drawn from 10 outlets. In a first step, the outlets are
the universe of N units is selected by attaching a nonzero listed in a random order (column A). Although turnover, or
inclusion probability πj to each unit j. The inclusion prob- total gross sales of the establishment, would be the preferred
abilities pj for each unit in the sample are assumed to be selection variable, it is often not available in the sampling
strictly positive and known in advance. This requires the frame. An alternative might be the use of the number of
availability of a sampling frame, that is, a list of all the units employees as a proxy for turnover. The table includes the
eligible to be sampled. A frame may suffer from overcover- cumulative sizes and the inclusion intervals (columns C and
age if it includes units that are not in the universe or includes D). Taking the cumulated (or total) size measure, which is
duplicates of units. It may have undercoverage if units in the 90 in this case, and dividing it by the sample size, 3, gives a
universe are missing from the frame. sampling interval of 30. Next, a random number between 1
4.12 In simple random sampling (SRS) and systematic and 30 is chosen, say 25. The sample will then consist of the
sampling, each unit is sampled with equal probability and outlets whose inclusion intervals cover the numbers 25, 25 +
πj = n/N. In SRS (without replacement), a random number is 30 = 55 and 25 + 2 × 30 = 85 (column E).
assigned to each unit in the frame and the n highest (or low- 4.18 PPS sampling has the advantage of selecting the
est) values are selected. In systematic sampling, the sam- sample in proportion to the relevant variable. It ensures a
pling units are selected at equal distances from each other in sample which reflects the heterogeneity of the population,
the frame, with random selection of only the first unit. These and the sample does not need to be rebalanced by reweight-
techniques are usually recommended in situations where the ing. Alternatively, if each unit could be given an equal chance
units are relatively homogeneous. of selection in the sample, then reweighting may be neces-
4.13 In probability proportional to size (PPS) sampling, sary. For example, SRS would give a large outlet the same
nx j
the inclusion probability π j = is proportional to chance of selection as a small independent shop, despite the
∑ k ∈N xk enormous differences in turnover, and so reweighting would
some auxiliary variable xj. Units for which the inclusion be needed.
probability is greater than one are selected with certainty,
while the inclusion probabilities for the remaining units are
calculated after excluding the large ones. This technique is
recommended if the units are of different sizes, so that larger
units have a greater probability of selection than smaller Table 4.1 Systematic Sample of 3 out of 10 Outlets,
Based on PPS Sampling
units.
4.14 In the context of a CPI, the auxiliary variable is Number of Cumulative x Inclusion Included when
typically the household expenditure on the goods and ser- Employees (x) Interval Starting Point
under PPS is 25
vices covered by the CPI. In practice, this variable is often
not available and alternative variables correlated to expen- (a) (b) (c) (d) (e)
diture must be used, such as retail turnover, population of Outlet 1 13 13 0–13
geographic areas, or number of outlets. The extent to which Outlet 2 2 15 14–15
this assumption about the alternative variables holds true Outlet 3 5 20 16–20
eventually determines the quality of the sampling design. Outlet 4 9 29 21–29 X
Outlet 5 1 30 30
4.15 In a CPI, the size of a sample is often fixed a priori. Outlet 6 25 55 31–55 X
It is therefore impractical to let the sample size be randomly Outlet 7 10 65 56–65
fixed by the sampling procedure. Several techniques exist Outlet 8 6 71 66–71
for drawing fixed sample size samples, for example, system- Outlet 9 11 82 72–82
Outlet 10 8 90 83–90 X
atic sampling or simple random sampling, and so on.
66
Sampling
4.19 Assuming that the prices used to compile the index involved. All units above the threshold are selected with cer-
have been obtained using a specific probability sampling tainty, whereas those below the threshold are excluded.
design, it can then be shown how common price indices 4.23 The disadvantage of cutoff sampling is that it does
are estimators for certain population price indices (see Balk not produce unbiased estimators, since the small units may
[2005] for details). If the size variable in PPS sampling display price movements that systematically differ from
corresponds to the expenditure on the variety in the base those of the larger units. Any estimator resulting from cut-
period, then the sample Jevons price index is an approxi- off sampling has zero variance because there is only one
mately unbiased estimator for the population geometric sample that can be drawn for a given threshold. With regard
price index. Similarly, if the size variable corresponds to the to mean square error, cut-off sampling might be a good
quantities in the base period, then the sample Dutot price choice if the variance reduction more than offsets the intro-
index is an approximately unbiased estimator for the popu- duction of a small bias.
lation Laspeyres price index. These results suggest that vari- 4.24 Cutoff sampling can be considered when the size
eties should be sampled based on expenditures if using a variable is highly skewed. The goal is to select the largest
Jevons index whereas for a Dutot index, varieties should be units that are not so numerous and disregard the many small
sampled based on sold quantities. Finally, the Carli index units. This design is also useful when the “below cutoff sec-
is an unbiased estimator for the population Laspeyres price tion” of the universe is considered insignificant and difficult
index if the size variable corresponds to the expenditures in to measure. A CPI practice considered as cutoff sampling
the base period (the different index formulas are discussed occurs when the price collector selects the most sold product
in Chapters 1 and 8). in an outlet, within a centrally defined specification. In this
4.20 These estimators are “approximately” unbiased case, the sample size is one (in each outlet).
in the sense that the bias tends to zero if the sample size 4.25 Table 4.2 shows the three largest outlets that have
increases. It can be shown that the finite sample bias of been selected using cutoff sampling, with the size of the out-
the Jevons index is always positive. Under SRS, the sample let being measured by its number of employees (column B).
Jevons price index is on average larger than the population This cutoff sampling technique allows covering 54 percent
Jevons price index. For the Dutot index, the sign of the finite of the total number of employees (column D). If the cutoff
sample bias is undetermined, although its magnitude is often rule is defined to cover at least 70 percent of the total num-
negligible. However, for the Jevons index the bias may be ber of employees, two additional outlets must be selected.
significant if sample sizes are very small. Thus, when using Instead of fixing the sample size, the cutoff rule can be
the geometric mean, very small sample sizes should be defined with regard to the cumulative size variable for a cer-
avoided (see Bradley [2007]). Increasing the sample sizes tain threshold. Contrary to PPS sampling, an outlet with a
will not only reduce finite sample bias, but also reduce the small number of employees will systematically be excluded
variance caused by sampling. As a rule of thumb and unless in cutoff sampling. At the same time, this technique ensures
price change variance is very low, at least eight to ten obser- that all the largest units are always covered.
vations for each elementary aggregate should be selected for
the compilation of the Jevons index. Quota Sampling
4.26 Quota sampling is another nonprobability sam-
pling technique used in the CPI. Many product groups, even
Nonprobability Sampling Techniques rather small ones, are quite heterogeneous in nature, and the
4.21 In many circumstances, detailed sampling frames price varies according to many subgroups or characteristics.
are not available, and nonprobability sampling techniques There may be different price movements within such a prod-
are applied, because they do not necessarily rely on sam- uct group, and a procedure to represent the group by one or a
pling frames. Moreover, such techniques allow, to a cer- few tightly specified item types may then carry a substantial
tain extent, more control over the samples than random risk of bias.
approaches. For example, the costs and the feasibility of
price collection play a key role when designing samples for
a CPI, especially given that the sample has to be priced con-
tinuously over the next periods. Examples of nonprobability
Table 4.2 Cutoff Sample of 3 out of 10 Outlets
sampling techniques that are often applied in a CPI context
include cutoff sampling, quota sampling, and the representa- Number of Cumulative x Cumulative x Included in
tive item method. Employees (x) (percent) Cutoff Sampling
When n = 3
Cutoff Sampling (a) (b) (c) (d) (e)

4.22 Cutoff sampling refers to the practice of choosing Outlet 6 25 25 28 X
the n largest sampling units with certainty and giving the Outlet 1 13 38 42 X
remaining units a zero chance of inclusion. To apply cut- Outlet 9 11 49 54 X
off sampling, a sampling frame is required that contains an Outlet 7 10 59 66
Outlet 4 9 68 76
appropriate variable measuring the size of the unit. More- Outlet 10 8 76 84
over, the quality of that variable has to be judged when Outlet 8 6 82 91
performing cutoff sampling. The term “cutoff” refers to the Outlet 3 5 87 97
threshold value between the included and the excluded units. Outlet 2 2 89 99
Outlet 5 1 90 100
Unlike with probability sampling, there is no randomness
67
4.27 In quota sampling, the sample selection is made Table 4.3 Stratification by Region, Outlet Type, and
using judgmental procedures with respect to known and Product Type
relevant characteristics when choosing products to price or Type of Outlet Product Region 1 Region 2
the type of outlet to select. The sample is drawn to have
Specialized Independent Specialized Independent
the same proportions as the total population or universe to Chains Shops Chains Shops
ensure that the sample is representative. Consequently, the
sample is self-weighted. Quota sampling is a stratified sam- Refrig- Small
erator Capacity
pling method with a sample allocation that is proportional to Capacity Large
the stratum weights and where the sampling within a stra- Capacity
tum is conducted in a judgmental way.
4.28 The following example illustrates the concept of
quota sampling. If there is a characteristic that allows divid-
ing a product group into three product types (for example, rural, or urban), by type of outlet or by outlet, and by item/
red delicious apples, granny smith apples, and Honeycrisp product. Samples are set up within each geographic area,
apples), each representing 30, 20, and 50 percent of house- for each of the different outlet types, and for each item/
hold expenditure, then the sampled prices will also represent product within outlets. For example, suppose that the CPI
the same proportions. If there are 10 prices to be collected, a classification contains the item refrigerators that can be
quota sampling approach recommends that three prices are stratified by region. In a given region, refrigerators are pri-
collected for the first product type, two for the second and marily sold in larger specialized shop chains and in several
five for the third. If the Dutot index is used, then quanti- small independent shops. In addition, refrigerators can be
ties should be used instead of expenditure for defining the classified according to their capacity. Such a stratification
importance of the strata. structure is illustrated in Table 4.3. Within each cell, prices
can be sampled using a two-stage approach. In the first step,
Representative Item Method specialized chains and independent shops that sell this type
4.29 In the representative item method, one or several of goods are selected within the different regions. In the
item specifications are defined to represent the item. For selected outlets, specific refrigerator models that meet the
example, spaghetti may be a representative item for “Pasta.” product-type specifications are then identified for continu-
In this example, the price changes for pasta will be measured ous pricing.
using the sampled price changes for spaghetti. According to 4.33 Formally, if a subcomponent of the CPI is com-
this method, only varieties matching the specification are posed of several strata and the subindex for stratum k is
priced and no varieties falling outside the specifications will labeled Ik, then the subcomponent’s estimate is obtained
enter the index. The sampling of representative items for a using the strata’s weights and indices:
CPI is usually purposive or judgmental, because of the lack
of adequate sampling frames that formally list all the pos- I=
∑ k ∈K
wk I k
(4.1)
sible product types. ∑ k ∈K
wk
4.30 If the representative item method is used for sam-
pling products, the number of items must be sufficiently where wk is the weight of stratum k. The weight of the stra-
large to properly represent the diversity of products that tum corresponds to the expenditure of that stratum, and not
can be found in a given product category. Prices that only to the (smaller) expenditure of the locations, outlets, or
belong to the same representative item may be relatively varieties that have been sampled within that stratum.
homogeneous, both with regard to price level and price 4.34 Ideally, the stratification should be designed to
change. What matters is the variance in price changes minimize the sampling error. The strata should be con-
between the different representative items that could be structed so that the variance of the price changes within
selected. Sampling variance can be reduced by specify- strata is low, while the variance between the strata can
ing more representative items in those product categories remain large. A low-price change variance within stratum
where the price change variance between the representa- also has the advantage that results obtained using differ-
tive items is large. ent price indices are likely to be similar. Consequently, the
choice of the elementary aggregate price index will have a
smaller impact on the final results.
Stratification 4.35 Stratified sampling avoids selection bias by ensur-
4.31 A common sampling technique used in the CPI ing that the universe is appropriately represented, as it
is to divide, in advance, the universe into disjoint homoge- controls for the fact that prices must be collected for each
neous subpopulations or strata. An independent sample of specified stratum. Without stratification, certain parts of the
appropriate size is then selected for each stratum using any universe may be completely ignored. In addition, the use of
of the probability or nonprobability sampling techniques explicit weights, if available, guarantees that the prices are
previously described in this chapter. Stratified systematic weighted according to the importance of each stratum. In
random sampling is a sampling technique in which system- price statistics, stratification thus helps reduce the bias of
atic random sampling is conducted independently within the estimates.
each of the predefined strata. 4.36 Stratification is a useful strategy to make the sam-
4.32 Stratification can be applied to all of the differ- ple more efficient. The allocation of the number of prices
ent samples relevant for the CPI. In practice, the CPI is across the strata is further discussed later in this chapter
typically stratified by geography (for example, region, city, (paragraphs 4.83–4.94). Furthermore, the availability of
68
Sampling
detailed subindices can be convenient for meeting specific discussed in the following paragraphs. The product level is
publication needs. The impact of stratification on sample related to the outlet level. At some point, the sample of items
sizes using different elementary aggregation formulas has must be matched to the sample of outlets to decide which
been analyzed by De Gregorio (2012). items are priced in which outlets. Time is an additional sam-
4.37 Although the universe can exhaustively be divided pling level that must be considered. If timing is relevant,
into a set of strata K, there are circumstances where only a it may be necessary to incorporate timing elements in the
sample of these strata S(K) is used in practice for index com- outlet or item specifications.
pilation. Only prices belonging to the sampled strata S(K)
will be collected. The strata can be selected purposively. One Sampling of Locations
strategy could be to select only the most important strata,
which is equivalent to running a cutoff sampling procedure. 4.41 The objective of the first sampling stage is to
The bias that results from this process depends on the weight select the locations where prices are going to be collected.
of the omitted strata and on the difference between the price A location is an area where outlets that sell products to
change of the omitted strata and the sampled prices. If the households can be found. Ideally, a large part of the CPI
weight of the omitted strata is small or their price change is product basket should be available to purchase within each
similar to the price change of the sampled prices, then the location. To obtain a representative sample of outlets, it
bias may be small. may be necessary to create locations by combining, for
4.38 If the strata composing the index structure are example, out-of-town shopping areas that mainly sell non-
selected using probability sampling, then the aggregation food items with a neighboring urban area in which food
of the stratum indices should be adjusted using sampling is available. The sampling of locations can be omitted if
weights. In particular, if expenditure is used as the size vari- outlets are directly sampled independently of where they
able in PPS sampling, the higher-level estimate is compiled are located.
as the simple arithmetic average of the sampled stratum 4.42 To define the list of possible locations, a decision
indices. In the example shown in Table 4.1, three outlets must be made on the geographical scope of the price col-
were selected using PPS sampling. If outlets 4, 6, and 10 lection areas. For example, price collection can be limited
each represent a stratum in the index structure, then the to the capital city only or to the main cities of a country
weight associated with each stratum does not correspond or of a region. Some countries restrict price collection to
to the actual weight of that outlet. Instead, an unbiased urban areas because of the challenges and costs of collecting
estimate is obtained by equally weighting the three stratum prices in rural areas. Alternatively, it can be decided to select
indices: locations belonging to both urban and rural areas. Under the
broadest geographical scope, locations must be sampled to
30 30 30 be representative of the national territory as a whole (as fur-
I= I Outlet_4 + I Outlet_6 + I Outlet_10 (4.2) ther discussed in Chapter 2).
90 90 90
4.43 The sampling of locations is typically conducted
4.39 A challenge arises in systematic PPS sampling using regional stratification. Location selection can then be
when the size of a stratum is larger than the interval value. done separately within each region. For instance, the strata
In such a situation, a distinction must be made between used for sampling can be aligned with the administrative
the larger strata that are selected with certainty (“take-all” subdivisions of a country and considered the distinction
strata) and the other strata that are sampled after the larger between urban and rural areas. In principle, stratification
units have been removed (the “take-some” strata). The works best if the locations of a stratum are homogeneous. In
higher-level index is then obtained as a weighted average that sense, the territory could also be partitioned by group-
of two subindices. The first subindex covers all the certainty ing together the locations that are situated close to each
strata, while the second subindex contains a sample of some other or that have similar sociodemographic characteristics.
of the lower weighted strata. In the first subindex, the actual If no regional stratification is used, locations are sampled
weights of the certainty strata must be used in the aggrega- directly at the national level.
tion, as these strata only represent themselves. In the second 4.44 Within each region, some locations may be
subindex, the remaining sampled strata are aggregated with selected with certainty, and the selection process is not ran-
equal weights. dom. This can be the case of the main cities. Locations can
also be selected randomly, using, for example, systematic
Sampling Stages in a Consumer PPS sampling. To apply this kind of technique, a variable
that measures the size of the locations must be available and
Price Index the number of sampled locations within a region must be
4.40 In practice, national statistical offices often adopt fixed in advance. As the information on expenditure made
four levels of sampling in a CPI: locations, items within by households in outlets situated in a given location is rarely
different sections of the expenditure classification, outlets available, proxy measures must be used. The number of
within locations, and item varieties.2 These four steps are households living in a location can be used as a size variable,
although the residence of a household does not necessarily
2
coincide with the place of purchase. Population size can be
The order of the sampling stages can differ. For instance, varieties could obtained from population registers or from a recent census.
be selected centrally and then outlets could be selected to price these vari-
eties. Alternatively, outlets could be selected first, and the choice of the The number of locations to be selected in each region can
specific varieties to be priced is only taken when the price collector visits be determined as the proportion of national expenditure,
the outlet. income, or a corresponding proxy measure, such as regional
69
gross domestic product, in that region, multiplied by the can be drawn up by the national statistical office using
total number of locations to be visited nationally. specifications. This is what is referred to as the representa-
4.45 There are operational implications for setting up a tive item method (as described in 4.29–4.30). The product
new price collection in the sampled locations, and the pro- group is thus represented by a sample of specified product
cess described previously may need to be modified for prac- types (items). The selection of representative items at a cen-
tical considerations. For example, because data collection tral level ensures that they are representative of consumer
costs are important, it may not be efficient to send a price behavior, rather than of one shopping location. Consistent
collector to a small isolated location where many of the items with the principle of a fixed basket, the representative items
in the CPI basket are not available for pricing. Location resa- are usually reviewed periodically at the same time as the
mpling might be needed, although it is likely to be conducted weights and chain linking is carried out. The disadvantage
on a less frequent basis than outlet or product resampling. of this method is that the national specification may often be
Sample coordination techniques can be used to adjust the unavailable in some locations and does not reflect regional
initial selection probabilities to increase the likelihood of an tastes and preferences.
overlap between the old and the new sample of locations. 4.51 A representative variety can be defined using either
loose or tight specifications. For example, it may be decided
to collect prices for “Tea,” which is a loose specification,
Sampling of Items as it encompasses many different tea varieties that could
4.46 Cutoff sampling can be applied to select the items be selected to be priced. Alternatively, the representative
that comprise the CPI basket. Expenditure data on items variety may be specified tightly, as a specific variety of tea
are typically obtained from a household budget survey (see defined by a brand, flavor, and package size. The discussion
Chapter 3 for more information on data sources). However, between loose and tight specifications has implications for
not all items for which expenditure data are available might the choice of the elementary aggregate formula (see Chapter
be included in the CPI basket. There may be low-expenditure 8 for more information).
items for which it would not be an efficient use of limited 4.52 If specifications are too tight, the price collectors
resources to collect prices (although they are within the scope may have difficulty in finding the varieties, and this will lead
of the CPI). Thresholds can be defined to identify the low- to fewer price quotes and a deficit in the sample. There is
expenditure items that will not be included in the CPI sample. also a risk that tight specifications will miss important parts
4.47 In practice, an expenditure share is estimated for of the market, and consequently, results can be biased. For
all the items that could be included in the basket. If this these reasons, the representative variety method is more
share is below a given threshold, the item could be excluded. suitable for homogeneous product groups. In heterogeneous
For example, it might be decided to exclude from the index groups, it is more likely that important segments of the item
items with an expenditure share lower than 0.1 percent for universe, with different price movements, will be left out.
food items and 0.2 percent for nonfood items. A lower mini- 4.53 Loose specifications give the price collector some
mum threshold for the food items might be set because the freedom in choosing locally popular items and varieties
prices of these products tend to display a higher variability and to adjust the sample to match local conditions. It will
and because prices for food products are usually less costly normally lead to greater representativity of the sample, as
to collect. If an item is excluded, its expenditure may be it will reflect regional tastes and preferences. However, this
redistributed to another expenditure group with similar con- approach may require training for price collectors on how to
tent and price development. select within the outlet a variety to be priced that matches
4.48 The cutoff process can be conducted at the national the loose specifications. This requires a more formal pro-
level or separately within a region. The list of items used cess selecting the varieties within an outlet. Even with loose
in each region may differ. This approach ensures that the specifications, the variety selected in the outlet must be
regional baskets are representative of the consumption described in sufficient detail by the price collector to ensure
structure within each region. Manual adjustments may also that the same variety is priced over time, and in case of a
be needed so that an item that falls below the threshold can replacement, a correct quality adjustment can be made.
nevertheless be included in the basket. For example, it may
not be so costly to collect a few additional prices if these are
collected together with prices for more important items. The Sampling of Outlets
additional costs may be acceptable given the advantage of 4.54 Different data sources can be used as sampling
achieving wider product coverage. frames for selecting outlets. In the CPI, an outlet should be
4.49 The descriptions (or detailed specifications) of the defined broadly to include any retail channel that sells goods
items that are included in the CPI basket can be general or or services to households, including different types of physi-
more detailed. This largely depends on how detailed expen- cal shops, online shops, or the offices of providers of utilities
diture data were recorded in the household budget survey. If even when a physical shop that customers can visit is not
information is available from other sources, a finer stratifi- available. Outlets to be visited in the field may be selected
cation can be developed. In any case, the universe of variet- only in locations that were previously identified as price col-
ies that fall within the scope of an item is often quite large. lection centers.
Consequently, additional sampling is required at this level 4.55 Business registers can be a starting point for select-
for the selection of the specific varieties to be priced over ing outlets. Business registers are databases that list the
time. business units resident in a given area and contain informa-
4.50 To provide effective control over the sample of tion such as location, type of business activity, turnover, and
prices collected in the field, a list of product types (items) employment. Business registers should be updated regularly
70
Sampling
for new businesses and for closures. One limitation of busi- type of product. Additional guidance may be provided to the
ness registers is that, in some cases, data are available only price collector by specifying the type of outlet or by defining
at the level of establishment or enterprise, and the informa- the area where outlets are to be selected, to ensure that the
tion at the local level is not always available or regularly overall sample of outlets remains representative of where
updated. Provided that some size measure is included in the households shop. This sampling method often involves vis-
database, such as number of employees, business registers its by the price collector to the outlet to gather the necessary
can be used for selecting a PPS sample. Similarly, tax data information on the outlet’s characteristics before deciding
such as value-added tax records may be used as a data source which outlets are included in the sample and before the first
that has a wide coverage and contains timely information on price collection is conducted. For example, if prices for chil-
outlets. Tax records typically refer to a legal unit that can be dren’s clothes are to be collected, a price collector may visit
composed of one or several outlets. With tax data, the selec- a clothing shop initially selected from a central list to verify
tion of random samples can usually be based on gross sales and check whether it sells clothes for children and not just
as the measure of size. for adults.
4.56 Administrative records kept by local government, 4.60 The sampling strategy for outlets typically relies
business associations, or marketplace managers are another on stratification by outlet type. Within each stratum, outlets
potential data source for outlet sample selection. These are sampled randomly or purposively if no appropriate sam-
records can be used to create sampling frames and might pling frame is available. For example, the number of out-
be particularly useful for sampling local markets. Depend- lets to be sampled in each stratum could be proportional to
ing on the availability of information on size, these sources the weight of that stratum. In any case, all significant outlet
might provide a frame for PPS sampling. Business tele- types should be covered. If relevant, the internet can also be
phone directories usually contain less information, such as considered as an outlet type or stratum from which specific
the business name, address, and activity, and do not include websites are selected.
any size information. Therefore, telephone directories are 4.61 As similar price trends can be expected in the out-
useful for simple random sampling or systematic sampling, lets that belong to the same chain, it may be useful to also
but not for PPS sampling unless additional information is stratify by retail chain. For example, the sample design could
made available, for example, by visiting the outlets. include a first stratum that contains all the major retail chains
4.57 Information on outlets can also be obtained from as substrata and a second stratum that covers the remaining
surveys designed for this purpose. For example, the objec- independent shops. In this case, outlets can be selected sepa-
tive of point-of-purchase surveys is to collect information rately for each retail chain and from the remaining universe
on the place where purchases are made by households and of independent shops. The index should be designed so that
on the amount of these purchases. In some cases, the house- the retail chains and other outlet types are properly repre-
hold budget survey includes information on the outlets or sented according to their market share.
type of outlet from where the households purchased the
goods and services recorded. Surveys that include this infor- Sampling of Varieties
mation could be used as a source of relevant data on outlets
for sampling frames. However, as these surveys are based 4.62 In a final step, the selection of specific varieties
on a sample, they are unlikely to provide an exhaustive list within outlets is usually made by the price collector in a pur-
of all outlets and the results obtained may become quickly posive way. The objective is to select the most sold variety
out of date. that fits the item specifications. This process essentially cor-
4.58 If the data sources described previously are not responds to a type of cutoff sampling. In addition to being a
available, it is necessary to enumerate the outlets within a high-volume seller, the variety is also expected to be avail-
selected location to provide supplementary information for able over time so that the same variety can easily be priced
the sampling frame or even to construct a sampling frame. during the following periods.
This enumeration can be made by price collectors and their 4.63 A compromise must sometimes be made between
supervisors, visiting each location and noting details (for representativity and comparability. On the one hand, it is
example, name, type of outlet, and items sold) of all retail important to select varieties with high sales, as this will
outlets found. This can be costly, and the price statistician ensure that the samples are representative of what house-
will need to compare the costs and benefits of a more rep- holds are buying. On the other hand, the future availability
resentative sample. To reduce costs, the number of outlets of the variety must be considered to improve comparability
enumerated per location may be limited by enumerating over time by minimizing the need for replacements. How-
only a subdivision of a location. The ideal size measure of ever, there is a risk of bias if the choice of varieties is mainly
an outlet is turnover. However, a proxy may be used when driven by the convenience to collect. Chapter 5 discusses the
an estimate of turnover is not available. For instance, the price collection in more detail.
approximate net retail floor space estimated by the outlet 4.64 With broad item specifications, many products
enumerators may provide an adequate alternative. Addition- available in an outlet can be selected for inclusion in the
ally, it is important that the assortment of goods and services CPI sample. A more formal approach may be adopted by
sold by the outlet is recorded, to allow matching the outlets designing a process that approximates PPS sampling to the
with the CPI product basket. sales of each product.3 To apply this technique, all possible
4.59 If none of the formal sampling techniques are fea-
sible, outlets must be selected in a more judgmental way. 3
This applies if the Jevons index is used for elementary aggregation. If the
For example, field staff can identify the largest retail and Dutot index is used, the size variable should be the number of sales and not
service outlets where households typically shop for a given the total sales revenue.
71
products or product types (items) must first be enumerated, For example, for airfares or other transport services, the
and selection probabilities are determined for these products price can be highly dependent on the day of the week, the
based on their sales, for example, using the information on time of the flight, and how long in advance the ticket is pur-
sales obtained directly from the respondents. Alternatively, chased. Such timing elements should be held constant to
shelf space can be used to estimate sales proportions in some ensure the comparability of the collected prices over time.
cases. Even with no information on sales, random sampling Moreover, the specifications including their time-dependent
can still be conducted by simply assuming that all products components should be defined to be representative of con-
have the same selection probability. sumer behavior.
4.65 If the data for the sales measure is taken during a
very short period, this may coincide with a special promo-
tion or sale with temporarily reduced prices. If a variety with Central Price Collection
a temporarily reduced price is given a larger probability for 4.71 For some products, prices can be collected cen-
selection, and if this price is likely to increase more than trally by CPI staff. These include those prices obtained from
average, an overestimating bias may result. It is thus essen- a specific data source or by a survey conducted centrally.
tial that the sampling takes place before the first price col- The same general principles apply as with price collection
lection or that sales values from an earlier period are used. in the field in the sense that outlets or respondents must be
Generally, annual sales values would be the best measure. chosen and specific varieties for regular pricing must be
4.66 Administrative data sources are sometimes avail- selected. However, there is no operational need to select
able with detailed expenditure data by product or by product specific price collection areas in advance, and a broader geo-
group. Such data sources can help build detailed stratifica- graphical coverage of price collection can often be achieved.
tions and serve as a sampling frame for selecting the indi- Moreover, it is possible at the head office to design specific
vidual varieties to be priced. sampling strategies adapted to the product or to the data
4.67 If no explicit weights are used in the elementary sources at hand. Some examples of central price collection
aggregate compilation, the sample will be implicitly weighted are discussed in the following text.
by the number of observations. For example, if prices of the
same two varieties are collected in five outlets, then each out-
let and variety implicitly have the same weight. The elemen- Scanner Data
tary aggregate is often the lowest level for which weights 4.72 When using scanner data, the product and the out-
are available. However, efforts can be made to improve the let sampling levels are closely related. A two-stage sampling
accuracy of this detailed subindex using implicit or explicit approach can be adopted where outlets are sampled first,
weights. The sample can be designed so that the proportions possibly belonging to the same retail chain or outlet type,
within an elementary aggregate represent those of the uni- and then varieties are selected within the sampled outlets.
verse with respect to the outlet type or outlet, if these aspects If price levels are similar between outlets or if data are not
are not explicitly weighted in the CPI aggregation structure. available by outlet, varieties can also be sampled at a more
Similarly, the sample should be well balanced with respect aggregated level, for instance at the level of an entire retail
to product type (item) or item characteristics. For example, chain. Eventually, the compilation methods can be adapted
if local beer has an approximate market share of 80 percent to use all the available data on both prices and quantities,
and imported beer of 20 percent, these proportions can be and no sampling is thus involved. Even if scanner data
reflected in the elementary aggregate. Chapter 8 discusses the cover “all” transactions, these data do not necessarily cover
use of weights within elementary aggregates in more detail. the entire universe. In principle, the remaining part of the
universe should also be covered if it is significant, if it is
expected to have different price developments, and if there
Sampling in Time are sufficient resources to maintain separate price indices
4.68 The timing of price collection is chosen purpo- not derived from scanner data. Weights must then be used at
sively in most cases. The main principle is that the prices some level of the index hierarchy so that each part is appro-
of each item should be collected each month at the same priately represented. Chapter 10 discusses the use of scanner
time, during the same week or the same day of the month. If data in detail.
there is some price variation within a day, it is important that
prices are collected always at the same time of the day. Price
collection can be limited to a certain week of the month, for Tariffs
example, during the middle of the month, or it can be spread 4.73 For many tariffs, there is often a national pricing
over several weeks of the month according to some pattern, strategy and the regional dimension discussed previously
for example, in different weeks in different regions or for becomes irrelevant. There may only be a limited number of
different product groups. providers (sometimes only one), and it is straightforward to
4.69 If prices are known to fluctuate considerably dur- select either all of them or at least the most important ones,
ing a month, then an increased price collection frequency and these can then be explicitly weighted in the index. For
should be considered. In this case, it can make sense to col- tariffs, it is often possible to obtain additional information
lect prices once a week or sometimes once a day instead of on products directly from the providers, authorities, regula-
once a month, as practical. This could be the case for energy tory bodies, or other administrative data sources. Consumer
products or for fresh food. profiles are sometimes defined (sampled) to price the tariff
4.70 Some products have a time-dependent component structure. Chapter 11 discusses the different price measure-
that should be incorporated into the product specifications. ment approaches for tariffs.
72
Sampling
Rents website can be larger than the number of prices collected in

4.74 Examples of sampling frames used for rental the physical outlets because of lower price collection costs,
dwellings are registers of dwellings, lists of addresses, or a wider product coverage, or an increased price collection
registers of tenants and landlords. Information may also frequency. Chapter 11 discusses the treatment of internet
be available in the microdata of a census of buildings and purchases.
dwellings. Stratified sampling would be the preferred option
to select rents. Typical stratification variables that corre- Sample Maintenance
late with rental change are the location, type, and size of
the dwelling, the type of rent (social/fixed/regulated rent or 4.77 The samples are drawn at a point in time and based
market rent), or the type of landlord. Ideally, each stratum on the information available at that moment. However, the
should be weighted according to its total rental expenditure situation in the market is seldom static. Products and out-
and random samples should be drawn within each stratum. lets gain or lose importance over time. New products con-
If no sampling frame is available and it is too costly to build stantly enter the market while other products may no longer
one, two-stage cluster sampling can be a practical option. be available. New outlets can open in a certain location and
The country or the region has first to be divided into detailed previously popular outlets close. This means that the sample
geographic subdivisions (“clusters”) from which a sample that was set up at an earlier period may not be representative
can be drawn. The in-scope rental dwellings of the selected of household expenditure at the current period.
clusters are then enumerated so that a random sample of 4.78 Over time, the quality of the sample can dete-
rents in each selected cluster can eventually be obtained. An riorate. To avoid this deterioration, it is recommended that
alternative approach is to collect rental data directly from samples be continuously reviewed and adjusted as needed
real estate agencies. Chapter 11 examines the treatment of (see Chapter 7 for more information on maintaining the
housing in a CPI. sample). If resources do not permit a systematic full review
of products and outlets, a rolling review can be set up, where
every year a different part of the basket is analyzed. The
Chains with National Pricing Strategies focus should be on products and outlet locations that are
4.75 Prices can be collected in a single outlet of a larger likely to be affected by major developments.
chain if it is believed that the same price applies in all the 4.79 In practice, feedback received from field staff can
other outlets of the same chain across the country. This be considered to maintain the ongoing representativity of
approach may however be subject to some caveats. First, the samples. Price collectors and their supervisors should
there is no guarantee that prices are always the same in all the be encouraged to report if the item specifications used for
outlets of the same chain. For example, sales or promotions price collection are no longer in line with the products sold.
could be outlet-specific. Second, it must be ensured that the Through their local knowledge, the staff in charge of price
varieties selected in one outlet are representative of the con- collection may identify new outlets that are becoming popu-
sumption habits in the outlets located in the other regions. lar and should be included in the CPI samples. An increasing
In any case, the prices must be appropriately weighted in the number of missing prices can also be an indication that the
CPI. If a price quote from a large retail chain is combined outlets or the product specifications are no longer appropri-
with prices from other outlets within an elementary aggre- ate, and action must be taken to update the samples. Some
gate, an appropriate representation of the retail chain can national statistical offices have dedicated product specialists
be achieved through implicit weighting by replicating the who monitor the markets of complex product groups and
price of the retail chain. Alternatively, specific elementary can assist with sampling and quality adjustments.
aggregates weighted according to the market shares of the 4.80 Replacements are one way of preserving the repre-
different outlets could be constructed. sentativity of the samples over time. In some cases, price col-
lectors find that an outlet at which they have been collecting
prices closes or no longer stocks one of the products being
Internet priced. Disappearing outlets and varieties within a specific
4.76 For prices collected on the internet, a distinction outlet are handled by replacing on a one-to-one basis when
must be made between prices collected online, so that there is the outlet closes or the variety ceases to be sold. A more
no need for a price collector to visit an outlet, or those prices proactive approach can also be adopted by anticipating the
that genuinely represent e-commerce transactions (online or expected disappearance of an outlet or a variety so that the
web-based outlets with no physical location). In the latter number of forced replacements is kept to a minimum.
case, one could still adopt the same sampling strategy as for 4.81 The criteria for selecting replacements will vary.
price collection in the field, by first selecting websites and An outlet can be replaced on a like-for-like basis by another
then selecting specific varieties to be priced. The selected outlet of the same type, in the same or similar location,
websites are typically organized according to different prod- selling the same or similar range of items and varieties. In
uct categories that can serve as strata for product sampling. replacing a variety, two strategies are usually adopted. If the
Instead of collecting prices for only a few varieties, web initial selection rule was “most sold” or “probability pro-
scraping techniques can be used to obtain significantly more portional to (sales) size” then the replacement may follow
data in an automated way (see Annex 5.6). Prices collected the same rule with the advantage of maintaining the repre-
on the internet can sometimes be highly volatile over time sentativity of the sample. However, the replacement variety
with some websites applying dynamic or personalized pric- may no longer be comparable to the variety that it replaces.
ing strategies. This implies that prices should be collected In this case, quality-adjustment methods must be applied to
more often. The number of price quotes collected on a ensure the index measures pure price changes (see Chapter 6
73
for more information on quality adjustments). There is a risk collection activities. If a variance and a cost function are
of bias if the differences in quality are not properly adjusted. known, an optimization program can be defined to minimize
Alternatively, the replacement may be made based on the the variance given the budget available for price collection.
variety most similar to the one which disappeared, thereby Alternatively, costs can be minimized under the constraint
reducing the need for quality adjustment. When replace- that a minimal variance must be achieved.
ments are performed, there is an inherent tension between 4.86 The general approach to distribute the number of
representativity and comparability. observations across strata was established by Neyman. If
4.82 The dynamic nature of the target universe can also each stratum has a known weight wk and samples are drawn
be addressed through sample rotation involving a full or a randomly and independently in each stratum, then the stan-
partial resampling. Sample rotation is a technique where dard deviation of the individual price changes (not the price
the length of time that outlets and varieties are included in levels) within a stratum can be denoted by sk. In addition,
the price surveys sample is limited. In a given period (for let ck be the cost to collect one unit in stratum k. If the total
example, every year or every two years), a defined propor- sample size is n, the following allocation is optimal:
tion (for example, 20 percent or 25 percent) are rotated out
of the sample and replaced with a new sample of outlets and wk sk / ck
nk = n (4.3)
varieties for that proportion thereafter. The process of resa-
mpling can be based on any of the methods and sampling
∑wsk k k / ck
frames that are used for initial sample selection. Resampling 4.87 This expression formally shows that the number
in the CPI can apply to outlets or even locations. Techni- of observations in a stratum should be proportional to the
cally, resampling involves an overlap period where the first weight of that stratum times the standard deviation of the
period of the new sample overlaps with the last period of the price changes and inversely proportional to the square root
old sample. of the unit cost. The expression can be simplified by assum-
ing that costs are identical in all strata. The underlying idea is
that more price observations are needed for those strata with
Variance Estimation and Optimal a higher weight and a significant variance in price changes.
Allocation Additionally, it is sufficient to allocate fewer prices to the
4.83 A CPI is a statistic with a complex sampling strata with a smaller weight and more similar price changes.
design that involves different stages and that often relies 4.88 In practice, there are challenges in systematically
on nonprobability sampling methods. Therefore, estimating applying this type of allocation. While stratum weights are
the variance of a CPI is not a trivial task. Strictly speak- in principle available, the estimation of the standard devia-
ing, variance estimation is only applicable in case of prob- tion of price changes may be more problematic. One dis-
ability sampling, as the sampling mechanism is known. To advantage of this approach is that it depends on the target
the extent that samples in a CPI are not probability-based, price change. For instance, the prices can be compared to the
variance estimates must rely on models in which some type previous month or to the same month of the previous year.
of random sampling is assumed. Different approaches for The optimal allocation can be different if it is derived for
variance estimation in a CPI have been developed, taking monthly price changes or for annual price changes. More-
into account national circumstances. It is useful to know over, the standard deviation of price changes of a stratum is
the sampling error of a CPI as this can inform the users not necessarily constant over time.
about the level of precision of the published indices. The 4.89 Data sources, such as previous CPI surveys, can be
interpretation of such error margins depends on the specific used to estimate the standard deviations of the price changes
approach adopted for variance estimation of each stratum. However, there is a risk that an estimation
4.84 A very general form of a CPI is I = ∑wk I k , where based only on past observed prices is biased. Considering
k the situation in which a stratum is represented by a unique
k denotes products, wk is the expenditure share of this prod- item with very tight specifications, the prices collected in
uct, and Ik is the product price index. If the estimation of the different outlets may all change at a similar rate, and
each product index was independent, the variance would be this will consequently lead to a low sample price change
V ( I ) = ∑wk2V ( I k ), where V(Ik) denotes the variance of the variance. The true population price change variance may be
k much larger than the estimated because of the unobserved
product index k. However, product indices are not statisti- price change behavior of the varieties that belong to the stra-
cally independent. For instance, the price change of differ- tum but are not included in the CPI sample.
ent products collected in the same outlet can be related. The 4.90 Costs also need to be considered when defining
sampling errors of the product indices are therefore corre- sample sizes. It may be more time-consuming and complex,
lated, which means that this expression probably underesti- and, hence, costly, to collect prices for some goods and ser-
mates the total sampling errors to some extent. vices than for others. For instance, for products where qual-
4.85 From an operational point of view, variance esti- ity adjustment plays an important role, the price collector
mation is a useful tool for deciding on the optimal allocation can be asked to record, in addition to the price, many tech-
of sample prices. Many resources are spent on price collec- nical characteristics. Moreover, some outlets have larger
tion to produce a CPI. It is therefore worthwhile to devote assortments, meaning that more prices can be collected in
some effort to allocating these resources in the most efficient one outlet, which then reduces the cost per collected price.
way. A better allocation of prices can often minimize the In practice, the prices could be classified in categories, such
sampling error and consequently improve the quality of the as “easy to collect” or “difficult to collect.” Such ordinal
CPI without increasing the resources associated with price information could be used to derive an optimal allocation.
74
Sampling
Table 4.4 Different Allocation Strategies strategy is to select a smaller number of outlets and a larger
Item Weight Price Number of Prices
sample of varieties within each outlet. In addition, such a
(percent) Change strategy is cost-efficient as it is often less costly to sample an
Option 1 Option 2 Option 3
Variance additional price in an already selected outlet than to increase
(a) (b) (c) (d) (e) (f) the outlet sample. For example, in some countries, car
prices do not vary much across outlets for the same model
Item 1 10 Medium 20 10 percent × 11 and specification, but there are many models with different
100 = 10
Item 2 40 Medium 20 40 43 prices. One car retailer could easily provide prices for sev-
Item 3 25 Low 20 25 18 eral models during the same collection period.
Item 4 20 Medium 20 20 21 4.94 The time dimension must also be considered for
Item 5 5 High 20 5 7 price collection. If prices vary a lot over time, it is preferable
Total 100 100 100 100
to collect more prices for the same product during a month.
For example, petrol prices normally show little variation
among outlets on the same day while prices can vary con-
A more advanced quantification of costs would be based on siderably even over a month. It would thus make sense to
estimates of the average time per collected price by product have a relatively small sample of outlets but follow prices
or product group, while considering the travel time from one several times in the same month. The time dimension is also
outlet to another. highly relevant for internet purchases. As it is probably less
4.91 In Table 4.4, a sample of 100 prices needs to be costly to consult websites than to visit physical outlets, it
distributed over five items. A first strategy consists of col- can be feasible to improve the temporal coverage of prices
lecting the same number of prices for each item (option 1). collected on the internet.
Instead of allocating the same number of price observations
to each stratum, an alternative approach consists of linking
the number of prices to the weight, collecting more prices Key Recommendations
in those strata that have a larger weight (option 2). This
will improve the sampling error of the subindices that have • When designing samples for the CPI, the index compiler
more impact on the higher-level aggregates because of their must consider the different sampling stages relevant for
weight. In addition, judgments can be made on the magni- a CPI: geographic and outlet, products, and time. All
tude of the variance. This means that the sample sizes should significant parts of that universe should be appropriately
be increased (or decreased) for strata with an expected higher represented unless there are excessive costs or estimation
(or lower) price change variance. As an approximation, the problems involved in doing so.
Neyman allocation formula is applied by assuming fac- • The CPI aggregation structure is organized according to
tors of 1 (low), 1.5 (medium), and 2 (high) for the standard a product classification, and possibly stratified by region
deviation. The result of the allocation can be found in the and by outlet type.
last column (option 3). Because the price change variance • Within each stratum, the collected prices are typically
is expected to be low for item 3 (as indicated in column C), the result of multistage sampling designs. Price collec-
sample size for this item is decreased whereas the number tion locations and items are defined beforehand. Outlets
of prices to be collected for the other items increases. Out of that sell the items of interest must then be chosen in the
these three options, the third allocation aims at reducing the selected locations. The specific varieties to be priced may
sampling variance while keeping the total number of prices only be selected when the price collector visits the outlet.
collected unchanged. Finally, the sample size per stratum • Either probability or nonprobability sampling techniques
should not drop below a fixed threshold to avoid formula can be used. Probability sampling is the preferred option
bias or to guarantee a sufficient precision if stratum indices if detailed sampling frames that include data on size (for
are separately published. Totals are shown in bold. example, total sales) are available.
4.92 To design optimal samples, it must be well under-
• Samples should be reasonably optimized, based on at
stood how the different sampling stages contribute to the
least a rudimentary analysis of sampling variance. Differ-
CPI variance. For example, consider a two-stage sampling
ent options can be considered:
design, where outlets are sampled first and then varieties
are selected within the sampled outlets. The sampling vari- • Best method would be to multiply each weight by a
ance of the price index can be decomposed into a sum of measure of price change dispersion in the group. Vari-
two terms that are linked to these two stages. The first term ance and cost considerations together call for alloca-
relates to the variance of the price changes between the out- tions where relatively many products are measured per
lets, and the second term relates to the variance of the price outlet and relatively few outlets are contained in the
changes of the varieties that are available within an outlet. sample.
4.93 If the price changes between the outlets are rather • Second-best method would be to set the sample sizes
homogeneous but the price changes of the different varieties approximately proportional to the weights of the prod-
that are sold in an outlet are more diverse, then the preferred uct groups.
75
PRICE COLLECTION AND VALIDATION 5
Introduction scanner data), and prices obtained via web scraping have
to be aggregated in a correct and coherent way taking into
5.1 Several factors will determine the choice of which account the weights of the different channels or sources.
price collection methods a national statistical office (NSO) Ideally, indices per data source should be weighted and
uses, taking into consideration efficiency, accuracy, and prices from different sources should not be mixed; however,
representativity of consumers’ purchasing patterns. For in practice, this will depend on the availability of data to
example, local price collection is costly but can have the develop weights.
advantage of covering a wide range of locations and items, 5.4 The activity of price collection needs to be care-
particularly for food, alcohol, tobacco, and durable goods fully planned and a number of organizational options can be
such as clothing, furniture, and electrical products. On the followed. The focus of this chapter is on traditional forms
other hand, central collection, either at the head office or in of price collection where price surveys are undertaken in
regional offices, can be cheaper and may be used for prod- outlets, using paper forms, handheld computers, or tablets,
ucts where there are national pricing policies, as for rail and supplemented by prices obtained centrally from postal,
fares, or where prices cannot be observed directly in retail email, or telephone inquiry. However, the general principles
outlets, such as for many professional services. With regard described in paragraphs 5.15–5.49 apply to all types of price
to representing consumers’ purchasing patterns, the price collection.
collection method also needs to reflect methods of shopping. 5.5 A separate section details the collection of prices on
For instance, internet purchases or goods purchased through the internet (paragraphs 5.194–5.208). The use of scanner
mail order and catalogs need to be properly reflected in the data is discussed in Chapter 10, considering the conceptual
sample. and practical issues associated with the use of these data.
5.2 Price collection will therefore require different 5.6 The relative advantages of central and local price
practical solutions in different countries according to local collection will depend on individual country circumstances.
circumstances. Thus, the most appropriate sampling and A further option involves the outsourcing of certain elements
survey methods, and the best data sources for price collec- of price collection. In reviewing these options, this chapter
tion will depend on the structure of retailing with regard covers the relationship between the field and the head office,
to the characteristics of outlets, their geographical spread, and the handling of the flow of information.
and the range of goods and services available to purchase,
plus the shopping behavior of the households covered by the Organization Options
index. The compiler should always be guided by the prin-
ciples and objectives of the price index being compiled, as Local Price Collection
addressed in earlier chapters. 5.7 Local price collection involves collectors visiting
5.3 Price collection is becoming increasingly multi- individual outlets to collect prices for several goods and ser-
modal with prices being web scraped from the internet or vices. This is the predominant method of price collection
obtained from scanner data, as well as being collected by in most countries, although more countries are now starting
hand from outlets and by telephone inquiry. Issues of coher- to use other methods. The range and number of outlets vis-
ence can arise when integrating price data from different ited and the types of goods and services priced vary among
sources, with relevant consequences. The internet prices countries.
charged by online retailers can differ from the prices charged 5.8 Although the precise method of local price collec-
by traditional outlets. Similarly, the internet prices obtained tion varies, each price collector is usually responsible for
from a retail chain’s website and the associated sales volume collecting prices from a certain location or from certain
can differ from the corresponding in-store prices. The prices types of outlets. Collectors will visit the same outlets in
collected from each source should be weighted according each collection period to attempt to price the same items.
to the respective values of sales via the different channels Through this type of arrangement, price collectors can build
to ensure the index properly represents the purchases of the up effective relationships with retailers and gain specialist
target population. Moreover, products representative of in- knowledge.
store sales may not be representative of internet sales and 5.9 There are several important criteria relating to the
vice versa. Constructing separate elementary aggregates to conduct of the collection:
represent different outlets or outlet types, and aggregating
the elementary aggregates using explicit weights relating (1) Collectors should be appropriately dressed and be po-
to the respective sales can help to ensure a balanced sam- lite—they are representing the NSO.
ple. For example, prices from a sample of bakeries (using (2) Collectors should carry identification to confirm their
price collectors), prices for bread from supermarkets (using role and status.
77
(3) Collectors should make themselves known to the re- a year on a set date. Internal checking procedures need to be
tailer or store manager when they arrive, and before in place for centrally collected prices, because these prices
they begin collecting prices. can often represent items with a relatively high expenditure
(4) Collectors should comply with any request from the weight. Credibility checking should be carried out to test the
outlet staff if it does not invalidate the sample. For in- reliability of the collected prices, for instance, by comparing
stance, if the store is very busy, the collector may com- whether the price change relative to the previous price looks
ply with a request to return later in the day. reasonable compared with historic evidence while taking
into account sales prices.
(5) The collection should be carried out as quickly as pos-
sible, causing minimal disruption to store business.
The Principles of Price Collection
Central Price Collection 5.12 The procedures for price collection should follow
the general concepts underlying the CPI and should reflect
5.10 Centrally collected prices are obtained from the the expenditure of the target population.
central offices of major retailing chains with national pricing
strategies. Branches of these chains may be excluded from
the local collection if data can be collected more effectively Defining the Price
centrally. Data suppliers may provide information on paper 5.13 Given that the aim of the CPI is to measure price infla-
forms, or by sending spreadsheets to the NSO by email, fax, tion, prices are defined according to what consumers pay in
or on external media such as a flash memory stick. Mail the marketplace (that is, the actual transaction prices of goods
order catalogs can also be treated in the same way. These and services purchased by the consumer, including all taxes).
prices are then combined with those for the same items from However, in practice, the advertised price is usually selected
the local collection. Prices can also be collected centrally because it can be readily observed. Practical difficulties stem
online or by using web scraping or scanner data (see para- from the existence of discounts offered when stock is sold off
graphs 5.194–5.208 and Annex 5.6 and, for scanner data, or when end-of-line, damaged, or special stock is brought in
Chapter 10). at sale time. In these cases, special procedures apply. Another
5.11 Price data for services or fees may be collected cen- complication that can arise is bargaining, where a price might
trally from organizations such as trade associations or national not be displayed or where goods and services are purchased
or local governmental departments. This is often the case for on the black market. Procedures to address these situations
tariffs, for example, for gas and electricity, or for telephone are discussed in paragraphs 5.24–5.29.
and internet charges, where data from administrative sources
on pricing structures and purchasing patterns are required Nontransaction Prices
(see Chapter 11).1 Whenever possible, centrally collected 5.14 There are some exceptions that deviate from the
prices are obtained from one central source, although con- stated aim of measuring actual transaction prices. The most
tact with regional or competing companies is needed if there notable example is the treatment of owner-occupied housing
are local variations. Data may be requested in writing, by services, where alternative conceptual approaches require
telephone, or may be delivered automatically because the different measurement methodologies, and the choice of con-
NSO is on a provider’s mailing list. Providers may send a ceptual treatment depends on the main purpose of the index.
full price list or tariff sheet, from which the relevant prices For some services, the transaction price may be represented
will be extracted by the consumer price index (CPI) staff, or by a tariff or daily or hourly rate. More details on owner-
just the prices of the items specified in the data request. It occupied housing and tariffs are discussed in Chapter 11.
is good practice to confirm price quotations by some form
of written documentation. When collecting prices from the Catalog and List Prices (Not Mail Order) 2
internet, screen prints can be useful as displayed prices can 5.15 Catalog or list prices provided by the supplier to
frequently change (see paragraphs 5.194–5.208 relating to the retail outlet are in many cases identical to the transaction
collecting prices online). Frequency of inquiry can vary price. However, the catalog or list price may only be the
across the range of items depending on when prices are recommended price and not the actual price the item is sold
known or expected to change. The most common frequen- at. Even where it is supposed to be the actual price charged,
cies are monthly, quarterly, and annual, but there are also retailers may not always comply with the prescribed prices.
instances of collecting as frequently as weekly or even daily. Although the use of catalog prices is contrary to the prin-
The frequency depends on price volatility. For regulated ciple of recording transaction prices, in practice it can be a
and fixed prices, collection could occur on the date when cost-effective method of obtaining prices. Therefore, cata-
the change becomes effective. For instance, this may be the log or list prices can be used, but their reliability should be
case where tariffs for gas, electricity, and water change once confirmed periodically.
1
Price Reductions and Related Issues
There are a number of common issues associated with the use of admin- 5.16 Transaction prices may differ from advertised
istrative sources for the use of listed prices in the CPI, including coverage,
definitions, data quality, and resilience of supply. For instance, does the prices, for example, if at time of purchase a discount is
source relate to the definition of the population being measured (transac-
tion prices and sales to private households)? What is the quality of the data
2
source (up to date with no errors)? How resilient is the supply (appropriate The pricing of goods and services ordered by mail can be conducted by
gateways available to access the data in an appropriate and reliable form on reference to mail order catalogs. The index will need to include posting and
a regular and timely basis)? packaging costs.
78
Price Collection and Validation
offered. In practice, however, discriminatory discounts, ever, before designating a price as a “sale” price, spe-
which are available only to a restricted group of households cial care should be taken to ascertain that there is a
(as opposed to nondiscriminatory discounts that are available genuine sale with price reductions on normal stock.
to all), are generally excluded on principle. For example, On occasion, stock is continually sold below the rec-
money-off coupons and loyalty rewards for previous expen- ommended retail price or advertised as a special offer
diture are normally ignored and the nondiscounted price is even though these prices are available all year. In such
recorded. Also, it may be difficult to obtain the price paid if cases, prices should not be considered as sale prices,
it is subject to individual bargaining. The following prac- but should still be collected. Special purchases of
tices are recommended to address different types of price end-of-season, damaged, or defective goods should
reductions. Price collectors should make extensive notes on not normally be priced, as they are likely not of the
the situations confronted so that decisions on how to treat same quality as, or comparable with, goods previously
specific cases can be reviewed and confirmed at the head priced and are unlikely to be available in the future. If
office. In general, if on the day scheduled for price collec- the special offer is limited to the first customers, the
tion, the price of an item is reduced due to sale or promotion, item should not be priced, as the offer is not avail-
the reduced price is collected even if the sale or promotion is able to everyone. Introductory special offers may be
only for one or two days. included if they are available to all. However, given
(1) Discounted prices: should only be included if gener- the need to price the same “basket” each month, such
ally available to everyone with no conditions attached; offers will not be chosen as representative items, un-
otherwise, the nondiscounted or unsubsidized price less they have sufficient volume of sales to be consid-
should be recorded. The general practice is to ignore ered representative and are introduced at the time of
money-off coupons and loyalty rewards. A judgment an update of the basket, or when a replacement item
needs to be made, however, relating to the interpre- needs to be chosen. Discounts on goods close to expiry
tation of “generally available.” For instance, reduced date should be disregarded or treated as specification
prices for payment by direct debit may be collected or quality changes.
depending on the extent to which consumers have ac- (4) Bonus offers, extras, and free gifts: prices for items
cess to, and use, of this service. A judgment is required temporarily bearing extra quantities (for example, 30
on the threshold to be set for access, above which percent extra free) should be adjusted to take account
the discounted price is included in the index. Alter- of the increased quantity unless it can be determined
natively, for tariffs, different payment methods may that the extra quantities involved will not be wanted
all be priced individually (for example, separate data by most consumers, will not have influenced the deci-
collection for payments for electricity by cash, direct sion to purchase, and will not be consumed. Similarly,
debit, and prepayment) and the prices weighted and “2 for the price of 1” offers should be included. The
aggregated to form a single price index for that item. reasoning behind this is that, if the CPI is tracking the
(2) Price discrimination: discounts systematically avail- price of a specific product, say a 330-milliliter can of
able only to a restricted group of households should a diet soda, and the offer is two cans for the price of
be disregarded because they are discriminatory, un- one, the consumer will always take the two cans rep-
less they are significant and are available either to resenting a 50 percent discount on a single can. How-
the vast majority of the population or to identifiable ever, an offer of three cans for the price of two would
subgroups who qualify for such discounts on the basis be disregarded because the offer requires the purchase
of demographic or other characteristics not requiring of two cans and the item specification is for a single
action by the individuals concerned at the time of pur- can. Money-off coupons for future purchases are usu-
chase. If included, they should be treated as stratifica- ally disregarded as these may not be used or wanted.
tion or coverage issues in item sampling. A judgment Free items with other purchases (for example, a free
is required from the compiler. Examples of price dis- gift with every product purchased) are generally dis-
crimination generally included in the CPI may include regarded. For example, free gifts such as plastic toys
lower prices offered to senior citizens (for example, in cereal boxes should be ignored because they are not
discounted travel or haircuts) and discounts for people included in the list for price observations; it is the price
who receive state benefits. An example where prices to be paid for the cereal in the box that is relevant.
are not universally available and judgment is required Also, such free gifts can be difficult to value. Simi-
is when a nominal or token membership fee is re- larly, receiving a free toothbrush when buying a bottle
quired by the retail outlet. In these cases, the take-up of mouthwash should be disregarded because it would
of such membership, which is widely available, needs be difficult to value the free gift and the consumer
to be considered with regard to thresholds and general may only want the mouthwash and not the toothbrush.
spending patterns of the consumers and the conditions The treatment of bonus offers, extras, and free gifts is
placed on membership that may make it restrictive subject to interpretation on a case-by-case basis. Col-
(for example, minimum levels of purchase). See para- lectors should be aware that temporary “special offer”
graphs 5.24–5.29 for a discussion on price bargaining. weight changes (X percent extra free) could become a
(3) Sale or special offer prices: should be recorded if they permanent weight change (for example, cans of alco-
are temporary reductions on items likely to be avail- holic drinks changing size from 440 milliliters to 500
able again at normal prices or are stock-clearing sales milliliters) and should feed the information back to the
(for example, January sales or summer sales). How- head office as they become aware of it. In this way,
79
CPI staff can issue new or amended guidance to price apply to the purchase of an individual product and are
collectors about item specifications. granted within a time period, and are expected to have
(5) Stamps: sometimes purchasers are given special a significant influence on the quantities purchasers are
stamps, which can be accumulated and subsequently willing to buy. Loyalty rebates or coupons associated
exchanged for goods and services. If a discount is with previous expenditure at the outlet, to be used for
available as an alternative to such stamps, then the similar or other purchases, should be disregarded as
discounted price should be recorded. Otherwise, the they are conceptually out of scope because they are
stamps should be disregarded. discriminatory relying on previous purchases. One-off
rebates should be disregarded as they do not relate to
(6) Trade-ins: in general, the price reduction obtained
the specific time period of the consumption and are un-
by trading in an old item (for example, a car) com-
likely to affect levels of consumption. They are viewed
pared with the nominal full price should be ignored.
more as a source of additional income.
This treatment follows convention, as the transaction
essentially relates to a second-hand good and only the (11) Credit card and other payment arrangements involv-
service charge levied by the outlet in buying and selling interest, service charges, or extra charges: charges
ing the good comes under the scope of the index. In incurred as a consequence of failing to pay within a
practice, however, the situation is not so clear-cut. For specified period of time from the purchase should be
instance, a garage may give a discount that is greater disregarded. For example, zero or positive interest
than the retail value of the traded-in car and, therefore, loans granted to finance a purchase should be disre-
in effect gives a genuine discount on the new car. In garded when determining the price. These charges
many cases, discounts from trade-ins are difficult to come under financial services.
evaluate. The trade-in value may be negotiable in each (12) Cash back: whether to include a cash back offer de-
case, and the full nominal price—which is used as the pends on the circumstances. Some large retailers offer
benchmark against which the discount is measured— credit card accounts to their customers. In this case, a
may not be known. It is therefore generally best to cash back offer could be considered a form of discount.
report the list price or asking price. Each time consumers use the retailer’s credit card, they
(7) Sales taxes: when a tax is not included in the price of could earn a percentage of the total amount paid in the
individual items in an outlet, but is added when the form of cash back. For example, if the card pays 2 per-
customer pays for the item, care must be taken to cent cash back and the customer spends £100 in the
record the price including tax. To make sure of this, retail outlet, they will earn £2. Most cash back cards
with items for which the price is normally quoted pre- credit the amount earned by the customer onto their
tax, and in areas where a general sales tax is added statement, thereby reducing their credit card bill. Like
to the bill, the price collection forms should require loyalty cards, data collectors should collect additional
the collector to indicate whether or not the price information to determine what proportion of consum-
recorded includes the tax, as a price check, so that it ers use the retailer’s credit card to judge whether the
can be added where necessary. “discount” provided by the cash back offer is widely
available or should be considered discriminatory. If
(8) Tips for services: if a compulsory service charge is in- deemed widely available, the CPI price should reflect
cluded, for example, on a restaurant bill, only the com- the net purchase price (price less 2 percent cash back).
pulsory amount should be included in the price, but not If discriminatory, these offers should be excluded. In
any additional discretionary tips. For services that are free other cases, cash back offers are provided by the bank
in principle, but that in practice can rarely be obtained that issues a credit card. The cash back offer provides
without a tip, or where tipping at a standard rate is the an incentive for consumers to use a particular credit
common practice, such tips may be added to the specified card for purchases but does not represent a discounted
price. However, there is no agreed-upon convention. price offered by the retailer and should be excluded.
(9) Regular rebates or refunds: should only be consid- Cash back offers can also be tied to loyalty cards. A
ered when attributable to the purchase of an individual percentage of the total amount spent accrues and can
identifiable product and granted within a time period be used as a discount on future purchases. These types
from the actual purchase, and are expected to have a of cash back offers are excluded because they apply to
significant influence on the quantity buyers purchase. future purchases and the price paid is not discounted in
For example, money-back deposits on bottles should any way.
be deducted from the price if these are a sufficient in- (13) Dual pricing for cash and for credit or debit card pur-
centive for returning the bottle, whereas money-back chases: some outlets may sell goods at different prices
offers on lawn mowers after a five-year period should depending on whether the item is paid for in cash or
be disregarded. In all cases, a consistent decision for if a bank debit or credit card is used. The primary
each item must be applied over time. Decisions about objective of the price collection is to achieve a rep-
the treatment of rebates are made on an individual resentative sample and continuity. The price collector
basis. They may reflect income rather than expenditure should determine the proportions of buyers who pay
changes and may require different treatment from that by the different methods and this should be used to
used by national accounts. judge what prices should be collected to ensure rep-
(10) Irregular rebates or refunds: as with regular rebates resentativeness. If the price collected on a particular
or refunds, these should only be considered when they month is based on cash payment, then a price based on
80
cash payment must be obtained in each of the follow- Unavoidable Costs That Are Not Part of
ing months. If a price collected on a particular month the Advertised Price
is based on the use of a bank debit or credit card, then
5.18 In some cases, the consumer has no choice but to
the prices collected in the following months should be
pay an extra cost, in addition to the advertised price of the
based on the use of a bank debit or credit card. Widely
product, to obtain the use of the product. In cases where
available reductions for cash payments may be in- most customers will treat and pay this cost as part of the
cluded, but care should be taken to ensure represen- purchase price, the extra cost should be added to the adver-
tative purchases are measured, treatment is consistent tised price to determine the price for CPI purposes.
from one period to the next, and practice adheres to the 5.19 A common example of this situation is the sale of
fixed-basket concept. Thus, where relevant, item de- large household appliances and furniture. Most consumers
scriptions should include method of payment and the would not be able to provide their own transport to carry
proportions that are paid by different methods should these products home. As a result, many stores arrange deliv-
be kept constant between reviews of the CPI basket. ery to the customer’s home for an additional cost. It can be
(14) Bank-specific credit card offers: in some countries, argued that most customers must pay for the delivery ser-
banks that issue credit cards will negotiate with retailers vice, and this additional charge should be included as part
to offer customers who use their credit cards a discount of the purchase price for CPI purposes. The concept still
on all purchases. For example, a bank can negotiate applies when the transport is provided by another business,
with a supermarket to offer all customers using their if this is part of the transaction. In this case, the transport
credit card a discount off the total purchase price. Judg- cost would be priced separately but added to the cost of the
ments need to be made whether such offers can be de- appliance when the index is being compiled. However, if
fined as generally available. Like discounts associated the customer arranges for the transportation under a separate
with loyalty cards, if the majority of consumers use that contract then the transportation should be priced separately
bank-issued credit card, it could be considered gener- under transport services.
ally available and the discount reflected in the collected 5.20 A similar situation can occur in marketplaces
prices. However, if it is judged to be discriminatory and when live poultry is bought for meat. If consumers regularly
not generally available, they should be ignored. buy live fowls and then go to other stalls to have the fowls
(15) Foreign currency denominated prices: the price col- slaughtered, then the two purchases can be combined to cal-
lector should collect prices of items quoted in foreign culate the cost of buying poultry meat.
currency from outlets selling goods and services in for-
eign currency, if their exclusion from the CPI basket Price Bargaining3
would undermine the representativeness of the index. 5.21 Bargaining relates to a situation where prices are
The price should then be converted into local currency individually negotiated between sellers and purchasers and
using an average exchange rate obtained by the head are not predetermined. In some marketplaces, the prices for
office from the central bank or appropriate dealers in a wide range of daily necessities are negotiated. Final trans-
the foreign-exchange market. The price in local cur- action prices and quantities will vary from one transaction
rency (obtained after conversion) should be used in the to another and cannot be determined until the purchase has
CPI. been made. Similarly, there will be variations between trans-
(16) Product with a local and foreign currency price: in actions in the quality of the goods being purchased which
some countries, there are outlets that sell products will impact on the price that is paid. For example, the bar-
in local and foreign currency (that is, a purchaser is gained price for fresh fruit is likely to be lower the longer the
given an option to pay for an item in local or foreign fruit is on display. Clearly, these special conditions require
currency). The local currency price is used when pur- special methods to determine purchasers’ prices for inclu-
chases are made in local currency. However, if more sion in the CPI.
purchases are made in foreign currency, the foreign 5.22 Where prices are determined by bargaining, stan-
currency price is converted to local currency as in the dard price survey methods, which consist of collecting
previous example. To determine whether more pur- prices directly from sellers, can generate erratic price indi-
chases are made in local or in foreign currency, the ces that do not reflect actual price movements in a market.
enumerator must ask the outlet staff which currency The prices that are collected should be “bargained prices”
is being used more frequently for purchasing the item, and will depend on the ability, willingness, and power of
and this currency should be used for pricing purposes, price collectors to bargain in the same way as genuine pur-
unless purchases are frequently made in both curren- chasers. In addition, prices can vary during the day and from
cies, in which case the price in both currencies is col-
lected and a weighted average used after conversion to 3
It can be argued from the viewpoint of the System of National Accounts
the local currency. that bargaining is a form of price discrimination. A purchaser is not free to
choose the purchase price because the seller can charge different prices for
5.17 The appropriate practice to be followed will be identical goods and services sold under the same circumstances. It follows
determined by individual circumstances, which might vary that “identical” products sold at different prices should be recognized as
among different countries. Price collectors should make having the same quality, and their prices must be averaged to obtain a single
price relative to calculate price indices. In practice, the variation in transac-
extensive notes on the situations confronted so that deci- tion price can rarely be associated with identifiable price-related categories
sions on how to treat specific cases can be quality assured of customers. Rather, purchasers may inadvertently buy at a higher price
and reviewed at the head office. than may be found elsewhere or could have been negotiated.
81
one day to the next, adding an extra dimension to the con- be used to check the number of quotes obtained
cept of representativeness. A number of survey methods and against the target number set by the head office.
price collection techniques have been developed to over- (3) Survey of trends in wholesale prices. A limited parallel
come the difficulties inherent in measuring prices that have collection of wholesale prices can be a useful supple-
been bargained. ment for items where the information obtained from the
(1) Survey by purchase of products, simulating a con- previous survey techniques is only partially successful,
sumer. The principle is that price collection should for example, where there is a deficit in the number of ob-
be carried out in conditions that simulate as closely servations obtained. Ideally, prices should be obtained
as possible situations in which real transactions take from the wholesalers where the relevant retailers get
place. Price collectors behave like regular consumers their goods. All factors should be observed that might
by actually purchasing items to be priced and spread- result in increases in the corresponding retail prices,
ing their purchases over the day to ensure representa- such as changes in taxes on retail activities, license fees,
tiveness. In each case, the field manager will need to and the rental for the market stall. On the assumption
carry out regular checks on quantities and prices ob- that these factors remain constant over time, the evolu-
tained by collectors. The following approaches may be tion of wholesale prices may be used as a proxy for the
taken: retail prices of relevant items. The price of an item for
(1.1) Price collectors buy items to determine the rel- the current period would be estimated by multiplying
evant price through bargaining. They should be the price of the previous period by the corresponding
trained to behave as normal purchasers and strive evolution in wholesale price.
to get the lowest possible price from selected out- 5.23 Determination of the prices paid by a purchaser
lets and sellers. Given the high turnover of sellers can be challenging where the final price is for a bundle
in many informal markets, the sample of sellers of items, for example, where a stall holder gives the pur-
should be partially renewed on a regular basis to chaser extra quantities as a bonus for buying a number
ensure that it remains representative. of goods. If the bonus comprises several categories of
(1.2) Price collectors buy items and, in addition, are items, including the item on which a transaction price
given an incentive to get the best price. For ex- was being directly negotiated, then the purchase has to
ample, a price ceiling may be set, and the col- be split into as many subtransactions as item catego-
lector may receive a proportion of the difference ries. In these cases, a common-sense approach is needed
between the ceiling and the bargained price. This regarding treatment in the CPI. For example, consum-
incentive system guards against potential difficul- ers who are living on a subsistence income consume all
ties caused by the collectors not getting the low- food purchases and the extra quantities involved will
est price because, unlike an ordinary customer, be consumed and should be included when calculating
they are not concerned with maximizing value the price paid. Purchasers will have actively bartered an
for money and are not constrained by income al- overall price for the total basket of purchases, including
though there might be a limit on the money avail- any “free” goods thrown in.
able for purchasing. 5.24 The method for determining the price paid by the
(2) Survey of purchasers. The prices purchasers have paid purchaser is illustrated in the following example: a pur-
are collected throughout the day immediately after the chaser wants to buy five kilograms of carrots and is offered
purchaser leaves the outlet or market stall, together a bonus consisting of 500 grams of carrots, a head of lettuce,
with a record of the quantity and quality of the prod- and three onions.
uct purchased. The extent of the haggling or bargaining 5.25 Three transactions can be identified, involving: 5.5 kilo-
should be determined (for example, opening and clos- grams of carrots, 100 grams of lettuce, and 200 grams of
ing prices) together with an indication of the relevant onions. The bonus should be valued at prices at which the
parameters determining the price. A form of incentive seller would have sold the items and the purchaser would
payment for survey participation may be needed where have bought them. The assumption made is that prices
there is reluctance among purchasers to submit to such would have been determined through bargaining on the same
time-consuming questions. Determining the quantity conditions as the price of the item that initiated the pur-
and quality of the items purchased can be difficult. chase (carrots). In this example, the opening value of five
(2.1) For the survey by purchase of products and the kilograms of carrots is 15,000 pesos and the closing value
survey of purchasers, all CPI basket items sub- 12,000 pesos, whereas the opening values of other food
ject to bargaining should be covered. The num- products included in the bonus are 990 pesos for a bunch
ber of prices collected needs to be sufficient to of 264 grams of lettuce and 4,620 pesos for a pile of onions
cover all relevant items and to provide a reliable of 4.4 kilograms. The actual closing price of carrots will be
guide to average price. This may be difficult to determined as shown in Table 5.1. The actual purchaser’s
determine beforehand, although previous price price of carrots is found to be 2.0967 pesos per gram or
collections should provide some guidance. It is 2,096.7 pesos per kilo.
suggested that price collectors engaging in a sur- 5.26 If the price collector does not know the closing
vey of purchasers are given a form on which to price at which lettuce and onions would have been sold by
record the number of quotations per stall or shop, the seller of the carrots, it can be estimated. This is done
as indicated by the various respondents. This can by collecting opening values and standard quantities from a
82
Table 5.1 Determining Purchaser Price When Bargaining Takes Place

Requested Item Bonus Items
Carrots Carrots Lettuce Onions
(a) Opening Value of Standard Quantities (local currency units) 15,000 15,000 990 4,620
(b) Standard Quantities (grams) 5,000 5,000 264 4,400
(c) Opening Unit Price of Standard Quantities (local currency 3.00 3.00 3.75 1.05
units per gram) (a/b)
(d) Opening Unit Price of Quantities and Bonus Quantities (local 3.00 3.00 3.75 1.05
currency units per gram) (as per (c))
(e) Quantities and Bonus Quantities Offered (grams) 5,000 500 100 200
(f) Opening Value of Quantities and Bonus Quantities (local 15,000 1,500 375 210
currency units) ((e) at unit price at (d))
(g) Closing Value of Items Received (local currency units) 12,000 1,200 300 168
(assumes bargaining ratio for lettuce and onions same as for carrots,
that is, 15,000/12,000 = 1.25)
(h) New Price (local currency units per gram) (g/e) 2.40 2.40 3.00 0.84
(i) Bargaining Ratio (D/H) (for example, old price for carrots/ 1.25 1.25 1.25 1.25
new price for carrots, that is, 15,000/12,000 = 1.25)
(j) Payment (local currency units) 12,000
(k) Estimated Closing Value of Bonus (local currency units) 1,668
(1,200 + 300 + 168 = 1,668)
(l) Actual Value of Requested Item (all carrots) (local currency units) 10,332
(12,000 − 1,668 = 10,332)
(m) Quantity Received of Requested Item (grams) 5,500
(n) Actual Purchaser’s Unit Price of Requested Item (local 2.10
currency units per gram) ((12,000 − 300 − 168)/5,500) = 2.0967)
(o) Improved Bargaining Ratio (d /n), that is, 3/2.10 = 1.43 1.43
sample of sellers in the same market or at different outlets Item Specifications

in the same area. The average opening price of an item is 5.28 The recording of item specifications is particularly
equal to the sum of opening values of the item divided by relevant for traditional forms of price collection where prod-
the sum of relevant standard quantities. For each bonus item ucts are sampled and where price collection involves price
(lettuce and onions), the resulting average opening price will collectors visiting retail outlets to collect prices or where
be divided by the bargaining ratio calculated on the item prices are collected by postal inquiry, email, or telephone
needed (carrots) to estimate a closing price for that bonus most particularly from central sources.
item. The value of each bonus item is obtained by multiply- 5.29 There are no firm rules, especially regarding the
ing the closing price by the quantity offered. If the packet use of loose or tight item specifications: each NSO may
of bonus items contains an item of the same quality as the choose their preferred methods. Considerations in deciding
requested item, that bonus item will be valued based on the on item specifications are the following:
closing value of the requested item.
(1) Loose specifications leave more discretion to the price
The Principle of the Fixed Basket collector, so the reliability and training of collectors are
factors to consider when deciding whether to use loose
5.27 The important principle underlying price collec- or tight specifications.
tion is the need to compare prices on a like-to-like basis
from one period to the next. This has consequences: (2) The specification should be more detailed for heteroge-
neous items where there is scope for significant differ-
(1) Where there is a choice of variety of product to be priced ence of varieties, and for items that by nature are highly
initially, an important consideration should be whether specified, such as cars and hi-tech goods.
that variety will be available to price over a reasonably (3) Loose specifications allow the index to more broadly reflect
long period. This is in addition to being typical of what regional differences in tastes and preferences, and account
is sold to customers. Note that tight specifications are for differences resulting from socioeconomic factors.
of no use if the described item cannot be found in the
(4) Tight specifications allow for the calculation of mean-
outlets.
ingful average prices.
(2) A record should be kept of additional information
(5) Average prices are useful to identify outliers and to assess
needed to ensure the unique identification of the variety
the accuracy of reported prices.
priced so that:
(6) Subject to satisfactory sample design, average prices
(2.1) The same variety continues to be priced in the allow comparisons of price levels, including between
case of subsequent price collection being carried regions or urban and rural areas. Tight specifications
out by a different person. facilitate the use of CPI prices in the computation of
(2.2) The identification and adjustment for any quality purchasing power parities.
change can be made when the variety disappears (7) However, specifications should not be so detailed that
and is replaced by a different one. the item cannot be found in an adequate number of out-
83
lets. Individual varieties sold differ from outlet to outlet Period-of-Time or Point-in-Time
and region to region, and the CPI should reflect these Price Collection
differences. Overly tight specifications do not reflect
5.33 Prices for a monthly (quarterly) CPI should aim
this diversity and can increase the number of missing
to reflect the average price of the reference period. To that
prices.
end, NSOs should strive to calculate an index based on
(8) Without tight specifications, average prices have little prices covering the whole period (for example, a month).
meaning or use. For example, an average price for a liter Ideally, price collection would be organized so that prices
of whole milk is more meaningful than an average price are collected from different outlets throughout the month.
for men’s shirts. This ensures that the prices used to calculate the index more
5.30 Responsibility for deciding whether to use loose or broadly reflect the average price for the reference month.
tight specifications and for specifying the items to be priced While resource constraints could restrict price collection to
should normally rest with the head office. In general, loose a specific period during the month, effort should be made
specifications can be useful for food, beverages, clothing, to maximize the price collection period to include as many
and personal items. Tight specifications tend to be useful for days in the month as possible. Regardless of the price col-
electronic and other items with high rates of turnover. Speci- lection period, the interval between successive price obser-
fications should be reviewed on a regular basis to determine vations must be held constant, for example, by visiting an
whether they continue to be relevant. The following could outlet during the same fixed time period each month or
signal the need to review and revise specifications: quarter. If the CPI is used to deflate income, expenditure,
or sales, the index should relate to the time period of these
• A large number of missing price quotations
money flows. For economic analysis, where the index will
• A large number of substitutions be used in conjunction with other economic statistics, most
• A wide variation in the distribution of collected price of which relate to a period rather than to a point in time, the
levels CPI should do the same.
• An observed distortion in the achieved sample, for exam- 5.34 Under the traditional approaches to price measure-
ple, where an item is stocked only by one retail chain indi- ment, where price collectors visit retail outlets and record
cating that the item is unlikely to be fully representative prices, spreading price collection over a period of time will
of what households buy, or where there is a reduction in result in a more even workload. This avoids some of the
the number of brands available possibly indicating a fall operational problems and inefficiencies associated with
in demand point-in-time price collection but will involve collecting
sufficient price quotations to obtain a reliable average price
5.31 Practical difficulties stem from the temporal level for each product over the period.
dimension in price collection. Issues such as the timing 5.35 From an operational perspective, the uneven work-
of entering prices into the index, the treatment of missing loads associated with point-in-time price collection when
prices, seasonality, and the frequency of price collection using traditional price collection techniques can be inefficient
are temporal issues that need to be addressed to maintain and have a negative impact on price collector performance at
the quality of the index. Aspects relating to the frequency peak times. For instance, a sizeable price collection team will
and timing of price collection are addressed in paragraphs be required for a short time in each collection period. This
5.35–5.47. has implications for the recruitment and deployment of price
collectors and for management of the fieldwork. Communi-
Frequency cation and planning between collectors and their supervisors
5.32 Decisions about the frequency of price collection to manage events like absence because of illness must be
are governed by several factors. The most important are: the prompt and effective. Staff at the head office will similarly
volatility of prices; the characteristics of the market from be confronted with a heavy workload of data checking and
which the prices are collected; the known regularity of price editing of price data over a short period of time.
changes; and the frequency and method of calculation of the 5.36 With the point-in-time approach, major price
CPI. The general principle is that each item should be priced as setters, notably the government, can influence the index
often as is necessary to ensure that the index reflects a reliable according to whether their price changes take effect on a
and meaningful measure of price change. Thus, the desired day just before or just after the day for which their price
frequency of price collection will vary by product, depend- information is obtained, or on the day of collection. Since
ing on how frequently the prices to be observed change. For prices are often collected centrally from such price setters,
instance, the prices of some products (for example, fees for it should be possible to obtain information from them about
government services and utilities) might need to be collected the amount and timing of price changes at the end of each
only once a year if it is known that prices are reviewed annu- month, so that in applying the period-of-time approach, an
ally at a regular point in time. However, it is advisable for average price for the whole month can be calculated. For
the price collector to occasionally check that the assumption example, if electricity charges are made quarterly and prices
of regular and predictable reviews still holds. In contrast, the increase partway through the three-month period, individ-
prices of products with more volatile prices, such as fresh ual customers’ payments could include zero, one, two, or
food, might have to be collected more frequently than the fre- three months at the higher rate depending on when the price
quency of index calculation and publication. Also, price col- change is implemented.
lection will need to reflect user needs with regard to whether 5.37 The point in time or period is chosen to represent a
the target index is a point-in-time or period-of-time index. prespecified reference period, normally the calendar month
84
in which the price observations take place. Whether col- meaningful measure of price change for the month. The cal-
lection is continuous or point-in-time, the interval between culation of the average monthly price should relate to the
successive price observations must be held constant, for periodicity of the collection, taking into account the appro-
example, by visiting an outlet during the same fixed time priate pricing periods. For example, if prices are increased
period each month or quarter. a third of the way through the period, then two-thirds of the
5.38 The sampling variance will differ according to average price over the month should reflect the higher pric-
whether a period or point-in-time index is compiled and the ing. In these cases, different areas or locations should be
frequency of collection. Regarding the timing and frequency scheduled for price collection at different times of the month
of price collection, the CPI compiler should consider the according to a regular pattern to be repeated each month.
trade-off between statistical accuracy and cost, particularly This makes the use of the collectors’ time more efficient and
when prices are collected by visits to local outlets, as this has the advantage of providing a spread of collection dates
can be a costly activity. The budget for price collection can for many representative items. Individual price observations
limit the available options. should be carried out at the same time each month so that
5.39 The timing of the publication of the resulting price the index does not move because of a change in the length of
indices can be a constraint on the price collection sched- interval between collection dates. Prices may vary by day of
ule and vice versa. For example, there may be legal con- the week (for example, depending when is market day) or by
straints on the timing of the publication of indices, such as time of the day (for example, fish is more expensive in the
a requirement to publish at a set time each month. In such morning when it is fresher), and prices should be collected
cases, prices must be collected to a schedule that allows on the same day, at the same time, if this is the case.
quality assurance, processing, and aggregation procedures 5.43 Preferably, days of the week and month should be
to be completed before the deadline. The quality manage- chosen taking into account when purchases are concentrated
ment of the price collection process is covered in paragraphs and where prices and goods in stock are known to be repre-
5.78–5.116. sentative of the whole month. In some countries, the results
of the household budget survey suggest that most house-
holds do the shopping on the day of the market or souk.
Timing of Price Collection However, retailers may be less prepared to cooperate when
5.40 The interval between price observations should be they are busy, so a balance needs to be struck between the
uniform for each outlet. Since the length of the month varies, ideal timing for collection and the impact on response rates.4
this uniformity must be defined carefully, for instance, not by An entirely fixed interval is impossible because of the vary-
date but by a formula such as “second Monday in the month.” ing length of a month and the timing of public or religious
5.41 Regular timing is important, particularly when holidays. One solution is to take sequences of four and five
inflation is high. Where there is a specific collection day, weeks, so maintaining a relatively stable monthly or quar-
the most volatile prices should be collected on that specific terly period where price collection takes place on a fixed day
day rather than on the days around it. Items where prices or days each month; another is to follow a rule such as col-
can be more volatile include fresh fruit and vegetables, fresh lecting prices on the regular market day or on a Wednesday
meat and fish, and fuels. In the case of food products sold in through Friday of the first full week in the month.
marketplaces, the time of day as well as the day of the week 5.44 Price collection days (and sometimes the time
matters. Prices of fresh fruits, vegetables, meat, and fish can of collection) need to be set in advance. The NSO should
be higher in the mornings when produce is fresh and lower explain the procedures used for setting collection dates
in the evenings, especially if there is limited refrigeration. and the underlying objectivity of the method to assure the
Thus, the precise timing of price collection is particularly public of the integrity of the index. Any data suppliers who
important for fresh produce. supply prices directly to head office staff need to know the
5.42 When inflation is low and stable there will be little collection date in advance to be able to prepare and supply
difference between the annual inflation rates based on point- the necessary price. To enhance transparency, it is recom-
in-time and period-of-time collections. For example, there is mended practice to include the price collection period in the
likely to be very little difference between the annual rate of CPI metadata.
change in the index from Monday, January 8, 2018 to Mon-
day, January 7, 2019 and the corresponding annual rate of Measuring Hyperinflation or Selected
change between the complete months of January 2018 and Large Price Changes
January 2019. This will not be the case if inflation is high or
the inflation rate changes significantly during the year. The 5.45 Special arrangements need to be put into place in
difference between January 1 and February 1 and between case of hyperinflation. Hyperinflation is an economic phe-
average January and average February inflation rates may nomenon that occurs when inflation increases very rapidly.
be different, particularly if so-called “sale” periods fall on Economists typically consider monthly inflation rates of
regular dates each year or are limited by laws. For certain above 50 percent as hyperinflation episodes. In these cir-
products with high index weights, where price changes are cumstances, it becomes even more important that the prices
sudden and tend to affect the whole market on about the
same day, the choice between point-in-time or period-of- 4
time price collection is especially influential. Examples are There is no rule of thumb on what constitutes an acceptable response rate
as this will depend on the sample design that has been adopted and the
fuels, electricity, and telecommunication prices. For these structure of the retail sector (most particularly, the variability of product
kinds of products, it can be argued that the case for taking an and item varieties available and their prices, and the range of outlets in
average price for the period is stronger. This ensures a more which they are stocked).
85
of individual items in individual outlets are collected at pre- of price collectors in the concepts and practices of price
cisely the same time each month, otherwise misleading fig- collection is an essential element in achieving a representa-
ures may result. The NSO should consider collecting prices tive and error-free sample of prices. Price collectors should
more frequently and, correspondingly, a more frequent com- be given help and practical support when undertaking price
pilation of the index. For example, where prices are nor- collection. The collection of prices should be supported by
mally collected on a quarterly basis, it may be sensible to good documentation. Also, the scheduling of price collec-
collect them more frequently. If this is not feasible, it may tion and the drawing up of contingency plans, for example,
be appropriate to adjust prices proportionately by some rel- in the case of price collector illness or other factors outside
evant indicator, such as a subset of the index, to provide an the control of the price collection team should be in place
approximation to a monthly index. An appropriate com- so that the price collection operation runs smoothly. Effec-
parator must be chosen because relative prices can change tive management of the price collection process should be
dramatically in periods of hyperinflation. The same consid- given the necessary attention and funding. More detailed
erations apply to collecting prices online. guidance on good practices is given in the paragraphs that
5.46 In some circumstances, rapid or frequent price follow.
changes may be associated with certain items only and action 5.52 The collection schedule should include time for
should be taken accordingly. For example, food prices may the price collector to travel around all required locations
rise disproportionately because of a bad harvest and it may within a reasonable number of working hours in a day. The
be sensible to increase the frequency of the index for food schedule needs to allow the price collector to perform all
items only, possibly publishing a separate index. Alterna- the necessary checking of prices, to answer queries from
tively, a simpler way of dealing with this situation may be the supervisor or the head office, and where necessary to
to monitor a small number of relevant prices on a regular revisit the outlet. The collection schedule also needs to
basis without producing a full price index. Such subindices allow for the transfer of data and forms between the data
can be published separately or used to adjust the index as collectors, regional offices (if applicable), and the head
mentioned previously, or to provide background briefing for office. If paper forms are being posted or hand-delivered,
analytical purposes. These items may be chosen according time must be allowed to ensure that all information reaches
to their importance for the household budget and whether its destination by the required deadline. If electronic trans-
they are particularly susceptible to big price increases. mission of data and forms is used, time needs to be allowed
to ensure that all the data arrives in the correct format and
to address potential problems if the files are corrupted.
The Practical Aspects of Managing Plans for recovery in case of technical breakdowns need
Price Collection to be in place.
5.47 As mentioned in the introduction, the most appro- 5.53 Public holidays can occur on days that would oth-
priate sampling and survey procedures will vary depending erwise be price collection days. In general, prices due to be
on the use of the CPI and on national circumstances. This also collected on the holiday should be collected as close as prac-
applies to the practical aspects of price collection. Paragraphs ticable to that day. This usually means adjusting the regular
5.51–5.116 describe aspects of planning and organization and price collection schedule to the immediately preceding or
provide indicative guidance on the processes and procedures following few days. Any adjustments to the regular schedule
that contribute to successful price collection in the field. should be made so that prices reflect the normal pattern of
buyer–seller behavior. The work schedule for index compi-
lation and publication also needs to be considered.
Practical Procedures for Local Price
Collection: Planning and Organization Dealing with Queries: Inquiry
5.48 The discussion that follows is based on period- Management System
of-time pricing rather than point-in-time pricing (see para- 5.54 Price collection queries need to be dealt with in a
graphs 5.36–5.42), but the concepts discussed generally timely and efficient manner both because of the tight sched-
apply to both collection methods. ules associated with the compilation of a CPI and because
5.49 The procedures governing the collection of prices of the difficulty to correct errors in pricing retrospectively
have the requirements of obtaining usable prices from the given the dynamic nature of retailing where prices and stock
outlets, and the practical problems faced in managing travel can change very quickly without warning. The progress
to and from the various locations, in transferring data, and in dealing with queries (for example, verify questionable
in validating the data back in the office. The overall opera- prices or seek additional details on incomplete specifica-
tion can only be achieved by cooperation between the price tions) needs to be monitored and the system for monitor-
collectors, their supervisors, and, of course, the retailers ing must be simple, in order to be effective and flexible to
selected for the price survey. the needs of the CPI production cycle and to any problems
5.50 An overview of local price collection is given in that may arise. An inquiry management system should be
Annex 5.1. It is in the form of a flow chart and shows the able to monitor progress and provide an effective audit trail.
different situations that confront the price collector and the Information recorded should include: date price collection
decisions or referrals to the head office for further action. form received and from whom; date due; collection date/
trip/period (if applicable); current progress (with date);
Planning the Collection Schedule dates when queries were raised and responses sent; and
5.51 Price collection is a complex process that needs to date completed. It is particularly important that decisions
be effectively managed by the head office. Proper training are recorded and signed off to provide price collectors with
86
confirmation for their own purposes of verification and as previous period price, data collectors are prompted to justify
audit trails for quality assurance purposes at the head office. or explain the price change. Price collectors are encouraged
Audit trails are an effective way to ensure that processes to add additional information to the “standard” description
were carried out correctly and to support the review of the to facilitate the unique identification of the product to be
effectiveness of those processes. The information gathered priced without a reliance on price. Such information might
should also feed into a quality management system for the include: brand, make, size, model name and number, refer-
computation of the CPI (see Chapter 13). ence number, distinctive features, and location in outlet (for
5.55 A simple inquiry management system might use example, “bottom shelf at rear of shop”).
emails to send the query. As these will be seen and read, they 5.61 Items should appear on the form in the same order
satisfy the notification requirement. A template form requir- in which prices are collected. A correctly ordered list will
ing the entry of simple ticks and dates could be created in reduce time-consuming swapping and searching between
a spreadsheet and a new copy provided for every collector pages. Even with a handheld computer or tablet, search-
each collection period. ing and pressing of navigation buttons will be less time-
5.56 An inquiry management system can be either consuming if the item list is ordered to match the layout of
paper or electronic. The two methods should be matched the outlet.
to the resources and infrastructure available. For instance, 5.62 For fruit and vegetables, weight and quantity
messages to and from price collectors to their supervisor should be part of the item description, but the amount actu-
might be on paper but messages between a regional office ally priced should be recorded separately so that a unit price
and the head office could be by email. can be calculated. Weight and quantity are needed because
5.57 Queries about collected prices are a two-way pro- the unit price is often lower if larger quantities are pur-
cess. The head office may want to question some prices col- chased. For example, if the price collector prices a bunch of
lected and request a check, while price collectors may ask bananas and the item description is one kilo of bananas, the
for further guidance on situations arising in the field such price collector should choose a bunch weighing approxi-
as the selection of a suitable replacement for a disappearing mately one kilo, weigh the bunch, and calculate the price
good. per kilo.
5.63 When an index basket is updated, the question-
Practical Collection Procedures and naire will need to list all items included in the old and the
Questionnaire Design new basket. Where an update of the outlet sample is planned
to coincide with the introduction of a new basket, it will be
Design of Price Collection Questionnaire necessary in the chain link period for the price collection to
5.58 Good design of the questionnaire form (or its elec- cover both the old sample of locations, outlets, and items
tronic equivalent) is essential for the successful collection of and the new sample of locations, outlets, and items.
prices (see Annex 5.2). Price collectors should be given the 5.64 An example of a typical price collection form is
appropriate direction and find it easy to use, and the format given in Annex 5.2. This example is of a form used by the
and layout should facilitate the extraction of data (for exam- collector for recording prices when visiting an outlet, either
ple, price, item description, or comments) by the head office in paper form or electronic. It is also possible to ask the outlet
for effective quality assurance. The price collection form staff concerned to complete the form themselves and to send
should cover: collection date and collector’s name; name or it to the NSO, although there are obvious potential prob-
location of the particular outlet; product name and specifica- lems in the form not being completed correctly. Such a form
tion of the actual item to be priced; price entry; and collector may therefore serve for reporting as well as for collecting. If
comments about the product or price movement or changes the form has space for recording prices over a sequence of
in the representative item being priced such as package size. months, the collector may keep the form and transcribe the
5.59 Whether recording price information obtained prices each month onto a separate report form, which is then
from personal visits on paper form or via handheld com- sent to the NSO. Where the form used for collection is also
puter or tablet, the price collector needs to be provided with used for reporting, there are two main options: either the
all the information required to successfully collect prices form has space for recording prices over a whole sequence
and, correspondingly, the head office will want from price of months, and the form is shuttled backward and forward
collectors all the information needed to assure the quality of monthly between the collector and the head office, or new
the prices collected. forms for collection and reporting are provided each month.
5.60 Including the previous period price on the collec- In the latter case, it can be useful if the form contains the
tion form continues to be a topic of debate. There are advan- prices recorded in the previous month alongside the spaces
tages and disadvantages to including the previous period for recording the current month’s prices, although this might
price on the collection form. The optional inclusion in the encourage the price collector or data provider to overly rely
form of the price collected in the previous period raises on the previously recorded price. The transfer of the prices
concerns that the price collector may be inclined to auto- to another form or system, especially when done manually,
matically record the previous price or be too influenced by it may lead to transcription errors and is best avoided. In the
when identifying the item in the outlet to be priced. However, example given in Annex 5.2, the previous month’s price is
this information is sometimes included to assist the collector shown (see paragraphs 5.185–5.193 on Computer-Assisted
in ensuring that the correct item and price are being recorded Data Collection for a more detailed discussion of the advan-
and, where paper forms are being used, to identify any tages and disadvantages of giving price history). For sim-
unusual price movements that need to be investigated. With plicity, the example assumes all prices entered on the form
electronic collection, if the collected price differs from the will include a sales tax.
87
5.65 Completion of the price collection forms, whether available, then in normal circumstances, this item should be
by paper or electronically, should comply with the following chosen as a “comparable” item and the item description suit-
guidelines: ably amended.5 If a different brand, size, or quality product
is available then this should be selected as a “new” item, but
(1) All prices should be entered in the collection sheet
only where no comparable items are available. The same
in full even if there is no price change. This is good
principles apply to other items such as clothes and fresh fruit
practice as it helps to ensure that the correct item is
and vegetables. With clothes, it is important that color, fab-
being priced, the price has been recorded correctly, and
ric, country of origin, logos, and size are specified to ensure
the price collector has not relied on information given
that the same item is priced each month. For fresh fruit and
by the outlet staff without checking. For instance, a
vegetables, useful attributes to record may be country of ori-
price may not have changed but the package size might
gin, class, and variety. For electrical equipment, the speci-
have been reduced, a practice sometimes used to make
fications and features given in the manufacturer’s catalog
price increases less transparent, but to the outlet staff,
may be important.
there has been no price change. A price should never be
5.70 A detailed description of the items being priced is
recorded as “no change.” All other entry fields on the
recorded to assist the price collector and the head office in
price collection sheet should have something entered if
choosing or confirming the suitability of a replacement for
only to indicate that they were not accidentally missed,
an item that has been withdrawn and in identifying changes
for instance, “Not Available (NA).”
in quality. The focus should be on recording price-determin-
(2) If a price is not available a reason must always be sup- ing characteristics. If the regular price collector is unable to
plied. The information given will be useful to the col- carry out the normal collection, full and accurate descrip-
lector’s supervisor and the index compilers as well as to tions will also enable a relief collector to carry out the col-
the price collector. The price collector may need to ask lection without any doubt concerning the correct items to
the outlet staff for an explanation and may also need to be priced.
choose a replacement, for example, if the outlet staff Most of the time, the item will be exactly as collected in
indicates that the item is no longer stocked. the previous month, and all that will be recorded is a current
5.66 No matter which method of local price collection price. However, if there is a change or uncertainty in the
is used, it is essential to provide some procedures to allow item, it is necessary for price collectors to use their judg-
the tracking of activities and formal sign-off, confirming ment and to inform the head office, bearing in mind that the
that processes have been completed and necessary actions head office staff is responsible for making the final decision.
undertaken on data checking and transfer to the head office. A precoded specification is less time-consuming and pro-
Audit trails are essential for work and quality management. vides better guidance to the price collector on what informa-
5.67 The price collection form should have space to tion should be reported. The codes should cover situations
provide full descriptions of the items being priced. Price col- faced regularly when collecting prices. The codes can be
lectors should be given a checklist or set of codes to record numeric or alphanumeric, and each one should indicate the
relevant information on changes relating to outlets, items, or action taken by the price collector or supervisor and will be
prices. The information needs to be systematically collected. associated with corresponding procedures to be followed in
For instance, codes to help with quality adjustment need to the calculation of the index by the head office. The form
reflect those characteristics that most influence price. Prior should include codes such as:
research, for example, based on the hedonic method (see • Comparable (C). The original item is no longer stocked
Chapter 6), can help to predetermine these characteristics. but a similar alternative has been collected that does not
5.68 Codes for managing the sample of outlets can be differ with regard to major price-determining attributes.
helpful and may include: The price is likely to be in a similar range although this
• Closed down: outlet permanently shut or closed down may not always be the case.
• Temporarily unavailable: outlet temporarily closed, but • Noncomparable (N). The original item has been replaced
likely to be open next month by a new item that is not really comparable but is equally
representative of that product group. If possible, the col-
• Refusal: owner or staff refuses to cooperate
lector should try to find the price of the “new” item in the
• Change of details: change of ownership or name or previous or price reference period.
change of purpose
• Sale or special offer (S). A price decrease because of a
genuine sale or special offer. This does not include dam-
Continuity and the Use of Price aged or out-of-date stock or clearance goods, which
Collection Codes should never be included. A price reduction where there is
5.69 As the index measures pure price changes, the no notice of a sale or special offer is not a “sale”; the item
same item should be priced every month to establish a true should still be priced, but without the S indicator code.
picture of price change. For example, if a jar of a super-
market’s own brand strawberry jam has been selected, that
brand and flavor should continue to be collected; if it is out 5
The replacement should be either (1) as similar as possible to the previous
of stock, another brand and flavor should not be used with- one; (2) the most popular “similar” product in the outlet; or (3) the “similar”
product that most likely will be available for future pricing. Unlike approach
out further investigation to establish whether this is a tem- (1), which leaves the sample “static” with the risk that it will be increas-
porary situation or likely to be permanent. In the latter case, ingly out of date and difficult to collect prices for, approaches (2) and (3)
and if another flavor of the same brand, size, and quality is have the advantage of introducing an element of sample replenishment.
88
• Recovery (R). A return to the normal selling price after a the specified quantity, as total and unit prices usually depend
sale or special offer. This does not need to be a return to on the number bought. If X eggs are to be priced and the
the price prior to the sale or special offer. A referral back price for the number is not quoted directly, the price of one
to the previous price collected and consultation with the egg can be obtained and multiplied by X to get the required
outlet staff will normally be required. price. However, the price collector must ensure that the unit
• Temporarily out of stock (T). Guidance will need to be price does not decrease with quantity: significant changes
given to the price collector concerning the meaning of in weight or quantity are to be avoided. Other examples are
“temporarily” (with regard to expected duration, which herbs, such as mint, or vegetables, such as cabbage leaves,
may vary for different items). If nonseasonal items are sold in bunches of variable size with no label giving the
missing for three or more months, depending upon coun- weight. In this instance, a number of bunches should be
try circumstances, enumerators should replace items weighed and priced to obtain an average price per unit of
immediately (for example, fashion clothing, if it is weight (for example, kilogram). It is necessary to purchase
unlikely that the identical item will come back into stock). bundles and weigh them at the head office when reliable
Typically, “T” indicators should not be used for more scales are not available at the outlet or market stall.
than three consecutive months, and in the fourth month, a 5.74 Certain food items, such as fruit or vegetables, are
replacement should be chosen. In food outlets, it is very more difficult to price as some outlets might price items per
unusual for items to go permanently out of stock. The col- number purchased while others might price by weight. For
lector should always try to check future availability with example, peppers may be priced either by weight or by unit
the retailer. no matter what the size; garlic may be priced per bulb, clove,
or by weight; and various types of berries may be priced
• Missing (M). Used where the outlet does not stock an item
either by weight or by container, which may differ in size or
or no longer intends to stock an item and there is no appro-
how full it is. In these instances, care needs to be taken with
priate alternative replacement. In these circumstances, it the product descriptions. The item description given on the
is recommended that the item is checked at subsequent price collection form should refer to the pack size, weight,
collections to ensure that a suitable replacement item has or quantity being priced. Collectors need to be aware of the
not come into stock. importance of collecting the same item from one month to
• Weight (W). A permanent weight or quantity change to the another, so that genuine price changes are recorded and not
product. quantity or quality price changes.
• Query (Q). Such a code may be used to supply extra retail
information to the head office (some examples are “10 Practical Collection Procedures: Quality
percent extra free,” “2 for the price of 1,” or a strange Management in the Field
price difference that is not covered by one of the other
indicators, such as a special edition issue of a magazine at 5.75 Adequate field procedures are required to ensure
an increased price). Arrangements need to be in place for that the quality of the price index is not compromised by
the head office to respond to these comments and to treat errors in price collection. Price collection needs to be care-
the price quotes accordingly. fully planned and managed, and effective instructions and
training are given to price collectors. In traditional price col-
5.71 Even if the retailer says there have been no price lection, most prices are collected through price collectors
changes since the previous month, the price collector should visiting individual outlets. Checks need to be carried out to
confirm prices. This will require some diplomacy, but it is ensure the accuracy of the data. This section gives guidance
important because the outlet staff may overlook a small on field procedures relating to local price collection and
number of price increases, or forget when the last increase provides an overview of quality management. It focuses on
has occurred. Also, checks need to be made to ensure that data validation in the field. Chapter 13 provides a broader
there have been no changes in the price-determining charac- look at the organization and management issues associated
teristics of the item such as package size or weight. with central price collection and with the complete process
5.72 As noted previously, a price should be recorded of producing a CPI.
only if the exact product being priced is on display and
immediately available for sale; however, for certain large Data Validation in the Field
items that must normally be ordered, such as furniture or 5.76 Data validation should be carried out throughout
cars, the price should be recorded as long as the retailer con- the entire process of compilation of a CPI, from the collec-
firms that it is available for delivery. tion of individual prices to their aggregation into indices.
5.77 The questionnaires, and the data collection soft-
Unit Prices ware if handheld computers or tablets are used, should facil-
5.73 Some food items, such as meat, fish, or cheese, itate quality assurance of the collected prices at or close to
are typically sold in variable weights and it is necessary the time of collection and record the results of the checks
to collect prices per unit of weight. The price per unit of made in the field as part of an audit trail.
weight should be taken from the package labeling or cal- 5.78 Collected prices can be compared to the prices
culated directly by the collector. Roughly, the same pack- of the same product sold in the same outlet previously col-
age size and type should be used each month, as the unit lected, and large movements can be checked for accuracy.
price might be lower for larger pack sizes or differ between Preferably, the head office should provide guidance on
package types. Other items, such as eggs, are often sold in what are viewed as acceptable ranges for price movements
specified quantities. For these, collectors record prices for based on previous price movements. If a price has changed
89
significantly or remained unchanged for a very long period, product description or inadequate reason given for price
the retailer should be asked to explain why. change). The audit report should be followed up by a formal
5.79 Where resources allow, field supervisors and inde- request to the price collector asking for corrective action and
pendent auditors should be deployed to support price col- confirmation that all necessary follow-up actions have been
lectors in providing accurate prices for input into the CPI. carried out.
Supervisors should check the validity of the prices and 5.84 Field supervisors should check for consistency
related information recorded by the price collectors, provide and credibility in price movements recorded by the collec-
assistance, and help when required as part of a collaborative tors under their oversight. For instance, if one location is
effort. The required level of the validity checks may vary reporting different price movements from the other loca-
depending on the nature of retailing and the data collection tions within the collection region, some explanation or
procedures. For example, the use of handheld computers a follow-up price collection will be required to check the
and tablets facilitates real-time data editing and the creation accuracy of the prices collected. Preferably, this should be
of price collection reports, reducing the chance of errors in done once data have been transferred into the computer sys-
price collection. tem and have been checked for errors. Tabulations of price
changes grouped by product or elementary aggregate should
Data Validation: Field Supervisors be provided to enable the supervisor to conduct these checks
5.80 Checks to ensure that data are complete and cor- efficiently. This will enable the supervisor to quickly iden-
rect should be carried out as early as possible in the col- tify extreme or inconsistent movements, which may indi-
lection and compilation processes. A return to the outlet to cate either errors in collection or unexpected behavior in the
re-input prices becomes increasingly less feasible as time market. These checks should be conducted regularly during
goes on, and there is a greater risk that the prices in the out- the price collection period.
lets will have changed since the initial collection. The use of
handheld computers or tablets by price collectors facilitates
much more detailed checking at the time of the initial collec- Quality Checks in the Head Office: Data Entry
tion of prices in the outlet than the equivalent paper system Queries and the Role of the Head Office
(see paragraphs 5.185–5.193 on electronic reporting). 5.85 Once the price collection has been completed and
5.81 Field supervisors have a number of important the prices submitted to the head office, a series of further
roles: training of data collectors when implementing new validation checks can be run. In determining the checks to
procedures or methods; one-on-one training of price collec- use, the validation checks carried out in the field should be
tors during joint field trips to correct any deviations from considered. For example, the use of handheld computers or
the procedures and tasks laid down in the price collection tablets will increase the potential for validation at the time of
documentation; and reviewing the work done by price col- price collection and reduce the need for detailed scrutiny at
lectors in previous days to verify the quality and to facilitate the head office. In addition, it would clearly not be produc-
the correction of errors. tive or cost-effective to repeat tests already carried out.
5.82 In an ideal system, field supervisors will be 5.86 The range of tests carried out on individual price
employed to regularly check that price collectors are adher- quotes can include:
ing to the price collection schedule and are undertaking (1) Price change. The price entered is compared with the
the required checks at the appropriate time. The supervisor price for the same product in the same outlet in the pre-
should check that price collectors are completing price col- vious month and triggers a query where the price dif-
lection forms correctly. A sample of collection forms from ference is outside preset percentage limits. These limits
each collector should be checked where it is not practical will vary, depending on the item or group of items, and
for supervisors to check every form. Checks may be made, may be determined by analyzing historical evidence
for example, on whether the price collector has attempted of price variation for the product or item concerned. If
to collect all prices from all outlets, that explanations have there is no valid price for the previous month because,
been given where prices were not obtained, and adequate for example, the item was out of stock, the check can
descriptions entered where replacements for disappear- be made against the price two months or three months
ing products have been priced. The supervisor may also be before.
required to check the accuracy with which data are trans- (2) Maximum/minimum prices. A query is raised if the
ferred from data collection forms to the computer. This is an price entered exceeds a maximum or is below a mini-
essential task associated with the quality assurance process mum price for the item of which the particular variety
and needs to be allocated to somebody other than the person is representative. The range may be derived from the
who initially inputs the data so that an independent check validated maximum and minimum values observed for
can be performed. that item in the previous month expanded by a standard
5.83 Supervisors should also be encouraged to visit out- scaling factor. This factor may vary between items,
lets and check the individual prices collected by price collec- again based on previous experience. Where necessary
tors. These checks can be organized either on a random basis and possible, the maximum/minimum price should take
or chosen on the basis of indicative information, such as account of any significant differences in average prices
extreme price variations. A typical audit report will include between, for example, regions.
the percentage error rate and a breakdown of whether the
errors are likely to have a high impact (for example, wrong 5.87 If a handheld computer or tablet is used (see para-
price, wrong item, or item available but listed as temporar- graphs 5.185–5.193), both tests (related to price change and
ily unavailable) or a low impact (for example, incomplete maximum/minimum prices) can be implemented easily to
90
take place at the time of collection; otherwise, they will need price collection will always be the main focus of the auditor.
to be conducted in the head office as soon as possible after There are several other areas, however, in which auditors
collection and prior to the computation of the index. A fail- can be called upon to contribute. Auditors may be required
ure in either test should prompt the collector to check and to help with the sampling of locations and items and check
correct or confirm the entry, and prompt for an explanatory that proposed collection locations contain an adequate range
comment. of outlets, and advise on economic conditions in these loca-
5.88 Queries raised may be either dealt with at the tions and on any dangerous areas. Auditors can carry out
head office or sent to the price collector for resolution. For product reviews. For example, if an item is causing diffi-
example, scrutiny of a form might show that a significant culty for price collectors, auditors can speak to collectors
price difference has arisen because the item priced was a and retailers to determine the reasons for these difficulties.
new product replacing another that has been discontinued. Auditors can also advise on changes to basket composition,
In this case, there may be no need to raise a query with the ensure that products suggested by the head office are avail-
price collector, unless there is evidence to suggest that label- able across the country, and suggest item descriptions. Fur-
ing the item “new product” is incorrect. thermore, auditors can provide reports on price collection in
5.89 If a price collection error is discovered and is too existing locations. For example, the head office may raise a
late in the process of computing a nonrevisable CPI for the query about an outlet in a location, and the auditors can visit
collection of the correct price, the head office will need to this outlet to find the answer to the question or to persuade a
reject the price and exclude that item from that month’s retailer to continue with the survey.
index and the price reference period, or treat it as a missing 5.95 The main purpose of audits is to ensure that each
price and impute a price using the price movements of simi- collector is following the procedures laid down for price
lar products. Where a CPI is revisable, it can be recompiled collection so that the risk of errors is reduced. However,
and the corrected figure published the following month. In there are other benefits of strategic importance with regard
some countries, the CPI is first published as a provisional to continuous quality improvement:
figure to facilitate the late take-on of data including the situ-
• Raising awareness of quality
ation just described.
5.90 Collectors should be encouraged to give feedback • The identification of the scope for introducing improve-
to the head office on their experiences of price collecting. ments to quality, including rectifying weaknesses in pro-
Collectors are a valuable source of information and often cedures, documentation, and price collection skills
give good early feedback on changes in the marketplace and
can often warn of size or product changes before the head Quality Checks of Local Price Collection:
office is able to obtain this information from other sources Back-Checking
such as newspaper advertisements. Collectors’ feedback can 5.96 Another approach to monitoring the standard of
be used to support observed price movements and to provide price collection is to carry out a back-check (that is, a retro-
supplementary briefing material. spective check of a proportion of the prices recorded during
5.91 Feedback can also form the basis of a newslet- the collection).
ter for collectors. Collectors’ shared experiences can guide 5.97 Back-checks can be used to:
other collectors on how to treat different situations or
• Assess the standard of competence of individual price
circumstances.
collectors
5.92 The periodic routine of collecting prices in the field
needs to be carefully planned and monitored, with arrange- • Audit the overall standard of price collection
ments in place to reflect local conditions. However, price • Identify general training needs or the specific needs of an
collectors should send in information when it is due, and late individual
submissions require follow-up. • Highlight any key issues including, for example, prob-
lems with documentation or instructions issued by the
Quality Checks of Local Price Collection: The head office
Role of Auditors • Identify areas where collection is problematic; for exam-
5.93 One way of monitoring the work being carried out ple, all collectors may have problems in certain types of
by price collectors and addressing any issues is to employ outlets, prompting the need for more detailed instructions
auditors to occasionally accompany collectors during the field from the head office
collection or to carry out a retrospective check on the col-
lected data. The function of the auditor is to check the validity 5.98 Back-checking should be done by an expert inde-
of the prices collected and to initiate corrective action that pendent of the process (preferably employed by the NSO),
may extend beyond correcting an individual price quote to such as an auditor. Back-checking is carried out by visit-
reviewing and updating instructions to price collectors and to ing the selected outlet and recollecting the prices and other
general retraining. The function can cover more than one geo- relevant information, such as attribute or description codes.
graphical area, but it does not normally extend to the supervi- This activity should be carried out close to the original col-
sor’s role in managing price collectors and the price collection lection period to avoid problems of price changes occurring
process. Sometimes the function of the auditor is combined in the interim. Back-checkers should seek permission from
with that of the field supervisor. The observations and com- the outlet staff beforehand and follow the general criteria of
ments of auditors are an essential part of quality management. conduct for local collection.
5.94 The range of tasks that an auditor carries out will 5.99 Performance criteria should be determined to
vary from one NSO to another. Monitoring the standard of which all back-check results can be compared. These criteria
91
should set, for example, the acceptable number of pricing country. In some cases, local or regional supervisors will do
errors per number of items checked. Well-defined criteria some of it; in other cases, it will be more appropriate for it
will enable performance of collectors or locations following all to be done centrally. Some of these tasks can be done by
a back-check. computer and others manually.
5.100 A back-check may include a range of tests to 5.105 Procedures should be in place to check that all
identify the following: documents, messages, or files are returned from the field, so
• Price difference—if the price is different, the auditor
that price collectors can be contacted about missing returns.
should check with outlet staff if there has been a price Initial checks should then be carried out to ensure that data
change since the original collection took place are complete and correct. For instance, checks should be run
to ensure that unexpected duplicate prices (that is, for the
• Insufficient item description—incomplete descriptions same item, in the same outlet, in the same location) are not
should be augmented to include all price-determining taken on, and that the location, outlet, and item identifier
characteristics codes, which accompany each price, exist and are valid. If
• Wrong item priced—such as incorrect size or brand being any prices fail these checks, a query should be raised with
chosen the price collector for clarification. Since some of the check-
• Items wrongly recorded as missing or temporarily out of ing may require reference back to the price collectors (or to
stock their supervisors or respondents when direct mail question-
naires are used), the timetable for producing the index must
5.101 A report should be sent to the head office for
allow for this communication to take place.
scrutiny once the back-check has been completed. The head
5.106 In deciding on what checks should take place in
office will then need to take appropriate action, which may
the head office, account should be taken of the validation
include, for example, retraining the price collector or send-
checks carried out in the field. The use of handheld com-
ing out supplementary instruction.
puters or tablets will increase the potential for validation at
Quality Checks Conducted Centrally the time of price collection and reduce the need for detailed
by the Head Office scrutiny at the head office. It would clearly not be produc-
5.102 Four kinds of regular checks are necessary at the tive or cost-effective to repeat all the tests already carried
head office: out locally, except as a secondary audit or random check that
those checks have been completed.
• Check that the price collectors’ reports are sent in when
they are due. If this is not done, it is necessary to find the Data Reports
reason and take appropriate action to obtain the reports. 5.107 Reports can help the head office staff identify
• Confirm that the reports contain what they are supposed prices for which the level or change stands out as different
to contain (for example, mandatory fields have not been from that reported for similar varieties elsewhere, or simply
left blank, numeric fields contain numbers, or nonnumeric where a change may need to be queried because it lies out-
fields do not contain numbers). side specified limits. Thus, a comparative analysis of other
• Review and edit each return. Substitutions may have to be prices collected for an item can be undertaken. A printout
made centrally, or those made by the collectors may have can list all prices that either fall well outside the range of
to be approved. Unusual or large price changes may need prices obtained for that representative item or for which the
to be queried. Items priced in multiple units or varying percentage change from last time falls outside a specified
weights may have to be converted to price per standard range, and a similar list can be compiled identifying out-
unit. Missing prices must be dealt with according to stan- liers based on the recent price behavior of the same item
dard rules relating to the cause. in the same outlet. The limits used will vary from item to
• Find and correct errors introduced when keying the num- item and can be amended in the light of experience. The CPI
bers into the computer or transcribing them onto work- compiler can then work through these lists, first ascertaining
sheets. Errors preferably should be avoided by eliminating whether there has been a keying-in error, and then exam-
the need to transcribe. ining whether any explanation furnished by the collector
adequately explains the divergent price behavior or whether
5.103 The way the data are organized in worksheets or a query should be sent back to the supervisor or collector.
in the computer may differ from the way they are arranged Again, the timetable for CPI compilation should allow for
on receipt. For instance, the data may arrive at the head office this, and anomalous observations should be discarded where
organized for price collection purposes by collector, outlet, an acceptable explanation or correction cannot be obtained
and item but will be entered on to a spreadsheet designed in time.
to reflect the computational needs of CPI compilation. The
5.108 Other reports may be produced regularly cover-
data in original format should be recorded for reference if
ing several periods (for example, three monthly) to detect
any problems with the data are disclosed during processing.
accumulated patterns, thus enabling broader problems to be
This facilitates operational management when dealing with
detected. For example:
queries. Furthermore, even if the same set of codes are used
in price collection and in the processing of collected prices, • One collector’s reports might show many more “outlet
other codes may have to be used for information that comes closed” remarks than those of other collectors, perhaps
in from the collectors in noncoded form. indicating either a motivational or training need on the
5.104 The organization of the quality checks con- part of that collector, or a change in retail trade patterns in
ducted centrally by the head office will vary from country to a particular area.
92
• Variety substitution for a particular representative item with complex index construction methods. This situation
might become more numerous than before, suggesting is associated with utilities and other services that account
a possible need for revision of the specification or the for relatively large expenditures and where there may be
choice of another representative item. only one or a few suppliers and prices are based on com-
• Where tight specifications list several brands and models plex tariffs.
of which one is to be chosen, but a large number of col- 5.111 Some of the techniques discussed in paragraphs
lected prices are for items not specified in the original list, 5.116–5.156 work best when applied to large quantities of
this suggests that the specified brands and models are no data and have the advantage of being automated with the
longer appropriate and that a review of the list is required. intervals identified for closer examination being generated
• The dispersion of price changes for a specific representa- by the prices data. Abnormal individual prices such as sale
tive item might be much larger than it used to be, rais- prices, or price movements such as sale recovery prices,
ing the question of whether it has been appropriately may be excluded from manual and automated procedures
specified. for the detection of outliers, particularly in the calculation
of upper and lower bounds, as they are not representative of
5.109 Routine computer-generated reports should the general trend in prices. Nevertheless, such prices should
enable to detect such problems. Two types of reports are be checked for accuracy, for instance by reference to pre-
particularly useful: index dispersion reports and price quote vious price history. Automated checking can be applied to
reports: seasonal sale prices and prices for seasonal items.
• Index dispersion report. This is a list of the current index 5.112 Automated checking essentially performs the
for each elementary aggregate, the number of valid quotes same basic filtering purpose for the identification of outliers
for each item, and the number of price relatives and their as the manual techniques described previously. It is some-
values. The ratios of current to previous valid prices can times referred to as statistical checking in contrast with the
be compared and queries generated, if these ratios fall out manual techniques which are sometimes referred to as non-
of acceptable ranges based on previous price behaviors statistical checking.
and considering any special circumstances such as the
introduction of discounted prices in seasonal sales. The
index dispersion reports can be used to identify quotes Data Validation and Editing
with price relatives that fall outside the range of the main
bulk of quotes. These suspect quotes can then be investi- Data Validation and Editing by the Head
gated, and appropriate action is taken. Office: Automated (Statistical) Checking
• Price quote report. This consists of a range of informa-
and the Use of Algorithms
tion on an item that the index dispersion report has high- 5.113 Data validation methods identify possible errors
lighted as warranting further investigation. Information and outliers for validation. Errors are incorrect prices, while
listed may include current price, recent previous prices, outliers can be defined as price movements that are excep-
and base price, together with locations and types of outlet. tionally large compared with most movements. The purpose
The report can be used to identify the quotes that require of data validation is to validate and confirm prices flagged
further investigation and to investigate rejected prices. as errors or outliers. Any errors should be corrected. Outli-
ers verified as correct should be used in the calculation of
5.110 While price collectors should examine every price the index.
they collect, it is not considered necessary nor feasible for 5.114 The main conceptual difference between auto-
collection supervisors and index compilers to subject every mated (statistical) checking and manual checking is that
collected price to the same level of scrutiny. It is recom- the automated technique calculates the limits for acceptable
mended that, to improve cost-effectiveness, a significance movement based on the data collected. These techniques
rating should be applied to determine how much time and have the benefit of automatically updating the acceptable
effort should be expended on examining and, where neces- limits in line with any overall change in price volatility, as
sary, editing individual prices. observed when new price data are received and the limits
• In general, prices from elementary aggregates with rela-
are recalculated. These techniques require a large amount
tively small price samples should receive more attention of data to provide reliable results and so are best suited to
from the index compiler. This is because, if the weights of data handlers and index compilers in regional offices and
the head office, where prices data from several collection
the elementary aggregates are broadly equal, each individ-
centers will have been collated and stored, rather than at the
ual price movement within these elementary aggregates
local level.
can have a much more significant influence on the index
5.115 Automated (statistical) checking compares each
calculation than any individual price movement from
price change with changes in the other items from a given
within an elementary aggregate with many price quotes.
price sample. The chosen price sample is usually the sample
• Price samples from elementary aggregates with high to which the item being checked belongs, but the sample for
expenditure weights should also be examined critically as testing may be a combination of price samples for similar
a higher expenditure weight will cause all price movements products. It can also be updated as more prices are received
within the sample to have a greater influence on the CPI. from the field. For each of the methods described in the fol-
• The highest risk is associated with elementary aggregates lowing text, the price ratios may measure the price change
with relatively large weights but few price quotes and over any time period: for instance, the change from the
93
previous period or the change from the same period in the the sample of 30 observations used in Example 1. A price
previous year. sample with at least one-third of the observations showing
no movements would not be unusual for many categories of
The Use of Median and Quartile Values items. If C is set to two, then 60 percent of the actual price
5.116 One method of setting the limits to determine movements would be flagged as possible errors, compared
whether a movement is a possible error is based on the with 30 percent in the unadjusted sample.
median and quartile values of the price ratios (R) from the 5.121 The index compiler should experiment with dif-
sample. The acceptable limits are set as a predefined mul- ferent values of C for different product groups or outlet
tiple of the range between the median and the quartiles. Any types to determine appropriate values for local use. It is rec-
observation with a price change outside this range is identi- ommended that a relatively low value of C be used. C need
fied as a possible error. The major benefit of a method like not be an integer and can be expressed with regard to frac-
this is that it is not affected by any single outlier value. A tions as well.
numerical example is provided in Annex 5.3. 5.122 The distribution of prices and price movements
5.117 The basic approach to estimating sensible upper is rarely normal; rather in most cases, a skewed distribution
and lower limits of acceptable price movement relies on the exists. Thus, the underlying assumption of a normal distri-
assumption that the observed price changes are normally bution is not valid, and the use of symmetrical upper and
distributed. Under this assumption, the distance between lower limits will result in a skewed distribution of prices
each of the first and third quartiles (RQ1 and RQ3 ) and the flagged up as possible errors or outliers. This is operation-
median (RM) will be the same: call this distance “DM.” Oper- ally inefficient and the examination of differing proportions
ating under this assumption, the proportion of price changes of “low” and of “high” prices and price movements could
that are likely to lie outside specified upper (LU) and lower lead to bias.
(LL) limits can be estimated from a normal distribution table.
The limits can be defined as A Modified Use of Median and Quartile Values
5.123 To use the previous method in practice, three
LU = RM + C × DM; and (5.1) modifications are recommended, as shown in paragraphs
LL = RM − C × DM 5.127–5.134.
5.124 Based on the simple price ratios, the distances
where C is a user-defined value. from the median represented by price decreases are not as
5.118 As discussed in paragraphs 5.126–5.134, a varia- large as the distances represented by price increases. As an
tion of this basic approach is recommended to allow for the example, consider a case where a product is on special offer
skewed distribution of price changes that can be observed at half price. This is represented by a price decrease of 50
in practice. percent. However, to return to the original price requires a
5.119 If C is defined as equal to one, then approximately 100 percent increase. To make the calculation of the dis-
50 percent of the observations will lie between the upper and tance from the center the same for extreme changes for price
lower limits. Using the standardized normal distribution, decreases and for price increases, the price ratios should be
this is equivalent to setting the limits at plus or minus 0.7 transformed. The transformed distance, Si, for the ith price
times the standard deviation (σ) from the median. Table 5.2 observation can be calculated as
provides approximate multiples of σ for selected values of C
and the associated percentage of the observations that will RM
Si = 1− if 0 < Ri< RM (5.2)
be flagged as possible errors and outliers. In practice, there Ri
are serious shortcomings with this method as described here. Ri
Si = −1 if Ri ≥ RM
5.120 In normal circumstances, the majority of obser- RM
vations for many products will not show any price move-
ment. Therefore, the values of the quartiles are likely to be where
very close to the median value. As a result, using small val-
ues for C is likely to cause the majority of price movements RM = median.
to be flagged as possible errors and outliers. To demonstrate
this effect, in Example 2 in Annex 5.3, 16 additional obser- 5.125 The observations with a price ratio lower than RM
vations indicating no price movements have been added to have now been transformed into the negative of the increase
required to return the price ratio to the value of RM. Any
observations with a price ratio equal to RM will have a trans-
formed price movement of zero. Observations with a price
Table 5.2 Selected Values of C and the Proportion of ratio greater than RM have been transformed to show changes
Observations Flagged as though they had increased from RM. The procedure is then
C σ Multiplier Expected Proportion of Observations carried out on the set of Si.
Flagged (percent) 5.126 In situations where the quartiles (RQ1 and RQ3)
1 0.68 50.00 are quite close in value to the median (RM), many small price
2 1.37 17.00 movements are likely to be identified as possible errors or
3 2.07 4.00 outliers. To reduce this problem, items with no price move-
4 2.75 0.70 ments may be removed from the calculations. If the accep-
6 4.00 0.14
tance interval is still very narrow, some minimum distance
94
should be set. A starting value is 5 percent for monthly where (EM − C*dQ1) is the lower bound, and (EM + C*dQ3) is
changes but it is up to the CPI compiler to choose, based on the upper bound of the interval. C is an extra variable that
past experience. may be introduced. The larger the C, the larger the accep-
5.127 The third modification is intended to overcome tance interval, and the fewer extremes and potential errors
the problem of using small samples. When using a small will be identified.
sample, the impact of one observation on the distances
between the quartiles and the median might be considered
too significant. In practice, the sample sizes for many ele- The Tukey Algorithm
mentary aggregates will be small. To improve the usefulness 5.132 The Tukey algorithm overcomes the problem
of this method, the samples from several similar elemen- of validating data when there are many observations with
tary aggregates can be combined. In this regard, elemen- no price change (that is, where many price relatives are
tary aggregates can be considered similar if their prices are equal to one indicating no price movement). The first step
believed to exhibit similar behavior. is to sort the sample of price relatives. The highest and
5.128 The Hidiroglou and Berthelot method, as lowest 5 percent are flagged for examination as possible
described in detail in Hidiroglou and Berthelot6 (1986), errors or outliers and removed from further calculation.
can easily be extended according to the description in para- All observations with no price movement are also removed
graphs 5.133–5.135. The variable si is independent of the from the sample before further calculations are done. The
price levels. To address the issue that the level of the prices next step is to calculate the arithmetic mean (AM) of the
can influence the acceptance interval, s may be transformed remaining observations (referred to as the Tukey sample).
into a new variable, E: This value is then used as the dividing value to separate
the observations into two smaller samples: an upper and
Ei = si ⋅ (max { pit −1 , pit }) , 0 ≤ U ≤ 1
U
(5.3) a lower set of price ratios. The arithmetic mean of each of
these two samples is then calculated as (AML, AMU). The
upper and lower Tukey limits (TU, TL) are then calculated
5.129 E is calculated as s multiplied by the largest of for the Tukey set as
the prices in period t or t − 1, raised to the power U. The
variable U determines to what degree the price level influ-
ences the acceptance interval. The larger U is, the larger the TU = AM + 2.5(AMU − AM) (5.6)
influence of the price level will be. If U = 0 the price level TL = AM − 2.5(AM − AML)
plays no role. This method is helpful, if the compiler wishes
to pay more attention to a price increase from 1,000 to 1,100 5.133 All observations that are greater than TU or less
than from 10 to 11. than TL are flagged as possible errors or outliers.
5.130 A further transformation can be made to ensure a 5.134 As this method excludes all observations with
minimum acceptance interval to avoid that too many price no price movement, the calculated limits are unlikely to be
changes being identified as possible errors for elementary close to the mean. Therefore, there will be no need to impose
aggregates (E) with only small price changes. In this case, a minimum difference. However, the problem of requiring
for each elementary aggregate, the median, EM, and the first a reasonably large number of observations in the sample
and third quartiles, EQ1 and EQ3, of the Ei’s are found, and the remains. Again, it may be necessary to combine the samples
following values calculated: of similar elementary aggregates. Example 3 in Annex 5.3
shows that five observations would have been flagged by
dQ1 = Max {EM − EQ1, |AEM|} (5.4) this method in comparison to 18 observations by the previ-
ous method based on the modified use of median and quar-
dQ3 = Max {EQ3 − EM, |AEM|} tile values (see 5.126–5.134).
5.135 As mentioned before, statistical methods
5.131 A is a constant that enters |AEM| to ensure a mini- of filtering have an advantage over simple filtering,
mum acceptance interval. A low value of A raises the prob- because the limits are set by the data and can be recal-
ability that (EM − EQ1) or (EQ3 − EM) determines dQ1 or dQ3, and culated over time. The disadvantage is that filtering can-
vice versa. For instance, if A is set to 0.05, |AEM| will be quite not be done until sufficient quantities of data have been
small so that (EM − EQ1) or (EQ3 − EM) are likely to determine collected, unless the index compiler uses approximations
dQ1 or dQ3, even if the dispersion of the Ei’s is relatively small. from past experience. The processes can be repeated as
If, on the other hand, the dispersion of the Ei’s becomes very additional prices are received. Compilers should aim to
small, |AEM| determines dQ1 and dQ3. Hence, A can be used set filters so that most of the records flagged as poten-
to avoid having too many price changes identified as pos- tial errors do turn out to be errors (or outliers requiring
sible errors for elementary aggregates with only small price explanation). The aim of all these methods of filtering
changes. The acceptance interval is finally defined as is to indicate which records require examination, not
to flag records for automatic deletion from the sample.
Each price movement should be checked for credibility
Acceptance interval = {EM − C*dQ1; EM + C*dQ3}, (5.5)
and representativeness. Only if the movement is an error
or unrepresentative should modification be considered.
6
Mike Hidiroglou and Jean-Marie Berthelot. 1986. “Statistical Editing and
There should not be a presumption that an outlier is
Imputation for Periodic Business Surveys.” Survey Methodology 12 (1): “wrong until proven right,” and outliers should not be
73–83. treated as incorrect prices.
95
Visual Data Validation 5.139 The tests for outliers are the same as those for
identifying potential errors. Outliers can be determined by
5.136 Using plot charts is helpful to spot the outliers
comparing the price movement relative to defined allowable
on the collected data and focus validation on these (see Fig-
limits. These may be either predetermined numerically or
ure 5.1 with arrows highlighting outliers). For visual data
predefined based on statistical tests.
validation, it is easy to use plot charts that are readily avail- 5.140 If, by exception, outliers are to be modified, they
able in spreadsheet software tools or can be programmed on are usually modified to lie on the predefined boundaries of
the information technology (IT) system. Visualization can acceptable movement or to be imputed by the movement
be more convenient than focusing validation to the change of a suitable sample of prices. Imputation by the average
of prices, especially when there are extreme price changes price change of the product group to which the item belongs
such as during sales periods or in the case of fresh fruit and yields a similar result as its exclusion (the same result
vegetables. If the plot chart is programmed into the IT system, within the elementary aggregate), but such imputations
from the outlier it is possible to make a direct link to the can have operational advantages as they employ proto-
observation. cols already in the calculation system for the imputation of
missing prices. An automatic adjustment should generally
Review of Outliers be avoided, and not be used to reduce volatility in an index,
for example, at the elementary aggregate level. The index
5.137 The detection of price observations that are out-
compiler should consider each case on its individual merits,
liers may be conducted through an examination of both
following agreed guidelines and deciding based on all rel-
price levels and of price movements. The movements will evant information. Prices should be modified or discarded
have been verified as being based on correctly collected and only if there is sufficient justification. The CPI protocols
recorded data but may not be representative of the behavior followed by the NSO may even forbid the modification or
of the section of the market that they are meant to represent. exclusion of outliers.
This leads to the concern that a different sample would have 5.141 Price collectors and their supervisors are
produced a significantly different and more representative responsible for providing as much information as pos-
average price movement. sible about the reasons for extreme price movements or
5.138 It is recommended to use resources efficiently on levels and why they accepted the price quote as valid. In
validation and checking of input data and to focus on iden- addition to checking for better accuracy, supervisors can
tifying the most important errors/outliers. The general rule also be instructed to compare the price movements for
should be to include verified prices. Excluding or modifying equivalent products obtained by all the collectors they
prices should be the exception. The aim should be to reflect supervise.
the reality.
Figure 5.1 Price Changes in Plot Chart during Sales Season
96
No Price Change the predefined movement limits described earlier, may be

5.142 If the price observations are collected in a way judged by the index compiler to be valid as a result of refer-
that prompts the respondent with the previously reported ring to other information such as market intelligence. Other
price, the respondent may report the same price as a matter potential errors might only be resolved after checking with
of convenience. This can happen even though the price may the respondent, if time allows.
have changed, or even when the particular product being sur- 5.148 If it is possible with individual price quotations
veyed is no longer available or has changed its price-deter- to resurvey the price or obtain a satisfactory explanation
mining characteristics. As many item prices do not change from the respondent, the query can be sent back to the price
frequently, this kind of error is unlikely to be spotted by collector and the data can be flagged as being verified and
normal checks. Often the situation comes to light when the then subsequently corrected if found to be in error. Even if
contact at the responding outlet changes and the new contact it is not possible to check with the respondent before the
has difficulty in finding something that corresponds to the computation deadline, the respondent could be questioned
price previously reported. It is advisable, therefore, to keep a during the next regular visit, as the answer may assist the
record of the last time a respondent reported a price change. NSO’s understanding of market behavior for the particular
product or retail sector. When a satisfactory explanation is
not available, the CPI procedures should provide guidelines
Missing Prices to aid the compiler in deciding how to treat the questionable
5.143 Treatment of missing prices is dealt with in more price. For instance, the compiler could omit the price, allow-
detail in Chapter 6. This section discusses ways of minimizing the processing system to impute a price, or modify the
ing the occurrence of missing observations. price to keep the price change within a predefined limit, but
5.144 It is important to maintain the relevance of the this is best avoided and should be the exception. If prices
sample of items priced. As part of the longer-term mainte- are modified without verification from the respondent, it is
nance of price samples, items and locations for which prices recommended that price collectors be informed of potential
are missing can be examined for common patterns. For problems during the next collection.
instance, if many retailers are missing the same item, there 5.149 NSOs can minimize problems caused by unusual
may be a general supply problem. This may be an indication prices and price movements by training price collectors
that an item will have to be replaced. If the number of regu- to recognize these situations, to check prices when first
larly missing items is growing, then the sample might need observed, and to collect relevant explanatory information
to be reviewed. If a particular outlet is recorded as having a during the initial price collecting visit. Avoiding return vis-
relatively large number of missing prices it may no longer its or calls keeps costs down and reduces the burden placed
be appropriate for the particular items assigned to it, or the on respondents.
varieties whose prices are collected in the outlet may need
to be reviewed. Checking by Impact or Data Output
5.145 When prices are collected using a questionnaire Checking
sent to the outlet, individual respondents generally follow a
regular pattern in terms of response. Some will return their 5.150 Filtering by impact, or output editing, is based on
price survey promptly, while others will take more time. calculating the impact that an individual price change has
Price collectors should be encouraged to become familiar on an index to which it contributes. This index can be an
with these patterns. If the system for recording the return elementary aggregate index, the total index, or some other
of these surveys also records the expected return date, then aggregate index. The impact that a price change has on an
unexpected nonreturns can be flagged even though the final index is its percentage change times its effective weight.
deadline for return of survey forms has not passed. These However, the exact calculation of the impact will depend on
respondents can be contacted in advance of the final dead- which formula has been applied for the elementary indices.
line to ensure that the survey form has not been forgotten. It is possible to set a maximum value for this impact, so
Early contact can reduce the number of prices still missing that all price changes that cause an impact greater than this
by the deadline. Respondents that provide prices for heavily can be flagged for review. The impact of a price change on
weighted items can also be monitored and contacted earlier. a higher-level index will also depend on the weight of the
elementary index in the aggregate.
5.151 At the lowest level, the appearance and disap-
Credibility Checking pearance of products in the sample cause the effective weight
5.146 Credibility checking tests the reasonableness of of an individual price to change substantially. The effective
the input data and the results obtained. Credibility check- weight is also affected if a price observation is used as an
ing of the results should take place after the checking of the imputation for other missing observations. The evaluation
numerical accuracy of the data at or shortly after price col- of effective weights in each period is possible, though com-
lection as described previously. These early checks are the plicated. To help highlight potential errors, nominal weights,
responsibility of the price collectors and their supervisors as a percentage of their sum, will usually provide a reason-
but also involve outlier detection at the head office. These able approximation. If the impact of 12-month changes is
early checks should discover all straightforward errors like required to highlight potential errors, approximations are the
incorrect coding (for example, wrongly attributing a price as only feasible filters, as the effective weights will vary over
a sale price) and the incorrect recording of prices. the period.
5.147 Addressing other potential errors is less straight- 5.152 One advantage of identifying potential errors in
forward. Results that fail a data check, such as exceeding this way is that it focuses on the results. Another advantage
97
is that this form of filtering helps the CPI compiler describe • Background to the NSO and the CPI
the contributions to change in the price indices. Much of • Use of the CPI and importance of accurately recording
this analysis is done after the indices have been calculated, prices
as the CPI staff often wishes to highlight in the statistical • The general principles of index compilation and price
press release those indices that contributed most to over- collection
all index changes. Sometimes the CPI staff findings that
some retail sectors have a relatively high contribution to the • How local price collection fits into the overall CPI compi-
overall price change may be considered counterintuitive. lation process
The change may also be traced back to an error, but it may • Instructions for retailer recruitment, getting permission to
be late in the production cycle and jeopardize the sched- enter outlets, and so on
uled release date. There is thus a case for identifying such • Practical price collection issues: for example, prod-
unusual contributions early as part of the data editing proce- uct identification or descriptions; pricing (for example,
dures rather than for analytical purposes. The disadvantage item descriptions, definition of price/sale price, rules
of this method is that in practice the final calculation of an relating to seasonal items, quantity conversions, quality
elementary index change may be rejected only after the CPI adjustment—when is an item or product equivalent?)
has been computed. • The timetable and administrative arrangements
5.159 Practical examples and practice collections should
Price Collector Training be an integral part of the learning process. For example, there
5.153 The training of local price collectors and clear should be an opportunity for:
instructions for them are vital elements in ensuring the qual- • Discussions about “equivalent” replacements using pho-
ity of the prices data and of the CPI. Collectors need to be tographs and item descriptions
properly trained, require adequate instructions, and must • Practice collections in office
have easy access to guidance because:
• Supervised practice collection in the field
• Price collection is of significant policy relevance.
5.160 Tests and evaluation of individual performance
• Quick judgments often need to be made. should be an integral part of the training. This could be
• Collectors often work remotely and on their own. achieved through:
• Instant communication is not always possible. • Written tests at the end of the training day
• Collectors work in a dynamic environment. • Evaluation by supervisors of practice collections in the
• Errors are difficult to rectify. field
5.154 Documents are needed to explain what is to be • Feedback to new collectors, including additional training
done, when it should be done, how it should be done, and needs
why it should be done. Reviewing the documentation also • Evaluation by collectors of training provided (essential
provides an opportunity to review the procedures. for ensuring training is relevant and effective)
5.155 Good documentation as part of an integrated
5.161 The evaluation of individual collector perfor-
quality management system is addressed in Chapter 13. The
mance is essential. Collectors must pass the required stan-
current chapter deals specifically with the documentation dards, against a checklist of tasks, before being allowed to
needs of price collectors and training. conduct a real collection.
5.156 Training for price collectors should enable them
to successfully perform all essential activities and deal with
potential difficulties including: Follow-Up Training and Refresher Training
5.162 The longer-term training is just as important as
• Persuade new outlets to become price providers
introductory training to the integrity of price collection,
• Understand and recognize occasions when prices pro- particularly with the evolution of the retail sector and CPI
vided are unacceptable methodology and when CPI baskets are updated. One way
• Record relevant information to describe the quality of facilitating this is for the price collector’s supervisor to:
change in a product
• Accompany the new price collector on a live price
• Recognize unusual price movements when checking their collection.
collected prices • Conduct a back-check of the prices collected to identify
any problems.
Training for Price Collectors • Produce an evaluation report that will provide a basis
5.157 Introductory training should be given to all price for further training of the price collector. The evalua-
collectors so that they gain the necessary skills before col- tion report can include a scorecard against a checklist of
required actions.
lecting prices for the CPI. It can also be a motivational tool.
5.158 A typical training schedule might consist of 5.163 Where resources allow, NSOs should conduct
a one-day training course at the head office (which may regular accompanied checks and background checks of all
include some refresher training for experienced collectors) price collectors in addition to special checks of performance
covering: issues. The information gathered can be used to compile
98
scorecards for individual price collectors, supervisors, and 5.168 The work instructions for supervisors of price
groups of price collectors. collectors should be in the form of a supplement to the price
5.164 Regular refresher training workshops should be collectors’ work instructions and should cover:
considered, especially where evidence from the field indi- • Checking the quality of the price collectors’ work
cates a need or where price collection procedures and con-
• Checking the accuracy and completeness of the prices
ventions change, or the CPI basket or sample of outlets has collected
been updated. These present an opportunity to: raise aware-
• Official recording of resource use (for example, cars
ness of the importance of collecting correct prices; provide
and bicycles for transport and funds for buying goods in
formal training on revised guidelines; resolve recurring
markets)
or recent problems; and provide price collectors with an
• Official procedures for maintenance of resources (for
opportunity to assist each other in managing problem situa-
example, testing the accuracy of scales)
tions encountered in the field, such as dealing with reluctant
respondents. • Creating complementary collection timetables for all col-
lectors within the supervisor’s area of responsibility
Training of Supervisors and the Head 5.169 Most of the documentation should be prepared
Office Staff by the head office, with input as appropriate from regional
offices, fieldwork supervisors, and data collectors. Cen-
5.165 Supervisors must be at least as well informed as trally prepared documentation will help ensure consistent
the price collectors. As the supervisor will normally be the practices in the field including between regions and should
first point of contact when a difficult situation is encoun- be readily available to all collectors and supervisors. The
tered during price collection, they also need a good under- documents can be available in paper or electronic form and
standing of the methodology and theory behind the selection should be accessible to the relevant staff.
of the product sample. Supervisors are part of the manage- 5.170 All documentation should be kept up to date.
ment team. Their training should cover: Effective documentation control systems should be in place.
With paper-based documentation, this could mean keeping
• Team management the instructions in a loose-leaf folder and issuing individual
• Performance appraisal (where this is normal office updates. The amendment pages should include version num-
practice) ber and date printed and be kept to a reasonable number for
• Project management ease of reference. Editorial access should be restricted and
password protected. A judgment will need to be made on
5.166 The head office staff should also be provided with when a redraft of a chapter or the complete working instruc-
a basic training in price collection. The benefits are threefold: tions is justified. Documentation is an essential part of a
• It gives the head office staff a better understanding of col-
quality management system.
5.171 An example of a documentation control template
lectors’ needs.
is given in Annex 5.4.
• It helps editing (staff at the head office will know what to
look out for).
• It supports disaster recovery (paragraphs 5.175–5.177) or
Disaster Recovery
business continuity (the head office staff will be able to 5.172 The prices data should be stored on a database
undertake price collection in an emergency). with hardware and software that is robust and supported
to minimize the business continuity risks associated with
running the existing system. But even with a resilient pro-
Documentation: Work Instructions duction system, contingency planning and operational conti-
5.167 Accessible, relevant, and up-to-date work nuity in the field and the head office are essential.
instructions are essential for both price collectors and their 5.173 In a world of rapidly changing statistical needs,
supervisors. Documentation should cover all aspects of the a statistical system should be able to respond quickly
and effectively to changing demands and should have
job and should in large part reflect what has been covered
the resilience needed to ensure continuity in the produc-
in training. Price collectors should be provided with work tion of statistics. This is not possible if systems are old,
instructions on: inflexible, and extensively tailored to past requirements.
• How to approach outlet staff Building a modern statistical infrastructure of methods,
tools to implement them, and a technical environment to
• How to ask questions to ensure that the required informa-
support the statistical processes is a significant compo-
tion is obtained
nent of achieving quality CPI compilation and computa-
• Appropriate personal behavior and dress codes tion. Contingency plans are needed when the unexpected
• Procedures for recording and passing on collected prices happens, for example, when there is a systems failure or
and other relevant information the price collection team is affected by an unexpected ill-
ness. Disaster recovery plans address these risks by, for
• Data checking
example, saving regular and frequent backup copies of the
• Creating collection schedules prices database on a secondary computer and by having
• Recognizing when recorded prices appear to be incorrect the capacity to collect prices when significant numbers of
99
price collectors are not available. Two strategies have been common and can also be an excellent channel for communi-
followed to accommodate a short-term unexpected deficit cating with the respondents.
in the numbers of price collectors:
• The collection of prices from a subsample of outlets chosen Collection by Telephone
to be representative of the full sample, thereby not need- 5.179 The prices for certain items, particularly services
ing so many price collectors. The price evolution from one such as electricians’ and plumbers’ charges and the cost of
period to the subsequent period can then be calculated from home security, may be obtained by telephoning the busi-
the subsample using matched pairs of price observations. ness or organization concerned. This applies when the out-
• The training of the head office staff in price collection (for lets provide standard items or services. However, even if
example, as part of their initial training), so that they can prices are obtained by telephone, the outlet should be visited
provide cover. The head office can also be given respon- occasionally. This helps to maintain cooperation through
sibility for price collection on a routine basis in a location personal contact and to ensure that there are no misunder-
close to the head office, with individual head office staff standings over the prices. This will be more important for
allocated the task of price collection on a rotating basis. An some outlets than others. For example, the price of hiring a
added benefit of this is that the head office staff become van may be less certain than the cost of an eye exam.
more familiar with the issues confronted by price collectors.
Computer-Assisted Data Collection
5.174 In cases where no fieldwork can be performed,
for example, where there is an all-out strike and no prices (CADC): The Use of Mobile Telephones,
can be collected by visiting outlets, an indicative figure may Handheld Computers, and Tablets
be possible using data from other sources, such as retailers 5.180 A number of NSOs have successfully used mobile
websites, but only if it can be established that like-for-like telephones, handheld computers, or tablets for local price
comparisons are being made and that the fixed-basket prin- collection. These technologies are now available at competi-
ciple is being adhered to. tive prices and the necessary infrastructures are generally in
place to make the use of CADC an attractive option.
Other Methods of Price Collection 5.181 A CADC system can lead to improvements in the
quality of CPI data, particularly as increased quality control
5.175 This chapter has so far focused in large part on at the point of data entry helps identify anomalies and ensure
traditional methods of price collection where price collec- that prices are correct. CADC has the potential to signifi-
tors visit outlets and record prices on paper forms. It now cantly improve the quality of the final CPI in the following
considers other methods of price collection. As noted at the ways:
beginning of the chapter, scanner data are given a separate
• Price history. The price collection program might allow
consideration in Chapter 10.
for a more comprehensive price history to be available
to the price collector, rather than just one previous price
Electronic Reporting included on paper forms. The availability of such data
5.176 Electronic reporting for centrally collected prices leads to less judgmental editing at the point of data col-
and use of handheld computers and tablets for local price lection and helps ensure the comparability of items, par-
collection can introduce greater efficiency into price collec- ticularly where prices for a particular item are variable.
tion and processing, and provide more scope for effective Some price statisticians have argued that machines should
quality assurance of prices and auditing, but both depend be programmed to reveal the price history only after a
on the introduction of effective quality control procedures. price quote has been entered so that collectors are not
Electronic reporting through electronic point of sale, com- overly influenced when locating the item to be priced or
monly referred to as scanner data, is another option. when choosing a replacement item. Others argue that an
5.177 Centrally collected data can be collected elec- early sight of the price history is useful information for
tronically in several ways. Once initial contact has been the price collectors as it assists them in their work.
made with data suppliers, a mutually convenient electronic • Quality checks in the field. The price collection program
data collecting procedure can be initiated. Options include: can include several automatic validity checks that can
• Emailing data collection spreadsheets between the NSO be used to identify where the price entered varies by a
and the retailer certain percentage (positive or negative) from the pre-
• Emailing of price lists at agreed times by retailers vious month’s price and the average price for that item
over a number of months and to flag up where data were
• Touch-tone dialing facilities for data to be supplied in an
not entered in all required fields (price, weight, indicator
agreed-upon format
code). These checks provide a useful marker for when
• The use of the internet (supplemented, if necessary, by a price needs to be double-checked. In a paper-based
telephone calls to clarify definitions and availability and system, such checks are carried out in the head office
whether the prices displayed on the internet are the same after the data have been collected, and audits can be car-
as those displayed in the corresponding outlets) (see also ried out after the collection period when prices may have
paragraphs 5.192–5.206 on collecting prices online and changed.
web scraping and Chapter 10 on scanner data)
• Transcription. There is a major risk of errors when transcrib-
5.178 The use of electronic websites/portals, where ing paper forms. This is not a risk when using CADC, where
respondents can report prices online, is becoming increasingly data can be transferred electronically to the head office.
100
5.182 The use of CADC also significantly reduces the 5.185 There is a short-term cost associated with the
time taken to make data available electronically at the head introduction and implementation of CADC for CPI price
office and between data collection and finalization. This can collection. Costs include:
be achieved through:
• The purchase of equipment.
• Transcription. Data collected on paper must be tran- • Upgrading “back-office” systems to enable interaction
scribed onto a computer for computation. This process is with the handheld computers, mobile telephones, or
time-consuming and resource-intensive. When data are tablets.
collected on handheld computers or tablets, the data can • The development of appropriate software for local price
be directly transferred electronically to the servers at the collection building on the experience of others. Costs
head office potentially in real time. depend on the functionality and sophistication of the pro-
• Transmission from regions. Electronic transmission will gram. Some NSOs have developed software for CADC
allow price collectors or regional offices to directly trans- that they may be ready to share.
mit an electronic data file to the head office thus avoiding
• Training of field staff and NSO staff on using the new
the need for postal or courier services or hand delivery
systems, including pilot price collection.
forms. This significantly increases the speed of data
transmission to the head office and reduces the cost of 5.186 There will also be longer-term costs associated
doing so. In addition, the head offices can look up the lat- with maintenance of the system and training of new staff,
est returns of price data from all regions immediately on but the additional expenditure on the latter is likely to not be
receipt and identify early any issues. significant as new staff must be trained whatever system of
• Quality checks in advance. As the functionality is available price collection is used.
to run certain quality checks in the field that would normally 5.187 When planning to use handheld computers,
be run in the office after data were transcribed, the time taken mobile telephones, or tablets for local price collection,
for quality checking centrally can be reduced, or, alterna- the decision needs to be made on whether the software is
tively, extra supplementary checks can be carried out. designed specifically for a certain hardware or not. If it is
hardware dependent, the life cycle of the software is usu-
5.183 These improvements to the speed of the process- ally dictated by the life cycle of the hardware. In data trans-
ing system can facilitate an earlier publication or provide mission, the issue of confidentiality needs to be addressed
opportunities to spend more time on analysis and interpreta- as well as how to secure the transmission action in practice
tion, the production of press releases and associated brief- with regard to reliability. The speed and reliability of any
ing, or the collection of more prices. network that is used should be tested from all price collec-
5.184 A CADC system enables certain checks that tion locations.
improve the efficiency of the CPI management. These include: 5.188 There are several advantages when using CADC,
• A check that all prices have been collected before the col- but as with all data collection methods it has risks and limi-
lector leaves the outlet. An electronic data collection form tations. With CADC the first risks are with the devices: the
can easily check whether all prices have been collected battery might not last the whole day of price collection,
and flag when they have not. This mitigates the risk of the especially if price collection is done in extreme conditions,
price collector inadvertently forgetting to price an item. and the data might be lost if the device breaks down during
• A check on when prices were imputed. Electronic data price collection period. Also, the devices have life cycles
collection can automatically record a date/time when the and if the software is coded to operate on a certain device
prices were entered in the machine. This is useful for vali- this may need to be rewritten for a replacement device. This
dation purposes. limitation causes costs when the devices come to the end of
their life cycle. There are also risks in data transfer, espe-
• Indicator codes. CADC provides the opportunity for
cially if the connections are poor, for example, if for any
additional features to be included in the data collection unpredicted situation the connection fails during download-
form. One such feature would be indicator codes (repre-
ing the data from CADC to the database. Some of these risks
sented by a single letter) that can be used to show when a
can be avoided using different solutions, like programming
price collected is for an item on sale, a replacement item,
the software independently from the device choice.
a missing item, a discontinued item, etc. This is a simple
tool to enhance the ease of validation and the manage-
ment of the item list (see paragraph 5.73). Collecting Prices Online and Web
Having price histories more readily at hand can: Scraping
5.189 A distinction should be made between collect-
• Make briefing of price collectors prior to fieldwork ing the prices of goods and services purchased online, to
more effective, for example, by a better appreciation of reflect the increasing importance of the internet as a chan-
when an “outlier” is a legitimate price change and vice nel for making purchases. A strategy to collect prices online
versa. through automation has the potential to reduce costs when
• Add to the quality assurance processes through assist- compared to the resource-demanding process of manual
ing with analysis when the index has been compiled price collection. Prices for online sales can differ from
and the briefing is being put together. These advantages the prices charged at physical outlets—even for the same
are particularly relevant when there can be significant retailer—and the profile of goods and services purchased
regional variations in price levels and trends. online can be different to other purchases. The prices in the
101
CPI must be representative and accurate. The motivation in 5.196 The sample of items to be priced should be repre-
adopting different data sources and extraction techniques is sentative of all online purchases and will be different from
important. The motivation for collecting prices online can online collection of prices charged by physical outlets. The
be twofold—efficiency and to ensure online purchases are prices recorded should represent the full cost of purchase.
properly representative. Online purchases may include standard extras such as deliv-
5.190 Retailers with an online presence—either as sell- ery charges. Unavoidable charges that are directly con-
ers of goods and services online, or as retailers who do not nected to the purchase of the priced product and which are
sell online but use the internet to list prices—should be not separately invoiced should be included in the price for
treated like any other retailer and be contacted first by the the purpose of CPI compilation. If the charge is separately
head office and be invited to participate in the price survey invoiced or relates to the purchase of a number of items,
even though this will not involve a physical interface. then the treatment is less clear-cut. One option is to include
these charges under transport services, but issues relating to
Online Collection of Prices the Classification of Individual Consumption According to
5.191 This section relates to collecting prices online Purpose (COICOP) arise (for more information on COICOP,
from publicly accessible websites, referring to goods and see Chapter 2). Another option is to follow the approach
services also sold in the corresponding physical outlets. It used for the Harmonised Index of Consumer Prices in Euro-
is a way of increasing the efficiency of price collection for pean Union countries, where such unavoidable charges that
a traditional sample of retail outlets and a fixed basket of are not part of the basic advertised price may be considered
goods and services. It does not relate to web scraping, the as an inseparable bundle of a good and a service and can be
collection of prices for online purchases or the use of scan- treated as one product (for additional information, see the
ner data. section on internet purchases in Chapter 11).
5.192 Instead of the traditional way of sending a price
collector to a retail outlet, extracting prices online directly Web Scraping
from websites will significantly reduce the price collection 5.197 Web scraping is the process of automated collec-
costs. Similarly, the response burden on data providers will tion of data from the internet through a set of computer soft-
be reduced to close to zero when extracting prices online ware techniques for extracting information from websites
directly from websites to replace postal or online question- (webpages) or using an application programming interface,
naires. Collecting prices online is relatively straightforward which is a set of routines, protocols, and tools for building
although care needs to be taken and checks put in place, software applications. Web scraping identifies and retrieves
especially when automated technical solutions are adopted. relevant data and downloads and organizes them in a suit-
The prices obtained online must represent the transaction able format for computing a CPI.
price in the physical outlet. Checks need to be made that the 5.198 There are technical measures applied in some
price advertised online is the same as the price advertised websites to avoid web scraping activities. These measures
on the retailer’s website and in the physical outlet and that block the scraper Internet Protocol address access to the
no overhead associated with buying online, such as delivery website or block the response to the scraper http browser
charges, are included.7 Also, online data need to contain suf- agent identification. The action is triggered after identifying
ficient information on characteristics to detect changes in an “abnormal” behavior (by analyzing the activity log) or
quality. When using online data sources, checks also need to by filtering access from some agents (through the robots.txt
be made that product code numbers, if used to identify the server configuration file).8
good or service, have not changed between price collection 5.199 Anti–web scraping devices by retailers reinforce
periods and that the codes are unique. the need to gain the cooperation of online retailers prior
5.193 Collecting prices online, which can be relatively to data scraping to avoid being blocked from data collec-
easy and cheap, will not always be a suitable substitute for tion. Retailers should always be informed about the nature,
price collectors. extension, and frequency of web scraping actions prior to
any web scraping taking place. NSOs should inform and ask
Collection of Prices for Online Purchases permission from retailers and agree to the most suitable web
5.194 Goods and services purchased on the internet scraping technique with the retailer’s management. Addi-
need to be properly reflected in the sample of prices used to tional measures, such as pauses between “scrapes,” may
compile the CPI. be required to maintain access given that the technology
5.195 Elementary product groups should be stratified employed to block access may be automated. There may
to reflect products purchased online. Sale information for also be legal constraints to web scraping.
weighting and the drawing of samples of internet purchases 5.200 When undertaking web scraping, the same
can be taken from HBSs, which should record information considerations apply as with online collection, most par-
about outlet type (including internet purchases), and from ticularly whether the aim is to download from publicly
information supplied by online retailers and market research accessible websites, the prices of goods and services sold in
companies.
8
Web Robots (also known as Web Wanderers, Crawlers, or Spiders) are
programs that traverse the internet automatically. Search engines use them
7
Charges relating to a delivery service arranged by the customer for spe- to index the web content. Website owners use the robots.txt file to give
cific items, or for bulk delivery with other items purchased, are legitimate instructions about their website to web robots; this is called the Robots
for inclusion in a CPI but should be recorded under a separate heading as Exclusion Protocol. For further information visit http://www.robotstxt.org/
indicated in COICOP (2018 COICOP 07.4 Transport services of goods). robotstxt.html.
102
physical outlets to increase the efficiency of price collection of these calculations and how a disregard of the differences
or whether to download prices paid for online purchases. in the frequency of price collection can lead to an underesti-
5.201 For web scraping to replace traditional price col- mate of inflation. However, the use of an empirical country-
lection, it needs to be demonstrated that prices, both online based data set is needed for a more accurate evaluation.
and in traditional outlets, are the same. While true for some 5.205 There is a variety of tools to aid the scraping
retailers and some products, it is almost certainly not true for activity. For example, many programs are available in com-
all retailers and all products. When prices on the web differ mon programming languages (for example, C, Python, or
from the prices in outlets, the price collection from a website JavaScript), for standalone applications or, most commonly,
should be seen as a different outlet type, which should be as add-ins to the browser. In most cases, web scraping tools
sampled along with traditional outlets. Traditional price col- are primarily designed to fulfill web application testing
lection would continue along with web scraping. It should and verification. For that reason, most of the tools used are
be noted that to integrate these prices into a CPI it is neces- implemented as browser add-ins (or plug-ins). This is not
sary to evaluate if there are some extra fees associated with the ideal situation with regard to IT architecture.
the purchase and that are not included in the prices listed on 5.206 When combining prices using traditional price
the website (see paragraphs 5.207–5.209). collection techniques with prices obtained from web scrap-
5.202 Integration of prices from different data sources ing, allowance needs to be made in the computation of aver-
needs to consider differences in sampling regimes (for age product prices for the different collection techniques
example, the relatively bigger samples facilitated by web being deployed, especially the frequency of price collection
scraping) and the relative values of sales. It is one of the where web scraping is sometimes carried out at a greater
reasons why online purchases are often treated as separate frequency than traditional price collection (see Annex 5.5).
elementary aggregates with separate weights.
5.203 Annex 5.6 provides more details on web scraping. Key Recommendations
Calculation of Average Price from • Price collection methods and organization decisions will
Different Data Sources in the depend upon country-specific circumstances, available
Elementary Aggregate resources, and could potentially vary by item.
5.204 Another aspect to be taken into consideration • Collected prices should reflect actual transaction prices
when combining traditional data sources with web-scraped including any tax and reflecting any discounts, sales, or
data refers to the different collection frequencies. Traditional promotions.
price collection deals with price “snapshots” scheduled in • Items should be priced as often as necessary to ensure that
such a way that the price series for a product in an outlet the index reflects a reliable and meaningful measure of
respects both frequency and equidistance with regard to time. price change.
In contrast, one of the perceived advantages of web-scraped • NSOs should strive to calculate an index based on
data is that prices can be collected daily during a certain prices covering the whole period (for example, month or
period, extract the average value and take that as the “snap- quarter).
shot” price for that week. A problem arises when trying to • The interval between price observations should be uni-
integrate data collected in “continuous daily” frequency into form for each outlet.
the regular monthly CPI. When using web-scraped data, care
• The price collection period should be made publicly
must be taken not to apply the raw collected data directly on
the calculation of the average monthly price at the elementary available and any changes announced well in advance.
aggregate level. Since the number and nature of observations • Proper training of price collectors is essential. Detailed
(probably many more than one per month per outlet/period documentation on data collection procedures and pro-
for price obtained by web scraping) are different, they should cesses should be drafted and made available to data col-
be transformed into compatible data. This can be accom- lectors and CPI staff.
plished simply by calculating an “outlet monthly price” for • Quality assurance procedures should be defined and
the product offer. This average price will then have the same implemented to ensure the accuracy of collected prices.
importance as other prices generated by the traditional snap-
• Data validation techniques are necessary to ensure the
shot approach. The consequences of using directly the online
accuracy and reliability of the collected prices.
prices will lead to a distorted estimate of inflation, since aver-
age prices will be calculated using a disproportionate number • Outlier detection methods should be defined and imple-
of prices that came from online collection, regardless of any mented. All questionable prices should be verified and
relative weight information. Annex 5.5 gives a brief example errors are corrected as necessary.
103
Annex 5.1
Consumer Price Index Price Collection Procedures
Figure A5.1 Planning and Organizing Price Collection
105
Annex 5.2
Consumer Price Index—Example of a Price Collection Form
Figure A5.2 Price Collection Form
107
Annex 5.3 Table A5.2 Parameters and Derived Limits (Example 1)

Parameter Series PR Series Si
Consumer Price Index— (a) (b) (c)
Automated Data Checking RQ1 0.94488 −0.03907

RM 0.98175 0
RQ3 1.05035 0.06988
DM 0.05274 0.05447
Example 1 C 2 2
Example 1 demonstrates the use of median and quartile val- LL 0.87628 −0.10894
ues to identify outliers. Table A5.1, column A, shows the LU 1.08723 0.10894
price ratios for the illustrative sample.
RM
Si = 1− , if 0 < R i< RM (A5.1)
Ri where the multiplier C is a user-defined value and has been
R set equal to two to limit the number of observations flagged
Si = i −1 , if Ri ≥ RM up as potential errors.
RM
The resulting upper and lower limits are shown in the
The first and third quartiles (RQ1 and RQ3) and the median “Series PR” in column B in Table A5.2.
(RM) can be obtained using the quartile function in Microsoft The price ratio series can be transformed to provide more
Excel. The average distance of the quartiles from the median equal weighting between negative and positive price move-
(DM) is defined as ments. The transformations are repeated here as
DM = (RQ3 − RQ1)/2 (A5.2) R

S i = 1− M , if 0 < Ri< RM
(A5.4)
Ri
The upper and lower limits are then calculated as
Ri
Si = −1, if Ri ≥ RM
LU = RM + C × DM; and (A5.3) RM
LL = RM − C × DM The transformed observations are shown in the “Series Si” in
column C in Table A5.1. The quartiles, median, and calcu-
Table A5.1 Price Relatives Showing Movement from lated limits for the transformed series are shown in column
Previous Period (Example 1) C in Table A5.2. The increased value for DM for the trans-
Price Ratio Si Flagged Using PR Flagged Using Si formed sample shows that the transformation has increased
Above 1.08723 or Above 0.10894 or
the distances for the price decreases while leaving the dis-
Below 0.87628 Below −0.10894 tances for positive movements the same.
Columns C and D in Table A5.1 show, respectively, the
(a) (b) (c) (d)
observations that would be flagged for further examination
0.81380 −0.20638 Extreme Extreme (indicated by the word “extreme”) for the original price
0.85250 −0.15161 Extreme Extreme ratios and the transformed price movements.
0.87600 −0.12072 Extreme Extreme
0.89900 −0.09205 OK OK
0.90860 −0.08051 OK OK Example 2
0.91350 −0.07471 OK OK
0.93390 −0.05124 OK OK Example 2 demonstrates the same statistical filtering method
0.94140 −0.04286 OK OK as in Example 1, but with 16 additional price ratios added to
0.95530 −0.02769 OK OK the sample. All the new price ratios show zero change. The
0.96080 −0.02180 OK OK same calculations are done but on a sample of 46 instead of
0.96580 −0.01651 OK OK
0.96680 −0.01546 OK OK 30 observations. Table A5.3 shows the sample of price ratios
0.97020 −0.01190 OK OK and the transformed price movements, as well as the obser-
0.97240 −0.00962 OK OK vations flagged for further observations. Table A5.4 shows
0.98170 −0.00005 OK OK the parameters and calculated limits.
0.98180 0.00005 OK OK
0.98430 0.00260 OK OK
A comparison of the results from the two examples
0.98690 0.00525 OK OK demonstrates the effect of having a significant number of
1.00340 0.02205 OK OK observations with no price movement. The distance from
1.00500 0.02368 OK OK the median (DM) is reduced and the number of observations
1.00610 0.02480 OK OK flagged for further examination is significantly increased.
1.03010 0.04925 OK OK
1.05710 0.07675 OK OK
1.08240 0.10252 OK OK Example 3
1.09090 0.11118 Extreme Extreme
1.09310 0.11342 Extreme Extreme Example 3 demonstrates the alternative statistical filtering
1.13000 0.15101 Extreme Extreme method, the Tukey algorithm. The enlarged sample from
1.15500 0.17647 Extreme Extreme Example 2 is used here to demonstrate the benefit of this
1.22960 0.25246 Extreme Extreme method when the sample has a large proportion of price ratios
1.23040 0.25327 Extreme Extreme
indicating no movement. Table A5.5 presents the intermediate
109
Table A5.3 Price Relatives Showing Movement from Table A5.5 Price Relatives Showing Movement from
Previous Period (Example 2) Previous Period (Example 3)
Price Ratio Si Flagged Using PR Flagged Using Si Price Less 5 Percent T Lower T Upper Flagged as Extreme
Relative Tails and Less Set Set
Above 1.03490 or Above 0.03598 or Above 1.17184 or
Zero Movement
Below 0.96510 Below −0.03598 Below 0.88332
(a) (b) (c) (d) (a) (b) (c) (d) (e)
0.81380 −0.22880 Extreme Extreme 0.8138 Extreme

0.85250 −0.17300 Extreme Extreme 0.8525 Extreme
0.87600 −0.14160 Extreme Extreme 0.8760 0.8760 0.8760 Extreme
0.89900 −0.11230 Extreme Extreme 0.8990 0.8990 0.8990 OK
0.96580 −0.03540 OK OK 0.9658 0.9658 0.9658 OK
0.96680 −0.03430 OK OK 0.9668 0.9668 0.9668 OK
0.97020 −0.03070 OK OK 0.9702 0.9702 0.9702 OK
0.97240 −0.02840 OK OK 0.9724 0.9724 0.9724 OK
0.98170 −0.01860 OK OK 0.9817 0.9817 0.9817 OK
0.98180 −0.01850 OK OK 0.9818 0.9818 0.9818 OK
0.98430 −0.01600 OK OK 0.9843 0.9843 0.9843 OK
0.98690 −0.01330 OK OK 0.9869 0.9869 0.9869 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00000 0.00000 OK OK 1.0000 OK
1.00340 0.00340 OK OK 1.0034 1.0034 1.0034 OK
1.00500 0.00500 OK OK 1.0050 1.0050 1.0050 OK
1.00610 0.00610 OK OK 1.0061 1.0061 1.0061 OK
1.03010 0.03010 OK OK 1.0301 1.0301 1.0301 OK
1.05710 0.05710 Extreme Extreme 1.0571 1.0571 1.0571 OK
1.22960 0.22960 Extreme Extreme 1.2296 Extreme
1.23040 0.23040 Extreme Extreme 1.2304 Extreme
Table A5.4 Parameters and Derived Limits (Example 2) The price ratios for the sample are shown in column A.
Parameter Series PR Series Si
The first step was to remove the highest and lowest 5 per-
cent of price ratios. Five percent of this sample equals 1.5
(a) (b) (c)
observations. This was rounded up to two observations
RQ1 0.96765 −0.03343 so the two highest and the two lowest price ratios were
RM 1.00000 0 removed. Observations with zero price movement were
RQ3 1.00255 0.00255 also removed. The remaining observations are shown in
DM 0.01745 0.01799
C 2 2 column B. The arithmetic mean (AM) of the remaining
LL 0.96510 −0.03598 set of observations was calculated. This value, along with
LU 1.03490 0.03598 other parameter calculations is shown in Table A5.6. The
arithmetic means of the lower and upper sets of data are
then calculated (labeled AML and AMU, respectively). The
data stages in addition to the basic sample and the indicator lower and upper data sets have been presented in columns
flagging extreme observations as possible errors. C and D, respectively, of Table A5.5 purely for explanatory
110
Table A5.6 Parameters and Derived Limits (Example 3) purposes. The Tukey lower and upper limits are then cal-
Parameter Value
culated as
(a) (b) TL = AM − 2.5(AM − AML) (A5.5)
AM 0.99429 TU = AM + 2.5(AMU − AM)
AML 0.94990
AMU 1.06531 The results are shown in Table A5.6.
TL 0.88332
TU 1.17184 Using this method, five observations would be selected for fur-
ther examination—many fewer than the 18 selected in Example 2.
111
Annex 5.4 only.” It provides checks, background (including explana-

tions for changes), and an audit trail. Two further benefits
Documentation Control accrue when combined with an electronic system:
Template • More efficient production of documentation as it helps

with initial compilation and updates, and reduces the need
to print and circulate paper copies
The documentation control template is an essential element • Better informed staff because they have immediate
of documentation, production, and dissemination and con- electronic access to the latest documentation, including
trol. Documentation control contributes to better quality desk instructions, with search facility by subject and
management as access to nonauthors is restricted to “read author
Figure A5.3 Documentation Control Template

Date Issued Documentation Reference Details of Change Reason for Change Name of Issuer
Day/Month/ Calculating CPI 2.1 Change in Process CPI Technical Board Has L. Smith,
Year Food Item Weights with Effect from . . . Agreed That in Future Weights Statistician, CPI
(nonseasonal) Should Be Taken from Program
(date) National Accounts
XX/XX/XX Calculating and 2.5 Change in Process Methodology Changes L. Smith,
Updating Price Index with Effect from . . . in Pricing Structures for Statistician, CPI
for Telecommunication Mobile Telephones—New Program
(date)
Services Methodology Agreed by CPI
Technical Board. Reflects
Changing Market
XX/XX/XX Desk Instructions for 3.1 Additional Checks Last Audit Indicated Current C. Brown,
Checking and Editing to Be Carried Out Checks Inadequate Resulting Operations
of Prices Based on Month- in Incorrect Prices Entering Manager, CPI
on-Month Price the CPI Program
Change
113
Annex 5.5 will differ according to the way these are computed at the
elementary aggregate level.
The Calculation of Average There are two methods of calculating the average price
(geometric mean) at the elementary aggregate level. The
Product Price When assumption is made that there are no expenditure weights (and
Combining Prices from each observation at a given point in time has the same weight)
and prices are collected to be representative of all sales.
Different Price Collection • Method 1: compute the geometric mean in two steps: first,
Methods and for Different an average price is calculated by kind of outlet/collection
method; second, a geometric average is calculated with
Price Collection Frequencies the two average prices per outlet
• Method 2: compute the geometric mean in a single stage
using all prices (not recommended)
When combining traditional and web scraping price col-
lection methods, there are two options for average product Taking into consideration the different collection frequen-
price computation. Web scraping can follow the same cal- cies, method 2 will undervalue price change (noting that in
endar collection that is used in traditional collection. With this example prices are falling) since it takes into account a
this option, the advantages of web scraping are disregarded. huge number of observations taken every day where the price
The second option will be to collect prices more frequently change is not significant and it is providing more “weight” to
with web scraping techniques. This will increase the num- the internet outlet with a price series that is more stable than
ber of observations used in the computation, potentially to the local outlet where the price being observed each 10
providing more reliability to the estimation. In the follow- days is recording a higher change. Due to these differences
ing simplified example relating to just one item, the price and the circumstance described, method 1 is recommended
for a specific product A is collected in a sampled local out- since it will provide the same “weight” to both kinds of out-
let in the framework of traditional price collection twice lets/collection methods. In summary, the prices can only be
along the considered time frame of 10 days. The price for averaged in a single step when there is an equal number of
the same product is collected on a daily basis using web price observations generated by each price collection for
scraping. These data are collected for two months and the each item over a given period. Averaging prices in a single
price at day 1 and at day 10 is the same for both outlet/col- step when the frequency of price collection varies between
lection types. However, the price averages and the indices different price collection methods leads to distorted results.
Table A5.7 Combining Prices from Different Price Collection Methods and for Different Price Collection Frequencies
Product A Month 1 Geometric
Mean
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9 Day 10
Local Shop (traditional collection) 500 480 489.898

Internet Website (web scraping) 500 500 500 497 497 490 489 483 480 480 491.536
Geometric Mean: Method 1 490.717
Geometric Mean: Method 2 491.263
Product A Month 2 Geometric Indices
Mean
Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9 Day 10
Local Shop (traditional collection) 450 430 439.886 89.79

Internet Website (web scraping) 450 452 452 448 445 445 445 440 435 430 444.146 90.25
Geometric Mean: Method 1 442.011 90.02
Geometric Mean: Method 2 443.433 90.17
115
Annex 5.6 • Web scraping may be time-consuming for large retailers

with many different webpages and products.
Web Scraping
The Process of Web Scraping
For those NSOs using web-scraped data for research and
Introduction production purposes, the process of performing web scrap-
The collection of prices is an integral component of the pro- ing has focused on two main methods:
duction of a CPI. Several data collection modes are available
• Web scraping performed within the NSO using statistical
and used by countries. These include personal visits, online,
telephone, administrative data, and transactions data. More software9
recently, with the growth of online retailing, pricing infor- • Web scraping services procured from a third-party/private
mation may be obtained directly from websites. Advances in company10
technology and automated scraping software have enabled The choice to perform web scraping within the NSO or con-
large-scale data collection from the internet. This is referred tracting with an external vendor will depend on the local
to as web scraping. context in which the NSO operates (for example, budgets,
NSO programming capacity, maintenance costs). With
Benefits of Web Scraping respect to those NSOs performing web scraping internally
Web scraping enables many more products to be priced, using statistical software, the process of web scraping infor-
and for these products to be priced more often than would mation from the internet can be broken down into three main
be possible using traditional data collection methods. steps:
Web scraping provides an opportunity to significantly • Confirm if the website allows scraping
enhance the sample of products and prices collected
• Scrape the website
by expanding product coverage. Other benefits of web
scraping include: • Clean the collected data
• Automated data collection at a reduced collection cost. There are two main ways to determine whether a website is
• Enhanced price representativity.
eligible for web scraping. First, staff setting up web scrapers
should check the terms and conditions section of a website for
• Increased price collection frequency (that is, daily versus “conditions of use.” Here, websites will often specify if web
once per month) allows for a more representative price scraping is allowed or prohibited. Additionally, a “robots.
for the period to be obtained. Secondary effects of using txt” file can be located within the root directory of a website.
an average period price include reduced volatility. These text files contain detailed information and may outline
• Faster identification of new and disappearing products. the conditions for web scraping possibly including who is
• Reduced respondent burden. allowed to web scrape, which information is available for
• Rich source of metadata (that is, product characteristics) scraping, and anything for which scraping is forbidden.
that can be extracted, stored, and potentially used for an Once it is determined that a website is available for scrap-
explicit quality adjustment (for example, hedonics). Such ing, a scraper is set up for the website. Each website is
metadata may also complement other data sources such as unique and the optimal scraping strategy may change from
transactions data. one site to the next. The scraper locates the website’s cat-
egory structure and identifies all the relevant categories to
Limitations of Web Scraping be scraped. The programmer defines the parts of the struc-
ture to be included and excluded. For example, a website
The lack of expenditure information from web scraping may list all the individual product categories, then additional
means that products/product groups cannot be weighted by categories such as “new products” or “all products” which
economic importance and limits the types of index formulas duplicate the products in the individual categories. These
that can be used to compile the index. Other potential limita- additional categories can be excluded by the programmer.
tions of web scraping are the following: The scraper then proceeds to download all products and
• Web scraping is limited to retail outlets that have an prices from the internet. An attempt is made to show as many
online presence (that is, potential for undercoverage). products on each page as possible by experimenting with
• Web scraping requires regular IT maintenance. Website
URL options on the website prior to setting up the scraper.
changes may cause the web scraping program to fail. There are two options available for pulling in the informa-
Businesses can block Internet Protocol addresses if they tion. First, data may be collected as text that essentially uses
detect the web scraping activity and wish to prevent it. scrapers to copy and paste from the website. In this case,
the process of cleaning the text is carried out following the
• If performed within the NSO, web scraping requires com- scraping. Alternatively, products and prices can be pulled in
pilers with intermediate programming knowledge to deal using designated HTML tags and classes which provide a
with the regular IT maintenance. more targeted approach to extracting and cleaning the data.
• Web scraping distinct prices for different geographic Another advantage of this approach is that product identifiers
regions may be difficult if the website detects the physi-
cal location of the computer’s Internet Protocol address.
This may require assumptions that retailers use national 9
Van Loon and Roels (2018).
pricing for regional price indices. 10
Bentley and Krsinich (2017).
117
Table A5.8 Web Scraping—Typical Data Structure Classification of Web-Scraped Data

Date Retailer Category Product ID Price The classification of web-scraped data involves similar
July 10, Retailer Children’s Brand XYZ— $45.00 considerations as described in Chapter 10 for the use of
2019 ABC Shirts Short Sleeve scanner data. Web-scraped data typically have some basic
Polo Shirt product text and category description that are required to be
July 10, Retailer Children’s Brand XYZ— $55.00 mapped to a NSO hierarchical classification (for example,
2019 ABC Shirts S/S Regular
Shirt
COICOP). Approaches considered by NSOs to solve these
July 10, Retailer Children’s Brand XY— $15.00 classification problems include:
2019 ABC Shirts Short Sleeve
Regular T-Shirt • Text string searches: Check for the presence or absence of
July 10, Retailer Children’s Brand XYZ— $65.00 keywords in the description string for classification.
2019 ABC Shirts Long Sleeve • Category mapping: Some of the data sets (or parts thereof)
Regular Shirt
July 10, Retailer Children’s Brand XYZ— $35.00
contain retailer categories for each product; if one of these
2019 ABC Shirts Short Sleeve categories sits within a classification, the category can be
Regular Shirt mapped to the classification.
• Manual mapping: A compiler looks at the description
string. This is the most feasible option for small data sets.
can occasionally be hidden in the HTML so pulling them in • Supervised learning algorithms: Provide training data
using the tags allows these to be added to the product descrip- (for example, using one or multiple methods previously
tion (as opposed to relying purely on the text description). mentioned) to statistical learning algorithm that identifies
However, the use of HTML tags is not easy for every website. patterns between text and training decision for automatic
For information collected as text, the raw data need to classification.
be cleaned post-scraping so that only the set of products
and prices remain. A pattern in the data needs to be uncov-
ered and coded to separate the products and prices from the Options to Define Individual Products
“noise” (including removing all the information prior to and An essential part of price measurement is accounting for
after the list of products). For products which are on sale, quality change and the introduction of new products. The
multiple prices may be listed. In these instances, the scraper CPI measures the price inflation in a basket of goods and
records the sale prices as the current price of the product. services priced at constant quality. If the quality of a product
With regard to the available information to construct price changes over time, then prices are adjusted so that the index
indices, Table A5.8 provides a summary of a typical meta- movements reflect pure price change. This has important
data scraped by NSOs. In summary, the data frame will typi- implications for web scraping.
cally include: The appearance and disappearance of products from a
• Date: specific day of the scrape (date) CPI sample has the potential to bias the index unless any
• Retailer: name of the retailer (text) corresponding changes in the quality of the sample are dealt
with appropriately. This poses a problem for the calculation
• Category: retailer’s website classifications (text)
of indices incorporating all (or most) web-scraped prices
• Product ID: text description of product (text) due to a large number of prices these data sets contain, the
• Price: specific price of product (numeric) high rate of product attrition, and the tendency for products
to have unusual price movements near the start and end of
Practical Considerations their life cycles.
One approach for dealing with this problem is to estimate
The basic information required to compile a price index includes the price change between two periods using the products
prices, expenditure information (or reasonable assumptions on available at both time points only, thus excluding the prices
substitution if no expenditure information is available), and of new and disappearing products. The matched-model
classifications (for both products and product groups). While method (as described in Chapter 6) involves discarding
web-scraped data (such as the example in Table A5.8) may information about new and disappearing products. However,
appear reasonably consistent with these requirements, it is it gains strength from the comprehensive coverage from the
important to ensure the conceptual requirements of the CPI (for census of products represented in the full web-scraped data
example, quality adjustment) are maintained. “Big data (trans- set. This can be considered as an application of the overlap
action, online, and administrative data) is ‘found data’ in the method, as it is almost certain that sales of new and disap-
sense that measuring CPI inflation is a secondary use of the pearing products would overlap with the sales of other prod-
data—the data were not created with this use in mind.”11 ucts that consumers would use as substitutes.
This subsection describes some of the main challenges An assumption behind the overlap method is that price
identified by NSOs when using web-scraped data, including: differences between products are reflective of quality differ-
• Classifying web-scraped data ences. This assumption seems reasonable in a competitive
• Options to define individual products marketplace and in normal circumstances. However, disap-
pearing products are sometimes sold at discounted prices to
• Index aggregation options
clear remaining stock (end of life cycle), and if not linked to
a product of comparable quality, may produce a “relaunch”
11
Bentley and Krsinich (2017, 6). problem and could potentially create a downward bias in the
118
index.12 This problem has been identified in various price amount of drift across most bilateral and multilateral indices
index studies on different types of products including high- with granular product definitions for clothing products.
technology goods,13 clothing,14 and personal care products.15
To overcome this problem, several practical strategies Country Case Study—Web Scraping
have been proposed, primarily focused on extracting char-
acteristics information (for example, brand and shirt type) As part of a broader project to modernize the CPI, one NSO
from text strings to form broader product definitions. Key began collecting prices using web scrapers beginning in
techniques proposed are the following: May of 2016. Web scrapers are currently programmed and
maintained by the NSO staff using Microsoft Excel (Visual
• Use of a broader product category (for example, chil- Basic for Applications), collecting approximately 500,000
dren’s shirts in Table A5.8) prices per week across more than 50 retailers. Web scrap-
• Text/regular expression functions: used to extract char- ing increases the sample of prices used to compile the
acteristics from semistructured text data (for example, index, thus providing a more meaningful measure of price
“XYZ” in Table A5.8 extracts the characteristic “brand”) change. From the second quarter of 2017, the NSO began
• Approximate (fuzzy) matching functions: used to approx- using web-scraped data in the calculation of the CPI. Variet-
imately match text strings using a penalty function ies are selected using the same methods as other forms of
price collection (field collection, online collection, or other
• Supervised learning algorithms:16 provide training data methods). An average price is calculated for each item over
to statistical learning algorithm that identifies patterns a given period.
between text and training decision for product classification Representative and stable products are selected. When
• Unsupervised learning algorithms:17 use characteristics transitioning a respondent from using traditional data col-
(for example, text string and price) and algorithms to lectors to web-scraped prices, an attempt is made to link the
automatically define “clusters” of products current field-collected product to the identical product on
the website. A determination regarding quality adjustment is
Options to Aggregate Prices made for each product to ensure that the new web-scraped
Web-scraped data do not contain quantity or expenditure product’s base period price is correct.
information. This would naturally restrict the choice of index For each web-scraped product in the sample, the price
formula to an unweighted price index formula (for example, for a given period (month or quarter) is an arithmetic aver-
Jevons), which is the current choice for many NSOs in the age of the prices which fall within the specified period. If a
CPI production. product disappears, a replacement product is selected from
The use of web-scraped data in compiling price indices among the other products within the relevant category from
continues to be the subject of extensive research. Some that respondent. A quality adjustment is performed to ensure
researchers have experimented with approximating expen- that only pure price change is shown. The combination of
diture weights based on data observed from the websites.18 category name, brand/product description, and price history
Using brand and product-type definitions and the number of of both the old and new product is sufficient to enable an
products as a proxy for quantity data, results demonstrate accurate quality adjustment to be applied.
that the web-scraped data approximate a benchmark index This NSO has adopted a phased approach to implement-
(scanner data) using the Geary–Khamis method. Additional ing new retailers and price index methods using web-scraped
published studies (for example, Metcalfe and others [2016]) data. With respect to respondents, an assessment is made
experiment with several bilateral and multilateral price for each respondent to be transitioned to web-scraped data,
index methods. Research findings showed a substantial taking into account the quality of web-scraped data over a
period of time, the correlation between online prices and
field-collected prices in each city, and the potential for col-
12
Chessa (2016). lection efficiencies and sample improvements. With respect
13
14
Silver and Heravi (2005). to price index methods, development work continues within
Chessa (2017). the NSO on text mining techniques (to form broader product
15
Chessa (2016).
16
Abe and Shinozaki (2018). definitions, especially for clothing) and price index methods
17
Metcalfe and others (2016). (both bilateral and multilateral index methods) that maxi-
18
Chessa and Griffioen (2016). mize and automate the use of web-scraped data.
119
TEMPORARILY AND PERMANENTLY MISSING
PRICES AND QUALITY CHANGE 6
Introduction (megabits per second, Mbps), and inclusion of a warranty
in the price of a dishwasher, all contribute to effective
6.1 Chapter 6 focuses on the treatment of temporarily decreases in price; consumers get more for their money.
and permanently missing varieties and their prices. While Similarly, quality decreases, for example, less legroom in
Chapter 5 focuses on the collection of data, Chapter 6 high- economy flights, when prices remain constant, are effec-
lights the important role of the price collector in the context tive increases in price. A volume change for an individual
of the treatment of missing prices and starts by providing variety may be comprised of a quantity and quality change.
an overview of the matched-model method (MMM). While The change in the variety’s nominal value of consumption
the MMM serves as the underlying method regarding the expenditure is the product of its price and volume change. It
treatment of missing prices, the chapter describes how the follows that the price change is the change in value divided
MMM can potentially fail, the consequences of this failure, by the change in volume.
and how to deal with the effects of such failure on price 6.6 National statistical offices (NSOs) go to great
measurement. lengths to ensure measured price changes are not influenced
6.2 Temporarily missing prices and the methods used by changes in the quality of items. By measuring the price
for the treatment of missing varieties are reviewed in this change of a fixed, constant-quality basket of goods and ser-
chapter. The concept of quality is defined and discussed. vices, NSOs use the MMM. When updating the basket, price
Explicit (direct) and implicit (indirect) methods for quality collectors visit selected outlets with broad details of an item
adjustment are identified and described. and identify the most popular, regularly stocked varieties
6.3 Some introductory notes are provided on general sold in each of the outlets. Next, they develop a detailed
measurement issues including the use of additive versus description of the variety including all the price-determining
multiplicative quality adjustments, price reference versus characteristics (for example, brand or size) and record the
current period quality adjustment, short-term versus long- price. This specification must be sufficiently detailed to
term comparisons, and geometric aggregation formula. include all price-determining characteristics to define a
Finally, this chapter considers the need of price measurement unique, specific variety. The detailed specification allows
in product markets with a rapid turnover of models,1 usually the price collector to easily identify the variety or model in
in the electronic and high-technology product markets. subsequent periods and record its matched price.
6.7 The measurement of changes in the level of con-
Background sumer prices using the MMM is appropriate when variety
prices are not missing. However, the use of the MMM is
6.4 The measurement of changes in the level of con- complicated by temporarily unavailable prices, for example
sumer prices is complicated by the appearance and disap- for one, two, or three months, because of a variety being out
pearance of new and old goods and services, as well as of stock and not yet replenished. A matched price is unavail-
changes in the quality of existing ones. If there were no able in these intervening months. The treatment of prices of
such complications, a representative sample could be taken temporarily missing varieties is considered in more detail
of the varieties of goods and services households consumed in paragraphs 6.52–6.72, but typically requires the missing
in a reference period 0, their prices recorded and compared variety’s price to be imputed for the month(s) in which it is
with the prices of the same matched varieties in subsequent missing using the price changes of similar goods or services,
periods. In this way, the prices of like would be compared or price changes drawn from a higher level of aggregation.
with like. In practice, some complications do exist. Varieties Actual prices are then compared with imputed prices for the
change in quality over time and replacements are of a differ- measurement of the consumer price index (CPI).
ent quality compared to the original. New and old models of 6.8 If varieties become permanently unavailable a
varieties appear and disappear. replacement variety is required that is preferably compa-
6.5 Changes in the quality of varieties should be treated rable with regard to the price-determining characteristics
as changes in the volume of the goods or service provided, of the missing variety. If the replacement is of a compa-
as opposed to changes in the price. For example, increases rable quality (that is, possesses the same price-determining
over time in the concentration of a detergent (number of characteristics), its price can be directly compared with the
washes per one kilogram packet), faster internet service last actual or imputed price for the missing variety. If the
replacement is noncomparable, for example, it is of a bet-
1
ter quality, the improvement in quality has to be explicitly
The term “model” of a good or service is used in this chapter mainly in the
context of high-technology goods such as laptops, household appliances,
quantified with regard to its “worth” or its contribution to
or cars. This use of the term “model” follows general usage and refers to a the price. Using this value, compilers make a price adjust-
specific variety whose characteristics are updated regularly. ment to reflect the difference in quality allowing the price
121
of like be compared with like. If a reliable explicit quality 6.14 The different methods for the treatment of missing
adjustment to the price is not possible, for data or resource prices, and the implied assumptions, are listed in Figure 6.3
reasons, implicit methods of quality adjustment are avail- and discussed in some detail in paragraphs 6.90–6.234. By
able. Details of explicit and implicit methods of quality definition, the prices of the unavailable varieties cannot be
adjustment are provided in paragraphs 6.90–6.188, and determined and the accuracy of some of the assumptions
methods for dealing with new and disappearing goods and about their price changes is difficult to establish. The match-
services are illustrated in Chapter 8. ing of prices of varieties allows for the measurement of price
6.9 Products that are “strongly” seasonal, that is, miss- changes unaffected by quality changes. When varieties are
ing in particular months when out of season but expected to replaced with new ones of a different quality, then a quality-
return in the next season, could also be treated as temporar- adjusted price is required. If the adjustment is inappropriate,
ily missing and imputed. Chapter 11 describes in more detail there is an error, and if it is inappropriate in a systematic
the different options that can be used for the treatment of direction, there is a bias. Careful quality-adjustment prac-
seasonal products. Strongly seasonal products include some tices are required to avoid error and bias. Such adjustments
fresh fruits, vegetables, and clothing. Also considered in are the subject of this chapter.
Chapter 11 is the treatment of “weakly” seasonal products:
available throughout the year but whose prices, sales, and
quality fluctuate. The prices of weakly seasonal products are Sampling Issues
not missing and are treated differently from strongly sea- 6.15 There are three sampling concerns when using the
sonal ones, as discussed in Chapter 11. MMM. First, the MMM and the use of replacements are
6.10 The matching of models facilitates the measure- designed to meet the needs of constant-quality price mea-
ment of constant-quality price change. When the matching surement and while the sample of varieties priced might
breaks down because of missing prices, temporary imputa- initially be designed to be representative of price changes
tions are necessary until the variety’s temporarily missing of the population of varieties, it is effectively following
price becomes available or a replacement can be introduced, a static sample of varieties that, over time, can become
thus helping to update the sample. However, there are prod- increasingly unrepresentative. The matching of prices of
uct markets where the matching breaks down on a regular identical varieties over time, by its nature, is likely to lead
basis because of high turnover of models, with new models to the monitoring of a sample of varieties increasingly
of different quality compared to the old ones, such as laptop unrepresentative of the population. The sample deterio-
computers. In this case, a failure to match and replace mod- rates over time because the MMM fails to incorporate new
els would lead to a seriously depleted and unrepresentative models/varieties into the sample, except as replacements
sample. Yet, a continual process of linking-in new replace- to obsolete ones. For example, substantial developments
ment varieties has been found to lead to a bias in CPI mea- in telecommunication hardware and services, reflected in
surement. Paragraphs 6.141–6.177 of this chapter outline an the growing number of models available, are excluded
alternative approach making use of hedonic regressions. from the sample covered by the CPI. This omission would
not be problematic if the (implicit) quality-adjusted price
changes of the excluded varieties were similar to those
Potential Errors in the Matched- of the included matched-model sample. However, this is
Model Method unlikely to be the case. The (quality-adjusted) prices of
old models being dropped may be relatively low and the
6.11 Three potential sources of error arise from the
(quality-adjusted) prices of new ones relatively high as
MMM: missing varieties, representativity of sample space,
part of a sales strategy of dumping old models at relatively
and new products.
low price to make way for the introduction of new models
priced relatively high.
Missing Varieties 6.16 A second sampling concern with the use of the
6.12 The first source of error in the MMM, and the MMM relates to the timing of the substitution and to when
focus of this chapter, occurs when a variety is no longer a replacement variety is chosen to replace an old one. In
available in the outlet. It may be temporarily out of stock, general, the prices of varieties continue to be monitored
discontinued, or one or more of the price-determining char- until they are no longer sold. This means that old varieties
acteristics may have changed. Whatever the reason, the vari- with limited sales continue to be monitored and included in
ety is effectively missing in the current period and a price the sample. Such varieties may show unusual price changes
cannot be collected. The variety’s price may be missing for as they approach the end of their life cycle because of the
other reasons: it may be a seasonal variety or one whose marketing strategies of firms. Firms typically identify
price does not need to be recorded so frequently, or it may gains to be made from different pricing strategies at dif-
be a custom-made good or service, supplied each time to the ferent times in the life cycle of products, particularly at
customer’s specification. the introduction and end of their life cycle. The (implicit
6.13 It is necessary to distinguish between varieties that or explicit) weight of end-of-cycle varieties in the index
are permanently and temporarily missing. Varieties that are would thus remain relatively high, being based on their
temporarily missing are varieties that are not available and sales share when they were sampled. Furthermore, new
not priced in the current period, but that are available and unmatched varieties with possibly relatively large sales
priced in subsequent periods. The treatment of varieties miss- would be ignored. Consequently, greater weight would be
ing because demand and supply are seasonal, as is the case given to the unusual price changes of matched varieties at
with some fruits and vegetables, is described in Chapter 11. the end of their life cycle.
122
Temporarily and Permanently Missing Prices and Quality Change
6.17 The final sampling concern with the use of the problem. The new product was not in the old basket and
MMM results from the price collector collecting prices until should be excluded. Although an index properly measuring
the variety is no longer available, thus forcing a replace- an old fixed basket would be appropriate in a definitional
ment. Data collectors replace the missing discontinued vari- sense, it would not be representative of what households
ety with the most popular or typically consumed variety. purchase. Such an index would thus be inappropriate. For a
This improves the coverage and representativity of the sam- cost of living index concerned with measuring the change in
ple. But it also makes reliable quality adjustments of prices expenditure necessary to maintain a constant level of utility,
between the old and new popular varieties more difficult. there is no doubt that it would be conceptually appropriate
The differences in quality are likely to be beyond those that to include the new product and any welfare gain from its
can be attributed to price differences in some overlap period, introduction, though as outlined in Chapter 8, this is highly
as one variety is in the final stages of its life cycle and the problematic in practice.
other in its first. Furthermore, the technical differences
between the varieties are likely to be of an order that makes
it more difficult to provide reliable, explicit estimates of the
Useful Concepts for the Treatment
effect of quality differences on prices. Finally, the (quality- of Missing Prices
adjusted) price changes of very old and very new varieties
are unlikely to meet assumptions of “similar price changes Multiplicative versus Additive Adjustment
to existing varieties or classes of varieties,” as required by 6.20 Explicit quality adjustments to prices can be made
the imputation methods. Many of the methods of dealing by adding a fixed amount (additive adjustment) or multipli-
with quality adjustment for unavailable varieties may be cation by a ratio (multiplicative adjustment). For example,
better served if the switch to a replacement variety is made consider m, an old variety, and n its replacement; for a price
earlier rather than later. Sampling issues are closely linked comparison over periods t, t + 1, and t + 2, the price of m is
to quality-adjustment methods. This topic is discussed in only available in periods t and t + 1 and n only available in
Chapter 8. periods t + 1 and t + 2. The measurement of constant-quality
6.18 This chapter references the need to permanently price change between periods t and t + 1 is based on variety
replace missing varieties to ensure that the sample of variet- m, pmt+1 / pmt , and the price change between periods t + 1 and
t +2 t +1
ies does not become unrepresentative. Samples of represen- t + 2 on variety n, pn / pn . Although not necessary for
tative varieties and outlets are generally updated when an the compilation of the price index, the previous calculation
index is updated. Where there is a lengthy period between can be elaborated in an equivalent, though more complex,
rebasing, the sample can become seriously deteriorated. It form that enables the nature of the quality adjustment from
is feasible to update/rotate the sample between periods of m to n to be identified and the multiplicative formulation
revising the index and Chapter 7 outlines how this can be demonstrated.
done in the context of maintaining the representativity of 6.21 A price relative over periods t, t + 1, and t + 2
t +1 t +1
the sample. Chapter 6 refers to the need for regular updat- requires an overlap ratio pn / pm to be used as a measure
ing of the sample which could be achieved through sample of the relative quality difference between the old variety and
rotation. its replacement. This ratio could then be multiplied by the
price of the old variety in period t, pmt to obtain the quality-
adjusted prices pn*t as outlined later in equation (6.6), and
New Products illustrated in Table 6.1:
6.19 Another potential source of error when using the 6.22 Such multiplicative formulations are generally
MMM arises when a new product is introduced into the mar- recommended, as the adjustment is invariant to the absolute
ketplace. When a really new product is introduced, there is value of the price. If the overlap ratio equals, for example,
an immediate gain in welfare or utility as demand switches 1.2, the new variety costs 20 percent more than the old.
from the old variety to the new variety. For example, the There may be some varieties for which the worth of the
introduction of mobile telephones represented a completely constituent parts is not considered to be in proportion to the
new good that led to an initial gain in utility or welfare to price. In other words, the constituent parts have their own,
consumers as they switched from the old (landlines) to the intrinsic, absolute, additive worth, which remains con-
new technology. This gain from the introduction of mobile stant over time. For example, retailers selling on websites
telephones, and subsequently of increasingly smarter tele- may include free shipping in the price. In some instances,
phones, would not be properly brought into the index by the cost of shipping may remain the same in the short to
waiting until the index was rebased, or by waiting for at least medium-term irrespective of what happens to the price of
two successive periods of prices for mobile telephones and the variety (exclusive of shipping). If the price no longer
linking the new price comparison to the old index. Subse- includes free shipping, this fall in quality should be valued
quent prices might be constant or even fall. The initial wel- as a fixed additive sum.
fare gain would be calculated from a comparison between
the price in the period of introduction and the hypothetical
price in the preceding period, during which supply would
be zero. The practical tools for estimating such a hypotheti- Table 6.1 Example of a Replacement Variety with Overlap
cal price are neither well developed nor practical for CPI t t+1 t+2
compilation, as referenced in more detail in Chapter 7. For Old Variety m pmt pmt+1
a CPI built on the concept of a base period and a fixed bas-
Replacement Variety n pn*t pnt+1 pnt+2
ket, this situation, strictly speaking, does not represent a
123
Price Reference versus Current Period depleted and degraded. Yet such one-on-one replacement is
Adjustment unlikely to be sufficient to maintain the representativity of
the sample, which may be based on variety selection when
6.23 Two variants of the approaches to quality adjust-
updating the index (Chapter 9) or sample rotation that may
ment are to make the adjustment either to the price in the
for some countries be many years ago. Since this initiation,
price reference period or to the price in the current period.
many newer varieties/products may have been introduced
For example, in the overlap method, described previously,
and old ones become obsolete. Maintaining the representa-
the implicit quality-adjustment coefficient was used to
tivity of the sample is addressed in Chapter 7 and a related
adjust pmt to pn*t . An alternative procedure would have been to
t +1 t +1 use of web-based and scanner data sources in Chapter 10.
multiply the ratio pm / pn by the price of the replacement
*t +2 The treatment of permanently missing variety prices was set
variety pn to obtain the quality-adjusted price pm . The
t +2
within the context of an elementary aggregate, where weights
first approach is more straightforward since, once the refer-
were neither available nor used. A modified Lowe price
ence period price has been adjusted, no subsequent adjust-
index number is used. The formulas are presented in more
ments are required. Each new price of a replacement can be
detail in Chapters 7 and 8 (see paragraphs 8.109–8.116).
compared with the adjusted reference price.
6.27 The example in Table 6.3 might be used to illus-
trate an individual elementary aggregate with geometric
Long-Term versus Short-Term mean prices compiled from several outlets. If a variety is
Comparisons temporarily not available, an imputation can be based on
6.24 Much of the analysis of quality adjustments in this a short-term month-on-month price relative, rather than a
Manual is undertaken by comparing prices between two adja- long-term price that might assume similar price movements
cent periods, say, month-on-month period t prices with those over several years. Similarly, for permanently missing prices
in a subsequent period t + 1. For long-term comparisons, where imputations are used to form an overlap comparison
the price reference period is taken as, for example, period for the missing variety price and its replacement, assump-
t and the index compiled by comparing prices in period t tions based on similar short-term month-on-month price
first with t + 1; then t with t + 2; t with t + 3, and so on. The movements are more reasonable than the less plausible ones
short-term framework allows long-term comparisons built based on long-term price movements.
up as the product of links: t first with t + 1; then t + 1 with 6.28 This mechanism facilitates the inclusion of new
t + 2; t + 2 with t +3, and so on; built up as a sequence specifications when old specifications become obsolete and
of links joined together by successive multiplication. This enables the index to better represent the dynamic changes
chapter focuses on short-term comparisons, for reasons of taking place in consumer choice. A direct comparison
their inherently better properties and for focus of exposition; between the price of a new replacement variety specifica-
In particular, as outlined in paragraphs 6.51–6.71, the short- tion in April 2020 with its old specification in 2019 is likely
term approach enables superior imputations to be made for to be challenging given the quality differences between the
temporarily missing prices and facilitates the incorporation two variety specifications over a long period.
of replacement varieties as and when an old variety’s price is 6.29 The short-term formula will differ (be improved)
permanently missing. The short-term approach is generally from its fixed-base long-term counterpart since the monthly
recommended. imputations will differ.
6.30 This chapter’s work has for the large part been
Treatment of Missing Variety Prices and Quality concerned with short-term price relatives compiled within
Adjustment within an Elementary Aggregate: an elementary aggregate, i. The larger picture of weighted
Short-Term Comparisons aggregation across elementary aggregates is for this context
6.25 This chapter uses a short-term framework of of missing prices and sample representativity, considered in
comparing period-on-period prices rather than a long-term Chapter 7, with illustrative calculations of the aggregation
framework of comparing the current period’s price with a formulas in Chapter 8, and the introduction of new weights
fixed price reference period. The use of matching is particu- in Chapter 9.
larly problematic for long-term price comparisons. For long-
term price comparisons a selection of representative models
in period 0, for example, the year (or a month in) 2020, have Aggregation Formula for Elementary
their prices compared with those in January 2021; in Febru- Price Indices
ary, for the 2020-February price relative; in March, for the 6.31 A ratio of geometric means (that is, the Jevons price
2020-March price relative; continuing for what may be, in index number formula) is used to measure price changes at
some countries, several years. The sample is increasingly the unweighted level of the elementary aggregate. Alterna-
depleted over time as 2020 varieties become obsolete. tive formulas include a ratio of arithmetic means (that is, the
6.26 The short-term approach has several advantages. Dutot price index number formula) and an arithmetic aver-
An illustration of the short-term against the long-term age of price ratios—the Carli price index number formula.
approach is included in Chapter 7. This chapter is concerned The Jevons price index formula is used here for reasons of
with temporarily and permanently missing variety prices. its better properties and focus of exposition. Chapter 9 pro-
Temporarily missing variety prices return to the sample, vides detail, an illustration, and the relative merits of the use
as do seasonal ones (as described in Chapter 11), and there of the Dutot and Carli price index number formula.
is no issue with maintaining the sample. However, perma- 6.32 With scanner and other such data, information on
nently missing variety prices need a replacement variety so prices, expenditure values, and quality characteristics will often
that, over time, the sample does not become increasingly be available for most individual models sold by major outlets.
124
This availability of data on transaction values allows weights to Table 6.2 Illustrative Variety Codes for Price Collector
be used at this detailed level of aggregation and thus the use of for Missing Values
weighted price index formulas as outlined in Chapter 9. Collection Code Description
T Temporarily Missing: The Variety Is Unavailable but

The Role of Price Collectors P
Is Expected to Be Available Again in the Near Future
Permanently Missing: The Variety Is No Longer
6.33 Price collectors have a critical role to play in the Available and Is Unlikely to Return
treatment of missing price observations. They observe and S Seasonally Missing: The Variety (Product Group)
record that a price is missing; whether it is temporarily or per- Is Strongly Seasonal and Is out of Season
C Comparable Replacement: A Replacement Variety
manently missing; if permanently missing, whether a compa- That Is Comparable to the Old Variety in All Major
rable or noncomparable replacement is available, and in the Aspects
latter case, the price and details of the replacement variety. NC Noncomparable Replacement: A Replacement
When selecting a sample of prices, the outlets are visited in a Variety That Is Not Comparable to the Old Variety
process referred to as initiation. During the initiation phase of
price collection, collectors identify the detailed specifications
of representative varieties sold. For example, for the general a replacement that matches as closely as possible the miss-
class of “large white bread, unsliced,” the more detailed, “large ing variety’s specification. After being unavailable for three
loaf, white, unsliced, Brand A, 800 gm” may be selected and months and with no indication that the variety will return, it
its details entered along with its price for subsequent periodic is treated as being permanently missing and a replacement
repricing. Ideally, price collectors should have in their pos- variety is sought. There may be particular products or cir-
session a checklist of these specifications when visiting out- cumstances in which the three-month rule can be relaxed,
lets in subsequent periods. The detailed specifications serve such as a temporary withdrawal of products for health rea-
many purposes including (1) to help identify the variety to be sons, national emergencies, or the logistics of restocking
priced; (2) to review and verify the variety’s specification to where there is a sound basis for a belief that the product will
ensure there have been no changes in the price-determining return in the near future, but not within three months.
characteristics; and (3) if the variety is noncomparable, to 6.37 Decisions on the treatment of missing prices are
use the specifications to identify a replacement variety to be made by CPI staff based on information provided by the
priced and record any changes in the price-determining speci- price collector and, in some instances, by a follow-up tele-
fications. Initiation activities occur only when it is necessary phone contact or visit to the outlet.
to select a new variety for pricing. 6.38 Variety prices may be missing for products because
6.34 The price collector also plays an important role they are seasonal and out of season. Products that are out of
in determining whether the missing price should be treated season, but expected to return in the next season, are treated
as temporarily or permanently missing. A variety’s price is differently from those considered in this chapter, and their
considered temporarily missing if the same variety is likely treatment is the subject of Chapter 11.
to return to the market within a reasonable time period. On 6.39 Data collection codes, such as those illustrated in
finding that the specified variety is not available for immedi- Table 6.2, should be used to justify or explain each missing
ate sale, the price collector should check with the manager price to ensure proper treatment. Metadata should be col-
or informed member of outlet staff whether it is temporar- lected on those products in which, for example, there is a
ily or permanently missing. If temporarily missing, record high level of missing prices of different forms and the extent
the expected duration: one, two, or more periods should be to which the prices are missing. Illustrative variety codes
recorded along with the reason for it being unavailable and are given in Table 6.2 and NSOs should build on the detail
an indication of the likelihood of its return. required to meet their specific needs.
6.35 Temporarily missing varieties have their prices
imputed; permanently missing ones require a replacement. The Treatment of Temporarily and
As these are different issues requiring different treatments,
it is important for the price collector to establish whether
Permanently Missing Variety Prices
the unavailability of the variety is temporary or permanent. 6.40 To measure aggregate price changes, a representa-
Consider the case of a monthly CPI. When a price is tempo- tive sample of varieties is selected from a sample of outlets,
rarily missing, it should be imputed using an overall mean along with detailed item descriptions that define each vari-
imputation, a targeted mean imputation, or a class mean ety and its specifications. For each selected variety, detailed
imputation. Some NSOs use a method referred to as car- item specifications, or descriptions, define a unique, specific
ryforward (repeating the last observed price). As stressed in variety that will be priced each period. The detailed specifi-
paragraph 6.65, this is not recommended. cations are included on the repricing form each period and
6.36 Permanent unavailability occurs when the variety serve as a prompt to help ensure that the same varieties are
is withdrawn from the market with no prospect of returning. being priced. Detailed checklists of variety descriptions
In some instances, it might be absent the next month and should be used, as any lack of clarity in the specifications
confirmed by the outlet manager or informed staff that it is may lead to errors. Attention should also be devoted to
not going to be replaced. With such information the price ensuring that the specifications used are not just to identify
collector should immediately look to collecting the price and the variety on a subsequent visit, for example its location
specifications of a replacement variety. In other cases, if a in the outlet, but contain all pertinent, price-determining
variety is out of stock, for example during three consecutive characteristics, otherwise it will not be possible to identify if
months, the price collector should be instructed to choose changes in quality have occurred.
125
6.41 The MMM succeeds in ensuring the prices of like 6.44 The treatment of missing price quotes is divided
being compared with like. The detailed variety specifica- into two types, depending on if these are for temporarily or
tions facilitate this process, and the method ensures that the permanently missing variety prices. These two types cannot
measurement of price change is not influenced by quality always be readily identified and treated accordingly. Some
changes. However, when a price is missing there is poten- form of mechanism or rule is required to enable a transi-
tial for mismeasurement. The treatment of missing prices tion from temporary to permanently missing variety prices.
depends on whether the variety is temporarily missing A price collector, supported by the head office, would
(that is, the product will be available in the near future) or regard a variety as permanently unavailable if verified by an
permanently missing (that is, the variety will not be avail- informed member of the outlet’s staff or, following a three-
able in the future). Each month, all imputed prices are used month period, the variety is no longer available and there is
in the calculation of the index. For example, an imputed no evidence that it will reappear. A price index suffers from
March price would be compared with the actual price col- sample depletion if an increasing number of temporary miss-
lected in February in order to calculate the price change ing prices are imputed over a lengthy period. Once judged
from February to March. If the variety returns in April, the to be a “permanently missing” variety type a replacement is
imputed March price is compared with its actual price col- found. If noncomparable, a quality adjustment is made, and
lected in April to calculate the March to April price change. imputations are no longer necessary.
All permanently missing prices require a replacement. 6.45 It may also be the case that a replacement, compa-
6.42 When a variety is missing in a month a number of rable or not, is unavailable (for example, for a videocassette
approaches may be used. Details of each method are pro- player when it became obsolete). The topic of introduc-
vided in separate sections of this chapter. While the termi- ing new goods and services and removing obsolete ones
nology may differ between NSOs, the methods include: is the subject of Chapter 7. Alternatively, a price collec-
tor may find the variety to be permanently missing in the
• Imputation. The price change for the missing variety is
outlet visited, and with no comparable or noncomparable
assumed to be the same as the price changes of all vari-
replacement, although the variety is sold in other outlets. An
eties in the item group, or targeted similar ones. Such
informed outlet staff may inform the price collector that the
imputations can be used for temporarily missing varieties.
product is no longer being stocked. For example, consider-
Permanently missing varieties, however, require a com-
ing bicycles in a sports shop, a price index for bicycles could
parable or noncomparable replacement.
be continued by imputing the price change for this outlet
• Direct comparison. If a variety is permanently missing using the price change in other outlets as described in para-
and a replacement variety is directly comparable, that is, graphs 6.51–6.71. Such a procedure depletes the sample and
it is so similar it can be assumed to have had more or in the longer term this should be remedied by a forced outlet
less the same quality characteristics as the missing one, replacement, sample rotation, or rebasing (see Chapter 7).
its price replaces the price of the unavailable variety. Any 6.46 Adjustments to prices are not a simple matter of
difference in price level between the new and the old is applying routine methods to prices in specified product
judged to be price change and not quality change. areas, and alternative approaches are suggested in this chap-
• Explicit quality adjustment. If a replacement variety is ter. Some approaches are more appropriate than others for
noncomparable (that is, there are identifiable quality dif- specific product areas. An understanding of the consumer
ferences) estimates of the impact of the quality differences market, technological features of the producing industry,
on the price enable quality-adjusted price comparisons to and alternative data sources will be required for the success-
be made between the old and the new varieties. ful implementation of quality adjustments.
• Implicit quality adjustment—overlap. If a replacement
variety is noncomparable and no information is available, The Treatment of Temporarily
or resources are too limited to allow reasonable explicit
estimates to be made of the impact of a quality change on Missing Price Observations
the price, the price difference between the old variety and Overall Mean Imputation
its replacement in the overlap period is then taken to be a
measure of the quality differential. 6.47 The overall mean imputation method uses the price
changes of other similar varieties as estimates of the price
6.43 Specific attention needs to be devoted to prod- change of the missing variety. Considering a Jevons elemen-
ucts with relatively big weights, where large proportions tary price index (that is, a geometric mean of price relatives,
of varieties are turned over. Some of the methods are not equivalent to the ratio for geometric means of prices),2 the
straightforward and require a higher level of expertise and price of the missing variety in the current period, t + 1, is
experience. Quality adjustment should be implemented by imputed by multiplying its price in the immediately preced-
developing a methodical approach on a product-by-product ing period t by the geometric mean of the price relatives
basis. Such concerns should not be used as excuses for fail- of the remaining matched varieties in the product group
ing to attempt to estimate quality-adjusted prices. Ignoring between these two periods. This method provides the same
quality change results in an implicit quality adjustment. This
approach assumes that any difference in price is pure price 2
Chapter 8 provides further examples of imputations and replacements
change and not changes because of a difference in quality. using the Jevons index, as compared with the Dutot (ratio of arithmetic
Such an implicit approach may not be appropriate and may means) and Carli indices (mean of the ratio of price relatives), and outlines
even be misleading. the relative merits of the three indices.
126
result as simply dropping the variety that is missing from 6.50 In practice, the use of the Jevons formula in this
both periods from the calculation. In practice, the series long-term form is not advised. Instead, a short-term formu-
is continued in the database by including the imputed lation is recommended as the product of month-on-month
prices; and, as described in Table 6.3, this forms a com- Jevons indices. The short-term cumulative Jevons index for
plete table of the variety prices in outlets A to F. The December 2019 = 100 to July 2020 is
imputations are based on assumptions of similar price
movements. 1 1
6.48 In the example in Table 6.3, a product with broad ∏ i =1( pi ) ∏ i =1( pi )
N Jan 2020 n N Feb 2020 n
specifications is sold in six outlets, A to F, with different I J (I Dec 2019

,I Jul 2020
)= 1
× 1
detailed outlet-specific specifications adopted for each ∏ i =1( p ) ∏i =1( pi )
N Dec 2019 n N Jan 2019 n
i
outlet. The price reference period is December 2019 with 1 1
successive prices collected for each outlet’s specification ∏ i =1( p
N Mar 2020 n
) ∏ i =1( pi
N
)
Jul 2020 n
i
in January, February, March, April, May, June, and July × 1
×....× 1
2020. The price collector finds the variety temporarily ∏ i =1( p
N Feb 2019 n
) ∏ i =1( pi
N
)
Jun 2019 n
missing in outlet F’s price collection for March 2020, and

i
likely to remain missing for the next month or two, but to (6.3)
return thereafter. In Table 6.3, the figures in bold for prices
of March to May 2020 in outlet F represent two alterna- 6.51 The long-term and short-term approaches in equa-
tive imputation methods, as explained in the following tions 6.2 and 6.3 provide the same result as the numerator
paragraphs. of each term on the right-hand side of equation 6.3 cancels
6.49 The Jevons index number formula is shown in with the denominator of the next. However, a major advan-
equation 6.2 as a direct or long-term index comparing, in its tage of the short-term formulation is that when an individual
second to last term, the geometric mean of the prices of each variety is missing, its price can be imputed using the month-
matched variety in the current month t with the geometric on-month price changes of similar or higher-level aggregates
mean of the prices in the price reference period, hereafter rather than the long-term comparison of the current month to
referred as period 0, and in the last term, for the example in the price reference period, which may be several years ago.
Table 6.3, July 2020 with the price reference period (= 100) 6.52 The right-hand side of equation 6.3 requires
of December 2019. monthly geometric mean prices to be calculated for the
numerators and denominators for each month-on-month
1 1 comparison.Table 6.3 shows the geometric means for the
 p t n ∏i =1( pi )
Jul 2020 n
N
price reference period (December 2019, January 2020, and
I j ( I 0 , I t ) = ∏  i0  ≡
nN
1 (6.2) February 2020); missing prices for March, April, and May;
 pi 
i =1 
∏ i =1( pi ) and prices returning in June and July. The first task is to
N Dec 2019 n
impute the missing price for March 2020. This is obtained

where p = price. using the ratio of geometric mean prices for each of the
Table 6.3 Temporarily Missing Price Observations and Imputed Prices

Outlets Price Reference Period
Dec.-19 Jan.-20 Feb.-20 Mar.-20 Apr.-20 May-20 Jun.-20 Jul.-20
A Supermarket 5.25 5.25 5.49 5.49 5.49 5.49 5.49 5.49

B Supermarket 5.10 5.10 5.10 5.25 5.25 5.25 5.25 5.25
C Supermarket 5.20 5.20 5.20 5.20 5.20 5.25 5.25 5.25
D Independent Trader 5.49 5.49 5.49 5.65 5.75 5.80 5.80 6.00
E Independent Trader 5.99 6.50 6.50 6.90 6.90 6.90 6.90 7.00
F Independent Trader 5.99 5.99 5.99 6.13 6.15 6.17 6.25 6.25
Overall Mean Imputation: F Mar:May
Geometric Mean: A:E 5.54 5.67 5.69 5.71
Geometric Mean: A:F 5.49 5.57 5.61 5.74 5.76 5.78 5.79 5.84
Short-Term Price Relatives: A:F 1.00000 1.01371 1.00748 1.02377 1.00352 1.00365 1.00198 1.00864
Long-Term Indices as Product 100.00 101.37 102.13 104.56 104.92 105.31 105.52 106.38
of Short Term
Targeted Imputation: Independent Traders
D Independent Trader 5.49 5.49 5.49 5.65 5.75 5.80 5.80 6.00
E Independent Trader 5.99 6.50 6.50 6.90 6.90 6.90 6.90 7.00
F Independent Trader 5.99 5.99 5.99 6.26 6.32 6.34 6.25 6.25
Geometric Mean: A:C & D:F 5.49 5.57 5.61 5.76 5.79 5.81 5.79 5.84
Short-Term Price Relatives: A:F 1.00000 1.01371 1.00718 1.02731 1.00440 1.00377 0.99753 1.00808
Long-Term Indices as Product 100.00 101.37 102.13 104.92 105.38 105.78 105.52 106.37
of Short Term
Bold: Imputed values
127
matched sample of outlets A to E in February and March to Table 6.4 Overall Mean and Targeted Mean Imputations
provide the short-term price relative:3 Overall Mean Imputation Targeted Mean Imputation
1 (a) (b)
∏ i= A ( p )
E Mar 2020 5
Dec.-19 100 100
PJ ( P Feb 2020 , P Mar 2020 ) =
i
1 Jan.-20 100.00 × 1.013712 = 101.37 100.00 × 1.013712 = 101.37
∏ (p )
E Feb 2020 5
Feb.-20 101.37 × 1.007478 = 102.13 101.37 × 1.007181 = 102.13
i= A i
Mar.-20 102.13 × 1.023765 = 104.56 102.13 × 1.026738 = 104.92
1 Apr.-20 104.56 × 1.003515 = 104.92 104.92 × 1.004396 = 105.38
(5.49×5.25×5.20×5.65× 6.90)5 May-20 104.92 × 1.003652 = 105.31 105.38 × 1.003766 = 105.78
= 1 Jun.-20 105.31 × 1.001977 = 105.52 105.78 × 0.997531 = 105.52
(5.49×5.10×5.20×5.49× 6.50)5 Jul.-20 105.52 × 1.008081 = 106.37 105.52 × 1.008081 = 106.37
5.67
= = 1.02347 (6.4)
5.54
a wider subset of the higher level of aggregation will depend
This change in the (geometric) mean price for the matched in part on the adequacy of the sample size for the subset
prices A to E is from February to March. The increase of of similar varieties and the homogeneity of the elementary
1.02347 when multiplied by the February price of 5.99 aggregate at the higher level.
yields an imputed March price of 1.02347 × 5.99 = 6.13. 6.57 Column B in Table 6.4 shows imputed prices for
6.53 The price collector subsequently finds April and the missing variety prices in outlet F for March, April, and
May prices for variety F to be temporarily missing. The May 2020, based on adjusting the preceding period’s price
respective imputed prices are: 1.0353 × 6.13 = 6.15 and by the price movements of the remaining matched pairs of
1.00351 × 6.15 = 6.17. The imputed prices are entered in prices at other independent traders, D and E, rather than all
Table 6.3, highlighted in bold, providing a complete table of outlets, presented in column A for comparison. Changes in
prices for outlets A to F over the price reference period and the geometric mean price relative as applied to adjust the
subsequent months. preceding period’s price are:
6.54 The short-term month-on-month price relative for
all outlets A to F for February to March 2020 is given as
follows: 5.99 × (5.65 × 6.90)1/2/(5.49 × 6.50)1/2 =
5.99 × 1.04522 = 6.26 for March 2020; and
1
∏ (p )
F Mar 2020 6
5.74 6.26 × (5.75 × 6.90)1/2/(5.65 × 6.90)1/2 =
PJ ( P Feb 2020 , P Mar 2020 ) = i= A i
= 6.26 × 1.00881 = 6.32 for April 2020; and
1
5.61
∏ (p )
F Feb 2020 6
i= A i 6.32 × (5.80 × 6.90)1/2/(5.75 × 6.90)1/2 =
(6.5) (6.6)
= 1.02317 6.32 × 1.00434 = 6.34 for May 2020
6.55 The short-term price relatives for outlets A to F cal-
culated from imputed prices, as in equation 6.5, provides the 6.58 The price index is compiled as the cumulative
same measure as the price relative calculated from outlets A product of the short-term price relatives, as shown in Tables
to E since the price for outlet F is computed from the price 6.3 and 6.4, again highlighted in bold.
change for outlets A to E. Other short-term price relatives 6.59 The higher levels used at this elementary stage of
are shown in Table 6.3. The long-term price index (Decem- aggregation are country-specific and follow the country’s
ber 2019 = 100) to July 2020 is shown in Table 6.4 and in CPI aggregation structure, as described in Chapter 8 (para-
Table 6.3 as the cumulative product of short-term relatives. graphs 8.9–8.10) and Figure 8.1. The higher level might be
a region and type of outlet. For example, in Figure 8.1, this
would be Brand A of whole grain bread sold in supermarkets
Targeted Mean Imputation in the Northern region; if there is an insufficient sample size
6.56 The overall mean imputation is based on assum- for Brand A, similar Brands A and B, or all brands, might
ing that the price change of the temporarily missing variety be used for all types of outlets in the region. Imputation of
is similar to the overall price change at a higher level of the missing price by the average change of the available
aggregation. A targeted form of the method would use price prices may be applied for elementary aggregates where the
movements of an elementary aggregate or an aggregate of prices can be expected to move in the same direction. The
similar varieties, that is, varieties expected to have simi- imputation can be made using all of the remaining prices in
lar short-term price changes. The sample of observations the elementary aggregate. This is numerically equivalent to
used for the targeting may be specific to a type of outlet omitting the variety for the immediate period, but it is neces-
and region and cluster of features, for example, “up-market” sary to make the imputation (see 6.69–6.71).
television sets. It would generally be a subset of varieties 6.60 An imputed price should always be directly com-
within a higher level of aggregation. The decision to target pared with the actual price on the variety’s return as this
the imputation using similar varieties in the subset or to use provides a self-correcting measure. For example, if the
imputation was not accurate and showed decreases in prices
over the period, when in fact the price of the variety sold
3
Geometric means over large samples are more accurately computed as the elsewhere or, if not sold, being held in storage, was increas-
1 N N  ing, then a direct comparison between the last imputed and
equivalent: exp  ∑lnpit − ∑lnpi0  .
N   the returning actual price would bring the index back to
 i=1 i=1 
128
its longer-term trend. The long-term price indices in Table short-term month-on-month price changes between the
6.3 using an overall imputation for June and July 2020 are missing variety’s price changes, and price changes of sim-
105.42 and 106.37, respectively. These are the same results ilar varieties and higher levels of aggregation, to choose
as those given in Table 6.3 for targeted imputations. Both between an overall or targeted imputation accordingly. In
series have self-corrected the imputations to return to the the much-simplified illustration of Table 6.3, price changes
long-term price changes as properly measured using actual of supermarkets are very different from independent trad-
prices in outlet F. The overlap method described in para- ers and a target imputation is illustrated for outlet F using
graphs 6.90–6.118 links in a replacement variety’s price independent traders. Outlet F might also have been imputed
change and can be used for permanently missing variet- using the price index at a higher level of aggregation or
ies. The overlap method does not have this self-correcting even a single outlet’s price. Similar principles of aggrega-
feature and should not be used for temporarily missing tion apply.
varieties. 6.64 Typically, a price collector reports a variety price
as temporarily missing; this is then passed to the head office
Carryforward Imputation for confirmation and then, perhaps, further at a higher level.
If confirmed, a decision is made as to use a targeted or over-
6.61 Carrying forward the last observed price should be
all imputation. If the overall mean imputation is chosen,
avoided and is acceptable only in the case of fixed or regu-
an appropriate computer routine is applied which enters
lated prices. Special care needs to be taken in periods of high
the imputed price with a designation as being imputed into
inflation or when markets are changing rapidly as a result of
the data system. If a targeted imputation is used, the CPI
a high rate of innovation and product turnover. While simple
compiler selects the variety rows regarded as likely to have
to apply, carrying forward the last observed price biases the
similar price changes, and the imputation is applied accord-
resulting index toward zero change. In addition, when the
ingly. For quality assurance, the computer routine should
price of the missing variety is recorded again, there is likely
record the decisions made and tabulate the number of tem-
to be a large compensating step change in the index to return
porarily missing varieties by elementary aggregate and their
to its proper value. The adverse effect on the index will be
treatment.
increasingly severe if the variety remains unpriced for some
6.65 Imputations are preferable to simply omitting
length of time. In the example in Table 6.3, if carryforward
the missing price observation for the calculation of the
imputation was used the missing prices in March 2020
index. For example, consider a variety priced at 4 in Jan-
would be imputed with the February 2020 price of 5.99,
uary, temporarily missing in February, and returning at 6
as would be the imputed prices in April and May 2020. In
in March. If the price change between January and Feb-
June, on the variety’s return, there would be a step increase
ruary for the remaining varieties in the elementary aggre-
in price from May to June of 5.99 to 6.25. In general, car-
gate was 25 percent (a price relative of 1.25). The imputed
rying forward is not an acceptable procedure or solution to
variety price for February would be 1.25 × 4 = 5. The
this problem. Exceptions may be well-established and well-
actual price is not known, but the imputation serves as a
advertised periodic increases of fixed or controlled prices
benchmark without impairing the 25 percent price index
and tariffs.
measure based on observed matched prices. The February
to March price relative for the missing variety is 6/5 =
General Considerations 1.20. This is referred to as a self-correcting imputation;
6.62 As a general principle, temporarily missing prices it self-corrects in the sense that the long-term February to
require an explicit imputation entered into the data compi- March calculation would allow the index to return to its
lation. The overall mean imputation should, by default, be appropriate level: 5/4 × 6/5 = 6/4 = 1.5. Similar principles
based on a higher level of aggregation; however, it also may apply for the treatment of seasonal products as outlined in
refer to a variety within a region or type of outlet. The higher Chapter 11. Simply omitting the February price, and bas-
level of aggregation may comprise more than one variety, ing the February to March price relative on matched prices
some with different price changes. For example, if the miss- in these two months, would not have this self-correcting
ing price observation is for “canned tuna,” where the higher property.
weighted level aggregate is “canned fish” which includes 6.66 Omitting a missing variety’s price from the calcu-
“canned tuna” and “canned salmon,” then subject to a suf- lation is equivalent to an overall imputation using the other
ficient sample size, the imputation should be based on price matched prices in the elementary aggregate. The entering
movements of “canned tuna.” of an imputed value allows for the flexibility of using a tar-
6.63 An overall mean imputation benefits from an geted imputation, if desired.
automation of its implementation, and the transparency of 6.67 The designation of an imputed value by a differ-
this methodology contributes to the integrity of the index. ent tab or color clearly shows in the prices database the
By using an overall imputation, NSOs guard themselves extent to which imputations are used, the length of their
against criticisms of influencing the CPI by their choice runs, and product codes where they are overused. Sum-
of “similar varieties,” particularly where there are miss- mary counts of imputed prices should be monitored by
ing prices for heavily weighted elementary aggregates. type, product code, frequency, and duration as part of qual-
However, this method should not be used when there are ity assurance. This is particularly important to identify if
strong a priori or empirical grounds to believe a targeted or where temporarily missing variety prices have continu-
imputation would improve the results. NSOs should have ally imputed values well over a three-month run, with the
retrospective monthly price data available at higher lev- possibility of continuing sample depletion as other variety
els of aggregation and be able to examine differences in prices go missing.
129
The Nature of Quality Change provided at off-peak times. The fact that peaks exist shows
that purchasers or users attach greater utility to the products
6.68 Understanding the meaning of quality change at these times and reflect supply-side pressure. Other differ-
requires well-defined conceptual and theoretical back- ences, including the conditions of sale and circumstances
ground, so that adjustments to prices for quality differences or environment in which the products are supplied or deliv-
are made against this framework, as described in this section. ered, can make an important contribution to differences in
6.69 Over time, the quality of what is produced changes. quality. For example, a retailer may attract customers by
For example, new automobiles typically have an increasing providing free delivery, financing, or better variety, by being
number of options and become more reliable, durable, safe, more accessible, by offering shorter order times, smaller tai-
powerful, and economical. Another example is smartphones, lor-made orders, clearer labeling, better support and advice,
as each new model includes faster processing speeds and more convenient car parking, or a wider range of brands,
power, more memory, improved screen resolution, and other or simply by operating in a more pleasant or fashionable
technological advances such as facial recognition. In match- environment. Although these sorts of benefits are not always
ing the prices of a sample of models selected in a price ref- specified in the variety description, such quality improve-
erence period with the same models in subsequent periods, ments should conceptually not be outside the scope of the
the quality mix remains constant to avoid affecting the price index.
measurement with quality differences. However, the result- 6.72 To consider how to adjust prices for quality
ing sample of models gives less emphasis to newer models changes, it is first necessary to define quality. While there
that have benefited from more recent technological change may be an intuition as to whether a variety consumed in one
and have different price changes given the quality of ser- period is better than its counterpart in the next, a theoretical
vices they provide. framework will help in establishing the basis for such com-
6.70 Observed changes in prices arise in theory from parisons. For example, a variety of clothing is sampled and,
several sources, including quality changes, changes in tastes after a few periods, is missing. One option is to replace it
and preferences, and changes in the technology of produc- with a similar variety, but the most comparable option may
ers. Price differences of similar products are often taken to have more cloth in it, or have a lining, a different color or
be measures of differences in quality. However, differences buttons, better stitching, or be considered more fashionable.
in prices are often observed for varieties of the same quality. There is a need to put a price estimate on the difference in
This may arise for a number of reasons: quality between the old and new varieties so that the price
of like can be compared with like. To propose or criticize
• Some consumers may be unaware of the availability of a quality-adjustment procedure requires some concept of
the same varieties at lower prices, as there may be “search what is ideally required and how it is done in practice.
costs” to exploring the market to discover lower priced 6.73 In Chapter 4 of Consumer Price Index Theory,
varieties. a cost of living index is defined as the ratio of the mini-
• There may be price discrimination because the seller is mum expenditure in the base and current period required to
able to charge different prices to different categories of achieve a given standard of living or utility. Quality adjust-
consumers, such as movie tickets for children and senior ments to prices involve attempting to measure the price
citizens. change for a product that has exhibited some change in its
• Prices may be sticky with some retailers changing their characteristics from an earlier period that provides a differ-
prices infrequently to avoid the costs of doing so, includ- ent level of utility to the consumer. Equating of the value
ing adverse customer reaction, or as strategic competitive of a quality change with the change in utility derived by
behavior, such as loss leaders, leading to different retail- the consumer, while falling naturally under a cost of living
ers changing prices at different times. index framework, is not exclusive to it. A cost of goods index
• Cases where there are parallel markets, an official market can also benefit from regarding quality in this way. While a
subject to government or official control at which prod- cost of goods index requires the pricing of a fixed basket of
ucts are rationed and an unofficial unregulated market. products, some varieties will become unavailable and the
The unofficial market may be at a lower price because it replacement varieties selected to maintain the sample may
avoids taxes and regulations, or at higher price since the not be of the same quality. A cost of goods index based on
official price is a subsidized one, but has limited, possibly a fixed-basket concept has the pragmatic need to adjust for
varying, quantities available for sale (System of National quality differences when a variety becomes unavailable, and
Accounts 2008, paragraphs 15.64–15.75). in the definition of a fixed-basket index does not preclude
differences in utility being used as a guideline. If variety
6.71 In addition to the changing mix of the character- A is better than its old version, variety B, it is because it
istics of varieties, there is also the practical problem of not delivers more utility to the consumer who in turn is willing
always being able to observe or quantify characteristics to pay more.
such as the style, reliability, ease of use, and safety of what 6.74 The definition of a quality change is based on
is produced. The same product provided at a different and equating some change in characteristics to a different level
more convenient location may result in a higher price and be of utility provided. For example, consider the case in which
considered to be of a higher quality. Furthermore, different a new variety with improved quality is made available, and
times of the day or periods of the year may also give rise the consumer has to choose between the old and new variet-
to quality differences, for example, electricity or transport ies in period t. If both varieties were offered to the consumer
provided at peak times must be treated as being of higher at the same price, pt = 100, the consumer would naturally
quality than the same amount of electricity or transport prefer the new variety. If the price of the old variety was
130
then progressively reduced until pt* = 75, at which the con- Informed outlet staff should be asked to confirm that the
sumer was indifferent between purchasing the old variety missing variety is permanently missing and help identify a
(at pt* = 75) and the new variety (at pt = 100), the consumer best-selling replacement variety, its specifications, and how
might select the old variety or the new one and would obtain these specifications differ from those of the old variety. For
the same utility from both. Any further decrease below pt* logistical reasons the selected replacement variety would
= 75 would cause the consumer to switch back to the old be expected to have high sales for the foreseeable future.
variety. The CPI staff at the head office would confirm or reject the
6.75 The difference between pt and pt* would be a mea- choice of replacement variety.
sure of the additional utility that the consumer placed on
the new variety as compared with the old. It would mea-
sure the maximum amount that the consumer was prepared Comparable Replacement
to pay for the new variety over and above the price of the 6.78 The price collector should use the missing vari-
old variety. In economic theory, if consumers are indiffer- ety’s specification and identify a comparable variety with
ent between two purchases, the utility derived from them the same specifications, for example, a washing machine
is the same. The difference between pt and pt* (75 and 100) with the same spin speed, capacity, brand, or equivalent
must therefore arise from the consumers’ valuation of the brand. If a comparable replacement exists, its detailed speci-
difference in utility they derive from the two varieties: their fications should be confirmed by the price collector against
quality difference. the specifications for the missing variety. Any changes in
6.76 The utility-based framework provides insights into the specification deemed to be not sufficiently price-deter-
the question of how consumers might choose between vari- mining should be noted for the head office to confirm, for
eties of different qualities. Consumers derive more utility example, color and trim.
from a variety of higher quality than from a variety of lower 6.79 The comparable replacement method requires the
quality, and thus they prefer it. But this does not explain why price collector to make a judgment that the replacement is of
consumers buy one variety rather than the other. For this, it a similar quality to the old variety and any price changes are
is also necessary to know the relative price of one variety unaffected by changes resulting from quality differences. In
with respect to the other, since if the lower-quality variety is Table 6.5A, there is a comparable variety F1 to replace vari-
cheaper, it may still be purchased (pt* ≤ 75 in the previous ety F, also from outlet F. The replacement variety is consid-
example) and to determine the price below which the old ered by the price collector and confirmed by the head office
quality would be purchased. as being directly comparable, and its prices in May and June
(respectively, 6.20 and 6.25) are entered into the data system
Permanently Missing Price as a continuation of the outlet F series. The price index is
Observations calculated using short-term price relatives built into a long-
term price index. The price index in July 2020 (December
6.77 Tables 6.5A–C illustrate the treatment of perma- 2019 = 100) is 105.52, a 5.52 percent increase over this
nently missing price observations. In these tables, prices are period. The price index remains as a constant-quality index
observed for outlets A to E over the seven months of Decem- since in May and June 2020 the prices of like quality variet-
ber 2019 to June 2020. For outlet F, the price observations ies continue to be compared with like.
for the variety is reported as permanently missing from 6.80 A common practice of manufacturers of electronic
May 2020, and the price collector has to find a replacement goods, such as televisions, household appliances, comput-
variety (the prices in bold for May and June are imputed ers and computer-related hardware and software, and of
as explained in the following text). The use of replace- automobiles is to have major quality changes in some years
ment varieties maintains the original sample sizes at the last but relatively minor ones in other years. A new “compara-
rebasing and the representativity of the varieties selected. ble” model would have a new model number with a new
Table 6.5A Illustration of Treatment of Comparable Replacements

Dec.-19 Jan.-20 Feb.-20 Mar.-20 Apr.-20 May-20 Jun.-20
A Supermarket 5.25 5.25 5.49 5.49 5.49 5.49 5.49

B Supermarket 5.10 5.10 5.10 5.25 5.25 5.25 5.25
C Supermarket 5.20 5.20 5.20 5.20 5.20 5.25 5.25
D Independent Trader 5.49 5.49 5.49 5.65 5.75 5.80 5.80
E Independent Trader 5.99 6.50 6.50 6.90 6.90 6.90 6.90
F Independent Trader 5.99 5.99 5.99 6.13 6.15 6.20 6.25
Permanently Missing: F, May, June; F1: Comparable Replacement
Geometric Mean: A:F 5.49 5.57 5.61 5.74 5.76 5.79 5.79
Short-Term Price Relatives: A:F 1.00000 1.01371 1.00748 1.02370 1.00347 1.00440 1.00134
Long-Term Indices as 100.00 101.37 102.13 104.55 104.91 105.37 105.52
Product of Short Term
Values in bold are imputed.
131
production run, though physically not much has changed. price change is considered as change due to quality dif-
The method of comparable replacement relies on the effi- ference. The most commonly used implicit method is the
cacy of the price collectors and head officer and, in turn, on overlap method. The replacement’s price change is linked to
the completeness of the specifications used as a description the old variety’s price change using an overlap period that
of the varieties. NSOs may tend toward designating replace- includes both the old and the replacement variety’s price.
ments as comparable since they are concerned with sample Where an overlap price for the replacement variety does not
sizes being reduced by dropping varieties, and the intensive exist, it might be imputed.
use of resources to introduce noncomparable replacements 6.84 Explicit (or direct) quality-adjustment methods
or make explicit estimates of quality differences. The use of directly estimate the value of the quality difference between
varieties of a comparable specification has practical advan- the old and new product and adjust one of the prices accord-
tages. However, if the quality of varieties is improving, the ingly. Pure price change is then estimated as the difference
preceding variety will have inferior quality compared to the in the adjusted prices. Explicit methods include quantity
current one. Continually ignoring small changes in the qual- adjustments, option/feature costs, and “patched” hedonic
ity of replacements can lead to an upward bias in the index. regression methods.
The extent of the problem will depend on the proportion of 6.85 Some of these methods are complex, costly, and
such occurrences, the extent to which comparable replace- difficult to apply. The methods used should as far as possible
ments are accepted as being so despite quality differences, be based on objective criteria.
and the weight attached to those varieties. Chapter 8 includes
proposals to monitor types of quality-adjustment methods
by product group providing a basis for a strategy for apply-
Implicit Methods of Quality
ing explicit adjustments where they are most needed. Adjustment
6.86 This section discusses the following implicit meth-
Noncomparable Replacements ods for obtaining adjusting for quality differences: overlap
method, class mean imputation, and link-to-show-no-change.
6.81 Noncomparable replacements result when the
price-determining characteristics of the replacement vari-
ety are different from those of the old variety. This means Overlap Method
that the collected price of the replacement variety cannot
be compared directly to the price of the old variety because The Use of an Overlap Price
the difference in these prices reflects not only pure price 6.87 A numerical illustration of the overlap method is
change, but also differences because of changes in quality. shown in Table 6.5B. In this example, in outlet F there is a
Noncomparable replacements require some form of quality preexisting old model F available up to April 2020, and a
adjustment. noncomparable new replacement model, F2, available from
6.82 Methods of quality adjustment for prices are clas- May 2020, with the same actual price of 5.25 in both May
sified into implicit (or indirect) quality-adjustment meth- and June. The prices are lower than would be expected from
ods and explicit (or direct) methods. Implicit and explicit the prices of F, but this is a noncomparable replacement: it is
methods are discussed in paragraphs 6.90–6.188. Both a replacement variety with a major share and is expected to
decompose the price change between the old variety and its remain on the market in the foreseeable future.
replacement into quality and pure price changes. 6.88 The overlap method requires a price for both the
6.83 Implicit (or indirect) quality-adjustment meth- old and new models in an overlap period. In the example in
ods estimate the pure price change component of the price Table 6.5B, F exists up to and including April, F2 exists in
difference between the old and new products based on the May, June, and thereafter. The question is how to determine
price changes observed for similar products. The difference an overlap price in this case. One source of information for
between the estimate of pure price change and the observed this overlap price (of 5.25 in April) is the price collector.
Table 6.5B Illustration of Treatment Using the Overlap Method, Noncomparable Replacements: Actual Preceding
Period Price
A Supermarket 5.25 5.25 5.49 5.49 5.49 5.49 5.49

B Supermarket 5.10 5.10 5.10 5.25 5.25 5.25 5.25
C Supermarket 5.20 5.20 5.20 5.20 5.20 5.25 5.25
F Independent Trader 5.99 5.99 5.99 6.13 6.15
Permanently Missing: F, May, June; F2: Noncomparable Replacement
F2 Noncomparable Replacement 5.25 5.25 5.25
Geometric Mean: April, May, June: A:E,F2 5.61 5.63 5.63
Short-Term Price Relatives 100.00 1.0137 1.0075 1.0237 1.0035 1.0030 1.0000
Long-Term Indices as Product of Short Term 100.00 101.37 102.13 104.55 104.91 105.23 105.23
132
Use the Actual Price of the Replacement in the 6.92 In Table 6.5B, the index from May onward no
Preceding Period, If It Exists longer uses F, but switches to F2. For this, there is a need
6.89 The price collector may have anticipated falling for the overlap prices: average prices up to and including
sales of a variety and the switch of consumers to a new model, April using the old variety F and for May, June, and onward,
brand, or variety, and recorded the price for the replacement using the new variety F2. The prices for outlet F in May
prior to its adoption, obtaining in this way an overlap price. and June (5.25 in both) are based on F2. The overlap in
In the example in Table 6.5B, the price collector would start April for the new model F2 is 5.25. This overlap price is
to record the price of the new variety F2 in April rather than used to calculate the index for May. Using the overlap price,
May. Price collectors should be trained to anticipate such the short-term price relative for April to May in outlet F is
changes, to corroborate them with informed outlet staff, 5.25/5.25 = 1.00000. This completes the price table. The
and to relay the information to the head office for possible geometric means are calculated as before, and their ratios
action. In the example, the price collector would have seen form the short-term price relatives, and, in turn, the cumu-
diminishing sales and a poorer positioning for display in the lative product of the price relatives, starting in December
outlet for the variety F, and it would have been apparent that it 2019 is the price index, at 105.23 for June 2020 (December
was to be replaced by F2. Outlet staff would confirm that 2019 = 100). The index from December to April is calcu-
F2 was to effectively replace F as a model aimed at that lated using A–F and the index beginning in May onward is
the same segment of the market, and the price for F2 and based on the prices of A–E and F2.
its quality characteristics should have been recorded along-
side that of F to provide an overlap April price to facilitate Imputed Overlap Prices
the introduction of F2 in May. As a general principle, the 6.93 In the previous example, the new replacement vari-
replacement of models is best not undertaken when the old ety may not have existed in April or the price collector may
model has limited sales and is at the end of its life cycle. not have been able to obtain a reliable estimate of its price.
6.90 Alternatively, the price collector may have asked In this situation, the overlap price in April can be imputed.
informed outlet staff in May whether the new model was The validity of this imputation is critical to the quality-
sold in the previous month to obtain an overlap price for adjustment methodology. In Table 6.5B, there was a price
April, or if sold in other outlets, whether there is a pricing of 6.15 for the old model and 5.25 for the new model for
agreement with the supplier that this outlet would have fol- April 2020. As shown in equation 6.6, this method implicitly
lowed had it been supplied to them, and what would have attributes the price difference in the common April overlap
been the price. The head office should confirm such details period as an indicator of the quality difference between the
by visit or contact with informed staff of outlet F. Table 6.5B old and new models.
shows 5.25 to be entered as an estimated price in April 2020 6.94 If the new model was not sold in April, an impu-
for the new model, to provide an overlap price in April for tation for the May price of the old model can be made to
the old and replacement models, F and F2. provide an overlap price. The imputation may be an overall
6.91 The price index is measured through and including mean or a targeted imputation following the principles out-
April 2020 using the prices of the old model: from Table lined for temporarily missing variety prices (as described in
6.5A, the geometric mean of the prices of the old model in paragraphs 6.52–6.72 and illustrated in Table 6.3). Table 6.5C
April 2020 is 5.76, and in March it is 5.74, with a short- illustrates a forward imputation of the old model’s price to
term price relative of 5.76/5.74 = 1.0035. The price index to provide an estimate of the price in May 2020 had it existed
April 2020 is the cumulated product of the old index’s price then. The imputed price is given in Table 6.5C as 6.17. It
relatives: for April 2020, it is 104.91 = 104.55 × 1.0035 is calculated by taking the ratio (relative) of the geometric
(December 2019 = 100). mean of the May prices of outlets A to E to the geometric
Table 6.5C Illustration of Treatment Using the Overlap Method, Noncomparable Replacements: Imputed Succeeding
Period Price
A Supermarket 5.25 5.25 5.49 5.49 5.49 5.49 5.49

B Supermarket 5.10 5.10 5.10 5.25 5.25 5.25 5.25
C Supermarket 5.20 5.20 5.20 5.20 5.20 5.25 5.25
F Independent Trader 5.99 5.99 5.99 6.13 6.15 6.17
Permanently Missing: F, May, June; F3: Noncomparable Replacement
F3 Noncomparable Replacement 5.25 5.25
Geometric Mean: May, June: A:F 5.76 5.78
Geometric Mean: May, June: A:E,F3 5.63 5.63
Short-Term Price Relatives 100.00 1.0137 1.0075 1.0237 1.0035 1.0037 1.0000
Long-Term Indices as Product of Short Term 100.00 101.37 102.13 104.55 104.91 105.30 105.30
133
mean of the April prices for the same outlets, 5.78/5.76 = considered as a means of quality adjustment where there is a
1.0037, and multiplying this by the old variety’s price in relatively high rate of frequent replacements, such as differ-
April. The index in Table 6.5C is calculated by multiplying ent models of automobiles launched each year.
the short-term May to June price relative for the new replace- 6.98 The class mean method is a form of targeted impu-
ment F3, 5.63/5.63 = 1.0000, by the value of the long-term tation, for the treatment of replacements for permanently
index for May, 105.30. The 5.63 are the geometric means of missing varieties, in which only the price changes of “com-
the prices in outlets A to E and that of the replacement F3. parable” replacements are used to impute the overlap price.
The comparable replacements are being limited to those that
Forward versus Backward Imputation have exactly the same price-determining characteristics, or
6.95 In Table 6.5B, the overlap took place in April, those varieties with replacements that have been declared
while in Table 6.5C, it was in May. In Table 6.5B, an actual comparable after review or have already been quality-
price was sought and found for the new (replacement) vari- adjusted through one of the “explicit” methods, as described
ety in April. There would be no equivalent May price for in paragraphs 6.120–6.188. For example, when the arrival
the permanently missing old variety. Given that an actual of a new model of a particular kind of motor vehicle forces
price may be preferred to an imputation based on the price price collectors to find replacements, some of the replace-
movements of varieties priced in other outlets, the use of a ments will be of comparable quality, and others comparable
backward price to provide an overlap in April can be used. with explicit quality adjustments, but imputed prices for
In Table 6.5C, a forward imputation is used for the old vari- an overlap month will be needed for the remaining ones.
ety’s price in May, which is also an acceptable means of Class mean imputation calculates imputed price relatives
making an imputation. But one could apply a backward using only the prices of comparable and, where appropriate,
imputation for the new (replacement) variety’s price in April explicitly quality-adjusted varieties or models. In general, it
to provide an overlap in April as the one in Table 6.5C. It is does not use the prices of the varieties or models that were
relatively straightforward to demonstrate algebraically that not replaced, because these are likely to be different from
the result would be the same if the index is calculated either those of the new models. The prices of old models tend to
way. Both methods impute the missing price using the price fall as they become obsolete, while the new models (rep-
movements of varieties in the outlets A to E for which prices resented by the replacements) tend to have a higher price
exist in both April and May. The backward imputation is before falling.
simply the inverse of the forward one. The ratio of prices 6.99 Class mean imputations rely on other explicit qual-
in the overlap periods, as shown by equation 6.6, is the ity adjustments and comparable replacements. The other
implicit measure of the quality differential between the old explicit quality adjustments may be from an available option
variety and the replacement variety, will be numerically the or feature prices and may be limited in nature, covering only
same for backward and forward imputations. As a result, some of the differences in product attributes, available for
either forward or backward imputations can be made. only a small proportion of unrepresentative model or variety
changes, and the availability of comparable replacements
may be limited. Given a substantial churn in the market and
Class Mean Imputation difficulties with such imputations and estimates, an alterna-
6.96 The overall imputation method has many advan- tive recommended approach is that of hedonic indices, as
tages considering the use of resources. It can be automated outlined in paragraphs 6.140–6.176.
as a default measure to readily link-in replacement variet- 6.100 In some cases, sufficiently large samples of com-
ies keeping the sample up to date. However, the (forward) parable substitutes or directly quality-adjusted varieties
imputation assumes that the price movements of existing are unavailable, or the quality adjustments and selection of
continuing varieties would be the same as that for the old comparable varieties are not deemed sufficiently reliable.
variety, if it had continued to exist. A backward imputation In that case, a targeted imputation may be considered. The
assumes the price movement of the new (replacement) vari- targeted mean is less ambitious because it seeks only to cap-
ety would be the same as existing continuing varieties had ture price changes of similar varieties, irrespective of their
the new variety existed in the period prior to its introduction. point in the life cycle. However, it is an improvement on the
Such assumptions are unlikely to be valid for high-technol- overall mean imputation, as long as sufficiently large sample
ogy goods being replaced at the end of their life cycle. An sizes are used.
alternative imputation procedure designed to mitigate this
problem is the class mean imputation which in principle, is Assumptions and Concerns on the Use of the
more suited to this context of imputing replacement variety Overlap Method
prices for permanently missing varieties, as opposed to tem- 6.101 The accuracy of the estimates obtained with the
porarily missing ones. overlap method depends on the validity of its underlying
6.97. The class mean imputation method is a specifically assumptions, as described in this section. If pmt and pmt+1
designed targeted imputation used to introduce a replace- denote the prices of an old variety m in periods t and t + 1,
t+2
ment when a variety’s price is permanently missing. The and pn denote the price of a new replacement variety n
class mean method of implicit quality adjustment to prices in period t + 2, an overlap can be made by imputing a price
arose from concerns that unusual prices were charged at the for the new replacement variety in period t + 1, pnt+1* . In the
start and end of a model’s life cycle. Thus, the price move- case where variety n replaces m and is of a different quality,
ment of continuing varieties appears to be a flawed proxy for the measured price index between periods t and t + 2 (shown
the pure price component of the difference between old and by the right-hand side expression in equation 6.7 is the
replacement varieties. A class mean imputation is mainly price change of the old to new variety between these two
134
periods, multiplied by (adjusted for) the price overlap for m than to compare the varieties. However, the replacement
to n in period t + 1 which the method implicitly takes to be a variety may have been sold in the previous period and outlet
measure of the quality difference. staff may have a record of its price.
6.102 A forward imputation is used in the previous 6.106 The underlying assumption in this case is that the
example for the price of the old variety in period t + 1. quality difference in any period is equal to the price differ-
Equation 6.7 also shows for this forward imputation that the ence at the time of the splice. The timing of the switch from
overlap method depends on the validity of the relative differ- the old variety m to the new variety n is thus crucial. Unfor-
ence in the prices of the old and new varieties in period t + 1 tunately, price collectors usually maintain a variety until the
as a measure of the quality difference between the varieties, decision to replace it is taken, so the switch may take place
and on the reliability of the imputed price, designated with at an unusual period of pricing, near the end of variety m’s
an asterisk (*), of the old variety m in period t + 1, pmt+1* , as life cycle and the start of variety n’s life cycle. This analysis
an estimate of pmt+1 . is more formally given in Annex 6.1.
6.107 Relative prices may not always reflect quality dif-
pmt+1* pnt+ 2 pnt+ 2 pmt+1* ferences. For example, a new replacement model or brand
I t ,t + 2 = × = × t +1 (6.7)
pmt pnt+1 pmt pn of an improved quality may be stocked and sold at the same
price as the old model, and the outlet competes in the market
6.103. Table 6.6 shows an old model m permanently in part by changing the quality of what is sold, as opposed
missing from March (t + 1) and replaced by a new model to the price. In other cases, retailers may reflect unusual
n in April (t + 2), with an overlap in March (t + 1). The pricing policies aimed at minority segments of the market.
price index for February (t) to April (t + 2), obtained using For example, the ratio of prices in an overlap period of a
the overlap method, is given by the first expression in equa- generic and a branded pharmaceutical drug may reflect the
tion 6.7 as the product of the old model’s price change needs of two different market segments, rather than quality.
between February and March and the new model’s price The overlap method can be used with a careful choice of the
change between March and April. This is equivalent to the overlap period. If possible, the overlap period should be a
second expression in equation 6.7, which is a direct price period before the use of the replacement, since in such peri-
comparison between the new and old models between Feb- ods the pricing may reflect a strategy to drop the old model
ruary and April with a quality adjustment as the value of to make way for the new one.
the relative prices in the overlap month March (t + 1). The 6.108 The overlap method is based on the law of one
overlap method implicitly values the quality difference as price: when a price difference is observed, it must arise from
the ratio of the two prices in the overlap period: a difference in quality or similar factors for which consumers
are willing to pay a premium, such as the timing of the sale,
pmMar* pnApr pnApr pmMar* location, convenience, or conditions. Economic theory would
I Feb,Apr = × = ×
pmFeb pnMar pmMar pnMar dictate that such price differences would not persist, given
that markets are made up of rational producers and consum-
30 38 38 30
= × = × = 1.163 (6.8) ers. However, as outlined in paragraph 6.74, there are many
288 35 28 35 reasons why identical varieties can be sold at different prices,
and the law of one price does not hold in practice. These
6.104 Moreover, for a longer-term price comparison, reasons include lack of information caused by search costs,
say January to June, the valuation of the quality difference price discrimination, and the existence of parallel markets.
remains as the price ratio in March, the time of the splice: 6.109 The overlap method is commonly used as a
default procedure for introducing replacements when vari-
pmFeb pmMar*  pnApr pnMay pnJun  eties are permanently missing. NSOs that rebase the CPI
I Feb,Apr = × Feb × Mar × Apr × May 
pmJan pm  pn pn pn  less frequently may experience sample deterioration with
many varieties becoming permanently missing and, with-
p Jun p Mar* 41 30 (6.9)
= nJan × mMar = × = 1.4057 out replacements, the sample becoming increasingly com-
pm pn 25 35 posed of imputed prices. Replacements serve to maintain the
sample composition and to update the representativity of the
6.105 The price of a missing variety is, by definition, varieties being priced.
not usually observed at the same time period as the price 6.110 In some cases, only noncomparable replacements
of the replacement variety, since the decision to replace the (that is, those of a different quality) of varieties with miss-
variety is only made after it has disappeared. Additionally, ing prices are available. In these cases, an explicit adjust-
the list of specifications is not always comprehensive, since ment to the price to account for the quality difference
the main aim is to identify the variety in the outlet rather should be made to compare the prices of the noncomparable
Table 6.6 Introducing a Noncomparable Replacement Table 6.6A Introducing a Noncomparable Replacement
via an Overlap to Illustrate Link-to-Show-No-Change
Model January February March April May June Model January February March April May June
Old (m) 25 28 30 Old (m) 25 28 28

New (n) 35 38 40 41 New (n) 35 38 40 41
Values in bold are imputed. Values in bold are imputed.
135
replacements in one period with the price of the now miss- That is, a (quality-adjusted) price change between February
ing varieties in a previous period. The widely used overlap (period t) and April (period t + 2) is measured as that for the
method has a major advantage of not requiring an explicit new variety between March (t + 1) and April (t + 2).
quality adjustment. Explicit quality adjustments (described 6.114 The bias is perpetuated through subsequent peri-
in paragraphs 6.119–6.187) are more resource-intensive ods of measurement. For example, the February to June
than the implicit overlap method. Furthermore, it is recom- index would still have its price change measured as the prod-
mended that the overlap method be automated with informa- uct of month-on-month price changes which would include
tion from the price collector, reported to the head office, and the February to March link that would show no price change.
used in a computational routine. The recommendation is that 6.115 The link-to-show-no-price-change method is not
in accepting the use of the overlap method, the CPI compiler recommended. As with carryforward, the method is par-
should ensure that relative prices at the time of the overlap ticularly harmful since it can be readily incorporated into a
reflect quality differences. For example, a much-improved regular automatic compilation routine and not noticed: the
new model of a smartphone may be launched at the same price of the replacement is automatically imputed to form
price as the old model, and the relative price would not the overlap price and the index compiled.
reflect the differences in quality. If the relative price does not
reflect quality differences and resources allow, an explicit
quality-adjustment method should be used.
Explicit Methods of Quality
6.111 The overlap method is implicitly employed when Adjustment
samples of varieties are rotated. That is, the old sample of 6.116 The methods described previously do not rely on
varieties is used to compute the category index price change explicit information on the value of the change in quality. This
between periods t − 1 and t, and the new sample is used section discusses methods that rely on obtaining an explicit
between t and t + 1. The “splicing” together of these index valuation of the quality difference: quantity adjustment; differ-
movements is justified by the assumption that—on a group- ences in production or option costs; and the hedonic approach.
to-group rather than variety-to-variety level—differences in
price levels at a common point in time accurately reflect dif- Quantity Adjustment
ferences in qualities. 6.117 Quantity adjustment is one of the most straight-
6.112 The bias in using the overlap method within an forward explicit quality-adjustment methods. It is applicable
elementary aggregate depends on (1) the ratio of missing when the size of the replacement variety differs from that of
to total observations, and (2) the difference between the the available variety, and any change in quantity is consid-
mean of price changes for existing varieties and the mean of ered a change in quality. In some situations, there is a readily
quality-adjusted replacement price changes. As noted pre- available quantity measure that can be used to compare the
viously in paragraph 6.91, imputations can be made either varieties. Examples are the number of units in a package
forward or backward. The bias decreases as either of these (for example, paper plates or vitamin pills) or the size or
terms decrease. A formal analysis is given in Annex 6.2. weight of a container (for example, kilogram of flour or liter
of cooking oil). Quantity adjustment to prices can be accom-
Link-to-Show-No-Price-Change plished by scaling the price of the old or new variety by the
6.113 Returning to the example in Table 6.6, repro- ratio of quantities. The index calculation system may do this
duced in Table 6.6A, where the new replacement variety is scaling adjustment automatically, by converting all prices in
noncomparable (that is, of a different quality), the link-to- the category to a price per unit of size, weight, or number.
show-no-price-change method imputes the price in March For example, if the weight of a candy bar is 450 grams in the
for the old model m to be the same as its February price, current period and 500 grams in the previous period, but the
28. The new variety is linked-in to show no price change in price remains unchanged, an adjustment is needed so that
the period of replacement: February to March. This method the index reflects the implicit price increase.
is used for the treatment of a noncomparable replacement 6.118 The specification of a variety is often to a specific
variety and should not be confused with carrying forward a size, for example, a one-kilogram packet of flour. If only two-
previous period price for the treatment of temporarily miss- kilogram packets are now sold in a specific outlet, the price col-
ing prices (described in paragraph 6.65). This method biases lector should choose a representative two-kilogram packet, but
the index downward when prices are rising and biases them mark the new specification as such and, after confirmation by
upward when (true, quality-adjusted) prices are falling. the head office, prices should continue to be collected for the
The link-to-show-no-price-change method attributes the two-kilogram packet. In this example, an adjustment would be
price difference between the new and old model in Febru- needed to ensure that the index reflects only pure price change
ary to quality difference. The new model is of a better qual- and not changes resulting from differences in quality (that is,
ity, valued at being worth an additional 7, from 28 to 35. the change in size). This is particularly important where price
The (quality-adjusted) price is therefore constant between variances are computed, or use is made of a Dutot price index
February and March. The February to April price change is number formula (that is, an elementary index formula sensi-
28 38 tive to the homogeneity of the varieties used).4
× = 1.0857 , an 8.57 percent increase in price.
28 35
4
In equation 6.6 A Dutot elementary price index number formula is a ratio of arithme-
tic means of matched prices. More weight is given to price changes with
pmt pt +2 pt +2 higher prices in the reference period. Thus, if the price of a container in the
pnt+1 = pnt* pmt+1* = pmt and I t ,t + 2 = t −1
× nt +1 = n t (6.10) price reference period doubles, while others remain the same, the implicit
pm pn pn weight given to its price change will also double (see Chapter 8).
136
6.119 When replacing a variety, changes in the size of constant [(2.25/0.75) = (1.5/0.5) = 3]; three per kilogram.
different varieties sold can be dealt similarly; however, there The approach can be outlined in a more elaborate manner as
are some caveats. For example, in the pharmaceutical con- illustrated by Figure 6.1. The concern here is with the part of
text, prices of bottles of pills of different sizes differ: a bottle the unbroken line between the (price, quantity) coordinates
of 100 pills, each having 50 milligrams of a drug, is not the (1.5, 0.5) and (2.25, 0.75), both of which have unit prices of
same as a bottle of 50 pills of 100 milligrams, even though 3 (price = 1.5/0.5 and 2.25/0.75). There should be no change
both bottles contain 5,000 milligrams of the same drug. If in quality-adjusted price. The symbol Δ denotes a change.
there is a change, for example, to a larger size container, and The slope of the line is β which is Δprice/Dsize = (2.25 −
a unit price decrease of 2 percent accompanies this change, 1.5)/(0.75 − 0.50) = 3 (that is, the change in price arising
then it should not be regarded as a price fall of 2 percent from a unit [kilogram] change in size). The quality-(size-)
if consumers gain less utility from the larger and more adjusted price in period t − 1 of the old m bag, to make it
inconvenient containers. In practice, it will be difficult to equivalent to the new bag, n, is
determine what proportion of the price fall is attributable to
quality and what proportion to price. A general recommen- pˆnt−1 = pmt−1 + β∆size = 1.5 + 3(0.75 − 0.5) = 2.25 (6.11)
dation is not to automatically interpret unit price changes
arising from packaging size changes as pure price changes,
if contrary information is available. The quality-adjusted price change shows no change, as
6.120 Consider another example shown in Table 6.7: a before:
branded bag of flour previously available in a 0.5-kilogram
bag priced at 1.5 is replaced by a 0.75-kilogram bag priced pnt 2.25
= = 1.00 (6.12)
at 2.25. The main concern here is with rescaling the quanti- ˆpnt−1 2.25
ties. The method would use the relative quantities of flour in
each bag for the adjustment. The price may have increased 6.121 The approach is outlined in this form so that
by [(2.25/1.5) × 100 = 150] 50 percent but the quality- it can be seen as a special case of the hedonic approach
adjusted price (that is, price adjusted by size) has remained (discussed in paragraphs 6.140–6.176), where the price is
related to a number of quality characteristics of which size
may be only one.
6.122 Now assume that the 0.5-kilogram bag was not
Table 6.7 Example of Size, Price, and Unit Price of
Bags of Flour
available (missing) and a 0.25-kilogram replacement packet
was used, priced at 0.75, as shown by the continuation to
Size First First Unit Second Second the coordinate (0.75, 0.25) of the unbroken line in Figure
(kilograms) Price Price Price Unit Price
6.1 and in Table 6.7; the quality-adjusted prices would again
0.25 0.75 3 1.25 5 not change. However, if the unit (kilogram) prices were 5,
0.50 1.50 3 1.50 3 3, and 3 for the 0.25-, 0.5-, and 0.75-kilogram bags, respec-
0.75 2.25 3 2.25 3 tively, as shown in Table 6.7 and in Figure 6.1 (including
Figure 6.1 Quality Adjustments for Different Sized Varieties
2.5
2
∆price
1.5
∆size
Price
β=∆price/∆size
1
0.5
0
0 0.25 0.5 0.75
Size in kilograms
137
the broken line), then the measure of quality-adjusted price z characteristic where ∆z = ( znt − zmt−1 ). The characteristics
change would depend on whether the 0.5-kilogram bag was
may be the size of the memory (RAM) of a computer when
replaced by the 0.25-kilogram one (a 67 percent increase) or
a specific model of computer is replaced by a model that
the 0.75-kilogram one (no change). This result is not satis-
is identical except for the amount of RAM it possesses.
factory because the choice of replacement size is arbitrary.
For example, the webpages of sellers of laptops allow buy-
The rationale behind the quality-adjustment process is to
ers to customize their purchase, an extra four gigabytes of
separate pure price change from changes caused by differ-
RAM, from eight to twelve gigabytes, for a specific brand
ences in quality (in this case, quantity changes).
and model of a laptop may cost an additional ¥70. Con-
sider that the standard laptop used for CPI measurement has
Differences in Feature/Option Costs eight gigabytes of memory and costs ¥900 and is not avail-
6.123 Consider an example of the price of an option able in the next period. The new standard model in period
being used to adjust for quality. Let the prices for a variety t has 12 gigabytes but costs the same ¥900, pnt . To com-
in periods t − 1 and t be 10,000 and 10,500, respectively, pare the (constant-quality) price of the new model with the
but assume the price in period t is for the variety with a old model in t − 1, the latter should have its price adjusted
new feature, as standard, that previously in period t − 1 had to include an extra four gigabytes of RAM. The price of
to be purchased as an “option” for an additional 300. Then an additional gigabyte of memory for this brand/model in
between periods t − 1 and t, the price change including the period t − 1 is 70/4 = 17.5, and its quality-adjusted price
feature in both periods would be 10,500/10,300 = 1.01942 inˆ period t − 1 is pmt−1 = 900 + 17.5× (12 − 8) = 970 . The
or 1.942 percent. period t (unchanged) price of ¥900 is now compared with
6.124 Option costs are useful in situations in which the its comparable period t − 1 price to yield a constant-quality
old and new varieties differ by quantifiable characteristics price change of 900/970 = 0.9278, which is a price fall of 7.22
that can be valued in monetary terms by reference to mar- percent, while the package price is constant.
ket prices. The valuation of a quantifiable product feature 6.128 The previous description of the calculation is
may be readily available from the comparison of different more complex than required: the adjustment is to simply add
product prices. This is especially convenient because some 70 to the old price, 70 + 900 = 970. However, it serves to
goods and services sold on the internet can be identified by demonstrate some limitations of this approach as a special
their brands and price-determining characteristics. case of the hedonic method, as outlined in the section on
6.125 Consider the addition of a feature to a product, hedonic prices.
for example, an automatic icemaker in the door of a refrig- 6.129 This calculation conveniently makes the quality
erator, and refrigerators for a particular brand may be sold adjustment to the old model’s price in period t − 1 so that
as standard or with a door-installed automatic icemaker. the new model’s price in future months can be directly com-
The price collector may always have collected prices on the pared with the quality-adjusted old price for the life of the
standard model, but this may no longer be in production, new specification. However, the required information on the
being replaced by a model with an installed automatic ice- value of the option cost (an extra four gigabytes) may only
maker. The cost of the option is thus known from before and be available in period t and not be applicable to a period
a continuing series can be developed by simply adjusting the t − 1 adjustment. NSOs should ideally keep a record of, for
old price in the price reference period to include the option example, web customizations of specified varieties along
price. However, this process may have problems. First, the with comparable/noncomparable replacements especially
cost of producing an option as standard may be lower than for products with a high degree of technical change and
when it was an option, and this saving may be passed on, turnover of models and maintain good relations with outlet
at least in part, to the consumer. The option cost method staff.
would thus understate the price increase. Further, by includ- 6.130 In the previous example, if the relationship
ing an option as standard, the consumer’s valuation of it may between price and RAM is linear, the previous formulation
fall since buyers cannot refuse it, and some consumers may is appropriate. Many webpages give the price of additional
attribute little value to the option. The overall effect would RAM as being independent of other features of PCs, and a
be that the estimate of the option cost, priced for those who linear adjustment is appropriate. A linear formulation values
choose it, is likely to be higher than the implicit average the worth of an additional fixed amount of RAM to be the
price consumers would pay for it as standard. Estimates of same, irrespective of the amount of RAM the computer pos-
the effect on price of this discrepancy should in principle be sesses or a number of other features.
made, though in practice they are difficult to quantify. 6.131 The relationship between price and the product
6.126 Quality differences are not necessarily positive, features may be nonlinear. Denote the price-determining
for example an airline may charge for a second piece of characteristics as z and assume there are k of them. The
baggage when previously it did not. Again, there will be an change in z is intended to reflect the service flow, but the
option price available for the additional piece of baggage so nonlinearity in the price–z relationship may reflect consum-
that the price of like—two pieces of baggage—is compared er’s decreasing marginal utility to the scale of the provision.
with like. In the previous example, the price a customer is willing to
6.127 Option cost adjustments can be seen to be simi- pay per gigabyte falls as increasing amounts of gigabytes
lar to quantity adjustments, except that instead of size being are purchased. For some features, there will be economies
the additional quality feature of the replacement, the added of scale: supplying much more of a feature makes the price
quality can be an individual option/feature. The compari- fall, possibly substantially; while for others, it may become
son is pˆnt pmt−1 where pˆmt−1 = pmt−1 + β∆z for an individual technically difficult, and more expensive, to compress
138
higher amounts of a feature into the available space. The and import price indices. As an example of the practical use
data should reveal some of this relationship, and caution is of this approach, an NSO uses production cost estimates to
needed against applying linear relationships outside of the value quality adjustments arising in from model changes in
range in which they are warranted. Further, data should give new vehicles. Allowable product changes for the purpose
some insight into the required adjustments for such nonlin- of quality adjustments include reliability, durability, safety,
ear relationships, though this may be better estimated using fuel economy, maneuverability, speed, acceleration/decel-
a regression formulation and nonlinear specification, as con- eration, carrying capacity, and related changes and addi-
sidered in the section on hedonic prices. tional parts required to accommodate the principal change
6.132 The similarity between the quantity adjustment in a component. Only price-determining characteristics are
and the option cost approaches is apparent since both relate included, and any characteristics or features that do not
price to some dimension of quality: the size or the option. affect or impact the price are excluded. Also excluded for
The option cost approach can be extended to more than one the CPI, unlike the producer price index, are changes man-
quality dimension. Both approaches rely on estimates of the dated by the government that provide no direct benefit to
change in price resulting from a unit change in the option or the purchaser, including modifications to meet air pollution
size: the β slope estimates. In the case of the quantity adjust- standards. When a new model of a specified automobile is
ment, this was taken from a variety identical to the one being introduced, its changes in quality components are identified,
replaced, aside from the fact that it was of a different size. valued, and added to the price of the old model so that the
The β slope estimate in this case was perfectly identified price of the old and new models can be compared on a like-
from the two pieces of information. It is as if the nature of to-like basis.5
the experiment controlled for changes in the other quality 6.135 A critical feature of this method is its reliance on
factors by comparing prices of what is essentially the same estimates of the retail margin for the new components. With
thing except for the quantity (size) change. the option cost approach, a consumer’s valuation of the new
6.133 This same reasoning applies to option costs. For feature was available. If only production cost data are avail-
example, two varieties are identical except for a single fea- able, estimates of the retail markup must account for the
ture. Their difference in price allows the value of the fea- (average) age of the models under consideration. Markups
ture to be determined. Yet sometimes the value of a feature will decrease as models come to the end of their life cycles.
or option has to be extracted from a much larger data set. Therefore, retail markups based on models at the start of
This may be because the quality dimension takes a relatively their life cycle should not be applied to the production costs
large range of possible numerical values without an imme- of models during their life cycle, and particularly at the end.
diately obvious consistent valuation. Consider the simple Moreover, estimates of the retail margin of a component
example of only one feature varying for a product, the speed may well not be available. A pragmatic practice in one NSO
of a computer. It is not a straightforward matter to determine is to use the proportionate retail markup on the vehicle. The
the value of an additional unit of speed. To complicate mat- proportionate retail markup is calculated based on the price
ters, there may be several quality dimensions to the variet- charged by the manufacturer to the dealer for the identical
ies and not all combinations of these may exist as varieties vehicle and the manufacturer’s suggested retail delivered
in the market in one period. Furthermore, the combinations price for the equipped vehicle.
existing in the second period being compared may be quite
different from those in the first. Considering these aspects Hedonic Approach: Patching
leads to a more general framework, known as the hedonic
approach. 6.136 The hedonic approach is an extension of the two
preceding approaches: the feature/option cost and the pro-
duction cost approach. First, in the hedonic approach, the
Differences in Production Costs change in price arising from a unit change in quality (that
6.134 An alternative approach to quality adjustment is, the quantity or option/feature) is estimated from a data
is to adjust the price of an old variety by an amount equal set comprising prices and quality characteristic values of a
to the resource costs of the additional features of the new larger number of varieties. Second, the quality characteristic
variety. An important source of such data is the manufac- set is extended to cover, in principle, all major characteris-
turers. In this approach, the NSO can ask the manufacturer tics that might determine price, rather than just the quantity
to provide data on direct and indirect production costs for or option/feature adjustment.
the embodied quality change which would include research 6.137 The hedonic approach is particularly useful when
and development (R&D) costs, assembly and installation the market does not reveal the price of the quality character-
associated with the change, the manufacturer’s established istics required for the adjustment. Markets reveal prices of
markup, the retail margin, and associated indirect taxes. varieties, not quality characteristics, and it is useful to con-
This method is similar to estimating market-equivalent sider varieties as tied bundles of characteristics. A sufficiently
option prices in the absence of market prices. This approach large data set of varieties with their characteristics and suf-
is most practicable in markets where there is a relatively ficient variability in the mix of characteristics between the
small number of manufacturers, and where updates of mod- varieties allows the hedonic regression to provide estimates
els are infrequent and predictable. Additionally, it can only of the implicit prices of the characteristics. For example, the
be successfully implemented if there is good communica-
tion between manufacturers and the NSO staff. It is particu-
larly suitable when the quality adjustments are also being 5
An illustration of the procedure used in the United States is available at:
undertaken to calculate the producer price index and export https://www.bls.gov/cpi/quality-adjustment/new-vehicles.pdf.
139
Figure 6.2 Scatter Diagram of Price against Capacity: Washing Machine Data
1,600
1,400
Price in Pounds Sterling
1,200
1,000
Price = - 436.23+117.3Capacity
R = 0.6556
800
600
400
200
4 6 8 10 12 14 16 18
Capacity (cotton, kg)
Source: example developed for Manual.
price of (clothes) washing machines will be listed, though a can be copied with relative ease, and automated using web
new (replacement) model for a brand may have a (cotton) scraping. Such detailed information is also available as
capacity load size not previously available, say 12 kilograms, scanner data (see Chapter 10).
instead of the preceding model’s 10 kilograms. To make an 6.140. Figure 6.2 is a scatter diagram relating the
explicit quality adjustment, the price of the additional two price (in pounds sterling, £) to the (cotton) capacity (kilo-
kilograms is required. The regression approach using a data gram) of models of washing machines sold in one coun-
set of many models’ prices and characteristics can estimate try (the data are from a well-known consumer magazine).
the price of additional kilogram of capacity from data for It is apparent that washing machines with larger capaci-
models of washing machines on their price, capacity, year ties have higher prices—a positive relationship. It is also
(age of model), color, running cost, and so forth. apparent from Figure 6.2 that there are several models of
6.138 Under the MMM, each price collector needs to washing machine with the same capacity but quite differ-
select a representative variety, record its price and specifica- ent prices, resulting from the fact that other features differ.
tions, and reprice the same variety in subsequent periods. For example, 12-kilogram capacity machine prices range
The extension required in the hedonic approach is that the from £754 to £1,349.
prices and price-determining characteristics should be col- 6.141 To estimate the value given to additional units
lected for all, or a large sample of models. The method is of capacity, an estimate of the slope of the line that best
particularly suitable when there are no immediately appar- fits the data is required. The equation of a straight line is
ent comparable replacements and the noncomparable ones
Price = βˆ0 + βˆ1 z1 .
vary in their characteristics over more than one variable. A
new model of car, household appliance, computer or related 6.142 The slope ˆβ1 is a measure of the change in
hardware and software, or telecommunications equipment, Price that arises from a one-unit change in the charac-
can differ from the old model in many respects, yet there is teristic, z1, Capacity. The ^ (hat) above denotes that it
a single price for each new and old model. This approach is is estimated from the data. The estimated slope is from
particularly necessary when there is a frequent turnover of the equation of a line that best fits the data (that is, that
models in the market, where new models with quite differ- best represents the underlying pattern of the relationship).
ent values for their characteristics are frequently replacing In Figure 6.2, the equation of the line that best fits the
old ones. data was derived using an ordinary least squares (OLS)
6.139 The requirement that data be collected on the regression. The intercept and slope of the line that best
prices and specifications of a large sample, if not all mod- fits the data are estimated as ones that minimize the sum
els, is not as demanding as it might appear. Extensive data of the squared differences between the individual prices
on prices and characteristics of models of consumer goods and their counterpart prices predicted by the line: the least
and services are generally readily available on websites (for squares criterion. Tools for regression analysis are avail-
example, many comparing prices and salient characteristics), able on standard statistical and econometric software, as
140
Table 6.8 Estimated (Linear) Equation of Price against Capacity: Washing

Machine Data
Coefficients Standard t Stat P -Value
Error
Intercept −436.229 86.310 −5.054 3.09E−06

Capacity 117.298 9.949 11.789 1.43937E−18
Table 6.9 Illustrative Hedonic Regression Estimates for Washing Machines

Linear Log-Linear
Coefficient Std. Error p-Value Coefficient Std. Error p-Value
(Intercept) −206.939 112.3 0.06986+ 5.217 0.17 < 2E−16***

Age −1.579 2.1 0.44971 −0.006 0.00 0.054218+
Capacity 81.024 13.2 5.32E−08*** 0.108 0.02 4.86E−07***
Warranty −138.651 48.7 0.00592** −0.265 0.07 0.000466***
Steel 144.036 74.0 0.05608+ 0.266 0.11 0.017575*
Energy Cost 10.103 3.4 0.00430** 0.019 0.01 0.000353***
LG −115.816 43.9 0.01044* −0.220 0.06 0.001167**
Steam 191.196 92.3 0.04233* 0.257 0.14 0.063056+
Hyg_AllergyP 63.627 40.2 0.11842 0.153 0.06 0.012198*
LED Display 49.409 52.9 0.35391 0.166 0.08 0.036833*
R2 0.701 0.721
F-Statistic 20.25 22.22
P -Value 1.06E–15 2.20E−16
***,**,*, and + denote statistically significant at 0.1, 1, 5, and 10 percent levels, respectively.
well as spreadsheets.6 The estimated (linear) equation in this 6.146 The coefficient of 0.174 has a useful direct inter-
instance is presented in Table 6.8. pretation: when multiplied by 100, it is the percentage
change in price arising from a one unit (kilogram) change in
Price = −436.229 + 117.298 Capacity R 2 = 0.65 (6.13) capacity. There is an estimated 17.4 percent change in price
for each additional kilogram of capacity.
6.147 The range of prices for a given capacity was noted
6.143 Formula 6.13 is the estimated regression equa- to be substantial which suggests that other quality character-
tion of Price on Capacity; although there are many other istics may be involved. Table 6.9 provides the results of a
price-determining variables, this regression equation only regression equation that relates price to a number of quality
includes Capacity, for illustration. In Table 6.9, the regres- characteristics as listed in the Column A.7 While the results
sion model is expanded to include other variables. are given for both linear and semilogarithmic regression
6.144 The coefficient on Capacity is the estimated slope specifications, the focus here is on the latter functional form.
of the line: the change in price (£117.30) resulting from a 6.148 A semilogarithmic hedonic multiple regression
one-kilogram change in Capacity. This can be used to esti- model is given by
mate quality-adjusted price changes for washing machines
of different capacities. The value of R 2 (that is, the adjusted
coefficient of determination) is 0.65; this indicates that ln p = β0 + z1β1 + z2 β2 + z3 β3 + ....zn βn + ε (6.15)
65 percent of price variation is explained by variation
in Capacity. A t-statistic to test the null hypothesis of the where ԑ is an error term assumed to have the usual properties
coefficient being zero was found to be 11.789: recourse to to satisfy OLS assumptions (see paragraph 6.171). For this
standard tables on t-statistics found the null hypothesis was semilogarithmic form, logarithms are taken only of the left-
rejected with a p-value of 1.43937E−18. The fact that the hand side variable (that is, Price). Each of the z characteris-
estimated coefficient differs from zero cannot be attributed tics enters the regression without having logarithms taken.
to sampling errors at this level of significance. There is a This has the advantage of allowing dummy variables for hav-
very low probability that the test has wrongly rejected the ing or not having a feature included on the right-hand side.
null hypothesis.
6.145 Hedonic regressions should generally be con-
ducted using a semilogarithmic formulation. The focus is 7
These include Age (months since model was launched); Capacity (for
thus on the semilogarithmic form. The estimated (semiloga- cottons, kilogram); Warranty (if five years, benchmark: two years); Steel
rithmic) regression equation in this instance is (stainless steel outer); (Annual) Energy Cost, in £; LG (manufactured,
benchmarked on Samsung); Steam (wash/refresh); Hygiene/Allergy
program; LED display (benchmarked on LCD); and price for 74 models
log (Price) = 4.776 + 0.174 Capacity R 2 = 0.61 (6.14) as advertised on November 2017 in Which? for three up-market brands:
Brand A, Brand B, and Brand C. A much larger general regression model
with more variables was first estimated but reduced to this smaller specific
6
The illustrative empirical work in this section was undertaken using R, model using standard econometric principles and practice. The White
though it would be equally applicable with any such standard statistical (studentized) Breusch–Pagan test for homoscedastic residuals was not
software including EViews, SAS, and STATA. rejected at conventional significance levels with a p-value of 0.2197.
141
Such dummy variables assume the value of one if the variety perception that these represented estimates of user value as
has the feature and zero otherwise. The taking of logarithms opposed to resource cost. The former is the relevant concept
of the first equation in 6.15 allows it to be transformed in in constructing a CPI, while for producer price index com-
the second equation to a linear form, and a conventional OLS pilation it is the latter. Yet hedonic coefficients may reflect
estimator can then be used to yield estimates of the logarithms both user value and resource cost, both supply and demand
of the coefficients. These are given as the coefficients for the influences. There is an identification problem, as referred in
semilogarithmic model in Table 6.9. The estimated coeffi- econometrics; the observed data do not permit the estima-
cients in Table 6.9 are based on a multiple regression model: tion of the underlying demand and supply parameters. What
for example, for Capacity, the estimated coefficient of 0.108 is being estimated is the actual point of intersection of the
is of the effect of a unit change in capacity on price, having demand curves of different consumers with varying tastes
controlled for the effect of other variables in the equation. and the supply curves of different producers with possible
The scatter diagram in Figure 6.2 clearly shows the inade- varying technologies of production.
quacy of relying on a single price-determining variable and 6.152 In many cases, the implicit quality adjustment to
this approach can be justified because it addresses this issue. prices arising from the use of the overlap method may be
The preceding estimated coefficient of 0.174 was based on inappropriate because the implicit assumptions are unlikely
only one variable and is different from this improved result. to be valid (as described in paragraphs 6.104–6.112). In such
6.149 When dummy variables are used, the coefficients, instances, the practical needs of reliable economic statistics
when multiplied by 100, are estimates of the percentage require explicit quality adjustments. However, the use of the
change in price, given by (eβ1 −1) × 100. For example, from hedonic approach may only be justified when the weight,
churn, and extent of the quality adjustment is substantial,
Table 6.9, Brand A models have a (e−0.219743 −1) × 100 = 19.73 due to the cost of implementing the method.
percent lower price than their benchmarked Brand B coun- 6.153 The proper use of hedonic regressions requires an
terpart, having controlled for other differences in their price- examination of the coefficients of the estimated equations to
determining characteristics as specified in the regression ensure their plausibility. It might be argued that the very multi-
equation.8 tude of distributions of tastes and technologies, along with the
6.150 The value R 2 = 0.721 is the proportion of variation interplay of supply and demand, that determine the estimated
in (the logarithm of) price explained by the estimated equa- coefficients, make it unlikely that “reasonable” estimates will
tion.9 A high value of R 2 can be misleading for the purpose arise from such regressions. For example, a firm may cut a
of quality adjustment, although such values indicate that the profit margin pertaining to a characteristic for reasons related
explanatory variables account for much of the price variation to long-term strategic plans; this may yield a coefficient on a
over a relatively large number of varieties in the period con- desirable characteristic that may even be negative. This situa-
cerned. This, of course, is not the same as implying a high tion does not invalidate the usefulness of examining hedonic
degree of prediction for an adjustment to a replacement variety coefficients as part of a strategy for evaluating estimated
of a single brand in a subsequent time period. Predicted values hedonic equations. First, there has been extensive empirical
depend for their accuracy not just on the fit of the equation, but work in this field and the results for individual coefficients
also on how far the characteristics of the variety whose price are, for the most part, quite reasonable. Over time, individual
is to be predicted are from the means of the sample. The more coefficients can show quite sensible patterns. Unreasonable
unusual the variety, the higher the prediction probability inter- coefficients on estimated equations are the exception and
val. Second, the value R 2 indicates the proportion of variation should be treated with some caution. Second, the CPI com-
in prices explained by the estimated equation. It may be that piler should rely more on an estimated equation whose coef-
0.90 is explained while 0.10 is not explained. If the dispersion ficients make sense, and which makes good predictions, than
in prices is very large, this still leaves a large absolute margin on one which may also predict well but whose coefficients
of prices unexplained. A high R 2 is a necessary, though not do not make sense. Third, if a coefficient for a characteristic
sufficient, condition for the use of hedonic adjustments. does not make sense, it may be due to multicollinearity, a data
problem, and should be examined using, for example, vari-
The Interpretation of Estimated Hedonic
ance inflation factors, to see if this is the case.
Coefficients
6.151 Concerning the interpretation of the coefficients
from hedonic regressions, there used to be a mistaken The Implementation of a Hedonic Quality
Adjustment
8
6.154 The implementation of hedonic methods to
There is some bias in these coefficients; and in the (semi)logarithmic estimate quality adjustments for matched noncomparable
equation, half the variance of each coefficient should be added to the coef-
ficient before using it. For the Brand A coefficient, the standard error from replacements can take two forms. The first form is referred to
Table 6.9 is 0.064687; its variance is (0.064687)2 = 0.00418: the adjust- as “patching”: undertaking a quality adjustment to the price
ment is to add 0.00418/2 to −0.219743, giving −0.21765; a lower price of of the old model to make it comparable with the new model.
−(e−0.21765 −1)×100 = 19.56 percent; see Triplett (2006) for further details. For many varieties, this can be seen as a one-off process for
9
More formally, where pit is the price of washing machine i in period t, individual varieties within the lifetime of updating a sample.
N
∑( p − pît ) / N , of
2
R2 is 1 minus the ratio of the variance of the residuals, t
i The second form is the more comprehensive process for
N
i=1
rapidly changing high-technology products with substantial
the equation to the variance of prices, ∑ ( pit − pît ) / N . The bar on the
2
i=1
changes in quality within relatively short periods.
term R2 denotes that an appropriate adjustment for degrees of freedom is 6.155 “Patching” is the term used in this Manual for intro-
made to this expression, which is necessary when comparing equations ducing noncomparable replacements (that is, replacements
with different numbers of explanatory variables. of a different quality), using hedonic regression estimates.
142
Table 6.10A Hedonic Regression Imputation of New t + 3 had a capacity of 12 kilograms. In this case, model
Variety’s Price n’s price in period t + 2 would be the predicted price,
Variety/ t t + 1 t + 2 t + 3 t + 4 p˘ = 117.3×12 − 436.23 = 971.37. The ratio of the actual
Period price of model m in period t + 2, for example, £750, to the
l plt plt +1 plt +2 plt +3 plt +4
predicted price in period t + 2, is the quality adjustment
shown for the overlap method in equation 6.6, though for
m pmt pmt+1 pmt+2 t +2
period t + 2 in this example, pm * , that is, 750 = 0.7721.
n p̆nt+2 pnt+3 pnt+4
pnt+ 2 971.37
The models are not comparable. The new model in period
t + 2 is more expensive even when its superior quality, its
capacity, has been considered.
Consider varieties l, m, and n in Table 6.10A where vari- 6.159 Given the availability of an estimate of the worth
ety l is available in all periods, the “old” variety m is only of an extra unit of capacity, an alternative approach would
available in periods t, t + 1, and t + 2, and the replacement be to simply add 2 × 117.3 to the price of m in period t + 2,
variety n is only available in period t + 3 and subsequently. rather than use predicted prices. Such use of individual coef-
The varieties are defined by their z quality characteristics; ficients is not recommended. In practice, a hedonic regres-
for example, for variety m, in period t these are zmt and the sion will include several explanatory price-determining
price of variety m is pmt . The example assumes that there is variables that may be linearly related and thus not strictly
no problem with comparing the prices of matched variety l independent. For example, larger (higher-capacity) washing
with characteristics Z1, as they have the same quality char- machines may also have higher spin speeds or be more likely
acteristics, but there is a problem when comparing varieties to have a steam feature. The estimated coefficient of each
m and n. Variety m’s replacement n is noncomparable, so such multicollinear variable would be imprecise, though the
pmt+2 cannot be directly compared with pmt+3 . An imputed predicted price of a regression equation that includes them
price is required in order to have prices for both the old and would be unbiased.
new varieties in the same period. This could be achieved by 6.160 With the option cost example, the quality adjust-
imputing the price of the new variety n in period t + 2 to ment might be for a single characteristic and an explicit
form an overlap in this period with the actual price of the old valuation of the price of further units of this characteristic
variety m, in that period, as illustrated in Table 6.5C. This is (for example, a gigabyte of storage for a computer, available
a backward imputation. In this case, as illustrated in Table from another source). Hedonic regressions are used where
6.10A, the overlap period is period t + 2. However, variety the market does not reveal the implicit shadow prices of
n does not have a recorded price in period t + 2, and it may individual characteristics; these shadow prices have to be
not have been sold then. The backward hedonic imputation estimated from price data for many varieties with differing
approach would predict the price of variety n in period t + 2 bundled sets of characteristics.
using a hedonic regression estimated in period t + 2 and 6.161 The hedonic method makes use of short-term
the characteristics of the new variety n, taken from period month-on-month comparisons. Predicting the price of vari-
t + 3 (that is, the predicted price of variety n in period ety n in period t + 2, if it was sold then, is only for this
t + 2, pˆ nt+2 —the hat over the price, pˆ , denotes a predicted one-off period as the new variety replaces the old, with a
value from the regression). The predicted prices are for the quality adjustment. Variety n’s characteristics are held con-
characteristics of the replacement variety n. This is an esti- stant for month-on-month comparisons from t + 2 onward,
mate of how much the price for the characteristics of the and variety m’s characteristics are held constant for month-
new replacement variety would have been if it had been sold on-month comparisons from period t up to, and including,
in period t + 2. period t + 2.
6.156 Where data are not available to support the 6.162 Alternatively, a forward imputation might have
monthly estimation of regression coefficients, as described been used, a procedure similar to that adopted in Table 6.5C.
in the previous paragraph, an alternative approach would be The price of variety m might be predicted from a hedonic
the hedonic quality-adjustment method.10 regression run on period t + 3 data, pˆmt+3 . As with the pre-
6.157 For short-term comparisons, an overlap method ceding methodology, a predicted price is only required for
is used with a price relative for t + 2 compared with t + 1 the overlap period, after which the replacement variety
given by pmt+ 2 / pmt+1 , and for t + 3 compared with t + 2 given forms the continuing index. It is not obvious which of the
by pnt+3 / p˘ nt+ 2 , and subsequently, without the need for an two approaches, predicting prices for m or n, is preferred.
imputation, by pn / pn .
t +4 t +3 Resources permitting, a geometric mean of the two would
6.158 The simple example outlined before using data be defensible, as would a clear rule from the outset as to
on washing machines is used here to illustrate the meth- the method applied based on some retrospective research on
odology. Assume that the linear regression equation 6.13 the outcome of using either method for particular product
was estimated using period t + 2 data, the old model m had groups.
a capacity of 10 kilogram, and the new model n in period 6.163 Table 6.5C shows that the backward and forward
imputation methods yield the same result when the imputa-
tion is based, for both methods, on the price movements of
10 varieties available in all periods. However, in this case, the
Jack Triplett. 2006. Handbook on Hedonic Indexes and Quality Adjust-
ment in Price Indexes. Paris: OECD. https://www.oecd-ilibrary.org/ backward prediction is based on a hedonic regression run
science-and-technology/handbook-on-hedonic-indexes-and-quality- in period t + 2 and the forward imputation on a hedonic
adjustments-in-price-indexes_9789264028159-en. regression run in period t + 3. The practical advantage of
143
running a hedonic regression in a prior period argues for a for a double imputation becomes more important as time
backward imputation, as in Table 6.10A, as the most fea- passes by, yet a double imputation requires monthly estima-
sible procedure. tion of hedonic regressions that hinder the advantage of this
6.164 A refinement to these approaches is to use pre- approach. If hedonic regressions are to be used on this long-
dicted values, for varieties m and n, in the overlap period, term basis, it is important that the regressions are reestimated
pmt+3 / pmt+ 2 . For this purpose, consider a misspecification regularly at a rate that will depend on the rate of the tech-
problem in the hedonic equation, for example, there may be nological innovations, and changes in consumer preference
an interaction effect between a brand dummy and a charac- specific to that product. For example, it may be that con-
teristic. Having a characteristic for a particular brand may be sumer’s valuations of characteristics of washing machines,
priced higher than all other brands, say a 5 percent premium. including spin speed, front-loaders, capacity, or number and
The use of pˆm / pˆ m would be misleading since the actual
t +3 t +2
types of wash programs, are fairly constant over time, even
price in the denominator would incorporate the premium, if the technology is changing rapidly. Frequent, say monthly,
while the one predicted from the hedonic regression would updating of estimated hedonic regression equations is not
not. It is stressed that, in adopting this approach, a recorded required. Prior empirical studies on the stability over time
actual price is being replaced by an imputation. This is not in hedonic characteristics would be valuable in this respect.
desirable, but neither is the omitted variable (interaction As a general principle, short-term hedonic imputations are
term) bias. The dual imputation approach is preferred when- preferred to long-term ones.
ever there are concerns about the suitability of the regres-
sion equation’s specification to fully model prices, as would Limitations of the Hedonic Approach
generally be the case. 6.168 The limitations and challenges of implementing
6.165 A further approach would be to not use a replace- the hedonic approach should be considered by the NSO:
ment variety. Variety m’s characteristics would be held con-
(1) First, the hedonic approach requires statistical exper-
stant in the comparison from period t + 2 onward. However,
tise for the estimation and maintenance of the hedonic
this would require a hedonic regression being run for each
regression equations. The availability of user-friendly
subsequent period, pˆ m , pˆ m . It would also lead to a con-
t+3 t+4
statistical/econometric software with regression tools
tinuing degradation of the sample as an obsolete old variety
makes this less problematic. Yet staff must possess
m would have its characteristics repeatedly priced into the
sufficient expertise and understanding of statistical
future, rather than being replaced by a new variety. For this
regression methodology applied to hedonic regression
reason, this method is not recommended.
equations, and the interpretation of the results and diag-
6.166 In the previous examples, short-term price com-
nostic statistics of regression models.
parisons are used and are preferable to long-term ones.
A long-term equivalent of Table 6.10A is shown in Table • Statistical and econometric software carry a range of
6.10C. A predicted price for any replacement variety n in its diagnostic tests to help judge if the final formulation
month of introduction is estimated for the reference period t of the model is satisfactory. These include R 2 as a
using a hedonic regression based on that period’s data. The measure of the overall explanatory power of the equa-
regression is estimated using period t prices and character- tion, and F-test and t-test statistics that test whether
istics, but the predicted prices are for the characteristics of the differences between the estimated coefficients
the replacement variety n in t + 3 and subsequently. It is of the explanatory (price-determining) variables are
an estimate of what the characteristics of the new replace- jointly and individually different from zero at speci-
ment variety would have been priced at had it been sold in fied levels of statistical significance. These statistics
period t. make use of the errors from the estimated regression
6.167 The long-term method has the significant advan- equation.
tage of only requiring a hedonic regression to be estimated • The regression equation can be used to predict prices
in the single reference period. For periods t + 3 and t + 4, for each variety by inserting the values of the char-
the price relatives are pn / pˆ n and pnt+4 / pˆ nt , respectively.
t +3 t
acteristics of the varieties against the estimated coef-
However, as time passes, such comparisons become less ficients of the explanatory variables. The differences
meaningful. For example, comparing the actual price this between the actual prices and these predicted results
month of a model of a laptop with one predicted 18 months are the residuals. Statistical/econometric software
ago using the hedonic approach, would estimate market calculate predicted values and residuals as a routine.
valuations of each characteristic which is then applied to A hedonic regression equation estimated using OLS
the characteristic set of a laptop sold now. Indeed, the need requires assumptions on the nature of the distribution
Table 6.10B Hedonic Regression Imputation of Old Table 6.10C Hedonic Regression Imputation of New
Variety’s Price Variety’s Price
Variety/ t t + 1 t + 2 t + 3 t + 4 Variety/ t t + 1 t + 2 t + 3 t + 4
Period Period
l plt plt +1 plt +2 plt +3 plt +4 l plt plt +1 plt +2 plt +3 plt +4
t +1
m pmt p t +1
p t +2
p̆ t +3 m p t
m
p m p t +2
m
m m m
n p t +3
n p t +4 n p̆nt pnt+3 pnt+4
m
144
of the residual errors. These include: (1) the error lar between the outlets. A similar principle applies for
term has a constant variance—if this assumption is the brands of varieties used in the sample for the he-
violated, the errors are heteroscedastic; consequently, donic regression. It should be borne in mind that high
standard tests of statistical significance can be biased R 2 statistics do not alone ensure reliable results. Such
and unreliable; (2) that explanatory variable(s) are high values arise from regressions in periods prior to
not correlated with the error term, they are not endog- their application and indicate the proportion of varia-
enous—this is particularly important when explanation in prices across many varieties and brands. They
tory price-determining characteristics are omitted are not a measure of the prediction error for a particular
from the hedonic regression: if an omitted variable variety, sold in a specific outlet, of a given brand in
is correlated with an included one, the estimated co- a subsequent period, though they can be an important
efficient on the included one is biased; and (3) the part of this.
price-determining explanatory independent variables (4) Fourth, the functional form and choice of variables
are not truly independent, but correlated with each to include in the model should be considered. Simple
other—multicollinearity; the coefficient estimates functional forms generally work well, though there is a
and their tests become sensitive to change in the class of more complex flexible-functional forms. These
model and data. While the estimated coefficients are include linear, semilogarithmic (logarithm of the left-
imprecise, the predicted prices in a hedonic regres- hand side), and double-logarithmic (logarithms of both
sion would be unbiased. sides) forms. Semilogarithmic models are often em-
• A full account of all OLS assumptions, consequences, ployed since many of the price-determining explanatory
means of detection of violation, and treatment, that variables are binary, 1 or 0, depending on whether or
may involve use of an alternative to OLS estimators, not a model has a particular feature (dummy variables).
can be found in any introductory econometrics/sta- The specification of a model should include all price-
tistical text. Modern software provides the appropri- determining characteristics. Typically, a study would
ate tests for, and means of overcoming, violations of start with a large number of explanatory variables and
these assumptions and thus, validation of the hedonic a general econometric model of the relationship, while
model used. It is recommended that the NSO devel- the final model would be more specific, having dropped
ops and publishes detailed metadata on the hedonic a number of variables. The dropping of variables would
regression model used and its supporting diagnostic depend on the result of experimenting with different
statistics to demonstrate the validity of the model and formulations, and analyzing their effects on diagnostic
satisfy the need for transparency. test statistics, including the overall fit of the model and
(2) Second, the estimated coefficients require regular up- the accordance of signs and magnitudes of coefficients
dating. Consider that the predicted price is for the new with prior expectations.
model in a reference period, as presented in Table 6.10C. (5) Fifth, the resources requirements for hedonic regres-
Although it might seem that there is no need to update sion should be considered. Hedonic regressions require
the estimated coefficients each period, the valuation of data on prices and price-determining characteristics
characteristics in the price reference period may be quite for varieties (models) sold. Extensive data sets may be
out of line with their valuation in the new period. For readily available on the internet or from scanner data,
example, quite dramatic falls in the price of storage and containing all pertinent price-determining characteris-
processing speed of computers, among other attributes, tics, either from the websites of individual retailers or
make the valuation of additional GBs of a new model, specialist websites comparing prices and features of
introduced a few years after the hedonic regression was laptops, household appliances, and many other such
estimated, a less meaningful exercise. Continuing to use goods and services. For example, the data used for the
the coefficients from some far-off period to adjust prices previous example of (patched) hedonic explicit quality
in the current period is similar to using out-of-date ref- adjustments for washing machines was taken from a
erence period weights. The comparison may be well de- website and was copied and pasted relatively quickly.
fined but have little meaning. There is a need to update Web scraping software can reduce even this workload
the hedonic regression estimates if they are considered substantially.
out of date, because of changing tastes or technology, (6) Finally, while data and software may not be problem-
and to splice the new estimated comparisons onto the atic, NSO staff resources will be required in devising the
old. The regular updating of hedonic estimates when specification, estimation, and validation of the estimated
using imputations or adjustments is thus recommended, hedonic model for each product. Such hedonic models
especially when there is evidence of instability in the should be estimated regularly prior to their use in the
parameter estimates of the hedonic regression over CPI and the results made available as part of the detailed
time. metadata for the purpose of transparency and feedback.
(3) Third, the sample of prices and characteristics used for In this regard, the resource requirements can be sub-
the hedonic adjustments should be suitable for the pur- stantial compared with an implicit overlap method. At
pose. If they are taken from a particular outlet or outlet least at first, hedonic methods should be applied only
type, trade group, or webpage, and then used to adjust to products with a relatively high weight and profile for
noncomparable prices for varieties sold in quite differ- which the implicit assumptions of alternative methods
ent outlets, there must at least be an intuition that the are found to be invalid and badly distort the results, es-
marginal price differences for characteristics are simi- pecially if they provide a reputational risk to the NSO.
145
6.169 Hedonic methods may also improve quality- use of resources for one product may lead to less resource-
adjustment procedures in the CPI by indicating which prod- intensive methods for others. The methods adopted for indi-
uct attributes do not appear to have material impacts on the vidual product groups may vary between countries as access
prices. That is, if a replacement variety differs from the old to data, relationships with the outlet managers, resources,
variety only in characteristics that have been rejected as expertise and features of the production, and market for the
price-determining variables in a hedonic study, this would product vary. Guidelines on choice of method arise directly
support a decision to treat the varieties as comparable. Care from the features of the methods outlined in this chapter. A
has to be exercised in such analysis because a feature of good understanding of the methods, and their implicit and
multicollinearity in regression estimates is an imprecision explicit assumptions, is essential to the choice of an appro-
of the estimated parameter estimates. This may give rise to priate method.
statistical tests that do not reject null hypotheses that are 6.171 Figure 6.3 provides a guide to the decision-mak-
false. However, econometric/statistical software provides ing process. Assume that the MMM is being used. If the
the tools to explore the nature and extent of multicollinear- variety is matched for repricing in a subsequent period, there
ity; these include variance inflation factors. The results from is no change in the specifications and no quality adjustment
variance inflation factors provide valuable information on is required. This is the simplest procedure. However, there
the nature and extent to which different explanatory vari- is a caveat: if the variety belongs to a product group where
ables (characteristics) are interrelated and this in turn can model replacement is rapid, and replacements are noncom-
help in the selection of replacement varieties. The results parable, the matched sample may become unrepresentative
from hedonic regressions thus have a role to play in identi- of the universe of transactions. Continued long-term match-
fying price-determining characteristics and may be useful in ing would deplete the sample. This is a matter for the fre-
the design of quality checklists in price collection. quent rebasing and maintenance of the sample (see Chapter 7).
6.172 Consider a variety found to be temporarily miss-
ing. If it was a seasonal product, its treatment would fol-
Choice between Quality-Adjustment low the principles and practices outlined in Chapter 11. If it
Methods was temporarily missing but not a seasonal product, a price
6.170 The choice of the method to be used for quality imputation is required, and if subsequently determined to be
adjustments is not straightforward. The CPI compiler must permanently missing (for example, either from information
consider the technology and market for each product and from outlet staff or use of a three-month rule) a replacement
devise appropriate methods, considering that the methods needs to be found. Overall or targeted price imputations for
selected for one product area might not be independent of temporarily missing prices may be used; the carryforward
those selected for other areas. Expertise built up using one method is not recommended except for controlled or regu-
method may encourage its use elsewhere, and intensive lated prices.
Figure 6.3 Guide to Treatment of Missing Prices
146
6.173 For permanently missing varieties, the selec- 6.177 If the old and replacement varieties are available
tion of a comparable variety is preferred, as is the use simultaneously, and if the quality difference cannot be quan-
of its price as a comparable replacement price which is tified, an implicit approach can be used whereby the price
then directly compared with the preceding variety price. difference between the old and replacement varieties in a
This direct price comparison would require that none of period in which they both exist is assumed to be attribut-
the price difference between the comparable replacement able to quality. This overlap method, in replacing the old
and the previous variety is attributable to quality, and con- variety by a new one, takes the ratio of prices in a period to
fidence that all price-determining factors are included in be a measure of their quality difference. It is implicitly used
the specification. In practice, varieties may be considered when new samples of varieties are taken. The assumption of
comparable if there are limited price-determining differ- relative prices equating quality differences at the time of the
ences, as might be the case with styling, color, even some splice is unlikely to hold if the old and replacement varieties
more substantial technical changes, including performance are at different stages in their life cycles and different pricing
and reliability, that may not be immediately apparent to the strategies are used at these stages. For example, there may
consumer. A decision on the comparability of a replace- be deep discounting of the old variety to clear inventories,
ment must be made by CPI staff with appropriate informa- and price skimming of market segments that will purchase
tion on product differences supplied by the price collector. new varieties at relatively high prices. As with comparable
A comparable replacement variety should also be repre- replacements, early substitutions are advised so that the over-
sentative and account for a reasonable proportion of sales. lap is at a time when varieties are at similar stages in their
Caution is required when replacing near obsolete varieties life cycles. It may well be the case, however, that overlap
with unusual pricing at the end of their life cycles with sim- prices are unavailable. In such cases, a range of imputation
ilar ones that account for relatively low sales, or with ones approaches is available to estimate an overlap price.
that have quite substantial sales but are at different points in 6.178 The quality differences between the replacement
their cycle. Strategies for ameliorating such effects are dis- and missing variety may be explicitly quantified. Explicit
cussed in paragraphs 6.182–6.222 and in Chapter 7, includ- estimates of quality differences are generally considered
ing early substitutions before pricing strategies become to be more reliable, although they are also more resource-
dissimilar. With comparable replacements, the price of the intensive, at least initially. Once an appropriate methodol-
old variety is directly compared with the price of the com- ogy has been developed, it can often be easily replicated.
parable replacement in the next period. General guidelines are more difficult to provide as the choice
6.174 Figure 6.3 considers the case where only non- depends on the factors already discussed in this chapter,
comparable replacements are available. If explicit estimates which are likely to make the estimates more reliable in each
of the price dimension of the quality differences are unavail- situation. Central to all of this is the quality of the data upon
able, and no replacement varieties are deemed comparable, which the estimates are based. Estimates based on objec-
implicit estimates might be used. One such method is the tive data are preferred. A relatively straightforward quality
continued use of imputations as they are applied to tempo- adjustment is when the quantity differs. The standardization
rarily missing varieties. Such use is not recommended as a of quantity units sold across outlets, for example, to price
default procedure. per kilogram, is relatively straightforward, though a change
6.175 The use of imputations has advantages resource- in the quantity of a variety included in the price—a quantity
wise, as it is relatively easy to employ and requires no judg- adjustment—may be more complicated than expected.
ment (unless it is a targeted mean imputation) and is therefore 6.179 The replacement variety may differ from the
objective. Targeted mean imputation is preferred to overall old one for having a different characteristic. Often it is the
mean imputation as long as the sample size upon which the price collector who is best placed to provide an estimate of
target is based is adequate. The bias from using imputations the price difference in quality of a noncomparable replace-
for permanently missing variety prices is directly related ment. For example, if a specified brand of a bottle of tomato
to the proportion of missing varieties and the difference ketchup used for pricing is missing in the current period, and
between quality-adjusted prices of available matched variet- a noncomparable replacement of the same brand is avail-
ies and the quality-adjusted prices of unavailable ones. The able, though the bottle has been restyled to now stand on its
nature and extent of the bias depends on whether short-term head, and the label has been reversed. The price collector
or long-term imputations are being used (the former being might note that other brands have both sizes sold with a 25
preferred) and on market conditions. Imputation, in practical percent price margin for the new one. The price collector
terms, produces the same result as deletion of the variety for in selecting a noncomparable replacement might also pro-
an elementary aggregate. The inclusion of imputed prices vide the basis for the head office staff to make an explicit
may give the illusion of larger sample sizes. Imputation quality adjustment. The head office staff might also make
should by no means be the overall catch-all strategy, and use of the internet to identify the percentage markup for a
NSOs are strongly advised against its use as a default device quality characteristic, for example, for additional memory
that may lead to serious sample degradation. for a computer or Bluetooth technology in an automobile.
6.176 Imputations can be used to extend the period of The option cost approach is applicable when a new feature
search for a replacement, though the absence of the old vari- is first sold as an option and then becomes a standard com-
ety and the unavailability of a replacement should indicate ponent included in the basic price. This requires that the old
to the CPI compiler that the weight for that variety might be and new varieties differ by easily identifiable characteristics
better attributed to a quite different variety. Such changes that are or have been separately priced as options. The use of
naturally take place on updating an index, as described in production cost estimates critically relies on the availability
Chapter 7. of suitable estimates for the price-cost margin.
147
6.180 The use of hedonic regressions for patching price down the implicit or explicit methods can be used. But
changes because of quality differences is most appropriate what should be done in the case of industries where the
where data on price and characteristics are available for a matching breaks down on a regular basis because of the high
range of models and where the characteristics are found to turnover in new models of different qualities than the old
predict and explain price variability well with regard to a ones? The matching of prices of identical models over
priori reasoning and econometric terms. Their use is appro- time, by its nature, is likely to lead to a seriously depleted
priate where the cost of an option or change in characteris- sample. There is both a dynamic universe of all varieties
tics cannot be separately identified and should be collected consumed and a static universe of the varieties selected
from the prices of varieties sold with different specifications for repricing. For example, if the sample is initiated in
in the market. The estimated regression coefficients are the December, by the subsequent May, for a long-term price
estimate of the contribution to price of a unit change in a comparison, the static universe will be matching prices
characteristic, having controlled for the effects of variations of those varieties available in the static universe in both
in the quantities of other characteristics. The estimates are December and May but will omit the unmatched new vari-
particularly suited to valuing changes in the quality of a eties introduced in January, February, March, April, and
variety when only a given set of characteristics changes and May, and the unmatched old ones available in December
the valuation is required for changes in these characteristics but unavailable in May. For December to May cumulative
only. The results from hedonic regressions may be used to month-on-month short-term comparisons, similar consider-
target the salient characteristics for variety selection. The ations apply. Although there will be improved imputations
synergy between the selection of prices according to charac- for temporarily missing variety prices and an improved
teristics defined as price determining by the hedonic regres- timelier introduction of replacements, the replacements
sion, and their subsequent use for quality adjustment, should only draw from the dynamic universe of new models on
reap rewards. The method should be applied where there are a one-on-one basis. This example refers to a December to
high ratios of noncomparable replacements, though not a January matched price comparison. For many countries,
frequent churn, and where the differences between the old matching may effectively continue for many years until
and new varieties can be well defined by its characteristics. the CPI is updated leaving an extremely degraded sample.
6.181 As previously discussed in this chapter, the use of Two empirical questions indicate whether there will be any
the link-to-show-no-price-change method for permanently significant bias. First, is sample depletion substantial? Sub-
missing and the carryforward method for temporarily miss- stantial depletion of the sample is a necessary condition for
ing variety prices are not generally advised for making qual- such bias. Second, are the unmatched new and old varieties
ity adjustment and imputations. likely to have quality-adjusted prices that substantially dif-
6.182 While Figure 6.3 is appropriate for the treatment of fer from those of the matched varieties in the current and
temporarily and permanently missing prices in the routine com- the reference periods?
pilation of a CPI, there is a context in which a quite different 6.184 The matching of prices of identical models over
strategy is required. The context is where there is a rapid turn- time may lead to the monitoring of a sample of models that
over or “churn” in the models or varieties sold. For example, is increasingly unrepresentative of the population of transac-
television sets are sold by several manufacturers each having a tions. Some of the old models that existed when the sample
range of models with different features. Over time many new was drawn are not available in the current period, and new
phases of technological development have occurred includ- models that enter the sample are not available in the refer-
ing the cathode ray tube, color televisions, wireless remotes, ence period. It may be that the models that are disappearing
plasma, LCD televisions, digital, high definition, larger screens, have relatively low prices, while the entrants have relatively
smart functions, 3D, LEDs, ultra HD resolution, OLED, and high ones. By ignoring these prices, a bias, known as sam-
roll-up OLED. New features and restyling extend the life cycle pling bias, is being introduced. Using old low-priced vari-
of each model in each phase. As with automobiles, comput- eties and ignoring new high-priced ones has the effect of
ers, computer-related hardware and software, telecommunica- biasing the index downward. For some products, the new
tions equipment, or household appliances, the product market variety may be introduced at a relatively low price though
is characterized by different manufacturers producing several the old one may continue at a relatively high price, serv-
varieties (models) each of different quality, such as screen size ing a minority segment of the market. In this case, the bias
for a computer or television set, aimed at different segments would take the opposite direction. The nature of the bias will
of the market. These will, over time, usually have a rapid turn- depend on the pricing strategies of firms for new and old
over in their quality characteristics. The previously outlined varieties. Some strategies for the introduction of new mod-
methods, if applied to these markets, may lead to a biased CPI. els, and implications for CPI measurement, are considered
Figure 6.3 notes that matching, class mean imputations, and in Annex 6.3.
hedonic price indices may be used, though there may be severe 6.185 This sampling bias exists for most products.
bias in the use of the former. The next section considers CPI However, the concern here is with product markets
measurement for these product markets. where the NSOs are finding the frequency of new variety
introductions and old variety obsolescence sufficiently
High Technology and Other Sectors high that they may have little confidence in their results.
Three procedures will be considered: an extensive use
with a Rapid Turnover of Models of the matched-model (overlap) technique, class mean
6.183 The measurement of price changes of variet- imputation, and the use of hedonic price indices (as
ies unaffected by quality changes is primarily achieved opposed to the partial, hedonic patching discussed in
by matching models; however, when the matching breaks paragraphs 6.136–6.150).
148
Matching and the Overlap Method for Table 6.11 Illustration of Rapid Model Turnover
Markets with Rapid Turnover of Models Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Jan.
6.186 The first approach to address markets with rapid 1 25 25 25 25 25 25 25 25 25 25 20
turnover of models is simply a more extensive use of the 1R 29.2 30 30
overlap approach outlined previously for permanently miss- 2 35 35 35 35 35 32
ing prices. In this case, it is adopted for permanently miss- 2R 37 37 37 37 37 37 40 40
3 30 30 30 30 30 30 30 30 30 30 27
ing varieties that occur frequently, as is usual for changes 3R 32.2 33 33
in models of electronic goods and automobiles. Matching 4 29 29 29 29 29 28
prices of a few representative varieties becomes a less fea- 4R 30 30 30 30 30 30 30 30
sible approach in this context. A backward imputation is 5 29 29 29 29 29 29 29 29 29 29 29 29 29
illustrated in Table 6.11 though, as outlined in the illustra- Values in bold are imputed.
tion for Table 6.5C, a forward imputation can be equally
justified and both methods provide the same results when
6.190 The price relative for model 1 from January in
the imputations are based on the price changes of overlap-
the preceding year to January in the current year shows a
ping matched samples.
(1 − 0.8219) × 100 = 17.81 percent price decrease. It is clear
6.187 Considering model 1 in Table 6.11, in Novem-
from Table 6.11 that the price for model 1 has been con-
ber there is no overlap price for the new model 1R, so its
stant up to October, and there was a price fall in November,
price is imputed “backward” by using the ratio of geometric
but this was to clear the market for the replacement. There
means of the December to November prices but only includ-
should be a November to December price increase for the
ing those for which matched models exist, that is, models 2R,
old model 1 to the new replacement 1R that reflects that part
4R, and 5. These are all constant-quality price comparisons;
of the price difference between model 1 and model 1R was
of like with like. As shown in expression 6.13, its imputed
not due to quality differences. But the imputation is based
price is 0.974 × 30 = 29.2:
on the constant price movements of models 4R and 5 and
1 1
a coincidental price increase in model 2R; it assumes that
(37 ×30× 29)3  37 30 29 3 short-term price movements of matched pairs will proxy the
=  × ×  = 0.974 price change of model 1. However, in this context, the con-
1  40 30 29 
(40×30× 29)3 stant price changes of matched models are an inappropriate
proxy badly biasing the measured price change downward.
6.188 The imputed price for the replacement model 2R At fault are first, the use of the unrepresentative price for
in June is based on the price changes of matched models 1, model 1 at the end of its life cycle in November, and second,
1 the inappropriate imputation for the replacement variety.
(25×30× 29)3 6.191 The January to January price decreases for mod-
3, and 5 for June and July, that is, = 1.00
1 els 2, 3, and 4, using replacements, and 5 are, respectively,
(25×30× 29) 3
1.2, 6.3, and 3.4 percent and no change for model 5. With a
and its imputed price: 1.00 × 30 = 30. The imputed prices for substantial churn of models, possibly more frequently than
annual, the bias from using overlaps can be substantial.
6.192 As an advantage, the method is simply an exten-
3R in November and 4R in June are 32.15 (rounded to 32.2,
sion of the linking-in of new products and can be readily
for simplicity of exposition) and 30.0, respectively. applied by an NSO, especially one with limited resources.
6.189 The overall price relatives for each model, and However, this approach biases the CPI because it bases the
their linked-in replacements, can now be computed as the imputations on price changes of matched varieties not sub-
product of short-term month-on-month price changes, that ject to the price changes that occur on the replacement of a
is, model 1 and its replacement 1R, for January to January, model.
using the overlap month of November: 6.193 The overlap method may be subject to bias if
applied to where there is substantial churn in the product
p1Feb p1Mar p1Apr p1May p1Jun p1Jull market and an active policy by the supplier of introducing
x x x x upgraded replacement models. The nature and extent of the
p1Jan p1Feb p1Mar p1Apr p1May p1Jun
bias depends on the pricing strategy. Table 6.11 illustrates
p1Aug p1Sep p1Oct p1Nov p1Dec p1Jan
R a policy of lower pricing at the end of the life cycle and a
x x x x x x Dec higher price at the start. Importantly, the example has no
p1Jul p1Aug p1Sep p1Oct p1Nov
Rimp p1R
price change for other matched models, from which the
p1Nov p1Dec p1Jan imputation was drawn, and thus a biased imputation at this
= x R
x R
(6.16a) critical overlap period. The bias is substantial and down-
p1Jan p1Nov
Rimp p1Dec
R ward. Alternative pricing strategies are given in Annex 6.3
along with their implication for bias from using the overlap
25 25 25 25 25 25 25 matching. It has been demonstrated in empirical studies that
× × × × × ×
25 25 25 25 25 25 25 the method can introduce substantial bias under quite rea-

25 25 20 30 30 sonable conditions. The nature and extent of the bias depend
× × × × × on business pricing strategies that may change over time and
25 25 25 29.2 30
are unpredictable, which is a concern for CPI compilation
20 30 30 in this important product area. The overlap method is thus
= × × = 0.8219 (6.16b)
25 29.2 30 149
not recommended for product markets with a high rate of adjustments may lead to bias and, second, the sampling will
model churn. be from a static matched/replacement universe likely to be
biased. With new models being continually introduced and
Use of a Class Mean Imputation old ones disappearing, the coverage of a matched sample
may deteriorate, and bias may be introduced as the price
6.194 The previous example showed that an imputation changes of new/old models differ from those of the matched
based on price movements of other matched models not at ones. What is required is a sample to be drawn in each month
the end of their life cycle could introduce a bias. An alter- and price indices constructed; but instead of controlling for
native, though more resource-intensive method, is to base quality differences by matching, they will be controlled for
the imputations not on price changes of matched varieties in the hedonic regression. Note that all the indices described
but to use, where possible, explicit quality adjustments for in the following text use a fresh sample of the data avail-
linked-in noncomparable replacement. For example, inter- able in each period. If there is a new variety in a period, it
net webpages of prices of similar products may show the is included in the data set and its quality differences con-
difference in characteristics and prices of the old models and trolled for by the regression. Similarly, if old varieties drop
the replacement models. The replacement may simply have out, they are still included in the data for the indices in the
a higher value of some performance characteristic or fea- periods in which they exist. Paragraphs 6.110–6.115 stress
ture for which the price is available as an option. If enough the need for caution in the use of hedonic regressions for
explicit quality adjustments can be made, imputations might quality adjustments.
be better made on the basis of only those models that have 6.198 Consider a price comparison between two adja-
had explicit quality adjustments to their price. cent time periods, periods t and t + 1. The models sampled
6.195 The nature of the high frequency of replace- do not have to be matched, they may simply be all recorded
ments makes the procedure resource-intensive and, in some models sold in the two periods, and they comprise a differ-
instances, not viable because of the absence of explicit infor- ent mix of qualities. The hedonic formulation regresses the
mation on prices of features or options. However, if enough price of model i, pi, on the k = 2, … , K characteristics of the
models have an explicit adjustment, an average of their price varieties zki. A single regression is estimated on the data in
change could be used to impute the price change of other the two time periods compared, the equation also including a
models being replaced. This is the basis of using class mean dummy variable Dt+1 being 1 in period t + 1, zero otherwise.
imputations. The method requires care that the linking-in of
replacement models does not take place at the end of the The Time Dummy Variable Approach
model’s life cycle when pricing might be abnormally low 6.199 A single hedonic regression equation is estimated
for a variety that relatively few consumers are purchasing. with observations across models over adjacent time periods,
This not only has a detrimental effect on the quality-adjust- including the reference period 0 and a subsequent period t.
ment methodology, but also on the representativity of the The logarithm of prices of individual models is regressed on
models upon which the prices change measurement is based. their characteristics and a dummy variable for time, taking
6.196 The class mean imputation method was outlined the values of Di1 = 1 if the model is sold in period 1 and 0
in paragraphs 6.96–6.100. It is similar in procedure to the otherwise. A log-linear specification is given by
overall mean imputation and is a form of targeted imputa-
tion. The “target” is measured price changes of replacements K
for permanently missing products. Only the price changes ln pit = ln β0t + ∑ zkt ,i ln βkt + ln εit (6.17)
of “comparable” replacements are used to impute the over- k =1
lap price, the replacements being limited to those that have 6.200 The δt are estimates of the proportionate change
exactly the same price-determining characteristics, or those in price arising from a change between the excluded refer-
varieties with replacements that have been declared compa- ence period t = 0 and successive periods t = 1, T having
rable after review and have already been quality-adjusted controlled for changes in the quality characteristics via the
through one of the “explicit” methods. For example, when .
term ΣkK=1 Z k0,,it x lnβ k
the arrival of a new model of a particular kind of automo-
bile forces price collectors to find replacements, some of the
( )
6.201 In principle, the index 100 × exp δ̆ˆ t requires an
adjustment for it to be a consistent (and almost unbiased)
replacements will be of comparable quality, while others can
approximation of the proportionate impact of the time
be made comparable with explicit quality adjustments, but
dummy variable.11 In practice, the adjustment usually has
the remaining ones will need imputed prices for an overlap
little effect.
month
6.202 The method implicitly restricts the coefficients on
the quality characteristics to be constant over time: for exam-
Hedonic Price Indices ple, for an adjacent period January and February regression,
6.197 It is important to distinguish between the use of for k = 1, … , K characteristics and where period 0 and t
hedonic regressions for patching and their use as hedonic are January and February, respectively, βk = βkJan = βkFeb .
price indices, which are measures of quality-adjusted price The (relative) valuation of a characteristic, for example, for
changes. Patching adjusts individual item prices for qual- a washing machine with an additional 100 revolutions per
ity differences when a noncomparable replacement is used minute spin speed, is the same in January as in February.
while hedonic price indices are measures of quality-adjusted
price changes. Hedonic price indices are suitable when the
pace and scale of replacements of varieties are substantial
because, first, an extensive use of these overlap quality 11
See Triplett (2006) for more details.
150
Feb
( )
The index, 100 × exp δ̆ˆ , is an estimate of the quality- solution is required. An index making symmetric use of
adjusted price change for February (January = 100). period 0 and period t characteristics values is intuitive:
The Characteristics/Repricing Approach zkτ

( )
K
( ) ∏ β˘ˆk exp ∑ zkτ ln β˘ˆkt
K
6.203 A hedonic regression is run to determine the
t
k =0
price-determining characteristics of models, for example in 0:t
I HGMC: = k =0
= (6.20)
exp ( ∑ z ln β˘ˆ )
zτ zkτ
∏ (β˘ˆ )
K K
a reference period 0. The average model in period 0 can then 0
k k
τ 0
k
be defined as a tied bundle of the averages of each price- k =0 k =0
determining characteristic. In the previous example for

washing machines, these would include: spin speed: 1,375 where zkτ = ( zk0 + zkt ) / 2 .
revolutions per minute; capacity (cotton load): 8.5 kilogram;
annual energy cost: £36.5; steam facility: 4 percent; Brand 6.207 Equations 6.18, 6.19, and 6.20 all use predicted
A: 15 percent; warranty period: 5.4 years; and run-time (cot- prices in both the denominator and numerator. This follows
ton): 18.8 minutes. These are the Z k averages for each of the recommendation in the following text to use dual impu-
the k price-determining characteristics. tations. However, this method also entails running hedonic
6.204 The average values of each characteristic are held regressions in each period. Yet a fortuitous result is that a
constant in each period but valued in turn using period 0 feature of the OLS estimator is the mean of actual prices
and period t hedonic regressions. One form of the (average) being equal to the mean of predicted prices:
characteristics approach is as a measure of the price change
of a set of average period 0 characteristics valued first, at 1 1 1 1
period t hedonic valuations, and second, at period 0 hedonic N0
∑ ln pˆ˘
0
0
i|zi0
=
N0
∑ ln p 0
0
i
and
Nt
∑ ln p˘ˆ t
i|zit
=
Nt
∑ ln p t
i
valuations. A ratio of the results is a constant (period 0 char-

t t
i∈ N i∈ N i∈ N i∈ N
acteristics) quality price index. The numerator, the period t Thus, while the numerator of equation 6.18 and denomina-
hedonic valuation, provides an answer to a counterfactual tors of equation 6.19 must be counterfactual—the valuing
question: what would be the estimated transaction price of a period 0 (t) average characteristics at period t (0) prices—
model with period 0 average characteristics, were it on the the denominator of equation 6.18 and numerator of 6.19 can
market in period t? use actual prices, since the means are the same for an OLS
6.205 A constant-quality hedonic geometric mean char- estimator. This leads to the important results that equation
acteristics price index from a log-linear hedonic regression 6.19 does not require a hedonic regression to be estimated in
equation is a ratio of geometric means with average charac- every current period t, only in the price reference period 0.
teristics held constant in the reference period 0, zk0 : This is an important result since it aids the practical work of
compilers who do not have to estimate a hedonic regression
zk0
( ) equation in each period, but maybe once every one or two
K
( ) exp ∑ zk0 ln βˆ˘kt
K
∏ βˆ˘k
t
years, depending on the amount of churn in the market and
0:t
I HGM = k =0
= k =0
(6.18) shifting technologies and preferences. The hedonic indices
exp ( ∑ z ln β˘ˆ )
C:z 0 zk0
∏ (βˆ˘ )
K K
0 0
0
k k k
from one regression can be chained to its preceding hedonic
k =0
k =0
indices, and so forth, using successive multiplication.
1 The Hedonic Imputation Approach

where zk0 =
N0
∑z 0
i,k . 6.208 In contrast to the characteristics approach, the impu-
i∈ N 0
tation approach works at the level of individual varieties/mod-
6.206 Equation 6.18 holds the (quality) characteristics els, rather than the average values of their characteristics. The
constant in period 0, though a similar index could be equally rationale for the imputation approach lies in the MMM. Con-
justified by valuing in each period a constant period t aver- sider a set of models transacted in period 0. The objective is
age quality set: to compare their period 0 prices with the prices of the same
matched models in period t. In this way, there is no contamina-
tion of the measure of price change by changes in the quality
zkt
( )
K
( ) exp ∑ zkt ln βˆ˘kt
K
∏ β˘ˆk mix of models transacted. However, for goods and services with
t
0:t
....I HGMC = k =0
= k =0
(6.19) a high model turnover, not all the period 0 models were sold
exp ( ∑ z ln β̆ˆ ) in period t—there is no corresponding period t price in many
:z t zkt
∏ (β˘ˆ )
K K
0 t 0
k k k ˆ
cases. The solution—in the numerator of equation 6.21—is to
k =0 k =0
t
predict the period t price of each i period 0 model, pi| zt0 .
6.209 A constant-quality hedonic geometric mean
Nt
imputation price index from a log-linear hedonic regression
N t i∑
where zkt = 1 zit, k . equation is a ratio of geometric means with characteristics
∈N t
held constant in the reference period 0, zk0 :
N 0 = N t are the number of matched observations (varieties)
 
in the sample. Neither a period 0 constant-characteristics
index nor a period t constant-characteristic quantity basket i∈ N
( ) N1
Π0 p˘ˆ it| z 0 0
exp  1 0 ∑ 1n p˘ˆ it| z 0 
 N i∈ N 0 i 
I HGMI:z (6.21)
0:t i
0 = =
can be considered to be superior, both acting as bounds for 1  
Π ( p˘ˆ ) exp  1 0 ∑ 1n p˘ˆ i0| z 0 
i
0
their theoretical counterparts. Some average or compromise i∈ N 0 i | zi0
N 0
 N i∈ N 0 i 
151
6.210 Alternatively, the value in the numerator of equa- prices only. Hedonic indices adjust for quality change not
tion 6.21 is the geometric mean of the period t price for the by any meticulous or time-consuming matching and, for
price-determining characteristics in that period, Z it, k . This is that matter, identification of replacements, but by applying a
compared, in the denominator, with the geometric mean of hedonic regression to value constant-quality characteristics.
the period 0 predicted price of the same period t price-deter- 6.215 Hedonic indices use data on matched and
mining characteristics, Z it, k . For each model, the quantities unmatched observations and, again importantly, can natu-
of characteristics are held constant in period t, Z it, k ; only the rally be applied to large monthly data sets, such as scanner
characteristic prices change: and web-scraped data, as opposed to a small sample of what
may have been in some long-past reference period, a repre-
  sentative variety.
( ) N1
Π pˆ˘ it| zt
i∈ N t
t
exp  1 t ∑ 1n pˆ˘ it| zt 
 N i∈ N t i 
6.216 An advantage of the imputation approach over
the dummy variable approach is that explicit weighting
I HGMI (6.22)
0:t i
= =
:zt 1  
Π ( pˆ˘ ) exp  1 t ∑ 1n p˘ˆ i0| zt  systems can be more readily, accurately, and intuitively
i
0
i∈ N t i | zit
N t  N i∈ Nt i  applied at this elementary level. For example, equa-
tion 6.22 may be defined for models i over a set of models
6.211 As with the characteristics approach, a compro- of television sets sold in period t. The formula gives equal
mise solution of whether period 0 or period t constant char- weight to each model sold. A major improvement would
acteristics should be used, is to apply an average of the two. be to apply to each model’s quality-adjusted price change
However, as with the characteristics approach, equation the weight of that price change, that is, the individual mod-
6.22 has the advantage of only requiring a single hedonic el’s share of transaction expenditure values, for example
regression to be estimated in the price reference period 0. from scanner data. Silver (2018) outlines the methodol-
If this is used, the regression should be reestimated every ogy for the imputation approach, in the context of house
year or so, the frequency being determined by the turnover price indices, to include quasi-superlative and superlative
of products. formulations. The weighted imputation approach also has
6.212 The three approaches have different, yet valid, a correspondence to a weighted characteristics approach,
intuitions. As long as the functional form of the aggregator and the more intuitive application of weights, if formulated
is aligned to the hedonic regression in the manner shown as in Table 6.12.
in Table 6.12, the imputation and characteristics approaches 6.217 Hedonic indices are particularly well suited for
yield the same result. This consolidation not only mark- large data sets, as web-scraped or scanner data (see Chap-
edly narrows down the choice between approaches, but also ter 10), for which there is no matching of varieties. It is at
validates the measure as one resulting from quite different initiation that a price collector selects a representative vari-
intuitions. ety and matches its characteristics in subsequent periods in
6.213 For a log-linear functional form of a hedonic order to track the price of this same variety. In doing so, the
regression, the requirements are that (1) for the characteris- sample of prices collected is highly restricted to what may
tics approach, zk0 and zkt are arithmetic means of character- be a single price. With hedonic indices, it is the varying
istic’s values, the right-hand side of the hedonic regression, values of the characteristics of the models that enable a con-
and (2) for the imputation approach, the ratio of average pre- stant-quality price change. There may be data sets in which
dicted prices is a ratio of geometric means, the left-hand side accurately matched sampled prices form part of the sampled
of the regression. data. In such a case there would be no need for predicted
6.214 The important feature of the hedonic indices is prices to be used for constant-quality price change. The
that they require no matching of individual models in the overall measure for this data set would contain: (1) actual
periods compared. Matching is required so that the price of price changes for the matched sample; (2) hedonic price
a model in period 0 can be compared with that in period changes for the period 0 models not sold in period t (as,
t, without a concern that the price change is affected by for the hedonic imputation approach, in equation 6.21); and
changes in quality. Such matching restricts the sample and, (3) hedonic price changes for the period t models not sold
importantly in this context of a high level of churn in models in period 0 (as, for the hedonic imputation approach, in
and where prices change when models change, can lead to equation 6.22). Each of these terms would be weighted by
bias. This was illustrated in Table 6.11. The price compari- their relative expenditure shares, if available. It is from the
son of matched models effectively removes from the sample measure of all three components that the difference between
price changes in the important period of a price compari- the MMM and hedonic indices becomes apparent.
son when models change. The imputation for November to
December for model 1 in Table 6.11 is based on matched The Difference between Hedonic Indices
and Matched Indices
Table 6.12 Equivalences of Hedonic Approaches 6.218 As already mentioned, an advantage of hedonic
Hedonic Regression: Characteristics Imputation Approach: indices over matched comparisons was the inclusion by the
Functional Form Approach: Form Form of Average of former of unmatched data. Consider a data set of prices
of Average of Predicted Prices
and characteristics over two successive time periods, peri-
Characteristics
ods 0 and t. Assume there are m matched models in both
Linear Arithmetic Mean Arithmetic Mean periods 0 and t, o old models in period 0, but disappearing
Log-Linear Arithmetic Mean Geometric Mean thereafter, and n new models appearing in period t, and
Log–log Geometric Mean Geometric Mean
subsequently, as shown in Table 6.13.
152
Table 6.13 Difference between Hedonic and Matched into the measure, but only insofar as there is one-on-one
Indices variety replacement. Furthermore, hedonic indices employ a
Period 0 Period t consistent basis for the explicit quality adjustment for non-
comparable replacements.
Matched Models (m) m m
Old Model (o) o
6.223 The deficiency of the MMM against a hedonic
New Model (n) n index has been shown previously with regard to the hedonic
imputation approach, though similar considerations apply
to a time dummy variable approach. Consider an adjacent
6.219 A constant-quality, period 0 to t, price index, from period time dummy variable hedonic index of the form of
a hedonic imputation approach, is made up of three terms: equation 6.17, with the index change captured by the coef-
ficient on the dummy variable for time. For example, a
• The change in the geometric mean price of the m matched sample of models of washing machines for periods t and
models, with no need for quality adjustment because they t + 1 would have in the regression the log of price on the
are matched. left and price-determining characteristics on the right-hand
• The change in the constant-quality geometric mean price side. On the right-hand side, a dummy variable would also
of the old models with actual prices in period 0 and coun- denote whether the observation is drawn from period t or
terfactual ones in period t. The counterfactual constant- t + 1. The hedonic regression includes matched, new, and
quality price in period t has to be estimated since there is old models and the quality adjustment is achieved through
only a price in period 0. A prediction is required of what K
each old model’s price in period 0 would have been had the term ∑z t
k ,i ln βkt in equation 6.17. A matched-model
it been sold in period t. A period t hedonic regression is k =1
measure of price change would again only measure the
estimated, and a predicted price estimated for each model price change for the more limited sample of matched mod-
by inserting its period 0 characteristic Z k0 values into the els but would not require a quality adjustment. The hedonic
right-hand side of the estimated regression equation. A dummy variable approach, with its inclusion of unmatched
geometric mean is compiled of these predicted values, old and new observations, will likely differ from a geomet-
1
 t No ric mean of matched prices changes, the extent of any dif-
∏i ∈o
 p̆ˆ 0  , and compared with the period 0 geometric
 i zi  ference depending, in this unweighted formulation, on the
1
mean, ∏( pt0 )No , as in equation 6.19. proportions of old and new varieties leaving and entering
i ∈o
the sample and on the price changes of old and new varieties
• The change in the constant-quality geometric mean price relative to those of matched ones. If the market for products
of the new model in period t. The counterfactual constant- is one in which old quality-adjusted prices are unusually
quality price in period 0 has to be estimated since there is low while new quality-adjusted prices are unusually high,
only a price in period t. A prediction is required of what then the matched index will understate price changes. Dif-
each new model’s price in period t would have been had ferent market behavior will lead to different forms of bias
it been sold in period 0. A period 0 hedonic regression is (see Annex 6.3).
estimated, and a predicted price estimated for each model
by inserting its period t characteristics Z kt values into the
right-hand side of the estimated regression. A geometric
1
Key Recommendations
mean is compiled of these predicted values, ∏( pˆ˘ )
i ∈n
0
i | zit
Nn
,
• All temporarily missing prices should be imputed using
1 one of the imputation methods described in the chap-
and compared with the period t geometric mean, ∏( p t No
i ) , ter. Methods include overall mean and target mean
as in equation 6.20. i ∈n
imputations.
6.220 The overall index can be phrased as a weighted • The imputation of temporarily missing prices is especially
average of these three elements with the matched compari- important when using the short-term formulas—modified
son having a weight of 2Nm/(2Nm + No + Nn), the old of Lowe and modified Young. Imputations, which are
No/(2Nm + No + Nn), and the new, Nn/(2Nm + No + Nn), self-correcting, avoid introducing any bias into the
though preferably the weights should be expenditure shares index.
rather than the numbers of each model. • Imputations can be made either forward or backward. The
6.221 The MMM effectively ignores the last two ele- results are equivalent, and the countries can choose which
ments of the bullet points in paragraph 6.222. This proce- is most appropriate.
dure would result in no bias if the imputed quality-adjusted
• The carryforward method should not be used, except
price change of new and old varieties were the same as that
for fixed or controlled prices. This method introduces a
for matched models. The MMM might be appropriate if
downward bias into the index.
the number of new and old models—or their expenditure
weights—is small relative to matched models. This would • NSOs should define a period during which nonseasonal
be the case for the hedonic patching of permanently miss- products can be considered temporarily missing. While
ing model prices outlined previously, but not for this context this threshold varies from country to country, the most
where there is a high and frequent turnover in models. commonly used threshold is three months but can be
6.222 Even if the MMM is used with replacements, longer.
something of the dynamic universe of models is brought • Permanently missing prices require a replacement variety.
153
• Quality change refers to changes in the price-determining straightforward, and many countries apply this method
characteristics when one variety replaces another. If these for changes in size. The other explicit methods require
differences are judged to be comparable (that is, they are data and experience making explicit quality adjustments.
deemed to be similar), the price of the old and the new • Implicit or indirect quality adjustments are the second-best
variety can be compared directly and any difference in approach; however, they could be preferred given a lack
price is reflected as price change. Should the differences of data and expertise required for the explicit methods.
be such that the old and the new variety are deemed to be Implicit methods include overlap pricing and imputation.
noncomparable, a quality adjustment is needed. Quality • The rapid turnover in the models sold of select products
adjustments ensure that the index reflects only pure price (for example, televisions, computers, telecommunications
change and not changes because of differences in quality. equipment, or appliances) requires a different strategy.
• Explicit or direct quality adjustments are preferred. They Over time, these products usually have a rapid turnover in
include quantity adjustment resulting from changes in their quality characteristics. While the MMM, class mean
size or quantity, changes in option costs, differences in imputations, and hedonic price indices may be used, the
production costs, and hedonics. Quantity adjustments are chapter notes that the MMM may lead to significant bias.
154
Annex 6.1 of price changes for existing varieties and the mean of qual-
ity-adjusted replacement price changes. The bias decreases
Overall Mean (or Targeted) as either (x/m) or the difference between r1 and r2 decreases.
Furthermore, the method is reliant on a comparison between
Imputation price changes for existing varieties and quality-adjusted
price changes for the replacement or unavailable compari-
t
son. This is more likely to be justified than a comparison
Consider i = 1 . . . m varieties in period t and pi as the price without the quality adjustment to prices. For example, sup-
of variety i in period t. All varieties continue into period pose there were m = 3 varieties, each with a price of 100 in
t + 1 except for the single variety m which is replaced by period t. Let the t + 1 prices be 120 for two varieties, but
variety n. pnt+1 is the price of a replacement variety n in assume the third, that is, x = 1, is unavailable and is replaced
period t + 1. Now n replaces m but is of a different qual- by a variety with a price of 140, of which 20 is attributable
ity. There are (m − 1) matched prices between periods t and to quality differences. Then the arithmetic bias as given in
t + 1 and a single replacement price. Let A(z) be the qual- equations A6.1.3 and A6.1.4, where x = 1 and m = 3, is
ity adjustment to pnt+1 which equates its quality services or
utility to pmt+1 , had it existed, such that the quality-adjusted 1 120 120  
price pm*t +1 = A( z ) pnt+1. For the imputation method to work,  (−20 + 140) / 100 −  +  / 2 = 0 (A6.1.5)

3 
100 100  
the average price change of the i = 1 … m varieties, includ-
ing the quality-adjusted price pm*t +1 , given on the left-hand
If the bias depended on the unadjusted price of 140 com-
side of equation A6.1.1, must equal the average price
pared with 100, the imputation would be prone to serious
change from just using the overall mean of the rest of the
error. In this calculation, the direction of the bias is given
i = 1 … m ‒ 1 varieties, on the right-hand side of equa-
by (r2 − r1) and does not depend on whether the quality is
tion A6.1.1. The discrepancy or bias from the method is the
improving or deteriorating, in other words, whether A(z) < 1
balancing term Q. It is the implicit adjustment that allows
or A(z) > 1. If A(z) < 1, a quality improvement, it is still pos-
the method to work. The arithmetic formulation given is
sible that r2 < r1 and for the bias to be negative.
based on Triplett (2006), though a similar geometric one
The analysis is framed with regard to a short-term price
can be readily formulated. The equation for one unavailable
change framework. That is, the short-term price changes
variety is given by
between the prices in a period and those in the preceding
1  pm*t +1 m−1 pit +1   1 m−1 pit +1  period are used for the imputation. This is different from
+∑ t =
  (m −1) ∑
+Q (A6.1.1) the long-term imputation where a base price is compared
m  pmt i =1 pi  
t 
i =1 pi  with prices in subsequent months, and where the implicit
assumptions are more restrictive.
1 pm*t +1 1 m−1
pit +1 Table A6.1 provides an illustration in which the (mean) price
Q=
m pmt
− ∑
m(m −1) i =1 pit change of varieties that continue to exist, r1, can vary for val-
ues between 1.0 and 1.5, corresponding to a variation between
and for x unavailable varieties by no price change and a 50 percent increase. The (mean) price
change of the quality-adjusted new varieties compared with the
1 m pm*t +1 x m− x
pit +1 varieties they are replacing is assumed not to change (that is, r2
Q= ∑ t
m i = m− x +1 pm
− ∑
m(m − x) i =1 pit
(A6.1.2) = 1.00). The bias is given for ratios of missing values of 0.01,
0.05, 0.1, 0.25, and 0.5, for both arithmetic means and geo-
The relationships are readily visualized if r1 is defined as the metric means. For example, if 50 percent of price quotes are
arithmetic mean of price changes of varieties that continue missing and the missing quality-adjusted prices do not change,
to be recorded and r2 of quality-adjusted unavailable variet- but the prices of existing varieties increase by 5 percent (r1 =
ies. For the arithmetic case, where 1.05), then the bias for the geometric mean is represented by
the proportional factor 0.9759 (that is, instead of 1.05), the index
 m= x  would be 0.9759 × 1.05 = 1.0247. For an arithmetic mean, the
r1  ∑ pit +1 / pit  ÷ (m − x) and bias is −0.025; instead of 1.05, it should be 1.025.
 i= p  quation A6.1.4 shows that the ratio x/m and the differ-
E
ence between r1 and r2 determine the bias. Table A6.1 shows
 m  that the bias can be quite substantial when x/m is relatively
r2 =  ∑ pi*t +1 / pit  ÷ x (A6.1.3)
 i = m− x +1  large. For example, for x/m = 0.25, an inflation rate of 5
 
percent for existing varieties translates to an index change of
then the bias of the arithmetic mean of ratios from substitut- 3.73 and 3.75 percent for the geometric and arithmetic for-
ing equation A6.1.3 in equation A6.1.2 is mulations, respectively, when r2 = 1.00 (that is, when qual-
ity-adjusted prices of unavailable varieties are constant).
x Instead of being 1.0373 or 1.0375, ignoring the unavailable
Q= (r2 − r1 ) (A6.1.4)
m varieties would give a result of 1.05. Even with 10 percent
missing (x/m = 0.1), an inflation rate of 5 percent for existing
which equals zero when r1 = r2. The bias depends on the ratio varieties translates to 4.45 and 4.5 percent for the geometric
of unavailable values and the difference between the mean and arithmetic formulations, respectively, when r2 = 1.00.
155
Table A6.1 Example—Bias from Implicit Quality Adjustment When the (Mean) Price Change of Quality-Adjusted New
Varieties Compared with the Varieties They Are Replacing Is Assumed Not to Change (r2 = 1.00)
r Geometric Mean Arithmetic Mean
1 Ratio of Missing Varieties, x/m Ratio of Missing Varieties, x/m
0.01 0.05 0.10 0.25 0.50 0.01 0.05 0.10 0.25 0.50
1 1.00000 1.00000 1.00000 1.00000 1.00000 0.00000 0.00000 0.00000 0.00000 0.00000
1.01 0.99990 0.99950 0.99901 0.99752 0.99504 −0.00010 −0.00050 −0.00100 −0.00250 −0.00500
1.02 0.99980 0.99901 0.99802 0.99506 0.99015 −0.00020 −0.00100 −0.00200 −0.00500 −0.01000
1.03 0.99970 0.99852 0.99705 0.99264 0.98533 −0.00030 −0.00150 −0.00300 −0.00750 −0.01500
1.04 0.99961 0.99804 0.99609 0.99024 0.98058 −0.00040 −0.00200 −0.00400 −0.01000 −0.02000
1.05 0.99951 0.99756 0.99513 0.98788 0.97590 −0.00050 −0.00250 −0.00500 −0.01250 −0.02500
1.1 0.99905 0.99525 0.99051 0.97645 0.95346 −0.00100 −0.00500 −0.01000 −0.02500 −0.05000
1.15 0.99860 0.99304 0.98612 0.96566 0.93251 −0.00150 −0.00750 −0.01500 −0.03750 −0.07500
1.2 0.99818 0.99093 0.98193 0.95544 0.91287 −0.00200 −0.01000 −0.02000 −0.05000 −0.10000
1.3 0.99738 0.98697 0.97411 0.93651 0.87706 −0.00300 −0.01500 −0.03000 −0.07500 −0.15000
1.5 0.99595 0.97993 0.96027 0.90360 0.81650 −0.00500 −0.02500 −0.05000 −0.12500 −0.25000
Considering a fairly low ratio of x/m, for example, 0.05, then change of an old variety compared with a new replacement
even when r2 = 1.00 and r1 = 1.20, Table A6.1 shows that variety may be higher than price changes of other varieties
the corrected rates of inflation should be 18.9 percent and in the product group. After the introduction of the new vari-
19 percent for the geometric and arithmetic formulations, ety, its prices may fall relative to others in the group. There
respectively. In competitive markets, r1 and r2 are unlikely to may be no law of one price change for differentiated variet-
differ by substantial amounts since r2 is a price comparison ies within a market.
between the new variety and the old variety after adjusting There is thus little in economic or marketing theory to
for quality differences. If r1 and r2 are the same, then there support any expectation of similar (quality-adjusted) price
would be no bias from the method even if x/m = 0.9. There changes for new and replacement varieties, as compared to
may, however, be more sampling error. It should be borne in other varieties in the product group. Some knowledge of the
mind that it is not appropriate to compare bias between the realities of the particular market under study would be help-
arithmetic and geometric means, at least in the form they ful when considering the suitability of this approach. Two
take in Table A6.1. The latter would have a lower mean, ren- aspects need to be considered in any decision to use the
dering comparisons of bias meaningless. imputation approach. The first is the proportion of replace-
An awareness of the market conditions relating to the ments; Table A6.1 provides guidance here. The second is the
products concerned is instructive in understanding likely expected difference between r1 and r2.. It is clear from the
differences between r1 and r2.. The concern is when prices previous discussion that there are markets in which they are
vary over the life cycle of the varieties. For instance, at unlikely to be similar. This is not to say the method should
the introduction of a new model, the price change may be not be used. It is a simple and expedient approach. What
quite different from price changes of other existing variet- arguably should not happen is that it is used by default, with-
ies. Thus, assumptions of similar price changes, even with out any prior evaluation of expected price changes and the
quality adjustment, might be inappropriate. For example, if timing of the switch. Furthermore, its use should be targeted,
new computers enter the market at prices equal to, or lower by selecting varieties expected to have similar price changes.
than, prices of previous models, but with greater speed and The selection of such varieties, however, should take account
capability, an assumption that r1 = r2 could not be justified. of the need to include a sufficiently large sample so that the
Or if new clothing enters the market at relatively high qual- estimate is not subject to undue sampling error.
ity-adjusted prices, while old, end-of-season, or out-of-style The manner in which these calculations are undertaken is
clothes are being discounted, there will be bias, as r1 differs also worth considering. In its simplest form, the pro forma
from r2. setting for the calculations (for example, on a spreadsheet),
Some of these differences arise because markets are com- would usually have each variety description and its prices
posed of different segments of consumers. Indeed, the very recorded on a monthly basis. The imputed prices of the miss-
training of consumer marketers involves consideration of ing varieties are inserted into the spreadsheet and are high-
developing different market segments and ascribing to each lighted to show that they are imputed. The need to highlight
appropriate pricing, product quality, promotion, and place such prices is, first, because they should not be used in sub-
(method of distribution). In addition, consumer marketers sequent imputations as if they were actual prices. Second, the
are taught to plan the marketing mix for the life cycle of inclusion of imputed values may give a false impression of a
varieties. Such planning allows for different inputs of each larger sample size than actually exists. Care should be taken
of these marketing mix variables at different points in the in any audit of the number of prices used in the compilation
life cycle. This includes “price skimming” during the period of the index to code such observations as “imputed.”
of introduction, when higher prices are charged to skim The method described previously is an illustration of a
off the surplus from segments of consumers willing to pay short-term imputation. As discussed in Chapter 8, there is a
more. The economic theory of price discrimination would strong case for using short-term imputations as against long-
also predict such behavior. Thus, the quality-adjusted price term ones.
156
Annex 6.2 A( z t +1 ) =
pmt
and similarly for future periods of the series
pnt
Quality Adjustment Using pt p*t +i
A( z t +1 ) = mt for mt −1 for i = 2, . . ., T
a Replacement and Price pn pm
Overlap But what if the assumption does not hold? What if the rela-
t
tive prices in period t, P t = pm t , do not equal A(zt) in
t pn
Consider pm as the price of variety m in period t, pnt+1 , the
price of a replacement variety n in period t + 1; n replaces m some future period, say A( z t +i ) = αi R t ? If αi = α, the com-
but is of a different quality. Let there be overlap prices for m parisons of prices between future successive periods, say,
and n in period t and let A(zt+1) be the quality adjustment to between t + 3 and t + 4, are unaffected, as would be expected,
pnt+1 which equates its quality to pmt+1 such that the quality- since variety n is effectively being compared with itself:
adjusted price pm*t +1 = A( z t +1 ) pnt+1. Now the quality adjust-
ment to prices in period t + 1 is defined as pm*t +1 = A( z t +1 ) pnt+1 pm*t + 4 pm*t +3 α R t pnt+ 4 pnt+ 4
/ = = (A6.2.3)
which is the adjustment to pn in period t + 1 that equates it to pmt−1 pmt−1 α R t pnt+3 pnt+3
pm in period t + 1 (had it existed then).
A desired measure of price changes between periods t − 1 However, if differences in the relative prices of the old and
and t + 1 is thus: replacement varieties vary over time, then:
pm*t + 4 pm*t +3 α4 pnt+ 4

( pm*t +1 / pmt+1 ) (A6.2.1) /
pmt−1 pmt−1
=
α3 pnt+3
(A6.2.4)
The overlap formulation equals this when: Note that the quality difference here is not related to the
technical specifications or resource costs, but to the relative
p*t +1 p t +1 p t +1 p t prices that consumers pay.
t −1
= A( z t +1 ) nt −1 = n t × tm−1 (A6.2.2)
pm pm pn pm
157
Annex 6.3 New and old models may coexist for some time and
existing models having their prices increased following the
The Nature and Extent of the introduction of a new model is of interest. Such a pricing
strategy can be used by a multiproduct firm that anticipates
Index Number Bias If Only the introduction of a new model. A multiproduct monopolist
can increase the prices of existing models because some of
Matched Varieties Are Used the demand for existing models that would usually be lost to
competitors, because of the price increase, will now not go
to the competitor’s products but will go instead to the firm’s
Sample degradation and differences in the (quality-adjusted) new model. The new model will cannibalize some of the
prices of unmatched new, unmatched old, and matched mod- existing model’s sales that would otherwise be lost because
els can lead to bias in matched-model price indices that take of the price increase in the existing model (Figure A6.2,
into account only models available in two consecutive periods. quadrant II). For example, the price of old, branded pharma-
The nature and extent of such bias depends on the frequency ceutical drugs can increase after the expiration of a patent
with which manufacturers turn over their models and the pric- and introduction of new generic models. This is due to price
ing strategy retailers employ over the life cycle of the models. discrimination with some market segments remaining with
For example, assuming that matched prices go down, if the particularly strong preferences for the old models willing
quality-adjusted prices of new unmatched models in period to pay higher prices (Figure A6.2, quadrant II). However, it
t = 2 are higher than their matched counterparts in period 2, has been argued that any such effect may be outweighed by
and if the quality-adjusted prices of old unmatched models in the need to cut the prices of the existing models to prevent
period 1 are lower than their matched counterparts in period the existing model’s sales cannibalizing sales of the new,
1, then there will be a larger fall in the matched-model index more profitable model (Figure A6.2, quadrant IV). A study
between periods 1 and 2 compared with a hedonic index that of computer processors and disk drives found that with the
uses all of the data. Similarly, if the quality-adjusted prices of introduction of products embodying new technologies the
unmatched new models are below matched ones in period 2, prices for older products decline rapidly to permit the older
and the quality-adjusted prices of old models above matched technology to compete with the newer one for a limited
ones in period 1, there will be a smaller fall in the matched- time, but the old technology is eventually driven out.
model index compared with a hedonic index that uses all of New models may have above average quality-adjusted
the data. The nature and extent of the matched models index prices in their period of introduction because firms “price
bias thus depends on the pricing strategy adopted for new and skim” market segments willing to pay a premium for the
old models. Indeed, if hedonic-adjusted prices are consis- new model over and above that due to its improved quality
tently above or below average prices for unmatched new and (Figure A6.2, quadrant I). Indeed, marketing texts advocate
old models some of the bias will cancel. price skimming as one of two “new product” pricing strate-
The case for old unmatched models having below aver- gies. The alternative strategy is “market-penetration” pric-
age quality-adjusted prices is based on an inventory-clearing ing for which a low initial price is set for a new model to
argument. For an old model near or at the end of its life attract a large number of buyers quickly to win market share
cycle retailers want to clear out the remaining inventory and take advantage of falling costs because of scale econ-
from both their warehouses and store shelves so they have omies. Such pricing may initially be possible because the
room to stock and display the replacement model. They do new model is based on new, lower cost components that can
not wish the old model to cannibalize some of the sales of provide a feature set that is comparable to existing models,
the new model which may have a higher price (profit) mar- but at a lower price point. In either event, quality-adjusted
gin. The extent of any such cannibalization will depend on prices of new models may have below average prices (Fig-
the cross-price elasticities between the new and old models. ure A6.2, quadrant III).
This inventory-clearing argument is noted in quadrant IV of Figure A6.2 summarizes these positions. The combina-
Figure A6.2. tion of above average prices for new models in quadrant
Figure A6.2 Matched-Model Price Index Bias and Pricing Strategies
New Unmatched Models Old Unmatched Models Pricing Matched-Model Price Index Bias
Quality-Adjusted Prices Market Skimming Multiproduct Monopoly Pricing Strategy; I and IV Upward
above Matched-Model Price Discrimination to Segments
Prices with Sticky Downward Pricing; Old
Technologies Reduce Prices
[quadrant I] [quadrant II] II and III Downward
Quality-Adjusted Prices Market-Penetration Pricing; Inventory Clearing I and II Countervailing
below Matched-Model Low Unit Costs, New
Prices Producing Technology
[quadrant III] [quadrant IV] III and IV Countervailing
159
I and below average prices for old models in quadrant IV digital cameras compared to film-based cameras. Given the
leads to an overall net downward bias of the matched-model current differential in product costs for the two technolo-
price index. Similarly, pricing in quadrants II and III leads gies, and where the two categories are in their respective
to matched-model indices that are biased upward. However, life cycles, relatively greater effort is likely to be placed
pricing in quadrants I and II leads to an indeterminate bias, on R&D in new models that reduce unit costs for digital
with countervailing effects from the new and old above aver- cameras compared to R&D in film-based camera models
age priced models. The outcome from pricing in quadrants that reduce unit costs. Products in a mature stage of their
III and IV is also indeterminate, with countervailing effects category life cycle, where R&D development is relatively
from the new and old below average priced models. small and product enhancing, as opposed to cost reducing,
It is possible to say something about the likely pric- may be more likely to have above average quality-adjusted
ing strategies of different products. Consider the case of prices for new models.
160
MAINTAINING THE SAMPLE 7
Introduction 7.4 Consider three universes:
7.1 When a new good or service is produced and con- • An intersection universe, which includes only matched
sumed, there is a need for it to be included in the index as products
soon as possible, especially if the good or service will have • A dynamic double universe, which includes all products
relatively high sales. New products might have quite dif- in the price reference period and all in the current period,
ferent price changes than existing ones, especially at the although they may be of different qualities
start of their life cycle. In the initial period of introduction • A replacement universe, which starts with the price
of a new product, producers and retailers often set higher reference period universe but also includes one-to-one
prices than might be attainable once the market settles into a replacements when a product from the sample in the price
competitive equilibrium. There is a related problem of obso- reference period is missing in the current period
lete products, as the price changes of such products may be
unusual. The products will be at the end of their life cycle 7.5 A sample of representative varieties that comprises
and may be priced at unusually low prices to clear the way only those varieties selected for pricing in the price refer-
for new models. ence period and having a matched sample provides an esti-
mator of the price change for the intersection universe.
7.6 It is difficult to ascertain the extent to which match-
Sample Maintenance and Matching ing from the intersection universe constrains the penetra-
tion of the sample into the dynamic double universe, since
7.2 Matching is designed to avoid price changes being national statistical offices (NSOs) generally do not collect
affected by quality changes. As described in Chapter 6, data for the dynamic double universe. Its extent will, in any
matching refers to comparing like with like. Its adoption event, vary between products. Scanner data have been used
constrains the sampling to a static universe of goods and to determine the coverage of intersection and replacement
services that exists in the price reference period. Therefore, universes for consumer durables, finding the intersection
items or varieties that exist in the reference period, but not in and replacement universes to be highly restricted.
the current period, are not matched, and similarly, those new 7.7 A first implication of this issue described in the pre-
goods and services existing in the current period but not in vious paragraph is that for permanently missing varieties,
the reference one. The challenge is that the products that are variety replacement is an opportunity to bring in a variety
not in the matched universe (that is, the new product appear- with a relatively large sales value to increase the coverage
ing after the reference period and the old products that dis- of the index. However, the selection of variety replacements
appeared in the current period) may be the ones whose price by price collectors puts coverage of the sample to some
changes differ substantially from existing matched ones. extent under the control of the price collectors. Guidelines
They may include different technologies and be subject and training on direct replacements in particular product
to different (quality-adjusted) strategic price changes. The groups may be needed. Where updating is infrequent, rotat-
same method used to maintain a constant-quality sample ing the sample has benefits. Sample rotation, as outlined in
may give rise to a sample biased away from technological paragraphs 7.18–7.24, is equivalent to initiating a new sam-
developments. Furthermore, when this sample is used to ple, but for specific product groups that maintain the same
make imputations (as discussed in Chapter 6) to the price weights until the next update. A particularly useful method
changes of replacement products, it reflects the technol- would be to refresh the sample for product groups with a
ogy of a sample not representative of current technological high level of sample churn, though as a rule, more frequent
changes. The matched-model method similarly constrains updating is advocated.
the incorporation of new products.
7.3 The problem described in paragraph 7.2 has been
outlined with regard to a variety having to “exist” in both Item Replacement or Substitution
periods being compared. The concern for price collection in 7.8 The price collectors often are best placed to select
an outlet is for the price collector to be able to collect a price replacement varieties for price collection. They are aware of
quote for the period for the comparable, matched variety the price-determining characteristics of the products being
selected and priced in the previous period. A variety may not produced and purchased, and their terms of sale. The selec-
be found by the price collector on an outlet shelf in a given tion of the replacement for price collection might be quite
month and thus not “exist” in the previous sense, but still be obvious to the price collector, especially if there is only a
consumed, sold by outlets not sampled or sold by the outlet slight, nominal improvement to the product. For example,
on a day of the month not sampled. Similarly, a price may the “improved” product can be simply a replacement vari-
be collected but there may be no or limited expenditure on it. ety sold instead of the previous one. The replacement could
161
have a different code or model number and will be known • If the replacement variety has relatively high sales, is of
to the price collector as simply a different color or packag- comparable quality, and at the same stage in its life cycle
ing. The specification list given to the price collector is a as the one being replaced, then its selection will minimize
critical prompt to identify when a variety has changed, and bias, but this replacement variety would also be at the end
it is important that all price-determining characteristics have of its life cycle and subject to replacement soon.
been included. • If there is more than one new (replacement) variety and
7.9 The price collector, supported by the consumer the most comparable one is selected in line with the old
price index (CPI) compilers and prompted by the specifica- technology, it will have low market share and unusual
tion list, takes on the role of identifying whether a variety is price changes.
of comparable quality or not. If it is judged to be comparable • Given the availability of advance market or consumption
when it is not, the quality difference is taken to be a price information, replacements undertaken before obsoles-
difference, and a bias will result if the unrecognized quality cence are likely to increase the sample’s share of the mar-
changes are in a consistent direction. Informed comparable ket, include varieties more representative of the market,
substitution requires general guidelines on what makes a and facilitate quality adjustment.
suitable replacement as well as product-specific information
on likely price-determining characteristics. It also requires
timely substitution to maximize the probability of an appro- Outlet Replacement
priate substitute being available. Chapters 5 and 6 provide
7.12 The problem of variety substitution is analogous to
further information on the need for and type of training to
the problem that arises when an outlet closes. It may be pos-
be provided to price collectors in this regard. The selection
sible to find a comparable outlet not already in the sample,
of replacement varieties is very much product-specific and
for example, a franchise, in the same chain. It may also be
guidelines for price collectors should focus on specific prod-
possible to find a noncomparable outlet for which, in prin-
uct groups and tailored to their needs.
ciple, an adjustment can be made, for example, for the better
7.10 The results from hedonic regressions can be use-
quality of service of the new (replacement) outlet. It is not
ful in the selection of varieties and their replacements. The
unusual for an outlet to close following the introduction of
results provide an indication of the major quality factors that
a new one, and in this case, there is an obvious replacement
make up the good or service, with regard to explaining price
outlet. However, if the new outlet has comparable prices
variation. The selection of varieties will be more representa-
but a better range of varieties, delivery, and service quality,
tive and the coefficients from hedonic regressions would be
there is a gain to purchasers from substituting one outlet for
more tailored to the sample because of their subsequent use
the other. Since such facilities have no direct price, it is dif-
to estimate quality-adjusted prices.
ficult to provide estimates of the value of such services for
7.11 Price collectors traditionally are required to find
an adjustment to be made for the better quality of service.
substitute varieties that are as similar as possible to the vari-
Simply rotating the new outlet into the sample via the over-
eties being replaced. This maximizes the likelihood that the
lap method would miss the quality difference, as outlined
old and replacement variety will be equivalent and minimizes
in Chapter 6. The index thus would have an upward bias,
the need to employ some method of quality adjustment.
which would be eliminated when rebasing. In such cases,
Replacement varieties should be chosen to make the sam-
substituting an old outlet for a new one that provides a simi-
pled varieties more representative of the dynamic universe.
lar standard of service would be preferable.
The inclusion of a popular replacement variety to refresh the
sample allows for a useful and accurate price comparison and
increases the chance of an appropriate quality adjustment. It Sample Rotation, Chaining, and Hedonic
is of little merit to substitute a new variety with limited sales Indices
for a missing variety with limited sales, as the index would 7.13 For some product groups, the product samples will
become more unrepresentative. If replacements are made for become quite out of date if left to the next update or revision
varieties at the end of their life with popular replacements for the sample to be reinitiated. This is especially so if the
varieties at the start of their life cycle, the quality adjustment rebasing is infrequent. Sample rotation is equivalent to ini-
will be substantial and substantive. More frequent sample tiating a new sample, but for a product group that maintains
rotation or directed replacements will be warranted for some the same weights until the next rebasing. Sample rotation
item areas. The selection of replacement varieties: is undertaken for specific product groups at different points
• Offers an opportunity to cut back on sample bias in the in time to save on the resources required if all the product
period of replacement. groups had their products rotated at the same time. The cri-
teria for choice of product groups to benefit from sample
• The more frequent the replacement, the less the bias.
rotation, and the timing of the rotation, should be clearly and
• If there is more than one new (replacement) variety in the openly scheduled in advance according to objective criteria.
market, there may still be bias since only the most popular 7.14 It is important to recognize the interrelationships
one will be selected, and it may be at a different stage in among the methods for addressing product rotation, product
its life cycle than others and priced differently. replacement, and quality adjustment. The rotation of CPI
• It is assumed that accurate quality adjustments are made product samples is a form of product substitution, although
on replacements. The less frequent the replacement, the not “forced” by a missing variety price but undertaken for a
more difficult this might be, because the latest replace- general group of items/varieties to update the sample. Rota-
ment variety on the market may have more substantial tion has the effect of making future forced replacements less
differences in quality than earlier ones. likely. The implicit assumptions in its use are equivalent to
162
Maintaining the Sample
those for the overlap adjustment method: price differences a new model that appears in period 3 can be introduced into
are an adequate proxy for the value of the quality change the index in the period 3 to period 4 link. However, the new
between products disappearing from the sample and replace- model’s effect on the price index in the initial period of intro-
ment products. Consider the initiation of a new sample of duction, period 3, is ignored for the period 2 to period 3 link.
items. Prices for the old and new samples are collected in Similar concerns arise for disappearing models. If the last
the same month and the new index is compiled based on the period a price is observed for a model is period 1, its effect
new sample, with the results being linked to the old. on the price index is lost for the period 1 to period 2 link.
7.15 As an example, assume that the initiation of a new This situation, and its resolution, is outlined in Chapter 6 in
sample is taking place in January. The prices of an old vari- the context of product groups with a high turnover of dif-
ety in December and January are, respectively, 10 and 11, ferentiated models with identifiable price-determining char-
a 10 percent increase, and those for the replacement vari- acteristics. Only subsamples of matched items exist between
ety in January and February are 16 and 18, respectively, an successive periods and these form biased price comparisons
increase of 12.5 percent. The new variety in January is of a since it is on the dumping of old models that unusually low
better quality than the old, and this difference in quality is prices, and on the introduction of new ones that unusually
estimated to be worth 16 − 11 = 5; that is, the price difference high prices exist. Hedonic price indices allow the price of
is assumed to be equal to the value of the quality difference, old model in period 1 to be used in the period 1 to 2 price
which is the assumption implicit in the overlap method. comparison and the price of the new model in period 3 to be
If the price of the old variety in December was compared used in the period 2 to 3 price comparison, because it uses
with the quality-adjusted price of the new variety in January the prices of all models in each period, though adjusted for
under this assumption, the price change would be the same: their differences in quality characteristics. In this case, there
10 percent ([16 − 5]/10 = 1.10). However, if the price differ- is no need for the use of the matched-model method.
ence in January was an unsuitable reflection of the quality 7.19 Hedonic indices are applicable if a new model/
difference (for example, the old variety was being sold at variety is not entirely new—it is an evolutionary product
a low price to clear the market for the new one), then the in the sense that it is providing more services than those of
implicit assumption underlying the overlap method would the old model. The price can be determined with regard to
not hold. In practice, the need to simultaneously replace and a different combination of the existing price-determining
update a large number of products requires the assumptions characteristics.
of the overlap method. This process is not error-free, and
in cases where the assumptions are likely to be particularly
untenable, explicit adjustments should be used, resources Incorporation of New Products
permitting (as discussed in Chapter 6).
7.16 Sample rotations to refresh the sample between New Products and How They Differ from
rebasing are expensive exercises. However, if rebasing is Quality Changes
infrequent, this might be appropriate for particular product 7.20 A new model of a product may provide more of a
groups where there is a substantial loss of products. The currently available set of service flows. Chapter 6 describes
need for a metadata system to facilitate such decisions is how new models often with quality improvements replace
outlined in paragraphs 7.83–7.86. The use of more frequent existing ones. A number of methods to incorporate the
sample rotation helps the process of quality adjustment in replacement models into the CPI and help to maintain the
two ways. First, the updated sample will include newer vari- CPI’s representativity were outlined. However, there may
eties, comparable replacements with substantial sales will be be many characteristics of the new model that go beyond the
more likely to be available, and noncomparable ones will be service flow of the existing model.
of a more similar quality, which will improve the accuracy 7.21 The first practical concern with defining a new
of explicit quality adjustments. Second, because the sample product’s quality changes against the existing model is that
has been rotated, there will be fewer missing varieties and the new model cannot be easily linked to the existing model
less need for quality adjustments. as a continuation of an existing resource base and service
7.17 An extension of more frequent sample rotation is to flow because of the nature of its “newness.” For example,
use a chained formulation in which the sample is reselected some forms of frozen foods, self-driving and all-electric and
each period. The prices of all varieties available in each suc- hybrid cars, computers, printers, and mobile phones, while
cessive linked comparison are compared: those available, extensions of existing products, have dimensions of service
for example, in both January and February are compared for that are new. Second, new products can generate a welfare
the January to February link, while those available in both gain to purchasers and surplus to producers when purchased/
February and March are compared for the February to March sold at the time of introduction and the simple introduction
link. The index for January to March is derived by succes- of the new product into the index, once two successive price
sive multiplication of the two monthly links. The principles quotes are available, misses this gain.
and methods of this chained formulation were outlined in 7.22 Many product markets are defined by a multitude
Chapter 6 in the context of products in which there is a rapid of brands and differentiated offerings along with a rapid
turnover of models, and such principles apply in this case. turnover in varieties. In some cases, there are core brands
Similarly, the use of hedonic indices and short-term com- and varieties that may be used for CPI price measurement
parisons might be useful in this context (see Chapter 6). as representative variety. However, the concern remains that
7.18 The chained formulation allows the price changes the rationale for the introduction of new or differentiated
of a new model to be included in the index as soon as the brands and varieties is to be distinct and not exact substi-
model can be priced for two successive periods. For example, tutes for existing ones. If the CPI misses both an increasing
163
variety of offerings and their distinctive quality improve- analytically appropriate, but by what is practically meaning-
ments, the index may misrepresent actual price inflation. ful for the needs of CPI compilation and calculation. Practi-
However, the magnitude and turnover in offerings in differ- cal needs are important in this context, especially because
entiated markets can make the definition and measurement the methods for providing reliable estimates of consumer
of quality change and “newness” impractical. surplus on a large-scale basis are not practically possible
7.23 A new product can be identified with regard to the given the substantial resource needs of data and economet-
absence of substitutes and the ensuing monopolistic power. ric expertise.
For example, some new movies, digital games, and toys
may have small cross-elasticities with other movies, games, Incorporation of New Products: Major
and toys; their shared service is to provide entertainment Concerns
and they are only similar in this respect. The same argu-
ment may apply to new books or breakfast cereals. There 7.27 There are two major concerns regarding the incor-
are many new forms of existing products that are not easily poration of new products into a CPI: their identification and
substitutable for similar products but can generate consumer detection, and the related decision on the need and timing
utility well above that of the preexisting counterparts and for their inclusion. This refers to both the weight and price
are not always sold at a higher price. changes of the new products.
7.24 A more practical classification that will meet the 7.28 For example, the levels of expenditure on mobile
needs of CPI compilation is to consider evolutionary and phones were in some countries at such a significant level
revolutionary products. Examples of evolutionary prod- that their early inclusion in CPIs became a priority. They
ucts would be new models of household appliances such as shifted from no expenditure to be a quite large proportion of
refrigerators and washing machines where improvements expenditure in their product group in a short time. Further-
in quality are introduced from time to time. Consequently, more, their price changes were atypical of other goods in
in theory at least, it should be possible to quality adjust their product group. Many new products can command sub-
for any differences between a preexisting and an evolu- stantial sales and be the subject of distinct pricing strategies
tionary product, as outlined in Chapter 6. Where the new at introduction prompted by a need to recuperate research
model is a one-for-one replacement with the old model, the and development expenditure and take advantage of an
sample is maintained. However, where there are many new opportunity to expand market share and make profits, as, for
brands/varieties spawned by a new innovation, the one-on- example, with high-technology goods, such as computers,
one replacement will not reflect the representativity of the pharmaceuticals, and entertainment gaming devices.
sample. Revolutionary products are products that are sub- 7.29 Waiting for a new product to be established or for
stantially different from preexisting ones. Revolutionary the rebasing of an index before incorporating new products
products represent a good or service that: may lead to errors in the measurement of price changes if the
unusual price movements are ignored at critical stages in the
• Was not included and could not be included in the price product’s life cycle. NSOs should have strategies in place
index during the initial selection of the current basket and for the early identification of new products and mechanisms
which is later available for possible inclusion in the index for their incorporation either at launch, if preceded by major
• Cannot be easily linked to the service flow or produc- marketing campaigns, or soon after, if there is evidence of
tion technology of existing goods and services (that is, it market acceptance. This should form part of the metadata
represents a distinct departure from previously available system. Waiting for a new product to achieve market matu-
products to the extent that it is a step change with regard rity may result in an implicit policy of ignoring the different
to technology or utility to the customer) price movements that accompany their introduction. New
• Has a recognizable and generally accepted new benefit to product prices may be very close to, or even lower than, the
consumers as a result of becoming available existing ones and serve to expand the new product’s mar-
ket share. There may be something in the “newness” of the
7.25 The last two points in paragraph 7.24 help to distin- technology and production that enables a price reduction for
guish a revolutionary product from an evolutionary product. a better product.
A revolutionary product is an entirely new good or service
that is not closely tied to a previously available product. A
revolutionary product tends to be a good or service that is Methods for Incorporating New Products
expected to satisfy some need in a new way and is unlikely 7.30 The methods outlined in paragraphs 7.32–7.52
to fit neatly into an existing CPI item category. For example, include those that fall under normal CPI procedures and
a mobile phone, although an extension of an existing flow those that would require exceptional treatment. In the case
of service (telecommunication), has a new dimension of of normal CPI procedures, the focus will be on evolutionary
service (it provides the opportunity to make “mobile” calls products, and the rebasing of the index, rotating of prod-
away from a fixed telephone) and is a distinct product from ucts, and introduction of new products as replacements for
existing landline telephone services (it is a step change in disappearing ones are considered. Much use is made of the
technology). It is therefore an example of a revolutionary overlap method where the price changes of new products are
product. linked/spliced onto the index. Of importance in this context
7.26 Quality adjustments to prices are therefore suit- of maintaining the representativity of the sample is a move,
able for evolutionary products, but unsuitable for revolu- where appropriate, away from one-for-one replacements and
tionary products. The definitions are designed to distinguish the introduction of appropriate weights for the new prod-
between the two types of products not with regard to what is ucts The overlap method invokes some highly restrictive
164
assumptions, and these will be explored in the context of there is the opportunity to reappraise the representative
explicit quality adjustments, all following the principles and varieties selected for regular pricing and replace them as
practices described in Chapter 6. The case of revolutionary necessary.
products is more problematic from a practical measurement
perspective. The focus on paragraphs 7.43–7.52 is less on
Sample Rotation (Reinitiation)
an understanding of how such new products can be incor-
7.33 In many countries, rebasing is infrequent. Rather
porated into the CPI, but is instead on the need for users to
than continuing to utilize a sample of varieties that have
be made aware of the conceptual limitations of the resulting
become increasingly unrepresentative, NSOs may select
measures.
new samples of outlets and of representative varieties
between regularly scheduled index updates. This does not
Evolutionary Products need to be done for all product groups in the same period,
Updating and Chaining the Index with different major groups of the Classification of Individ-
7.31 A new product may be readily incorporated into ual Consumption According to Purpose (COICOP) having
the index at the time of updating or revising the index. If the their samples updated periodically, and perhaps with dif-
new product has, or is likely to have, substantial sales, and ferent frequencies, according to needs and resources. Dur-
is not a replacement for an existing one, or is likely to have ing this process, the weights remain constant until the next
a much higher or lower market share than the existing one it rebasing. This should involve a reappraisal of the product
is replacing, then new weights are necessary to reflect this. groups, locations, outlets, and representative varieties within
New weights are fully available when updating the index. outlets, and the elementary aggregates. This may be under-
There will be a delay in the new product’s full inclusion, and taken on a phased basis to reduce the workload with more
the extent of the delay will depend on how close its intro- frequent rotations directed to product groups experiencing
duction is to the next rebasing, the frequency with which rapid changes. However, unlike rebasing, there is neither
the index is updated, and the time lapse between determin- a comprehensive introduction of new weights nor a sam-
ing the new expenditure weights from a household budget pling basis for identifying new locations, outlets, and prod-
survey (as described in Chapter 3) and their use in index uct selection. A continuing matching of prices of products
compilation (see Chapter 9 for more information). How- that are unrepresentative of expenditure patterns is also not
ever, even if the index was updated annually and chained, it desirable. Where resources do not permit a regular rebasing,
would take until the annual update before weights could be sample rotation provides a viable mechanism for making
assigned, and even then, there might be a further delay in the some in-roads into maintaining sample representativity, but
sampling and finalizing the survey results for the weights. it is not a complete solution.
Annual updating allows for a relatively early introduction of
new products and is advised when the weights are not keep- Sample Rotation (Reinitiation) in High-Turnover
ing pace with innovations in the product market. Product Groups and Hedonic Regressions
7.32 Updating provides two further opportunities to 7.34 In product groups where new products are continu-
maintain/refresh the sample: ously appearing and old ones disappearing, the sample of
products can quickly become outdated and unrepresenta-
• The first is to develop a new set of elementary aggre- tive of what consumers are purchasing. The existing sample
gates. Some elementary aggregates may no longer have within a class may cover a broad and representative range of
a sufficiently substantive weight and expected longev- varieties that were available in the market during the price
ity to merit inclusion in the CPI, according to some cut- reference period but not be representative of all the varieties
off criterion, as outlined in Chapter 4, while others may currently purchased. In such a case, the sample of variet-
have now passed the cutoff and be expected to continue ies within each elementary aggregate can be totally resa-
to gain importance. The formal inclusion of new evolu- mpled to reflect current spending patterns. One approach to
tionary and revolutionary products at this juncture, along facilitate the sample rotation or reinitiation process uses an
with their weights, and removal of “obsolete” elementary overlap. The old and new sample prices are collected in an
aggregates, provides an opportunity for an NSO to for- overlapping period.
mally announce and integrate new products in the CPI 7.35 Care should be exercised in the application of this
along with their weights. The price changes of the new technique for product areas where there is a rapid change
CPI basket are chain linked onto the price changes of the in the turnover of models. For example, new generations
old, as described in Chapter 9. of electronic goods, such as television sets, might be incor-
• The second is the reinitialization of the sample of out- porated into the price index using the overlap method. The
lets and representative varieties within outlets, as out- new products would have their price change measured when
lined in Chapter 4. On rebasing, the sample of outlets for two successive price quotes are available, possibly at the
the existing elementary aggregates can be revisited with same price at the start of the new model’s life cycle, and
the purpose of introducing new outlets, especially those this would be linked to the price change of the old model it
invoking new retailing technologies, including internet is replacing near the end of its life cycle. The method might
purchases, and remove obsolete outlets. There might also miss any effective fall in (quality-adjusted) prices from new
be a need for a reweighting and switching of the sample technological developments.
toward specific locations, for example, because of new 7.36 Modern data sources including scanner data and
metropolitan areas being developed because of new trans- web-scraped data, as outlined in Chapter 10 and Annex 5.6
portation links or job opportunities. Within each outlet, of Chapter 5, enable a continuous resampling of products
165
with high rates of technological change. They also include With sample augmentation, new items or varieties are intro-
price-determining characteristics data. For example, for duced into the index. To illustrate sample augmentation,
washing machines, the spin speeds, capacity, dimensions, take the case where a new canned fish, tuna packed in water,
programs, brand, and many more salient price-determining has been recently introduced to the retail market and has
characteristics are provided alongside the price for each significant popularity in the shops. This new variety can be
model. In Chapter 6, hedonic regressions were advocated as added to the existing sample in the elementary aggregate for
one option for avoiding the pitfalls of overusing the overlap canned fish as shown in Table 7.1.
method. 7.40 In the example in Table 7.1, period 1 displays the
7.37 It is not necessary for NSOs to wait until a product sample currently used. The NSO then decides to add the
is obsolete before the new one is introduced. It is quite feasi- additional variety (tuna in water) in period 2. But prices for
ble for NSOs to preempt the obsolescence of the old product two periods are required before there is a matched pair. The
and to direct an early substitution. In some product groups, period 2 elementary aggregate index (140.6) is computed
the arrival of a new product is well advertised in advance of using the geometric mean of short-term price relatives for
the launch, while in others it is feasible for an NSO to have the original five varieties (1.0307) multiplied by the previ-
more general procedures for directed substitutions. With- ous period elementary aggregate index (136.4). In period 3
out such a strategy and infrequent rotation and rebasing, a the new variety’s price is available in both periods and the
country’s CPI would be open to serious new product bias (as elementary aggregate index (143.3) is calculated by taking
described in Chapter 12). the geometric mean of short-term price relatives for the six
available varieties (1.0190) multiplied by the period 2 price
Forced Replacement index (140.6). To estimate the period 2 variety level index
7.38 It is quite straightforward to extend the principles for “tuna in water,” the same value is assigned as the ele-
of replacements outlined in Chapter 6 to more than one mentary aggregate index (140.6). This implicitly assumes
representative variety simply by using the weighted (or that the price trend for the new variety from the price ref-
unweighted) price change of more than one replacement erence period is the same as that for all the other varieties
variety. Indeed, existing samples may be supplemented by within the elementary aggregate. If the long-term price rela-
new varieties even when a replacement is not motivated by tive method is used to calculate the elementary aggregate
a permanently missing variety. A comparison at the elemen- index, then the base price for tuna in water is estimated by
tary aggregate level between, for example, prices in 2020 dividing the first price of the new variety (60.00) by the
and prices in June 2021 may be undertaken in two stages: long-term price change (1.406) to get a base price of 42.66.
first, by comparing average prices for several matched rep- The aggregate index is calculated as the geometric mean of
resentative varieties in 2020 with average prices of compa- the variety indices.
rable representative varieties in May 2021; multiplied by, 7.41 The example in Table 7.1 is for augmenting the
second, a comparison of average prices in May 2021 com- sample using a single additional variety.
pared with June 2021. The basket of representative varieties
in the May to June 2021 stage may include new representa- Revolutionary Products
tive varieties in addition to the replacements for the ones 7.42 Revolutionary new products are often high-profile
used in 2020 to May 2021 stage. In introducing such repre- and a failure to properly integrate them into a CPI in a timely
sentative varieties there is an implicit weighting, and care manner can damage the credibility of the index.
has to be exercised to ensure it is meaningful. At the elemen- 7.43 For evolutionary products, the sample of products
tary level of aggregation, the Jevons index is the ratio of was outdated, and new varieties were selected within exist-
geometric means, which is equal to the geometric mean of ing elementary aggregates. Occasionally, new revolutionary
price relatives (for more information see Chapter 6 of the products arise that are not covered within the scope of exist-
publication Consumer Price Index Theory). Equal (implicit) ing elementary aggregates but fall within the more widely
weight is given by the Jevons index to each variety’s price defined COICOP classes. They may be primarily sold by a
relative. Explicit weights may also be used. new outlet, or type of outlet, and there will be no previous
products to match them against and make a quality adjust-
Sample Augmentation ment to prices since, by definition, they are substantially
7.39 Sample augmentation does not require a represen- different from preexisting goods. Further, there is no refer-
tative item to be missing, as was the motivation in Chapter 6. ence period weight to attach to the new outlet and product.
Table 7.1 Example of Sample Augmentation

Variety Period 1 Period 2 Period 3
Price Price Price Price Price Index Price Price Price

Index Relative Relative Index
Canned Mackerel (in oil) 125.0 50.0 51.0 1.0200 127.5 51.0 1.0000 127.5
Canned Anchovies (in oil) 133.3 45.0 47.0 1.0444 139.2 48.0 1.0213 142.2
Canned Tuna (in oil) 150.0 50.0 52.0 1.0400 156.0 52.0 1.0000 156.0
Canned Salmon (in oil) 145.5 55.0 55.0 1.0000 145.5 57.0 1.0364 150.8
Canned Herring (in oil) 130.0 40.0 42.0 1.0500 136.5 43.0 1.0238 139.8
Canned Tuna (in water) — — 60.0 140.6 62.0 1.0333 145.3
Elementary Aggregate Index 136.4 1.0307 140.6 1.0190 143.3
166
7.44 Adding a new elementary aggregate and redis- index is derived using the index for the new elementary
tributing the weight for the COICOP class to all the ele- aggregates along with the new weights for the elementary
mentary aggregates (or varieties) is a way of including indices ([99.6 × 0.2] + [94.0 × 0.1] + [98.0 × 0.7] = 97.9).
such products. Table 7.2 illustrates how to redistribute the 7.51 This use of the overlap method is similar to the
weight. inclusion of new elementary aggregates on rebasing.
7.45 As an example, assume that currently desktop com-
puters have a weight of 60 percent and laptops 40 percent,
respectively, within subclass 08.1.3.1 (computers, laptops,
Introducing New Items and Higher-Level
and tablets) of COICOP 2018 and that the latest informa- Weights in the Consumer Price Index in
tion from importers of information processing equipment between Basket Revisions
indicates that sales of desktop computers to households 7.52 The example of adding an elementary aggregate
(that is, where sales to businesses have been identified and presented in Table 7.2 provides a method of introducing
excluded) now have a market share of 20 percent, laptops a new revolutionary product index, in this case for tablets
10 percent, and tablets 70 percent. The NSO can use this within the COICOP subclass 08.1.3.1 (computers, lap-
information to introduce a new elementary aggregate for tops, and tablets). The relevant index now includes the
computers, laptops, and tablets. The weights at the subclass contribution to price change of the tablets item within the
levels, including subclass 08.1.3.1, remain fixed for aggre- subclass. The subclass, however, is likely to be underrep-
gating to the class-, group-, and division-level indices while resented within the corresponding class (08.1.3 Informa-
the relative weighting of the elementary aggregates within tion Processing Equipment), group (08.1 Information and
the subclass level are allowed to change as new aggregates Communication Equipment), division (0.8 Information
are added. Thus, there is a two-tier aggregation system in and Communication), and the all-items CPI, because its
which the weights at the subclass level remain fixed at the weight does not reflect the increased expenditure result-
weight reference period level and the weights within the ing from the introduction of tablets. In most instances,
subclasses at the elementary aggregate level are changed NSOs are hesitant to change the weights for the subclass,
when new varieties are added, though are constrained to add class, group, and division until a new set of weights for
to the unchanged subclass weight. all items can be obtained from a recent household budget
7.46 An overlap approach similar to sample rotation is survey. NSOs will typically include the tablet in subclass
used in this situation, where a new sample is selected, and 08.1.3.1 without changing the weights for higher-level
an elementary aggregate is added. Prices are collected for aggregates.
both the old and new samples in the same period and the 7.53 If the NSO has no plans for conducting a house-
old sample is used for compiling the current period index hold budget survey in the near future to update weights for
(period 2) and the new sample for the next period (period 3). the subclass, class, group, and division, there are alterna-
This is illustrated in Table 7.2. tive sources it can consider for updating weights. To update the
7.47 For each of the two elementary aggregates in weights for high-level aggregates, the NSO first needs to
period 1, a new variety to price in period 2 is selected estimate weights for each subclass within the class where
together with a sample of tablets (that is, the new revolu- the new product is added so that these subclass indices can
tionary product). In period 2, the old sample is used to cal- be aggregated to the class level. Likewise, new weights are
culate the elementary aggregate indices and to compile the needed for each class and group to compile the group and
subclass-level index. Thus, the indices for computers and division-level indices. In the example for tablets, weights
laptops are aggregated using the old weights for the elemen- are needed for each of the three microclasses (that is desk-
tary indices to derive the period 2 class index ([101.5 × 0.6] + top computers, laptops, and tablets). Administrative sources
[94.3 × 0.4] = 98.6). may provide sales revenues from value-added tax data for
7.48 In period 3, the elementary- and subclass-level groups 08.1 and 08.2. Importers of computers, laptops, and
indices are compiled using the new sample of products and tablets may also be able to provide revenue data for group
varieties along with the new set of weights for each compo- 08.1. Regulatory authorities can be a source for revenue
nent. The computers in period 3 (99.6) are calculated by using data on internet service fees. Alternatively, the national
the geometric mean price relative for the three new varieties accounts may have expenditure data already compiled for
(0.9811) multiplied by the period 2 price index for computers these groups. The next step is to use the revenue informa-
(101.5). The same calculation is used to derive the, period 3, tion to calculate relative shares for each group and to use
elementary index for laptops (0.9967 × 94.3 = 94.0). the share weights to aggregate group indices to the division
7.49 The new elementary index for tablets has no level. Note that the previous procedure will still not be com-
period 2 index to use so the period 2 subclass-level index pletely satisfactory if it does not incorporate, into division
(98.6) is used as the tablets index, on the assumption that 08, increases or decreases in the overall share of commu-
the elementary index for tablets would have changed by the nications expenditures as a result of the revolutionary new
same percentage, on average, as the other products within product class.
the subclass. Note that this value is also used as the start- 7.54 Table 7.3 contains an example of introducing new
ing index for each of the variety indices within tablets. The weights at the aggregate level when the new items for tablets
period 3 tablets elementary index is calculated as 98.0 by have been introduced. The new share weights for these three
using the elementary-level price relative (0.9933) multi- groups in column C sum to 100. The new weights are intro-
plied by 98.6. duced into the calculation of the index in period 1. The com-
7.50 The aggregate index is calculated as the geomet- ponent price indices for period 1 (columns D and E) below
ric mean of the variety indices. The period 3 subclass-level the total CPI level are the same, but they are aggregated
167
Table 7.2 Example of Introducing a New Elementary Aggregate

Computers Computers
Variety Period 1 Period 2 Weight in Variety Period 2 Period 3 Weight in

Class Class
Price Price Price Price Price Price Price Price Price Price
Index Relative Index Index Relative Index
Brand A 110.0 50.00 50.00 1.0000 110.0 Brand A 101.5 100.00 100.00 1.0000 101.5
(Model 240 w (Model 2000)
memory)
Brand B 105.0 30.00 27.50 0.9167 96.3 Brand F 101.5 90.00 85.00 0.9444 95.9
(Model 960) (Model 1500)
Brand C 104.5 27.50 26.00 0.9455 98.8 Brand A 101.5 75.00 75.00 1.0000 101.5
Elementary 106.5 0.9534 101.5 0.6 Elementary 0.9811 99.6 0.2
Aggregate Aggregate
Index Index
Laptops Laptops
Variety Period 1 Period 2 Weight in Variety Period 2 Period 3 Weight in

Class Class
Price Price Price Price Price Price Price Price Price Price
Index Relative Index Index Relative Index
Brand D 100.0 2000.00 1950.00 0.9750 97.5 Brand D 94.3 2500.00 2500.00 1.0000 94.3
Brand E 95.0 1250.00 1200.00 0.9600 91.2 Brand C 94.3 1500.00 1490.00 0.9933 93.7
Elementary 97.5 0.9675 94.3 0.4 Elementary 0.9967 94.0 0.1
Aggregate Aggregate
Index Index
Tablets
Variety Period 2 Period 3 Weight

in Class
Price Price Price Price Price
Index Relative Index
Brand G 98.6 250.00 245.00 0.9800 96.7

(Model 14K25)
Brand C 98.6 180.00 180.00 1.0000 98.6
(Model 2J564)
Brand H 98.6 200.00 200.00 1.0000 98.6
(Model
MI7J2369)
Elementary 0.9933 98.0 0.7
Aggregate
Index
Subclass Index Subclass Index
Period 1 Period 2 Period 2 Period 3
08.1.3.1 102.9 98.6 08.1.3.1 97.9

Computers, Computers,
Laptops, Laptops, and
and Tablets Tablets
See Chapter 8, section on the calculation of higher-level indices for more details (beginning paragraph 8.89).
using different weights. As a result, the index for division using the new weights and the price relatives between
08 (Information and Communication) differs between col- period 1 and period 2 (column F) are calculated. The price
umn D (197.9) and column E (192.0). The all-items CPI also relative for division 08 (0.999229) is applied to the period
differs because of the different weights—column D (386.6) 1 published index for 08 Information and Communication
using the old weights and column E (393.4) using the new (197.9) to derive the period 2 index (197.8). Likewise, the
weights. As the difference between these values is solely price relative for the total CPI (0.997954) in column F is
due to weighting effects, the reweighted index value should applied to the period 1 published index for the all-items CPI
not be published—it should only be used to calculate the (386.6) to derive the period 2 all-items CPI (385.8). These
current period price change. Footnotes should be provided same calculations, using the price change in the reweighted
to explain the introduction of the new weights. 08 Information and Communication index and total CPI,
7.55 To derive the period 2 indices for division 08 and are repeated for all future periods (see columns H and I in
the all-items CPI in column G, the indices are compiled Table 7.3).
168
Table 7.3 Example of Introducing New Weights for Higher-Level Aggregates

COICOP Code Old Weights New Weights Period 1 Period 1 to Period 2 Period 2 to Period 3
Period 2 Period 3
Index Price Relative Index Price Relative Index
Old Weights New Weights New Weights New Weights
(a) (b) (c) (d) (e) (f) (g) (h) (i)
Total CPI 100.00 386.6 385.8 386.8

Total CPI Using New Weights 100.00 393.4 0.9980 392.6 1.0026 393.6
01 Food and Nonalcoholic Beverages 67.29 58.23 345.2 345.2 342.5 343.7
02 Alcoholic Beverages and Tobacco 0.54 0.77 453.2 453.2 455.1 454.1
03 Clothing and Footwear 1.37 1.70 376.9 376.9 375.0 376.0
04 Housing, Water, Electricity, Gas, and 7.91 8.45 572.4 572.4 576.8 577.2
Other Fuels
05 Furnishings, Household Equipment, 2.54 2.71 401.5 401.5 402.0 402.8
and Routine Household Maintenance
06 Health 1.47 2.12 215.1 215.1 217.3 217.3
07 Transport 6.06 8.25 685.9 685.9 687.7 686.5
08 Communication 2.14 197.9 197.7 197.5
Communication Using New Weights 4.15 192.0 0.9992 191.9 0.9986 191.6
08.1 Information and Communication 0.47 0.24 201.1 201.1 201.1 201.1
Equipment
08.2 Software 6.54 37.35 179.8 179.8 177.9 177.5
08.3 Information and Communication 92.99 62.41 199.2 199.2 200.1 199.9
Services
09 Recreation and Culture 1.79 2.55 348.2 348.2 349.6 350.3
10 Education 1.43 1.98 433.7 433.7 433.7 435.9
11 Restaurants and Hotels 3.41 5.67 411.2 411.2 413.2 415.5
12 Miscellaneous Services 4.05 3.42 391.3 391.3 393.1 397.8
See Chapter 8, section on the calculation of higher-level indices for more details (beginning paragraph 8.89).
here. The measured CPI would not include the benefit of

The Overlap Method and the the fall in price from the consumer switching to the cheaper
Incorporation of New Products and taxi service.
Outlets into a Consumer Price Index 7.58 Now assume that the new taxi service was incorpo-
rated into the CPI not as a replacement, but within the same
Evolutionary Products—Similar Service Flows elementary aggregate as an additional representative variety—
7.56 The methods previously discussed in this chapter a sample augmentation. It was recognized that this new service
introduce a new product into the CPI as soon as two succes- was different from the existing one, but it had its own market
sive period prices are available. For example, consider the niche substantial enough to merit inclusion as a new representa-
digital economy whereby many purchases of existing goods tive variety. Again, the CPI compiler would wait until prices for
and services can be made in a manner that may enhance the new taxi service were available for two successive periods
search, choice, and convenience of purchase as well as pro- and, using the overlap method, link the new price index into the
viding the same good or service at a cheaper price. While existing classification for taxi services. And again, the cheaper
much of this can be disputed for many products, there has taxi service would not be registered in the CPI measurement as
been a substantial shift in expenditure toward such digital a price fall for those who switched to it.
services. 7.59 Alternatively, it might be considered that the new
7.57 Consider a simple stylized example of a taxi ser- taxi service is so different from the existing one that it falls
vice accessed by an application on a mobile phone in which within a brand-new elementary aggregate, but still within
the consumer enters the destination, is matched with driver COICOP 2018 code 07.3.2.2—passenger transport by taxi
and vehicle, confirms the pick-up and destination, and is and hired car with driver. The new taxi service would have
automatically charged. Assume, for this stylized example, its own weight and be incorporated into the index by adding
the taxi service is cheaper, and the overall quality in all a new elementary aggregate and redistributing the weight.
other respects is equal to that of a traditional taxi, at least Again, prices for two successive periods are required and
on aggregate. The use of the overlap method to replace the this procedure precludes the CPI from taking into account
existing taxi service with this new replacement would not any effective fall in price arising from the availability of a
take account of the effective fall in price experienced by substitute taxi service that is effectively cheaper.
consumers. The overlap method treats the difference in price 7.60 An appropriate treatment in this stylized example
between the new and old taxi service in the overlap period would be to treat the new taxi provider as a comparable
as an indication of the difference in quality, as described replacement: the assumption is that the same quality of ser-
in Chapter 6: the new service is treated as being cheaper vice is maintained. If this assumption is valid, the price fall
because it has a poorer quality, though this is not the case is captured by the index.
169
7.61 The inclusion of new products using the overlap in period 1, is estimated for period 2. Furthermore, delays
method is, as outlined previously, a normal part of CPI com- in the introduction of the new product into the CPI might
pilation being undertaken on rebasing, when new elemen- generate more bias. Typically, the price of a revolutionary
tary aggregates are introduced and outlets and representative product declines rapidly after introduction, so there would
varieties for elementary aggregates within them, are reini- be an upward bias in delaying its introduction, again to the
tialized. Price changes of the “old” sample are measured up detriment of the credibility of the CPI regarding what may
to and including the overlap period, for example, December be high-profile products. Exiting products have the opposite
2020 for an annually chained CPI, and then from December welfare effect of new products and excluding the welfare
2020 to include successive months in 2021, for the new sam- loss of the exiting products would result in a downward
ple. For example, if online purchases were introduced in the biased CPI.
CPI in the rebasing from December 2020, the price relatives
for December 2020 to January 2021 would include the price Evolutionary and Revolutionary Products: A
changes from online outlets, while that from November to Continuum
December 2020 would exclude them. If prices in online out- 7.66 Evolutionary new products can be linked, with
lets were cheaper than those in brick-and-mortar stores, the a quality adjustment, to the preexisting technology. This
price fall experienced by consumers switching their expen- would be the case, for example, for household appliances,
diture from one to the other would not be registered. where the spin speed, running cost, reliability, and other
7.62 A further example would be from an “outcomes” price-determining features are improved. This might be as
approach to medical services. Consider that an outcome is to a replacement for an existing representative model with an
fix a medical problem and a new medical procedure makes explicit quality adjustment or integrated into the dynamic
this cheaper. The overlap method used the price change of sample of models of washing machines sold each month
the old procedure up to the introduction of the new, and then using a database, scanner or web-based, that includes the
links-in the price change of the new. The prices of the old quality characteristics of the models. Chapters 6 and 10
procedure may be constant in the months prior to the new outline the hedonic methodology and data requirements for
method and the price changes of the new procedure may integrating such new evolutionary products into a CPI while
also be constant when introduced, although at a lower level. avoiding the unrealistic assumptions implicit in the use of
7.63 The underlying measurement flaw is that while the overlap linking-in of a new model that makes its use likely
fixed basket is regularly updated to maintain the represen- to misrepresent quality-adjusted price changes, as also out-
tativity of the CPI, it fails to reflect price change benefits lined in Chapter 6.
resulting from switching to these new products. 7.67 However, evolutionary new products may combine
both a continuing service and a “newness.” For example,
Revolutionary New Products and mobile phones could be considered an evolutionary good in
Welfare (Utility) Gains the sense that they continued the provision of a communica-
7.64 Revolutionary new products might be introduced tion service previously provided by landlines, but had a sub-
on rebasing or, if attracting a high share of expenditure, stantive revolutionary characteristic on introduction, their
through sample augmentation. In both instances, the overlap wireless mobility, and this defined their revolutionary new-
method while including the new product would not effec- ness. This service flow was further developed as the mobile
tively capture the benefits experienced by the consumer phone became a platform for the extensive range of applica-
from its purchase. tions it is commonly used for. The introduction of mobile
7.65 The successful introduction of a revolutionary phones into the CPI when updating or through sample aug-
new product leads to a welfare (utility) gain to consumers mentation using the overlap method could not be justified as
reflected in, and evidenced by, a switch in consumer expen- a simple continuation of an existing service. Similarly, color
diture to the new product. Consumers are better off as a result television was a continuation of a black-and-white service
of their purchase of the new product. There is no preexisting flow, but there is no simple metric of “more of the service
service flow for a quality adjustment to be made. The intro- flow” that could be used to encapsulate the switchover to
duction of the new product through either sample augmenta- such a new service feature.
tion or a new elementary aggregate when updating using the
overlap method misses the welfare gain at the moment of Distinguishing between New Elementary
entry of the new good. It is only when two successive prices Aggregates and New Products
are available that the revolutionary new product is included, 7.68 New elementary aggregates are introduced on
and then it is too late to capture the initial welfare gain. Con- rebasing using the overlap method. The distinction between
sider the case of a new good to be introduced into a CPI, in introducing a new product when updating the index and the
period 3. A conceptually sound approach to its incorporation introducing a new elementary aggregate should be made.
into the index is to impute its price for period 2, that is, to For example, “charges for undertaking and other funeral
estimate its (Hicks) reservation (or choke) price. This is the services,” under COICOP class 13.9.0, were newly intro-
price that would drive the demand for the good down to zero duced into a CPI on updating. Its inclusion in the CPI would
in the period prior to its introduction. The fall in the reserva- neither be as a new revolutionary product providing a previ-
tion price in period 2 to the actual price in period 3 might ously unavailable service to consumers nor as an evolution-
be substantial, though neglected in the CPI measurement. ary product providing a different, usually improved, existing
The measured CPI would be biased upward, in this welfare service flow, for example, at a lower price. It is simply an
sense. An analogous approach applies to disappearing prod- updating of the basket of products consumers typically pur-
ucts, where the reservation price for a product last appearing chased resulting from an increased expenditure share over
170
and above a threshold cutoff, meriting inclusion. There is 7.72 Within the range of evolutionary and revolution-
neither a switch in consumers’ expenditure away from an ary products, CPI compilers should identify evolutionary
existing product to a new variant nor a welfare gain from new products whose main features are the extension of a
consumers purchasing a revolutionary new product. preexisting service flow, although they may result from a
new technology, and integrate the measures into the exist-
How to Respond to Limitations in the Use of the ing elementary aggregates using the techniques of explicit
Overlap Method Regarding New Goods quality adjustment outlined in Chapter 6. There should be a
7.69 When updating the CPI, the overlap method can recognition of the different aspects of the evolutionary prod-
be useful because of its virtues in updating the fixed basket, ucts that would not be taken into account and these would be
along with new weights, outlets, products, and representa- documented and explained to the user, much in the manner
tive varieties. However, publicizing the incorporation of a that the shortcomings in the introduction of revolutionary
new product, for example, a technological innovation into products.
the CPI using the overlap method may mislead users who 7.73 There may be specific instances where a relatively
may be expecting a fall in prices because of a switch to the homogeneous service flow is still provided by the evolution-
new product but find that when introduced into the index, ary product. For example, specific varieties or services may
price changes are not reflected as such.1 The overlap method now be sold through online or mobile applications. These
would not provide a measure that could be explained in this may be cheaper and have better search options than would be
manner. Simple statements to the effect that new goods are possible using brick-and-mortar stores. The introduction of
being incorporated into the CPI by use of the overlap method such new outlets using an overlap method would not enable
through either forced substitution, sample augmentation, the effective price fall to be included in the CPI. There is a
sample rotation, or rebasing may damage the credibility of different, positive service aspect of buying some items at
the CPI. An important response is transparency. To explain brick-and-mortar locations that are not available online. It
in the CPI dissemination publication that although the new is possible that the service through a brick-and-mortar store
products are being introduced so that its future successive is considered superior for some outlets/products. Again, the
price changes are reflected in the CPI, the price changes quality difference would not be reflected by simply linking-
from consumers switching from one to the other will not. in an overlap price change.
7.70 Given the practical problems in introducing revo- 7.74 If the purchases were considered homogeneous,
lutionary new products into the CPI, NSOs may opt to define a unit value index would appropriately reflect these differ-
a CPI to exclude the implicit price falls and welfare gains ences. When there is price variation for the same quality of
arising from the introduction of revolutionary new products. good or service, the price relatives used for index number
“Welfare gains” and “reservation prices” are not consider- calculation should be defined as the ratio of the weighted
ations employed in the usual interpretation of CPIs. While average price of that good or service in the two periods, the
there are studies on the estimation of such reservation prices, weights being the relative quantities sold at each price, a
their practical use for CPI compilation is problematic.2 unit value index. For example, suppose that a certain quan-
7.71 However, this is not to justify the statistical prac- tity of a particular good or service is sold at a lower price to
tice of an inability to include welfare gains on practical a particular category of purchaser without any difference in
grounds. A reality of revolutionary new products is that the the nature of the good or service offered, location, timing or
standard of living increases and this increase, when matched conditions of sale, or other factors. A subsequent increase
against changes in the nominal income, implies a price fall. in the proportion sold at the lower price lowers the aver-
When the CPI for the product group concerned does not age price paid by purchasers for quantities of a good or ser-
reflect this fall, a recognition and explanation given in the vice whose quality is the same and remains unchanged, by
CPIs metadata and “frequently asked questions” can rein- assumption.
force the credibility of the CPI. Revolutionary new prod- 7.75 If the new good or service has a quality dimension
ucts, and evolutionary new products whose main innovation different from the existing one, a judgment must be made as
is a revolutionary service, should be introduced through to the advantages of properly including the price fall in the
sample augmentation and rebasing in a timely manner with index resulting from the cheaper online purchase, but not
adequate explanation as to the limitation of the measure. taking account of the other quality dimensions, as against
simply introducing the outlet using the overlap method and
ignoring effective decreases in average prices paid.
7.76 The simple use of the overlap method creates a dis-
1
It may be the case that a price increase is expected with the introduction
tinct bias in the CPI as it overstates price inflation in ignor-
of the new product, but consumers are aware that the CPI measurement ing the switch to cheaper prices. The benefits of the digital
may mitigate this because of the effect of the quality innovation on this economy, in this respect, would be absent from the official
price increase.
2
statistics. CPI compilers would point to the integration of
Erwin Diewert, and Robert Feenstra. 2017. “Estimating the Benefits such outlets into the CPI sample, but this would only con-
and Costs of New and Disappearing Products.” Discussion Paper Series
17-10, University of British Columbia, Vancouver School of Economics; ceal the deficiency of the methodology; indeed, it might be
Erwin Diewert, Kevin J. Fox, and Paul Schreyer. 2017. “The Digital argued that based on the CPI data the new online outlets
Economy, New Products and Consumer Welfare.” Discussion Paper 17- have no impact on consumer prices.
09, University of British Columbia, Vancouver School of Economics; 7.77 The unit value’s assumption of homogeneity of
and Marshall Reinsdorf, and Paul Schreyer. 2017. “Measuring Consumer
Inflation in a Digital Economy.” Paper presented at the Fifth IMF Statistical
consumer services in the two types of outlets may be argued
Forum, Washington, DC, November 16–17. https://www.imf.org/en/News/ to open up a higher level of bias, the nature of which will
Seminars/Conferences/2017/05/03/5th-statistical-forum. vary depending on the product considered. Analytical
171
frameworks for unit value indices with quality adjustments including implicit and explicit approaches and, further, the
have been developed with this problem in mind.3 detailed use and nature of these methods for particular prod-
7.78 The recommendation is to provide as part of the uct groups may change over time. Metadata may need to be
published metadata detailed information on methods used to updated regularly. Such a monitoring system should serve
maintain the sample. This would include statistics on the use to alleviate the monthly degradation of the price reference
of temporary and permanently missing varieties and their period sample for each product group. The term “metadata”
replacements, and the methods employed for their replace- is used to include this aspect of quality assurance.
ment, as outlined in Chapter 6. Specific attention should be 7.82 A strategy for the maintenance of the sample must
given to product areas where there is a high churn or expan- be linked to sample representativity, and it requires building
sion in model turnover and introduction of new products, as a statistical metadata system. The CPI compilation requires
outlined in this chapter. the continual development of market information and the
7.79 The metadata should be further extended to recording and evaluation of the sample development on a
include an audit of new products that are believed to have product-by-product basis. Such a metadata system relates to
a sufficient impact on the standard of living of households. the variety of procedures for quality adjustments to prices
The broad principles of how such new products are treated discussed in Chapter 6 and how their suitability might vary
and any practical limitations in such treatment should be on a case-by-case basis, all of which require documentation.
highlighted. The inclusion of the products should for the 7.83 Metadata facilitate identifying those product
large part be focused on traditional criteria such as those groups where the sample is deteriorating and how it might
product areas with relatively high (up-to-date) expenditure be replenished. The metadata help counter the degradation
weights and should exclude free products.4 The exclusion of the static sample selected at the last rebasing by monitor-
should not just be confined to expenditure weights but draw ing the use of appropriate replacements when variety prices
attention to relatively low-expenditure new products that are go missing. It also can be used to ensure the static sample
considered to be responsible for substantial increases in the remains representative of the universe of transactions and is
standards of living. The document should be a living docu- not biased against the introduction of new varieties and new
ment updated regularly. products.
7.80 Impact studies and methodological developments 7.84 The metadata should monitor and document the
are likely to have a cross-country relevance. Therefore, extent of temporarily and permanently missing prices and
methodologies, and results, from single-country studies may methods used for quality-adjusted prices. Price indices
be applied more generally, or at least cited so that users have for specific products, such as personal computers, may be
some indication of the impact of specific new products. The derived using specific compilation and estimation routines
Expert Group on Consumer Price Indices5 and the Ottawa and metadata are required to document such procedures.
Group on Price Indices6 may provide useful references for Because the rationale for the employment of different meth-
developments in such work. ods of treating missing variety prices and quality changes is
usually specific to the features of the product groups con-
Information Requirements cerned, information is required on such features. For exam-
ple, data can be maintained on market features, such as the
Maintaining the Sample dates for the introduction of new products and the nature
7.81 Metadata are systematic, descriptive information of their technological change. Information can also be col-
about data content and organization that help those who lected and kept on out-of-sample developments (that is,
operate the statistics production systems to remember what market developments outside the static sample), such as the
tasks they should perform and how they should perform obsolescence of existing products and the emergence of new
them. Such data also serve to encourage transparency in the products onto the market. New technologies, products, and
methods used and help ensure that they are understood and a proliferation of new varieties, accompanied by emerging
continued as staff members leave and others join. In the con- brands, may grow to become responsible for a major part of
text of Chapter 6, a variety of methods might be employed the market. To increase transparency in the procedures used
for the treatment of missing prices and their replacement and allow the effort to be directed where it is most needed:
• NSOs should monitor the incidence of temporarily and
3
Such work includes Mick Silver. 2011. “An Index Number Formula Prob- permanently missing variety prices at perhaps a two-digit
lem: The Aggregation of Broadly Comparable Items.” Journal of Official COICOP level, as appropriate, and if the incidence is
Statistics 27 (4): 553–67. high for a particular product group, at the three- or four-
4
As discussed in Reinsdorf and Schreyer (2017).
5
The Expert Group on CPI is established by the Conference of European digit level. The advantage of a top-down approach is that
Statisticians. The meetings of the Group take place every second year, resources are saved by monitoring at the detailed level
jointly organized by the United Nations Economic Commission for only those product groups that are problematic.
Europe and the International Labour Organization. The proceedings of the
Expert Group meetings are available from the United Nations Economic
• The ratios of temporarily missing prices, comparable
Commission for Europe website, www.unece.org/statistics. replacements, and noncomparable replacements to the
6
The Ottawa Group is a city group set up in 1994 under the auspices of the overall number of variety prices, and the methods for
United Nations Statistical Commission. It is dedicated to methodological dealing with each of these three circumstances, should
work and applied research in CPI and other price statistics, and provides also be monitored. Against each product group, the
a forum for specialists and practitioners who work for or advise NSOs
or international organizations to exchange experiences and thoughts. weight of the product concerned should be listed so that
Information about the Ottawa Group and proceedings from its meetings are a disproportionate effort is not given to relatively low-
available from the Group’s website, www.ottawagroup.org. weighted products.
172
• The metadata system should be directed to the periodic • A strategy for the use of alternative data sources for
monitoring, preferably on a monthly/quarterly basis, of prices and weights including scanner data, web-based
the methods used for treating temporarily and perma- prices, monthly billing statements, and methodologies
nently missing prices as recommended in Chapter 6. and software for their use. Such alternative sources are
• Such metadata should extend to the CPI compilers devel- likely to directly benefit the maintenance of the repre-
oping market expertise on selected high-profile, heav- sentativity of a sample especially as there is a move-
ily weighted product groups where the consumer goods ment away from the static sample of matched-model
and services provided are sufficiently complex to impose method.
on the measurement system the need for special con- 7.86 While these metadata reflect best practice, devel-
sideration: product groups such as telecommunications, oping such detailed and comprehensive metadata requires
computers and computer-related hardware and software, significant resources. NSOs should begin developing
electronics, property and rentals, internet purchases, and detailed metadata on the incidence of temporarily and per-
so forth, as outlined in Chapter 11. manently missing prices. Understanding which items and
7.85 The metadata might also include: varieties are missing, the duration the prices are missing,
and why these prices are missing will serve as the first step
• Product-specific information, such as the timing of the for maintaining the sample. As resources permit, a plan to
introduction of new models, pricing policies, especially develop additional metadata over time can be developed and
in months when no changes were made, and popularity of implemented.
models and brands according to different data sources.
• Information arising from contacts with market research
organizations, retailers, manufacturers, and trade associa-
Key Recommendations
tions for products for which replacement levels are high.
• The samples of outlets, items, and varieties should be
The development of such contacts may lead, for example,
reviewed and updated as needed between regular index
to option cost estimates, which can be easily introduced.
updates. This ensures that samples remain representative.
Where possible, staff should be encouraged to learn more
about specific product groups whose weights are rela- • A rotation policy could be implemented for the outlet
tively high and where product replacement is common. sample, depending upon the availability of replacement
Contacts with organizations concerning such product outlets. For example, each year, 20 or 25 percent of the
groups will allow staff to better judge the validity of the outlet sample is replaced by new outlets. This reduces the
assumptions underlying implicit quality adjustments. respondent burden and allows for the selection of new
items and varieties.
• Product groups likely to be undergoing regular techno-
logical change should be identified. The system should • The variety sample should be reviewed on a routine basis
attempt to ascertain the pace at which models change and, as well to ensure that the selected varieties remain the
where possible, the timing. most popular with regard to sales volume.
• Price-determining characteristics for those products • New products should be introduced into the index as quickly
undergoing technological change, especially if quality- as possible to ensure the index remains representative.
adjustment procedures used make use of hedonic regres- − For evolutionary products, the new product replaces
sions. Information may be included from market research the obsolete product and maintains the same weights
organizations, responding businesses, wholesalers, trade until the next index update. For example, streaming
associations, and other such bodies. This should contribute video services replaces the rental of DVD movies.
to the statistical metadata system and be particularly useful − For revolutionary products, covered by an existing ele-
in providing subsequent guidelines on product selection. mentary aggregate, the new item can easily be included.
• An analysis is needed of what have in the past been judged If not covered by an existing elementary aggregate and
to be “comparable” replacements with regard to the fac- data on sales to households are available, a new ele-
tors that distinguish the replacement and old product. The mentary aggregate can be developed for the new item.
analysis should identify whether different price collectors If no sales data are available, the new item can tem-
are making similar judgments. porarily be included in a related elementary aggregate
• When hedonic regressions are used either for partial until the next weight update.
patching of missing prices or as indices, information on • Detailed metadata should describe methods used to main-
the specification, estimated parameters, and diagnostic tain the samples of outlets and varieties. These metadata
tests of the regression equations should be kept along should include a description of how new products (both
with notes as to why the final formulation was chosen and evolutionary and revolutionary products) are included in
used along with the data. the index.
173
CALCULATING CONSUMER PRICE
INDICES IN PRACTICE 8
Introduction 8.4 The chapter also discusses the calculation of higher-
level indices. The focus is on the ongoing production of a
8.1 The purpose of this chapter is to provide a general monthly price index in which the elementary price indices
description of the ways in which consumer price indices are averaged, or aggregated, to obtain higher-level indices.
(CPIs) are calculated in practice. The methods used in differ- Price updating of weights, chain linking, and reweighting
ent countries are not exactly the same, but they have much are discussed in Chapter 9. Data editing procedures are dis-
in common. There is clearly interest from both compilers cussed in Chapter 5 on price collection. Statistical tools and
and users of CPIs in knowing how most national statistical methods for index analysis such as contributions to price
offices (NSOs) actually calculate their CPIs. change appear in Chapters 9 and 14.
8.2 As a result of the greater insights into the proper- 8.5 While the purpose of this chapter is the compila-
ties and behavior of price indices that have been achieved tion of CPIs at the various levels of aggregation, NSOs must
in recent years, it is now recognized that some traditional keep in mind that the end goal of producing the indices is to
methods may not necessarily be optimal from a conceptual disseminate and publish CPIs of high quality. To this end,
and theoretical viewpoint. Concerns have also been voiced the sampling process for selecting the items that are included
in a number of countries about possible biases that may be in the indices and the price observations that are represen-
affecting CPIs. These biases and concerns are considered in tative of the product varieties in the consumer markets are
Chapter 13. The methods used to compile CPIs are inevita- critically important in determining the quality of the indices
bly constrained by the resources available, not merely for at the elementary and aggregate levels. In this regard, the
collecting and processing prices, but also for gathering the sampling procedures presented in Chapter 4 are very impor-
expenditure data needed for weighting purposes. In some tant to attain this end goal.
countries, the methods used may be severely constrained by
lack of resources. Nonetheless, there are still methods that
should be avoided at all costs because they result in severe The Calculation of Price Indices
bias in the indices. for Elementary Aggregates
8.3 The calculation of CPIs usually proceeds in two
stages. First, price indices are estimated for the elementary 8.6 CPIs are typically calculated in two steps. In the first
aggregates. These are referred to as the elementary price step, the elementary price indices for each of the elementary
indices. The elementary aggregate is the lowest level of aggregates are calculated. In the second step, higher-level
groups of goods or services for which expenditure weights indices are calculated by taking weighted averages of the
are assigned and kept constant for a period of one year or elementary price indices. The elementary aggregates and
more. An elementary aggregate should consist of a relatively their price indices are the basic building blocks of the CPI.
homogeneous set of goods or services, with similar end
uses and similar expected price movements. More detailed Construction of Elementary Aggregates
weights to reflect the relative importance of individual price 8.7 Elementary aggregates are groups of relatively
observations within elementary aggregates may be applied homogeneous goods and services (that is, similar in char-
and updated more frequently. In the second stage, these ele- acteristics, content, price, or price change). They may cover
mentary price indices are aggregated to obtain higher-level the whole country or separate regions within the country.
price indices using the expenditure shares of the elementary Likewise, elementary aggregates may be distinguished for
aggregates as weights. This chapter starts by explaining how different types of outlets. The nature of the elementary
the elementary aggregates are constructed, and what eco- aggregates depends on circumstances and the availability of
nomic and statistical criteria need to be taken into consider- information, such as detailed expenditure data. Elementary
ation in defining the aggregates. The index number formulas aggregates may therefore be defined differently in different
most commonly used to calculate the elementary indices are countries. Some key points, however, should be noted:
then presented, and their properties and behavior illustrated
using numerical examples. The advantages and disadvan- • Elementary aggregates should consist of groups of goods
tages of the various formulas are considered, together with or services that are as similar as possible, and preferably
some alternative formulas that might be used instead. The fairly homogeneous in construction and content.
problems created by disappearing and new varieties (that is, • Elementary aggregates should consist of varieties that
one variety with another of similar or different quality) are may be expected to have similar price movements. The
also explained, as well as the different ways of imputing val- objective should be to try to minimize the dispersion of
ues for missing prices. price movements within the aggregate.
175
• Elementary aggregates should be appropriate to serve as

Figure 8.1 Illustrative Aggregation Structure of a CPI3
strata for sampling purposes in the light of the sampling
regime planned for the data collection.
8.8 Each elementary aggregate, whether relating to the
whole country or to an individual region or group of out-
lets, will typically contain a very large number of individual
goods or services, or varieties. In practice, only a small num-
ber can be selected for pricing. When selecting the varieties,
the following considerations need to be made:
• The varieties selected should be ones for which price
movements are believed to be representative of most of
the products within the elementary aggregate.
• The number of varieties within each elementary aggregate
for which prices are collected should be large enough for
the estimated price index to be statistically reliable. The
minimum number required will vary between elementary
aggregates depending on the nature of the products and
their price behavior. However, there should be at least
eight to ten observations for calculating the elementary
index as discussed in Chapter 4.
• The objective is to try to track the price of the same vari-
ety over time for as long as the variety continues to be
representative. The varieties selected should therefore be
ones that are expected to remain on the market for some
time, so that like can be compared with like, and problems
associated with replacement of varieties be reduced.
The Aggregation Structure

8.9 The aggregation structure for a CPI is illustrated in
Figure 8.1. Using a classification of consumers’ expendi-
ture such as the Classification of Individual Consumption
According to Purpose (COICOP), the entire set of consump- selected in practice. Finally, for each representative product,
tion goods and services covered by the overall CPI can be several specific varieties should be selected for price collec-
divided into divisions, such as “Food and Nonalcoholic Bev- tion, such as particular brands of wheat bread. Again, the
erages.” Each division is further divided into groups, such as number of sampled varieties selected may vary depending
“Food.” Groups are divided into classes, such as “Cereals on the nature of the representative product.3
and Cereal Products.” Classes are further divided into sub- 8.10 Methods used to calculate the elementary indices
classes, such as “Cereals.” Many countries use an even finer from the individual price observations are discussed in para-
classification by further disaggregating below the level of graphs 8.15–8.88. Working upward from the elementary
the subclasses. For CPI purposes, each subclass can then be price indices, all indices above the elementary aggregate
further divided into more homogeneous microclasses,1 such level are higher-level indices that can be calculated from
as “Basmati Rice.” The microclass could be the equivalent the elementary price indices using the elementary aggregate
of the basic headings used in the International Compari- expenditure as weights. The aggregation structure should be
son Program,2 which calculates purchasing power parities consistent, so that the weight at each level above the elemen-
between countries. Finally, the microclass may be further tary aggregate is always equal to the sum of its components.
subdivided by dividing according to region or type of outlet, The price index at each higher level of aggregation can be
as in Figure 8.1. In some cases, a particular microclass can- calculated based on the weights and price indices for its com-
not be, or does not need to be, further subdivided, in which ponents, that is, the lower-level or elementary indices. This
case the microclass becomes the elementary aggregate. applies for indices with fixed weights. If the weight structure
Within each elementary aggregate, one or more products is updated and the index series based on the new weights is
are selected to represent all the products in the elementary chain linked, the linked index for the previous year is not
aggregate. For example, the elementary aggregate consist- consistent in aggregation. The individual elementary price
ing of “Bread” sold in supermarkets in the Northern region indices not only should be designed to be sufficiently reli-
covers all types of bread, from which “Wheat Bread” and able to be published separately, but they should also remain
“Whole Grain Bread” are selected as representative prod- the basic building blocks of all higher-level indices.
ucts. Of course, more representative products might be
3
The Classification of Individual Consumption According to Purpose
1
In this Manual, levels below the subclass are referred to as microclasses. 2018 structure breakdown includes divisions (food and beverages), group
2
Additional information available at https://www.worldbank.org/en/programs/ (food), class (cereals and cereal products), and subclass (bread and bakery
icp. products).
176
Calculating Consumer Price Indices in Practice
Weights within Elementary Aggregates the continuity of the price series for the individual varieties
8.11 The ideal index number formula to use for CPI cal- for one reason or another. The treatment of disappearing and
culations would have weights for each price observation used replacement varieties is taken up in paragraphs 8.51–8.74. It
to compile the elementary price indices, as well as weights for is also assumed that there are no explicit weights available.
aggregating elementary indices to higher-level price indices. 8.17 The properties of the three indices used to compile
In some countries, this approach has been achieved through elementary aggregates (Jevons, Dutot, and Carli) are exam-
comprehensive sampling procedures or the use of scanner ined and explained in some detail in Chapter 6 of Consumer
data for select item groups (for example, food). Those coun- Price Index Theory where it is shown that the Jevons is pre-
tries that have weights at this level use fixed-basket indices ferred in most circumstances when weights are not available.
that are discussed in paragraphs 8.89–8.136. Also, having Here, the purpose is to illustrate how they perform in practice,
weights for both the weight reference period and the current to compare the results obtained by using the different formu-
period would be ideal to produce one of the preferred tar- las, and to summarize their strengths and weaknesses. These
get indices for CPI compilation (Fisher, Törnqvist, or Walsh widely used formulas that have been, or still are, in use by
price indices). Several countries that have access to scanner NSOs to calculate elementary price indices are illustrated in
data use the prices and quantities for the individual observa- Tables 8.1–8.3 by using average prices, averages of price rela-
tions to derive elementary aggregate indices of their CPI. tives, and long-term versus short-term price relative methods.
8.12 In most cases, the price indices for elementary aggre- It should be noted, however, that these are not the only pos-
gates are calculated without the use of explicit weights. The sibilities and some alternative formulas are considered later.
elementary aggregate is simply the lowest level at which reli- The first is the Jevons index for i = 1 … n varieties. It is
able expenditure weighting information is available. In this defined as the unweighted geometric mean of the price rela-
case, the elementary index must be calculated as an unweighted tives, which is identical to the ratio of the unweighted geomet-
average of the prices of which it consists. However, even in ric mean prices, for the two periods, 0 and t, to be compared:
this case, it should be noted that when the varieties are selected 1 1
with probabilities proportional to the size of some relevant
variables such as sales (as described in Chapter 4), weights are I 0:t
 p t n
= ∏ i0  =
∏( p )
t n
i
(8.1)
implicitly introduced by the sampling procedure.
J
 p  1
8.13 For certain elementary aggregates informa-

i
∏( p )
0 n
i
tion about sales of particular varieties, market shares, and The Jevons price index in 8.1 is calculated by comparing
regional weights may be used as explicit weights within an directly the prices of the two periods, 0 and t. Indices that are
elementary aggregate. When possible, weights that reflect calculated by comparing the price of the reference period and
the relative importance of the sampled varieties should be the current period directly are referred to as direct indices.
used, even if the weights are only approximate. 8.18 Assuming the time from 0 to t exists for a number
8.14 For example, assume that the number of suppli- of periods, 0, 1, 2, … t − 1, t, it is possible to calculate the
ers of a certain product such as fuel for cars is limited. The index by first calculating the price indices from period to
market shares of the suppliers may be known from business period, and then multiplying, or chaining, these together to
survey statistics and can be used as weights in the calcula- obtain the price index from 0 to t:
tion of an elementary aggregate price index for car fuel.
Alternatively, prices for water may be collected from a 1 1 1 1
number of local water supply services where the population  p t n Π ( pi1 )n Π ( pi2 )n Π ( pit )n
in each local region is known. The expenditure weights for
0:t
I JC = Π  io  = ....
 pi  1 1 1
each region may then be used to weight the price in each Π ( pi0 )n Π ( pi1 )n Π ( pit −1 )n
region in order to obtain the elementary aggregate price 1

1
index for water. The calculation of weighted elementary Π( p )
t n
p t  n
= Π  (8.2)
i
indices is discussed in more detail in paragraphs 8.75–8.88. = i

1
 p 0

Π( p 0 n
i ) i
Calculation of Elementary Price Indices

8.15 Various methods and formulas may be used to calcu- A price index calculated by multiplying the period-to-period,
late elementary price indices. This section provides a summary or short-term, price indices, is referred to as a chained or
of the methods that have been most commonly used and the chain-linked price index. When calculating the Jevons
advantages and disadvantages that NSOs must evaluate when index in 8.2 the numerators and denominators of periods
choosing a formula at the elementary level. Chapter 6 of Con- 1, 2, … , t − 1 cancel out leaving only the prices of period 0
sumer Price Index Theory provides a more detailed discussion. and t, so that the resulting chained index is identical to the
8.16 The methods most commonly used are illustrated in direct version of the index in 8.1.
a numerical example in Tables 8.1–8.3. In these examples, an 8.19 The second elementary index formula is the Dutot
elementary aggregate consists of seven varieties of an item index, defined as the ratio of unweighted arithmetic mean
that could be collected from several outlets, and it is assumed prices:
that prices are collected for all seven varieties in all months,
so that there is a complete set of prices. There are no dis- 1
appearing varieties, no missing prices, and no replacement n
∑ pit
I 0:t
= (8.3)
varieties. This is quite a strong assumption since many of the D
1
problems encountered in practice are attributable to breaks in n
∑ pi
0
177
Table 8.14 Jevons and Dutot Price Indices Using Averages of Prices
Base Jan. Feb. Mar. Apr. May Jun. Jul.
Item A Prices
Variety 1 2.36 2.09 1.93 2.59 2.05 2.85 2.59 2.36

Variety 2 5.02 5.38 5.12 5.52 4.08 4.08 5.52 5.02
Variety 3 5.34 5.07 5.09 5.88 6.29 5.86 5.88 5.34
Variety 4 6.00 5.73 4.27 6.00 4.75 5.27 6.60 6.00
Variety 5 6.12 6.39 5.50 6.12 5.86 6.29 6.74 6.12
Variety 6 2.80 2.72 2.82 3.08 2.85 2.05 3.08 2.80
Variety 7 6.21 5.45 6.95 6.21 5.27 4.75 6.84 6.21
Geometric Mean Price 4.55 4.38 4.20 4.81 4.17 4.17 5.01 4.55
L-T Price Relative 1.000 0.963 0.923 1.056 0.917 0.917 1.100 1.000
S-T Price Relative 0.963 0.959 1.143 0.868 1.000 1.200 0.909
Arithmetic Mean Price 4.84 4.69 4.53 5.06 4.45 4.45 5.32 4.84
L-T Price Relative 1.000 0.970 0.936 1.046 0.920 0.920 1.100 1.000
S-T Price Relative 0.970 0.965 1.117 0.880 1.000 1.196 0.909
Jevons Index (L-T ratio of geometric mean prices) 100.0 96.3 92.4 105.6 91.7 91.7 110.0 100.0
Dutot Index (L-T ratio of arithmetic mean prices) 100.0 97.0 93.6 104.6 92.0 92.0 110.0 100.0
Jevons Index (chained S-T ratio of geometric mean prices) 100.0 96.3 92.4 105.6 91.7 91.7 110.0 100.0
Dutot Index (chained S-T ratio of arithmetic mean prices) 100.0 97.0 93.6 104.6 92.0 92.0 110.0 100.0
Table 8.2 Jevons and Carli Price Indices Using Averages of Long-Term Price Relatives
Base Jan. Feb. Mar. Apr. May Jun. Jul.
Item A Long-Term (L-T) Price Relatives
Variety 1 1.000 0.886 0.818 1.097 0.869 1.208 1.100 1.000

Variety 2 1.000 1.072 1.020 1.100 0.813 0.813 1.100 1.000
Variety 3 1.000 0.949 0.953 1.101 1.178 1.097 1.100 1.000
Variety 4 1.000 0.955 0.712 1.000 0.792 0.878 1.100 1.000
Variety 5 1.000 1.044 0.899 1.000 0.958 1.028 1.100 1.000
Variety 6 1.000 0.971 1.007 1.100 1.018 0.732 1.100 1.000
Variety 7 1.000 0.878 1.119 1.000 0.849 0.765 1.100 1.000
Geometric Mean of L-T Price Relatives 1.000 0.963 0.924 1.056 0.917 0.917 1.100 1.000
Jevons Index (L-T geometric changes) 100.0 96.3 92.4 105.6 91.7 91.7 110.0 100.0
Arithmetic Mean of L-T Price Relatives 1.000 0.965 0.933 1.057 0.925 0.932 1.100 1.000
Carli Index (L-T arithmetic changes) 100.0 96.5 93.3 105.7 92.5 93.2 110.0 100.0
Table 8.3 Jevons and Carli Price Indices Using Chained Short-Term Price Relatives
Jan. Feb. Mar. Apr. May Jun. Jul.
Elementary Aggregate A
Variety 1 0.886 0.923 1.342 0.792 1.390 0.911 0.909

Variety 2 1.072 0.952 1.078 0.739 1.000 1.353 0.909
Variety 3 0.949 1.004 1.155 1.070 0.932 1.003 0.909
Variety 4 0.955 0.745 1.405 0.792 1.109 1.253 0.909
Variety 5 1.044 0.861 1.113 0.958 1.073 1.070 0.909
Variety 6 0.971 1.037 1.092 0.925 0.719 1.501 0.909
Variety 7 0.878 1.275 0.894 0.849 0.901 1.438 0.909
Geometric Mean of S-T Price Relatives 0.963 0.959 1.143 0.868 1.000
Jevons Index (chained S-T geometric changes) 96.3 92.4 105.6 91.7 91.7 110.0 100.0
Arithmetic Mean of S-T Aggregate Relatives 0.965 0.971 1.154 0.875 1.018 1.219 0.909
Carli Index (chained S-T arithmetic changes) 96.5 93.7 108.1 94.6 96.3 117.4 106.7
A chained Dutot price index is calculated as  pi1  1  pi2  1  pit 

1    … 
1 1 1 1
0:t
I Cc =
n
∑ 0 
 p  n ∑ 1 
 p  n ∑ 
 p t −1  (8.6)
0:t n
∑ pi1 n
∑ pi2
n
∑ pit n
∑ pit i i i
I Dc = …. = (8.4) The chained Carli should be avoided because it has a known,

1 1 1 1
n
∑ pi0 n
∑ i n∑ i
p 1
p t −1
n
∑i
p 0
and potentially substantial, upward bias.4
8.20 Table 8.1 shows the comparison of the Dutot and
The third is the Carli index, defined as the unweighted arith- Jevons indices using the monthly average prices. The first
metic mean of the price relatives, or price ratios. The direct calculation for the Dutot index uses the average prices in the
Carli and the chained Carli, respectively, are calculated as
4
 pt  The results in the following tables are rounded to three decimals for aggre-
1
I 0:t
= ∑  i0  (8.5) gate price relatives and one decimal for price indices. The actual calcula-
C
n  pi  tions are derived in an Excel spreadsheet.
178
long-term formula (direct approach) where each month’s (t) 8.25 Another important property of the indices is that
average is compared to the initial base price (0) (that is, the the Jevons and the Dutot indices are transitive, whereas the
base price reference period). The Dutot index is also cal- Carli is not. Transitivity means that the chained monthly
culated using the short-term relatives (chained approach) indices are identical to the corresponding direct indices. This
where the month-to-month changes in average prices are property is important in practice, because many elementary
used to move forward the previous month’s index level as price indices are in fact calculated as chained indices that
shown in Table 8.3. The results are the same for both the link together the month-on-month indices. The intransitiv-
direct and chained approaches in the Dutot calculations. ity of the Carli index is illustrated dramatically in Table
Similarly, in Table 8.1 the Jevons index uses the geomet- 8.3 when each of the individual prices in the final month
ric average prices in the long-term and short-term formulas (July) return to the same level as it was in base month (as
to derive the price index levels that are the same for both observed in Table 8.1), but the chained Carli registers an
the long-term and short-term method. The Jevons indices increase of 6.7 percent over the base month. Similarly, in
do, however, differ from those calculated using the Dutot June, although each individual price is exactly 10 percent
formula. higher than the base month, the chained Carli registers an
8.21 In Table 8.2, the Jevons and Carli indices are cal- increase of 17.4 percent. These results would be regarded as
culated using the averages of long-term price relatives from problematic in the case of a direct index, but even in the case
the price reference period (base price). The results for the of a chained index, the results seem so intuitively unreason-
Carli indices are different from those of both the Jevons able as to undermine the credibility of the chained Carli.
and Dutot indices. The Jevons indices are exactly the same The price movements between April and May illustrate the
whether calculated using ratio average prices or average of effects of “price bouncing” in which the same seven prices
price relatives. are observed in both periods but they are switched between
8.22 The properties and behavior of the different index the different varieties. The monthly Carli index (short-term
formulas are summarized in paragraphs 8.21–8.48 (see also and long-term) increases from April to May whereas both
Chapter 6 of Consumer Price Index Theory). First, the dif- the Dutot and the Jevons indices are unchanged.
ferences between the results obtained by using the differ- 8.26 One general property of geometric means should
ent formulas tend to increase as the variance of the price be noted when using the Jevons index. If any single obser-
relatives, or ratios, increases. The greater the dispersion of vation out of a set of observations is zero, their geometric
the price movements, the more critical the choice of index mean is undefined, whatever the values of the other obser-
formula, and method, becomes. If the elementary aggregates vations. The Jevons index is sensitive to extreme falls in
are defined in such a way that the price variations within the prices and it may be necessary to impose upper and lower
aggregate are minimized, the results obtained become less bounds on the individual price ratios of, for example, 10 and
sensitive to the choice of formula. 0.1, respectively, when using the Jevons index. This range
8.23 Certain features displayed by the data in Tables 8.1 should be determined after assessing the typical size of price
and 8.2 are systematic and predictable; they follow from the movements and may vary across different product groups.
mathematical properties of the indices. For example, it is Of course, extreme observations often result from errors, so
well known that an arithmetic mean is always greater than, extreme price movements should be carefully checked. It
or equal to, the corresponding geometric mean, the equality is not recommended to replace a zero price by an arbitrary
holding only in the trivial case in which the numbers being low value in the Jevons index as this could lead to unstable
averaged are all the same. The direct Carli indices are there- results. If the Jevons index is used and the price moves from
fore all greater than the Jevons indices, except in the price positive to zero, a practical solution is to split the aggre-
reference period, in June when all prices increased by 10 gate into two and estimate weights for each part. The zero
percent above their base prices, and the end period (July) subindex multiplied by the positive weight plus the nonzero
when all prices return to their base price values. In general, Jevons subindex multiplied by the remaining weight is well
the Dutot may be greater or less than the Jevons but tends to defined, and the price change is taken into account.
be less than the Carli. 8.27 The message emerging from this brief illustration
8.24 The Carli and Jevons indices depend only on the of the behavior of just three possible formulas is that dif-
price relatives and are unaffected by the price level. The ferent index numbers and methods can deliver very differ-
Dutot index, in contrast, is influenced by the price level. In ent results. With the knowledge of these interrelationships,
the Dutot index, price changes are implicitly weighted by one can infer that the chained Carli formula is not recom-
the price in the base (price reference) period, so that price mended. However, this information in itself is not sufficient
changes on more expensive products are assigned a higher to determine which formula should be used, even though it
importance than similar price changes for cheaper products makes it possible to make a more informed and reasoned
(this can be seen from equation 8.3). In Tables 8.1 and 8.3, choice. It is necessary to appeal to other criteria to settle the
this is illustrated in the development of the March index choice of formula. There are two main approaches that may
where prices for varieties 4, 5, and 7, which have the high- be used, the axiomatic and the economic approaches, which
est base prices, are the same as in the price reference month are presented in paragraphs 8.28–8.41. First, however, it is
and mitigate the 10 percent price increases of varieties 1, useful to consider the sampling properties of the elementary
2, 3, and 6 from the price reference month. The monthly indices.
Dutot price index is 104.6 versus 105.6 in the Jevons, and
105.7 in the Carli. Because of the relatively high base prices Sampling Properties of Elementary Price Indices
for varieties 4, 5, and 7, that results in a lower level for the 8.28 The interpretation of the elementary price indi-
Dutot index. ces is related to the way in which the sample of goods
179
and services is drawn. Hence, if goods and services in the whereas the Carli index does not. For example, although
sample are selected with probabilities proportional to the the prices in July have returned to the same levels as the
population expenditure shares in the price reference period, index reference period, the chained Carli registers 106.7.
then: This illustrates the fact that the chained Carli may have a
significant built-in upward bias.
• The sample (unweighted) Carli index provides an unbi-
ased estimate of the population Laspeyres price index 8.31 Many other axioms or tests can be devised, but the
(see equation 8.11). previous ones5 illustrate the approach and also throw light
• The sample (unweighted) Jevons index provides an unbi- on some important features of the elementary indices under
ased estimate of the population geometric Laspeyres price consideration in this Manual and provide evidence of the
index (see equation 8.14). preference for the Jevons index.
8.32 The sets of products covered by elementary aggre-
8.29 If goods and services are sampled with probabili- gates are meant to be as homogeneous as possible. If they
ties proportional to population quantity shares in the price are not fairly homogeneous, the failure of the Dutot index to
reference period, the sample (unweighted) Dutot index satisfy the units of measurement or commensurability test
would provide an estimate of the population Laspeyres can be a serious disadvantage. Although defined as the ratio
price index. However, if the basket for the Laspeyres index of the unweighted arithmetic average prices, the Dutot index
contains different kinds of products whose quantities are not may also be interpreted as a weighted arithmetic average of
additive, the quantity shares, and hence the probabilities, are the price relatives in which each relative is weighted by its
undefined. price in the base price period.6 This can be seen by rewriting
formula 8.3 as
Axiomatic Approach to Elementary
Price Indices 1  pt 
∑ pi0  i0 
8.30 As explained in Chapters 3 and 6 in Consumer n  pi 
Price Index Theory, one way to decide upon an appropri- I D0:t = (8.7)
1
ate index formula is to require it to satisfy certain specified
n
∑ pi0
axioms or tests. The tests throw light on the properties that
different kinds of indices have, some of which may not be However, if the products are not homogeneous, the relative
intuitively obvious. Four basic tests are cited to illustrate the prices of the different varieties may depend quite arbitrarily
axiomatic approach: on the quantity units in which they are measured.
• Proportionality test. If all prices are λ times the prices 8.33 Consider, for example, salt and pepper, which
in the price reference period, the index should equal λ. are found within the same class of Classification of Indi-
The data for June in Tables 8.1–8.3, when every price vidual Consumption According to Purpose. Suppose the
is 10 percent higher than in the price reference period, unit of measurement for pepper is changed from grams to
show that all three direct indices satisfy this test. A special ounces, while leaving the units in which salt is measured
case of this test is the identity test, which requires that if (for example, kilos) unchanged. As an ounce of pepper is
the price of every variety is the same as in the reference equal to 28.35 grams, the “price” of pepper increases by
period, the index should be equal to unity, as in the last over 28 times, which effectively increases the weight given
month (July) in the example. to pepper in the Dutot index by over 28 times. The price
of pepper relative to salt is inherently arbitrary, depending
• Changes in the units of measurement test (commensura- entirely on the choice of units in which to measure the two
bility test). The price index should not change if the quan- goods. In general, when there are different kinds of products
tity units in which the products are measured are changed, within the elementary aggregate, the use of the Dutot index
for example, if the prices are expressed per liter rather is not acceptable.
than per pint. The Dutot index fails this test, as explained 8.34 The use of the Dutot index is acceptable only
in paragraphs 8.29–8.33, but the Carli and Jevons indices when the set of varieties covered is homogeneous, or at least
satisfy the test. nearly homogeneous. For example, it may be acceptable
• Time reversal test. If all the data for the two periods are for a set of apple prices even though the apples may be of
interchanged, the resulting price index should equal the different varieties, but not for the prices of several differ-
reciprocal of the original price index. The chained Carli ent kinds of fruits, such as apples, pineapples, and bananas,
index fails this test, but the Dutot and the Jevons indices some of which may be much more expensive per variety
both satisfy the test. The failure of the chained Carli to sat- or per kilo than others. Even when the varieties are fairly
isfy the test is not immediately obvious from the example homogeneous and measured in the same units, the Dutot’s
but can easily be verified by calculating the index back- implicit weights may still not be satisfactory. More weight is
ward from June to the index reference period. In this case, given to the price changes for the more expensive varieties,
the chained Carli from June backward is 97.0 whereas
the reciprocal of the forward chained Carli is (1/1.174) ×
100 = 85.2. 5
Note that an index that satisfies the transitivity test and the identity test
• Transitivity test. The chained index between two periods automatically also satisfies the time reversal test.
6
Although the Jevons index is nonlinear, it can be approximated as a
should equal the direct index between the same two peri- weighted average of price relatives, where the weights correspond to the
ods. It can be seen from the example in Tables 8.1–8.3 square root of the inverse price relatives (see J. Mehrhoff. 2007. “A Linear
that the Jevons and the Dutot indices both satisfy this test, Approximation to the Jevons Index.” in: y.d. Lippe, P.M.).
180
but in practice, they may well account for only small shares a random sample would provide an estimate of the COLI
of the total expenditure within the aggregate. Consumers are if the varieties are selected with probabilities proportional
unlikely to buy varieties at high prices if the same varieties to the population expenditure shares. If the varieties were
are available at lower prices. selected with probabilities proportional to the population
8.35 It may be concluded that from an axiomatic view- quantity shares (assuming the quantities are additive), the
point, both the Carli and the Dutot indices, although they sample Dutot would provide an estimate of the underlying
have been, and still are, used by some NSOs, have disad- COLI.
vantages. The Carli index fails the time reversal and tran- 8.39 The second case occurs when consumers are
sitivity tests. In principle, it should not matter whether it assumed to vary the quantities they consume in inverse
is chosen to measure price changes forward or backward proportion to the changes in relative prices. The cross-
in time. It would be expected to give the same answer, but elasticities of demand between the different varieties are
this is not the case for the chained Carli indices that may all unity, the expenditure shares being the same in both
be subject to a significant upward bias. The Dutot index is periods. The underlying preferences are described by a
meaningful for a set of homogeneous varieties but becomes “Cobb–Douglas” utility function. With these preferences,
increasingly arbitrary as the set of products becomes more the Jevons index calculated for a random sample would pro-
diverse. On the other hand, the Jevons index satisfies all vide an unbiased estimate of the COLI, provided that the
the tests listed in paragraph 8.28 and emerges as the pre- varieties are selected with probabilities proportional to the
ferred index when the set of tests is enlarged, as shown pre- population expenditure shares.
viously in paragraphs 8.28–8.29. From an axiomatic point 8.40 On the basis of the economic approach, the choice
of view, the Jevons index is clearly the index with the best between the sample Jevons and the sample Carli rests on
properties. which is likely to approximate more closely the underlying
COLI: in other words, on whether the (unknown) cross-
elasticities of demand are likely to be closer to unity or zero,
Economic Approach to Elementary on average. In practice, the cross-elasticities of demand could
Price Indices take on any value ranging up to plus infinity for an elemen-
8.36 In the economic approach, the objective is to tary aggregate consisting of a set of strictly homogeneous
estimate an economic index, that is, a cost of living index varieties (that is, perfect substitutes). It should be noted that
(COLI) for the elementary aggregate (see Chapter 6 of in the limit when the products really are homogeneous, there
Consumer Price Index Theory). The varieties for which is no index number problem, and the price “index” is given
prices are collected are treated as if they constituted a bas- by the ratio of the unit values in the two periods. It may be
ket of goods and services purchased by consumers, from conjectured that the average cross-elasticity is likely to be
which the consumers derive utility. A COLI measures closer to unity than zero for most elementary aggregates,
the minimum amount by which consumers would have especially since these should be constructed in such a way as
to change their expenditures in order to keep their utility to group together similar varieties that are close substitutes
level unchanged, allowing consumers to make substitu- for each other. Thus, in general, the Jevons index is likely to
tions between the varieties in response to changes in the provide a closer approximation to the COLI than the Carli.
relative prices of varieties. In this case, the Carli index must be viewed as having an
8.37 The economic approach is based on several upward bias.
assumptions about consumer behavior, market conditions, 8.41 The use of the Jevons index in the context of the
and the representativity of the sample. These assumptions economic approach implies that the quantities are assumed
do not always hold in reality. At the detailed level of ele- to vary over time in response to changes in relative prices.
mentary aggregates, special conditions will often prevail As a result of the inverse relation of movements in prices
and change over time, and the information available about and quantities, the expenditure shares are constant over
outlets, products, and market conditions may be incom- time. Carli and Dutot, on the other hand, keep the quanti-
plete. Thus, although the economic approach may be use- ties fixed while the expenditure shares vary in response to
ful in providing a possible economic interpretation of the change in relative prices.
index, conclusions should be made with caution. In general, 8.42 The Jevons index does not imply that expenditure
in the decision of how to calculate the elementary indices shares remain constant. Obviously, the Jevons index can
one should be careful not to put too much weight on a strict be calculated whatever changes do or do not occur in the
economic interpretation of the index formula at the expense expenditure shares in practice. What the economic approach
of the statistical considerations. shows is that if the expenditure shares remain constant (or
8.38 In the absence of information about quantities roughly constant), then the Jevons index can be expected
or expenditure within an elementary aggregate, an eco- to provide a good estimate of the underlying COLI. Simi-
nomic index can only be estimated when certain special larly, if the relative quantities remain constant, then the Carli
conditions are assumed to prevail. There are two special index can be expected to provide a good estimate, but the
cases of some interest. The first case is when consumers Carli index does not actually imply that quantities remain
continue to consume the same relative quantities what- fixed.
ever the relative prices. Consumers prefer not to make 8.43 It may be concluded that, on the basis of the eco-
any substitutions in response to changes in relative prices. nomic approach as well as the axiomatic approach, the
The cross-elasticities of demand are zero. The underlying Jevons emerges as the preferred index, although there may
preferences are described in the economics literature as be cases in which little or no substitution takes place within
“Leontief.” In this first case, the Carli index calculated for the elementary aggregate and the direct Carli or Dutot
181
indices might be used. The chained Carli should be avoided will require successive old and new prices for the present
altogether. The Dutot index may be used provided the ele- and the preceding months. For a chained index, the replace-
mentary aggregate consists of homogeneous products. In ment variety for a missing observation would also need to
general, the index compiler should use the Jevons index for have prices for the current and previous period. However, if
the elementary aggregates. the previous price is not available, it will have an impact on
the index for two months, since the substitute observation
Chained versus Direct Indices for cannot be used until the subsequent month. It also is pos-
Elementary Aggregates sible to impute the price of the missing variety in the first
missing month so that the next period price can be compared
8.44 In a direct elementary index, the prices of the cur- to the imputed price.
rent period are compared directly with those of the price ref- 8.49 A missing price does not present such a problem
erence period. In a chained index, prices in each period are in the case of a direct index. In a direct index a single, non-
compared with those in the previous period, the resulting estimated missing observation will only have an impact on
short-term indices being chained together to obtain the long- the index in the current period. For example, for a compari-
term index, as illustrated in Tables 8.1–8.3. son between periods 0 and 3, a missing price of the replace-
8.45 Provided that prices are recorded for the same ment in period 2 means that the chained index excludes
set of varieties in every period, as in Table 8.1, any index the variety for the last link of the index in periods 2 and
formula defined as the ratio of the average prices will 3. By comparison, the direct index includes it in period 3
be transitive: that is, the same result is obtained whether since a direct index will be based on varieties whose prices
the index is calculated as a direct index or as a chained are available in periods 0 and 3 (unless an imputation is
index. In a chained index, successive numerators and made). In general, however, the use of a chained index can
denominators will cancel out, leaving only the average make the estimation of missing prices and the introduction
price in the last period divided by the average price in the of replacements easier from a computational point of view,
reference period, which is the same as the direct index. whereas it may be inferred that a direct index will limit the
Both the Dutot and the Jevons indices are therefore tran- usefulness of overlap methods for dealing with missing
sitive. As already noted, however, a chained Carli index observations.
is not transitive and should not be used because of its 8.50 The direct and the chained approaches also pro-
upward bias. duce different by-products that may be used for monitoring
8.46 Although the chained and direct versions of the price data. For each elementary aggregate, a chained index
Jevons and Dutot indices are identical when there are no approach gives the latest monthly price change, which can
breaks in the series for the individual varieties, they offer be useful for both data editing and imputation of missing
different ways of dealing with new and disappearing vari- prices. By the same token, however, a direct index derives
eties, missing prices, and quality adjustments. In practice, average price levels for each elementary aggregate in each
products continually need to be excluded from the index and period, and this information may be a useful by-product.
new ones included, in which case the direct and the chained Nevertheless, because the availability of computing power
indices may differ if the imputations for missing prices are at a low cost and of spreadsheets allows such by-products
made differently. to be calculated whether a direct or a chained approach is
8.47 When a replacement variety must be included applied, the choice of formula should not be dictated by con-
in a direct index, it will often be necessary to estimate the siderations regarding by-products.
price of the new variety in the price reference (base) period,
which may be some time in the past. The same happens
if, as a result of an update of the sample, new varieties are Consistency in Aggregation
linked into the index. If no information exists on the price 8.51 Consistency in aggregation means that if an index
of the replacement variety in the price reference period, it is calculated stepwise by aggregating lower-level indices
will be necessary to estimate it using price ratios calculated to obtain indices at progressively higher levels of aggrega-
for the varieties that remain in the elementary aggregate, a tion, the same overall result should be obtained as if the
subset of these varieties, or some other indicator. However, calculation had been made in one step. For example, aggre-
the direct approach should only be used for a limited period gating the elementary aggregate indices to the all-items
of time. Otherwise, most of the reference prices would end index gives the same result as aggregating the group-level
up being imputed, which would be an undesirable outcome. indices to the all-items index. For presentational purposes,
This effectively rules out the use of the Carli index over this is an advantage. If the elementary aggregates are calcu-
a long period of time, as the Carli should only be used in lated using one formula and the elementary aggregates are
its direct form and not in chained form as previously dis- averaged to obtain the higher-level indices using another
cussed. This implies that, in practice, the direct Carli may formula, the resulting CPI is not consistent in aggregation.
be used only if the overall index is chain linked annually, However, consistency in aggregation is not necessarily the
or biannually. most important criterion, and it is unachievable when the
8.48 In a chained index, if a variety becomes perma- amount of information available on quantities and expen-
nently missing, a replacement variety can be linked into the diture is not the same at the different levels of aggregation.
index as part of the ongoing index calculation by including In addition, there may be different degrees of substitution
the variety in the monthly index as soon as prices for two within elementary aggregates as compared to the degree
successive months are obtained. Similarly, if the sample is of substitution between products in different elementary
updated and new products must be linked into the index, this aggregates.
182
8.52 The Carli index would be consistent in aggregation unpriced for some length of time. In general, to carryfor-
with the Laspeyres index if the varieties were to be selected ward is not an acceptable procedure or solution to the
with probabilities proportional to expenditures in the refer- problem of missing prices.
ence period. This is typically not the case. The Dutot and 8.57 Imputation of the missing price by the average
the Jevons indices are not consistent in aggregation with a change of the available prices may be applied for elemen-
higher-level Laspeyres. As explained in paragraphs 8.88– tary aggregates where the prices can be expected to move
8.94, however, the CPIs actually calculated by NSOs are in the same direction. The imputation can be made using
usually not true Laspeyres indices, even though they may be all the remaining prices in the elementary aggregate. As
based on fixed baskets of goods and services. If the higher- already noted, this is numerically equivalent to omitting
level index were to be defined as a geometric Laspeyres, the variety for the immediate period, but it is useful to
consistency in aggregation could be achieved by using the make the imputation so that if the price becomes avail-
Jevons index for the elementary indices at the lower level, able again in a later period the sample size is not reduced
provided that the individual varieties are sampled with prob- in that period. In some cases, depending on the homoge-
abilities proportional to expenditure. Although unfamiliar, a neity of the elementary aggregate, it may be preferable to
geometric Laspeyres has desirable properties from an eco- use only a subset of varieties from the elementary aggre-
nomic point of view and is considered again later. gate to estimate the missing price. In some instances, this
may even be a single comparable variety from a similar
Missing Price Observations type of outlet whose price change can be expected to be
8.53 The price of a variety may fail to be collected in similar to the missing one. See Chapter 6 on imputation
some period either because the variety is missing tempo- methods.
rarily or because it has permanently disappeared. The two 8.58 Tables 8.4A and 8.4B illustrate the calculation of
classes of missing prices require different treatment as noted the price index for the elementary aggregate where the price
in Chapter 6. Temporary unavailability may occur for sea- for variety 6 is missing in March. The long-term (direct)
sonal varieties (particularly for fruit, vegetables, and cloth- indices are therefore calculated based on the six varieties
ing), because of supply shortages, or possibly because of with reported prices. The short-term (chained) indices are
some collection difficulty (for example, an outlet was closed calculated based on all seven prices from January to Feb-
or a price collector was ill). The treatment of seasonal variet- ruary and from April to July. From February to March and
ies raises several particular problems. These are dealt with from March to April the monthly indices are calculated
in Chapter 11. based on six varieties only.
8.59 The average prices (both arithmetic and geomet-
Treatment of Temporarily Missing Prices ric) are calculated using the six available prices for the base
8.54 In the case of temporarily missing observations for period, February, March, and April in Table 8.4A. The direct
nonseasonal varieties, one of four actions may be taken: Jevons and Dutot indices use the average of the six prices in
• Omit the variety for which the price is missing so that March and the base period to derive the March index (104.9
a matched sample is maintained (like is compared with and 104.1, respectively). This calculation uses a matched
like) even though the sample is depleted sample for the prices available in each period (March and
the base period) to derive the averages. In April, all seven
• Carryforward the last observed price
prices are again available so the direct indices are derived by
• Impute the missing price by the average price change for comparing the averages of the seven prices to their average
the prices that are available in the elementary aggregate in the base period.
• Impute the missing price by the price change for a par- 8.60 For the chained Jevons and Dutot indices that use
ticular comparable variety from another similar outlet the short-term price relatives, the average prices for the six
8.55 Omitting an observation from the calculation of varieties available in March are compared to the average
an elementary index is equivalent to assuming that the prices of the six available varieties in February. The result-
price would have moved in the same way as the average ing price relatives are multiplied by the February indices to
change in the prices of the varieties that remain included derive the March indices (106.4 for the Jevons and 104.8 for
in the index. Omitting an observation changes the implicit the Dutot). The same holds true for April’s compilation—the
weights attached to the other prices in the elementary average of the six prices that were available in both March
aggregate. and April are used to derive the April indices (91.4 for the
8.56 Carrying forward the last observed price is not Jevons and 91.8 for the Dutot).
recommended, except in the case of fixed or regulated 8.61 For both the Jevons and the Dutot indices, the
prices. Special care needs to be taken in periods of high direct and chained indices now differ from March onward.
inflation or when markets are changing rapidly as a result The first link in the chained index (January to February) is
of a high rate of innovation and product turnover. While the same as the direct index, so the two indices are identical
simple to apply, carrying forward the last observed price numerically. The direct index for March completely ignores
biases the resulting index toward zero change. In addition, the price increase of variety 6 between January and Feb-
when the price of the missing variety is recorded again, ruary, while this is counted in the chained index. As a result,
there is likely to be a compensating step change in the the direct index is lower than the chained index for March.
index to return to its proper value. The adverse effect on On the other hand, in April, when all prices are again avail-
the index will be increasingly severe if the variety remains able, the direct index captures the price development for the
183
Table 8.4A Jevons and Dutot Elementary Price Indices Using Averages with Missing Prices
Base Match Jan. Feb. Mar. Apr. May Jun. Jul.
Variety 1 2.36 2.36 2.09 1.93 2.59 2.05 2.85 2.59 2.36
Variety 2 5.02 5.02 5.38 5.12 5.52 4.08 4.08 5.52 5.02
Variety 3 5.34 5.34 5.07 5.09 5.88 6.29 5.86 5.88 5.34
Variety 4 6.00 6.00 5.73 4.27 6.00 4.75 5.27 6.60 6.00
Variety 5 6.12 6.12 6.39 5.50 6.12 5.86 6.29 6.74 6.12
Variety 6 2.80 2.72 2.82 2.85 2.05 3.08 2.80
Variety 7 6.21 6.21 5.45 6.95 6.21 5.27 4.75 6.84 6.21
Geometric Mean Price (seven observations) 4.55 4.38 4.20 4.17 4.17 5.01 4.55
Geometric Mean Price (six matched observations) 4.93 4.49 5.17 4.45
L-T Aggregate Relative 1.000 0.963 0.924 1.049 0.917 0.917 1.100 1.000
Jevons Index (direct) 100.0 96.3 92.4 104.9 91.7 91.7 110.0 100.0
Geometric Mean S−T Aggregate Relatives 0.963 0.959 1.152 0.859 1.000 1.200 0.909
Jevons Index (chained averages) 100.0 96.3 92.4 106.4 91.4 91.4 109.7 99.7
Arithmetic Mean Price (seven observations) 4.84 4.69 4.53 4.45 4.45 5.32 4.84
Arithmetic Mean Price (six matched observations) 5.18 4.81 5.39 4.72
Dutot Index (direct) 100.0 97.0 93.6 104.1 92.0 92.0 110.0 100.0
S−T Aggregate Relatives 0.970 0.965 1.120 0.876 1.000 1.196 0.909
Dutot Index (chained averages) 100.0 97.0 93.6 104.8 91.8 91.8 109.7 99.7
Note: The text in gray refers to six matched observations whereas the text in bold refers to seven matched observations.
Table 8.4B Jevons and Carli Elementary Price Indices Using Relatives with Missing Prices
Variety 1 0.886 0.818 1.097 0.869 1.208 1.100 1.000
Variety 2 1.072 1.020 1.100 0.813 0.813 1.100 1.000
Variety 3 0.949 0.953 1.101 1.178 1.097 1.100 1.000
Variety 4 0.955 0.712 1.000 0.792 0.878 1.100 1.000
Variety 5 1.044 0.899 1.000 0.958 1.028 1.100 1.000
Variety 6 0.971 1.007 1.018 0.732 1.100 1.000
Variety 7 0.878 1.119 1.000 0.849 0.765 1.100 1.000
Geometric Mean Price Relatives (seven observations) 0.963 0.924 0.917 0.917 1.100 1.000
Geometric Average Price Relatives (six observations) 1.049
Jevons Index (mean L-T price relative) 96.3 92.4 104.9 91.7 91.7 110.0 100.0
Arithmetic Mean Price Relative (seven observations) 0.965 0.933 0.925 0.932 1.100 1.000
Arithmetic Mean Price Relative (six observations) 1.050
Carli Index (average L-T price relative) 96.5 93.3 105.0 92.5 93.2 110.0 100.0
Elementary Aggregate A S-T Price Relatives
Variety 1 0.886 0.923 1.342 0.792 1.390 0.911 0.909
Variety 2 1.072 0.952 1.078 0.739 1.000 1.353 0.909
Variety 3 0.949 1.004 1.155 1.070 0.932 1.003 0.909
Variety 4 0.955 0.745 1.405 0.792 1.109 1.253 0.909
Variety 5 1.044 0.861 1.113 0.958 1.073 1.070 0.909
Variety 6 0.971 1.037 0.719 1.501 0.909
Variety 7 0.878 1.275 0.894 0.849 0.901 1.438 0.909
Geometric Mean Aggregate Relatives (seven observations) 0.963 0.959 1.000 1.200 0.909
Geometric Mean Aggregate Relatives (six matched observations) 1.153 0.859
Jevons Index (chained S-T price relatives) 96.3 92.4 106.4 91.4 91.4 109.7 99.7
Arithmetic Mean Aggregate Relatives (seven observations) 0.965 0.971 1.018 1.219 0.909
Arithmetic Mean Aggregate Relatives (six matched observations) 1.164 0.866
Carli Index (chained S-T aggregate relatives) 96.5 93.7 109.1 94.5 96.2 117.3 106.6
full sample, whereas the chained index only tracks the long- 8.63 As Tables 8.4A and 8.4B demonstrate, the Jevons,
term development in the six-price sample. Dutot, and Carli direct indices return to 100.0 in the final
8.62 Table 8.4B shows the compilation of the Jevons period when all prices return to their base period levels.
and Carli indices using the long-term and short-term average The chained versions do not, with the Carli showing a large
of price relative methods. The long-term Carli index shows upward drift by the end month and the Jevons and Dutot
similar effects in March and April as those for the Jevons with a slight downward drift.
index in missing the long-term price change for variety 6. 8.64 The problem with the chained index will be
The short-term Carli, however, shows a significant upward resolved if the missing price is imputed using the average
bias as it increased to 106.6 when all the prices return to short-term change of the other observations in the elemen-
their base period levels in July. tary aggregate. In Table 8.5A, the missing price for variety 6
184
in March is imputed using the geometric average of the price 8.65 The calculations in Tables 8.5A and 8.5B show
changes of the remaining varieties from February to March. that when the missing price for variety 6 is imputed using
While the imputation might be calculated using long-term the short-term price change of the other varieties, the trend
relatives (that is, comparing the prices of the present period of the Jevons, Dutot, and Carli indices reflect the changes
with the base period prices), the imputation of missing prices for all the observations using the direct and long-term rela-
should be made based on the price change from the preced- tive methods. For the Jevons and Dutot indices, the chained
ing to the present period, as shown in the table. Imputation method gives the same results as the direct method. How-
based on the average price change from the base period to ever, the chained Carli is significantly upward biased dem-
the present period should not be used as it ignores the infor- onstrating that this method should not be used for index
mation about the price change of the missing variety that has compilation.
already been included in the index. The treatment of imputa-
tions is discussed in more detail in Chapter 6.
Table 8.5A Jevons and Dutot Elementary Price Indices Using Averages with Imputed Prices
Base Match Jan. Feb. Mar. Apr. May Jun. Jul.
Elementary Aggregate A Prices

Variety 1 2.36 2.36 2.09 1.93 2.59 2.05 2.85 2.59 2.36
Variety 2 5.02 5.02 5.38 5.12 5.52 4.08 4.08 5.52 5.02
Variety 3 5.34 5.34 5.07 5.09 5.88 6.29 5.86 5.88 5.34
Variety 4 6.00 6.00 5.73 4.27 6.00 4.75 5.27 6.60 6.00
Variety 5 6.12 6.12 6.39 5.50 6.12 5.86 6.29 6.74 6.12
Variety 6 2.80 2.72 2.82 3.25 2.85 2.05 3.08 2.80
Variety 7 6.21 6.21 5.45 6.95 6.21 5.27 4.75 6.84 6.21
Geometric Mean Price (seven observations) 4.55 4.38 4.20 4.84 4.17 4.17 5.01 4.55
Geometric Mean Price (six observations) 4.93 4.49 5.17
Jevons Index (direct) 100.0 96.3 92.4 106.4 91.7 91.7 110.0 100.0
Geometric Mean S−T Aggregate Relative 0.963 0.959 1.152 0.862 1.000 1.200 0.909
Jevons Index (chained averages) 100.0 96.3 92.4 106.4 91.7 91.7 110.0 100.0
Variety 6 (imputed price) 2.80 2.72 2.82 3.16 2.85 2.05 3.08 2.80
Arithmetic Mean Price (seven observations) 4.84 4.69 4.53 5.07 4.45 4.45 5.32 4.84
Arithmetic Mean Price (six observations) 5.18 4.81 5.39
Dutot Index (direct) 100.0 97.0 93.6 104.8 92.0 92.0 110.0 100.0
S−T Aggregate Relatives 0.970 0.965 1.120 0.878 1.000 1.196 0.909
Dutot Index (chained averages) 100.0 97.0 93.6 104.8 92.0 92.0 110.0 100.0
Table 8.5B Jevons and Dutot Elementary Price Indices Using Relatives with Imputed Prices
Variety 1 0.888 0.816 1.100 0.869 1.207 1.100 1.000
Variety 2 1.072 1.019 1.100 0.813 0.813 1.100 1.000
Variety 3 0.949 0.953 1.100 1.178 1.097 1.100 1.000
Variety 4 0.955 0.712 1.000 0.792 0.878 1.100 1.000
Variety 5 1.044 0.898 1.000 0.957 1.028 1.100 1.000
Variety 6 0.974 1.008 1.160 1.018 0.733 1.100 1.000
Variety 7 0.877 1.118 1.000 0.848 0.765 1.100 1.000
Geometric Mean Price Relatives (seven observations) 0.963 0.924 1.064 0.917 0.917 1.100 1.000
Jevons Index (average L-T price relatives) 96.3 92.4 106.4 91.7 91.7 110.0 100.0
Variety 6 (imputed L-T price relative) 0.963 0.924 1.173 0.917 0.917 1.100 1.000
Arithmetic Mean Aggregate Relatives (seven observations) 0.933 1.067 0.925 0.932 1.100 1.000 0.933
Carli Index (average L-T price relatives) 93.3 106.7 92.5 93.2 110.0 100.0 93.3
Variety 1 0.886 0.923 1.342 0.792 1.390 0.909 0.911
Variety 2 1.072 0.952 1.078 0.739 1.000 1.353 0.909
Variety 3 0.949 1.004 1.155 1.070 0.932 1.003 0.908
Variety 4 0.955 0.745 1.405 0.792 1.109 1.252 0.909
Variety 5 1.044 0.861 1.113 0.958 1.073 1.072 0.908
Variety 6 0.971 1.037 1.152 0.877 0.719 1.502 0.909
Variety 7 0.878 1.275 0.894 0.849 0.901 1.440 0.908
Geometric Mean Price Relatives (seven observations) 0.963 0.959 1.152 0.862 1.000 1.200 0.909
Jevons Index (chained S-T price relatives ) 96.3 92.4 106.4 91.7 91.7 110.0 100.0
Variety 6 (imputed S-T price relative) 0.963 0.959 1.164 0.782 1.000 1.200 0.909
Arithmetic Mean Aggregate Price Relatives (seven observations) 0.965 0.971 1.164 0.868 1.018 1.219 0.909
Carli Index (chained S-T price relatives ) 96.5 93.7 109.2 94.7 96.4 117.5 106.8
185
Treatment of Permanently price movement between periods t − 1 and t, based on the
Disappeared Varieties matched set of prices in those two periods, onto the value of
8.66 Varieties may disappear permanently for a num- the chained index for period t − 1. In the example, no further
ber of reasons. The variety may disappear from the market imputation is required after April, and the subsequent move-
because new varieties have been introduced or the outlets ment of the index is unaffected by the imputed price change
from which the price has been collected have stopped sell- between March and April.
ing the product. Where varieties disappear permanently, a 8.71 For the Dutot index, the short-term relative of aver-
replacement variety must be sampled and included in the age prices is used to make imputations. In the Dutot example
index. The replacement variety should ideally be one that in Table 8.7, the average base price used in the direct calcu-
accounts for a significant proportion of sales, is likely to lation must be adjusted for the relative difference between
continue to be sold for some time, and is likely to be repre- the old sample’s average price and the new sample’s aver-
sentative of the sampled price changes of the market that the age price. When using the long-term Dutot index based on
old variety covered. In practice when selecting replacement an arithmetic mean of prices, the imputation of base price
varieties, compromises must be found between representa- is made using the new sample average price and long-term
tivity, comparability over time, and similarity. elementary index to estimate the average base price. The
8.67 The timing of the introduction of replacement vari- trend of the index is affected by the level of the base prices
eties is important. Many new products are initially sold at where the movement of the observation with largest base
high prices that then gradually drop over time, especially price has the most importance in the trend of the elementary
as the volume of sales increases. Alternatively, some prod- index. In the Jevons and Carli indices, each observation is
ucts may be introduced at artificially low prices to stimulate equally important, and estimation of the base prices is not
demand. In such cases, delaying the introduction of a new or affected by the level of the other observations in the sample.
replacement variety until a large volume of sales is achieved 8.72 The adjusted base price in this example is derived
may miss some systematic price changes that ought to be by dividing the new average price level by the long-term
captured by CPIs. It is desirable to avoid making replace- price change of the elementary index. From another per-
ments when sales of the varieties they replace are signifi- spective, the adjusted base price is estimated by applying the
cantly discounted in order to clear out inventory. In such ratio of the new sample’s average price to the old sample’s
cases, the disappearing variety’s price should be returned to average price to the old base price. This implicitly assumes
its last nondiscounted price as the new variety is introduced. that the difference in the average prices reflects the differ-
8.68 To include the new variety in the index, an imputed ence in quality.
price needs to be calculated. The imputation will differ 8.73 The situation is somewhat simpler when there is
based on the formula used. For the Jevons index, the geo- an overlap month in which prices are collected for both the
metric average of short-term relatives is used, while for the disappearing and the replacement variety. In that case, it is
Carli index, the arithmetic average of short-term relatives is possible to link the price series for the new variety to the
used. For the Dutot index, the short-term relative of average price series for the old variety that it replaces. Linking with
prices is used. If a direct index is being calculated from aver- overlapping prices involves making an implicit adjustment
age prices, the imputed price must be included in calculating for the difference in quality between the two varieties, as
the average prices in the current month. For the Jevons and it assumes again that the relative prices of the new and old
Carli indices, the base price can be estimated by using the varieties reflect their relative qualities. For perfect or nearly
price ratio of the new variety price to the imputed price of perfect markets, this may be a valid assumption, but for
the old variety as the relative quality difference. This ratio is certain markets and products, it may not be so reasonable.
then applied to the base price of the old variety. A different The question of when to use overlapping prices is dealt with
method must be used for estimating the Dutot base price that in detail in Chapter 6. The overlap method is illustrated in
involves estimating the average base price using the long- Table 8.8.
term price change of the elementary aggregate. 8.74 In the example in Table 8.8 overlapping prices are
8.69 Table 8.6 shows an example where variety A dis- obtained for varieties A and D in March. There is now an
appears after March and variety D is included as a replace- overlapping sample for March—one using varieties A, B, C,
ment from April onward. Varieties A and D are not available and the other using varieties B, C, and D. A monthly chain
on the market at the same time and their price series do not Jevons index of geometric mean prices will be based on the
overlap. The base price estimation in the examples applies prices of varieties A, B, and C until March, and from April
to the Jevons and Carli price indices. The methods for the onward on the prices of varieties B, C, and D. The replace-
Dutot price index are shown in Table 8.7. ment variety is not included until prices for two successive
8.70 If a chained index is calculated, the imputation periods are obtained. Thus, the monthly chain index has the
method ensures that the inclusion of the new variety does advantage that it is not necessary to carry out any explicit
not, in itself, affect the index and an adjustment of the base imputation of a reference (base) price for the new variety.
price is not necessary. In the case of a chained index, imput- The same approach applies to the Dutot chain index.
ing the missing price by the average change of the avail- 8.75 If a direct index is calculated as the ratio of the
able prices gives the same result as if the variety is simply arithmetic (geometric) mean prices, the price in the price
omitted from the index calculation. However, by storing reference period needs to be adjusted by deflation of the new
the imputed price as an observation, it can be used with a average in March by the long-term index so that the March
reported price for index calculation in the subsequent month index level is maintained and the new sample does not affect
as previously demonstrated in Table 8.5A. Thus, the chained the long-term price change through March. If a new refer-
index is compiled by simply chaining the month-to-month ence price of variety D for January was imputed, different
186
Table 8.6 Replacing Varieties with No Overlapping Prices: Jevons and Carli Price Indices
January February March April May
Elementary Aggregate B Prices

Variety A 6.00 7.00 5.00
Variety B 3.00 2.00 4.00 5.00 6.00
Variety C 7.00 8.00 9.00 10.00 9.00
Variety D 9.00 8.00
Geometric Mean 5.01 4.82 5.65 7.66 7.56
Average of L-T Price Relatives 0.992 1.151 1.360 1.540
(a) No Imputations for Missing Prices (price indices calculated directly from monthly averages)
Jevons Index—The Ratio of Geometric Mean Prices = Geometric Mean of Price Relatives
Direct Index 100.0 96.1 112.6 152.9 150.8
Month-to-Month Change 0.961 1.171 1.357 0.986
Chained m/m Index 100.0 96.1 112.6 152.9 150.8
Carli Index—The Arithmetic Average of Price Relatives
Direct Index 100.0 99.2 115.1 136.0 154.0
Chained m/m Index 100.0 99.2 127.0 149.9 149.3
(B) Imputation for Missing Prices
Jevons Index—The Ratio of Geometric Mean Prices = Geometric Mean of Price Relatives
Impute the Price of Variety A in April Using the S-T Relative of Average Prices: 5.00 × [(5 × 10)/(4 × 9)]0.5 = 5.89
The April Average Price Is Derived as (5.89 × 5 × 10)1/3 = 6.65
The April Index Is Derived Using the January Geometric Average Price (6.65/5.01) = 1.327 × 100 = 132.7
A January Base Price Is Set for Variety D to Be Equal to the Base Price of Variety A Adjusted for the Quality Difference = Relative Price
Difference Between Varieties D and A in April: 6 × (9/5.89) = 9.17. By Taking the Geometric Average of the Base Prices of B, C, and D,
One Then Obtains the adjusted Average of 5.77. The May Average Price Is (6 × 9 × 8)1/3 = 7.56
The May Index Is Then Calculated as (7.56/5.77) × 100 = 130.9

Variety A 6.00 7.00 5.00 5.89
Variety B 3.00 2.00 4.00 5.00 6.00
Variety C 7.00 8.00 9.00 10.00 9.00
Variety D 9.17 9.00 8.00
Geometric Mean 5.01 4.82 5.65 6.65
Adjusted Average 5.77 7.66 7.56
Direct Index 100.0 96.1 112.6 132.7 130.9
The Month-to-Month Changes Are Calculated from the Geometric Mean of Price Changes of Varieties A, B, C from January through April.
The Monthly Change in May Is Calculated Using the Geometric Mean of Price Changes for Varieties B, C, D from April to May
Chained m/m Index 100.00 96.1 112.6 132.7 130.9
Carli Index—The Arithmetic Mean of Price Relatives
The April Price of Variety A Is Missing, and the Average S-T Price Relative in April Is Derived from Varieties B and C:
(5/4 + 10/9) × 0.5 = 1.181
Impute the Price of Variety A in April as 5.00 × 1.181 = 5.90 , So That L-T Relative Is (5.90/6) = 0.984
The Average L-T Relative in April Is (0.984 + 1.667 + 1.429)/3 = 1.360 × 100 =136.0, the April Index
A January Base Price Is Set for Variety D to Be Equal to the Base Price of Variety A Adjusted for the Quality Difference = Relative Price
Difference between Varieties D and A in April: 6 × (9/5.90) = 9.15. The L-T Relative for Variety D in May Is (8/9.15) = 0.8745. The May
Index Is (1/3 × (2.000 + 1.2857 + 0.8745)) = 138.67

Variety A 6.00 7.00 5.00 5.90
Variety B 3.00 2.00 4.00 5.00 6.00
Variety C 7.00 8.00 9.00 10.00 9.00
Variety D 9.15 9.00 8.00
Price Relatives
Variety A 1.167 0.833 0.984
Variety B 0.667 1.333 1.667 2.000
Variety C 1.143 1.286 1.429 1.285
Variety D 0.874
Direct Index 100.00 99.2 115.1 136.0 138.7
results would be obtained because the price changes are the Dutot (39 and 43 percent) versus the Jevons (33 and 33
implicitly weighted by the relative reference period prices percent).7
in the Dutot index, which is not the case for the Carli or
the Jevons indices. The April and May index change in the 7
The new sample starts in March as the price reference. The Dutot implicit
Dutot index is lower than the Jevons because the declines weights are 17.4 percent (4/23), 39.1 percent (9/23), and 43.5 percent
in price of varieties C and D have larger implicit weights in (10/23) percent, respectively, for varieties B, C, and D.
187
Table 8.7 Replacing Varieties with No Overlapping Prices: Dutot Index

Variety A 6.00 7.00 5.00
Variety B 3.00 2.00 4.00 5.00 6.00
Variety C 7.00 8.00 9.00 10.00 9.00
Variety D 9.00 8.00
Arithmetic Average 5.33 5.67 6.00 8.00 7.67
(a) No Imputations for Missing Prices
Dutot Index—The Ratio of Arithmetic Mean Prices
Direct Index 100.00 106.25 112.50 150.00 143.75
Chained m/m Index 100.00 106.25 112.50 150.00 143.75
(b) Imputation for Missing Prices
Impute the Price of Variety A in April Using the S-T Relative of Average Prices: 5.00 × (5 + 10)/(4 + 9) = 5.77
The April Average Price Is Derived as (5.77 + 5 + 10)/3 = 6.92
The April Index Is Derived Using the January Average Price (6.92/5.33) = 1.2981 × 100 = 129.81
A New Imputed Average Price Is Calculated for January by Taking the April Arithmetic Mean Price of Varieties B, C, and D (5+10+9)/3 = 8
and Deflating the Value Using the April L-T Price Change (8/1.2981) = 6.16
The May Index Is Then Calculated as (7.67/6.16) × 100 = 124.40
Variety A 6.00 7.00 5.00 5.77
Variety B 3.00 2.00 4.00 5.00 6.00
Variety C 7.00 8.00 9.00 10.00 9.00
Variety D
9.00 8.00
Arithmetic Mean 5.33 5.67 6.00 6.92
Adjusted Average 6.16 8.00 7.67
Direct Index 100.00 106.25 112.50 129.81 124.40
The Month-to-Month Changes Are Calculated from the Average Price for Varieties A, B, C from January through April. The Monthly
Change in May Is Calculated on the Average Price for Varieties B, C, D in April and May
Chained m/m Index 100.00 106.25 112.50 129.81 124.40
Table 8.8. Disappearing and Replacement Varieties with Overlapping Prices


Variety A 6.00 7.00 5.00
Variety B 3.00 2.00 4.00 5.00 6.00
Variety C 7.00 8.00 9.00 10.00 9.00
Variety D 10.00 9.00 8.00
Geometric Average Price A,B,C; (B,C,D) 5.01 4.82 5.65 (7.11) (7.66) (7.56)
Arithmetic Average Price A,B,C; (B,C,D) 5.33 5.67 6.00 (7.67) (8.00) (7.67)
Jevons Index—The Ratio of Geometric Mean Prices = Geometric Mean of Price Ratios
Chain the Monthly Indices Based on Matched Prices
Month-to-Month Change 1.0000 0.9615 1.1713 1.0774 0.9869
Chained m/m Index 100.00 96.15 112.62 121.33 119.75
For the Direct Index, a New Imputed Average Price Is Calculated for January by Taking the Average Price of Varieties B, C, and D in March
(4 × 9 × 10)1/3 = 7.11 and Deflating by the March L-T Index (1.1262) to Derive the Adjusted Base Price (6.31). This Calculation Maintains
the Level of the March Index
The Adjusted Base Price Is Used to Compile the April and May Indices
Direct Index 100.00 96.15 112.62 121.32 119.68
Chain the Monthly Indices Based on Matched Prices
Month-to-Month Change 1.0000 1.0625 1.0588 1.0435 0.9583
Chained m/m Index 100.00 106.25 112.50 117.39 112.50
For the Direct Index, a New Imputed Average Price Is Calculated for January by Taking Average Price of Varieties B, C, and D in March
(4 + 9 + 10)/3 = 7.67 and Deflating by the March L-T Relative (1.1250) to Derive the Adjusted Base Price (6.81). This Calculation Maintains
the Level of the March Index. This Adjusted Base Price Is Used to Compile the April and May Indices
Direct index 100.00 106.25 112.50 117.39 112.50
Carli Index—The Arithmetic Mean of Price Relatives
Variety A 1.0000 1.1667 0.8333
Variety B 1.0000 0.6667 1.3333 1.6667 2.0000
Variety C 1.0000 1.1429 1.2857 1.4286 1.2857
Variety D 1.0000 1.0357 0.9206
L-T relative 0.9921 1.1508 1.3770 1.4021
Average L-T Relative for Elementary Index in March Is (0.8333 + 1.333 + 1.2857)/3 = 1.1508 × 100 =115.08, the March index
Impute the Price of Variety D in January as 10.00 /1.1508 = 8.69 , Keeping the L-T Relative as 1.1508 So That the Introduction of Variety D
Does Not Affect the March Index Level. The New L-T Relatives for Variety D in April and May Are 1.3770 (9.00/8.69) and 0.9206 (8.00/8.69)
The Average L-T Relative for Varieties B, C, and D Are Used to Calculate the April and May Indices
Direct Index 100.00 99.21 115.08 137.70 140.21
188
8.76 If the index is calculated as a direct Carli, the Jan- Table 8.9 Calculation of a Weighted Elementary
uary base period price for variety D must be imputed by Index
dividing the price of variety D in March (10.00) by the long- Weight December January February Price Relative
term index change for March (1.1508). This deflation of Dec.–Feb.
the variety D price maintains the index level in March. The Variety A 0.80 7 7 9 1.2857
long-term relative for replacement variety D in April and Variety B 0.17 20 20 10 0.0500
May is calculated by dividing the prices by the estimated Variety C 0.03 28 28 12 0.4286
base price (8.69) of variety D in January. Weighted Arithmetic Mean of Price Relatives Index
((9/7) × 0.8 + (10/20) × 0.17 + (12/28) × 0.03) × 100 = 112.64
Weighted Geometric Mean of Price Relatives
Calculation of Elementary Price ((9/7)0.8 × (10/20)0.17 × (12/28)0.03) × 100 = 105.95
Indices Using Weights
8.77 The Jevons, Dutot, and Carli indices are all cal-
culated without the use of explicit weights. However, as
already mentioned, in certain cases weighting information calculated using a weighted arithmetic average of the price
may be available that could be exploited in the calculation observations:
of the elementary price indices. Weights within elementary
aggregates may be updated independently and possibly more  pt  pib ⋅ qib
often than the elementary aggregate weights themselves. 0:t
I EA = ∑wib ⋅  i0  , wib = (8.8)
8.78 Sources of weights include scanner data for  pi  ∑ pib ⋅ qib
selected divisions, such as food and beverages. Because
scanner data include quantities, the relative importance of where I EA
0:t
is the elementary aggregate price index
sampled varieties can be calculated. Weights can also be pi0 is the base price observed for variety i
developed by outlet or outlet type within an elementary wib is the weight for the variety in the weight refer-
aggregate. For example, for bread and bakery products, ence period
scanner data could provide data to develop weights for dif- 8.81 The geometric index is calculated using a weighted
ferent grocery stores. In some countries, the household bud- geometric average of the price observations:
get survey (HBS) includes a question asking respondents to
identify the type of outlet where an expenditure on a par-  pt  i
wb
ticular item was made. These data could be used to develop I 0:t
= ∏ i0  (8.9)
weights for the different outlet types identified. Another
ge
 p 
i
potential source of data for developing weights for outlets

or outlet type would be estimates of market shares obtained 8.82 Table 8.9 provides an example of calculations of
from business or trade groups and marketing firms. A spe- fixed-base elementary aggregate indices. The group consists
cial situation occurs in the case of tariff prices. A tariff is a of three varieties for which prices are collected monthly. The
list of prices for the purchase of a particular kind of good or expenditure shares are estimated to be 0.80, 0.17, and 0.03.
service under different terms and conditions. One example 8.83 One option is to calculate the index as the weighted
is electricity, where one price is charged during the day- arithmetic mean of the price relatives, which gives an index
time while a lower price is charged at night. Similarly, a of 112.64. The individual price changes are weighted accord-
telephone company may charge a lower price for a call at ing to their explicit weights, irrespective of the price levels.
the weekend than in the rest of the week. Another example The index may also be calculated as the weighted geometric
may be bus tickets sold at one price to ordinary passengers mean of the price relatives, which yields an index of 105.95.
and at lower prices to children or pensioners. In such cases,
one option, depending upon the availability of data, would Other Formulas for Elementary Price Indices
be to assign weights to the different tariffs or prices in 8.84 Another type of average is the harmonic mean. In
order to calculate the price index for the elementary aggre- the present context, there are two possible versions: either
gate. Another option would be to calculate a unit value, as the harmonic mean of price relatives or the ratio of har-
described in paragraph 8.85. However, changes in the tariff monic mean prices. The harmonic mean of price relatives
structure can be more difficult to capture. The treatment of is defined as
tariffs is further discussed in Chapter 11.
8.79 The increasing use of electronic points of sale 1
in many countries, in which both prices and quantities are 0:t
I HR = (8.10)
1 pi0
scanned as the purchases are made, means that valuable new
n
∑ pt
sources of information are increasingly available to NSOs. i
This could lead to significant changes in the ways in which

price data are collected and processed for CPI purposes. The The ratio of harmonic mean prices is defined as
treatment of scanner data is examined in Chapter 11.
8.80 If the weight reference period expenditure for all n
the individual varieties within an elementary aggregate, ∑p 0
or estimates thereof, were to be available, the elementary 0:t

I RH = i
(8.11)
n
price index could itself be calculated as a fixed-basket price ∑ pit
index, or as a geometric price index. The arithmetic index is
189
Formula 8.11, like the Dutot index, fails the commensurabil- Formulas Applicable to Scanner Data
ity test and would only be an acceptable possibility when 8.88 As noted at the beginning of this chapter, it
the varieties are all fairly homogeneous. Neither formula is preferable to introduce weighting information as it
appears to be used much in practice, perhaps because the becomes available rather than continuing to rely on simple
harmonic mean is not a familiar concept and would not unweighted indices such as Carli and Jevons. Advances in
be easy to explain to users. Nevertheless, at an aggregate technology, both in the retail outlets and in the computing
level, the widely used Paasche index is a weighted harmonic power available to NSOs, suggest that traditional elemen-
average. tary price indices may eventually be replaced by superla-
8.85 The ranking of the three common types of mean tive indices, at least for some elementary aggregates in some
is always arithmetic ≥ geometric ≥ harmonic. It is shown in countries. A superlative index is a type of index formula that
Chapter 6 of Consumer Price Index Theory that, in practice, can be expected to approximate the COLI. An index is said
the Carli index (the arithmetic mean of the price ratios) is to be exact when it equals the true COLI for consumers
likely to exceed the Jevons index (the geometric mean) by whose preferences can be represented by a particular func-
roughly the same amount that the Jevons exceeds the har- tional form. A superlative index is then defined as an index
monic mean. The harmonic mean of the price relatives has that is exact for a flexible functional form that can provide
the same kind of axiomatic properties as the Carli index, but a second-order approximation to other twice-differentiable
with opposite tendencies and biases. It fails the transitivity, functions around the same point. The Fisher, the Törnqvist,
time reversal, and price bouncing tests. and the Walsh price indices are examples of superlative indi-
8.86 In recent years, attention has focused on formulas ces. Superlative indices are generally symmetric indices.
that can take account of the substitution that may take place The methodology must be kept under review in the light of
within an elementary aggregate. As already explained, the the resources available.
Carli and the Jevons indices may be expected to approxi- 8.89 Scanner data obtained from electronic points of
mate a COLI if the cross-elasticities of substitution are close sale have become an increasingly important source of data
to 0 and 1, respectively, on average. A more flexible for- for CPI compilation. Their main advantage is that the num-
mula that allows for different elasticities of substitution is ber of price observations can be enormously increased and
the unweighted Lloyd–Moulton index: that both price and quantity information is available in real
time. There are, however, many practical considerations to
1
be taken into account, which are discussed in other chapters
 1−σ  1−σ
 1  Pi t   of this Manual, particularly in Chapter 10. To date, scanner
0:t
I LM =  ∑  0    (8.12)
 n  Pi   data has been used for selected components of the CPI, pri-
 marily for goods.
8.90 Access to the detailed and comprehensive quantity
where σ is the elasticity of substitution. The Carli and and expenditure information within an elementary aggre-
the Jevons indices can be viewed as special cases of the gate means that there are no constraints on the type of index
Lloyd–Moulton in which σ = 0 and σ = 1. The advantage number that may be employed. Not only Laspeyres and
of the Lloyd–Moulton formula is that σ is unrestricted. Paasche but superlative indices such as Fisher and Törnqvist
Provided a satisfactory estimate can be made of σ, the can be calculated. However, the frequent weight and price
resulting elementary price index is likely to approximate changes that are prevalent in the scanner data cause several
the underlying COLI. The Lloyd–Moulton index reduces problems with index estimation. Scanner data application
“substitution bias” when the objective is to estimate the and formulas are discussed in more detail in Chapter 10.
COLI. The difficulty is the need to estimate elasticities
of substitution, a task that will require substantial devel-
opment and maintenance work. The formula is described The Calculation of Higher-Level
in more detail in Chapter 4 of Consumer Price Index Indices
Theory.
8.91 As shown in Figure 8.1, the elementary indices are
Unit Value Indices the starting point (building blocks) for calculating the CPI.
8.87 The unit value index is simple in form. The unit These indices are then aggregated to successively higher
value in each period is calculated by dividing total expen- levels (for example, city, region, class, or group), to derive
diture on some product by the related total quantity. The the national all-items index. These higher-level indices are
quantities must be strictly additive in an economic sense, derived by aggregations using weights that are generally
which implies that they should relate to a single homoge- derived from an HBS, although other sources are presented
neous product. The unit value index is then defined as the in Chapter 3. The aggregation formulas can take several
ratio of unit values in the current period to that in the refer- forms such as arithmetic and geometric depending on the
ence period. It is not a price index as normally understood, target index. Fixed-basket indices tend to use arithmetic
as it is essentially a measure of the change in the average aggregations while for the superlative indices, the Törn-
price of a single product when that product is sold at differ- qvist index uses geometric aggregations, the Walsh uses an
ent prices to different consumers, perhaps at different times arithmetic aggregation, and the Fisher is the geometric aver-
within the same period. Unit values, and unit value indices, age of two arithmetic average price indices (Paasche and
should not be calculated for sets of heterogeneous products. Laspeyres).
Unit value methods are discussed in more detail in Chap- 8.92 An NSO must decide on the target index at
ters 10 and 11. which to aim. The target index takes into consideration the
190
observable price, quantity, and expenditure data that can be superlative index. Even when the target index is the Laspey-
used to calculate the index in practice. The advantages of res, there may be a considerable gap between what is actu-
having a target index are the following: ally calculated and what the NSO considers to be its target.
Chapters 2–4 of Consumer Price Index Theory present the
• Providing reference and guidance for compilation of the
alternatives from a theoretical point of view. It is also shown
CPI
that some combination of an arithmetic index such as the
• Ability to quantify the size of any bias, the differences Laspeyres index and a geometric index such as geometric
between what is actually measured and what should be Laspeyres index may approximate the superlative Fisher
measured and Törnqvist indices.8 Such an approach may be the ideal
• Use in identifying and making improvements to the CPI solution when both of these indices or their proxies can be
• Being able to document the sources and methods used produced in real time. What many NSOs tend to do in prac-
in the CPI and how they approximate the target index tice varies; many use the Laspeyres index as their target, but
compilation a few have chosen the Fisher, Törnqvist, or Walsh as their
targets.
8.93 NSOs must consider what kind of index they
would choose to calculate in the ideal hypothetical situa-
tion in which they had complete information about prices Reference Periods
and quantities in both time periods compared. What is the 8.97 It is useful to recall that three kinds of reference
purpose of the index? Should it measure both inflation and periods may be distinguished:
maintaining economic welfare? If so, the best CPI measure • Weight reference period. The period covered by the
could be a COLI, and a superlative index such as a Fisher, expenditure data used to calculate the weights. Usually,
Törnqvist, or Walsh would have to serve as the theoretical the weight reference period is a period of 12 consecutive
target. A superlative index may be expected to approximate months.
the underlying COLI.
• Price reference period. The period whose prices are used
8.94 In many countries the purpose of the CPI is to mea-
as denominators in the index calculation.
sure inflation and to adjust wages, income, or social pay-
ments for changes in a fixed basket of goods and services, as • Index reference period. The period for which the index is
discussed in Chapter 2. Thus, the concept of a basket index set to 100.
might be preferred, sometimes also referred to as a cost of 8.98 The three periods are generally different. For
goods index. A basket index is one that measures the change example, a CPI might have 2016 as the weight reference
in the total value of a given basket of goods and services year, December 2018 as the price reference month, and
between two time periods. This general category of index is the year 2015 as the index reference period. The weights
described here as a Lowe index (see Chapter 2 of Consumer typically refer to a whole year, or even two or three years,
Price Index Theory). It should be noted that, in general, whereas the periods for which prices are compared are typi-
there is no necessity for the basket to be the actual basket cally months, quarters, or a year. The weights are usually
in one or other of the two periods compared. If the target estimated on the basis of an expenditure survey that was
index is to be a basket index, the preferred basket might be conducted some time before the price reference period. For
one that attaches equal importance to the baskets in both these reasons, the weight reference period and the price ref-
periods; for example, the Walsh index. Thus, the same index erence period are invariably different periods in practice.
may emerge as the preferred target in both the cost of goods The price reference period should immediately precede the
and the cost of living approaches. introduction of the updated index. For example, if January
8.95 In Chapters 2–4 of Consumer Price Index Theory 2019 is the first month for the updated CPI, the price ref-
the superlative indices Fisher, Törnqvist, and Walsh show erence period would be December 2018 or the year 2018,
up as being “best” in all the approaches to index number depending on whether December or the full year is used as
theory. These three indices, and the Marshall–Edgeworth the price reference period.
price index, while not superlative, give very similar results 8.99 The index reference period is often a year, but it
so that for any practical reason it will not make any differ- could be a month or some other period. An index series may
ence which one is chosen as the preferred target index. In also be re-referenced to another period by simply dividing
practice, an NSO may prefer to designate a basket index that the series by the value of the index in that period, without
uses the actual basket in the earlier of the two periods as changing the rate of change of the index. For the CPI, the
its target index on grounds of simplicity and practicality. In expression “base period” can mean any of the three ref-
other words, the Laspeyres index may be the preferred target erence periods and is ambiguous. The expression “base
index. Similarly, if the quantities in both periods are avail- period” should only be used when it is absolutely clear in
able, the Walsh index, which is also a fixed-basket index, context exactly which period is referred to.
might be the target. The target indices, whether Fisher,
Törnqvist, or Walsh, can be calculated retrospectively as
additional expenditure estimates become available. The
retrospective indices can then be used to assess the perfor-
mance of the CPI and estimate the potential bias from the
target index. 8
Janice Lent, and Alan H. Dorfman. 2009. “Using a Weighted Average of
8.96 The theoretical target index is a matter of choice. Jevons and Laspeyres Indexes to Approximate a Superlative Index.” Jour-
In practice, it is likely to be either a Laspeyres or some nal of Official Statistics 25 (1): 129–49.
191
Higher-Level Price Indices as Weighted weights are used in a harmonic mean formula. Whether the
Averages of Elementary Price Indices Lowe or Young index should be used depends on how much
price change occurs between the weight and price reference
8.100 A higher-level index is an index for some expen-
period as well as the target index. This is discussed in more
diture aggregate above the level of an elementary aggregate,
detail in Chapter 9.
including the overall CPI. The inputs into the calculation of
8.104 As noted, a more frequent version of equation
the higher-level indices are the following:
8.11 would be that of a Young or a Lowe index, where the
• The elementary aggregate price indices weights are derived from an earlier period, b < 0. This is
• The expenditure shares of the elementary aggregates often the case because it may take a year or longer to com-
pile the expenditure weights from the HBS before they are
8.101 The higher-level indices are calculated simply available for use in the CPI. The Young index is
as weighted averages of the elementary price indices. The
weights typically remain fixed for a sequence of at least  pt  pib ⋅ qib
12 months. Some countries revise the weights at the begin- IY0:t = ∑wib ⋅  i0  , wib = (8.14)
ning of each year to try to approximate as closely as possible
 pi  ∑ pib ⋅ qib
to current consumption patterns, but many countries continue
to use the same weights for several years and the weights 8.105 The Young index is general in the sense that the
may be changed only every five years or so. The use of fixed shares are not restricted to refer to any particular period but
weights has a considerable practical advantage that the index may refer to any period or an average of different periods,
can make repeated use of the same weights. This saves both for example. The Young index is a fixed-weight index where
time and resources. Revising the weights can be both time- the focus is that the weights should be as representative as
consuming and costly, especially if it requires a new HBS to possible for the average value shares of the period covered
be carried out. The disadvantage is that as the weights become by the index. A fixed-weight index is not necessarily a fixed-
older, they are less representative of consumer purchasing basket index (that is, it does not necessarily measure the
patterns and usually result in substitution bias in the index. change in the value of an actual basket such as the Lowe
index). The Young index measures the development in the
Examples of Fixed-Basket Price Indices cost of a period 0 set of purchases with period b value pro-
8.102 When describing their calculation methods, some portions between the expenditure components. This does
NSOs note that the Laspeyres price index is used for the not correspond to the changing value of any actual basket,
calculation of higher-level aggregate indices; however, it is unless the expenditure shares have remained unchanged
not possible to calculate a Laspeyres index in practice. The from b to 0. In the special case where b equals 0, it reduces
Laspeyres price index is defined as to the Laspeyres.
8.106 In the case of the Lowe index the weights from
period b are updated for price change between b and 0. The
∑p t
⋅ qi0  t
0  pi  pi0 ⋅ qi0
∑ i  p 0  i Lowe index is
i
I L0:t = = w ⋅   , w 0
= (8.13)
∑p 0 0
i ⋅ qi i ∑ pi0 ⋅ qi0
 pt  pi0 ⋅ qib
where wi0 indicates the expenditure shares for the indi- 0:t
I Lo = ∑wib:0 ⋅  i0  , wib:0 = (8.15)
vidual varieties in the reference period. As the quantities
 pi  ∑ pi0 ⋅ qib
are often unknown, the index usually will have to be cal-
culated by weighting together the individual price relatives 8.107 By price updating, the weights are aligned to the
by their expenditure share in the price reference period, wi0. same reference period as the prices. If the NSO decides to
The expenditure shares are usually derived by consumption price update the weights, the resulting index will be a Lowe
expenditure estimates from an HBS. The available weight- index. The Lowe index is a fixed-basket index, which from
ing data may refer to an earlier period than the price refer- period to period measures the value of the same (annual)
ence period but may still provide a reasonable estimate. basket of goods and services. Because it uses the fixed bas-
8.103 While NSOs often refer to the Laspeyres formula ket of an earlier period, the Lowe index is sometimes loosely
for compiling their CPI, this is not the case. In fact, the most described as a “Laspeyres-type” index, but this description
commonly used formulas for compiling the CPI are either is unwarranted. A true Laspeyres index requires the basket
the Lowe or Young formulas.9 If the weights are derived to be that purchased in the price reference month, whereas
from expenditure in period 0, the price reference period, as in most CPIs the basket refers to a period different from the
in equation 8.11, the index is a Laspeyres price index. If the price reference month. When the weights are annual and the
weights from an early weight reference period b (that is, b prices are monthly, it is not possible, even retrospectively, to
< 0) are used directly in the index as expenditure shares in calculate a monthly Laspeyres price index.
period 0, the index is known as a Young index. If the weights
are updated for price change from b to 0, which keeps the Typical Calculation Methods for Higher-
quantity shares fixed, the index is called a Lowe index. This Level Indices
is similar to the attribution given to the more noted Laspeyres 8.108 The most common method for calculating higher-
index where the b = 0 and the Paasche index where period t level indices in the CPI is not done using individual prices
or quantities. Instead, a higher-level index is calculated
9
The Young and Lowe indices are named for the nineteenth-century index by averaging the elementary price indices by their prede-
number pioneers who advocated these indices. termined weights. Using weights instead of quantities,
192
equation (8.11) can be expressed as the following formula expensive new one. Moreover, when the weights are known
for aggregating price indices: in advance of the price collection, the index can be calcu-
lated immediately after the prices have been collected and
I L0:t = ∑ wbj I 0j :t , ∑ wbj = 1 (8.16) processed.
8.115 The longer the same weights are used, however,
the less representative they become of current consump-
where IL0:t denotes the all-items CPI, or any higher-level tion patterns, especially in periods of rapid technological
index, from period 0 to t, and wjb is the weight attached to each changes when new kinds of goods and services are continu-
of the elementary price indices where the weights sum to 1. ally appearing on the market and old ones disappearing.
Ij0:t is the corresponding elementary price index. The elemen- This may undermine the credibility of an index that purports
tary indices are identified by the subscript j. to measure the rate of change in the total cost of a basket of
8.109 The weight reference period (b) usually will refer goods and services typically consumed by households. Such
to a year, or an average of several years, that precedes the a basket needs to be representative not only of the house-
price refence period (0). If the weights are used as they holds covered by the index but also of expenditure patterns
stand, without price-updating, the resulting index will cor- at the time the price changes occur.
respond to a Young index. If the weights are price-updated 8.116 Similarly, if the objective is to compile a COLI,
from period b to period 0, the resulting index will corre- the continuing use of the same fixed basket is likely to
spond to a Lowe price index. become increasingly unsatisfactory the longer the same bas-
8.110 Provided the elementary aggregate indices are cal- ket is used. The longer the same basket is used, the greater
culated using a transitive formula such as Jevons or Dutot, the upward bias in the index is likely to become. It is well
but not Carli, and that there are no new or disappearing vari- known that the Laspeyres index has an upward bias com-
eties from period 0 to t, equation 8.16 is equivalent to pared with a COLI. However, a Lowe index between peri-
ods 0 and t with weights from an earlier period b will tend
0:t
I MLo = ∑ wbj :0 I 0j :t −1 I tj−1:t , ∑ wbj :0 = 1 (8.17) to exceed the Laspeyres between 0 and t by an amount that
increases the further back in time period b is (see Chapter 2
8.111 The difference is that equation 8.16 is based on of Consumer Price Index Theory).
the direct elementary indices from 0 to t, while (8.17) uses 8.117 There are several possible ways of minimizing or
the chained elementary indices. I tj−1:t is the short-term price avoiding the potential biases from the use of fixed-weight
index for the elementary aggregate between t − 1 and t. A indices. These are outlined in the following sections.
CPI calculated according to (8.17) in this manual is referred
to as a modified Lowe index if the weights are price-updated Annual Chaining
from the weight reference period to the price reference 8.118 One way in which to minimize the potential
period. If the weights are used as they stand, the index is biases from the use of fixed-weight indices is obviously to
referred to as a modified Young index. keep the weights and the index reference period as up to date
8.112 The recommended procedure is to use the short- as possible by frequent rebasing and chaining. Quite a few
term price index formulation, instead of basing the aggrega- countries have adopted this strategy and revise their weights
tion on long-term elementary price indices compiled in a annually. In any case, as noted earlier, it would be impos-
single stage. sible to deal with the changing universe of products with-
8.113 There are two ways that the modified index can out some chaining of the price series within the elementary
be compiled. First, chaining the monthly elementary indi- aggregates, even if the weights attached to the elementary
ces into long-term price indices and compiling the higher- aggregates remain fixed. Annual chaining eliminates the
level indices by aggregating the elementary indices using need to choose a weight reference period, as the weight ref-
the expenditure shares as weights. Alternatively, the modi- erence period is always the previous year (t − 1), or possibly
fied index can be compiled by each month multiplying the the preceding year (t − 2).
expenditure weights with the elementary indices to form the 8.119 Annual chaining with current weights. When the
long-term weighted relatives up to period t – 1. These can weights are changed annually, it is possible to replace the
then be multiplied with the elementary price indices from original weights based on the previous year, or years, by
t – 1 to t and the resulting series are aggregated into higher- those of the current year, if the index is revised retrospec-
level price indices. The two methods give identical results tively as soon as information on the current year’s expendi-
and it is up to countries to decide which to apply. tures becomes available. The long-term movements in the
CPI are then based on the revised series. This method could
provide unbiased or less-biased results.
Some Alternatives to Fixed-Weight
Indices Other Index Formulas
8.114 Monthly CPIs are typically arithmetic weighted 8.120 When the weights are revised less frequently, say
averages of the price indices for the elementary aggregates, every five years, another possibility would be to use a differ-
in which the weights are kept fixed over a number of periods ent index formula for the higher-level indices instead of an
that may range from 12 months to many years (but no more arithmetic average of the elementary price indices. An alter-
than five). The repeated use of the same weights relating native method for aggregating elementary indices would be
to some past period b simplifies calculation procedures and geometric aggregation. Geometric aggregation is similar to
reduces data collection requirements. It is also cheaper to arithmetic aggregation but involves weighting each elemen-
keep using the results from an old HBS than to conduct an tary index by the exponent of its share weight.
193
8.121 The geometric version of the Laspeyres index is or Lowe > Laspeyres > Fisher > geometric Lowe or geo-
defined as: metric Young > geometric Laspeyres > Paasche. The Törn-
wi0 qvist and Walsh price indices provide the same results as
∏ ( pit )
w0
 pt  i
I = ∏  i0  =
0:t
, ∑ wi0 = 1 (8.18) the Fisher, so that they will place in the same position as the
GL
 pi  0
Fisher. There are several factors to consider in the choice of
∏ ( pi )
0 wi
the aggregation method:
where the weights, wi0, are again the expenditure shares in the • The country target index. The target index for the CPI
price reference period. When the weights are all equal, equa- is one factor to consider in deciding on the aggregation
tion (8.18) reduces to the Jevons index. If the expenditure method. The CPI can be compiled as a fixed-basket cost
shares do not change much between the weight reference of goods index. The target index could be a Laspeyres
period and the current period, then the geometric Laspey- price index in which the quantity of items purchased is
res index approximates a Törnqvist index. Using equation assumed to be fixed. Also, it could be a Walsh or Mar-
(8.18) the Geometric Young index, IGY 0:t
, can be derived using shall–Edgeworth index where the fixed weights are aver-
the weights wjb, and the Geometric Lowe index, IGLo 0:t
, can be ages of the base and current periods. This is a traditional
derived using the weights wjb:0. approach for the CPI and the traditional arithmetic aggre-
8.122 The geometric version of the Young index is gation method would be used. The purpose of the index
defined as: might be a COLI with the Fisher or Törnqvist price index
wbj
as the target index. In this case, a geometric aggregation
0:t
I GY = ∏ ( I 0j :t ) , ∑ wbj = 1 (8.19) could be used. The Walsh index, which uses arithmetic
aggregation, represents an equally good estimate of the
8.123 The geometric version of the Lowe index is COLI.
defined as: • Timeliness and availability of source weight data.
Another factor in the aggregation formula decision is the
0:t
I GLo
wbj :0
= ∏ ( I 0j :t ) , ∑ wbj :0 = 1 (8.20) timeliness of the expenditure weights. In most cases, the
weights for the current period will take several months
to become available and for the most part may only be
8.124 Another form of aggregation that yields the same available following an HBS. If the HBS or other source is
result as equation (8.18) is to convert the elementary indi- conducted on an ongoing basis, a superlative index such
ces to natural logarithms and use linear weighting of the as Fisher, Törnqvist, or Walsh can be calculated on a time-
logarithms. In this case, the result of the aggregation must lier basis, particularly if the CPI can be revised as the new
be converted from natural logarithm to a real number (the weights become available.
antilog or exponential function). This formula is most suited
• Elasticity of substitution. If the cross-elasticity of substi-
for compilation purposes using spreadsheets or other similar
software. tution is approximately 1, it implies that the expenditure
shares are not changing as the relative change in prices
is offset by the relative change in quantities purchased.
0:t
I GY = exp  ∑ wbj ln ( I 0j :t ) (8.21)
  In such a case where the expenditure shares remain
unchanged, a geometric Young index would provide a
Again, note that if the weight reference period refers to close approximation to a superlative index and geometric
period b, the index is a geometric Young index; if the refer- aggregation would be justified. If, on the other hand, the
ence period is 0, the index is a geometric Laspeyres index, cross-elasticity of substitution is close to zero, it implies
and if the reference period is the average of periods 0 and t, that there is no change in quantities purchased as rela-
it is a Törnqvist index. Recent empirical research discussed tive prices change. In such a case, the Laspeyres, Lowe,
in Chapter 2 of Consumer Price Index Theory has indicated or Young price indices that assume quantities (or shares)
that taking a geometric-average of an upward biased fixed- remain fixed would be appropriate and arithmetic aggre-
weight arithmetic index and a downward biased fixed-weight gation would be justified.
geometric index may closely approximate the Fisher index.
The reason for this close fit is that the possible upward bias • Consistency in aggregation. To maintain consistency in
in the arithmetic index is offset by a possible downward bias aggregation, the same type of formula would be used
in the geometric index. throughout the aggregation process. Thus, if a Dutot or
direct Carli index is used at the elementary level, then the
Arithmetic versus Geometric Aggregation for arithmetic aggregation of indices should occur at higher
Higher-Level Indices levels. If a Jevons index is used at the elementary level,
8.125 Given that both arithmetic and geometric aggre- then the corresponding higher-level indices would use
gation can be used for compiling higher-level indices in the geometric aggregation. This criterion should not be the
CPI, the question arises as to which is the most appropriate. only one used in determining the method of aggregation.
An index using arithmetic aggregation will normally pro- • Public understanding of the different averaging meth-
duce an index level that is greater than one using geometric ods. For decades the traditional definition for the CPI
aggregation for the same data points. The exception is when has been a fixed-basket index of constant quality. Pub-
all prices change at the same rate, so both will give the same lic perception has been directed to understanding the
result. Chapter 2 of Consumer Price Index Theory suggests fixed-basket concept where a historical basket is priced
the following ordering in the levels of price indices: Young at today’s prices. This also involves the understanding
194
of deriving higher-level indices as weighted arithmetic 8.129 Assume that each link runs from December to
averages of component indices. This concept is widely December. The long-term index for month m of year Y with
understood by the public. The use of geometric aggrega- December of year 0 as index reference period is then calcu-
tions and superlative indices has not been widely pre- lated using the formula:
sented and discussed with the public and thus is not well
understood. The result has been that most NSOs pro- Y −1 
I Dec 0:mY =  ∏ I Decy−1:Decy  I DecY −1:mY (8.22)
duce a fixed-basket index using arithmetic aggregation,  y =1 
although a few have begun producing a superlative ver-
sion of the CPI, but often with a lag as the new weight
= I Dec 0:Dec1 I Dec1:Dec 2 ...I Decy−2:DecY −1 I DecY −1:mY
data become available.
8.126 Which of these factors are the most important 8.130 In the actual practice of the European NSO, a fac-
must be decided by the NSOs. To the extent that traditional tor scaling the index from December year 0 to the average
fixed-basket indices are being produced and the public has a of year 0 is multiplied onto the right-hand side of equation
good understanding of these concepts, then NSOs will tend 8.22 to have a full year as the reference period. The long-
to stick with the traditional, and most commonly used, arith- term movement of the index depends on the long-term links
metic aggregation. As NSOs move to implementing one of only, as the short-term links are successively replaced by
the superlative indices as their target index, then geometric their long-term counterparts. For example, let the short-term
aggregation could become more prevalent and public educa- indices for January to December 2018 be calculated as
tion about the target index and the methods for producing it
should become a priority. I Dec2018:m 2019 = ∑w j
2018(Dec2018) Dec2018:m 2019
I (8.23)
j
Retrospective Superlative Indices

where w j ( ) are the weights from 2018 price updated
18 Dec18
8.127 It is possible to calculate a superlative price index
retrospectively. Superlative indices, such as Fisher, Törn- to December 2018, and PjDec18:m19 is the price index from
qvist, and Walsh, treat both periods symmetrically and require December 2018 to month 1 of 2019. In subsequent months
expenditure data for both periods. Although the CPI may (2, 3, …, 12), the price index is calculated with December
have to be a Lowe or Young index when it is first published, 2018 as the base. When weights for 2019 become available,
it may be possible to estimate a superlative index later when this is replaced by the long-term link:
much more information becomes available about consum-
ers’ expenditure period by period. In practice, currently one I Dec2018:Dec 2019 = ∑w j
2019(Dec2018) Dec2018:Dec2019
I (8.24)
country publishes a Walsh index, while another publishes a j
Törnqvist index. The publication of revised or supplementary

where w j ( ) are the weights from 2019 price backdated to
19 Dec18
CPIs raises matters of statistical policy, although users readily
accept revisions in other fields of economic statistics. More- December 2018. The same set of weights from 2019 price
over, users are already confronted with more than one CPI in updated to December 2019 is used in the new short-term
the European Union where the harmonised index for Euro- link for 2020:
pean Union purposes may differ from the national CPI. Thus,
the publication of supplementary indices that throw light on I Dec2019:m 2020 = ∑w j
2019(Dec2019) Dec2019:m 2020
I (8.25)
the properties of the main index and which may be of consid-
j
erable interest to some users seems justified and acceptable.

When weights for 2020 become available, this is replaced
Use of Long-Term and Short-Term Links to by the long-term link:
Calculate the CPI
8.128 Consider a long-term chain index in which the I Dec2019:Dec 2020 = ∑w j
2020(Dec2019) Dec2019:Dec2010
I
j (8.26)
weights are changed annually. The Walsh index requires
weights from the previous and current years. In any given
8.131 Using this method, the movement of the long-
year, the current monthly indices are first calculated using the
latest set of available weights, which cannot be those of the term index is determined by contemporaneous weights. The
current year. However, when the weights for the year in ques- method is conceptually attractive because the weights that
tion become available subsequently, the monthly indices can are most relevant for most users are those based on con-
then be recalculated based on the weights for the current year. sumption patterns at the time the price changes actually
The resulting series can then be used in the long-term chain take place. The method takes the process of chaining to
index, rather than the original indices first published. Thus, the its logical conclusion, at least assuming the indices are not
movements of the long-term chain index from, say, any one chained more frequently than once a year. As the method
December to the following December are based on weights uses weights that are continually revised to ensure that they
of that same year, the weights being changed each December. are representative of current consumer behavior, the result-
This method has been developed by a European NSO.10 ing index also largely avoids the substitution bias that occurs
when the weights are based on the consumption patterns of
10 some period in the past. The method may therefore appeal to
P. Bäckström, and M. Sammar. 2012. “The Use of Superlative Index
Links in the Swedish CPI.” Paper presented at the Meeting of the Group of NSOs whose objective is to estimate a COLI. This method
Experts on CPI, Geneva. https://www.unece.org/index.php. also provides better deflators for national accounts.
195
8.132 A North American NSO11 publishes a chained where I N0:t is the national all-items index from the period 0 to t
index using the Törnqvist price index and must use short-
term links to calculate the current month index. The weights I I0:t is national index for item I from period 0 to t
for the current period are lagged by about a year so that the wIb is the national weight for item I in the weight refer-
current calculations are made using a different formula. In ence period (sum of all area weights for that item)
this case, the short-term link is estimated using the Lloyd– n is the number of items in the CPI
Moulton price index from equation 8.10, where the estimate
of the elasticity of substitution is based on historical patterns. Numerical Examples
8.133 Finally, it may be noted that these methods involve 8.137 Table 8.10 illustrates the calculation of higher-level
some revision of the index first published. In some countries, indices using arithmetic aggregation where the weight and the
there is opposition to revising a CPI once it has been first pub- price reference periods are identical (that is, b = 0). The index
lished, although it is standard practice for other economic sta- consists of five elementary aggregate indices and two interme-
tistics, including the national accounts, to be revised as more diate higher-level indices, G and H. The overall index and the
information and more up-to-date information become available. higher-level indices are all calculated using (8.29). Thus, for
example, the overall index for April can be calculated from the
two intermediate higher-level indices for April as
Calculation of Geographic and National
Indices I Jan:Apr = (0.6×103.91) + (0.4×101.79) = 103.06
8.134 CPIs are often calculated for individual geo-
graphic areas within a country and then aggregated to pro- or directly from the five elementary indices as
vide a national index based on the price movements in the
individual areas. The aggregation approach is the same
where elementary aggregates are combined using weights I Jan:Apr = (0.2×108.75) + (0.25×100) + (0.15×104)
for each item index in the geographic area to derive the all- + (0.1×107.14) + (0.3×100) = 103.06
items CPI for the area. The elementary item indices are then
aggregated using their area weights to derive the national 8.138 Table 8.10 illustrates the calculation of higher-
item index. The formula for aggregation of items across level indices using geometric aggregation where the weight
areas to derive a national item index is and the price reference periods are identical (that is, b = 0).
k k The index consists of the same five elementary price indi-
I I0:t = ∑wbj , a ( I 0j ,:ta ) / ∑wbj , a (8.27) ces and two intermediate higher-level indices, G and H. The
a =1 a =1 overall index and the higher-level indices are all calculated
using (8.29). Thus, for example, the overall index for April
where I I0:t is national index for item I from period 0 to t can be calculated from the two intermediate higher-level
I 0j ,:ta is the area index for item j in area a from the period indices for April as
0 to t
wbj , a is the weight for item j in area a from the weight 0.6
reference period b I (Jan:Apr) = [(103.85) + (101.74)0.4 ] = 103.00
k is the number of areas in the CPI
8.135 The national all-items index can be compiled by
the aggregation of items across areas using their area weights: or directly from the five elementary indices as
n k n k
I N0:t = ∑∑wbj , a ( I i0,:at ) / ∑∑wbj , a (8.28) (108.75)0.2 + (100)0.25 + (104)0.15 
 
j =1 a =1 j =1 a =1 I Jan:Apr =  
+ (107.14)0.1 + (100)0.3 
 
where I is the national all-items index from the period 0 to t
0:t
N = 103.00
I is the area index for item j in area a from the period
0:t
j ,a
0 to t
8.139 The calculation of geographic area indices is
wbj , a is the weight for item j in area a from the weight
similar to those for the national index. The items and their
reference period b
weights should be derived for an HBS that covers each area.
n is the number of items in the CPI
If the HBS sample does not support independent estimates
k is the number of areas in the CPI
for each area, NSOs will often use the national item shares
8.136 The same result is obtained if the national item
in the areas and collect price information for varieties in the
indices are aggregated using the national item weights:
areas. Such an approach assumes that there are no significant
n n
differences in the purchasing patterns among areas. This is
I N0:t = ∑wIb ( I I0:t ) / ∑wIb (8.29) a second-best solution as producing independent consump-
I =1 I =1
tion estimates for each geographic area is preferred. The
assumption of similar purchasing patterns across areas is
often flawed. Usually, capital cities have quite different pur-
11
David M. Friedman. 2016. “A New Estimation System for the US CPI—
chasing patterns from those of regional centers.
Capabilities and Impacts.” Presentation at the Meeting of the Group of 8.140 Another issue that arises when consumption
Experts on CPI, Geneva. http://www.unece.org/index.php. expenditure weights are not available for the geographic areas
196
Table 8.10 Aggregation of Elementary Price Indices

Weight January February March April May June
Month-to-Month Elementary Price Indices

EA A 100.00 102.50 104.88 101.16 101.15 100.00
EA B 100.00 100.00 91.67 109.09 101.67 108.20
EA C 100.00 104.00 96.15 104.00 101.92 103.78
EA D 100.00 92.86 107.69 107.14 100.00 102.67
EA E 100.00 101.67 100.00 98.36 103.33 106.45
Direct or Chained Monthly Elementary Price Indices with January = 100
EA A 0.20 100.00 102.50 107.50 108.75 110.00 110.00
EA B 0.25 100.00 100.00 91.67 100.00 101.67 110.00
EA C 0.15 100.00 104.00 100.00 104.00 106.00 110.00
EA D 0.10 100.00 92.86 100.00 107.14 107.14 110.00
EA E 0.30 100.00 101.67 101.67 100.00 103.33 110.00
Total 100.00 100.89 99.92 103.06 105.03 110.00
Higher-Level Indices (arithmetic)
G=A+B+C 0.6 100.00 101.83 99.03 103.91 105.52 110.00
H=D+E 0.4 100.00 99.47 101.25 101.79 104.28 110.00
Total 100.00 100.89 99.92 103.06 105.03 110.00
Higher-Level Indices (geometric)
G=A+B+C 0.6 100.00 101.82 98.79 103.85 105.47 110.00
H=D+E 0.4 100.00 99.39 101.25 101.74 104.27 110.00
Total 100.00 100.84 99.77 103.00 104.99 110.00
EA = Elementary aggregate
Table 8.11 Aggregation of Elementary Price Indices (Arithmetic) across Locations

Weight January February March April May June
Area 1
EA A 0.12 100.00 102.50 107.50 108.75 110.00 110.00
EA B 0.15 100.00 100.00 91.67 100.00 101.67 110.00
EA C 0.09 100.00 104.00 100.00 104.00 106.00 110.00
EA D 0.06 100.00 92.86 100.00 107.14 107.14 110.00
EA E 0.18 100.00 101.67 101.67 100.00 103.33 110.00
All Items 0.60 100.00 100.89 99.92 103.06 105.03 110.00
Area 2
EA A 0.08 100.00 103.53 108.04 108.21 110.82 110.55
EA B 0.10 100.00 101.00 92.13 99.50 102.43 102.17
EA C 0.06 100.00 105.04 100.50 103.48 106.79 106.53
EA D 0.04 100.00 93.79 100.50 106.61 107.94 107.67
EA E 0.12 100.00 102.69 102.18 99.50 104.11 103.85
All Items 0.40 100.00 101.90 100.42 102.55 105.82 105.55
National
EA A 0.20 100.00 102.91 107.72 108.53 110.33 110.22
EA B 0.25 100.00 100.40 91.85 99.80 101.97 106.87
EA C 0.15 100.00 104.42 100.20 103.79 106.32 108.61
EA D 0.10 100.00 93.23 100.20 106.93 107.46 109.07
EA E 0.30 100.00 102.08 101.87 99.80 103.64 107.54
All Items 1.00 100.00 101.29 100.12 102.86 105.34 108.22
EA = Elementary aggregate
is which weights should be used. Often there is a tendency to indices in two geographic areas. The area all-items indices
use weights based on population. Again, population weights are calculated using equation 8.16 in which the weights
have a potential bias because one assumes that they are rep- are the items’ share within the area. The national-level
resentative of distribution of consumption expenditure across item indices are all calculated using equation 8.20. The
areas, which is frequently not the case. The second-best national all-items index can be calculated using either
solution would be to use area income estimates that should equation 8.28 or 8.29. Thus, for example, the area 2 index
have a closer approximation to consumption than does the for April is calculated from the five item-level indices for
population. April as
8.141 Table 8.11 illustrates the calculation of higher-
level indices using arithmetic aggregation where the I aJan:Apr = [(0.08×108.21) + (0.10×99.50) + (0.06×103.48)
weight and the price reference periods are identical (that
is, b = 0). The index consists of five elementary aggregate + (0.04×1006.61) + (0.12×99.50)] / 0.4 = 102.55
197
The national item index for item A is calculated from the average prices or the average of price relatives. This is not the
two area indices for item A: case when using the arithmetic mean.
• The Dutot index should be used only for homogeneous
I IJan:Apr = (0.12×108.75) + (0.08×108.21) / 0.20 = 108.53 elementary aggregates since the price relatives are implic-
itly weighted by the price level in the reference period.
The national all-items index is calculated using equation • The permanently missing goods or services should be
8.15 as replaced, and appropriate methods used to adjust for any
changes because of the difference in quality. This ensures
that the index remains representative and reflects only pure
I Jan:Apr = [(0.2×108.53) + (0.25×99.8) + (0.15×103.79) price change, not changes due to differences in quality.
+ (0.1×106.933) + (0.3×99.8)] / 1.0 = 102.86 • The chained Carli formula for elementary aggregates
(arithmetic mean of price relatives) should not be used.
Key Recommendations The chained Carli has a well-known upward bias.
• It is recommended to calculate the elementary price indi-
• Elementary aggregates should be constructed to include ces by chaining the short-term (month-to-month) price
groups of relatively homogeneous goods and services (that indices. This short-term formula is preferred because it is
is, similar in characteristics, content, price, or price change). more flexible and has a number of advantages, including
• Elementary aggregates should be designed to be suf- (1) facilitation of the introduction of new outlets, items,
ficiently reliable for publication. This promotes greater and varieties; (2) imputations of temporarily missing
transparency and enhances user confidence in the data. prices should be made using the short-term change; and
• Temporarily missing price observations should be imputed (3) facilitation of data verification since outliers in short-
using all available prices or a subset of the available prices. term changes are more readily identifiable than those in
The imputation of temporarily missing prices is especially long-term ones.
important when using the modified Lowe or modified • Higher-level price indices should be calculated using the
Young index. When using these formulas, a price in the short-term index formulation (modified Lowe and modi-
previous period is needed. Without imputation, the sample fied Young). Two methods for calculating the index using
of prices used to calculate the index deteriorates. the short-term formula are described in the chapter and
• In general, the Jevons formula should be used for the calcu- either method is acceptable.
lation of elementary indices because of its better statistical • When compiling a national index, elementary aggregates
properties. With the Jevons formula, the results are identical for each region should be summed using their area expen-
whether the elementary index is calculated using the ratio of diture weights to derive the national item index.
198
UPDATING CPI WEIGHTS AND LINKING
NEW TO PREVIOUS CPI SERIES 9
Introduction 9.3 There are several important features of a chain index:
• The chain index formula allows weights to be updated
9.1 Consumer price index (CPI) weights should be and facilitates the introduction of new items and sub-indi-
updated on a periodic basis (see Chapter 3). The preferred ces and the removal of obsolete ones.
interval is at least once every five years. This chapter discusses
the processes and procedures for introducing a new basket • In order to be able to chain the old and the new series, an
and corresponding weights in the CPI as well as approaches overlapping period (k) is needed in which the index has to
for linking the new and previous CPI series following the be calculated using both the old and the new set of weights.
introduction of new weights. It includes examples of the steps • A chain index may have two or more links. In each link,
involved in linking CPI series with different price reference the index may be calculated as a fixed weight index using
periods, methods to keep the current index reference period equation 9.1, or indeed using any other index formula.
or shift to a new one and explores whether interim or partial The chaining period may be a month or a year, provided
weight updates might be implemented in the period between the weights and indices refer to the same period.
major surveys such as a Household Budget Survey (HBS). • Chaining is intended to ensure that the individual indices
on all levels show the correct development through time.
Calculating a Chain Index • Chaining leads to non-additivity so that chained indices
at the lower-level cannot be aggregated into indices at
9.2 Assume that a series of fixed weight Young indi- higher level using the latest set of weights. If, on the other
ces has been calculated with period 0 as the price refer- hand, the index reference period is changed and the index
ence period and that in a subsequent period, k, a new set of series prior to the chaining period is rescaled to the new
weights has to be introduced in the index. The new set of index reference period, this series cannot be aggregated to
weights may, or may not, have been price-updated from the higher-level indices by use of the new weights.
new weight reference period to period k. A chain index is
then calculated as: 9.4 An example of the calculation of a chain index is
presented in Table 9.1. From 2008 to December 2016 the
index is calculated with 2008 as weight and price reference
I 0:t = I 0:k ∑ wkj I kj :t −1 I tj−1:t period. From December 2016 onwards, a new set of weights
is introduced. The weights may refer to the year 2014, for
= I ∑ w j I j
0:k k k :t
example, and may or may not have been price-updated to
December 2016. A new fixed weight index series is then cal-
= I 0:k I k :t (9.1) culated with December 2016 as the price reference month.
Table 9.1 Calculation of a Chain Index

Weight 2008 Nov. 2016 Dec. 2016 Weight Dec. 2016 Jan. 2017 Feb. 2017 Mar. 2017
2008 2016
2008 = 100 December 2016 = 100

1. Elementary price indices
A 0.20 100.00 120.00 121.00 0.25 100.00 100.00 100.00 102.00
B 0.25 100.00 115.00 117.00 0.20 100.00 102.00 103.00 104.00
C 0.15 100.00 132.00 133.00 0.10 100.00 98.00 98.00 97.00
D 0.10 100.00 142.00 143.00 0.18 100.00 101.00 104.00 104.00
E 0.30 100.00 110.00 124.00 0.27 100.00 103.00 105.00 106.00
Total 100.00 119.75 124.90 100.00 101.19 102.47 103.34
2. Higher-level indices
G=A+B+C 0.60 100.00 120.92 122.33 0.55 100.00 100.36 100.73 101.82
H=D+E 0.40 100.00 118.00 128.75 0.45 100.00 102.20 104.60 105.20
Total 100.00 119.75 124.90 100.00 101.19 102.47 103.34
3. Chaining of higher-level indices to 2008 = 100
G=A+B+C 0.60 100.00 120.92 122.33 0.55 122.33 122.78 123.22 124.56
H=D+E 0.40 100.00 118.00 128.75 0.45 128.75 131.58 134.67 135.45
Total 100.00 119.75 124.90 124.90 126.39 127.99 129.07
199
Finally, the new index series is linked onto the old index the CPI weights: (i) price-updating the weights to the price ref-
with year 2008 = 100 by multiplication to get a continuous erence period (Lowe index) to keep the implied quantities fixed
index from 2008 to March 2017. The chained higher-level at the weight reference period levels; or (ii) simply introducing
indices in Table 9.1 are calculated as: the new weight structure (Young index) which keeps the expen-
diture shares fixed. This chapter examines what the choice of
I 2008:t = I 2008:Dec 2016 ∑ w2008(Dec 2016) Dec 2016:t
Ij (9.2) the updating method should be, using some objective criteria.
j
9.9 The target index for the CPI can be a Laspeyres price
index which is easy to explain to users. It is a fixed quan-
9.5 Because of the lack of additivity, the overall chain tity basket price index, with quantities held fixed in the price
index for March 2017 (129.07), for example, cannot be reference period. Alternatively, the target index for the CPI
calculated as the weighted arithmetic mean of the chained could also be the Fisher, Törnqvist, or Walsh price indices.
higher-level indices G and H using the weights that was Chapter 4 in Consumer Price Index Theory demonstrates that
introduced from December 2016. these three indices produce essentially the same results in
practice. The Laspeyres index formula, however, is not gen-
Updating Weights for Price Change: erally used in practice. This is because the weight reference
period, the period to which the expenditure weights refer, is
Pros and Cons generally earlier than the price reference period of the CPI.
Approaches to Updating Weights To derive an index that commences from its price reference
period but keeps quantity weights fixed based on the earlier
9.6 An issue that National Statistical Offices (NSOs) face weight reference period, many NSOs would price update the
in conducting revisions of the CPI is the timeliness of the data earlier weights. In this way, the resulting Lowe index was
on weights and their introduction in the CPI. Typically, a HBS often referred to as a “Laspeyres-type index.” This raises the
or other expenditure survey collects data during a prior year question of which index currently produced by NSOs would
to the current time period. While the expenditure data from best approximate the target indices: the Lowe index with
the HBS are the most commonly used source for develop- price-updated weights, or the Young index that simply uses
ing weights, alternative sources can be used. For example, the weights from the weight reference period.
national accounts data. Alternative sources for developing 9.10 For example, the European Union Harmonized
weights are discussed in more detail in Chapter 3. Thus, the Index of Consumer Prices (HICP)1 is defined as a Laspeyres-
period to which the weights refer precedes the current period type index where the weight reference period is year t-1 and
and the NSO cannot produce one of the target indices that the price reference period is December of year t-1. In prac-
require weights for the current period. When the weights are tice, year t-1 expenditure data are not available, thus weights
introduced, the price reference period precedes the current are derived using preliminary national accounts data refer-
period and the time period over which the weights will be used ring to year t-2. The observed year t-2 expenditure may, or
extends beyond the current period. Thus, there is a continuum may not, be price updated between year t-2 and year t-1.
for the life cycle of the CPI that begins at a point in the past and The objective is to obtain the best possible estimate of the
continues to a point in the future, as illustrated in Figure 9.1. expenditure shares for the year t-1. If goods and services are
9.7 The index must have a period to which the weights perfect complements (i.e. there is no substitution between
refer, the weight reference period b, when the expenditure them and they are consumed in fixed proportions), the best
data are collected over a period of months or quarters, usually approximation would be the price-updated weights. If goods
for a year. The results of the survey are processed, compiled and services are substitutes at such a rate that expenditure
to provide estimates of expenditures by detailed products to on one product relative to another is independent of the rela-
establish elementary aggregates, and introduced in the CPI. In tive prices, the preferred approach would be not to price-
addition, the index must have a starting point, the price refer- update. It is possible to evaluate retrospectively which of
ence period 0, to which future prices will be compared in cur- the two options performs better by comparing the outcomes
rent periods t. The CPI will continue through time to its end with that from using actual expenditure shares when they
point T after which a new set of weights, products, and prices become available. The degree of substitution may vary
will be introduced. Finally, the index must have a index refer- across products, so the choice between these two options
ence period which is the period when the index is set to 100. is not necessarily straightforward. In any case, the estimate
9.8 When introducing the new weights, NSOs must decide of the expenditure shares for the year t-1 are always price-
which method to use. There are two main options for updating updated to the price reference period, December of year t-1.
9.11
Figure 9.1 The CPI Life Cycle Should NSOs Use the Lowe or the Young
Index (To Price Update or Not)?
Weight Price
Current End of 9.12 In Chapter 3 of Consumer Price Index Theory various
Reference Reference
Period Period
Period Index Link axioms are used to measure the performance of different price
index formulas. From an axiomatic point of view, the Lowe
1
b 0 t T See also Harmonized Index of Consumer Prices (HICP) Methodological
Manual, Section 3.5, https://ec.europa.eu/eurostat/web/products-manuals-
and-guidelines/-/KS-GQ-17-015
200
UPDATING CPI WEIGHTS AND LINKING NEW TO PREVIOUS CPI SERIES
index performs better than the Young index. Of the twelve tests of time between the weight reference period and the period
used, the Lowe passes all, while the Young index passes ten, when the weights are introduced in the index. Thus, whether
failing the time reversal and the circularity tests. Thus, some the weights are price-updated or not, they should be reviewed
NSOs consider the Lowe index preferable to the Young index. and updated as frequently as possible to reduce potential bias.
9.13 The recent research cited suggests that price updat- When rapid changes take place in relative quantities as well
ing weights before introducing them in the CPI may not be as relative prices, NSOs are effectively obliged to change the
the best approach for the CPI based on the performance of expenditure weights more often. Price-updating on its own
the resulting index formula when compared to the preferred cannot cope with this situation. The weighting basis must be
target indices. Price-updating the expenditure shares does updated with respect to both quantities and prices, which, in
not imply that the resulting weights are necessarily more effect, implies that new weights have to be included.
up to date. When there is a strong inverse relation between
movements of price and quantities, price-updating on its
own could produce unreliable results. For example, assume Recent Research on Price-updating
the price of beef increases relative to the price of chicken. If Expenditure Weights
the quantities are held fixed while the prices are updated, the 9.19 Economists make certain assumptions about nor-
resulting expenditure weights for beef would increase sub- mal consumer behavior based on observation and economic
stantially while the expenditure weights for chicken would theory. Both indicate that in response to relative price change
decrease. In fact, consumers would spend less on beef and consumers will adjust the quantities of goods and services
spend more on the relative less expensive chicken; however, they purchase. When the price of a product rises relative to
because the quantities are fixed and the changes in relative other similar products, consumers will usually reduce the
expenditure are not reflected by price updating. amount purchased of the product with the relatively higher
9.14 By price updating, the weights are aligned to the price increase and purchase more of the similar products with
same reference period as the prices. If the NSO decides to the relatively lower price increases. The opposite phenom-
price update the weights, the resulting index will be a Lowe enon will occur as relative prices fall—consumers purchase
index. The Lowe index is a fixed basket index, which from more of the product with the relative decline in prices and
period to period measure the value of the same (annual) bas- less of the products with the relative price increases. This con-
ket of goods and services. sumer reaction to price movements, known as substitution, is
9.15 Not price-updating the weights results in the calcu- the theory underlying the downward sloping demand curve.
lation of a Young index. The Young index keeps the expen- 9.20 Given this behavior, Chapter 4 in Consumer Price
diture shares fixed in the expenditure survey period b. The Index Theory outlines why the Laspeyres price index will
Young index is a fixed weight index where focus is that the be an upper bound to the true cost-of-living index (COLI)
weights should be as representative as possible for the aver- and the Paasche price index will be a lower bound. It also
age value shares of the period covered by the index. shows subsequently that the three target indices—Fisher,
9.16 By keeping the expenditure shares constant from Törnqvist, and Walsh—are very close approximations to
the weight reference period to the price reference period one another and to the true COLI.
the underlying quantities are assumed to vary in response 9.21 The Lowe index, like Laspeyres, assumes that con-
to changes in relative prices. Hence, if households tend to sumers do not substitute away from items with relatively
keep constant expenditure shares by substituting from goods large price increases; the relative quantities are fixed at the
or services with relative price increases to goods or ser- weight reference period (b). In practical terms this means
vices with relative price decreases, the period b expenditure that items with relatively large increases in price gain an
shares will be good estimates of the expenditure shares in the unduly large expenditure share in period 0 relative to period
price reference period when the weights are introduced in b (where b precedes 0). Balk and Diewert (2003), Balk
the index. In turn, if expenditure shares stay unchanged, the (2010), and Chapter 4 in Consumer Price Index Theory
Young index will be a good estimate of a target superlative show that the Lowe index has an upward substitution bias
index. However, if quantities tend to remain constant (i.e., compared to the true COLI. This bias increases the longer
the households do not substitute between goods and services the period between the weight reference period (b) and the
in response to relative price changes), the Young index will price reference period (0). Chapter 4 in Consumer Price
be biased downwards compared to a superlative target index. Index Theory also notes that the Lowe index is upwardly
9.17 Whether a Young or Lowe index is the better esti- biased relative to the Laspeyres index. Thus, price updating
mate of a superlative target index depends on whether the the weights results in an index number that is upward biased
original or the price-updated weights are the better estimate relative to the target indices, as well as the Laspeyres index.
of the average expenditure shares from 0 to t. Normal con- 9.22 The dispersion of price change over the period
sumer behavior suggests that in general some substitution affects the magnitude of the substitution bias. In the unlikely
should be expected, so that the Lowe index will tend to be event that all prices change in the same proportion, there will
biased upward compared to a superlative target index. As be no bias from using the Lowe index rather than Laspeyres.
the Young index allows for some substitution from b to 0, The price-updated weights will be the same as the weights in
while Lowe does not, it may be argued that the traditional the weight reference period. However, if prices are trending
Laspeyres bias to some degree is reduced in the Young index upward with normal consumer substitution behavior, the dis-
as compared to the Lowe index. Thus, to omit price-updating persion in prices is expected to increase and price-updating
may be one practical way in which to reduce this type of bias. would have a substantial effect on the weights. This implies
9.18 Both quantities and expenditure shares change the bias in the Lowe index would be larger than if there were
through time and progressively more, the longer the elapse little price change. In general, prices tend to trend upward
201
over time and therefore it is clear that price updating weights Detailed Methods for Updating
from period b to 0 will result in an upward bias compared to
the target indices (and the Laspeyres index also). Weights
9.23 An alternative approach would be to use the weights
from the weight reference period directly. Boldsen Hansen
Decide on Price Reference Period for
(2006) argues for using the Young index. Keeping the expen- Calculating the Updated CPI
diture shares fixed at the weight reference period level means Annual Price Reference Periods
that there has been no change in weights between b and 0. 9.27 As the new weights are developed from the HBS or
Unchanging weights is consistent with consumer substitution alternative sources for weights such as national accounts data,
behavior that exhibits unitary elasticity—consumer shifts in the NSO must decide on what time period to use for the new
reducing the quantities purchased are in the same proportion price reference period. Often the choice is between a single
as the increase in prices. period of less than a year (month or quarter) that is close to
9.24 Whether the Young index is biased relative to the current period or a yearly average. In either case, the price
the Laspeyres index will depend on the long-term trend in reference period should immediately precede the introduction
prices and the elasticity of substitution.2 In general, the long- of the new index series. For example, an index introduced
term trend in prices for most items has been positive. Given in January should have a price reference of either the month
this trend, if the elasticity is lower than one (inelastic), the prior, December, or an average of the previous year.
Young index may have a downward bias compared to the 9.28 For infrequently updated CPIs, a single year is
Laspeyres. This occurs because consumers on average do preferred as the price reference period. If a single month (or
not substitute as much in response to price changes as the quarter) is used, the prices of some seasonal products will
Young index implies. They tend to purchase the relatively be unavailable or unusually high or low and a large number
higher priced items in greater quantities than implied by uni- of unusual or imputed prices may have to be used for the
tary elasticity. If the elasticity is greater than one (elastic), price reference period. Furthermore, if a single unusual price
the Young index may have an upward bias compared to the reference month is to act as the reference period, the initial
Laspeyres, because consumers tend to substitute more than result of the index change may be a distorted one. Preferably
assumed. This is presented in more detail in Chapter 2 of the price reference period for infrequently updated CPIs is a
Consumer Price Index Theory. whole year in which seasonal prices would be appropriately
9.25 Recent studies of the potential bias in the Young represented. In some months there will be no sales, for exam-
and Lowe indices indicate that measured price changes using ple, of a seasonal fruit, but an average price for the whole
the Lowe index generally exceed those of the Young index year would still be available for the price reference period.
and both exceed the target indices. Boldsen Hansen (2007) For a homogeneous item specification, say a large Bartlett
using Danish CPI data, Greenlees and Williams (2010) and class I pear grown in California, it is recommended that, data
Armknecht and Silver (2013) using United States (US) CPI permitting, the average price used in the reference period be
data, Pike et al (2009) with New Zealand CPI data, and a unit value, that is, the sum of the value purchased over the
Huang, Wimalaratne, and Pollard (2016) using Canadian whole year divided by the sum of the quantity purchased as
CPI data all verify that the CPI price changes measured measured in the HBS. The advantage of the unit value is that
using the Lowe exceed those of the Young index, and both the resulting average price gives accordingly less weight to
are greater than those of the Törnqvist index. prices in months in which there were fewer purchases.
9.26 Armknecht and Silver (2013) also suggests that 9.29 If the NSO chooses the annual period for the price
there are practical solutions that might be used to simulate reference period, it will need annual average prices for all
a target index. This study provides evidence from the meth- items of the rebased series. Such prices will be collected
odology proposed by Lent and Dorfman (2009) that using alongside those of the existing index and in many cases may
the geometric average of arithmetic indices, such as Young be for the same specification. The existing index’s prices, for
or Lowe, with geometric indices such as geometric Lowe or example for 2017, will be collected and used for the monthly
geometric Young, respectively, can closely approximate the index compiled and published for 2017, however the prices for
Fisher or Törnqvist indices. NSOs should be able to pro- the supplementary items/products for the new index should be
duce these four indices using the expenditure data available collected alongside them for the rebased index to commence
from the HBS. First, the NSO can calculate the Young and in January 2018, or soon afterwards, with year 2017 as the
Lowe indices with the HBS weights and the price-updated index reference period (100). The price collection for the items
weights. Next, the NSO can calculate the geometric Young that belong to the old index but not to the new index can be
and geometric Lowe by using the geometric aggregation stopped once the rebased index has been released.
formula (e.g., by taking a weighted average of natural loga- 9.30 It is important to stress the practical arrange-
rithms of price relatives and converting the logarithm back ments required for price collecting the specifications for
to an index number). The NSO can then test which combi- the new index alongside the old. Many items sold in the
nation of the resulting arithmetic and geometric versions same outlets sampled for the existing index will remain
provide the best approximation to the Fisher or Törnqvist the same for the revised index and thus involve no addi-
indices. tional workload. There may be some changes to items that
replace an old one, for example brown sauce, with a new
one, say ketchup, in the same outlet which again requires
2
Often the Laspeyres price index is the target index for countries’ CPI so limited additional resources. However, the very need for
the comparison is made first to the more standard target then to the superla- the rebase to capture changes in item/variety specifications
tive index targets (Fisher, Törnqvist, or Young).
202
and include new outlets in place of old ones—a purpose the price reference period. There is a continuing flow of
of the rebasing—will require prices to be collected for price data that may include imputations, and a relatively
some new items in new outlets. Resources have to be small number of changes in specifications or products;
committed in advance to plan for this. The sampling of the major exercise is to bring the new weights to bear on
products and broad item specifications and the associated the flow of price data and link this to the existing chain.
outlet selection provides the basis for determining the At the same time, the annual update of the price refer-
new item/variety specifications to be sampled and the old ence period remains an opportunity for re-sampling in
ones replaced. order to ensure that the basket remains representative,
9.31 Some of the items for the new index will be the and for possibly implementing other methodological
same as the old; some will be different, but available in the improvements.
same outlet; and some will be available in new outlets, with 9.35 For infrequently updated CPIs (i.e., those updated
the existing or new items. These provide the prices in the every five years or less frequently), the use of a single-
price reference period from which the new index is mea- month price reference period is not ideal; however it is often
sured. An average of the existing index for 2017 and the the case that the resources of an NSO are only sufficient to
rebased index 2017 = 100 provides the linking factors that allow for a price reference period of less than a year. Often
enable the new index to be a continuation of the existing this is monthly, though it follows from the reasons given
index. in paragraphs 9.28 and 9.29 that six-months is preferred to
9.32 It is usually the case that the overlap year and quarterly, and quarterly to monthly. The principles for using
prices collected for the rebased index are for a period fol- a monthly price reference period are similar to those of an
lowing the weight reference (survey) period. This is because annual one, the issue being to mitigate as much as possible
CPI compilers have to wait until the results of the HBS its shortcomings.
have been compiled and new elementary aggregates deter- 9.36 A primary shortcoming of using a monthly refer-
mined to know what to price. The purpose of rebasing is ence period is that out-of-season items in the price reference
to update the products and items priced and this needs to period will have no observed or economically meaningful
be informed by results from the survey. For example, the price. The decision as to which month to use for the price
HBS may be undertaken in 2015; during 2016 (i) the results reference month should take account when seasonal items
would be compiled to yield expenditure shares by product with relatively high weights are in-season. If not in season,
category (and possibly region); (ii) the results are validated/ an imputed price will have to be used and consideration
developed using ancillary data as appropriate; (iii) elemen- should be given to the validity of imputation methods for
tary aggregate products are selected (e.g., using cut-off sam- out-of-season items in this context. For example, if the
pling) and their broad specifications developed; (iv) use of carryfor method is used, the month’s imputed price in the
centrally-determined prices for selected products are deter- price reference period for the out-of-season item may be
mined (e.g., electricity, water, insurance); and (v) an out- unduly low. As mentioned above, the short-term price rela-
let sample selected for the remaining products. Pricing will tive method is preferred since it avoids a need for long-run
commence with an initiation of the sample by visits to the price comparisons with this one-month price reference
outlets to determine and price the appropriate representative period.
attributes for the selected item. This may commence in Jan-
uary 2017 or later depending on the resources of the NSO.
A whole year’s price data, say for 2017, will be collected for Developing Update Factors
the new products and their specifications alongside the old. 9.37 While an NSO can maintain the same weights and
The rebased index will be a Lowe index with a weight refer- price reference period, most NSOs choose to use a more cur-
ence period of 2015 and a price and index reference period rent period to introduce the new weight structure. Assume
of 2017 = 100. the NSO decides to introduce the new weights from the 2015
9.33 NSOs that calculates the elementary indices as HBS in the January 2018 CPI with a price reference date
chained monthly indices in which prices are compared to of December 2017, it has two choices for the weights: (a)
previous month’s prices, have the advantage of being able to introduce the 2015 weights directly, or (b) price update the
readily introduce replacement items/varieties so the task of weights to December 2017. As noted, the first choice will
collecting specifications for the new index alongside those result in a Young index while the second will result in a Lowe
of the existing one will be less arduous. The long-term price index.
relative method (i.e., the current period price compared to 9.38 Table 9.2 presents the approach for price updat-
the average price from the price reference period), makes ing the weights from the 2015 weight reference period to
such procedures more difficult because adjustments to base December 2017. A single month is used here for illustrative
prices are often required. purposes although, for reasons outlined in paragraphs 9.28
and 9.29, the whole year 2017 could also be used. The NSO
Monthly (Or Quarterly) Reference Period begins the rebased CPI using the price data for December
9.34 NSOs that use chained CPIs and annually 2017 and January 2018. It also produces the old CPI with
update their weights with a relatively small-time lag 2010=100 for December 2017. These data will be used for
between the weight and price reference period should linking the old and new series as discussed in paragraphs
use a single month as the price reference period. For 9.73 to 9.88.
instance, the weights that refer to year y-2 are available 9.39 The first step in updating the weights is to develop
and finalized in year y-1 so that they can be introduced the update factors that measure the price change from 2015
in the January index of year y with December year y-1 as to December 2017. This process requires a measure of price
203
Table 9.2. Updating Weights for Price Change from Weight Reference Period
COICOP Code Description Expenditure Avg. CPI CPI Dec. Update Updated Normalized
Share 2015 2015 2017 Factor Weight Weight
(A) (B) (C) (D) (E) (F) (G) (H)
Col E/D Col. FxC
01 Food and Non-alcoholic Beverages 100.000 206.883 100.000

01.1 FOOD 93.306 197.80 95.61
01.1.1 Bread & Cereals 21.419 40.08 19.37
01.1.101 Rice (white) 1.406 150.7 318.1 2.111 2.969 1.435
01.1.102 Rice (Brown) 3.361 151.7 224.6 1.480 4.975 2.405
01.1.103 Flour 2.578 134.8 320.2 2.375 6.121 2.959
01.1.104 Bread 6.864 127.2 222.4 1.748 12.006 5.803
01.1.105 Biscuits (Salted) 0.813 113.0 140.1 1.240 1.008 0.487
01.1.106 Cakes, Pastry, etc. 1.034 131.2 233.7 1.781 1.842 0.891
01.1.107 Chow mein 1.716 125.1 309.8 2.476 4.247 2.053
01.1.109 Macaroni 1.284 105.8 200.3 1.893 2.429 1.174
01.1.110 Oat flakes 0.450 100.6 225.7 2.244 1.010 0.488
01.1.111 Sago 0.341 103.5 222.9 2.154 0.735 0.355
01.1.112 Tennis Rolls 1.392 126.0 219.6 1.743 2.426 1.173
01.1.113 Whole Wheat Bread* 0.180 127.2 222.4 1.749 0.315 0.152
01.1.2 Meat 17.632 39.94 19.31
01.1.201 Stew Beef 1.940 110.5 254.8 2.306 4.474 2.162
01.1.205 Chicken (live) 1.038 112.6 229.7 2.040 2.117 1.023
01.1.206 Chicken (frozen) 10.202 110.2 252.5 2.291 23.370 11.296
01.1.207 Pork Leg 0.610 138.8 483.4 3.483 2.123 1.026
01.1.210 Corned beef* 0.866 111.9 253.7 2.114 1.832 0.949
01.1.211 Duck 0.217 107.3 148.3 1.382 0.300 0.145
01.1.212 Liver 0.207 115.4 198.6 1.720 0.356 0.172
01.1.213 Mutton 0.271 106.8 256.7 2.404 0.651 0.315
01.1.214 Sausages (pork & chicken) 1.823 120.3 233.0 1.937 3.531 1.707
01.1.215 Brisket* 0.458 110.5 254.8 2.306 1.056 0.511
01.1.3 Fish & Seafood 5.982 12.66 6.12
… … … … … … … …
change for the elementary aggregates in the new basket. A have collected prices or started test elementary aggregate indi-
large majority of the items in the new basket are the same as ces during 2015. Such cases are the exception rather than the
in the previous basket and continue in the rebased CPI. Their rule, and it is unlikely that all new items in the basket will have
price change can be measured either by using a matched prices or indices available in December 2017.
sample of (geometric) average prices for 2015 compared to 9.41 For the new items, the CPI compiler can identify
those in December 2017 or elementary aggregate price indi- existing elementary aggregates whose price trends might be
ces over the same period. It is very difficult to have the same reflective of the new elementary aggregates in the basket.
set of matched price observations in both periods because Otherwise, the price index for the higher-level class aggre-
product substitutions have most likely occurred over the gate can be used. In Table 9.2, there are three new items and
two-year period. Therefore, NSOs will find it easier to use for each of these the CPI compiler needs to select a represen-
the elementary aggregate indices already available for the tative elementary aggregate price index. As in the previous
CPI elementary aggregates that continue. example, brisket uses the stew beef index as representing its
trend. For most other new items, the same approach is taken:
whole wheat bread uses the bread index. The only item for
, 2010=100 / I i , 2010=100
Factori Dec 2017 = I iDec 2017 Avg 2015
(9.3)
which the class index is used is corned beef.
9.42 For each elementary aggregate index in Table 9.2,
9.40 There is an issue, however, for deriving the update the 2015 monthly indices were averaged for the calendar
factors for the new elementary aggregates in the basket. As pre- year and appear in column D.4 The December 2017 indices
viously noted, NSOs will not have price data for the new items are in column E. The update factors are calculated in col-
from 2015, nor will they have price indices to use for calculating umn F as the December 2017 index (column E) divided by
the weight update factors for the new elementary aggregates.3 the 2015 average index (column D). Next, the 2015 weights
In some instances the CPI compiler may have anticipated that (column C) are multiplied by the update factors (column
certain items were becoming important and most likely would F) to derive the updated weights (column G). Finally, the
appear in the new basket. For these items, the compiler may updated weights are normalized (divided by the total of the
4
The calculations for updating the weights are only made for the item indi-
ces where the new weights are applied. The class and higher-level indices
3
A similar issue may arise with changes to the classification system or will be calculated using the weights updated to December 2017 in para-
introduction of new geographical areas into the CPI. graphs 9.47–9.56.
204
updated weights in the first row of column G and converted weight reference period for the HBS) is set to 100 (column
to percentages) to sum to 100 (column H). D) by dividing all the elementary indices by their 2015 aver-
age and expressing them as an index. In columns E, F, and G,
Updated weight: wi
2015( Dec 2017)
= Factori Dec 2017 wiAvg 2015 the new weights are used to aggregate the rebased elemen-
tary indices to higher levels starting with the subclass levels.
k
2015( Dec 2017)
= ∑wi (
2015 Dec 2017)
Aggregate updated weight: wagg
i=1 I it , 2015=100 = I it , 2010=100 / I i2015 Avg , 2010=100 (9.4)
( )
n
2015( Dec 2017) 2015( Dec 2017)
Percentage shares: siDec 2017 = wi / ∑wi At×100
the elementary aggregate level the new weights are
i =1
applied to derive all higher-level indices, up to the overall CPI.
( )×100
n
2015( Dec 2017) 2015( Dec 2017)
siDec 2017 = wi / ∑wi k k
i =1
t , 2015=100
I agg = ∑wiAvg 2015 I it , 2015=100 / ∑wiAvg 2015 (9.5)
i =1 i =1
9.43 Note that the update factors are only calculated at
the elementary aggregate level and are used only to update 9.47 The CPI compiler must complete this aggrega-
the elementary aggregate weights. The higher-level weights tion for all months from January 2015 through January 2018.
are calculated by aggregating the lower-level weights: (i) sub- Table 9.3 only shows the aggregates for three months—
class weights are the sum of the elementary aggregate weights December 2016, December 2017, and January 2018. The
within the subclass, (ii) class weights are the sum of the sub- revised index series will now be available from 2015
classes within the class, (iii) group weights are the sum of the through January 2018.
classes within the group, and (iv) the division weights are the 9.48 The argument for keeping the weight reference
sum of the groups within the division. If the NSO calculates period and price reference period the same is that the result-
an update factor at the subclass level, it will not get the same ing price index will approximate a true Laspeyres index;
result as when summing the elementary aggregate weights. however, this procedure is not without problems. Consider
The new aggregated weights reflect the new basket’s structure. for example a weight reference period of 2015 for which the
Using the aggregate indices to calculate update factors at the new sample of products/items have prices collected in 2017
subclass level or above reflects the structure of the old basket. with an aim to commence the index in January 2019. Instead
9.44 Another important point is that, when price updat- of using 2017 as the new overlap index reference period,
ing the weights, the elementary aggregates that have rela- 2015 = 100 is used. As explained below, this is achieved by
tively larger price changes from the weight reference period backcasting the 2017 prices to 2015, a procedure that may
receive a larger share of the weight and those with relatively involve some imputations. Nonetheless, if this method is
lower price changes receive a smaller share. For example, used, what is derived is a Laspeyres index with the same
the elementary aggregate with the largest price change is weight and price reference periods. However, the interest in
pork leg (update factor of 3.484) and the updated weight compiling the new rebased index is for indices from Janu-
is 1.026 versus 0.610 in the weight reference period. The ary 2016 onward. The period-to-period comparisons use
elementary aggregate with the smallest price change is Laspeyres price indices, but the resulting measure will be
biscuits (update factor of 1.240) and the updated weight is a Lowe index comparison, not Laspeyres. The ratio of the
0.813 verses 0.487 in the weight reference period. In sum, two Laspeyres indices used in calculating the price change
all the elementary aggregates that exceed the average price results in a Lowe formula as follows:
changes receive greater weight than in the weight reference
period while all the elementary aggregates with less than ∑ pit qi0 ∑ pit −1qi0 ∑ pit qi0
I L0:t / I L0:t −1 = = (9.6)
average price changes receive less weight. ∑ pi0 qi0 0 0
∑ pi qi ∑ pit −1qi0
Introducing New Weights where I L0:t is the Laspeyres price index for period t
9.45 The NSO has three options on introducing the new I L0:t −1 is the Laspeyres price index for period t − 1
basket and weights. First, new weights can start with the
9.49 The resulting price change does not yield a Laspey-
weight reference period and price reference period being
res price index but a Lowe index where the quantities of
the same (Laspeyres). The second option is to introduce the
period 0 are valued at the prices in period t and t–1. A true
expenditure weights directly in a new (subsequent) price ref-
Laspeyres index would value the fixed quantities of period
erence period (Young). The third option is to introduce price- t t −1 t −1 t −1
updated weights in a new price reference period (Lowe). Each t-1 at the prices of period t-1 and t (i.e., ∑ pi qi / ∑ pi qi ).
of these options is demonstrated in paragraphs 9.46 to 9.54. Thus, going through the process of calculating the Laspey-
res index by backcasting still results in comparisons that
Introduce New Weights With the Same result in Lowe index price changes.
Weight and Price Reference Periods— 9.50 The NSO often has difficulty in obtaining price
Laspeyres Index measures for new items that are introduced with the new
9.46 The NSO can use the weight reference period as the weights. Because of the time lag between the HBS and the
price reference period. In addition, the NSO would also need development of the CPI weights, the required price data to
to reset the index reference period of the elementary indices use as the reference prices would usually be two to three
to 100 using the same period as the weight and price refer- years old. Retail outlets will find it very difficult to provide
ence periods. In Table 9.3 the average price for 2015 (the accurate prices for the period covered by the weights. This
205
Table 9.3. Updated CPI with the Same Weight and Price Reference Periods
COICOP Code Description Expenditure Avg. CPI CPI Dec. CPI Dec. CPI Jan.
Share 2015 2015 2016 2017 2018
(A) (B) (C) (D) (E) (F) (G)
01 Food and Non-alcoholic Beverages 100.00 100.00 153.20 206.41 209.24

01.1 FOOD 93.31 100.0 155.7 211.5 214.4
01.1.1 Bread & Cereals 21.42 100.0 143.6 187.1 189.5
01.1.101 Rice (white) 1.406 100.0 155.5 211.1 213.2
01.1.102 Rice (Brown) 3.361 100.0 124.0 148.0 150.2
01.1.103 Flour 2.578 100.0 168.7 237.5 242.2
01.1.104 Bread 6.864 100.0 137.5 174.9 176.7
01.1.105 Biscuits (Salted) 0.813 100.0 112.0 123.9 125.8
01.1.106 Cakes, Pastry, etc. 1.034 100.0 139.1 178.1 181.7
01.1.107 Chow Mein 1.716 100.0 173.8 247.6 250.1
01.1.109 Macaroni 1.284 100.0 144.6 189.3 190.2
01.1.110 Oat flakes 0.450 100.0 162.2 224.4 227.8
01.1.111 Sago 0.341 100.0 157.7 215.5 218.7
01.1.112 Tennis Rolls 1.392 100.0 137.2 174.3 176.0
01.1.113 Whole Wheat Bread* 0.180 100.0 137.5 174.9 178.4
01.1.2 Meat 17.63 100.0 161.6 226.5 226.9
01.1.201 Stew Beef 1.940 100.0 165.3 230.6 232.9
01.1.205 Chicken (live) 1.038 100.0 152.0 204.0 207.0
01.1.206 Chicken (frozen) 10.202 100.0 164.5 229.1 233.6
01.1.207 Pork Leg 0.610 100.0 224.2 348.3 351.8
01.1.210 Corned beef* 0.866 100.0 128.5 226.8 159.3
01.1.211 Duck 0.217 100.0 119.1 138.2 141.0
01.1.212 Liver 0.207 100.0 136.0 172.0 173.8
01.1.213 Mutton 0.271 100.0 170.2 240.4 241.6
01.1.214 Sausages (pork & chicken) 1.823 100.0 146.8 193.7 196.6
01.1.215 Brisket* 0.458 100.0 165.3 230.6 234.0
01.1.3 Fish & Seafood 5.98 100.0 155.8 211.6 214.5
… … … … … … …
can be problematic for an index that uses the long-term price 9.54 Assume the CPI compiler has decided to introduce
relative method (i.e., the current period price compared to the new weights directly in December 2017 to use in the
the average price from the price reference period). compilation of the January 2018 index (skipping the com-
9.51 The alternative is to use price indices to measure pilations in columns D to F of Table 9.3). The new price
the change in prices for the time lapse between the weight reference period is December 2017. The price relatives for
reference period and the introduction of the new weights. For January 2018 are used to estimate the January 2018 elemen-
items that are already in the CPI basket, their index can be tary index levels and the new weights (from Table 9.4,
re-referenced to the weight reference period. For example, if column C) are used directly with the elementary indices to
the current time period is late in 2017 and the weight refer- derive the higher-level aggregates.
ence period is 2015, the NSO can use the annual average 9.55 The new index starts with December 2017 = 100
price index for 2015 to re-reference the series so that the as shown in column D of Table 9.4. The January 2018 ele-
2015 average is set to 100. Or it could use the prices for the mentary indices appear in column E. The 2015 CPI weights
item from 2015 (if they are available) to calculate the 2015 are used to aggregate the elementary indices to the subclass,
average price and compare current period prices to the 2015 class, group, and division levels as discussed in the previous
average. example.
9.52 For new items to the CPI, the NSO will not have
the prices or an index for the item. In this case, an index for  m It 
I it , Dec 2017=100 = ∏ j =1 j Dec 2017  ×100 (9.7)
a related item or for the subclass in which the new item is  Ii 
included may be used to reflect the price change from the
weight reference period. 9.56 The aggregate index changes are different from
those in the previous example because the weights reflect
that the expenditure shares are kept fixed at their 2015 levels,
Introduce the New Weights Directly in a New but not used in the index calculation until December 2017.
Price Reference Period—Young Index
9.53 The second option is for the NSO to introduce the k k
new weights directly in a new price reference period. As dis- t , Dec 2017 =100
I agg = ∑wiAvg 2015 I it , Dec 2017=100 / ∑wiAvg 2015 (9.8)
i =1 i =1
cussed in paragraphs 9.12 to 9.18, this may be the preferred
approach if there has been significant price change between In Table 9.3, the price change for Food and Non-alcoholic
the weight reference period and the introduction of the new Beverages is 1.37 percent ([209.24 ÷ 206.41] x 100 – 100),
weights. compared to 1.36 percent (column E) in Table 9.4.
206
Introduce the Updated Weights in a New Price to December 2017 with the same date as the new price ref-
Reference Period—Lowe Index erence period. The price changes for these two indices will
9.57 The third option is for the NSO to introduce the be the same because they keep the same fixed quantities. As
new weights updated for price change between 2015 and prices change through time, the expenditures and their shares
December 2017 (Table 9.2, column H) in the new price ref- will change providing greater importance to those items with
erence period. The new price reference period is Decem- larger than average price changes. This fact can be seen in
ber 2017. The price relatives for January 2018 are used to Table 9.2. The aggregate price relative between 2015 and
estimate the January 2018 elementary index levels and the December 2017 is 2.069 (206.883 ÷ 100). All items with price
price-updated weights are used with the elementary indices changes (update factors in column F) greater than this value
to derive the higher-level aggregates. have larger final weights (column H) when compared to the
9.58 Referring to Table 9.4, the revised index starts with 2015 weights (column 3). The Young index produced in the
December 2017 = 100 as shown in column D. The price-updated second approach (column E of Table 9.4) holds the expen-
weights from 2015 to December 2017 (column F) are used to diture shares fixed and allows for substitution of quantities
aggregate the elementary indices to the subclass, class, group, inversely proportional to price changes. For this example, the
and division levels as discussed in the previous examples. Young index has a smaller price change between December
2017 and January 2018 than the Lowe index, as confirmed by
= ∑ i =1 wi2015( Dec 2017 ) I iDec 2017:t several empirical studies cited earlier (see paragraphs 9.19 –
∑
k k
t , Dec 2017 =100
I agg wi2015( Dec 2017 )(9.9)
i =1
9.26). In general, however, it cannot be said a priori if either
The aggregate index changes are the same as those in Table 9.3 the Young or the Lowe index is higher.
because the weights reflect the fact that the quantities (not
expenditure) shares are fixed at their 2015 levels. In Table 9.4, Decide on the Index Reference Period
the price change for Food and Non-alcoholic Beverages using 9.60 The NSO has the choice of setting a new index
the price-updated weights is 1.37 percent (column G). reference period or keeping the old index reference period.
9.59 In the first approach (Table 9.2), the NSO estimated In the previous examples, new reference periods were set. In
a Laspeyres index with the quantity shares fixed at 2015 lev- the first case, the NSO used the HBS weight reference period
els and the price reference period of 2015. The Laspeyres of the 2015 annual average as the weight, price, and index
index requires that the price and weight reference periods are reference periods (Laspeyres index). In the second example,
the same. In the third approach (Table 9.4, columns F-G), the a new price and index reference period was established for
NSO compiled a Lowe index using price-updated weights December 2017 with the weight reference period of 2015
Table 9.4. Updated CPI with a New Price Reference Period

COICOP Description Expenditure CPI Dec. CPI Jan. Expenditure CPI Jan.
Code Share 2015 2017 2018 Share Dec 2017 2018
(A) (B) (C) (D) (E) (F) (G)
01 Food and Non-alcoholic Beverages 100.00 100.0 101.36 100.00 101.37

01.1 FOOD 93.31 100.0 101.38 95.61 101.39
01.1.1 Bread & Cereals 21.42 100.0 101.26 19.37 101.27
01.1.101 Rice (white) 1.406 100.0 101.00 1.435 101.00
01.1.102 Rice (Brown) 3.361 100.0 101.50 2.405 101.50
01.1.103 Flour 2.578 100.0 102.00 2.959 102.00
01.1.104 Bread 6.864 100.0 101.00 5.803 101.00
01.1.105 Biscuits (Salted) 0.813 100.0 101.50 0.487 101.50
01.1.106 Cakes, Pastry, etc. 1.034 100.0 102.00 0.891 102.00
01.1.107 Chow Mein 1.716 100.0 101.00 2.053 101.00
01.1.109 Macaroni 1.284 100.0 100.50 1.174 100.50
01.1.110 Oat flakes 0.450 100.0 101.50 0.488 101.50
01.1.111 Sago 0.341 100.0 101.50 0.355 101.50
01.1.112 Tennis Rolls 1.392 100.0 101.00 1.173 101.00
01.1.2 Meat 17.63 100.0 100.17 19.31 100.16
01.1.201 Stew Beef 1.940 100.0 101.00 2.162 101.00
01.1.205 Chicken (live) 1.038 100.0 101.50 1.023 101.50
01.1.206 Chicken (frozen) 10.202 100.0 102.00 11.296 102.00
01.1.207 Pork Leg 0.610 100.0 101.00 1.026 101.00
01.1.210 Corned beef* 0.866 100.0 70.20 0.949 70.20
01.1.211 Duck 0.217 100.0 102.00 0.145 102.00
01.1.212 Liver 0.207 100.0 101.00 0.172 101.00
01.1.213 Mutton 0.271 100.0 100.50 0.315 100.50
01.1.215 Brisket* 0.458 100.0 101.50 0.511 101.50
01.1.3 Fish & Seafood 5.98 100.0 101.28 6.12 101.37
… … … … … … …
207
(Lowe index or Young index). The previous index reference geometric averages of prices and price relatives rather than
period was 2010 and the NSO also has the option to main- arithmetic averages whenever weights are not available, and
tain the 2010 index reference period. In such an instance, identifies several problems with the arithmetic formulas that
the weight (2015), price (December 2017), and index (2010) are avoided when using geometric averages.
reference periods might all be different.
9.61 Many NSOs change the index reference period to Higher-level Indices
correspond to the price reference period. Often this is the case 9.66 The higher-level price indices are compiled by either
in countries where CPI revisions have historically occurred aggregating the elementary (item) level price indices or by
on a ten-year cycle or longer. Defining a new index refer- aggregating the elementary price relatives. For aggregating
ence period often is notice to users that a new basket and CPI indices, the NSO uses the item weights from the introduction of
procedures have been put in place. Users may not pay close the new CPI series. The weights for the Laspeyres and Young
attention to NSO announcements about the CPI being revised, indices refer to the weights from the weight reference period,
but when they find the new CPI index level on a different ref- while those for the Lowe index refer to the price reference
erence period, they take note and inquire about getting histori- period. These weights and their elementary (item) indices are
cal data or the revised CPI on the old index reference period. used to calculate weighted average price indices for the current
9.62 If the NSO chooses to keep the old index reference period. As noted earlier, the reference period of the weights and
period, users may not notice that the CPI has been revised. the price reference period used in the index formula determine
Most users are concerned with the overall all-items CPI and whether the index is Laspeyres, Young, or Lowe.
perhaps some of the major division indices, and do not nec- 9.67 As noted above, aggregating price relatives requires
essarily use any of the detailed indices at the group or class a different set of weights each month to derive the higher-
level. These users may not realize that the CPI has been level aggregates. The price relatives have a price reference in
revised and that some of the detail has changed because the previous period. Therefore, the aggregation weights must
new products entering, old products leaving, and, perhaps, refer to the previous period. A good rule to remember is that
an updated classification system or improved methodolo- when calculating price movements, the price period in the
gies. However, maintaining the old reference period is prob- denominator of the index formula should be the same as the
ably helpful to most users whose only interest is to monitor price period implicit in the weights.
changes in the overall CPI. Any changes in weights or meth- 9.68 The weight required is one that reflects the price ref-
odologies should be clearly explained and announced to erence period weights that are price-updated to the previous
users well in advance to avoid any confusion. It should be period. These weights may be referred to as cost weights and
clear to users that a revised index has been disseminated. All reflect what the cost would be to purchase the same quantity of
relevant metadata should be updated accordingly. the item at the current period’s prices. This weight is derived
9.63 If the NSO decides to change the index reference each period by using the price relatives for the current period
period, the old series should be linked to the new series so to bring forward the previous cost weight. These weights and
that users have the appearance of a continuous CPI series. their elementary (item) price relatives are used to calculate
Nonetheless, users should be advised that the new series is weighted average aggregate price relatives for the current
not strictly comparable with the old because of the change in period. The aggregate price relative is multiplied by the pre-
basket, weights, and often methodology. If the NSO keeps the vious period aggregate index to derive the current aggregate
old reference period, the new series must be linked to the old index.
series to give users a continuous series. Linking techniques to 9.69 The NSOs could also use geometric estimators
form a time series for the CPI are presented in paragraphs 9.73 to derive higher-level indices. Using a Geometric Young,
to 9.88. Geometric Laspeyres, or a Geometric Lowe formula could
help reduce the substitution bias inherent in the use of fixed
Formula Used for Estimation base indices. They also provide for consistency in aggrega-
tion when geometric indices are compiled at the elementary
Elementary Aggregate Indices level.
9.64 The process of updating the weights from the
weight reference period until their introduction in the CPI Linking the Previous CPI to the New
involved using a long-term price relative as the update
factor. The period-to-period estimation of the CPI can be
Price Index Reference Period
accomplished using either a long-term relative approach or a 9.70 When the NSO introduces a new basket and
short-term (two-stage or chained) approach. The long-term weights, the new series is not completely comparable to
relative approach is straightforward in that the elementary the previous series. Nonetheless, users typically want a
aggregate index for the period is derived directly as either CPI time series that covers a long period of time and pro-
the ratio of average prices in the current period to the aver- vides historical context. In order to provide such a series,
age prices in the price reference period or the average of the the NSO will need to link or splice the two series together.
long-term price relative for each observation from the price For example, if the previous series had an index reference
reference period to the current period. period of 2012 and the new series has a reference period
9.65 Chapter 8 of this Manual and Chapter 6 in Con- of 2017, there is likely to be a large difference in the index
sumer Price Index Theory strongly urge the NSOs to use levels. The CPI with 2012 = 100 will have registered price
208
changes over the period from 2012 to the end of Decem- presented here as an example of the linking process. Assume
ber 2017. The new series with 2017 = 100 will show little December 2017 is the link period and price reference period
change comparatively with the new index level close to 100. when the new weights for 2015 will be introduced. Also,
Thus, it becomes important to have procedures for NSOs to the last weight update occurred in December 2012 when the
use, in this case, to adjust the old series to reflect the level of weights for 2010 were introduced. The new CPI and the old
the new series. Alternatively, the NSO could use procedures CPI series should be produced for December 2017. If that
to adjust the new series to the level of the old series. These is the case, the linking of the series is straightforward. In
approaches are presented in paragraphs 9.74 to 9.88. December 2017 each of the new CPI indices has a value
9.71 The NSO may choose to start the new series using of 100. For the previous CPI series, each index will have a
the new price reference period. In Chapter 3 the recommen- value that could be different from the new series. The goal
dation is that when a new index is introduced there should of the linking process is to set the old indices’ levels to those
be an overlap period for the two indices so that they can of the new index. Since the new indices all have a value of
be linked together. The overlap period is used to develop 100 and the NSO wants the old indices to have that same
adjustment factors that may be applied to the old series to value, the NSO can simply reference the old series to 100
bring it to the same level as the new series. by dividing each item, subclass, class, group, etc. index by
its value for December 2017. The NSO can also derive an
adjustment factor for each of the new CPI series that users
Price Reference Period is a Single Month can apply to the new series going forward in time to raise
(Or Quarter) the new series level to that of the old series if the index ref-
9.72 At minimum, a single common period is required erence period is to remain the same as that for the old CPI
as an overlap period between the indices. When updating (December 2012 = 100).
weights less frequently (e.g., every five years), a single com- 9.74 Table 9.5 provides an example of linking the previ-
mon period is not the ideal method. However, some NSOs ous and revised CPI using a single period overlap. The table
update the CPI weights each year so that the time lapse contains the old CPI for a Division (Food and Non-alcoholic
between the weight reference period and the link month is Beverages), a Group (Food), two classes (Bread & Cereals
short. The single period link could be used in these instances. and Meat)5, and 12 items. Column D contains the old CPI
The NSO should be aware of possible distortions when intro-
ducing methodological changes. If the changes result in a
different seasonal pattern of the linking month, this would
lead to a permanent shift in the level of the index series. Para-
graphs 9.98 to 9.113 provide a detailed discussion on annual 5
These classes refer to COICOP 1999. In COICOP 2018, these classes have
weight updates. been revised to Cereals and Cereal Products and Meat, Fresh, Chilled, or Frozen.
9.73 Although not a preferred method, some NSOs 6
These classes refer to COICOP 1999. In COICOP 2018, these classes have
update the weights using a single period overlap which is been revised to Cereals and Cereal Products and Meat, Fresh, Chilled, or Frozen.
Table 9.5. Linking CPI Series Using a Single Period Overlap on a New Index Reference Period6
COICOP Description Expenditure Share Old CPI for New CPI Linking Factor Linking Factor
Code 2015 Dec. 2017 for Dec. 2017 Backward Forward
Dec. 2012=100 Dec. 2017=100 (Old Series) (New Series)
(A) (B) (C) (D) (E) (F) (G)
Col. E / D Col. D / E
01 Food and Non-alcoholic Bev. 100.00 119.88 100.00 0.8341 1.1988

01.1 FOOD 94.83 119.80 100.00 0.8347 1.1980
01.1.1 Bread & Cereals 22.41 132.49 100.00 0.7548 1.3249
01.1.101 Rice (white) 5.759 150.7 100.0 0.6636 1.5070
01.1.102 Rice (Brown) 0.381 151.7 100.0 0.6591 1.5173
01.1.103 Flour 4.356 134.8 100.0 0.7416 1.3484
01.1.104 Bread 6.167 127.2 100.0 0.7863 1.2718
01.1.105 Biscuits (Salted) 1.083 113.0 100.0 0.8846 1.1305
01.1.106 Cakes, Pastry, etc. 0.375 131.2 100.0 0.7622 1.3121
01.1.107 Chow Mein 1.370 125.1 100.0 0.7992 1.2513
01.1.109 Macaroni 0.426 105.8 100.0 0.9447 1.0585
01.1.110 Oat flakes 0.751 100.6 100.0 0.9945 1.0056
01.1.111 Sago 0.535 103.5 100.0 0.9664 1.0347
01.1.112 Tennis Rolls 0.589 126.0 100.0 0.7938 1.2598
01.1.2 Meat 17.358 111.87 100.00 0.8939 1.1187
209
in the overlap month (December 2017) and column E presents index levels (column D) to the price reference index levels
the new index in the overlap month (and new price reference (column E).
period) so that all series are equal to 100. There are two ways
to link the old series to the new series level going backward Backward Linking Factori = 1/ Link period indexi
in time. The first is to re-reference the old series by dividing
each old series by the overlap period index (December 2017 Forward Linking Factori = Link period indexi / 100
value). The second method is to calculate a “linking fac-
tor” that can be applied to each of the old series historically.
This linking factor is the reciprocal of the December 2017 Price Reference Period is a Yearly Average
index level and appears in column F. Multiplying each series 9.76 Most NSOs will establish a new price reference
by their link factor backward in time has the same effect as period using an annual average from a previous year. The
dividing by the December 2017 value. These methods are simplest and easiest method is to link the series with data for
applied to all index series in the old CPI at all levels— the month preceding the introduction of the new series (link
Division, Group, Class, Subclass, and Item. month). This involves re-referencing the old series at each
9.75 If the NSO or users want to continue the old level to the annual average index for the new price reference
series CPI going forward in time, they can produce a set period. However, there will be a discontinuity between the
of forward linking factors to use in future months as the index level for the new index and that for the re-referenced
new CPI is released. The forward linking factor raises the index level for the old series in the link month. This reflects
level of the new CPI series to that of the old series thus the difference in price trends between the old and new
keeping the series on the old reference period. The forward series as the new weights are being introduced. Table 9.6
linking factors (column G) are simply the ratios of the old shows the method for linking a new series to the old using a
Table 9.6. Linking Old and New Index Series to a Previous Annual Average6
CPI (2012=100) Forward Link CPI (2017=100) Backward Link 12-month % change
(A) (B) (C) (D) (E) (F) (G)
Year Month All Items 2017=100 All Items 2017=100 old series new series
2016 Jan 123.2 94.7 94.7

Feb 124.7 95.8 95.8
Mar 125.1 96.1 96.1
Apr 125.6 96.5 96.5
May 125.8 96.7 96.7
Jun 126.5 97.2 97.2
Jul 126.5 97.2 97.2
Aug 126.7 97.4 97.4
Sep 126.8 97.4 97.4
Oct 127.3 97.8 97.8
Nov 127.8 98.2 98.2
Dec 127.6 98.1 98.1
AVG 126.1 96.9 96.9
2017 Jan 128.1 98.4 98.5 4.0 4.0
Feb 128.1 98.4 98.5 2.7 2.8
Mar 128.4 98.7 98.8 2.6 2.8
Apr 129.4 99.4 99.6 3.0 3.2
May 129.6 99.6 99.8 3.0 3.2
Jun 130.2 100.1 100.1 2.9 3.0
Jul 130.6 100.4 100.3 3.2 3.2
Aug 131.5 101.1 100.7 3.8 3.4
Sep 131.5 101.1 100.9 3.7 3.6
Oct 131.4 101.0 101.1 3.2 3.4
Nov 131.4 101.0 100.9 2.8 2.7
Dec 131.4 101.0 100.8 3.0 2.8
AVG 130.133 100.000 100.000 3.2 3.2
Link Factors 1.00172 0.768443
2018 Jan 101.9 101.7 3.5 3.2
Feb 102.1 101.9 3.7 3.5
Mar 101.9 101.7 3.3 2.9
Apr 102.2 102.0 2.8 2.4
May 102.2 102.0 2.6 2.2
Jun 102.8 102.6 2.7 2.5
Jul 103.0 102.8 2.6 2.5
Aug 103.0 102.8 1.9 2.1
Sep 103.4 103.2 2.3 2.3
Oct 103.6 103.4 2.6 2.3
Nov 103.7 103.5 2.7 2.6
Dec 104.7 104.5 3.7 3.7
210
previous year’s annual average index as the new index refer- approach is to calculate a forward linking factor to apply to
ence period. There are three steps involved: (i) re-reference each of the new index series every month in the future.
the old index series to the new index reference period; (ii)
compile the new index series in the link month using the I wJan 2018,
= 2017
2017 =100
= Linking factori
new weight structure; and (iii) link the new series to the old
series by using forward linking factors or, if using the short-  n  (9.14)
×∑wi2017 ( I iJan 2018 / I iDec 2017 )
term price relative method, start the new series indices at the  i =1 
level of the old series in the link month.
2017 , 2017 =100 2017 , 2017 =100
Step 1: Calculate the annual average index from the Linking factori = I wDec
= 2012 / I wDec
= 2017
old series for the new index reference period which in this
example is 2017. The annual average index for 2017 with
The forward linking factor is calculated by dividing the
2012 = 100 appears in column B. The new index starts in
link month index level of the old series (December 2017)
January 2018 with 2017 = 100 and the link month is Decem-
by the link month index level of the new series: (131.4/130.133) ×
ber 2017. The old series is then re-referenced to the new ref-
100/100.8 = 1.00172. In the example, the old series link
erence period by dividing the monthly indices by the 2017
month index is in column C and the new series link month
annual average which appears in column B. The link month
index is in column D. The forward linking factor appears in
index for the old series is December 2017.
column C.
9.77 An important point to note when new series are
I wt,=2017
2012
=100
= I wt,=2012
2012
=100
I wAvg 2017 ,
= 2012
2012=100
(9.10) introduced is the calculation of the 12-month (year-over-
year) inflation rates. Series with annual overlaps provide two
The re-referencing of the old series could also be done sets of indices to use for the annual inflation rates. The old
by using a backward linking factor that is the reciprocal of series will have inflation rates calculated through the link
the new reference period average for the old index (column month. In Table 9.6, column F shows the inflation rates that
B: 1/1.30133 = 0.768443). The backward linking factor is would have been calculated using the old series (note that
in column D and the backward linked series is in column the inflation rates are the same for the original and rebased
E (note that column E is identical to the series in column series). In the example, the NSO has not published the new
C). Whereas the forward linking factor is based on a single series; it has only published the old series on the 2012 = 100
month (monthly overlap), the backward linking factor is base. The inflation rates during 2018 should then be calcu-
based on an entire year (annual overlap). lated using the rebased series (2017 = 100 in column D), as
Step 2: The indices for the new series are compiled for all appears in column G.
months of 2017 using the new weights and price reference
period. These appear in column D of Table 9.6. Keeping the Old Index Reference Period
 n  9.78 An alternative method NSOs may pursue is linking
I wJan 2017 , 2017 =100
= ∑wi2017 ( I iJan 2017 / I iAvg 2017 ) ×100 (9.11) the new CPI series to the old CPI series. This approach is
= 2017  i =1 
similar to linking the new series forward in time to main-
 n  tain the same index reference period as the old series. The
I wDec 2017 , 2017 =100
= ∑wi2017 ( I iDec 2017 / I iAvg 2017 ) × 100 (9.12) only difference is that the new index on an updated price
= 2017  i =1 
reference and weight reference period is not released to the
Step 3: Because there is a difference in the index levels public. Instead, linking factors are applied to each index
for the two series in the December 2017 link month, the new before release. For example, if the NSO kept the index ref-
series should be linked to the re-referenced old series index erence period as 2012 = 100 while starting compilation of
level in the link month. This approach maintains a continu- the new index with 2017 = 100, the forward linking factor
ous series that will reflect the short-term price movements of in column B of Table 9.7 would be applied to the new index
the old series up to the link month and the short-term price level in column D beginning in January 2018 and continue
movements for the new series following the link month. to be used until the next revision. The forward linking factor
The easier approach is to simply use the link month level is simply the old index series in the new indices’ reference
of the re-referenced old series as the starting point for the period (2017 = 100) expressed as a price relative (column B).
new series.
2017:t , 2012=100
I wJan
= 2017 = Linking factori
I Jan 2018, 2017 =100
=I Dec 2017 , 2017 =100 (9.15)
w= 2017 w= 2012 ×( I wJan 2017:t ,
= 2017
2017 =100
)
( )
n
× ∑wi2017 , Dec17 ( I iJan 2018 / I iDec 2017 ) (9.13) Linking factori = I wAvg 2017 , 2012=100
/ 100
i =1 = 2012
 I Dec 2017  Dec 2017 , 2017=100 9.79 Each period, the new index series is being tied to
Where: wi2017 , Dec 2017 = wi2017  i  I
 I iAvg 2017  w= 2017 the old index levels for item and higher-level aggregates.
Because the new index series on a new index reference
From the example, all the indices for December 2017 in the period is not published, the annual rates of change in col-
new series will be set to the index levels of the re-referenced umn F are all calculated using the published series with
old series in column C. This method is consistent with using 2012=100. If the short-term relative compilation method is
the short-term price relative compilation procedure. The other used, each index starts with its level in December 2017.
211
Table 9.7. Linking New Series to an Old Index Reference Period

CPI (2012=100) Forward Link CPI (2017=100) Backward Link 12-month % change
(A) (B) (C) (D) (E) (F)
Year Month All Items 2012=100 All Items 2017=100 2012=100
2016 Jan 123.2 94.7

Feb 124.7 95.8
Mar 125.1 96.1
Apr 125.6 96.5
May 125.8 96.7
Jun 126.5 97.2
Jul 126.5 97.2
Aug 126.7 97.4
Sep 126.8 97.4
Oct 127.3 97.8
Nov 127.8 98.2
Dec 127.6 98.1
AVG 126.1 96.9
2017 Jan 128.1 128.2 98.5 4.0
Feb 128.1 128.2 98.5 2.7
Mar 128.4 128.6 98.8 2.6
Apr 129.4 129.6 99.6 3.0
May 129.6 129.9 99.8 3.0
Jun 130.2 130.3 100.1 2.9
Jul 130.6 130.5 100.3 3.2
Aug 131.5 131.0 100.7 3.8
Sep 131.5 131.3 100.9 3.7
Oct 131.4 131.6 101.1 3.2
Nov 131.4 131.3 100.9 2.8
Dec 131.4 131.2 100.8 3.0
AVG 130.133 130.133 100.000 3.2
Link Factors 1.301333 0.768443
2018 Jan 132.4 101.7 3.3
Feb 132.6 101.9 3.5
Mar 132.4 101.7 3.1
Apr 132.8 102.0 2.6
May 132.8 102.0 2.4
Jun 133.5 102.6 2.6
Jul 133.8 102.8 2.5
Aug 133.8 102.8 1.8
Sep 134.3 103.2 2.1
Oct 134.6 103.4 2.4
Nov 134.7 103.5 2.5
Dec 136.0 104.5 3.5
Example of Calculating New Series 9.81 This aggregation results in a class index that is dif-
at the First Level of Aggregation ferent from the one obtained by simply re-referencing the
old CPI class index. In addition, the new CPI monthly class
9.80 As another example, assume the NSO keeps the
indices adjusted to the 2012 index reference period will be
2012 = 100 reference period while introducing the new CPI
different because the new weights have changed the levels
with 2017 = 100, but the revision starts at the lowest level
of the monthly indices.
of aggregation. In this case, the NSO will have to revise ele-
9.82 The example in the first box of Table 9.8 shows
mentary price indices with the first pricing period in 2017.
the monthly item and subclass indices for Oils and Fats
Assuming the CPI series is monthly, each monthly elemen-
(COICOP 2018 01.1.5) in 2017 on the old reference period
tary aggregate index (first level of aggregation) in 2017 must
(2012 = 100). It also has the annual averages7 for 2017 and
be divided by the 2017 annual average to re-reference the
indices to 2017 = 100. The re-referenced item indices (ele-
mentary aggregates) are aggregated using the new weights
to derive the higher-level indices. 7
Standard international practice for calculating annual average indices is
to use a simple arithmetic average of the monthly indices. While geometric
I im , 2017=100 = I im , 2017 / I iAvg 2017 (9.16) averaging could be used, the results can be different when aggregate indi-
ces are calculated using arithmetic aggregation. If geometric aggregation
m m were used at the higher level, then geometric annual averages should be
I wm=, 2017
2017
=100
= ∑ ( wi2017 I im, , 2017=100 ) / ∑wi2017 (9.17) used. Consistency in aggregation is important for calculating the annual
i =1 i =1 averages.
212
the re-referenced monthly indices for the class index in col- column G of the first box, again because the new indices are
umn G (i.e., dividing the monthly class indices by the annual compiled using different weights and index levels than those
average of 111.8). in the old CPI.
9.83 In the second box of Table 9.8 the monthly item
indices for the four items in the class are re-referenced to
2017 = 100 and these indices are aggregated (columns B-E) Aggregation Across Linked Series
to the class level (column F) using the new weights for 2017. 9.85 When the new series is introduced there is a
Note that the new class indices with 2017 = 100 in column break in the comparability of the historical indices. The
F of the second box are different from those in column G7 new and previous series are no longer strictly comparable
of the first box. The reason for this is that the higher-level because of the change in weights, item structures, and,
indices are compiled using different weights and index lev- in the case of re-referencing the indices, the level of the
els than those in the old CPI. indices. Each old series is re-referenced to the new index
9.84 The class indices in column G of the second box reference period. The re-referenced series will no longer
represent the new CPI on the weight and price reference for yield the same results as before the rebasing when aggre-
2017. If the NSO was introducing the new series, then the gated. This is demonstrated in column H of Table 9.8. In
indices beginning in January 2018 would be linked to the the first box of column H the new CPI series for the item
December 2017 index level in column G as shown in Table 9.6. indices are aggregated using the old item weights. There
The NSO, however, is keeping the old 2012 index reference are differences between the rebased old series (column G)
period. So it will need to use the forward linking factors to and the re-aggregated new series using old weights (col-
adjust the new CPI levels to the same level as the old CPI. As umn H) in April, May, September, and December demon-
noted previously in the discussion of Table 9.7, the forward strating the effects of the weight differences. In the second
linking factor is the old index level in the overlap period box of Table 9.8, column H, the old item indices are
(2017 average) expressed as a price relative (i.e., divided by aggregated using the new weights. There are differences
100) which is 1.118 (111.8 ÷ 100). This factor is applied to in every month reflecting the effects of both the weight
the monthly class level indices in column G of the second changes and the index level differences. Had there been a
box to tie them to the old index levels that appear in column new item added to the class, for example lard, this could
H. Note that these values are different from the values in also account for a difference.
Table 9.8 Aggregating new CPI Series using an Annual Period Overlap
2017 Butter, fresh Margarine Peanut Vegetable Old CPI Old CPI Class Old CPI aggregate
(item) (item) butter Oil (item) Oils & Fats index rebase to item indices using
(item) 2012=100 (class) 2017=100 rebased series
(A) (B) (C) (D) (E) (F) (G) (H)
Item weights (2012) 0.307 2.669 1.155 2.600 6.731

January 108.2 105.8 108.3 119.4 111.6 99.8 99.8
February 110.5 101.5 98.0 114.5 106.3 95.1 95.1
March 112.8 103.6 100.1 116.9 108.6 97.1 97.1
April 107.1 105.8 97.0 119.4 109.6 98.0 97.9
May 113.9 100.4 103.2 113.3 106.5 95.2 95.3
June 115.1 106.8 104.2 120.6 112.1 100.2 100.2
July 112.8 107.9 102.1 121.8 112.5 100.6 100.6
August 113.9 107.9 102.1 120.6 112.1 100.2 100.2
September 120.8 106.8 103.2 127.8 114.9 102.8 102.6
October 115.1 113.2 109.4 126.6 117.8 105.4 105.4
November 117.4 110.0 104.2 124.2 114.8 102.7 102.7
December 119.6 112.2 106.3 121.8 115.2 103.0 103.1
Average 113.9 106.8 103.2 120.6 111.8 100.0 100.0
Re-referenced indices (2017 = 100) New CPI Linked CPI Old base with
Item weights (2017) 0.269 1.435 1.194 2.674 5.573 2012=100 new weights
January 95.0 99.0 105.0 99.0 100.1 111.9 113.0
February 97.0 95.0 95.0 95.0 95.1 106.3 107.4
March 99.0 97.0 97.0 97.0 97.1 108.6 109.7
April 94.0 99.0 94.0 99.0 97.7 109.2 110.5
May 100.0 94.0 100.0 94.0 95.6 106.9 107.9
June 101.0 100.0 101.0 100.0 100.3 112.1 113.3
July 99.0 101.0 99.0 101.0 100.5 112.4 113.6
August 100.0 101.0 99.0 100.0 100.0 111.9 113.0
September 106.0 100.0 100.0 106.0 103.2 115.4 116.8
October 101.0 106.0 106.0 105.0 105.3 117.7 118.9
November 103.0 103.0 101.0 103.0 102.6 114.7 115.9
December 105.0 105.0 103.0 101.0 102.7 114.8 115.9
Average 100.0 100.0 100.0 100.0 100.0 111.8 113.0
213
Frequency of Weight Updates 9.88 If HBSs are conducted on a two- or three-year

cycle, the CPI compiler should give serious consideration
Data Sources for More Frequent to updating the weights as the latest HBS results become
Weight Updates available. In fact, many countries update their CPI weights
every year or every two years to minimize the substitution
9.86 The primary source for more frequent updating bias that is problematic with fixed-base indices. Such fre-
of the CPI would be the HBS; however, in some countries, quent updating of the weights keeps the CPI basket more
national accounts data are used as the source for develop- representative of consumer purchases.
ing new index weights. Countries that have a continuous 9.89 As discussed in Chapter 3, other sources for more
HBS conducted every quarter or every year can use the data frequent updating are the national accounts where data on
from the HBS to verify any significant shifts in the CPI bas- household final consumption expenditure are often available
ket and the weights. The NSO can develop a concordance on an annual basis. A retail sales survey could also provide
between the HBS and the CPI basket and review the changes indications of changes in purchasing patterns for goods and
in the basket and weight shares on a regular basis. As the services sold at retail outlets.
shares shift, the NSO could update the CPI basket as dis- 9.90 The shortcoming of these two data sources is that
cussed in Chapter 3, paragraph 3.80: they do not have the detail that is found in the HBS. Typically,
these sources may have data at the COICOP class, group, or
Even if weights are updated only every five years, it is even division levels. In such cases, a full reweighting at the
desirable to review the weights in between to ensure item level will not be possible. The NSO will need to decide
that they remain sufficiently reliable and representa- if it wants to introduce new weights at the higher level, such
tive. The review, which may be limited to weights at the as the class, group, or division levels, while maintaining the
elementary index level and their major components, weighting pattern fixed at the lower levels. The introduc-
should examine whether there are indications that im- tion of weights for the higher-level aggregates is referred to
portant changes may have taken place in the consump- as “partial reweighting” or “partial weight updates”. More
tion pattern since the weight reference period. generally, this is also an illustration how national accounts
and HBS data can be combined in practice in order to derive
9.87 There is no specific guidance on what constitutes CPI weights.
an important change, but the CPI compiler should look at 9.91 Table 9.9 presents an example of partial weight
all changes in shares that exceed ±0.5 percent on an annual updates. Assuming that the CPI compiler was able to use
basis. Such shifts may indicate significant economic changes data from the national accounts and retail sales survey to
in the pattern of consumer purchases. derive a new weight distribution at the class level, column D
Table 9.9 Partial Weight Updates at the COICOP Class Level

COICOP Description Expenditure High-Level Allocation with Updated Shares Allocation with Updated Shares
Code Share 2015 Shares 2017 2015 Shares 2017 base updated 2015 Shares 2017 base
(A) (B) (C) (D) (E) (F) (G) (H)
01 Food and Non-alcoholic Bev. 100.00 100.00 100.00 100.00

01.1 FOOD 93.31 91.90 91.90 91.90
01.1.1 Bread & Cereals 21.42 21.90 21.90 21.90
01.1.101 Rice (white) 1.406 0.06567 1.438 0.07406 1.622
01.1.102 Rice (Brown) 3.361 0.15692 3.437 0.12411 2.718
01.1.103 Flour 2.578 0.12036 2.636 0.15272 3.345
01.1.104 Bread 6.864 0.32048 7.019 0.29953 6.560
01.1.105 Biscuits (Salted) 0.813 0.03795 0.831 0.02514 0.551
01.1.106 Cakes, Pastry, etc. 1.034 0.04829 1.058 0.04596 1.007
01.1.107 Chow Mein 1.716 0.08009 1.754 0.10596 2.321
01.1.109 Macaroni 1.284 0.05993 1.312 0.06061 1.327
01.1.110 Oat flakes 0.450 0.02101 0.460 0.02519 0.552
01.1.111 Sago 0.341 0.01592 0.349 0.01833 0.401
01.1.112 Tennis Rolls 1.392 0.06498 1.423 0.06053 1.326
01.1.2 Meat 17.63 16.50 16.50 16.50
01.1.201 Stew Beef 1.940 0.11004 1.816 0.11237 1.854
01.1.205 Chicken (live) 1.038 0.05885 0.971 0.05317 0.877
01.1.206 Chicken (frozen) 10.202 0.57862 9.547 0.58705 9.686
01.1.207 Pork Leg 0.610 0.03457 0.570 0.05334 0.880
01.1.210 Corned beef* 0.866 0.04913 0.811 0.04601 0.759
01.1.211 Duck 0.217 0.01232 0.203 0.00754 0.124
01.1.212 Liver 0.207 0.01174 0.194 0.00894 0.148
01.1.213 Mutton 0.271 0.01536 0.253 0.01636 0.270
01.1.215 Brisket* 0.458 0.02598 0.429 0.02653 0.438
01.1.3 Fish & Seafood 5.98 6.70 6.70 6.70
214
contains the compiler derived share weights. The weights beginning of the year. It also will use the last pricing period
at the item level are not available so the compiler chooses of the previous year as the price reference period.8 The NSO
to keep the same allocation of the original weights at the must decide whether to price update the weights to the price
item level. This approach assumes the shares will not have reference period or simply introduce them with the new price
changed at the item level and is consistent with a Young reference period. In this example, the NSO would need to
index. Column E contains the item allocation share for each update the weights from the 2015 annual average to December
class. These are calculated by taking the 2015 expenditure 2017. The same procedures described for Tables 9.7 and 9.8
shares for each item (column C) and dividing them by the apply: (i) re-reference the old series (e.g., with December 2016
class share (column C). The item allocation share in column = 100) to the new reference (overlap) period (e.g., December
E for each item is multiplied by the new class share to derive 2017); (ii) convert the old index levels in the link month to
the updated item shares in column F. The compiler can use price relatives by dividing each by 100 to derive the forward
these updated shares in the CPI to revise the weights. linking factors; and (iii) use the linking factor for tying the new
9.92 Another alternative that the NSO could choose series to the old series level going forward in time.
for allocating the weights is to use shares derived from the 9.97 If this linking process continues for multiple years,
weights updated for price change from the original period the linking factors for each year must be derived from the
when the weights were introduced. This approach assumes indices on the fixed index reference period or made cumula-
the weights have changed because of price change but the tive by chaining the annual series through time. With the
quantities have remained the same. It is consistent with a annual linking, there are a series of one-year links going
Lowe index. Column G contains the item allocation share forward in time. Each annual link starts an index with a
for each class using the updated weights. new price and index reference period of 100. For the first
weight introduction, the annual index starts with 100; for
m
ŝsi2017 = Sc2017 ( si2015 ) the second, it too will start with 100 as will the third. If the
∑s
i =1
2015
i (9.18)
linking factors are calculated at the end of each new annual
index, they will only tie the index level to that of the pre-
These are calculated by taking the 2017 cost weights for vious annual index. For example, the NSO introduced a
each item (not shown in Table 9.9) and dividing it by the new set of weights with a price and index reference period
class cost weight for 2017. of December 2015 at the end of January 2016. In January
k
2017, new weights are introduced with December 2016 =
100. In order to keep the previous index reference period of
Sc2017 = wc2017 ∑w
c =1
2017
c (9.19)
December 2015 = 100, forward linking factors are needed
and compiled using the index values for the overlap period
9.93 The item allocation share in column G for each in December 2016. These reflect the index level of the pre-
item is multiplied by the new class share to derive the vious index period, December 2015. When yet another set
updated item shares in column H. The NSO can use these of weights are introduced in January 2018 with December
updated shares based on the 2017 cost weights in the CPI to 2017 = 100, the linking factors for the overlap period will
revise the weights. refer to the index with December 2016 = 100. In this sit-
9.94 The NSO can use either method depending on the uation, in order to adjust the new index level to that with
type of index they are compiling—Young, Lowe, or Laspey- December 2015 =100, the linking factors for December
res. Once the new weights at the item level are available, 2016 and December 2017 must each be used (chained) to
the NSO can introduce them using the methods described in get the correct index levels. The series of one-year indices
paragraphs 9.45 to 9.54. The resulting revised indices should must be chained together because each only represents the
be linked to the previous series using the methods described change over a yearly period. To get the long-term change,
in the previous section on linking the previous CPI to the these annual changes need to be made cumulative.9
new index reference period (paragraphs 9.70 – 9.85). 9.98 The aggregation to the higher-levels must always
be based on the initial indices using the most recent weights
that have not yet been multiplied by the linking factors. The
Annual Updating and Linking chain-linked indices will not be consistent in aggregation.
9.95 If data are available for updating the weights on an The published indices may have a reference period of 2015 =
annual basis, the updating procedures for the NSO to use are 100 but they are initially compiled on a more recent ref-
the same as those already presented. The primary difference erence period (e.g., December of the previous year). The
is that the process is done every year versus every five to ten linking factors are then applied to the new index to adjust
years. The NSO must determine when introducing the new to the level of the chain-linked indices. Because there are
weights what the new price reference period will be and the differences in the index levels, aggregations using the new
date for the weight update. The weight reference period is weights on the chain-linked indices give different results.
still annual (e.g., 2015) so the price reference period could
be the annual average for 2017 or December 2017. If this
is the case, the NSO will most likely decide not to update
the index reference period, and they will use procedures for 8
The new weights can be introduced in any time period. Normally, the price
keeping the same index base (e.g., 2010 = 100) described in reference period will either be the date of the weight reference period, a
calendar year, or the period prior to the introductory period.
paragraphs 9.82 to 9.83. 9
The alternative approach to using linking factors, as discussed earlier, is
9.96 If the NSO chooses to update the weights annu- to apply the short-term price relatives each month to update the previous
ally with the most recent weight data, it usually does it at the month’s indices.
215
For users to derive the correct indices for their own special samples that were priced in the previous year. Different sea-
aggregations, they must first unchain the published indices sonal patterns in the samples, combined with linking over a
by dividing them with the chain-linked index of the link single month, could lead to distortions in the annual rates of
month before they can be aggregated with the weights that the chain-linked index due to a permanent shift in the index
are applied during that year. level. Generally, this is the case when the index values of
9.99 Because of the introduction of new weights every the old and new sample strongly differ in the linking month.
year, the annual rates will not only capture changes in prices Some form of parallel calculations may be needed in order
but also changes in the weight structure. In such a situa- to measure the impact of such changes. Ideally, the NSO
tion, the contribution of each item to the all-items annual should test alternative linking approaches in order to mini-
rate of change is ill-defined and different approaches have mize any statistical distortions in the index series.
been proposed for measuring contributions or impacts (para- 9.101 The resulting annually chained indices can in
graphs 9.108 to 9.118 describe how to calculate contribu- principle be presented in any index reference period. The
tions to change).10 If the weights are introduced in January index reference period corresponds to the period for which
with the previous December month having the role of the the index is set to 100. A price index expressed for instance
new price reference period, the distortion on the annual rates as 2010 = 100 can be re-referenced or re-scaled to 2015 =
caused by the different weights vanishes in the December 100 by simply dividing the 2010 indices by the arithme-
month of that year. tic average of the 2010 indices for the 12 months of 2015.
9.100 When a methodological change is introduced Apart from rounding errors, re-referencing should have no
together with the annual weight updates, annual rates of impact on monthly or annual rates of change. The index
change can sometimes be affected by the fact that indices 12 reference period may correspond to the first price reference
months apart are calculated according to different methods. period used in the series. Another option would be to update
The samples that can be renewed together with the weight the index reference period from time to time, say every 10
update may also have different seasonal patterns than the years, or when major methodological changes are taking
place. When the index reference period is changed, it may
be useful to continue providing the results in the old ref-
10
erence period because some users may still need the index
The contribution to change must be broken down into two components. levels expressed in the old reference period.
The first component is the change from the relevant month in the previous
year to the link month using the old weights and added to this is the (second 9.102 Table 9.10 demonstrates the process for linking
component) change from the link month to the relevant month in the cur- annual indices for multiple periods. The new index refer-
rent year using the new weights. ence period in January 2015 was December 2014 = 100
Table 9.10 Linking Annual Indices for Multiple Periods with Chained Linking Factors
COICOP Description CPI Dec. CPI Dec. 2015 Linking Factor CPI Dec. 2016 CPI Dec. 2016 Linking Factor CPI Dec. 2017 CPI Dec. 2017
Code 2014=100 with Dec. Forward (2016 with Dec. with Dec. Forward (2017 with Dec. with Dec.
2014=100 Series) 2015=100 2014=100 Series) 2016=100 2014=100
(A) (B) (C) (D) (E) (F) (G) (H) (I) (J)
01 Food and Non- 100.00 108.99 1.089859 103.69 113.01 1.130084 104.33 117.90
alcoholic Bev.
01.1 FOOD 100.00 108.91 1.089085 106.29 115.76 1.157578 106.87 123.71
01.1.1 Bread & Cereals 100.00 120.45 1.204490 103.75 124.96 1.249616 115.37 144.16
01.1.101 Rice (white) 100.0 137.0 1.370003 135.4 185.4 1.854301 171.19 317.44
01.1.102 Rice (Brown) 100.0 137.9 1.379356 95.6 131.8 1.318197 121.70 160.42
01.1.103 Flour 100.0 122.6 1.225822 136.2 167.0 1.670136 154.19 257.52
01.1.104 Bread 100.0 115.6 1.156216 94.7 109.4 1.094450 101.04 110.58
01.1.105 Biscuits (Salted) 100.0 102.8 1.027706 59.6 61.3 0.612767 56.57 34.67
01.1.106 Cakes, Pastry, etc. 100.0 119.3 1.192797 99.4 118.6 1.186214 109.51 129.91
01.1.107 Chow Mein 100.0 113.8 1.137564 131.8 150.0 1.499635 138.45 207.62
01.1.109 Macaroni 100.0 96.2 0.962266 85.3 82.0 0.820352 75.74 62.13
01.1.110 Oat flakes 100.0 91.4 0.914144 96.0 87.8 0.877843 81.04 71.14
01.1.111 Sago 100.0 94.1 0.940671 94.9 89.2 0.892387 82.39 73.52
01.1.112 Tennis Rolls 100.0 114.5 1.145302 93.4 107.0 1.070218 98.80 105.74
01.1.113 Whole Wheat Bread* 100.0 115.6 1.156216 94.7 109.4 1.094450 101.04 110.58
01.1.2 Meat 100.00 101.70 1.016968 108.42 110.26 1.102645 101.80 112.25
01.1.201 Stew Beef 100.0 100.5 1.004750 108.4 109.0 1.089502 100.58 109.59
01.1.205 Chicken (live) 100.0 102.3 1.023498 97.7 100.0 1.000247 92.34 92.37
01.1.206 Chicken (frozen) 100.0 100.2 1.002088 107.4 107.7 1.076706 99.40 107.03
01.1.207 Pork Leg 100.0 126.2 1.261643 205.7 259.5 2.595448 239.62 621.91
01.1.210 Corned beef 100.0 101.7 1.016968 100.7 102.4 1.023703 94.51 96.75
01.1.211 Duck 100.0 97.5 0.975450 63.1 61.5 0.615484 56.82 34.97
01.1.212 Liver 100.0 104.9 1.049449 84.5 88.7 0.886868 81.88 72.61
01.1.213 Mutton 100.0 97.1 0.970904 109.2 106.1 1.060580 97.91 103.85
01.1.214 Sausages (pork & 100.0 109.3 1.093397 99.1 108.4 1.083866 100.06 108.46
chicken)
01.1.215 Brisket* 100.0 100.5 1.004750 108.4 109.0 1.089502 100.58 109.59
01.1.3 Fish & Seafood 100.00 96.93 0.969334 96.98 94.00 0.940030 86.78 81.58
216
(column C). The NSO proceeded to update the weights provide exactly the published rate of change. However, esti-
in January 2016, January 2017, and January 2018, while mates are possible as shown in the section that follows on
maintaining the original index reference period of Decem- annual weight updates. When the comparison period t-m to
ber 2014. In January 2015, new weights were introduced t crosses over a link period k, then the contribution must be
from 2013 and the new price reference period was set to calculated separately for each period (t-m to k and k to t)
December 2014 = 100. This is also the overlap month, so the and combined.
indices are converted to long-term relatives and serve as the
linking factors (column E) to tie to the next year’s indices. Annual Weight Updates11,12
In January 2016 weights were introduced from 2014 and 9.107 When expenditure weights are annually updated,
the new price reference period was set to December 2015 = price indices spanning more than a year are chain-linked and
100. The link factors from the previous year (column E) are the formula to compute contributions to inflation between
applied to the new index levels for December 2015 (column month m of year (y-1) and month m of year y needs to be
F) to derive the current year index levels (column G) with modified. The annual inflation rate denoted by 𝜋𝑇𝑂𝑇𝑦,𝑚
December 2015 = 100. These indices are then converted to can thus be decomposed as follows, assuming that each year
linking factors to use with the 2016 indices (column H). The the new weights are introduced with the December to Janu-
same actions were taken in January 2017 with new weights ary link:
from 2015 and a new index using December 2016 = 100.
y,m y −1, m
The previous year linking factors (column H) were applied y,m ITOT y, m − ITOT
πTOT ≡
to the December 2017 index (column I) to derive the index y −1, m
ITOT
for December 2017 with December 2014 = 100 (column J).
 y,m   y −1, m 
9.103 The index levels in columns F and I reflect the y −1,12
ITOT  w y−1,12 I j  − ITOT
y −2 ,12
 w y−2,12 I j
levels of each index referencing the previous December = y −1, m
ITOT
∑j  j I y−1,12  I y−1,m ∑j  j I y−2,12 
(2015 and 2016, respectively). They are not calculated on  j  TOT  j 
the December 2014 reference period. This is accomplished y −1,12
ITOT   I y−1,12   y −1,12
ITOT
by linking each of these periods together successively. As = y −1, m
ITOT
∑j W jy−1,12  ITOT
y −1,12 
 −1 +
  I y−1, m
can be seen in Table 9.9, the chaining of the linking factors   j  TOT

is similar to chaining the annual indices of the three-year =
y−2 ,12
ITOT  I jy−1,12 
∑ W jy−2 ,12 I y−2 ,12 
period (December 2014 to December 2016). For example,
y−1,m
ITOT j  j 
the December 2017 index for the Food Group (COICOP I y −2 ,12  I jy−1, m 
1999 01.1) of 123.71 is the product of chaining the price −
I
TOT
y −1, m ∑j W jy−2,12 I y−2,12 
changes (indices converted to price relatives) for the three TOT  j 
years (1.089085 × 1.0629 × 1.0687) to the December 2014
reference period index of 100. y −1,12   y , m 
ITOT W y−1,12  I j 
= y −1, m
ITOT
∑j  j  I y−1,12 
Calculating Cntributions to Change   j 
(9.21)
9.104 Contributions to change help explain those I y−2,12  I jy−1,12 − I jy−1, m 

groups of goods and services that contribute most to infla- + TOT
y −1, m
ITOT
∑W j
y −2 ,12
 I jy−2,12


tion. These data are useful to better understand the sources of j 
inflation and can contribute to greater transparency. Whether
the weights are fixed for a period of time or updated annu- 9.108 The contribution of component j to the overall
ally, different formulas would apply. annual rate of inflation in month m of year y can be writ-
ten as follows. By construction, these contributions sum to
Fixed Weight Indices overall inflation.
9.105 The formula used to calculate the contributions
   
from the aggregates to the total index is as follows:  y−1,12   I y , m − I y−1,12 
 ITOT 
=  y−1, m *W jy−1,12  * 
j j
Contrib jy , m 
wit0 * ( I it − I it −m )  ITOT   I jy−1,12 
Cit = (9.20)  2
( ) 
   
t −m
ITOT  (1)   ( 3) 
(9.22)
where    
 y−2,12   I y−1,12 − I y−1, m 
Cit : contribution of aggregate i in period t I y −2 ,12   
+  TOT * W  *  j y−2,12j 
wit0: weight of aggregate i in period t0  ITOT y −1, m j   I
 (5)   j 
I it : index of aggregate i in period t ( 4)   (6) 
I it −m: index of aggregate i in period t-m
t −m
ITOT : total index in period t-m
9.106 The addition of individual contributions is equal 11

to the published rate of change. It should be noted that this For detailed methodology, see http://www.oecd.org/sdd/prices-ppp/OECD-
calculation-contributions-annual-inflation.pdf
formula may only be applied if the weights remain con- 12
See also Harmonized Index of Consumer Prices (HICP) Methodological
stant. When the weights change, in the case of a chained Manual, Section 8.5.3, https://ec.europa.eu/eurostat/web/products-manuals-
index, the additivity of individual contributions does not and-guidelines/-/KS-GQ-17-015
217
Notations: With m = 1, the formula 9.27 gives the effect of a monthly

Wjy−1,12 is the expenditure weight corresponding to change; with m = 12, it gives the effect of the change over
COICOP component j in December of year (y-1) and used the past 12 months.
for the link from December of year (y-1) until December 9.112 If the weights have been updated, then a sub-
of year y; index, j, from t-m enters the higher-level index with a weight of:
ITOT, is the overall price index in month m of year y;
Ij, is the price index of component j in month m of year y; wkj I kj :t −m wkj ( I 0j :t −m I 0j :k )
= (9.27)
9.109 The first part of the above formula (terms 1 to 3) consid- I k :t −m ( I 0:t−m I 0:k )
ers price developments that occurred between December of year
9.113 The effect on the higher-level index of a change
(y-1) and month m of year y, whereas its second part (terms 4 to
in a sub-index then is:
6) takes into account price developments that occurred between
month m of year (y-1) and December of year (y-1). This cut wkj
in December of year (y-1) is introduced due to the change in Effect =
I k :t −m
( I kj :t − I kj :t−m )
expenditure weights that is introduced after this month.
9.110 Note that when m = 12 (December), the second  I 0j :t − I 0j :t −m  (9.28)
wkj 
part of the formula is equal to zero and the first part simpli- = 
0:k  
fies as follows, which leads to the same contribution formula ( I 0:t−m I )  I 0:j k
as in the fixed basket case when y0 = y − 1 and m0 = 12:
It is assumed that t-m lies in the same link (i.e. t-m refers
I jy ,12 − I jy−1,12 to a period later than k). If the effect of a sub-index on a
Contrib jy ,12 = W jy−1,12 * higher-level index is to be calculated across a chain, the cal-
I y−1,12 (9.23)
( 2) j culation needs to be carried out in two steps: one with the
≡ π yj ,12 :( 3) old series up to the link period, and one from the link period
to period t.
9.111 Alternative approaches to calculating the contri- 9.114 The calculation of the effect of a change in a sub-
butions to change are possible. First, when the weights are index on a higher-level index is illustrated in Table 9.11. The
fixed, the relative change of the index from t-m to t can then index is calculated in one link so that equation 9.26 may
be written as: be applied for the decomposition. For instance, the effect
in percentage points of the increase for housing from Janu-

I 0:t
−1 =
∑w I b 0:t −m t −m:t
j j j I
−1 (9.24) ary 2018 to January 2019 can be calculated as 0.25/118.6 ×
I 0:t −m
∑w I b 0:t −m
j j
(120.0 – 110.0) = 2.11 percentage points. This means that,
of the increase of 10.03 per cent in the all-items index, 2.11
Hence, a sub-index, j, from t-m to t enters the higher-level percentage points can be attributed to the increase in the
index with a weight of: index for housing.
wbj I 0j :t −m wbj I 0j :t −m Introducing New Classification

=
∑ wbj I 0j :t−m I 0:t −m (9.25)
Systems
9.115 Chapter 2 recommends that countries adopt the
The effect on the higher-level index of a change in a sub-
United Nations’ COICOP. Many countries that have been
index can then be calculated as:
using a national product classification system will want to
introduce the COICOP at the time of their next CPI revi-
sion in order to meet the recommendations of this Manual.
wbj I 0j :t −m  I 0j :t  wbj The introduction of COICOP may cause a break with the old
 I 0:t −m ( j
Effect =  − 1  = I 0:t − I 0j :t −m ) (9.26) classification system. For example, COICOP 2018 classifies
I 0:t −m  I 0j :t −m
household final consumption expenditure into 13 Divisions
Table 9.11 Decomposition of Index Changes

Weight Index Change in % Effect (contribution)
from Jan. 18
2015 Jan. 18 Jan. 19 % points of % of total
to Jan. 19
total change change
Food 0.30 100.0 120.0 130.0 8.33 2.53 25.21

Clothing 0.10 100.0 130.0 145.0 11.54 1.26 12.61
Housing 0.25 100.0 110.0 120.0 9.09 2.11 21.01
Transport 0.20 100.0 125.0 130.0 4.00 0.84 8.40
Miscellaneous 0.15 100.0 114.0 140.0 22.81 3.29 32.77
Items 1.00 100.0 118.6 130.5 10.03 10.03 100.00
218
as shown in Annex 2. Countries who have not adopted in the basket and may be replaced with Bluetooth speakers
COICOP often have classifications with 9 or 10 Divisions. and other audio media. In addition, electronic items such
Both classification systems will cover the detailed items, but as tablets and e-readers may be added. In some instances,
they are often classified in different Divisions. One com- owner-occupied housing may now be represented whereas
mon area of difference is food eaten at restaurants, cafes, only rental units were included previously.
canteens, and kiosks. COICOP includes these expenditures 9.120 When the NSO does not change the index refer-
in Division 11 Restaurants and Accommodation Services, ence period (i.e., keeps the old reference period), it is less
while the national classification may include these in Divi- obvious to users that a change has taken place. This is partic-
sion 01 Food and Non-Alcoholic Beverages. ularly true when it has been several years since the last CPI
9.116 When the NSO is ready to introduce the revised revision. If the weights are updated annually, users may be
CPI using the COICOP, there will be a difference in the old aware of the change in weighting; however, when revisions
and new classification structures for the published indices. only take place every five to ten years, the NSO needs to
Some Divisions such as Clothing and Footwear and Health make extra efforts to ensure that users are aware of changes
may be the same. Others such as Furnishings, Household and discontinuities in series.
Equipment, and Routine Household Maintenance (where
televisions and video equipment are moved to Entertain-
ment).will have differences in the detailed components
Expanding CPI Geographic
9.117 In most instances, the old CPI includes the same Coverage
detailed components as the COICOP, but the components 9.121 Many countries have CPIs with limited geo-
appear in different divisions. If this is the case, it may be graphic coverage—capital city, two or three of the largest
a straightforward task to develop a concordance between areas, large and medium-sized cities—when they are first
the old classification and the COICOP. The old index series developed. Over time as the CPI becomes more important in
at the detailed level can be recompiled using the old series economic planning and inflation monitoring, and population
weights to reflect the COICOP structure at the 13 Division centers expand, efforts are made to expand the CPI to cover
level. The recompiled old series can then be linked with more geographic areas including all urban and rural areas.
the new COICOP series to form a time series going back Generally, this expansion takes place by increasing the
to the point of the previous revision. For example, assume coverage of the HBS so that representative baskets can be
the last revision was on a 2010 =100 reference period and developed for more geographic areas.13 As new geographic
introduced in January 2013. These series can be recompiled areas are added to the CPI, the comparability between the
to form COICOP Division level indices for January 2013 previous index and the revised index becomes questionable.
through December 2017. The new COICOP series is intro- In these cases, linking the aggregate indices to the previous
duced in January 2018 with 2015 = 100 reference period. measures needs to be done with caution and users advised
The old series can be linked to the new series at the Division about the differences in coverage.
and Total CPI levels using 2015 = 100 as the overlap period. 9.122 In some countries, NSOs first produce a CPI for
The old series can be re-referenced to the 2015 annual aver- the capital city only and, as resources become available for
age = 100 if a new reference base is introduced. subsequent CPI revision, they expand to cover more areas.
9.118 Alternatively, if the NSO decides to keep the As the revised CPI is introduced, estimates should be pro-
2010 =100 reference period, the changes in the revised CPI vided for both the capital city and for the new areas in addi-
can be applied to the old index series at the COICOP Divi- tion to the all-items index. In such a case, the capital city
sion level. In this latter instance, the new series must be index for the revised CPI can be linked to the previous capi-
compiled for both December 2017 and January 2018 on the tal city index with meaningful results in that the geographic
2015 = 100 reference period and the monthly price relative coverage has not changed even though the structure of con-
for January 2018 is applied to the recompiled old series at sumer spending patterns may differ.
the Division and Total CPI levels on the 2010 = 100 refer- 9.123 However, linking the revised CPI with expanded
ence period. This process is carried out each month going geographic coverage (perhaps with two or more new areas)
forward in time. Note that the new weights are applied to the needs to be done with caution. The NSO assumes when
revised index levels with 2015 = 100. They are not applied it produces such a linked series that the historical price
and are not relevant for the old series with 2010 = 100. The changes for the new coverage would be the same as that
new weights refer to the CPI structure in 2015, not the struc- for old coverage. The NSO can evaluate the differences in
ture from 2010. the baskets among the new areas with that for the old area
9.119 It is important for the NSO to advise users about to determine how similar or different the baskets and price
changes that are made to the CPI, particularly when a revi- trends might be. If there are significant differences between
sion takes place. When the NSO updates the index reference the areas (e.g., the capital city may have substantially more
period, users will generally take note of the change and ask weight for housing, clothing, and education, and substan-
about the difference between the old and revised series. The tially less for food and transport than the new areas), then
NSO should provide users with the information about all
the changes taking place including new weights, new item
structure, and improvements in methods and procedures.
Users also need to be aware that the series are not strictly 13
In some cases, the expansion may take place by using an existing basket
comparable. Some items may have been dropped from cov- such as that for the capital city or urban areas in conjunction with broader
erage and others added. For example, radios, tape record- data collection in targeted geographic locations. In such instances it is
ers, and stereo systems may no longer be important items assumed the basket is very similar across areas as described in Chapter 3.
219
the NSO should not link the aggregate old and new series. Key Recommendations
Rather, they should continue to produce the CPI series for
comparable geographic areas and link these series to form a • NSOs should update the basket, update the weights, and
time series. The new geographic areas and the new aggre- link the revised CPI to the previous series to form a con-
gate CPI should be published separately as new series and tinues time series of data.
not linked with old series. • Reduce the lag between the weight reference period and
9.124 At the same time, if the CPI already had broad the price reference period.
coverage such as all urban areas and the weight structure
• NSOs should carefully consider the advantages and dis-
included all urban areas, then the aggregate series for the
advantages of price updating the weights, based on the
all-urban CPI might be linked together as the city sample
target of the CPI and its primary uses.
is expanded. For example, the previous CPI could have
included five urban areas (but their weights represented • If a CPI is updated less frequently, a one-year price refer-
all urban areas) and now the number of urban areas in the ence period is preferred.
city sample is expanded to include eight urban areas whose • If a CPI is updated on a more frequent basis (annu-
weights also cover all urban areas. In this instance, the NSO ally, biannually), a one-month price reference period is
could still derive meaningful comparisons for price change used.
for all urban areas by linking the all urban index for the old • If resource constraints do not allow for a one-year price
series with that for the new series using the larger sample. reference period, a quarterly price reference period would
The basket has changed so the series are not strictly com- be the second-best option and a monthly price reference
parable, but the new, larger sample would have more preci- period the third best option.
sion because of the increased sample size compared to the • A one-year index reference period is preferred for all
old. If city indices were also produced, those with the same CPIs.
geographic coverage could be linked to form a time series.
9.125 If the NSO went one step further and included total • When using a one year overlap to link indices, the new
national coverage by adding a sample of rural areas to the index compiled as an overlap series for linking pur-
CPI, it would need to do an analysis of the differences in the poses should be used internally and not disseminated.
baskets and price trends before linking the old urban series to When calculating the 12-month (year on year) change,
the new national coverage. In this instance the NSO would the change should be calculated on the published linked
not link the urban and national series together; the national index and not the internally compiled new index used
series should start as a new CPI with broader coverage. for linking.
220
SCANNER DATA 10
Introduction their locations and include both quantities sold and revenue
received by the retailer for these items. This information
10.1 The environment in which statistical agencies oper- has the potential to: improve the accuracy of the prices used
ate is changing. New opportunities to access and interrogate to compile the CPI by calculating unit values for homoge-
big data are becoming available, increasing the potential to neous products (see paragraphs 10.26–10.61); improve the
provide new insights and possibilities for the compilation samples of items priced, with the potential to use a census of
of consumer price indices (CPIs). The statistical landscape items for the product categories and outlets covered by the
is becoming more complex, expectations of decision mak- scanner data, for example, the product categories in the data
ers are growing, and national statistical offices (NSOs) are provided by a particular supermarket chain; and use quantity
being challenged to deliver the best possible statistical pro- or revenue information to weight items according to their
gram in more efficient and innovative ways. economic importance. Scanner data will typically not cover
10.2 The launch of barcode scanner technology during the entire universe that is in scope of the CPI. For example,
the 1970s, and its growth in the twentieth century, enabled in most countries, scanner data do not cover services, rents,
retailers to capture detailed information on transactions at automobiles, restaurants, or cafes. In addition, this informa-
the point of sale. Scanner data are high in volume and con- tion may only be available for large retail chains but not for
tain information about individual transactions, including small independent stores or other types of outlets.
date, quantities and values, and detailed product character- 10.6 While scanner data sets present opportunities to
istics. The data provided to NSOs are typically aggregated improve the accuracy of the CPI, there are also challenges
across consumers and across time (week or month). It is that need to be addressed before NSOs can use scanner data
a rich data source that can potentially be used to improve to compile the CPI.
the accuracy of the CPI and reduce the costs of physically
collecting data; however, it may increase the costs for data
analysis and processing. Also, scanner data give opportuni- Obtaining Scanner Data Sets
ties for improved quality in other aspects, such as publishing 10.7 Scanner data have existed for several decades
more detail and new data products. and their value in the compilation of official statistics has
10.3 This chapter discusses the opportunities and chal- become more and more evident over time. One challenge
lenges presented when using scanner data to compile the CPI faced by NSOs is obtaining the scanner data sets. Two main
and aims to provide guidance to NSOs on different practices options are available. NSOs may seek the supply of scanner
for the treatment of scanner data. It outlines several practical data sets directly from retail businesses or from third-party
considerations regarding the acquisition of scanner data sets, data providers. Both options present benefits and challenges.
the assessment and preparation of the data, and implementa- 10.8 Several NSOs have successfully negotiated the
tion issues. New methods that have been developed to con- supply of scanner data directly from retail businesses and
struct price indices from scanner data, so-called multilateral used these data in the compilation of their CPI. Direct col-
methods, are presented. This chapter concludes with a dis- lection of data sets from the retail business by NSOs has
cussion of the assessment of the new methods and the empir- potential benefits. These include the ability to negotiate:
ical results, communication with users and stakeholders, and
• The supply of the data set at no (or minimal) cost
publication and dissemination of the “new” price indices.
10.4 Given the dynamic nature of using scanner data • The scope of items included in the data set
for CPI compilation, this chapter provides an overview of • The level of item aggregation to ensure homogeneous
country practices. Work on the treatment of scanner data information
continues, especially with regard to developing elemen- • The temporal coverage and detail (day, week, or month)
tary aggregate formulas appropriate for compiling an index • An agreed timetable for the supply of the data set to meet
using scanner data. The treatment of scanner data has been CPI processing requirements
included on the CPI research agenda for further discussion
and development. • A contact officer within the retail business who is familiar
with the data set to answer NSO data queries
Practical Considerations 10.9 Negotiating the supply of scanner data sets directly
with retail businesses presents challenges as well. The pri-
Introduction mary challenge is that the bilateral negotiation of scanner
10.5 The availability of scanner data provides oppor- data sets concerns data that may be regarded as confidential
tunities to improve the CPI. Scanner data sets typically because they contain information on turnover and quantities
contain complete coverage of items sold by a retailer at all at items level. Another factor is the legal and institutional
221
setting that governs the relationship between NSOs and the burden on retailers and enhance the timely delivery of
retailers. In some countries, it may be necessary that the data to the NSO. IT system development requires human
(statistical) law stipulates which data are to be supplied, and financial resources. Several NSOs have documented the
whereas in other countries a verbal agreement between the challenges presented by the need to develop an IT system.
parties is sufficient. Experiences in countries using scanner The solution is dependent on individual NSO circumstances.
data suggest these negotiations will take at least six months NSO resources will be required for an IT system if the NSO
to complete. The negotiations relate to a wide range of top- is to utilize scanner data to compile the CPI, irrespective of
ics: from information technology (IT) systems and reporting the data provider.
formats to confidentiality concerns. An agreement reached 10.14 Given these large investments, it is important for
between the NSO and retail business is typically formal- the NSOs to gain experience with test data (for example,
ized in a memorandum of understanding (or similar) which from market research companies) and engage with other
documents the roles and obligations of each party and aims NSOs with experience. The system needs to be designed to
to ensure an ongoing supply of scanner data to the NSO last some time, while at the same time being able to adapt
according to the agreed timetable. to evolving methodological developments and being scal-
10.10 An alternate approach to obtaining scanner data able to cater for the large volume of data received from an
sets directly from retail businesses is to obtain these data sets increasing number of data providers. It is recommended
from intermediaries or market research companies. Market that, regardless of the methods that are implemented to pro-
research companies possess scanner data sets obtained by duce the CPI, the system remains easily adaptable to newer
some NSOs for CPI assessment and compilation purposes calculation methods as they evolve.
(Krsinich 2015). Market research companies have no legal
obligation to provide scanner data; however, for a fee they Classifying Scanner Data
may be willing to provide older scanner data that would 10.15 Scanner data sets generally include product clas-
allow the NSO to explore and become more familiar with sifications that are unique to the individual retailer. The NSO
the data. A better understanding of these data may clarify will most likely receive data sets that contain different prod-
the requirements before starting negotiations with retailers uct classifications which need to be mapped to a single CPI
or market research companies. The primary benefit of this classification. The classification of scanner data sets may
approach is the ability to negotiate the supply of multiple require significant NSO resources. The largest investment
data sets relating to a diverse set of products with a single or of resources is needed when the data sets are first received
small number of data providers. by the NSO. However, there is a need for ongoing classifica-
10.11 The experience of NSOs using scanner data sets tion resources as new products enter the data set.
to compile their CPI suggests that obtaining data sets directly 10.16 Classifying scanner data items to the CPI classi-
from retail businesses is preferred. However, obtaining scan- fication has been addressed by NSOs in various ways, many
ner data sets from market research companies is beneficial of them using the classification of the individual retailer.
in those cases where securing data sets directly from retail Such classifications provide important information and can
businesses is not possible or resources are not available to be very useful if they are at the same (or more) level of detail
negotiate bilateral data supply agreements. Accessing scan- than the lowest level of the Classification of Individual Con-
ner data from market research companies, in most cases, sumption According to Purpose (COICOP) If the correspon-
requires resources to purchase these data. dence is 1:1 or n:1 (retailer:COICOP), then scanner data can
be mapped automatically. In other cases, scanner data either
need to be classified by the NSO or the data are excluded.
Assessing and Preparing Scanner Data From time to time, the retailer may also change its classifi-
for Use cation. The IT system and the classification process should
10.12 If the NSO is successful in securing access to be designed to be flexible so that changes in retailer classifi-
scanner data sets, these data sets should then be turned into cations can be handled in a timely manner.
information that can be effectively and efficiently used to 10.17 NSOs have attempted to find a solution given
compile the CPI. NSOs need to overcome several challenges their circumstances. Some countries have classified scanner
to achieve these objectives. data items to the CPI classification by purchasing market
research metadata (Müller 2010). One European NSO uses
Developing an IT System the most detailed classification provided by the retailers and
10.13 Scanner data are by their very nature big data. then checks if the mapping is correct and makes appropri-
The size of the files depends on the characteristics of the ate changes if required (van der Grient and de Haan 2010).
underlying data. For instance, files with daily data by outlet Some NSOs have, for various reasons, undertaken the entire
will be more voluminous than files with weekly data aggre- classification of scanner data items to their CPI classifica-
gated at the retail chain level. The use of this information by tion internally (Howard and others 2015).
the NSO requires an IT/computing system that can acquire, 10.18 Several NSOs have been exploring the use of
store, and process the large scanner data sets if the informa- machine learning algorithms for classifying scanner data
tion is to be used to compile the CPI. The IT system needs (see Van Loon and Roels 2018). These methods use an
to be able to acquire and process data sets that have dif- input data set either of prelabeled items (supervised learn-
ferent classification structures, formats, and contents. This ing) or of unlabeled items (unsupervised learning) to pre-
is because retail businesses (and third-party data providers) dict the correct taxonomy label for each item. The resulting
normally develop unique systems for their own internal model can then be used to classify new data sets. Machine
reporting purposes. These data sets can be used to minimize learning methods are particularly promising where there is a
222
Scanner Data
mismatch between the product classifications used by retail- Implementation—From Confrontation to

ers and the classification used for CPI compilation. As with New Methods
all classification methods, ongoing maintenance is required
to ensure that items with new features that have not previ- The Benefits and Challenges of Using Scanner
ously been identified are classified appropriately. Data
10.19 The challenge of classifying scanner data items 10.27 The use of information contained in scanner data
to the CPI classification primarily arises when scanner sets to compile the CPI can represent a significant change
data sets have been secured directly from retail businesses. to the data collection practices and the price index methods
Obtaining scanner data sets from market research compa- traditionally employed by NSOs. This suggests that these
nies may enable the NSO to negotiate the supply of scanner changes need to be carefully managed, both with regard to
data that have already been classified according to the CPI the impact on the statistical program as well as communica-
classification. This is viewed by some NSOs as a particular tion with users, key stakeholders, and staff. NSO staff need
advantage of obtaining scanner data from market research to understand how to manage these data because scanner
companies. data sets are so much larger than the traditional CPI data sets.
10.20 The reliability of the classification system needs Traditionally, the influence of each price could be traced,
to be continuously monitored. Errors that are made at this and often needed to be visible. In dealing with scanner data,
stage will be reflected in the resulting subindices that may such attention to detail may not be feasible, and these data
then be compiled based on wrongly classified items. may require a more top-down approach. Communication
with users and key stakeholders is important. It is important
Quality Assurance of the Scanner Data Sets for data users to fully understand how scanner data are used
10.21 Compared with traditional price collection in out- in the compilation process. This enhances transparency and
lets, scanner data sets are a new data source to compile the user confidence.
CPI. As is the case with any change in data source, the com- 10.28 Scanner data potentially enhance the accuracy
pilers of statistical series should undertake a range of checks of the CPI in several ways and provide significantly more
to ensure the new data source provides the foundation from data at lower cost. The scanner data sets can be used to (1)
which to produce fit-for-purpose statistics. compare and validate price data; (2) replace field-collected
10.22 The checks should become routine and per- prices (including a better treatment of sales, promotions,
formed automatically each production run. Because scan- and discounts); (3) expand pricing samples; (4) expand the
ner data sets are new, it is important that NSOs gain some period over which prices are collected; (5) weight items at
experience with them before using these data in production. the lowest levels of the CPI to reflect their economic impor-
The experience gained will facilitate setting the values for tance; and (6) implement new improved index calculation
the checks. methods and enable process automation. Each of these
10.23 These scanner data checks can be classified as enhancements is explained in the following text.
either global checks or detailed checks. Global checks occur
when the data enter the production process and are part of Using Scanner Data Sets for Data Validation
the acceptance procedure. Detailed checks typically occur and Quality Assurance
toward the end of the production process. 10.29 The availability of scanner data provides NSOs
10.24 Global checks relate to broad quality measures with the opportunity to validate or quality assure the data
generally applied at the time the NSO receives the data set. used to construct the CPI. Scanner data sets contain vari-
These checks ensure the data set is broadly consistent with ety quantities sold and revenue received by the retailer for
data sets received by the NSO from the same data provider these varieties for some period of time, usually a week or
in previous periods. The checks may relate to the format of a month. This information enables NSOs to calculate a
the data set, the total number of items within the data set, price for an individual variety by dividing a variety’s rev-
and the total revenue by outlet. These global checks should enue by the quantity sold. This price, referred to as a unit
highlight significant errors with the data set. value, represents the average price experienced by consum-
10.25 Detailed checks are generally applied at the vari- ers over a period of time. Note, however, that revenue data
ety or item group level. These checks aim to highlight signif- may not align perfectly with the purpose and concept of the
icant changes in the quantities sold, revenue, and the prices national CPI because it may include expenditure by nonresi-
of the items within the data set. These detailed checks have dent households, businesses, or even government (Fenwick
traditionally been referred to as micro-editing of price data. 2014).
Unexpected changes in the development of prices, turnover, 10.30 For a homogeneous item, the unit value more
or quantities will trigger these checks. Processing scanner accurately reflects prices paid by consumers over the whole
data means processing much larger sets of data and it may period than point-in-time pricing (Balk 1998). Unit values
require a different approach than processing traditionally contain discounts and the effects of these discounts on the
collected data. quantity of varieties sold. The period for which unit values
10.26 Both the global and detailed checks should be are calculated is important with regard to the accuracy of
automated to generate reports for analysis by NSO staff. the unit value. Diewert and others (2016) argue that unit
These checks may require contact with the data provider, value prices used for constructing the CPI should be for the
as well as comparing the data with alternate price informa- same period as the index to be constructed, rather than for
tion sources (for example, flyers and online prices). The a subperiod.
final compiled indices should be reviewed and validated to 10.31 It is acknowledged that NSOs may use a subpe-
ensure plausibility. riod of the reference period due to data supply timeliness
223
and publication deadlines. The bias and variance this intro- 10.37 An essential part of price measurement is
duces can be assessed by comparing indices compiled using accounting for quality change and the introduction of new
a subperiod of data with indices compiled using the full ref- items. This has been achieved by NSOs when field officers
erence period (Krsinich 2015). visit retail outlets with the aim of measuring price change
10.32 Price analysts can compare the prices collected in for identical or equivalent items in successive periods and
the field to those calculated from the scanner data sets. This identifying new items. As the characteristics of varieties are
analysis provides insight into any biases introduced to the altered, the NSO field officers collect descriptive informa-
CPI from point-in-time pricing compared with unit values. tion that enables the effects of a change in quality to be sepa-
An analysis of the variety’s revenue and quantities sold can rated from the price change, so that the CPI measures only
be used by the NSO price analysts to highlight where CPI pure price change.
samples could be improved. 10.38 Accounting for quality change is particularly
challenging when using scanner data. Scanner data sets tend
Using Scanner Data Sets to Replace Field- to exhibit a high level of churn in the varieties available from
Collected Prices month to month. There are new models (and versions of
10.33 In most countries, the majority of the prices used models) of items becoming available in the market and old
to compile the CPI are collected by visits to sampled retail models dropping out of the market as they become obsolete.
businesses. These visits are made by NSO field officers who Calculating quality-adjusted prices is therefore difficult.
observe point-in-time prices as well as discuss discounts,
special offers, and volume-selling items with the respon- Using Scanner Data Sets to Update Pricing
dent. The field officers record this information during the Samples
visit, often in handheld electronic devices. The regular visits 10.39 The collection of point-in-time prices by NSO
to outlets enable the NSO field officers to actively monitor field price collectors visiting retail outlets is resource-
market developments and observe quality change. intensive. A census of items cannot practically be priced
10.34 Replacing field-collected prices by prices (unit each period resulting in the need for some sort of sampling
values) from scanner data generally results in NSO resource approach. For example, sampled products may be selected
savings, because NSO field officers are no longer required to for inclusion in the CPI basket by NSO field price collectors
visit businesses where prices were collected. The potential who discuss with the respondent which items are volume
for NSO resource savings is influenced by the size of the sellers or examine the shelf space of the products and make
field officer reductions and the increase in resources required judgments about their relative importance. NSO field staff
by the NSO to manage and process scanner data sets. then aim to select a representative basket of items for pric-
10.35 Unit values should relate to a single homoge- ing. This is a purposive sampling approach.
neous variety whose specification remains constant over 10.40 Nonprobability or purposive sampling has tradi-
time because changes in the composition of varieties sold tionally been used because sampling frames for items pur-
and the quality of these varieties should not be reflected as chased were not available and detailed quantity or revenue
price changes. These requirements present some challenges data to measure the economic importance of the items was
when replacing field-collected prices with information from lacking (see Chapter 4 for further information on sampling).
scanner data sets. Negotiation between the NSO and data Nonprobability or purposive sampling can lead to biases
provider is needed to ensure access to data at the appropri- when the selected items are not representative of the prod-
ate level of item aggregation (or disaggregation) required uct population.
to support the production of unit values for use in compil- 10.41 This traditional approach to sampling can be
ing the CPI. The direct supply of product characteristics replaced by more scientific sampling methods due to the
can facilitate the classification of items. Such information, availability of scanner data. Since scanner data typically
if available, could then be used to perform explicit quality is a census of products, scanner data sets can be used as
adjustment. a sampling frame for updating pricing samples. A pricing
10.36 Several NSOs have experience in producing sample is usually two-dimensional; it is a combination of a
unit value data from scanner data sets. At the most detailed sample of outlets and a sample of items/product varieties.
level, items in scanner data sets are typically identified by If all the stores from a retail chain are covered, the scan-
barcode or the corresponding Global Trade Item Number ner data set can be used as a sampling frame for both the
(GTIN) or its subvariants, the Universal Product Code, and outlet and item dimension (see also Chapters 4 and 5). For
the European Article Number. While standardized identi- example, a two-stage sampling approach could be adopted
fiers such as GTIN allow for the tracking of items across by first selecting outlets and then selecting items within the
different retailers, they may be too detailed, differentiat- sampled outlets.
ing varieties by characteristics, such as packaging, which 10.42 Revenue shares for each product (or product/out-
are considered irrelevant to consumers (Dalen 2017). Item let combination) can be used to determine the significance
churn will then be overestimated and there is a potential of each product within a product group. Products are then
problem of relaunches which may impede the calculation of selected for inclusion in the CPI “basket” based on revenue
the CPI. For example, when using GTIN as the item iden- share either through sampling proportional to revenue or
tifier, the price change of a homogeneous variety whose cutoff sampling (de Haan and others 1999).
GTIN changes at the same time will not be measured. In 10.43 Over time, however, products in the sample can
some countries, the use of stock keeping unit (SKU) rather lose relevance or even cease to exist. In these situations, a
than GTIN has proven to be successful (Howard and others replacement product is needed to maintain the relevance of
2015). the sample. Relevance tests can be used to highlight those
224
Scanner Data
items in the samples that have become unsuitable, and also the index structure below the elementary aggregate level
highlight and rank suitable items as replacements. with separate elementary price indices for each retail chain.
10.44 The main principle behind these relevance tests This raises several issues. The first issue is whether the
is that the products should have a stable revenue share (that stores belonging to the chain in question should be viewed
is, consistent revenue share compared to other products), as separate outlets. If this is the case, unit values for the
within the CPI product group. These product groups are sampled items should be calculated at the store level. On
referred to as the elementary aggregate. The stable revenue the other hand, when the service levels are similar across
share is important as items can have large sales when intro- stores belonging to a retail chain, it may be useful to calcu-
duced into the market due to novelty or introductory sales late unit values at the chain level (Ivancic and Fox 2013). In
prices, have insignificant revenue thereafter, and hence not this case, the retail chain scanner data directly represent all
be representative of the broader market. the stores of that chain. When using chain-level data, one
10.45 To mitigate these problems, possible replacement must ensure that each retail chain is weighted in the final
products’ revenue must have been stable and significant for index. Some retail chains operate different types of stores
a specified period (for example, three to six months) before with different assortments of products and price levels. In
they can be considered for inclusion in the price samples. this case, a stratification can be introduced that distinguishes
CPI analysts can then manually review all items which are between the types of stores belonging to a retail chain. Some
flagged for replacement and select items from a list ranked NSOs do not have a choice if they receive scanner data at
according to average monthly revenue share over the previ- the chain level.
ous six months. 10.51 A regional disaggregation of the scanner data may
10.46 Many food and household items will have variet- be needed if regional CPIs are compiled. The chains, stores,
ies of the same base item which have similar if not identical products, and prices included in the resulting indices should
price evolution. A specific brand of canned tuna, for exam- be representative for the respective region(s). Alternatively,
ple, is available in many flavors and CPI compilers will be scanner data can be directly exploited at the national level.
aware that prices for the different flavors from the same It has then to be determined whether the resulting scanner
brand will behave similarly, going on sale at the same time data price indices, which are representative for the country
and changing price at the same time. Having a single flavor as a whole, can also be used to compile the regional CPIs.
in the sample will hence represent the price movement for a 10.52 The next issue to be considered by the NSO
much more significant portion of the market than that single is to what extent existing sampling procedures should be
flavor’s revenues would suggest. improved. Suppose the NSO formerly used sampling of
10.47 The sampling process used to ensure product items proportional to revenue from the scanner data. This
samples are representative is usually manually driven, procedure can also be used to sample items from chain-
requiring the CPI analysts to select a replacement from this specific elementary price indices, where items are either
ranked list of potential products that pass certain eligibility defined (and unit values calculated) at the outlet level or the
criteria. This scanner data sampling approach requires addi- chain level. If the NSO wants to significantly increase the
tional CPI analyst resources which, ideally, are offset by the sample sizes to make use of a substantial part of the price
reduced field collection resources. information contained in the scanner data set, sampling
procedures need to be reconsidered.
Using Scanner Data to Update Index Structures 10.53 Another issue is how to integrate the chain-
and to Apply Weights specific elementary price indices from scanner data with
10.48 Variety samples have traditionally been small. price information from other sources. Because these ele-
When the additional CPI analyst resources are indeed off- mentary price indices are different from the elementary price
set by the reduced field collection resources, the NSO could indices in the traditional index structure, the scanner data
decide to expand the variety samples without changing the price indices have to be aggregated up to a level—perhaps
price index formula at the elementary aggregate level or the the lowest level of product aggregation the NSO publishes
sampling procedure. price indices—where they can be combined with price indi-
10.49 It may be worthwhile, however, to reconsider the ces from other sources. If a retailer-specific classification is
index structure and the sampling procedure when the NSO used, it must be as detailed as the lowest level of product
obtains scanner data directly from retail chains. Tradition- aggregation that the NSO publishes. If not, the data need to
ally, an elementary aggregate index is compiled from prices be reclassified accordingly. In other words, two aggregation
collected at outlets that belong to different retail chains (or steps are required: aggregation of the chain-specific elemen-
independent stores). When the NSO wants to use much tary price indices up to some higher-level product category,
more price information from a retail chain than before, it and aggregation of the resulting scanner data indices with
seems preferable to treat elementary aggregate1-chain com- price indices at that level pertaining to other retail chains
binations as separate strata in the index compilation process. and independent stores.
10.50 If the NSO chooses to use the classification sys- 10.54 The revenue data provides the opportunity for
tem provided by the retailer, it will be necessary to expand NSOs to weight price indices more frequently using more
timely data. This can be achieved in various ways, depend-
1
ing on the availability to the NSO of scanner data for mul-
The elementary aggregate refers to the lowest level for which expenditure tiple chains. It is suggested that the weights to combine the
weights are available. Because scanner data contains detailed and timely
quantity information, the elementary price indices are stratum indices that
price indices from scanner data be updated annually, using
are aggregated to obtain the higher-level indices in the CPI classification product revenue data from the previous 12 months. Combin-
structure. ing the scanner data indices with the price indices compiled
225
from other sources requires expenditure data for the latter a census of varieties from scanner data sets is prohibitively
indices, which may be difficult to come by or estimate. expensive, however, and cannot be undertaken to meet the
10.55 With the household budget survey, detailed data CPI production timeframes. Automating CPI compilation
by item (or item/outlet combination) are not available. The processes is required.
majority of NSOs therefore apply unweighted price index 10.61 Also, when using a census of varieties, not a
methods at the lowest levels of the CPI: the prices or price sample, a weighted index number formula is preferred.
changes of the items sampled from an elementary aggregate Again, variety turnover poses a significant problem. To
are combined without explicitly weighting the items accord- maximize the number of matches in the data, chaining at
ing to their economic importance. In most cases, the Jevons high frequency will be needed. This can lead to a significant
index formula is used by NSOs. drift in the index. Multilateral price index number methods,
10.56 Scanner data sets contain revenue data at the most which are drift-free by construction, are currently consid-
detailed variety level. These data can be used to sample ered a suitable method to handle a census of items and vari-
varieties proportional to their revenue, as mentioned previ- eties from scanner data; however, work continues on how
ously, but this raises a few issues. The inclusion probabili- best to apply multilateral methods to compile elementary
ties serve as implicit weights. That is, the elementary price aggregates using scanner data. The scanner data offer many
index will be an implicitly weighted index, and the inclusion opportunities for new research and developments.
probabilities should correspond with the target/population
index aimed at (Balk 2005). Moreover, the revenue distribu- Multilateral Price Index Methods
tion within an item category as observed in scanner data is
often highly skewed. Consequently, sampling proportional Introduction
to revenue is likely to select some high-revenue varieties 10.62 Scanner data can be implemented in the CPI
with a probability of one, and the high-revenue items in this using traditional sample-based methods. Prices formerly
“self-selecting” subsample should be explicitly weighted— observed by price collectors visiting outlets can be replaced
without explicit or implicit2 weighting of these items the by unit values from scanner data without changing the sam-
(unweighted) sample-based Jevons index cannot be an unbi- pling design or the price index number formula used. If the
ased estimator of a weighted geometric target index. NSO decides to use all the available data rather than tak-
10.57 It is preferable to reflect the items’ economic ing samples, the preferred approach, multilateral price index
importance explicitly via a weighted index number formula number methods are most suitable. Multilateral methods
rather than implicitly via the inclusion probabilities in an were originally developed to compare price levels across
unweighted index. countries, but they can be adapted to price comparisons over
time. These methods are particularly useful for scanner data,
Using Scanner Data Sets to Implement New CPI where item turnover is often high and promotional sales
Compilation Methods occur frequently.
10.58 The approaches outlined previously enable the 10.63 The most important multilateral price index
NSO to continue using sample-based methods to compile number methods are described in the following text; for
their CPI. Improvements to the accuracy of the CPI will a comprehensive discussion, see Chapter 7 of Consumer
be achieved because the prices (unit values) are more rep- Price Index Theory. After defining the variety, a short
resentative of those actually paid by consumers. Also, the overview of traditional bilateral price indices and chaining
varieties sampled reflect volume sellers, and the weights is provided.
used to produce aggregate measures of price change are
based on more timely information and can be updated more
frequently. Defining the Variety
10.59 The major challenge faced by NSOs implement- 10.64 Before applying any index calculation method,
ing these approaches relates to the increase in resources the individual variety to be priced must be defined. The
needed. Maintaining a sample-based approach, especially basic principle is to compare like with like and to track the
when the variety samples are extended, requires significant price of the same variety over time. The article code level
manual intervention, primarily because variety turnover usually represents the most detailed level of homogeneity
can be large. When one European country first introduced in the data. In addition to this product dimension, one must
scanner data for supermarkets into the CPI, a Lowe index also consider the outlet and the time dimensions. Often, the
was used (Schut and others 2002). The idea was to mimic same article sold at different moments of time, in the same
traditional methods and processes on a sample of about or similar outlets can be considered to define a variety which
10,000 items (barcodes) from each supermarket chain. This is sufficiently homogeneous so that an average transaction
approach was very demanding with regard to the manual price (unit value) can be calculated for the variety.
selection of items to replace disappearing items and with 10.65 In some cases, the article codes such as GTINs are
regard to quality adjustments when deemed necessary. stable and long-lived. Some countries have access to retail-
10.60 Ideally, acknowledging practical constraints, an ers’ product codes, for instance, an internal SKU code that
NSO would use all the available information in scanner is already more aggregated than the GTIN code. However,
data sets rather than taking samples. Manually processing in other cases, these article codes level may be too detailed
for price index calculation. In some product categories, such
as clothing and footwear, the article codes frequently appear
2
Implicit weights result when samples are selected using probabilities pro- and disappear making it difficult to match them across time
portional to size sampling methods. and therefore price changes are not adequately measured.
226
Scanner Data
The different strategies to cope with the article code changes 10.70 In the situation with no expenditure information,
are described in the paragraphs that follow.3 Chapter 8 of this Manual recommends the use of the Jevons
10.66 One approach is to group together different indi- price index, the unweighted geometric mean of price relatives:
vidual articles with similar characteristics. NSOs can choose
to define the variety more broadly or more narrowly. It is 1 1
important to create groupings so that consumers are more I J0:t

 p t n
= ∏ i0  =
∏ (p )
i∈S
t n
i
(10.1)
or less indifferent between the different individual articles  p  1
within these groupings. Calculating unit values at this level

i∈S i
∏ (p )
i∈S
0 n
i
not only makes it possible to capture substitution effects

between comparable articles but also facilitates the inclu- 10.71 The NSO traditionally draws a sample of items
sion of new articles entering the market. NSOs face trade- from the entire universe S to reduce CPI production cost.
offs in balancing these groupings. If a grouping is too broad, Without access to scanner data, S is unknown and a detailed
this can lead to unit value bias (and high volatility) as the sampling frame is lacking. Most CPI samples have therefore
individual articles are not strictly comparable. On the other been drawn purposively, with the risk of introducing bias in
hand, defining the grouping too narrowly may lead to a lack the index.
of matching between outgoing and new or returning arti- 10.72 Since scanner data contains expenditure infor-
cles. Decisions made at this stage can significantly impact mation for a census of items, the construction of superla-
the price indices that are eventually obtained. This could be tive price indices is possible on the entire set S. The focus
especially relevant for technology products, especially mod- here is on the Törnqvist rather than the Fisher index or other
els with high turnover (see Chapter 6 for more details). If superlative index formulas. While the Fisher and Törnqvist
feasible, the sensitivity of the definitions of the groupings on produce very similar results, the Törnqvist index allows for
the results should be tested. simpler expressions. The Törnqvist price index is given by
10.67 The practical construction of these groupings
si0 + sit
can be challenging. NSOs require information about article  pt  2
characteristics, including brand and size, as well as internal IT0:t = ∏  i0  (10.2)
classification codes used by retailers. Some retailers may i∈S
 p 
i
only provide characteristics in a specific text string while
others may have several different variables that describe the where s i = p i q i / ∑ i∈S p i q i and s i = p i q i / ∑ i∈S p i q i denote
0 0 0 0 0 t t t t t
characteristics of the different articles. Characteristics gath- the expenditure shares in periods 0 and t.
ered in one single variable (text string) require some form 10.73 In a dynamic universe, there are new and dis-
of text mining to make it useful for classification. Not all appearing items so that not all items can be matched over
characteristics are equally important and affect the price time. The sets of items in periods t (t = 0,..., T ) are denoted
in the same degree. The clustering of individual articles by S t with size N t . To maximize the number of matches
should be defined by the most important price-determining in the data, chaining matched-model superlative price indi-
characteristics. ces seems useful, for example, chaining period-on-period
10.68 Another approach would be to impute the prices Törnqvist price indices:
for the new and disappearing articles in the periods when
they are not available. Prices could, for instance, be imputed sit− 1 t
, M + si , M
with the help of a hedonic function. Instead of using the char-  pt  2
acteristics of the articles to form the groupings, this informa-

ITt−1:t = ∏  t −i 1  (10.3)
 p 
t−1,t 
tion is now used to estimate missing prices. However, such i∈SM i
a strategy is only appropriate for index number formulas,

such as the Törnqvist or the Gini, Eltetö, Köves, and Szulc where SiM t−1
and SiM
t
are the expenditure
t−1,t
shares in the two
(GEKS)–Törnqvist, which are responsive to imputations for periods with respect to the set SiM = S t −1 ∩ S t of matched
missing prices. items that are available in both period t − 1 and period t.
10.74 However, empirical work showed that high-
Bilateral Price Indices and Chaining frequency chaining of weighted price indices, including
superlative price indices, can lead to strong chain drift. In
10.69 Suppose first that the set of varieties sold is fixed case of promotional sales with reduced prices, the quantities
over time (that is, that the price statistician is dealing with a purchased often increase substantially. But when the prices
static universe). This fixed set of items is denoted by S and return to their original level, the quantities purchased of
its size by N. The sample period consists of (T + 1) time storable goods may not return to their “normal” level. This
periods t = 0,..., T . The prices (unit values) of item i ∈ S type of asymmetric behavior can cause chain drift in super-
(i = 1,..., N ) in periods 0 and t (t = 1,..., T ) are denoted by lative price indices, which is typically downward. Ivancic
p i and p i ; q i and q i are the corresponding quantities sold.
0 t 0 t
(2007), using market research scanner data on goods sold
The aim is to construct price indices that compare period 0, in supermarkets, found a downward drift in chained Fisher
the starting period of the time series, with each period t. price indices (see also Ivancic and others 2009, 2011). Drift
in chained matched-model superlative price indices has
been documented for durable goods as well. Here, the drift
is likely due to seasonal fluctuations in prices and quantities.
3
A similar problem can also be encountered with web-scraped data (see De Haan and Krsinich (2014), using scanner data, found
Chapter 5, Annex 5.6). a downward drift in chained Törnqvist price indices for
227
Table 10.1 A Numerical Example of Chain Drift

Period P1 P2 Q1 Q2 Share 1 Share 2 Average Share Average Share Period-to-Period Chained
(percent) (percent) in t and t − 1 for in t and t − 1 for Törnqvist Törnqvist
Item 1 (percent) Item 2 (percent)
1 3 4 12 10 47.4 52.6 100.00

2 3 2 12 30 37.5 62.5 42.4 57.6 67.10 67.10
3 1 4 40 5 66.7 33.3 52.1 47.9 78.66 52.78
4 3 4 5 10 27.3 72.7 47.0 53.0 167.53 88.42
5 3 4 12 10 47.4 52.6 37.3 62.7 100.00 88.42
6 3 2 12 30 37.5 62.5 42.4 57.6 67.10 59.33
7 1 4 40 5 66.7 33.3 52.1 47.9 78.66 46.67
8 3 4 5 10 27.3 72.7 47.0 53.0 167.53 78.18
9 3 4 12 10 47.4 52.6 37.3 62.7 100.00 78.18
consumer electronics goods sold. Silver and Heravi (2005) items’ average expenditure shares in adjacent months. This
presented evidence of downward bias in chained Fisher approach, sometimes referred to as the “dynamic approach”
indices using scanner data on televisions. (see Eurostat 2017), has been implemented by several Euro-
10.75 Table 10.1 shows a numerical example of down- pean countries (for example, see van der Grient and de
ward drift in the chained Törnqvist price index. There are Haan 2010, 2011). This method reduces the risk of chain
two items and nine periods distinguished. The “regular” drift because weights are used implicitly for the sampling
prices of items 1 and 2 are 3.00 and 4.00, respectively, but of items, but not explicitly in the index calculations. This
the price of item 1 is temporarily reduced in periods 3 and method has the advantage that it relies on the usual bilateral
7, while the price of item 2 is reduced in periods 2 and 6. index methods while at the same time making the best use
Notice that in the last period (period 9) the prices and quan- of the information contained in the scanner data sets. The
tities are exactly the same as those in the first period. Never- method is therefore easy to explain to users. Yet, this is not
theless, the period-on-period chained Törnqvist price index an optimal situation. A more advanced solution would be to
ends up at 78.18, thus measuring a price decline of almost explicitly weight the items and construct weighted multilat-
22 percent. For the multilateral indices, the index equals one eral price indices.
(as will the direct Törnqvist). This downward drift for the
period-on-period chained Törnqvist stems from the fact that
because the quantities do not immediately return to their Multilateral Methods
“normal” level after the discount, the price change from the 10.78 Multilateral methods produce transitive price
normal price to the reduced price has a bigger weight than indices. For price comparisons across time, this means the
the following price change from the reduced price back to indices are independent of the choice of base period, can be
the normal price. written in chained form, and are therefore free from chain
10.76 One way to avoid chain drift due to promotional drift. Multilateral methods have in common that price indices
sales for storable goods would be not to weight the items are constructed simultaneously for the entire sample period.
and to construct a time series by chaining period-on-period 10.79 Two types of multilateral methods can be defined.
matched-model Jevons price indices: The first type starts from matched-model price comparisons
1
between any pair of time periods across the entire sample
 p t  N Mt−1,t period and then “transitivizes” this set of bilateral price indi-
I t −1:t
= ∏  t −i 1  (10.4) ces. The best-known method is GEKS (Eltetö and Köves
 p 
t−1,t 
J
i∈SM i 1964; Gini 1931; Szulc 1964). An alternate method is based
t −1, t on spanning trees (Hill 1999a, 1999b), where a spanning
where N M is the number of matched items between tree is a supplier of paths between countries. For a certain
periods t − 1 and t. This is not to say that the use of the spanning tree, the bilateral indices are chain linked to con-
chained matched-model Jevons index is without problems. struct price comparisons between any pair of countries or,
For instance, clearance sales can put downward pressure on adapted to our context, time periods. It is not clear, however,
the index. In order to mitigate this problem, a dump filter what the theoretical and practical advantages are over the
can be implemented. The filter removes an item if both the easier-to-construct GEKS indices. The second type of multi-
price and the quantities sold of an item fall sharply. A down- lateral method attains transitivity in another way, which will
ward drift may also arise for fashion goods, such as clothing, be explained in the following text, and includes the Geary–
that exit the sample at low clearance prices and never return. Khamis method (Geary 1958; Khamis 1972) and the Coun-
Clothing requires special treatment as noted in Chapter 11. try Product Dummy method (Summers 1973).
10.77 The lack of weighting is problematic as well.
Item expenditures are usually highly skewed, and so the GEKS Method
many low-expenditure items will be given the same weight 10.80 The GEKS index between period 0 and period t
as the few high-expenditure items. A crude form of implicit is calculated as the geometric average of the ratios of the
weighting can be attained by simply excluding low-expen- matched-model bilateral price indices I 0, j and I 0, k , con-
diture items (that is, by giving them an inclusion probabil- structed with the same index number formula, where each
ity of zero), for example, using a threshold based on the period l is taken as the base. Provided that the bilateral
228
Scanner Data
indices satisfy the time reversal test, the GEKS index can T N −1
be written as (de Haan and Van der Grient 2011; Ivancic ln p ti = α + ∑ δ t D ti + ∑ γ i D i + ε ti , (10.8)
and others 2011): t =1 i =1
1 where D i is a dummy variable that has the value of one if

T  I 0,l T +1 T 1
the observation relates to item i and zero otherwise, and D i
t
I 0:t
= ∏ t ,l  = ∏( I 0,l × I l ,t )T +1 (10.5)
GEKS
 
l =0  I l =0
is a dummy variable with the value one if the observation
relates to period t and zero otherwise; dummies for item n
10.81 The time reversal test requires that when the base and period 0 are excluded to identify the model.
period and the comparison period are reversed, the result 10.86 Diewert (2005) proposed to estimate model
should be equal to the reciprocal of the original index. In (10.8) by Weighted Least Squares regression with the items’
its standard form, the GEKS method uses bilateral Fisher expenditure shares in each period serving as weights. Expo-
indices, which satisfy the test, but other choices are possible, nentiating the estimated time dummy parameter δˆt yields
including bilateral Törnqvist indices. GEKS–Törnqvist indi- the TPD index between periods 0 and t; ITPD0, t
= exp (δˆt ). The
ces are also referred to as Caves, Christensen, and Diewert weighted TPD method can be written as a system of equa-
indices. tions that is similar to the geometric GK-style system defined
by (10.6) and (10.7), and thus, the TPD can be solved itera-
Geary–Khamis Method tively as well as via direct regression methods (Rao 2005):
10.82 The Geary–Khamis (GK) method, when applied
to comparisons over time, gives rise to the following price  pt 
sit
index: ∏i∈S t  exp (i γˆ )

−1
0:t
I TPD =
i
(10.9)
   
−1
 p 0  i
s0
∑ i∈S t pit qit ∑ t sit  pit   ∏i∈S 0  exp (i γˆ )
t  i∈S  p   i
 i  
0:t
I GK =
∑ i∈S t p i qi
=

  (10.6) siτ
∑ i∈S 0 pi0 qi0   
−1  −1
 pτ ∑ τ∈Si siτ
0  0 
 ∑ 0 si0  pi  
exp (γî ) = ∏  0,it  (10.10)
∑ i∈S 0 i i  i∈S  pi  
p q  I
τ ∈ Si TPD 

   
 
10.87 Equation 10.10 shows that the exponentiated
10.83 The numerator of equation 10.6 is a price index item fixed effect estimates γˆ i , or reference prices, are equal

(using period t quantities) with “reference prices” p i to the expenditure-share weighted geometric averages of the
that are fixed across the sample period. The index should be deflated prices with the TPD index serving as deflator. Both
equal to one in the starting period 0, so it will be necessary GK and TPD explicitly arrive at reference prices. In the case
to normalize the index by dividing by its period 0 value, of GK, this means that the index is consistent with national
which is the denominator of equation 10.6. The reference accounts, as it is additively decomposable. TPD, being a
prices are given by geometric index, is not.
10.88 Notice that the TPD index (10.9) can be viewed as
τ
 piτ  a normalized geometric Paasche index with imputed period

∑ τ ∈Si i  I τ 0,tau 
q
(10.7) 0 prices based on the reference prices (10.10). Similarly, the
GK
pi = GK index (10.6) can be viewed as a normalized (ordinary)
∑ τ ∈S qiτ i
Paasche index with imputed period 0 prices based on the
reference prices (10.7).
where S i is the set of time periods in which item i is actually
sold and for which prices are available. Equation 10.7 shows

that p i equals a weighted arithmetic average of the deflated Lack of Matching and Quality Adjustment
observed prices, with each period’s share in the total number
of sales of the item across the entire sample period serving Implicit Quality Adjustment
as weights. 10.89 Like GEKS, GK and TPD are matched-model
10.84 Since the GK index acts as the deflator in equa- methods in the sense that items with a single observation
tion 10.7, equations 10.6 and 10.7 define a system of equa- in the entire sample period do not affect the index. Items
tions which must be solved simultaneously. This can be contribute to aggregate price change only when price rela-
done iteratively (which may be simpler to implement), but tives can be calculated from the prices observed in both
there are other ways to solve the system (Balk 2008). periods compared, unless information on characteristics
would be available to perform explicit quality adjust-
Time Product Dummy Method ments. One implication of the matched-model method
10.85 This is a regression-based approach. Assuming is that items introduced in the most recent period T are
n different items are observed in the entire sample period ignored.
0, …, T (most of which will typically not be sold in every 10.90 Implicit quality adjustment can also be illustrated
time period), the time product dummy (TPD) regression by using an alternate interpretation of the GK index. Divid-
model for the pooled data is ing the value index of the product category by the ratio of
229
“quality-adjusted quantities” defines a quality-adjusted unit the estimated item fixed effects exp(γ̆i ) now acting as refer-
value index (de Haan 2004, 2015): K
ence prices. The formula exp(∑β˘k ( zik − zbk )) can also be
∑ i∈S t
pit qit ∑ i∈S t
pit qit k =1
used to estimate the quality-adjustment factors λi / b in equa-

0:t ∑ i∈S 0
pi0 qi0 ∑ i∈S t
λi / b qit tion 10.11. The resulting explicitly quality-adjusted unit
I QAUV
=
∑ λi / b qit
=
∑ pi0 qi0 value index—or “hedonic variant” of the GK index—is
i ∈S t i∈S 0
expected to be very close to the TDH index (de Haan and
∑ i ∈S 0
λi / b qi0 ∑ i∈S 0
λi / b qi0 Krsinich 2018).
−1 10.95 De Haan and others (2016) compared the TPD
  t  
−1
 t  pi   and TDH methods. They argued that the TPD model suf-
 ∑ i∈S t si   
fers from overfitting because it has too many parameters and
  λi / b  
=  “distorts the regression residuals towards zero.” Under cer-
−1 (10.11)
  0 
− 1 tain pricing strategies of retailers, such as price skimming
 0  pi  
 ∑ i∈S 0 si    (new items) and dumping (old items), the TPD index can
  λi / b   be quite different from the TDH index. Similarly, the GK

index can be quite different from its hedonic counterpart, the
10.91 If λi / b = 1 for all i, equation 10.11 simplifies to quality-adjusted unit value index.
the ordinary unit value index. The quality-adjustment fac- 10.96 If relaunches of homogeneous items with
tors λi / b aim to express the quantities purchased of each different barcodes or SKUs are a major cause of a low
item i with regard to quantities of an arbitrary item b. The matching rate, then defining items by their characteristics
ratios pit / λi / b and pi0 / λi / b in the second expression of rather than barcode or SKU could be an option (Chessa
(10.11) are quality-adjusted prices. In the static universe 2016). However, scanner data sets provided by retailers
case (with no new and disappearing items), if the quality- typically include rather broad item descriptions, from
adjustment factor of an item corresponds to its base price, which it may only be possible to extract a few charac-
the quality-adjusted unit value index turns into the Paasche teristics, such as package size and brand name. In that
price index. Von Auer (2014) argued that many conventional case, the prices, calculated as unit values across all the
price indices can be viewed as, what he called, a generalized barcodes of SKUs that belong to the various “groups,”
unit value index. can suffer from unit value bias.
10.92 A comparison of equations 10.6 and 10.11 shows 10.97 Unlike TPD and GK, GEKS does not aim at
that the GK index can be viewed as a quality-adjusted unit implicitly adjusting for quality change. A potential advan-
value index where the quality-adjustment factors are mea- tage of GEKS over TPD and GK, however, is that the “miss-
sured by the reference prices in equation 10.7. Similarly, the ing prices” of the unmatched new and disappearing items
TPD index in 10.9 can be viewed as its geometric counter- can be imputed. It is therefore possible to estimate explicitly
part where the quality-adjustment factors are measured by quality-adjusted GEKS indices by replacing the bilateral
the reference prices (10.10). Whether the reference prices in matched-model Törnqvist price indices by bilateral hedonic
the GK and TPD indices properly reflect quality differences imputation Törnqvist indices, for example, as proposed by
is likely to depend on the market circumstances. de Haan and Krsinich (2014) or de Haan (2019). This means
there is no need to define the items by their characteristics;
Explicit Quality Adjustment barcode or SKU will suffice to distinguish between items.
10.93 Data on item characteristics permitting, explicit The hedonic imputations for the unmatched items adjust for
quality adjustment is preferred, in particular through quality changes and also deal with the relaunch problem.
hedonic regression methods. A useful starting point is the De Haan (2019) suggests using the same characteristics
multilateral time dummy hedonic (TDH) model: information in the hedonic regressions that would be used to
define the “groups” for dealing with relaunches in the TPD
T K
and GK methods.
ln pit = α + ∑ δ t Dit + ∑ βk zik + εit (10.12) 10.98 Missing information on important characteristics
t =1 k =1
makes the use of hedonic quality adjustment, or explicit
where zik is the quantity of characteristic k (k = 1,..., K ) for quality adjustment in general, problematic as this can lead
item i. Notice that, as pointed out by Aizcorbe and others to an omitted variables bias. Also, as mentioned previously,
(2003) and Krsinich (2016), the TPD model (10.8) arises it may give rise to unit value bias in the “group approach.”
from the TDH model (10.12) by replacing the hedonic A few NSOs have been exploring the use of web scraping
K to observe quality characteristics from retailers’ or manu-
effects exp(∑βk zik ) by item-specific fixed effects exp(γi ). facturers’ websites to enrich scanner data sets. Scanner data
k =1 obtained from a market research company may already con-
Similar to the estimation of the TPD model, it is assumed tain detailed information on item characteristics. One NSO,
that equation 10.12 is estimated by expenditure-share for example, produces quality-adjusted GEKS price indices
weighted regression on the pooled data of all time periods for consumer electronics products based on scanner data
t = 0,..., T .
sets from one market research company that include many
10.94 The resulting weighted TDH index, I TDH
0, t
= exp(δˆt ), item characteristics (Krsinich 2015). All these methods are
can be expressed in a similar way as the TPD index (10.9), rather data-intensive as they require prices, turnover, and
K
with the estimated hedonic effects exp( β˘ z ) instead of ∑k =1
k ik
price-determining characteristics at a detailed level.
230
Scanner Data
Revisions in Multilateral Indices Both the recalculated and published indices are candidates
10.99 When new data become available, previously for the index on which a new index series can be linked. In
estimated multilateral indices change when new data are empirical research, and in applications by two NSOs, the first
processed. This is problematic as the CPI is not revisable. option has been applied. The second option (that is, linking
Also, as time passes, recent price movements will be increas- onto the published index numbers) has been recently sug-
ingly affected by prices and price changes in the distant past. gested by Chessa (2019). Linking onto the published index
In the context of the GEKS index, this is known as a loss of numbers has advantages. For example, year-on-year rates
characteristicity. Different methods have been proposed to (inflation figures) calculated from shifted windows will also
extend a multilateral time series without revising published be the published figures if the link period corresponds to the
index numbers (and mitigate the loss of characteristicity). period which is 12 months ago. This increases the transpar-
The methods can be characterized by a number of choices: ency of splicing methods. Moreover, each year-on-year rate
window adjustment (rolling window or extending window), is derived from a transitive index series and is in that sense
the link period, and the index in the link period (Chessa free from chain drift. This is not the case when splicing on
2019). the latest recalculated index which can therefore still lead to
10.100 Rolling window methods estimate multilateral some drift (see Chessa 2019).
indices on a rolling window with fixed length, say T + 1, 10.103 The choice of window length is a point of con-
which is shifted forward each period. Table 10.2 illustrates cern. Ivancic and others (2011) advocated a 13-month (or
a rolling time window of 13 periods. The multilateral index five-quarter) window as this is the shortest window that can
compiled in period 13 covers periods 1–13. The multilat- deal with strongly seasonal goods. However, recent research
eral index compiled in period 14 covers periods 2–14 and so suggests that indices for strongly seasonal items can be
on. The results of the latest window are then linked onto a significantly improved with a 25-month (or nine-quarter)
previously calculated index. For example, movement splice window (Chessa 2020). It is possible to construct weighted
links the most recent period-to-period index change onto GEKS indices which consider the reliability of the bilat-
the index of the previous period (that is the latest published eral price indices (Rao 2001b). Melser (2018) proposed a
index). In Table 10.2, the movement splice index of period weighted GEKS method where the weights depend on the
14 is obtained by linking onto period 13 the price change degree of matching of the items, for example, with regard
between periods 13 and 14 derived from the corresponding to expenditure shares. Here, the choice of window length is
multilateral index. less important since bilateral indices with a lower degree of
10.101 An alternative to movement splice is a window matching will be down weighted.
splice, which links the full period index change onto the lat- 10.104 The annually chained direct extension method
est calculated index of T periods ago. The movement splice (Chessa 2016) constructs multilateral index series of, say,
was proposed by Ivancic and others (2011) for the GEKS 13 months, starting in, for example, December and ending in
method, and the window splice was proposed by Krsinich December of the next year, and chain links them in Decem-
(2016) for the TPD method. However, each splicing method ber of each year to obtain a long-term time series. The
can be combined with any multilateral method. These splic- length of the estimation window for the short-term indices
ing methods link index changes onto a single link period. is extended each month—the index for January in the short-
Diewert and Fox (2017) proposed a mean splice by taking term series is estimated on two months of data (which is a
the geometric mean of the price indices obtained from using bilateral rather than multilateral comparison), and so forth,
every possible link period. This makes the result indepen- until in December 13 months of data is used.
dent of the choice of link period. 10.105 A potential weakness of the direct extension
10.102 Chessa (2019) pointed out that there are in fact method is that the price indices for the first couple of months
two main options for splicing methods (apart from move- of each year are based on sparse data and expected to be
ment splicing, which has the index published in the previous volatile. Also, December acts as the short-term index refer-
period as the only link option). Successive window shifts ence period and is given special importance. If, for some
generate a sequence of recalculated or “revised” indices reason, December is an unusual month, the results may be
alongside the initially published index in the same period. adversely affected. To mitigate these problems, Lamboray
Table 10.2 Movement Splice Linking with Rolling Window of 13 Months

Period 1 2 3 4 5 . . . 11 12 13 14 15
First 100 100.7 100.6 101.6 102.7 . . . 104.3 106 103.8
Compilation Round in Period 13
Second* 100 100.2 101.1 102.2 . . . 103.8 105.5 103.3 104.6
Third 100 101 102 . . . 103.5 105.3 103.2 104.4 104.1
Published Index (movement splice) 100 100.5 100.6 101.6 102.7 . . . 104.3 106 103.8 105.1 104.8
The splicing starts in period 14 (shown in bold). The published indices for periods 1–13 are obtained at the first compilation round. The published index in
period 14 is obtained by applying the change between period 13 and period 14 indices of the second compilation round to the published index of period 13
(103 8 × 104.6/103 3 = 105.1). The published index in period 15 is obtained by applying the change between period 14 and period 15 indices of the third
compilation round to the published index of period 14(105 1 × 104 1/104 4 = 104 8).
231
(2017) suggested combining the annually chained direct undertake similar assessments in their local context. A com-
extension method with a rolling window approach. prehensive discussion of the various multilateral methods
using the economic approach to index number theory can
be found in Chapter 7 of Consumer Price Index Theory;
Implementation of Multilateral Methods see also Diewert and Fox (2017). The most important result
is that GEKS deals appropriately with substitution effects
Assessing Multilateral Methods whereas GK and TPD are appropriate only under restric-
10.106 The implementation of new data sources and tive assumptions about consumer preferences. However, in
methods in any statistical series requires careful consider- practice, GK and TPD generate very similar results. Another
ation of the statistical impacts as well as the benefits and NSO developed a generic processing method based on the
costs. Only a handful of NSOs have implemented multilat- GK (see Chessa 2016). In a first step, the method was only
eral price indices in the CPI. implemented for mobile phones before being applied to
10.107 A suggested criteria to assess multilateral meth- more products and retailers in the following years.
ods should consider a broad concept of statistical quality. 10.112 In addition, expert peer review of the proposed
A possible framework could include seven dimensions of multilateral methods may be appropriate in circumstances
statistical quality: where CPI users may expect NSOs to demonstrate broader
• Institutional environment—pertains to the institutional endorsement of the proposed changes.
and organizational context in which a statistical producer
operates Calculating Indices
• Relevance—pertains to how well a statistic meets user Operational Choices
needs 10.113 The matched-model property of (nonhedonic)
• Timeliness—pertains to how quickly and frequently the multilateral indices implies that without any manual inter-
statistic is published vention, the results depend on the choice of item identifier.
• Accuracy—pertains to how well a statistic measures the For example, when using the barcode as item identifier, the
desired concept price change of a homogeneous item whose barcode changes
• Coherence—pertains to how consistent the statistic is at the same time—a “relaunch”—will not be measured. As
with sources of related information mentioned earlier, the use of SKU codes mitigates the prob-
lem since SKU generally consists of multiple barcodes for
• Interpretability—pertains to the information available to
similar items and is more stable than barcode. Nevertheless,
provide insight into the statistic
even SKU may be too detailed.
• Accessibility—pertains to ease of access to the statistic 10.114 If a relatively small number of observable attri-
10.108 Of note, multilateral methods are more compli- butes with discrete values suffice to define homogeneous
cated than standard bilateral indices and present communi- items, items could be defined by cross-classifying the sets
cation challenges for NSOs. High value should be placed on of categorical variables for each attribute and prices cal-
transparency to explain the statistics published, and describ- culated as unit values across all the barcodes/SKUs. Most
ing and justifying the methods used. This is of critical likely there will still be new and disappearing items (cells)
importance for the trust in the published CPI. Two aspects across the sample period. To maximize the degree of match-
of interpretability need consideration: first, to what extent ing without introducing chain drift, a multilateral method
the methods themselves are easy for index practitioners and could be applied (Chessa 2016). A potential issue is that the
users to understand; and second, whether it is easy to under- available characteristics information may be limited, espe-
stand the price movements each index produces, especially cially when the characteristics are extracted from variety
which products have the greatest influence on these move- descriptions in scanner data, which are often rather broad.
ments and why. In this case, unit value bias is likely to arise. Also, if the
10.109 This framework can be used by an NSO to characteristics information is deemed sufficient, it may be
determine the benefits and challenges of using multilateral better to construct hedonic indices.
index methods considering country-specific circumstances. 10.115 While taking a (cutoff) sample that ignores
10.110 Multilateral index methods to compile the items, however defined, with small expenditure shares
CPI can also be assessed from a theoretical perspective. would in many cases not significantly affect the results, it is
The assessment can use approaches previously applied to not necessary when using a weighted multilateral method.
bilateral and spatial indices. The bilateral price indices are Most of the issues discussed earlier in this chapter, such as
assessed both from axiomatic/test approaches (Chapter 3 of the choice of calculating unit values at the store or at the
Consumer Price Index Theory) and economic approaches chain level, and the need to have an index structure that
(Chapters 4 and 5 of Consumer Price Index Theory). Similar facilitates the use of scanner data, apply here as well.
approaches to assessing multilateral indices in a spatial con-
text have been developed and presented in several papers, Producing Empirical Results
especially Diewert (1999b) and Balk (2001). 10.116 The purpose of producing empirical results
10.111 A description of the theoretical assessments of multilateral methods is twofold: examining the perfor-
of multilateral price index methods in the present tempo- mance of various methods in local contexts as well as dem-
ral context using the axiomatic/test and economic theory onstrating to CPI users the likely impacts of moving from
approaches is available (ABS 2016a; Zhang and others current CPI data sources and methods to new approaches.
2019). This assessment can be used as a basis for NSOs to Ideally, these multilateral methods should be examined
232
Scanner Data
against each other, as well as in comparison to the official • Engaging with leading academics both to review the pro-
CPI. These comparisons should be undertaken at the low- posed changes and to encourage their support
est level of the published CPI and at various aggregation
10.119 Following this consultation, which could take a
levels, including the all-items CPI. For some empirical evi-
couple of years, the NSO should publish a position paper that
dence on multilateral price index methods, see Chapter 5
both responds to the topics raised as part of the consultation
of ABS (2016a), Chapter 3 of ABS (2017), and Chessa and
process and articulates how the NSO will proceed with the
others (2017).
use of scanner data to compile the CPI, including the rationale
10.117 Several insights can be obtained from produc-
and empirical results that support this approach. The position
ing empirical results. Often these insights further reinforce
paper should clearly state the data sources and methods to be
the theoretical arguments for utilizing multilateral index
employed and provide a timetable for the implementation of
methods to compile the CPI. This may include the impact of
changes.
using contemporaneous information for weighting purposes
that capture consumer behavior, including substitution, over
time. The empirical results should be communicated to CPI Publication and Dissemination
users and stakeholders. 10.120 Following the publication of the position
paper, it is suggested the NSO compile the CPI using both
Communicating with Users and Stakeholders current and new data sources and methods in parallel for a
10.118 The use of scanner data to compile the CPI period of approximately six months. This transition period
potentially represents quite a significant change to the data allows the NSO to refine processes and procedures to com-
sources and methods employed by NSOs. These changes pile the CPI using the new methods, as well as to compare
need to be carefully communicated to CPI users and stake- the empirical results of the two approaches. This transi-
holders. A suggested set of activities includes: tion period is often the first opportunity for the NSO to use
• Publishing information papers that outline the proposed the new data sources and methods in real time following
new methods and data sources the CPI processing and publication timetable. It is at the
NSO’s discretion whether the results of the parallel pro-
• Conducting face-to-face meetings with key stakeholders cessing are made public.
(for example, central banks, treasury, and finance minis- 10.121 The first period for which the new data sources
tries) and other interested parties, including members of and methods are used to compile the CPI should be announced
the public well in advance and should include detailed metadata for the
• Using media releases and briefing of economic journalists media and other key data users. This will ensure that the
to help inform the public of proposed changes methodological changes implemented to the CPI are well
• Encouraging stakeholders and the public to provide sub- understood. Further information relating to the CPI dissemi-
missions to the NSO for consideration nation can be found in Chapter 14 of this Manual.
233
SELECTED SPECIAL CASES 11
Introduction products generally do not require any direct intervention
by the index compiler. The seasonally fluctuating prices of
11.1 Certain products have proven to be challenging for weakly seasonal products will typically be captured in the
consumer price index (CPI) compilers with regard to devel- index although they are not without problems for CPI users.
oping weights and collecting prices. Chapter 11 focuses on For instance, when the weakly seasonal product is “out of
selected special cases and provides detailed advice for some season,” its price may be unusually high or low and the
of the more challenging products and issues facing compil- annual basket will reflect these unusual price fluctuations,
ers. These include the treatment of seasonal products, inter- leading to seasonal fluctuations in the overall index. This
net purchases, housing, second-hand goods, own-account volatility can cause “statistical noise” with the analysis and
production, tariffs, telecommunications, transport services, understanding of inflation. For some purposes, users want a
health, education, social protection, and financial services. CPI that measures the underlying price change and not these
11.2 Wherever possible, the chapter identifies the seasonal fluctuations. Measures of “core inflation” address
preferred approach for the treatment of each special case. these issues (see Chapters 12 and 14). It is the strongly sea-
However, at this time there is no preferred approach for sonal products that pose the bigger problem for the index
the treatment of owner-occupied housing, and the section compiler. If a price is available in only one of two periods
on this topic provides an overview of the different methods being compared, then it is not possible to calculate a price
for its treatment and describes the advantages and disadvan- relative for the product.
tages of each method. 11.6 There are two preferred approaches for the treat-
ment of strongly seasonal products. Ignoring the issue and
excluding all the challenging products from a CPI is not
Seasonal Products a solution in the context of an index whose purpose is to
reflect changes in all consumer prices. If these products have
Introduction some importance in the index basket then there is no justifi-
11.3 Seasonal products are those products that are either cation for ignoring them.
not available in the market during certain seasons or periods 11.7 This section describes the alternative ways of deal-
of the year or are available throughout the year but with reg- ing with strongly seasonal products. Essentially, there are
ular fluctuations in their quantities and prices that are linked two methods: a fixed-weight approach that uses the annual
to the season or time of the year. weight for the seasonal product in all months using an
11.4 Climate, traditions, and institutional arrangements imputed price in the out-of-season months, and a seasonal-
are the main causes of regular variations in the supply and weight approach where the weight is zero for out-of-season
demand for products. Fresh fruit and vegetables often have months and the annual weight is used for in-season months.
particularly marked seasonal purchasing and consumption The 2003 International Labor Organization Resolution
patterns, and certain fruit and vegetables may not be avail- states that the way these (strongly seasonal) products are
able at all at certain times of the year. Other products that can dealt with should be determined by the main purpose of the
display some seasonality include other fresh foods, clothing, indices, national circumstances, and the practicalities of
water, electricity, and fuels. The list of seasonal products is compilation. For instance, the relative importance given to
not uniform across countries. For example, oranges may be measuring month-on-month inflation and how the alterna-
available for purchase year-round in some countries, but tive methods of treatment perform in practice may inform a
only at certain times of the year or at a premium price in decision on which method to use.
other countries. Similarly, seasonality can vary between dif- 11.8 There is no completely satisfactory way of deal-
ferent regions within the same country. Certain religious and ing with strongly seasonal products. Index number theory
other festivals can also be associated with goods or services can provide reasonably effective solutions where the focus
whose consumption is limited wholly or partially to the fes- is on comparing prices in one month with prices in the same
tival period, such as Christmas trees, or products that are in month a year earlier. But the estimate of month-on-month
high demand or especially produced, such as certain gifts inflation can vary, depending on the approach used, mak-
given at the end of Ramadan. ing the analysis of short-term inflation trends difficult. For
11.5 In the compilation of a CPI, a useful distinction example, since under the fixed-weight approach the absence
can be made between “strongly” seasonal products, that of price quotes in particular months means that no month-
are available only part of the year when “in season,” and on-month price ratio can be compiled without imputing a
“weakly” seasonal ones, that are available throughout the price, the result will depend on the method of imputation.
year but their prices and availability for purchase fluctu- Alternatively, a zero weight can be attached to the miss-
ate significantly with the time of the year. Weakly seasonal ing product, as in the seasonal-weight approach, but the
235
resulting monthly variations in the composition of the CPI Fixed-Weight Approach

basket hinder month-on-month comparisons of inflation. In 11.13 The fixed-weight approach, where the weights
general, very large seasonal fluctuations in volumes com- remain constant over all months of the year and that imputes
bined with large systematic changes in prices can make a price when the product is out of season and not available for
month-on-month price index comparisons behave rather price collection, is theoretically consistent with the concept
poorly. of a fixed basket. However, it raises the issue of the choice
11.9 The difficulties for CPI compilation that are raised of imputation method for the unobservable out-of-season
by the existence of strongly seasonal items, and their sea- prices. The most commonly used approach is to impute a
sonal unavailability can be tackled by choosing one of two price using the last available observed price (that is reliable)
main approaches: and multiplying this last available price by the amount of
• The fixed-weight approach (also referred to as the strict price inflation for the most similar or “comparable” group
annual weights method). Allocating fixed annual weights, of items that has taken place since the time of this last avail-
on the assumption that seasonal products are to be treated able observed price. For instance, a similar item or group of
in the same way as all other products. Prices will need to varieties that is in season might be used. The missing prices
be imputed in the out-of-season period either from the last can be estimated using the monthly rate of change in the
observed price or from what is considered the “typical” or prices of the set of products within the same class of the
“normal” price. The main imputation methods are overall Classification of Individual Consumption According to Pur-
mean imputation or class mean imputation (see Chapter 6 pose (COICOP) or using an appropriate subset. Measured
for more details on imputation methods). inflation of higher-level aggregates is likely to be influenced
• The seasonal-weight approach.1 Allocating seasonal to a large extent by factors that are not so relevant for the
weights according to the consumption pattern found in the seasonal product. Using products from the same COICOP
weight reference period. Items are given a zero weight in the subclass, class, or group also enhances comparability with
out-of-season period, and the expenditure on other selected the seasonal-weight approach, thus increasing statistical
items is adjusted so that the basket weights sum to 100. integrity.
11.14 The basis for the imputation could be an elemen-
11.10 The methods are discussed in more detail in tary index that uses Dutot or Jevons averaging, depending
paragraphs 11.13–11.38. A common predicament is the on the formula used for the CPI at the elementary aggregate
determination of the in-season and out-of-season periods. level, or it could even be the all-items CPI (that is, at the
A conservative approach is recommended in determining aggregate level) if there is no similar group of items.
the in-season period to avoid abnormal prices from enter- 11.15 The use of the carryforward method (repeating
ing the index, for instance, high prices being charged for the last collected price) is discouraged. Using the carry-
the first supply appearing in the outlets of a fresh crop (for forward method may introduce bias into the index, as
example, summer fruit) where quantities may be low and month-on-month changes would be zero (as described in
demand and prices high. The chances of this happening are Chapter 6).
minimized in the fixed-weight approach by imputing from 11.16 The main drawback of the fixed-weight approach
what is considered a “typical” or “normal” price. There is is that the annual fixed weights will not be representative
no universally accepted standard definition of what consti- of the monthly consumption pattern. For example, oranges
tutes a “typical” or “normal” price. In the European Union might be available for sale only in some months, but the
(EU) context, in the Harmonised Index of Consumer Prices fixed-weight approach assumes that they are sold through-
(HICP), the average price in the previous season or the regu- out the year and introduces artificial prices that do not exist
lar price observed before the sales period is used. The aim is in the market but are based on the price movements of a
to prevent exceptional prices from having an impact on the comparable product, such as bananas, that in reality may
indices during the full out-of-season period. be of limited comparability. The limited comparability of
11.11 Even though existing index number theory cannot the comparable product can lead to an extreme fluctuation
deal satisfactorily with seasonal products in the context of in the oranges subindex when oranges return to the market
constructing month-on-month indices of consumer prices, it and are available again for direct pricing. In practice, the
can better deal with seasonal products if the focus is changed fixed-weight approach normally gives reasonably smooth
from month-on-month CPIs to CPIs that compare the prices 12-month rates of change. The imputation of prices by refer-
of one month with the prices of the same month in a previ- ence to the most similar group of products or items normally
ous year, that are individual annual inflation rates. reflects reasonably well the price inflation expectation when
11.12 The use of seasonal adjustment techniques to the product returns to the market and thus is often consid-
extract the seasonal components of a CPI for analysis are ered appropriate for the medium- and long-term measure-
not covered in this chapter and should be used only for ana- ment of inflation. However, the use of carryforward prices
lytical purposes (for more information, see Chapter 14). It can, of course, bias the month-on-month changes to zero.
should also be noted that, in general, the imputation of miss- 11.17 For clothing, in the first month, the seasonal item
ing prices does not eliminate seasonal fluctuations in prices. is missing; the price should be returned to the last “normal”
The CPI should reflect actual prices paid by consumers and price. The “normal” price refers to the last collected price
these fluctuations should be included. before the seasonal discount or sale period. In the second
month, the imputation should be made on this last “normal”
price. Imputing the sale or discounted price can introduce
1
Referred to in EU HICP as the class-confined seasonal weights method. a downward bias in the clothing index. For other seasonal
236
Selected Special Cases
items, such as fresh fruits and vegetables, it may also be year round to serve a specific purpose, so it is argued that
preferable to return the final in-season price to a “normal” they cannot be substituted by either a summer or winter
price before beginning the imputation. Alternatively, the dress. However, warm and lightweight clothes have the
imputation could be made on the last available price begin- same purpose of protecting the body and are substitutable
ning with the first month the seasonal item is missing. Impu- according to the temperature conditions.
tation methods are described in the section that follows.2
11.21 There are many examples of consumption prod-
11.18 The advantage of the fixed-weight approach
ucts that are available seasonally. The example in Tables 11.1
where missing prices are imputed is that it is easy to explain
and 11.2 illustrates the method of imputing prices for out-
and implement and keeps the annual fixed-basket methodol-
of-season products based on price movements of related
ogy more or less intact.
in-season products. In this example, there are two seasonal
11.19 An additional recommendation is that the method
products but only one is available during each collection
applied to deal with seasonal products should ideally be
period. In addition, there are occasions when neither is avail-
“self-correcting.” For example, if the item after the out-
able. The two products are children’s summer nightwear
of-season period reappears at the same price as it left the
and children’s winter nightwear. As both products serve the
market, then the method should ensure that this will take
same purpose, they are substitutes for each other within their
the index back to 100; similarly, if the price reappears 10
respective seasons. This allows the compiler to assume that
percent higher, it should take the index to 110.
the price movements of the available product would be suit-
able to impute price movements for the unavailable product.
Imputation of Prices under the Fixed- This assumption would also hold in situations where both
Weight Approach: Overall Mean or Class products are available in the same pricing period (that is, the
Mean Imputation seasons overlap for a short time). There would simply be no
11.20 The principle behind imputation is that it makes need to impute any prices during these periods.
use of the best available information to provide an unbiased 11.22 A well-constructed CPI structure would have
estimate of price and price movement. There are essentially products that could be considered substitutable or very
three choices: similar in nature as neighboring components. This allows
the index structure to determine the appropriate price move-
• Overall mean imputation. Impute the missing price by ments to be used in computing price movements for unavail-
reference to the average price change for the prices that able seasonal products.
are available in the elementary aggregate. This assumes 11.23 Table 11.1 shows the price relatives for a
that the price change of the missing variety, if it had monthly index for most of a year; in this case, the short-
been available in the outlet, would have been equal to term monthly price relatives noting the same can be applied
the average change in prices in the elementary aggregate. using long-term price relatives. Only the structure for one
This may be a reasonable assumption if the elementary small portion of the entire CPI structure is displayed. There
aggregate is relatively homogeneous. This method of are four elementary aggregates shown: Underwear, Socks,
imputation is equivalent to omitting the variety from the Summer nightwear, and Winter nightwear. The prices that
elementary index, no matter whether a Jevons, Carli, or have been collected and for which short-term price rela-
Dutot method of aggregation is used at the elementary tives can normally be calculated from actual nonimputed
aggregate level. This imputation method, as well as the prices are the figures not in italics. Underwear and Socks
next one, is discussed in Chapter 6. are available (and priced) all year round. Both seasonal
• Targeted mean imputation. Impute the missing price by elementary aggregates (Summer and Winter nightwear)
reference to the average price change for the price of a are grouped together under an aggregate component (sub-
“comparable” item from a similar outlet. This represents index) labeled Nightwear. No other elementary aggregates
a more precise match between the missing variety and the are included in this aggregate component. Consequently,
varieties or item supplying an imputed price. It is nor- the structure implies that each seasonal nightwear elemen-
mally preferable to impute using the average price change tary aggregate is the best index from which to impute price
in the elementary aggregate unless the imputations are movements for the other. Price relatives calculated from
unreliable because of small sample sizes. prices imputed in this situation are in italics. When there are
• Counter seasonal imputation. Impute by using only the no prices available for either of the two seasonal products,
prices of seasonal products in the COICOP subclass, other components must provide suitable price movements
class, or group. The reasoning underlying counterseasonal to be used in the imputation process. If the structure has
imputation is that both in-season and out-of-season prod- been organized to group comparable components as neigh-
ucts serve similar purposes, although in different condi- bors, then the best components for providing appropriate
tions. In contrast, nonseasonal products within the same price movements are likely to be similar to the component
COICOP subclass, class, or group may have different of the seasonal elementary aggregates. In this example, the
uses. For instance, formal ball gowns are available all components Underwear and Socks are judged to be similar
to Nightwear. Therefore, a weighted average of the price
2
movements of Underwear and Socks will be considered as
Paul A. Armknecht, and Fenella Maitland-Smith. 1999. “Price Imputation the best imputation of an appropriate price movement for
and Other Techniques for Dealing with Missing Observations, Seasonality
and Quality Change in Price Indices.” Working Paper No. 99/78, Interna- both Summer nightwear and Winter nightwear. Price rela-
tional Monetary Fund, Washington, DC. http://www.imf.org/external/pubs/ tives calculated from prices imputed in this situation have
ft/wp/1999/wp9978.pdf. been shaded in bold italics.
237
Table 11.1 Short-Term Price Relatives Based on Monthly Indices for Four Selected Elementary Aggregates
CPI Structure 2019
Feb. Mar. Apr. May Jun. Jul. Aug. Sep.
Children’s Clothing
Subindex: Underwear, Nightwear, and Socks
Elementary Aggregates:
Underwear 1.010 1.000 1.050 1.005 1.010 1.000 1.030 1.020
Socks 1.000 1.005 1.000 1.030 1.020 1.005 1.000 0.999
Nightwear
Winter Nightwear 1.010 0.970 1.036 1.100 1.050 0.980 0.950 1.014
Summer Nightwear 1.010 0.970 1.036 1.100 1.050 0.980 0.950 1.014
Table 11.2 Monthly Price Indices Using Imputed Prices for Missing Values of Winter and Summer Nightwear
CPI Structure Weights 2019
(base period)
Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep.
Children’s Clothing
Subindex: Underwear, Nightwear, and Socks 1000 118.3 119.2 118.3 122.6 127.1 130.2 129.5 129.4 131.3
Underwear 500 119.0 120.2 120.2 126.2 126.8 128.1 128.1 131.9 134.6
Socks 200 120.0 120.0 120.6 120.6 124.2 126.7 127.3 127.3 127.2
Lower-Level Index: Nightwear 300 116.0 117.2 113.6 117.8 129.5 136.0 133.3 126.6 128.4
Winter Nightwear 150 116.0 117.2 113.6 117.7 129.5 136.0 133.3 126.6 128.3
Summer Nightwear 150 116.0 117.2 113.6 117.7 129.5 136.0 133.3 126.6 128.4
11.24 The example commences with pricing conducted calculations would then proceed as normal. The resulting
in February. As this is the winter season (in the Northern indices are given in Table 11.2. The figures in italics are
Hemisphere), no prices are available for Summer nightwear. based on imputed prices as in Table 11.1.
Under this method, the price movement for Summer night- 11.28 This method can also be applied to other prod-
wear is imputed to be the same as for Winter nightwear. The ucts, such as fresh fruit and vegetables, as long as the CPI
same situation occurs in March. has an appropriate structure. For instance, seasonal fruits
11.25 However, in April no prices were collected for should be imputed from price movements of other in-season
either summer or winter nightwear items. Price movements fruits, and seasonal vegetables should be imputed from other
in the components judged to be most similar to the aggre- in-season vegetables.
gate nightwear component must be used to impute a price
movement for all nightwear items. To first aggregate the Seasonal-Weight Approach
indices (or long-term relatives), then base the imputation on
month-on-month change of the aggregate index of those two 11.29 The seasonal-weight approach gives a zero
components. In other words, calculate the indices of Table weight to an item when out of season and reverts to the
11.2 first, adding a line for a subindex of underwear and annual fixed weights when the product is in season and
socks, then use the month-on-month change of the subindex available for pricing. It is usually applied within a COICOP
for underwear and socks to impute the relative for missing subclass for which the weight is held constant. When the
nightwear. out-of-season item is assigned a zero weight, the upper-level
11.26 In May, summer nightwear was again available weight is redistributed proportionally to the available in-
and so prices were collected. The price relatives for the season items. For example, take the seasonal food category
summer nightwear items are calculated from the collected that covers fresh fruit and vegetables. Only the item weights
prices. The price movements for summer nightwear items within fresh fruit and vegetables can vary between zero and
in May are measured as the change from the imputed prices the annual fixed weight, though the section weights are fixed
for April to the actual prices collected in May. This method so that, at least at the upper levels, the principle of the fixed
does, therefore, self-correct if the imputed prices do start to basket is maintained.
diverge from the “true” level. 11.30 The underlying assumption is that total expendi-
11.27 While this example has shown price imputa- ture on the relevant COICOP subclass (often synonymous
tion occurring at the elementary aggregate level, this has with an elementary aggregate) does not vary between dif-
been done for simplicity in presenting the method. It can ferent times of the year, the main tendency being for expen-
be thought of as an example where there is just one price diture to switch between similar items only (for example,
observation in each of the elementary aggregates. In prac- consumers will tend to buy other fruits if oranges are not
tice, most CPI systems would impute price movements for available).
individual items in the price samples within an elementary 11.31 The seasonal-weight approach has the advantage
aggregate using the same figures for the imputation. Index of minimizing the practice of price imputation, as prices
238
will not be imputed in those months when the product is not on measured inflation assessed for reasonableness before a
available for purchase. Prices are observed only in months particular method is implemented. The use of the index will
where weights are not zero. For products that have prices also be an important consideration. For example, if the main
observed in two consecutive months, the monthly changes purpose of the index is for indexation of payments and sala-
of the product price indices are computed using matched ries, then the fact that the seasonal-weight approach means
samples. When a product has a positive weight after the that the month-on-month price change will reflect not only
weight has been zero for some months, the product index changes in price relatives but also changes in consump-
is compiled by matching the price observation in the first tion baskets may be considered a significant disadvantage
month of the new season with the observations from the (that is, users might question the meaning of comparing the
last month of the previous season and applying the relative price of this month’s basket with the price of last month’s
change to the last index of the previous season to give the basket that might be very different). More generally, if the
price index for the first month of the new season. user focus is on month-on-month change, then the seasonal-
11.32 The seasonal-weight approach has two key weight approach is recommended, despite the problems of
disadvantages: interpretation, as the annual weights attached to each month-
on-month price relative under the fixed-weight approach can
• It is conceptually inconsistent with a fixed-basket index.
be misleading. If user focus is on long-term index changes,
• Month-on-month price changes reflect not only changes then the use of an annual basket and annual expenditure
in price relatives but also changes in the weights (from shares is the most appropriate.
zero to the fixed annual weights approach). This makes it
difficult to interpret month-on-month changes in the price
index. Monthly Approach: Maximum
11.33 In addition, concerns have been expressed about
Overlap—Not an Alternative
the variability in the precise timing of the seasons from one 11.39 The maximum overlap approach deals directly
year to the next, meaning that the imputation of prices is not with the seasonality problem, but strictly speaking is not an
totally avoided. For instance, if unusual weather conditions alternative to the fixed or seasonal weight options for deal-
delay the appearance in the market of oranges, then prices ing with particular products. Rather it is a method of index
would need to be imputed for those months that oranges are construction that overcomes the challenges of seasonal prod-
unavailable but where a nonzero weight has been allocated. ucts by considering only those observations present in both
11.34 It is strongly recommended that the set of prod- compared months. In this case, it would seem reasonable at
ucts defined as seasonal should not vary from year to year, first glance to prefer a chained index, that compares prices
unless strongly justified on the grounds of necessity to keep in adjacent months, because a fixed-base index could fail
the sample representative. to effectively follow closely market developments. This is
because of the dynamics in the retail market resulting from
seasonality, the introduction of new products, and the disap-
Fixed or Seasonal Weights? pearance of older ones. The expenditure shares for month
11.35 Comparing the fixed- and seasonal-weight m + 1 are calculated excluding the products not priced in
approaches, the two approaches provide similar results. that month.
The imputation of prices based on the price movements of 11.40 The main disadvantage of the maximum over-
similar products—the fixed-weight approach—is a form of lap price index is that it can have a significant downward
reweighting where more weight is given to the price move- bias because of the chained nature of the index. Seasonal
ments of products that are available for pricing. products tend to enter the market at relatively high prices,
11.36 An “ideal” solution to the treatment of seasonal which drop in the subsequent months. The initial high prices
items does not exist, particularly where the existence of are not always captured by the maximum overlap index.
seasonal products (especially strongly seasonal products) When these products first become available, they come into
means that it is impossible to compute a completely satis- the market at relatively high prices and then, in subsequent
factory month-on-month index that accurately measures months, their prices drop substantially. The effects of these
month-on-month price change. initially high prices (compared to the relatively low prices
11.37 If the focus of the CPI is the accurate measure- that prevailed in the last month that the products were avail-
ment of annual price inflation, then the challenges associated able in the previous year) are not captured by the maximum
with strongly seasonal products are minimized (although overlap month-on-month indices, so the resulting indices
changes in seasonal patterns from one year to the next con- build up a significant downward bias. The downward bias is
tinue be problematic).3 most pronounced in the Paasche indices that use the quanti-
11.38 The advantages and disadvantages of each ties or volumes of the current month. Those quantities are
method should be considered and, if possible, the impact relatively large compared to the initial month when the
products become available, reflecting the effects of lower
prices when the quantity in the market increases.
3
The focus on annual price inflation leads to the possibility of a further 11.41 The maximum overlap method is not recom-
solution—the construction of 12 year-on-year monthly Lowe indices (or mended when there are particularly large price variations.
geometric Young indices or four quarter-over-quarter indices), using sea- The month-on-month maximum overlap method is gener-
sonal baskets that are appropriate for each month or quarter. From a practi-
cal point of view, annual household budget surveys (HBSs) would have
ally not recommended because of the possible chain drift
to be augmented to cover seasonal products in more detail and this would bias. However, it can be checked if there is a problem with
entail higher costs. the use of the method in practice by comparing the product
239
of 12 consecutive month-on-month maximum overlap indi- been observed but were not (and have not been included in
ces with the corresponding direct year-on-year comparison. the index or the imputation of missing prices) can affect the
If the product of the 12 links is reasonably close to the cor- stability of the price index. Thus, the fact that seasonal pat-
responding year-on-year index, then the month-on-month terns change and the need to confront whether to accept this
overlap method can be used. or force data into a predetermined seasonal pattern poses
a dilemma to CPI compilers. It is particularly important
Implementation Problems that when data are published, the national statistical office
(NSO) should provide background explanations for both
The Identification of Seasonal Items in the the monthly changes and the 12-month rate of change, and
Basket and the Corresponding Length of the whether this is influenced by a change in the timing of sales
Seasonal Cycle and in what direction and by what magnitude.
11.42 The most important characteristics that can be
associated with a strongly seasonal product are: it is not Domestic Products Being Replaced with
available for purchase during certain months (or the quan- Imported Products When Out of Season
tity available for purchase is negligible); there is a signifi- 11.45 When domestically produced, the product
cant variation in demand during the year; and there are description is especially important. The CPI compiler will
corresponding large fluctuations in price. The variations in need to make a judgment on whether an imported apple is
supply and demand, mean that prices cannot be observed of the same quality as a domestic one and if so whether it
during the out-of-season months. Seasonal products are can be treated as an equivalent or “comparable” fruit. This
associated with variations both in availability of products judgment should be based not just on the variety of apple, its
and in demand and this can lead to pronounced variations in general condition, and flavor, but also on whether the market
stocks in outlets, expenditure levels, and prices. Any period perceives it to be the same and treats it as the same.
of nonavailability for pricing in outlets should have some 11.46 To correctly define those months that the price
typical annual cyclical pattern. If for some reason a product index of the seasonal product will be calculated from
becomes unavailable in a month where it would normally observed prices, particularly in the situation where the meth-
be available and sold at a normal price, this does not make odology requires a fixed seasonal pattern, the NSO should
it a seasonal product. The out-of-season period of a product examine in detail the historical data on prices and avail-
includes the months in that no prices are observed or used in ability for each product and undertake a sensitivity analysis
the compilation for that product, because it is not available to identify working rules for determining the inclusion and
for pricing or only available in small quantities in very few exclusion of products in the index in specific months. These
outlets at abnormal prices. The inclusion of abnormally high working rules include:
“start-of-season” prices collected from relatively few out- • The data on availability are needed to decide whether it
lets can potentially introduce a downward bias in the index is likely that in any specific month the product will keep
for both the fixed-weight and seasonal-weight approaches. returning to the market every year.
It may also cause difficulties for price collectors trying to
• The price data will show how the inclusion of the prod-
find the products.
uct in the index for the first month that the product returns
The Geographical Dimension to the market may contribute to the volatility of the index.
11.43 The seasonal availability of products may vary For example, if the entry price is particularly high (for
between regions within a large country. There may also be a example, more than 250 percent higher than the product's
difference in supply (and demand) between markets in rural last actual price), then it would be reasonable to consider
and urban areas, or between market stalls and more upmar- not including the product for that month, especially as it is
ket outlets catering to higher incomes. For example, some improbable that the product was widely purchased at such
prices may be collectable in the capital city even though the a high price.
product is generally not available elsewhere. Where this is • Strongly seasonal products enter the market at high prices
the case, different seasonal cycles may be applied, particu- that decline toward the end of the season but can go up
larly if the differences are marked and regional or urban/ sharply before exiting the market. This will also have
rural subindices are being computed. The decision should an impact that may vary depending on the calculation
take account of the extent of the differences and the needs approach used.
of users.
Common Seasonal Products
Lack of Well-Defined Seasons and Their Treatment
11.44 In practice, seasonal products do not always
appear and disappear in the same month every year. This Fruits and Vegetables
presents a problem both for year-on-year monthly compari- 11.47 The prices of seasonal fruits and vegetables are
sons of monthly indices, and for year-on-year comparisons of strongly dependent on supply conditions, such as weather,
quarterly indices, but is particularly pronounced for the for- and are thus liable to show extreme price movements and
mer. The consequence of neglecting the importance of care- shifts in seasons, as well as seasonal unavailability. How-
fully predefined seasons can contribute to a method, failing ever, changes in specification or quality are not common
to correctly reflect the price changes in the price index in the and therefore do not cause the same challenges as are expe-
short term, particularly the seasonal-weight approach. Deci- rienced when dealing with, for example, seasonal clothing
sions made regarding the relevance of prices that could have (see paragraphs 11.49–11.51).
240
Clothing • Impute a price during the period of unavailability, for the

11.48 The treatment of seasonal clothing can be chal- second and subsequent months, by applying the monthly
lenging, especially because fashion is involved. In many movements for clothing items for which prices are avail-
countries, the price patterns of clothing are the result not able to the last normal price.
just of seasonal availability (for example, winter coats or • When the normal season resumes, select a replacement as
bathing suits) but, especially for fashion goods, of discounts similar as possible to the variety that was priced during
in seasonal sales. In addition, the problem of extreme price the previous season and compare its price directly to the
movements is combined not only with seasonal unavailabil- final imputed price of the old variety. It is important that
ity, but by annual changes in the products because of the the index reflects the full extent of the price difference
effect of fashion. New stocks of clothes that arrive in the between the last month that a price was imputed and the
outlets at the beginning of their season may have styles that first month of the new season. If the replacement has a
are different from those that disappeared at the end of the different level of quality, then the price should be adjusted
previous season, so there is the question of whether the new using one of the quality-adjustment methods presented in
styles are different in quality from the old styles. In addition, Chapter 6.
prices generally fall over the course of the season: prices at • Continue price collection using this new variety for the
the end of a season are often discounted as outlets clear their new season.
stocks in readiness for the next season. For seasonal items,
comparisons must be made between the new replacement Chapter 5 includes further guidance on the treatment of sale
products and the old products that disappeared at the end of prices.
the previous season, maybe six months earlier. Dealing with
these issues can be challenging: Intractable Problems and Serious
• Specification changes. Since it can be difficult to make Challenges
like-to-like observations for many categories of clothing, 11.50 In the context of constructing a month-on-month
the rules for judging whether a replacement is comparable index that accurately reflects consumer price inflation, there
to an original item may need to be less rigid or strict. For is no fully effective method of dealing with large monthly
example, this season’s high-fashion coat can be viewed as fluctuations in prices and quantities generated by strong sea-
comparable to last season’s model, unless there are obvi- sonality. The fixed- and seasonal-weight approaches yield
ous differences in important characteristics such as textile the same results. If the user focus is on month-on-month
replacing leather. For high-fashion items, only changes in change, then it has been argued that the seasonal-weight
compositional and material characteristics, if significant, approach is probably the preferred method, as it avoids the
should be treated as quality changes (see Chapter 6). The problem associated with the fixed-weight approach where
primary characteristic of most high-fashion items is the the annual weights attached to each month-on-month price
product brand, which incorporates the fashion element. relative can be misleading, but the fixed basket has been
Keeping the same or an equivalent brand is often the key compromised. From a presentational point of view, indices
to measuring the fashion element. showing the 12-month inflation rate assist in trend analy-
• Discounted end-of-season prices. In the case of a prod- sis but do not easily identify changes in the monthly trends.
uct showing temporary discounts or promotions, where Additionally, month-on-month indices using maximum
it seems likely that the price will return to its normal overlap are prone to downward bias.
level after a short period, then the discounted price 11.51 It should be noted that when estimating the con-
should be included, and no special method is needed tribution of a subcomponent to the change in the all-items
to handle it. If, however, the discounting is seasonal in CPI, the way this is calculated is different for a seasonal
nature and intended to help the outlet clear stocks of component. This is dealt with in Chapters 9 and 14.
old styles, then the issue is more problematic because,
unless specific procedures are applied to ensure a return Internet Purchases
to normal prices, the index will be subject to a system-
atic downward bias. The exclusion of such situations Introduction
is an important consideration in defining the in-season 11.52 The traditional concept of consumers visiting out-
period. Products that have flaws or are irregular should lets to purchase goods and services has been changing over
not be included. recent years. The importance of internet purchases has been
11.49 As a result of the complications discussed in growing significantly. Conceptually, the CPI should broadly
the previous paragraphs, some NSOs employ the follow- represent the expenditure patterns of the reference popula-
ing procedures during the period covering the time when tion; therefore, online purchases should be included in the
a seasonal clothing item is not available. This adopts the CPI to maintain the relevance of the index by properly rep-
fixed-weight approach and imputes missing prices by resenting consumer purchasing habits.
applying price movements, using the overall mean or the 11.53 Internet purchases are a form of electronic com-
targeted mean (see Chapter 6) to the last available normal merce that allows consumers to directly buy goods or ser-
price. vices from a seller over the internet using a web browser.
Consumers find a product of interest by visiting the web-
• In the first month that the item or variety is unavailable, site of the retailer directly or by searching among alterna-
record the price at its “normal” level; in other words, in tive vendors using a shopping search engine that displays
the first month impute the last available normal price. the product’s availability and price at different e-retailers.
241
11.54 In principle, internet purchases also include in- of delivery charges and credit card surcharges might need to
app purchases,4 purchases via social media, and other similar be addressed. When the prices are the same, CPI price col-
avenues that provide a mobile, desktop, and internet-based lection could simply switch from collecting prices in store
sharing application. These services can be extended to allow to collecting prices online to save on collection costs. Where
the consumer to get the product name and price informa- prices and price movements are similar in both modes of
tion for items in a photo. By clicking on a specific item, collection, the different market shares of online versus in
consumers are taken to its product page, and if they click store will not play a role in the calculation of average price
the “Shop Now” button, they are taken to the brand's web- movements. The fundamental question is whether prices
site to complete their purchase. While these services have online do in fact match the prices in the physical outlet. If
been excluded because of the practical difficulties of iden- not, online purchases should be treated as a distinct outlet
tifying these forms of purchase and estimating the volume type requiring separate and proper representation in the CPI.
and value of purchases, there is a need to collect informa- 11.60 A study by Cavallo (2017) undertook a com-
tion from households on how they make purchases on the parison of online versus offline prices in large multichannel
internet. The household budget survey (HBS) can be used to retailers across 10 countries. The findings indicate that the
collect these data. price levels, in general, were identical in approximately 72
11.55 The growth of internet purchases facilitates more percent of the cases across sampled countries; however, for
efficient price collection as collecting prices online can be clothing and electronics, identical prices were found in 83
relatively easy and cheap compared to sending price collec- percent and 92 percent of the cases, respectively. Based on
tors to outlets. However, identifying and measuring online these findings, there is scope for the internet to be used as an
purchases to develop weights and augment outlet samples alternative mode for price collection when the same prices
can be a challenge. are available online and in the traditional physical outlet.
11.56 The practical measurement challenges of includ- However, when using the internet for price collection the
ing internet purchases in the CPI share some common issues NSO should confirm that the online prices do in fact match
with other types of shopping. Whether internet purchases the in-store prices and that the strong correlation between
are any different from any other form of retailing is debat- movements of in-store and online prices continues. The rela-
able. Perhaps the main distinction from more traditional tionship between in-store and online prices can change over
forms of shopping is the practicality of measurement, that time, so this should be monitored.
is, estimating the expenditure weights and collecting price 11.61 The simplest method for determining whether
observations proportional to sales. a website matches in-store prices is to concurrently price
11.57 As mentioned in Chapter 5, prices for goods and in-store and online items for a few periods to confirm that
services sold by web-based outlets can be collected in the both the price and price change are the same. Another option
same way as the online collection of prices from the web- would be to meet with the respondent in the outlet and dis-
sites of outlets with a corresponding physical location. The cuss whether the prices online are similar to those in the
sample of items to be priced should be representative of physical outlet. If prices and price changes are considered
online purchases and may have different prices from the different and the expenditure is considered representative
ones charged online by physical outlets. The prices recorded and significant across both purchase channels then ideally
should represent the full cost of purchase, including any both modes of purchase should be priced.
tax and surcharges. Online purchases may include delivery 11.62 From a more strategic viewpoint, the NSO should
charges. For CPI compilation, charges that are directly con- decide if internet purchases in general should be consid-
nected to the purchase of the priced product and that are not ered as purchases from a distinctive type of outlet, in which
separately invoiced should be included in the price. If the case they should be fully integrated into the structure of
charge is separately invoiced or relates to the purchase of a the elementary aggregates and into the CPI outlet and item
number of items, then it should be included under transport samples. It would follow that internet purchases should be
services. treated as a separate stratum for sample selection and ele-
11.58 This chapter explores the conceptual issues and mentary aggregation.
challenges as well as the practicalities of incorporating
internet purchases into the CPI. The underlying principles
of measurement are the same as with collecting prices from Estimating Expenditure Weights for
physical outlets. Internet Purchases and Selecting
Samples
Coherence and Data Integration 11.63 The requirement that a broadly representative
basket of goods and services is used to compile the CPI and
11.59 Internet purchases can be made through retailers that the same basket is repriced month after month can be
that have both an online and physical presence (multichan- particularly challenging for internet prices. This is espe-
nel retailers) or just an online presence (web-based retail- cially so when having to account for changes in the charac-
ers). For multichannel retailers, there may not be a need to teristics of products, and their disappearance from a website
price both methods of purchasing if there is no difference in from one month to the next, along with the appearance of
price, terms and conditions of purchase, or product avail- new or updated items and varieties. Internet purchases rep-
ability. For the online presence of the outlet, the collection resent a share of the data universe available for the products
that are to be sampled and, as such, represent a particular
4
In-app purchasing refers to the buying of goods and services from inside outlet type. Online outlet and product selection are based on
an application on a mobile device, such as a smartphone or tablet. the same criteria and methods used for traditional collection,
242
drawing on information on modes of purchase for different of internet purchases, therefore, requires a different way of
products as recorded in an HBS and on sales information thinking and special consideration, especially with regard to
supplied by, for instance, the online retailers. The sales the domestic concept. Additionally, internet purchases con-
information should exclude sales to businesses and should tinue to grow in importance.
conform with the geographical concept followed by the CPI 11.68 Many countries have carefully considered how to
(see paragraphs 11.65–11.72); however, as this can be diffi- include the expenditure (and prices) made on goods and ser-
cult to separately identify from sales data supplied by retail- vices via the internet. For the purchase of goods, the expen-
ers, information from HBS is often preferred, despite the diture and prices should be reflected in the country where
fact that it is sample-based and can be out of date. HBS data the goods are delivered.
can be updated using commercial information on trends in 11.69 Services purchased on the internet can be more
internet purchases, but care must be taken to ensure consis- problematic for CPI compilation because there are both tan-
tency of weights to reflect expenditure patterns in the weight gible and digital services. Tangible services would include
reference period. traditional services such as transportation, hotels, entrance
to cultural/sporting events, or education. Digital services
Geographical Coverage would include telecommunications, broadcasting (for
example, streaming or downloading music, movies, or tele-
11.64 Internet purchases can be made from either vision content), and other services (for example, software).
domestic websites or foreign websites. The treatment of If the service is consumed in the economic territory where
these transactions can cause challenges for producing a the household is resident, it should be included in the CPI;
CPI and is dependent on the geographical coverage of the however, if the service is consumed outside of the economic
CPI. The CPI follows either the national or domestic con- territory of the country, it would be excluded. For example,
cept, as described in Chapter 2. The concept followed deter- if a household reserves a hotel room that will be used and
mines from a geographical perspective what expenditure paid for in another country, it would be considered out of
is included in the weights and what goods and services are scope. For digital services, because the service is being con-
priced. sumed within the economic territory of a country, the expen-
diture and prices should be included in the country where
The National versus the Domestic Concept the household resides.
11.65 The national concept means that the CPI should 11.70 The estimation of expenditure weights can be a
cover all expenditure (and prices) relating to the resident more challenging measurement issue. When expenditure
households of the country, regardless of where the expendi- or alternative sources of data are not readily available, but
ture takes place. The national concept aligns with the System internet purchases are known to be substantial, an NSO
of National Accounts (SNA).5 The weights for expenditure could investigate augmenting the CPI basket using an esti-
abroad can be computed from the HBS, at least in theory, mated weight until the precise weight can be established.
but measuring prices paid abroad can pose significant practi- By doing this, the relevance of the CPI basket is better
cal and operational problems especially for purchases from maintained.
physical outlets as opposed to purchases made online. 11.71 For item and variety selection, online retailers
11.66 The domestic concept means that the CPI should often identify their top-selling products on their websites.
cover all the expenditure made by households within the This can help with identifying those products to price; how-
economic territory of the country, including the nonbusiness ever, the bestseller or most popular item lists may be skewed
expenditure made by foreign visitors. It excludes the expen- by the web-based retailer to sell specific items. CPI staff
diture abroad of the resident households. should carefully review these lists to identify those items
11.67 As discussed in Chapter 2, the treatment of pur- that represent the most popular or best seller.
chases made online requires special consideration. In prin-
ciple, the domestic and national concepts could provide Determining the Location of the Online Retailer
guidance on how to treat the expenditure made on goods and and of the Transaction
services, including digital downloads, purchased online. In 11.72 The location of the purchaser, the location of the
many cases, however, internet-based outlets may be based website, and the location of the transaction may be spread
(registered) abroad and this expenditure would be consid- across different countries. Several alternative locations may
ered cross-border shopping. For those countries following be considered in determining whether an online transaction
the national concept, the approach is clear. Strictly speaking, complies with the national or domestic concept:
under the domestic concept, this expenditure would not be
included because it would be defined as expenditure abroad; • Purchaser’s domicile at the time of the order
in practice, this requires a broader interpretation. The nature • Address where the product is delivered to
• Purchaser’s address
5 • Billing address
Another issue raised by the increasing use of the internet for purchases is
the distinction between goods and services. The Impact of Globalisation on • The location where the product is consumed
National Accounts (UNECE 2015) includes a chapter on e-commerce. The
• Country where tax, most particularly value-added tax
chapter includes a table (Table 13.1) on whether to classify different types of
internet purchases as goods or services according to the 2008 SNA concepts (VAT) or goods and services tax (GST), is paid
and the Sixth Edition of the IMF’s Balance of Payments and International
Investment Position Manual. In the discussions about the problems in mea-
11.73 As noted previously, internet purchases force
suring globalization the issue of having suitable price indices for deflation of a broader interpretation of the domestic concept. In this
flows of goods and services is often mentioned as a major challenge. case, a pragmatic approach needs to be adopted rather than
243
establishing and adhering to rules and regulations designed services) are recorded in the HICP of the country where
to maintain strict consistency with national accounts. For the consumer usually resides.
this reason, alternative workable approaches are some- • The price to be recorded should be the full price, includ-
times used. Feldman and Sandberg (2012) suggest using ing any compulsory additional costs, provided that these
the address where the product is delivered to as a practical costs can be attributed solely to the purchase of the prod-
solution. The country of delivery is where the consumer and uct concerned.
the product come together. If the purchaser's country is not
listed as the standard location for the delivery of purchases
from the website, then it may be assumed that the seller’s Price Collection
location is in another country. Following this approach, 11.76 With regard to price collection and the mainte-
downloads of music, e-books, or software should be classi- nance of a fixed basket, collecting prices online poses simi-
fied as domestic purchases as these are normally delivered lar, but more difficult to resolve, challenges to collecting
instantaneously to the user. Although, conceptually this prices in physical outlets, such as nonavailability of prod-
deviates from the concept of a foreign internet purchase, ucts and the issues of substitution and quality adjustment,
as it will most likely capture purchases from some foreign- as well as the treatment of delivery charges.7 The standard
based websites, practically it is considered by some as the principles of price collection apply. Under traditional price
next best alternative. However, it can lead to inconsisten- collection, if a product is sold out or is no longer available,
cies between a CPI and the coverage of both the national the price collector can speak to the outlet staff and choose
accounts and of the balance of payments. The approach a replacement item and make any necessary quality adjust-
adopted by Eurostat for the EU HICP uses administrative ments. With internet price collection, it is unlikely that there
rules relating to taxation as the basis for determining how will be an opportunity for interaction with the retailer, and
to treat internet purchases.6 Eurostat's recommendations on the availability of an item can often only be determined by
the treatment in the EU of cross-border internet purchases attempting a purchase. Moreover, if an item is unavailable
borrow from the VAT rules as applied to all member states. it can be difficult to determine online whether it is perma-
The VAT rules consider the fact that an increasing number nently or temporarily unavailable and, in the former case,
of products are electronic (for example, an e-book), rather the detailed product information needed to select a replace-
than a physical good or service in the traditional sense. The ment and make any necessary quality adjustment is gen-
rules make a distinction between goods, digital services, and erally not accessible given the limited information that is
other services. For goods ordered on the internet, the place contained on a retailer’s website. NSOs should investigate
of delivery determines the VAT rate, and usually, this will potential websites for online price collection to ensure that
be the country of residence of the purchaser. For digital ser- there are a sufficient number of products and enough prod-
vices, the VAT rate is determined by the country where the uct detail to price consistently and to constant quality. For
purchaser normally resides. For nondigital services, even example, for products like clothing, websites should contain
if booked online, such as flights and package holidays, the enough characteristics, such as brand, material, style, and
VAT is determined by the country where the tangible service cut, to facilitate the selection of a suitable replacement for
is provided. The recommendations also cover purchases an article of clothing that is discontinued and, where neces-
ordered over the telephone or by mail order catalog that are sary, to perform a quality adjustment.
treated in the same way as internet purchases because of the 11.77 Platform websites pose additional challenges
similarities between different forms of remote purchase. for the collection of prices. Many websites sell products
11.74 In the EU, the scope of the HICP follows the directly to consumers, while others serve more as a virtual
domestic concept, as it covers all household final monetary marketplace selling products from a variety of different
consumption expenditure (as defined on the EU Regulations web-based and online retailers. If the website sells prod-
on HICP and described in Chapter 2) in the economic terri- ucts directly to consumers, it should be treated as an outlet
tory, regardless of the nationality or normal residence of the for purposes of price collection. If the website is more of a
consumer. virtual marketplace, the individual retailers can be regarded
11.75 According to Eurostat, applying the VAT rules to as separate outlets. This would be similar to how physical
CPI compilation concerning internet purchases results in the shopping malls are treated in the CPI. A shopping mall is
following: not treated as a single outlet; rather the individual retail-
• The expenditure and the prices for goods purchased ers within a shopping mall are treated as separate outlets.
through the internet are recorded in the HICP of the coun- A platform website serving as a virtual marketplace can be
try where the product is delivered. treated in the same way.
• The expenditure and the prices for tangible services pur-
chased through the internet are recorded in the HICP of Treatment of Additional Costs for Internet
the country where the service is provided. Purchases
• The expenditure and the prices for digital services (for 11.78 Often, when purchasing goods and services on the
example, communication, broadcasting, or electronic internet, there are additional costs associated with buying
7
As mentioned in Chapter 5, delivery charges should be included in the cost
6
http://ec.europa.eu/eurostat/documents/272892/7048317/HICP- of the online purchase where these are not separately invoiced and where
recommendation-on-internet+purchases-December-2016/. they are inseparable from the purchase.
244
that product. These costs can include, for example, delivery conceptual approach and the practical constraints relating
charges or credit card fees. In some cases, these are sepa- to data availability. The treatment of rented accommoda-
rately invoiced and in other cases, they are not. It can also tion is more straightforward, and the costs borne by tenants
be noted that some additional costs only become apparent at are often used to impute owner-occupied housing services
the point of purchase. It can be argued that if the cost is ines- costs.
capable to the consumer, that is, if ownership of the product
cannot be transferred unless these costs are incurred by the
household, then it should be included in the CPI as part of Owner-Occupied Housing Services
the transaction. The delivery charges may vary according 11.82 The treatment of owner-occupied housing ser-
to the geographical location of the purchaser and this varia- vices costs in CPIs is arguably one of the most difficult
tion will need to be included in the national CPI and in a issues faced by CPI compilers. Depending on the propor-
regional or city CPI. The approach can also be applied when tion of the reference population that are owner-occupiers,
the delivery is provided by another business, if this is part of the alternative conceptual treatments can have a significant
the transaction. In this case, the cost of transport would be impact on the CPI, affecting both weights and, especially in
priced separately and included under transportation services the short term, measured inflation.
for purposes of index compilation. 11.83 Ideally, the approach chosen should align with
11.79 Delivery charges. COICOP 2018 classifies deliv- the conceptual basis that best satisfies the main use of the
ery charges separately from the purchase price. This will CPI. However, the data requirements may be such that it
allow the CPI to more accurately reflect the appropriate price is not feasible to adopt the preferred treatment. Also, the
change. In COICOP 1999, delivery charges are included in CPI may be multipurpose and it may be difficult to ascer-
the price paid for the product. If the delivery fee changed, tain the main use of the index. The dual use of CPIs as both
this would be reflected as a price change for the item. For macroeconomic indicators and for indexation purposes (as
example, if the price of a table was $500 in period t1 and described in Chapter 2) can lead to clear tensions in design-
remained unchanged in period t2 at $500, and the delivery ing an appropriate treatment for owner-occupied housing
fees increased from $50 to $70, under COICOP 1999, the services costs that suits all needs. In these circumstances,
final price would show a change from $550 to $570 because it may be necessary to adopt a treatment that is not entirely
the CPI compiler includes the delivery fee in the price of the consistent with the approach adopted for other items in the
item. Consequently, even though the price of the item has CPI. National house market structures and practical mea-
not changed, the index is showing a movement that relates surement issues are also important considerations in deter-
not to the product itself but to a transport service.8 mining which approach to apply in practice.
11.80 Credit card fees should be captured if it is an 11.84 The inclusion of owner-occupied housing ser-
inescapable cost of buying the item and incurred by most vices costs will improve the representativeness and rel-
households buying that product. Some websites charge mul- evance of a CPI. When used in an international context, for
tiple fees for the use of different credit cards. If this is the example, to measure economic convergence, the inclusion
case, markup rates (percent charged to use a specific credit of owner-occupied housing services costs should, in prin-
card) should be obtained where possible to weight the dif- ciple, enhance the comparability of CPIs across countries.
ferent fees; otherwise, an estimate could be derived by However, for this purpose two major concerns have been
taking an average of the different fees. For example, if an expressed. First, given the relatively high weight for owner-
airline charged a fee of 1.3 percent for the use of a credit occupied housing in most countries, including owner-
card and 0.6 percent for the use of a debit card, and it was occupied housing services costs could impact the inflation
estimated that the take-up rates were 70 percent and 30 per- rate and its volatility. Second, it could add to cross-country
cent, respectively, the compiler would calculate a weighted divergence in inflation rates and may make it more diffi-
percentage fee of 1.09 percent. Alternatively, if no take-up cult for a country to meet inflation convergence criteria set
rates are available the compiler could average the two card for a block of countries, because of differences between the
fees together generating 0.95 percent. This rate can then be relative importance of each country’s owner-occupied hous-
applied to the price of the airline ticket. ing market and divergences in the respective movement in
house prices across countries.
11.85 Depending on the methodological approach used,
Housing data on owner-occupied housing services costs may not be
Introduction as timely as other data for the CPI, and the compilation of
representative owner-occupied housing services costs indi-
11.81 This section first describes the treatment of ces might only be feasible on a quarterly basis. Both aspects
owner-occupied housing services costs and then of the costs will have an impact on the underlying statistical quality and
borne by tenants. The treatment of owner-occupied hous- usability of the CPI.
ing services costs in the CPI depends on the agreed-upon 11.86 Once a decision has been made to include owner-
occupied housing services costs in the CPI, all the previous
considerations need to be taken into account by an NSO in
8
Although this classification is different from the SNA, which defines the deciding which approach to adopt. The criteria for choos-
“purchaser’s price” to include any transport charges paid separately by the ing which approach will include: alignment with user needs
purchaser to take delivery at the required time and place (2008 SNA, para-
graph 6.64) when this is not feasible many countries’ national accounts
and the main purpose of the CPI; consistency with the rest
deviate from this approach and include delivery charges under transport of the CPI; alignment with international practices (although
services rather than allocating the cost to the price of the product. in the case of housing services costs there is no single
245
agreed methodology); public acceptance; and practicality of capital gains.9 From an aggregate perspective, this implies
implementation. computing the proportion of homeowners who have a mort-
gage and the average size of mortgage and apply to this
The Different Conceptual Approaches the mortgage rate. The other rate in the user cost formula
must be a riskless nominal interest rate that is applicable to
The Use Approach: Valuing the Flow of Services
capital.
11.87 The general objective of the use approach is to 11.91 In general, NSOs do not use the full user cost
measure the change over time in the value of the flow of approach. This partly reflects the conceptual and method-
shelter services consumed by owner-occupiers. Detailed ological complexity of the measure, which may also make it
approaches fall under one of two headings: user cost and difficult to obtain widespread public understanding and sup-
rental equivalence. port for the approach. For this reason, the methodology is
11.88 The user cost approach attempts to measure the not discussed in detail in this Manual. It is, however, worth
changes in the cost to owner-occupiers of using the dwell- noting that both the weights and the ongoing measures of
ing. In the weight reference period, these costs comprise two price change are significantly influenced by the relative rate
elements: recurring actual costs, such as those for repairs of change in house prices. Since the user cost formula is
and maintenance, and property taxes; and the opportunity typically dominated by capital gains and interest rates, the
cost of investing in the dwelling rather than use the amount user cost weight can be negative (implying a negative price
invested for some other purpose. At its simplest, and where for user cost) if house price inflation exceeds nominal inter-
houses are purchased outright, this opportunity cost is est rates.
represented by the rate of return available on alternative 11.92 In practice, it is possible to avoid some of these
assets. More usually, house purchase will be at least par- difficulties by adopting a variant or a narrower definition of
tially financed through mortgage borrowing. In this case, user cost. For example, some countries have adopted a vari-
the opportunity cost can be viewed as an average of inter- ant of the user cost approach focusing on gross mortgage
est rates on mortgages and the alternative assets, weighted interest payments and depreciation, in part, because these
by the proportion of the purchase price borrowed and paid items are readily recognizable as key costs by homeowners.
outright, respectively. The mortgage interest may be viewed as the cost of retain-
11.89 Estimation of the base period weights for recur- ing housing shelter today, while the depreciation element
ring actual costs, such as expenditure on repairs and mainte- represents current expenditure that would be required to
nance, is generally obtainable from the HBS. However, care offset the deterioration and obsolescence in dwellings that
must be taken to distinguish between routine repairs and would otherwise occur over time. Methodologies for calcu-
maintenance that maintain a dwelling in its original condi- lating actual average mortgage interest payments for index
tion, and alterations and additions, that represent significant households are described in the section on the payments
functional improvements in the dwelling (for example, con- approach to owner-occupied housing services costs (para-
verting an attic into a room or building an extension). Alter- graphs 11.128–11.141).
ations and additions are outside the user cost approach. In 11.93 The use of mortgage interest payments in the user
practice, distinguishing between major repairs and mainte- cost approach and the payments approach, as described in
nance, on the one hand, and alterations and additions, on the paragraphs 11.128–11.141, may pose conceptual and practi-
other hand, may prove difficult when relying on information cal difficulties for some users depending on the stated pur-
gathered from an HBS. In contrast, the construction of price pose of the CPI. For almost all items in the CPI basket, an
measures for these items generally presents few difficulties. increase in price represents an increase in living costs for
11.90 Estimation of the base period weight for opportu- the target household population in aggregate. However, this
nity costs is more complex and will require modeling. One
is not necessarily the case for mortgage interest payments.
simplistic approach is to assume that all owner-occupiers An increase in interest rates generally benefits savers, which
purchased their dwellings outright at the beginning of the would include a considerable share of the target population.
period and sold them at the end. During the period their An increase in savings interest rates will leave this share
opportunity costs comprise the amount of interest forgone of the target population commensurately better off. For
(that is, the amount of interest they might have earned by pensioners in particular, who tend to be savers rather than
investing this amount elsewhere), transaction costs, and borrowers, indexation of pension entitlements to a CPI that
depreciation. Offsetting these costs would be any capital is based in part on mortgage interest charges may perplex
gains earned on the sale of the dwellings. Construction of the general public.10 Explaining to those financially affected
the required measures of price change is also quite complex when interest rates fall why mortgage interest payments are
and, particularly for the depreciation element, a good deal in the scope of the CPI whereas savings interest is out of
of imputation is required, including allowing for dwelling scope may prove difficult. A more fundamental concern,
purchases partly financed by mortgage borrowing. A typical
formula for user cost (UC) is
9
An alternative to this approach would be to use ex ante or forecasted costs
UC = rM + iE + D + RC − K (11.1) for a suitable period. Randal Verbrugge. 2008. “The Puzzling Divergence
of Rents and User Costs, 1980–2004.” Review of Income and Wealth,
where M and E represent mortgage debt and equity in the Series 54, Number 4, Bureau of Labor Statistics, Washington, DC, dis-
cusses various methodologies.
home, and r and i represent mortgage interest rates and the 10
In general, the primary purpose of indexation of payments is to hold
rate of riskless return available on alternative assets, respec- payment recipient’s purchasing power constant rather than eliminate the
tively. D is depreciation, RC other recurring costs, and K effects of financial shocks on payments.
246
from a macroeconomic perspective, is that including mort- rents weighted to reflect the current composition of the stock
gage interest rates in a CPI diminishes its relevance and of owner-occupied housing, that can then be applied to the
usefulness for monetary policy purposes, as interest rates rental equivalents in the weight and price reference periods,
are one of the main macroeconomic levers for controlling or from acquiring on a regular basis from an expert such as a
inflation. real estate agent or broker, the current equivalent rents for a
11.94 A way of overcoming the issue of indexation for sample of houses with different characteristics that are rep-
savers and borrowers under a user cost approach is to com- resentative of the owner-occupied housing stock.12
pile population subgroup indices alongside the official CPI. 11.97 The rental equivalence approach is considered as
The population subgroups can be derived through income, a viable option by many countries, but there is a requirement
wealth, or stages of the life cycle, and can be weighted using to have a transparent rental market and reliable information
HBS data. A separate index for that portion of the popula- on rents by type of accommodation, location, and other
tion receiving indexed pension entitlements can be created rent-determining factors. A number of countries use this
excluding mortgage interest charges thereby removing the approach for conceptual and practical reasons. For many
payments that are not relevant for that subgroup of the popu- developing economies, newly constructed dwellings (some-
lation. Such disaggregated estimates can assist in the formu- times self-builds) on family-owned land, or older dwellings
lation of policy. More generally, it is open to countries to that have been significantly upgraded, make up a signifi-
compile supplementary CPIs aimed at measuring the infla- cant proportion of the housing stock. Differences in quality
tion experience of different segments of the population. between newly constructed dwellings can be significant and
Many would argue that, if a certain subgroup of the popula- housing mobility may be low resulting in limited markets
tion is to be compensated for increases in the living costs, for the sale of dwellings. This can be further complicated
then, in principle, expenditure weights should be constructed by a lack of formal transfers of ownership or of transpar-
for this subgroup accepting that this reenforces the continu- ent property rights. The lack of this information, normally
ation of the current expenditure patterns of the subgroup.11 readily available in more developed economies, makes the
11.95 Depreciation is a gradual process and so is best rental equivalence approach a practical option for incorpo-
represented by the amount that needs to be estimated each rating owner-occupied housing services costs in a CPI for
year as opposed to actual expenditure (that will typically be developing economies. Many developed economies also use
large but infrequent). The base period weight for deprecia- this approach in their CPIs, for instance, if their purpose is
tion may be estimated from the current market value of the to compile a cost of living index (COLI), or their aim is to
owner-occupied housing stock excluding land values, mul- more closely align with the national accounts.
tiplied by an average rate of depreciation. The latter may be 11.98 Deriving the weight for rental equivalence
derived from national accounts estimates of housing capital requires estimating how much owner-occupiers would have
consumption. Imputed this way, the appropriate price indi- paid in the weighting base period to rent their dwellings.
cator should ideally be an index of house prices, excluding This is not something that owner-occupiers can normally be
land, rather than an index of the costs of renovation work. expected to estimate and reliably reported in an HBS. How-
Discussions on how this could be achieved are included in ever, in principle, it can be estimated by matching the dwell-
paragraphs 11.128–11.141. ings of owner-occupiers with comparable dwellings that are
11.96 The rental equivalence approach attempts to being rented and applying those rents to the owner-occupied
measure the change in the price of the housing services con- dwellings. In practice, this raises a number of problems, par-
sumed by owner-occupiers by estimating the market value ticularly in countries where the overall size of the private
of those services. In other words, it is based on estimating rental market is small or where rented housing is of a differ-
how much owner-occupiers would have to pay to rent their ent type from owner-occupied housing with regard to gen-
dwelling. Under this approach, it would be inappropriate to eral quality, age, size, and location. Direct imputation from
also include those input costs normally borne by landlords actual rents may also be inappropriate if the rental market is
such as dwelling insurance, major repair and maintenance, subject to price control.13
and property taxes, as this would involve an element of 11.99 In those countries where the reference population
double counting. The rental equivalence approach is recom- for the CPI corresponds to all resident households, greater
mended in the 2008 SNA (2008 SNA paragraphs 6.117 and collaboration with the national accounts helps address esti-
15.141) for measuring housing services to be included in mation problems such as sample sizes, lower-level weights,
the household final consumption expenditure estimates and and specific market variations. The corresponding price
is also used in constructing international comparisons of series for the rents of owner-occupiers can be derived from
living standards. The price indicator for imputed rents can an actual rent index, except where such rents are subject to
be sourced from either a readily available price series for price control. Depending on both the relative significance
of owner-occupiers to renters and the composition of the
11
The HBS provides data about household characteristics such as income
and number of members. This is useful for ensuring that the expenditure
12
corresponds to those of the CPI reference population and can also be used The use of a private sector data source should be done with care as the
for producing CPIs for different population subgroups. Depending on its methodology employed by the private entity could be potentially predis-
design, the HBS may provide information on the types of outlet where posed to bias and less appropriate than one produced by an independent
purchases are made and on the varieties purchased. This information can authority.
13
be used to construct elementary aggregates at a finer level of detail and In addition, it is also argued that owner-occupiers may be considered to
to improve the sample design for outlets and items for price collection to derive significant additional utility from features such as security of tenure
reflect the differences by population in the varieties of goods and services and the ability to modify the dwelling, implying a need to make additional
purchased and the types of outlets used. adjustments to the initial imputations.
247
housing stock with regard to dwelling characteristics, any compiler will need to adjust the values of the rents collected
existing rent surveys may need to be modified to meet the in the rent survey.
requirements of an owners’ equivalent rent series. If the total
value of owners’ equivalent rent is significantly larger than The Payments Approach
actual rents, the absolute size of the existing price sample 11.103 The scope for a payment index is defined by
may be insufficient. If the characteristics of owner-occupied reference to actual expenditure made by households to gain
dwellings differ significantly from the overall rental market, access to consumer goods and services. Thus, the payments
the existing rent survey may also require stratifying more approach covers the expenditure actually incurred in occu-
finely (for example, by type and size of dwelling, and by pying a dwelling. The set of expenditure specific to owner-
location) if the sample size is large enough to do so. The occupiers in the weight reference period includes:14
price measures for the different strata can then be given
• Down payments or deposits on newly purchased dwellings
different weights when calculating the actual rents and the
owners’ equivalent rent series. • Legal and real estate agency fees payable on property
11.100 While it may be acceptable to include subsidized transfers
and controlled prices in the actual rent series, these should • Repayments of mortgage principal
not be used in calculating the owners’ equivalent rent series. • Mortgage interest payments
Given the increased significance of rent prices in the overall • Alterations and additions to the dwelling
index, it may also be necessary to pay greater attention to
the measurement of price change for individual properties • Insurance of the dwelling
when tenancies change. As a change in tenancy often pres- • Repair and maintenance of the dwelling
ents landlords with an opportunity to refurbish properties • Property rates and taxes
and to increase rents regardless of any refurbishment, the
practice of regarding the total price change as quality change 11.104 While it is conceivable to include all these items
should be avoided. For these reasons, a stock-based index in the index, it is generally agreed that at least some repre-
is most appropriate under the user cost approach. A flow- sent capital transactions that ought to be excluded from a
based rental index compiled from new rental agreements CPI. For example, while down payments and repayments of
may behave quite differently from a stock-based rental the mortgage principal result in a running down of household
index. A stock-based rental index is generally more stable cash reserves, they also result in the creation of a real asset
and, as such, is more representative of the owner-occupied (at least part of a dwelling) or in the reduction of a liability
sector which, by definition, enjoys security of tenure. Fur- (the amount of mortgage debt outstanding). Similarly, any
thermore, the rent series may need to be quality-adjusted to cash expenditure on alterations and additions result in a run-
take account of ongoing depreciation to housing structures, ning down of cash reserves offset by increases in dwelling
depending on the treatment in the CPI of owner-occupiers’ values. In other words, those transactions that result in no
repair and maintenance costs. net change to household balance sheets should be excluded.
11.101 Whereas the rental equivalence approach has the 11.105 The remaining items can be regarded as current
advantage of relative simplicity, requiring only more than expenditure that does not result in any offsetting adjust-
a suitable rental price index and appropriate weights, it is ments to household balance sheets. It is therefore considered
a method based on notional or imputed prices rather than appropriate that these items be included in a payments-based
actual transaction prices. This could be viewed both as a CPI. By defining a payments index in this way, the aggre-
significant conceptual departure from how other items are gate payments equal a household’s source of funds. A house-
treated in a CPI and as an overreliance on imputed rather hold’s total source of funds comprises income after tax (for
than actual prices. example, wages, transfers, property income, and insurance
11.102 Double counting can also be an issue when claims) and net savings (as a balancing item). It is for this
the rental equivalence approach is used. If expenditure on reason that a payments-based CPI is commonly considered
repairs, maintenance, local property taxes, water charges, to be the best construct for assessing changes in net money
and so on are included in rents, these costs should not be income over time.
included elsewhere in the index. It is also important that 11.106 It is sometimes argued that the payments
HBSs determine if the renter household receives any addi- approach is more consistent with the traditional approach
tional services, such as electricity or use of facilities outside to CPI compilation, this is a carryover from a time when
the housing unit, such as off-street parking. It is also impor- the CPI was mostly used as a compensation tool. It is more
tant to learn if the household must pay any additional costs, easily understood by the public and measures costs directly,
such as taxes, that the owner of the dwelling does not pay. thereby avoiding imputation.
The value of any supplementary items provided and the cost 11.107 There are also disadvantages to using the pay-
of any items borne by the tenant should be allocated to their ments approach. The items of direct expenditure by pur-
proper CPI category. For example, the dwelling may have chasers of property include mortgage interest payments,15
water supplied at cost by the landlord: in this case, CPIs
may either leave landlord-supplied water in the rent or move 14
This is not the recommended approach used in the national accounts, and
an estimated value for it from the rent index to an index some of the items listed are not included in household final consumption
for water, but this must be done consistently in the weights expenditure.
15
In the national accounts, the payment of interest is not included in
and the rent survey. Leaving supplementary items in the rent household final consumption expenditure (except the part that is considered
avoids the need to adjust the weight but leaves the potential FISIM), but is counted as an interest payment in the household primary
problem that if the landlord ceases to provide them, the CPI income account.
248
repayment of capital, and large repairs (associated with for real estate agency fees and insurance is discussed in
depreciation from wear and tear as properties get older). All paragraphs 11.378–11.384 and 11.385‒11.403, respectively.
of these expenditure items are generally considered out of Indices for repair and maintenance, and property rates and
the scope of a CPI because they represent capital formation. taxes, are not considered particularly problematic and so are
A CPI should only relate to final consumption expenditure not discussed in this chapter. The remainder of this section
and should exclude financial transactions and investment. addresses the construction of price measures for mortgage
The acquisition of a house will normally represent a substan- interest charges.
tial capital asset over a period of time, a point that emerges 11.112 The construction of price indices for mortgage
when comparing the position of owner-occupiers with ten- interest charges is complex. The degree of complexity will
ants who rent. It can therefore be argued that the capital vary from country to country depending on the operation of
element of mortgage repayments should be regarded as an domestic financial markets and the existence of any income
investment or saving rather than final consumption expendi- tax provisions applying to mortgage interest payments.
ture and should therefore be excluded from the index. What follows therefore is a description of an overall objec-
11.108 When using the payments approach, the ques- tive for producing the required index in the most straight-
tion arises over whether the weight and price indicator forward cases. The methodology will need to be modified to
should be net of any tax allowances for mortgage interest account for additional complexities that may be encountered
payments. It is recommended that, in accordance with the in some countries.
principle that a CPI should be based on the amounts actually 11.113 The general approach to construct price indices
paid, the weight and price indicator should both be based on for mortgage interest charges may be summarized, briefly,
payments after tax relief. as follows. Under a fixed-basket approach, the objective of
11.109 The adoption of the payments approach is that the index is to measure the change over time in the inter-
it requires a large volume of data that may not be avail- est that would be payable on a set of mortgages equivalent
able to the compiler. Mortgage interest payments will be to those existing in the weight reference period. This base
affected by changes in both interest rates and house prices stock of mortgages will vary widely in age, from those taken
in different ways. Changes in interest rates will affect all up in the weight reference period to those taken up many
of those buying a house apart from those on fixed-interest years previously. In compiling a fixed-base index, the distri-
loans, while changes in house prices will affect only those bution of mortgages by age must be held constant.
buying a house in the current period. Thus, a price indicator 11.114 The amount of interest payable on a mortgage
consisting of the current interest applied to a standard-sized is determined by applying some rate of interest, expressed
mortgage to a standard-sized house would not be appropri- as a percentage, to the monetary value of debt. Therefore,
ate. An appropriate indicator involves two components: the changes in mortgage interest charges over time can, in prin-
rate of interest and the average amount of mortgage debt ciple, be measured by periodically collecting information on
outstanding. To calculate the average outstanding debt at a representative selection of mortgage interest rates, using
any point in time can be problematic as it consists of a large these to derive an average interest rate, and then applying
number of individual debts, some from mortgages taken out this to an appropriate debt figure. At least for standard vari-
recently and others from mortgages taken out some time ago able rate mortgages, interest due on the revalued stock of
at historic prices and with some of the debt paid back. It is base period mortgages may be derived simply with refer-
unlikely that all countries will have the necessary data to ence to current mortgage interest rates.
apply this method. 11.115 The main challenge in constructing a price index
11.110 It can be argued that the main disadvantage of for mortgage interest charges is determining the appropri-
the payments approach is that it includes a major (explicit) ate debt figure in each of the comparison periods. Since the
cost of owning a home, namely mortgage interest, but it does real value of any monetary amount of debt varies over time
not include a major offsetting (implicit or imputed) benefit, according to changes in the purchasing power of money, it
namely possible price appreciation or capital gains on the is not appropriate to use the actual base period monetary
home. The neglect of this benefit is particularly troublesome value of debt in calculations for subsequent periods. Rather,
when there is moderate or high inflation in the economy: the it is necessary first to update that monetary value in each
observed mortgage interest cost can increase compared with comparison period so that it remains constant in real terms
other costs and give a very misleading picture of the home- (that is, so that the quantities underpinning the base period
owner’s true long-term costs of living in the home (because amount are held constant).
the offsetting benefit is neglected). But the counterargument 11.116 To update the monetary value of debt, it is nec-
is that the capital gains of owning a house that appreciates in essary to form at least a theoretical view of the quantities
value are of limited relevance when people must bear such underpinning the amount of debt in the base period. The
costs from current income, and the value of the underlying amount of mortgage debt outstanding for a single household
asset can only be materialized if the home is sold and the in the base period depends on the original house purchase
proceeds are not used to purchase another property for own price and loan-to-value ratio, and on the rate of repayment
occupation. This is, perhaps, a case where the measurement of principal since the house was purchased. An equivalent
approach that is adopted will vary depending on the use and value of debt can be calculated in subsequent comparison
purpose of the CPI. periods by holding constant the age of the debt, the original
11.111 Estimation of gross expenditure on the items value of the debt (as some fixed proportion of the total value
listed in paragraph 11.104 in the weight reference period is of the dwelling when the mortgage was initially entered
readily achievable using HBS data, as the items are generally into), and the rate of repayment of the principal (as some
reportable by households. The construction of price indices proportion of the original debt), and applying these factors
249
to house prices for periods corresponding to the age of the Table 11.3 Calculation of a Mortgage Interest Charges
debt. Series
11.117 To illustrate how the update of the monetary Year Quarter Debt Index Nominal Mortgage Interest
value of debt can be done in practice, suppose a household Interest Charges Index
purchased a dwelling five years earlier than base period for Rates Index
$100,000 and financed 50 percent by mortgage. If, between (a) (b) (c) = (a) × (b)/100
the time of purchase and the base period, the household Y0 Q1 100.0 100.0 100.0
repaid 20 percent of this debt, then the outstanding debt on Q2 100.7 98.5 99.2
which base period interest charges were calculated would Q3 101.4 100.8 102.2
have been $40,000. Considering a subsequent comparison Q4 101.9 101.5 103.4
period and supposing that it is known that house prices dou-
bled between the period when the household was originally
purchased and the period five years prior to the comparison
period, then the equivalent amount of outstanding debt in mortgages (for example, banks or building societies) for an
the comparison period would be calculated by first taking age profile of their current mortgage portfolio. This type of
50 percent of the revalued house price (of $200,000) to give data is normally available from financial institutions and is
$100,000, and then reducing this by the principal repayment generally reliable.
rate (of 20 percent) to give $80,000. 11.121 To calculate mortgage interest charges, a nomi-
11.118 Under the assumptions described in paragraph nal mortgage interest rate index series is applied to the
11.117, the comparison period value of outstanding debt aggregate level of outstanding debt. A nominal mortgage
may be estimated directly from the base period value of interest rate index series is obtained by calculating average
outstanding debt based on house price movements between quarterly interest rates on variable rate mortgages from a
five years prior to the base period and five years prior to sample of lending institutions (starting in period Y0:Q1) and
the comparison period. In other words, while preservation presenting them in index number form. The nominal inter-
of original debt/equity ratios and rates of repayment of prin- est rate series can then be combined with the debt series to
cipal help in understanding the approach, estimates of these calculate the final mortgage interest rate charges series, as
variables are not strictly required to calculate the required illustrated in Table 11.3.
comparison period debt. All that is required is the value of 11.122 The construction of equivalent indices for fixed-
the outstanding debt in the base period, the age of that debt interest mortgages is more complicated since an interest
and a suitable measure of changes in dwelling prices. charges index needs to be calculated separately for each
11.119 Supposing that all mortgages are of the variable age cohort of debt to reflect the fact that interest payable
rate type, and that average nominal interest rates rose from today, for example, on a loan four years old, depends on
5 percent in the base period to 7.5 percent in the comparison the interest rate prevailing four years ago. This requires the
period, interest payments in the two periods can be calcu- compilation of a nominal fixed-interest rate index extending
lated as $2,000 and $6,000, respectively, and so the mort- back as far as the dwelling price series. To the extent that the
gage interest payments index for the comparison period is interest rates charged on fixed-interest loans also depend on
300.0. An identical result may be found directly from index the duration of the loan, calculation of the nominal fixed-
number series for debt and nominal interest rates. The mort- interest rate series is also more complex. The additional
gage interest charges index equals the debt index multiplied complexity of these indices may make the construction of
by the nominal interest rate index divided by 100. In this a mortgage interest charges index impractical for countries
example, the debt index equals 200.0 and the nominal inter- where fixed-interest rate mortgages predominate.
est rate index equals 150.0. Therefore, the mortgage interest 11.123 The construction of the index for mortgage inter-
rate index equals (200.0 × 150.0)/100 or 300.0.16 est payments is based on the assumption that the purpose
11.120 While the single-household example shown of the mortgage is to finance the purchase of the dwelling
previously is useful in explaining the basic concepts, it is (hence revaluation of debt by changes in dwelling prices).
necessary to devise a methodology that can be employed to However, it is increasingly common, particularly in devel-
calculate a mortgage interest charges index for the overall oped economies, for households to draw down on the equity
reference population. The main complication when moving they have in their home. That is, households may take new
from the single-household to the many-household case is the or additional mortgages, or withdraw part of the principal
fact that the age of the debt will vary across households. already paid to finance other activities, for example, to pur-
Given the importance of revaluing base period debt to main- chase a large consumer durable such as a car or a boat, to go
tain a constant age, this is no trivial matter. While it is con- on holiday, or to purchase stocks and bonds. If these alterna-
ceivable that information on the age of mortgage debt could tive uses of the funds made available by way of mortgages
be collected in HBSs, the additional respondent burden and are significant, it may be appropriate to regard at least some
the generally small number of households reporting mort- proportion of mortgage interest charges as the cost of a gen-
gages often serve to make estimates from this source unreli- eral financial service rather than a housing cost. For that pro-
able. Another option is to approach a sample of providers of portion of the debt deemed to be used for other purposes, it
would be more appropriate to use a general index of price
16 inflation for debt revaluation purposes.
This simple example illustrates the very important point that percentages
(for example, interest rates or taxes) are not prices and cannot be used as if 11.124 Despite being a relatively simple method of
they were. Percentages must be applied to some monetary value to deter- representing owner-occupied housing services costs, the
mine a monetary price. payments approach is open to conceptual challenge. As
250
Figure 11.1 The Acquisitions Approach for Owner-Occupied Housing
New Dwellings
(A)
(D)
Owner- Other
occupier Existing
Dwelling Dwelling
Stock Stock
(C)
(B)
Self-builds
discussed in the user cost approach, including mortgage • Legal and real estate agency fees payable on property
interest payments could be problematic for monetary policy transfers
purposes. However, another issue is that if both house prices • Repair and maintenance of dwellings
and mortgage interest rates remain unchanged for a period,
• Insurance of dwellings
the mortgage interest charges index does not usually remain
unchanged for the same period. This is because the mort- • Property rates and taxes
gage debt index will continue to change, as changing house
11.127 The construction of price indices for real estate
prices prior to the stable period continue to work their way
agency fees and insurance is discussed in paragraphs
through the various debt age cohorts. Conceptually, this may
11.378–11.384 and 11.385–11.403, respectively. Indices for
prove unsatisfactory to many users, who generally expect
repair and maintenance, and property rates and taxes are not
CPIs to reflect current interest rate movements only.
considered particularly problematic and are not discussed in
The Acquisitions Approach this chapter. Paragraphs 11.141–11.145 address the issues
11.125 The scope for an acquisition index is defined as involved in constructing measures for dwelling purchase,
all consumer goods and services acquired by households for construction, and alterations and additions. The advantages
own consumption. Those countries that compile their CPIs of the acquisitions approach compared to the user cost and
on an acquisitions basis have generally concluded that the payments approaches are discussed in paragraphs 11.88 and
principal purpose of their CPI is to provide a measure of 11.103
price inflation for the whole household sector. Based on the 11.128 As CPIs are constructed to measure price change
view that price inflation is a phenomenon peculiar to the for a group of households in aggregate (that is, the reference
operation of markets, the domain is also normally restricted or target population), like for other second-hand goods a net
to those consumer goods and services acquired in monetary expenditure should be used for those transactions that take
transactions. Consumer goods and services provided at no place between those households. In the case of an index cov-
cost to households by governments and nonprofit institu- ering all private households, the weight should only reflect
tions serving households (NPISHs) are excluded (as dis- net additions to the household sector owner-occupied hous-
cussed in Chapter 2). ing stock. In practice, net additions will mainly comprise
11.126 The expenditure of owner-occupiers of hous- those dwellings purchased from businesses (that is, newly
ing that could be included in an acquisitions index are the constructed dwellings, company houses, or rental dwell-
following:17 ings) and those purchased from or transferred from the gov-
ernment sector plus any purchases, for owner-occupation,
• Net purchases of dwellings (that is, purchases less sales
of rental dwellings from reference population households. If
by the reference population)
the CPI is constructed for some subgroup of the population
• Direct construction of new dwellings (for example, wage and salary earners), the weight should
• Alterations and additions to existing dwellings also include purchases from other household types.
11.129 The acquisition approach can be represented
17
This is not the recommended approach used in national accounts, and
in Figure 11.1. The circle on the left represents the exist-
some of these items are not included in household final consumption ing owner-occupied dwelling stock. The circle on the right
expenditure. represents all other existing dwelling stock (that is, rented
251
dwellings and vacant dwellings). The rectangle on top repre- treatment of owner-occupied housing in the CPI. Housing
sents newly constructed dwellings. The rectangle on the bot- represents both an asset and a source of shelter services.
tom represents self-builds. The flows of properties from one Although households recognize the likelihood of making
sector to another are represented by arrows. There are also capital gains when they purchase housing and invariably
internal flows within the existing owner-occupied dwell- regard their dwelling as an asset, they also commonly cite
ing stock (as one owner-occupier buys an existing dwelling the primary motivation for the purchase of a dwelling as
from another) and within the other existing dwelling stock. being to gain access to a service (that is, shelter and security
11.130 In CPI compilation, a normal procedure is to of tenure). From the households’ perspective, therefore, the
account expenditure that is used in weighting in “net” terms. costs borne by owner-occupiers in respect of their princi-
For most products, the reference population makes pur- pal dwelling represent a mix of investment and consump-
chases, not sales; however, for some product groups, sales tion expenditure, and the total exclusion of these costs from
are significant. This happens with the purchase of dwellings, an acquisitions-based CPI would be counterintuitive and
where it would be necessary to net out the sales from the could lead to criticism by the population at large. This is
purchases. These same principles apply to all second-hand particularly so in those countries where the rental sector is
goods. relatively small, with limited opportunities for substitution
11.131 According to the net acquisition approach, sales between owner-occupation and renting. In these circum-
of dwellings between households have a negative weight stances it might be argued that the consumption element
and purchases have a positive weight. Any sale or purchase dominates.
between households has no impact unless an intermedi- 11.134 The challenge confronting CPI compilers is how
ary, such as a real estate agent is involved. Thus, it is the to separate these two elements (that is, investment and con-
net acquisition cost approach that is followed. Under the sumption expenditure) so as to include only the consumption
net acquisitions approach, only four flows are considered, of housing services in the CPI. Although there is no single
depicted by A, B, C, and D in Figure 11.1. A is the flow agreed-upon technique, one approach is to regard the cost of
of new dwellings to the owner-occupied sector (that is, the the land as representing the investment element and the cost
expenditure owner-occupiers make to acquire new dwell- of the structure as representing the consumption element.
ings). B is the flow of self-builds to the owner-occupied sec- The rationale for this is that, while the structure may deterio-
tor (by definition, self-builds must flow to this sector only). C rate over time and hence be “consumed,” the land generally
is the flow of existing dwelling stock to the owner-occupied remains at constant quality. As the land (or location element)
sector. And D is the counterflow of existing dwellings from accounts for most of the variation in observable prices for
the owner-occupied sector to the nonowner-occupied sec- otherwise identical dwellings sold at the same point in time,
tor. Under the net acquisitions approach, the net expenditure the exclusion of land values may also be seen as an attempt
by owner-occupiers on dwelling acquisitions is given by the to exclude asset price inflation from the CPI. Measures of
following formula (note that this refers to the dwelling pur- asset price inflation are, of course, useful in their own right.
chase component only, and other costs, such as major reno- The conceptual simplicity of this approach appeals to some
vations and transfer costs, are not included): NSOs as a suitable method for compiling owner-occupied
house price indices.
11.135 Derivation of weight reference period expendi-
Acquisition = A + B + C − D (11.2) ture on the net acquisition of dwellings (excluding land), the
construction of new dwellings, and alterations and additions
11.132 In many countries, the flows C and D will be to existing dwellings pose some problems. Although HBSs
similar and will largely cancel each other out. Thus, the may yield reliable estimates of the amount households
owner-occupied housing net acquisition weight will largely spend on alterations and additions and on construction of
be determined by A and B, the purchase of new dwellings dwellings, it is unlikely that they will provide reliable esti-
and self-builds. In exceptional circumstances, for example, mates of net expenditure on existing dwellings exclusive of
after a particularly severe downturn in the property cycle, it the value of the land.
may be that the rate of owner-occupied acquisition of new 11.136 An alternative approach to derive weights for
and existing dwellings becomes very low. If there is a net the expenditure on the net acquisition of dwellings is to
flow of existing dwellings from the owner-occupied sec- combine data from population censuses and housing and
tor to the nonoccupied sector (that is, if D > C), then the building activity surveys. Population censuses normally
owner-occupied housing net acquisition weight may fall to collect information on housing tenure, from which aver-
zero (negative weights are not permissible). Conversely, if age annual growth in the number of owner-occupied
there is a construction boom, A and B may grow very large. households represents a good proxy for net additions to
Property boom and bust cycles will therefore have consider- the housing stock. Building activity surveys are also con-
able effect on the owner-occupied housing weight under the ducted in most countries, providing data on the total value
net acquisitions approach, with implications for the stability of dwellings constructed. These data can be used to esti-
of the CPI. mate the average value of new dwellings, which can then
11.133 In the national accounts, acquisition less dis- be applied to the estimated volumes derived from the pop-
posal of housing is recorded as gross fixed capital forma- ulation census. The suitability of this approach would need
tion, which would suggest purchases of dwellings should to be assessed by each country and may be complicated if
be excluded from household final consumption expenditure. the CPI relates only to some subset of the total population,
While this is unambiguously the case for housing purchased for example, by excluding the very rich who will purchase
for rental, the case is less clear-cut when it comes to the expensive homes.
252
11.137 A price index is required to measure the change information to compile satisfactory dwelling construction
in price over time in existing dwelling structures, newly price indices is not always available. In this situation, house
constructed dwellings, and alterations and additions. As the price indices that include the land component may be a nec-
appropriate price for existing dwelling structures is current essary compromise.
replacement cost, an index measuring changes in prices
of newly constructed dwellings is also appropriate for this Practical Options for Measuring Owner-Occupied
purpose. Given that the prices for both newly constructed Housing under an Acquisitions Approach
dwellings, and alterations and additions are, in principle, 11.141 In some countries, a significant proportion of
determined by costs of building materials, labor costs, and newly constructed dwellings are of the type referred to as
builders’ profits, it may also be satisfactory to construct a “project homes.” These are dwellings that builders con-
single price sample for all elements. The requirement for struct on a regular basis from a suite of standard designs
a separate price sample for alterations and additions will maintained for this purpose. This practice is most feasible
depend on the relative significance of this activity and in countries where a significant proportion of new dwelling
whether the material and labor components differ signifi- construction takes place in new developments (that is, land
cantly from those for a complete dwelling (for example, if recently developed or redeveloped specifically for residen-
alterations and additions are predominantly to kitchens and tial housing). If project home construction is significant in
bathrooms). In all cases, it is important that the price indices scale, then it is possible to select a sample of these project
are adjusted for the mix of these components to eliminate homes for pricing over time, safe in the knowledge that the
price variations that reflect changes in the characteristics of prices provided will be actual transaction prices (in this case
newly constructed dwellings. priced net of any site preparation costs to ensure the fixed-
11.138 The type of dwelling constructed in each coun- basket approach is adhered to). Even if project homes do not
try will significantly influence the complexity and cost of account for most new dwellings constructed, they may still
compiling appropriate price measures. If each newly con- provide a representative measure of overall price change.
structed dwelling is essentially unique (that is, individu- 11.142 In pricing project homes, it is necessary to moni-
ally designed to meet site or other requirements) it will be tor the selected sample to ensure that the selected plans
necessary to adopt “model pricing.” This approach requires remain representative and to detect changes in quality aris-
the selection of a sample of building firms, the identifica- ing from modifications in design and changes to basic inclu-
tion of samples of recently constructed dwellings, and the sions. Whenever a change is made to the plans, the change
collection of prices for constructing identical dwellings in in overall quality needs to be estimated. For physically mea-
subsequent periods (exclusive of site preparation costs that surable characteristics, such as a small increase in the over-
will vary from site to site). This is likely to entail significant all size of the dwelling, it may be assumed that the change in
costs for the respondents. Moreover, care needs to be taken quality is proportional to the change in the relevant quantity.
to ensure that the supplied prices truly reflect all prevailing Other changes, such as the addition of insulation, the inclu-
market conditions. That is, prices need to reflect the amount sion of a free driveway, and so on, will need to be valued,
builders could realistically expect to be able to charge in the preferably with regard to the current value to the consumer.
current market rather than the prices they would like to be These could be estimated by obtaining information on the
able to charge based on conditions prevailing in some prior amounts that consumers would have to pay if the items were
period. provided separately (the option cost method). An alternative
11.139 It should be noted that in many countries, pur- is to ask the builder if a cash rebate is available in lieu of
chasing a new or existing dwelling can have tax implica- the additional features. Where plans are modified to meet
tions. The transaction may be liable for “stamp duty” and changed legal requirements, the consumer has no choice
“taxes on transfers” or a related registration fee. In this in purchase, and it may be acceptable to classify the full
case, the national tax authorities may be able to provide change in price as pure price movement (even though there
comprehensive information on both the number and the may be some discernible change in quality).
value of new and existing dwelling purchases. If detailed 11.143 In some countries, attached and semidetached
dwelling characteristic information is available, including dwellings such as apartments, flats, and townhouses, make
the exact location, the floor area of the dwelling, and the up a significant share of additions to dwelling stock and the
plot size of the land, it may be possible to decompose each price index needs to be mix-adjusted to adhere to the prin-
individual dwelling price into implicit structure and land ciple of a fixed basket. As mentioned previously, measur-
prices, using hedonic methods (described in Chapter 6), to ing owner-occupied housing excluding land and to constant
arrive at aggregate owner-occupied expenditure on new and, quality can be a challenge, and a matched-model approach
where relevant, existing dwellings, that exclude land price. may not be possible because of the heterogeneous nature
If this level of detailed information is not available in the of the buildings being constructed. In these instances, the
transaction data, it may still prove possible to estimate the component cost approach or hedonic models may be viable
aggregate expenditure on dwellings (excluding land) from alternatives depending on the type and amount of data avail-
combining the transaction numbers with standardized build- able. A component costs approach is most commonly used
ing costs (typically building costs per square meter) that in producer price indices and entails the aggregation of a
may be available from construction representative bodies, basket of representative items used to construct the type of
chartered surveyors, building insurers, or other third parties. dwelling being measured. The assumption with this measure
11.140 Whereas having price indices for new dwellings, is that price change is predominately influenced by changes
exclusive of the land cost, is ideal under the net acquisition in the price or cost of goods and services used to construct
approach, in practice this is not always possible. Sufficient the dwelling. Depending on the state of the owner-occupied
253
Table 11.4 Relationship between the Choice of Owner-Occupied Housing (OOH) Approach and CPI Purposes
Approaches to OOH Primary Purpose of CPI OOH Price Definition Notes
User Cost Measures the Change Through Opportunity Cost Associated Approach Consistent with a COLI.
Time of the Total Value of All with the Use of a Dwelling by a Uses Imputed Prices/ Rental
Goods and Services That Are Household for Own Occupancy Equivalence. Can Be Difficult to
Actually Consumed by Households Implement in Practice
Rental Measures the Change Through Implicit Cost That Owner- Approach More Consistent with a COLI.
Equivalence Time of the Total Value of All Occupier Households Would Approach Is Recommended in 2008
Goods and Services That Are Have to Pay to Rent Their SNA. Uses Imputed Prices. Requires a
Actually Consumed by Households Dwelling Suitable Private Rental Index
Payments Measures the Change Through Cash Outlays Associated with More Appropriate for the Evaluation
Time of the Total Payments Owner-Occupied Dwellings of Money Income. Consistent with a
for All Goods and Services by COLI. No Imputed Prices
Households
(Net) Acquisition Measures the Change Through Acquisition and Maintenance Approach More Suitable for Measuring
Time of the Total Expenditure Cost of a Dwelling, Usually on a Monetary Inflation. Consistent with
by Households for Acquiring Net to the Sector Basis Treatment of Other CPI Items. No
and Goods and Services for Imputed Prices. Prone to Volatility in
Consumption the Housing Market
housing market, this may not always be representative of expenditure on rents using standard methods. The HBS
consumer final prices and adjustments in the index compila- should also inquire whether the household rents other hous-
tion method may need to occur, for example, for changes in ing in addition to its main residence, perhaps near to a job or
profit margins. If undertaking a component cost approach, it in a holiday location.
is important to ensure the selected components continue to 11.148 It is important that the HBS determines if any
reflect current building standards, materials, and estimates additional services are included in the rent. This is to ensure
for both builder and developer profit margins. that no double counting takes place and that a consistent
11.144 As mentioned before, there is no single agreed- approach is taken on where the expenditure is accounted for
upon approach that is internationally recommended or in the CPI.
applied for the treatment of owner-occupied housing ser- 11.149 Data should also be collected on the HBS
vices. Rather, there are four approaches that have gained on whether common expenses are charged. Common
recognition, each with its own conceptual basis. While the expenses include payment for common area lighting,
Manual refers to four approaches, in fact, there are three cleaning, building porters, maintaining elevators, and
main approaches as rental equivalence approximates the other services. These payments are paid monthly along
user cost approach and would be considered a variant of this with the rent. In most cases, it is not possible to separate
approach. These four approaches are summarized in Table what portion of the payment goes for lighting or cleaning
11.2. The precise approach to adopt in any given country or to pay a building porter. Because these fees are manda-
depends very much on the primary purpose of the CPI and tory, many countries regard them as part of the rent index.
on practical issues such as data availability. Data are needed to develop separate weights for actual
11.145 In general terms, the most prevalent approaches rent and the common fees.
for the treatment of owner-occupied housing services are
the rental equivalence approach and the net acquisitions Prices—Rent Surveys
approach. The former is relatively straightforward to apply 11.150 The rent to be recorded is the amount that the
where there is a suitable rental market. The latter, although household pays, including taxes and excluding any sub-
more complex, is consistent with the treatment of most other sidies. If the rent is subsidized or taxed, the amount the
goods and services in the CPI and is not directly affected by household pays will not be the same as what the landlord
methods of financing for house purchases. receives. Although information on rents may be collected
from any knowledgeable respondent, the ideal respondent
is the dwelling’s occupants. The goal is to collect an actual
Rented Accommodation rent paid. Collecting rent data from a landlord, or the land-
11.146 Unlike owner-occupiers, renters buy shelter lord’s representative, can be challenging in some countries
services from others who own the dwellings that the rent- because the landlord may not report the actual rent paid by
ers occupy. Consequently, there is a market transaction to the tenant out of fear the data will be used for taxation pur-
observe and the cost of rented accommodation is, in prin- poses. To respect the fixed-basket principle of the CPI, an
ciple, relatively easy to observe in the market. effort should be made to continue collecting prices from the
same dwelling units over time. Despite the perceived diffi-
Weights culties, a longitudinal survey of rental units will often be the
11.147 Obtaining the weight for renter-occupied hous- best solution to ensure that the proper rents are collected.18
ing in the CPI basket is relatively straightforward. HBSs
typically collect data from households at their place of resi-
dence. The surveys obtain the rents from those residences 18
With longitudinal studies, the same units are observed at regular intervals
that are renter-occupied, and the NSOs estimate the annual over long periods of time.
254
11.151 Alternatively, it may be possible to collect rents each month) is a strategy used for increasing the sample
from an existing continuous household survey, such as a without necessarily having to allocate more resources to the
labor force survey. In this case, a specific module on rent can collection exercise. Ideally, a panel survey is used to collect
be distributed to those tenants who participate in the main rent on a monthly basis. Many countries divide the sample
survey. The respondent should be knowledgeable about the into six panels and collect rent from each household twice
rent paid. For example, some occupants may not be the ones each year. Another option may be to divide the sample into
who actually pay the rent and may be unsuitable as respon- 12 representative panels and collect rent prices from each
dents. The important feature of any such survey is that it household one time each year.
provides rents at multiple time intervals but, because people 11.156 Where rent changes are not obtained monthly
may move, it will not necessarily follow the rents of a fixed for the full sample of dwellings, the monthly rent index from
set of housing units. In these circumstances, the rent data months t to t − 1, ∆tRent
−1→ t
, can be derived from a subsample
should be stratified by rent-determining characteristics so of rental units for that month. To obtain the full estimates of
that the average rent increase for a house with preassigned rent change over longer periods, the values from the differ-
characteristics, or for a fixed basket of house types, can be ent monthly subsamples are chained together in succession.
determined. 11.157 To calculate the monthly rent index with a sub-
11.152 If rents change rather infrequently, it may be sample of units, ∆tRent
−1→ t
, the same set of rental units is priced
more efficient to use a relatively large sample that is divided at intervals m months apart. The monthly change obtained at
into representative subsamples to collect the rents less fre- time t is the mth root of the sum of the weighted sample rents
quently than every month from each household. The CPI in period t to the sum for the same dwellings in period t − m:
requires monthly data on rent. Given the nature of rental
contracts and the reality that rent does not change every −1:t
∆tRent =m
∑  w × r 
i i i
t
(11.3)
month, the panel survey approach satisfies the monthly data
needs of the CPI and minimizes respondent burden.
∑  w × r 
i i i
t −m
11.153 Where a survey of rental dwellings is under- where

taken, a sample can be drawn from any frame that contains
−1:t
∆tRent is the change in rents from period t − 1 to period t
the residential units of an area. This may be the population ri is the rent of sample dwelling i in period t
t
census (if it has a shelter component), postal lists, or street rit −m is the rent of sample dwelling i in period t − m (the
directories. last time a rent for that dwelling was collected)
11.154 If the population census can provide information wi is the optional weight of sample rented dwelling i
on the average rent or dwelling value by geographic area,
11.158 Sample weights may be ignored for simplicity
sampling of the areas by probability proportional to rent will
if the sampled rental dwellings represent roughly the same
increase the final sample’s representativeness. Where a sam-
number of rental dwellings in the sampling frame. Equiva-
pling frame is not readily available, a grid can be placed on
lently, the weights in the formula 11.3 can be set to “1” in
a map of the area and a sample of cells in the grid can be
most cases, and this will provide a mechanism to handle
selected. The grid method can also be used as a second stage
nonresponse and sampling anomalies appropriately.
of sampling after selecting larger areas using available infor-
11.159 Geometric, rather than arithmetic, averages can
mation on rents. Several countries, including developing
be taken in formula 11.3 to avoid the defects of the Dutot
countries, have used satellite images of the selected areas to
index that it is not invariant to changes in the units of mea-
obtain an accurate view of the housing in the cells or target
surement of the dwellings or, in the present context, very
areas. Ideally, all the dwellings in the selected cells should
expensive dwellings will get a large implicit weight. Alter-
be enumerated and from this, a random sample should be
natively, expensive dwellings can be treated as outliers and
selected for the price collector to visit to determine if they
be excluded from the calculation.
are tenant-occupied and, if so, to obtain their rents and initi-
ate them into the sample. The initial selection of target areas Accounting for Missing Observations
for enumeration should be small enough to be manageable 11.160 When a dwelling fails to respond because of a
and relatively homogeneous; and, on the other hand, should temporary failure to collect the information (for example,
be large enough to allow for an initial sample selection that the data collector was not able to contact the respondent),
takes into consideration nonresponse and the fact that some but data relating to the other dwellings were collected that
dwellings will be owner-occupied. Where an enumeration month, the collected dwellings should be given the weight
exercise is not possible, for example, because of the cost, of the missing unit. For example, if there are three assigned
judgmental or purposive sampling may be used but, in this dwellings in a cell each with a weight of 1 (because they are
case, it is particularly important that the individual cells on equally weighted) and two are collected and one is not, the
the grid are relatively homogeneous (that is, that there is rent of the missing dwelling can be imputed from the two
little variation in the type, size, and quality of the dwellings) that are collected as follows:
to minimize the potential for drawing an unrepresentative t t
sample. rcollected1 + rcollected2
rˆmissing
t t −m
= rmissing × t −m t −m (11.4)
rcollected1 + rcollected2
Calculating the Rent Index When Data Collection
Is Less Frequent than Monthly 11.161 If all observations in an elementary aggregate
11.155 As previously mentioned, spreading price obser- are missing (for example, because of a data collection prob-
vations by pricing the rent for a particular dwelling less fre- lem), they can be omitted entirely from the calculation for
quently than monthly (that is, pricing a subsample of rents that month; however, their rents must be estimated to be
255
used the next period. If a rental dwelling is a nonresponse in has depreciated and hence is generally not as desirable as it
period t, its imputed rent for period t is as follows: once was. However, an older property does not necessarily
become less desirable especially when it is properly main-
tained (for example, a “heritage property”). In the case where
×(∆tRent )
m
rˆmissing
t t −m
= rmissing −1:t
(11.5)
the money is spent on a property for the repairs needed for
the property to avoid physical deterioration19 as it gets older,
This imputed rent is not used until period t + m. Notice that care should be taken not to double count these ongoing and
it cannot be computed until after period t, when the value for long-term maintenance costs associated with physical dete-
−1:t
∆tRent is known. rioration as they will often be reflected in the overall level of
11.162 If dwellings become unsuitable for inclusion in rents charged. Only those costs of repairs and maintenance
the CPI (for example, they become uninhabitable because of that the tenant is responsible for paying should be included
a fire) they can be dropped from the sample (if unlikely to be in the CPI and these can be obtained through the regular CPI
repaired) or, if there is reason to believe they will be repaired survey of prices as with any other household expenditure.
and returned to use, they can be treated as nonresponse. A study of rental agreements between landlord and tenant
11.163 Dwellings that cease to be rental dwellings (for should provide the information needed to identify whether
example, they become owner-occupied or are converted to the dwelling is depreciating, and which maintenance costs
nonhousing use) should be dropped from the sample, but it should be covered by the price survey.
is good practice to find a replacement dwelling nearby, if 11.168 Empirical estimates of the net physical dete-
feasible. Until the replacement dwelling can be used in the rioration of dwellings generally, including those of owner-
index, the old dwelling should be treated as a nonresponse. occupiers, mostly indicate that it can be small, at least in
11.164 Geometric imputation is more desirable if a Jevons the short term, so that where physical deterioration is not
index is being constructed at the elementary level, as geometric measured either directly or indirectly, ignoring it should not
imputation is consistent with the Jevons formula. The use of geo- be too problematic for users of a CPI who are interested in
metric weighting also avoids the problem previously referred to inflation trends in the short term. The compiler should nev-
of large expensive houses dominating the calculation. ertheless be aware of this phenomenon.
11.169 As mentioned earlier, for owner-occupied dwell-
Updating the Sample ings there is no internationally accepted method of treatment
11.165 The rent sample, like all CPI surveys, needs to of the quality change associated with physical deterioration
be kept up to date. This is especially important to ensure or the cost of major repairs.
that the sample reflects new construction of rental dwell-
ings. An entirely new sample can be drawn, run in parallel
with the old one, after which the old one can be dropped. Second-Hand Goods
The new sample should be based on a new sampling frame. Introduction
Deploying an entirely new sample at one time can be quite
expensive. An alternative is to replace part of the sample 11.170 The 2003 International Labor Organization
each year. If the rent sample uses several panels, the usual Resolution on CPIs states that the expenditure weights for
way to do this is to rotate out one or more panels per year second-hand goods should be based either on the net expen-
and replace them with equally representative panels. For diture of the reference population on such goods or on the
example, the January/July panel can be replaced one year gross expenditure, depending on the purpose of the index.
and the April/October one the next year. The new area cells A CPI is generally understood to be a price index that mea-
would be selected and assigned to panels all at one time, sures the changes in the prices of consumption goods and
but the data collection work would then be limited to one services acquired and should use weights consistent with
panel at a time. Rents for the old and new panels need to be this concept. Chapter 2 reiterates that second-hand goods are
collected at the same time so that the new panel is spliced in in the scope of a CPI and introduce the “net expenditure”
using the overlap method: the CPI uses the old panel for the concept (that is, total purchases less sales) for weights. The
last time while collecting the initial rents for the new panel. latter is in line with the concept of household final consump-
11.166 The index is a chaining of the rent change values: tion expenditure as defined in the 2008 SNA.
11.171 The use of gross expenditure as weights for
changes in the prices of second-hand goods is inconsistent
t
I Rent t −1
= I Rent −1→ t
×∆tRent (11.6) with the SNA.20 The SNA states that sales of second-hand
Depreciation, Major Home Improvements, 19

The 2008 SNA distinguishes between physical deterioration and deprecia-
and Quality Change tion (2008 SNA, paragraph 6.240). Physical deterioration is one of the rea-
11.167 Sample rental dwellings can change between vis- sons for depreciation or consumption of fixed capital (the decline in value of
its from CPI data collectors. Sample dwellings that undergo fixed asset during the accounting period), but the two are not synonymous.
20
The SNA does not directly treat consumer durables as fixed assets or their
dramatic changes, either improvements or deteriorations, acquisition as gross fixed capital formation (2008 SNA, paragraphs 10.32–
are best dropped from the sample, at least temporarily. More 10.37). The one exception is dwellings. However, through a recognition that
subtle changes affect all dwellings: they get older and depre- consumer durables are goods that can be used repeatedly or continuously
ciate. Regular maintenance (for example, replacing the roof) over time, the SNA does indirectly recognize them as durable goods, despite
their purchase being treated as consumption. By implication, durables are
offsets this phenomenon to some extent at least. Some coun- subject to the SNA guidelines relating to expenditure on gross fixed capital
tries make an explicit quality adjustment to observed rents formation. These guidelines explicitly define the latter as acquisitions less
to take account of the fact that a rental dwelling that is older disposals of fixed assets (2008 SNA, paragraphs 10.38–10.42).
256
goods (that is, partly used durables) have to be accounted • Purchases by households from units in other sectors
for and are treated as negative expenditure to be consistent including the rest of the world. The price changes are
with the treatment of the original purchases (2008 SNA, weighted by the values of the purchases or expenditure
paragraph 10.38). Furthermore, the use of gross expendi- incurred.
ture weights and the inclusion of disposals also would be • Sales by households to units in other sectors including the
inconsistent with an index based on acquisitions and could rest of the world. The price changes are given negative
clearly overstate, potentially by a substantial amount, weights equal to the values of the receipts from the sales.
the resources that are devoted by households to acquir-
ing second-hand goods. It is therefore not appropriate for 11.174 One consequence of using gross expenditure for
either a COLI or a non-COLI (for example, a cost of goods weights for second-hand goods will be that the weights car-
index).21 ried by second-hand goods in the index would be greater
11.172 In nearly all cases, net weights should be used in than if net expenditure were to be used, and too large com-
the CPI compilation.22 Regarding the weights, the following pared with other goods and services, as it would overstate
list details four different scenarios for the net expenditure in relative terms the amount of household resources taken
concept: up by transactions in second-hand goods. If the price of a
second-hand good increased, the index would reflect the
• Directly from another household. The net expenditure is
purchasing household being worse off but not the sell-
zero as the transaction is between households. It follows
ing household being better off. Similarly, from a national
that these purchases should be excluded from a CPI.
accounts’ perspective, there is no justification for including
• From another household via a dealer. In this case, dealers acquisitions but arbitrarily excluding disposals. Such a pro-
purchase second-hand goods from households and resell cedure would be illogical and inconsistent with the SNA.
them. Theoretically, these purchases should be included Thus, putting aside some highly specialized uses, in gen-
with a “net” weight reflecting the difference between the eral, there does not appear to be a circumstance in which use
buying and the selling price that is deemed to represent of gross expenditure can be justified in a mainstream price
the service the dealer is giving the buyer. index for household consumption, especially one intended
• Directly from another sector (that is, from an establish- to be used to adjust rates of compensation for changes in the
ment, government, NPISH, or from abroad). The appro- cost of living.
priate net weight is household purchases from these other 11.175 From the previous discussion, guidelines for
sectors less any sales to them. the measurement of second-hand goods in a price index can
• From an enterprise or from abroad via a dealer. Follow- be derived with practical implications that are most pro-
ing the same principles as applied previously, the appro- nounced for durable second-hand goods, such as houses,
priate net weight consists of household purchases from where the treatment is dependent on whether a use, pay-
dealers minus household sales to dealers plus the aggre- ments, or acquisition approach is adopted in the CPI. For
gate value of dealers’ margins on the products that they further details relating to owner-occupied housing services,
buy from and resell to households.23 see paragraphs 11.89–11.149 relating to the alternative treat-
ments of owner-occupied housing. Where the second-hand
11.173 For both a COLI and a cost of goods index the good is semidurable, such as for second-hand clothes and
wording can be simplified and “household” transactions in cars, its treatment is not practically affected by whether the
second-hand goods can be divided into three groups:24 use, payments, or acquisition approach is adopted.
• Transactions between households. The net expenditure is 11.176 Some goods such as cars may be sold by
zero. The changes in the prices of the goods concerned households to dealers who subsequently resell them
carry no weight and have no effect on a CPI. at a higher price to other households. This implies that
households’ net expenditure on goods that are purchased
indirectly from other households via a dealer should be
21
positive. It can be argued that the net expenditure should
A cost of goods index measures the percentage change in expenditure a be treated as purchases of intermediate services provided
household requires to purchase a fixed basket of goods and services. As its
name implies, it seeks to measure the change in cost. In contrast, a COLI by the dealers and not as purchases of second-hand goods.
measures the percentage change in expenditure needed to maintain a house- The weight carried in the CPI is the same whichever inter-
hold’s standard of living. As its name implies, its objective is to measure the pretation is adopted although the estimated price changes
change in the cost of living. might be different.
22
There is one important exception that occurs when the user cost approach
is used for the measurement of owner-occupied housing costs. The loan
11.177 Weights are generally derived from HBSs that
costs associated with house purchase enter the index regardless of whether should cover second-hand goods and be designed to gen-
a newly built house is being purchased or a house from another owner- erate the information required to separate the weights of
occupier. But under an acquisition approach owner-occupied housing second-hand items from new items. Explicit weighting is
would be based on a net acquisition cost basis, that is, the cost of purchasing recommended even when based on approximate estimates
all newly built houses or converted dwellings or existing dwellings newly
acquired by the household sector (for example, private purchases of of expenditure. Administrative records of, for example,
houses previously owned by the government) less disposals of houses (for the value of imported second-hand cars and of changes in
example, demolitions or sales of private houses to government).
23
second-hand car registrations, may provide an alternative
This category covers, for instance, both imports and domestic sales of source of information to compute approximate weights
former company cars.
24
The payment for the services of estate agents and other intermediaries
when adequate information is not available from the HBS.
such as housing agents, auctioneers, and salesroom operators come under Weights can be estimated “top-down” by taking total sales
COICOP 2018 13.9.0.9 (Other Services, not elsewhere classified). and making an approximate distribution between old and
257
new goods or “bottom-up” by computing separate estimates are able to collect as far as possible from the same retailers
and summing them. each month. It is likely that selecting the locations for price
11.178 There are several practical difficulties in pric- collection and the sample of actual outlets selling second-
ing second-hand goods. These exist whether using the tra- hand goods will be judgmental. From a practical perspective
ditional approach of sending price collectors to outlets or there are two main approaches:
pricing second-hand goods online. First, it can be difficult
• Periodically conducting a full enumeration of the relevant
to determine what prices are actually paid, since bargain-
outlets selling second-hand goods at a specific location
ing is more common with second-hand goods than new
and then selecting one at random for price collection each
ones, for instance reflecting concerns about the condition of
month. The difficulty with this approach is that it is not
the item. Second, two sampling problems are particularly
only resource-intensive but also may not be efficient in
challenging:
circumstances where it is unlikely that the item priced in
• The initial selection of a representative sample an outlet at the previous price collection will still be avail-
• Following the prices of a fixed basket of items able the following month as it is likely to have been sold.
By definition, second-hand goods are unique. In addi-
tion, it is not a practical option when outlets are move-
Initial Sample Selection able, which can often be the case for street traders selling
second-hand semidurable goods (for example, unbranded
Items clothing).
11.179 The factors that need to be considered when • A form of quota sampling where the price collector visits
choosing a representative sample of a specific second-hand a preselected location, for instance a market that is known
good include: to sell goods in a midprice and quality range, taking a
• The items chosen must be generally available and easy to “random” walk around the stalls until a second-hand item
find by price collectors. meeting the required description is found. This may be
• The item description must be sufficiently detailed for the the preferred method, especially under the circumstances
item to be easily identified both initially and on repeated described previously and is a departure from the conven-
visits to physical and online or web-based outlets to pro- tional CPI practice where the same outlet is visited each
vide sufficient guidance when a replacement needs to be month and the same good is priced. Outlets and other
chosen. points of sale for specific types of second-hand goods,
such as clothing, can often be identified by their specific
• The item must be representative of the second-hand
location on a market day. Choosing the appropriate day
market.
of the week for the collection of prices from street traders
11.180 For most second-hand goods the selection of is also important because they may change their position
precisely which items to be priced is likely to be purposive. (and prices) on different days.
To facilitate this, the CPI compiler will need to identify
11.183 In practice, a mixture of the two approaches
not only which characteristics physically differentiate the
described previously might be desirable and possible. Under
items, but also which factors have a significant impact on
both approaches, stratification by outlet type, for example,
price, including, most particularly, the age and condition of
shop, market stall, street trader, and online, is recommended
the good concerned. The latter may need to be deduced by
to keep the sample representative.
observation based on a number of price-determining charac-
teristics. For instance, with clothing there may be noticeable
thinning of the material. The attributes may be determined in Following the Prices of a Basket of Items
part indirectly, for instance in the case of a car by reference 11.184 In ideal circumstances, after initial selec-
to its age. These should make up part of the item description tion and pricing in an outlet, the identical item should be
and be used when selecting an item to price each month. priced each month. In practice, this is almost impossible to
11.181 Specifications should be kept up to date to achieve for second-hand goods because, for instance, the
reflect the second-hand market so that problems relating to identical T-shirt is unlikely to be on sale in two consecutive
sample depletion and “forced” replacements are minimized. price collection periods as it is likely to have been sold in
Forced replacements are to be avoided, particularly for cars, the intervening period. In addition, the price collectors are
as explicit forms of quality adjustment are problematic for unlikely to be able to identify and know whether they are
second-hand goods. For example, option costing, which is pricing exactly the same good. A more practical approach,
commonly employed for new cars, cannot be applied to sec- and a departure from normal price collection practice, is to
ond-hand cars, and hedonic regression techniques require a reselect an item each month, in this case a T-shirt, that most
rich data source. In these circumstances, expert judgment resembles the one priced in the previous month. The selec-
is often used to inform quality adjustment for differences tion of the new item is based on the price-determining char-
in technical specifications, mileage, or condition of vehicle. acteristics previously identified. To facilitate this approach,
the price collectors will need a checklist of characteristics
Outlets for each item requiring a price quotation. This can be com-
11.182 The outlet needs to be representative of where pleted during each price collection and a price adjustment
second-hand goods are purchased, whether from shops, made at the head office when there is a deviation from the
fixed markets, street traders, or online. It also must be cho- stated price-determining characteristics. Advice from the
sen considering the need for continuity, so price collectors retailer can be sought to assist this process. An example of
258
a price collection form is given in Annex 11.1 for a man’s likely to supply the retail outlets that are used by the refer-
branded T-shirt. ence population covered by the CPI, taking into account,
for instance, the geographical coverage and the type of
Quality Adjustment retail outlet, including the income bracket of its customers.
The product descriptions of the second-hand goods that are
11.185 An overdetailed checklist of characteristics priced are normally more oriented to the wholesale rather
should be avoided, as it can result in difficulties in locating than retail market to be easier for the wholesaler to com-
products matching the exact definitions because of sellers not plete the form. For example, the wholesaler may be asked to
being able to guarantee that the second-hand goods that they give the indicative prices (that is, the current price they have
acquire for resale will not vary from one period to another. been getting or expects to get) for a “bale” of “average qual-
This may in turn require a large number of quality adjust- ity” T-shirts with designer logos. In this case, “bales” are
ments as replacement items have to be constantly chosen. the standard quantity purchased wholesale that will contain
Where a matching item cannot be located, price collectors some good, average, and low-quality T-shirts mounting to a
should be asked to collect the price of a product that most bale being of “average quality” overall. It is then assumed
closely meets the specifications. The CPI compiler will then that retail prices move in line with these wholesale prices.
need to make a quality adjustment to the price to reflect the An example of the type of questionnaire sent to wholesalers
value of the difference (if any) between the specified item is given in Annex 11.2.
and its replacement. The option most suited to second-hand
goods is expert judgment. This relies on product experts, Prices of New Goods
often in consultation with outlet staff, determining the value 11.189 The use of new goods as a proxy for old should
of a change in specification. This role is often undertaken by be kept regularly under review to identify anything that
price collectors. might challenge the historic price relationship, for instance,
11.186 When there is a planned change of model, where new safety regulations or environmental laws reduce
for instance in pricing second-hand automobiles, the use the value of second-hand cars compared with new ones.
of overlap pricing is often seen as a way of making qual- 11.190 If the price trend of the corresponding new good
ity adjustments, as the prices of both old and replacement is used as a proxy, then the expenditure weight used in com-
models can be collected in the overlap month. Their price piling the index must reflect expenditure on both new and
ratios can be presumed to reflect the market value of rela- second-hand goods. One approach is to include the weight of
tive quality when linking price quotes for different models. the second-hand good with that of the new good. Another is
This removes any need for explicit quality adjustment. But to keep the second-hand item separate with its own weight,
the compiler should beware the possibility of computing a compute an index, and then combine it with the price index
flat index as a result of a high turnover in models rather than for the new good. The latter approach is more transparent.
any real stagnation in the second-hand prices. Chapter 6 pro-
vides more guidance about quality adjustment.
Own-Account Production
Alternative Approaches
11.187 Given the potential for finding challenging prac- Introduction
tical sampling problems associated with collecting prices of 11.191 Households can engage in various kinds of pro-
second-hand goods, some NSOs do not collect prices directly ductive activities that may be either aimed at the market or
but instead adopt one of two possible alternative approaches, intended to produce goods or services for own consump-
depending on circumstances and the goods involved. One tion (2008 SNA, paragraphs 1.41–1.42). When households
approach is to ask a dealer in second-hand goods to com- engage in production for the market, the associated business
plete a questionnaire with the current price that they would transactions are all outside the scope of a CPI. Expenditure
expect to achieve. This is sometimes done for clothing, for incurred for business purposes are excluded, even though
instance, where market traders purchase bundles of second- they involve purchases of goods and services that might
hand clothes from suppliers (normally from other market have been used instead to satisfy the personal needs and
traders who act as wholesalers). It is then assumed that retail wants of members of the household. In practice, households
prices will move in parallel to wholesale prices. Another also produce goods and services directly for their own con-
approach is to assume that the price trends for second-hand sumption and this can account for a significant proportion
goods are the same as those of the corresponding new goods of a country's household final consumption expenditure. For
obtained from the main CPI price collection. The latter is example, households grow vegetables, fruit, and flowers or
most likely to be the practice where purchases of second- other crops for their own use, and owner-occupiers produce
hand goods have been historically much less important than housing services for their own consumption. Goods and
purchases of new goods. Both cases use other price move- services produced by households for their own consump-
ments as a proxy for the price movements of the correspond- tion are called own-account production. According to the
ing second-hand goods. SNA (2008 SNA, paragraphs 6.27), all services produced for
own consumption are excluded from the production bound-
Wholesalers’ Prices ary, except services produced by employing paid domestic
11.188 Under this approach, the prices collected for or household staff and the housing services produced by
second-hand goods are wholesale prices and these are used owner-occupiers. This means that only those services pro-
as a proxy for consumer prices. Wholesalers are selected duced by employing paid domestic staff and the housing
to be representative of the types of wholesalers who are services produced by owner-occupiers would be included
259
in the CPI. Excluded services produced for own consump- owner-occupied housing in the CPI depending on the
tion include, for example, the preparation of meals, the care different conceptual approach adopted (that is, use,
of children, the sick or the elderly, the cleaning and main- payments, or acquisition). However, other durable
tenance of durables and dwellings, or the transportation of goods (for example, stoves, washing machines, or cars
household members. that are commonly referred to as “consumer durables”)
11.192 Many of the goods or services purchased by are treated in both the CPI and the SNA as though
households do not provide utility directly but are used as they are fully consumed at the time they are purchased
inputs into the production of other goods and services that rather than being depreciated over time, so their price
do provide utility: for example, raw foodstuffs, fertilizers, is included in the CPI in the period in which they are
cleaning materials, paints, electricity, coal, oil, or petrol. purchased.
11.193 For purchased final consumption goods and − In practice, in the context of own-account production,
services, the measurement of prices is not a problem as it can be difficult to draw a clear distinction between
the price is determined at the time of purchase. However, intermediate and final consumption, as the same goods
for those goods and services produced and consumed by and services may be used for either purpose. There are
households, there is a problem because no purchases are numerous examples. Basic constituents of foods, such
involved and there are no direct prices to measure or asso- as flour, cooking oils, raw meat, and vegetables, may
ciated expenditure for the construction of weights. For be processed into bread, cakes, or meals, with the assis-
example, conceptually, the purchase of seeds and fertilizer tance of other inputs including fuels, the services pro-
to grow vegetables and fruits that are then consumed is, in vided by consumer durables, such as refrigerators and
principle, intermediate consumption whereas a CPI is based cookers, and the labor of members of the household.
on final consumption. This is sometimes referred to as the Inputs of materials, equipment, and labor are used to
own-account production dilemma. The CPI compiler has clean, maintain, and repair dwellings. Inputs of seeds,
two options: measure output prices indirectly or measure fertilizers, insecticides, equipment, and labor are used
some (but not all) input prices and use them as a proxy for to produce vegetables or flowers, and so on. Some of
the prices of goods and services consumed. The 2008 SNA the production activities taking place within house-
recommends the indirect measurement. Although this seems holds’ activities, for example, gardening or cooking,
a simple and conceptually acceptable solution, exceptions may provide satisfaction in themselves. Others, such
may be made for one or two kinds of household production as cleaning, may be regarded as chores that reduce util-
that are particularly important and whose outputs can read- ity. In any case, the goods or services used as inputs
ily be identified: most particularly, subsistence agriculture into these productive activities do not provide utility in
and housing services produced for own consumption. themselves.
− Utility is derived from consuming the outputs from
Background household production undertaken for own consump-
tion. It is necessary, therefore, to decide whether a CPI
11.194 How goods and services produced for own should try to measure the changes in the prices of the
consumption are treated in the CPI depends on the scope outputs or the prices of the inputs. In principle, it seems
and use of index. Assuming for the purpose of illustration desirable to measure the output prices, but there are
that in principle a CPI should cover final consumption by objections to this procedure.
households (that is, not restricted to monetary transactions).
− On a conceptual level, it is difficult to decide what are
The scope and purpose of the index and implications for the
treatment of own-account production are considered begin- the real final outputs from many of the different house-
ning with paragraph 11.191. hold production activities, such as growing vegetables
11.195 A distinction can be made between: or raising livestock. Even if they could be satisfactorily
identified conceptually, they would have to be mea-
• Intermediate consumption. In the context of household sured and priced. There are no prices to be observed, as
own-account production, this refers to the goods and there are no monetary transactions. Prices would have
services that households use in the process of producing to be imputed for them and such prices would be not
other goods and services. They are not part of the final only hypothetical but inevitably very speculative. Their
consumption of households. use in CPIs is not a realistic possibility in general and
• Final consumption. This refers to goods and services pro- almost certainly would not be acceptable to most users
duced for direct consumption where utility is derived by who are primarily interested in the market prices paid
the household through the act of consumption. by households.
− Goods and services to be used repeatedly in production − The practical alternative is to treat the goods and ser-
over extended periods (more than one year), provide vices acquired by households on the market for use as
benefits to the owner over the lifetime of the goods. inputs into the various kinds of household production
These goods are called fixed assets (2008 SNA, para- activities as if they were themselves final consumption
graph 10.33) and, in general, are recorded as gross fixed goods and services. They provide utility indirectly on
capital formation. For example, a house will normally the assumption that they are used exclusively to pro-
provide shelter for many years. Similarly, a stove may duce goods and services that are directly consumed by
provide cooking services for decades. A house is treated households.
as gross fixed capital formation in the SNA and there − There are two important areas of own-account produc-
are four primary approaches used for the treatment of tion that warrant special consideration: subsistence
260
agriculture and housing services produced for own 11.198 To understand the differences in the indices
consumption. referred in paragraphs 11.195 and 11.196, consider, for
example, that drought affects a country with a high pro-
• In the national accounts, an attempt is made to record the portion of subsistence households. Assume nonsubsistence
value of subsistence agriculture, which is the agricul- households are included within the scope of the compen-
tural output produced for own consumption (2008 SNA, sation index and have no subsistence consumption. These
paragraphs 24.47–24.49). In some countries, subsistence households are also affected by the drought but to a lesser
agriculture may account for a large part of the production extent. Assume that the price of basic food products
and consumption of agricultural produce. The national increases sharply, that there is relatively little change in the
accounts require such outputs to be valued at their mar- prices of imported manufactured goods, and that the prices
ket prices (2008 SNA, paragraph 6.124). A CPI may record of farm inputs fall because of a sharp fall in demand from
either the actual input prices or the imputed output prices, subsistence farmers.25 In this example, the compensation
but not both. If the imputed output prices for subsistence index will rise because basic food products cost more, but
agriculture are included in a CPI, the prices of the pur- the rise will be moderated because the price of imported
chased inputs should be excluded. This could remove from manufactured goods has not increased to the same extent.
the index most of the market transactions made by such The rise in the monetary transactions index will be moder-
households. Expenditure on inputs may constitute the prin- ated further because of the fall in the price of farm inputs.
cipal contact that the households have with the market and The general consumption index will show the greatest
through which they experience the effects of inflation. It increase because the weight given to basic food products
therefore seems preferable to record the actual prices of the will be much higher, reflecting production for consumption,
inputs and not the imputed prices of the outputs in CPIs. and no weight will be given to farm inputs.
• The treatment of owner-occupied housing services is 11.199 Which index in this example is “correct”? If
particularly challenging as described in paragraphs properly constructed, all three are “correct” in the sense
11.102–11.161. that each addresses a user’s need. They are different sim-
ply because they serve different purposes and have differ-
ent uses. The general consumption index would give the
The Scope and Choice of Index best picture of how the drought was affecting the whole
11.196 While the general purpose of a CPI is to measure of the country. The fact that the price of farm inputs had
changes in the prices of consumption goods and services, fallen would not reflect the problem faced by subsistence
the precise scope of a CPI with regard to the goods and farmers who had to find ways of providing food for their
services and households covered should be determined by families and were unable to benefit from the fall in farm
what is intended to be the main use of the index. Consider- input prices. The monetary transactions index would give
ing the household coverage and the item scope of the CPI, the most relevant measure, for the country as a whole, of
subsistence households generally have a weak connection the change in prices of goods exchanged within the market.
to the formal economy, as a major part of their consumption For the purposes of monetary policy, the general consump-
is from their own or bartered production. However, indices tion index would overstate the level of inflation but for the
designed for the indexation of wages or state benefits would purposes of understanding, the impact of the drought on the
exclude subsistence households from their scope. For sim- price of consumption products it would not. The converse is
plicity of the following discussion, the latter indices will be true of the monetary transactions index. The compensation
referred to as “compensation indices.” Indices designed to index would not reflect the full impact of the drought on all
measure price changes covering all monetary transactions households, but it would reflect the impact of the drought
will include purchases by subsistence households but will on indexation households. The general consumption index
not include production for own consumption. These can be would overstate the impact of the drought on indexation
referred to as “monetary transactions indices.” There is a households and the monetary transactions index would
third type of index that includes production for own con- understate the impact; both would be biased if used in the
sumption within its scope. These can be referred to as “gen- context of indexation.
eral consumption indices.” 11.200 Each NSO needs to decide which of the three
11.197 For countries where production for own con- types of index is appropriate in its own national circum-
sumption represents a major part of total household final stances. In countries with well-developed price statistics
consumption, the three types of index will behave differ- systems, the NSO could consider compiling alternative indi-
ently when the prices of basic food, imported manufactured ces to meet different user needs. They would need to educate
goods, and farm inputs diverge. In such circumstances, users about their respective uses. In most countries, the NSO
an index that excludes subsistence households will not be opts for one general-purpose index. If the primary use of the
representative of the whole population as it is designed to index is to use in monetary policy decisions and serve as a
be representative of the price experience of a more limited macroeconomic measure of inflation, it is not appropriate to
group of households. Similarly, the monetary transactions
index and the general consumption index will give divergent
results. Each index provides a different picture of what is 25
Many different scenarios could be considered. This example is designed
happening within the economy. The first does not attempt to to show that all three index types are sensitive to the treatment of subsis-
tence households and subsistence consumption. If there are changes in the
be representative of the whole population and the remaining relative prices of the selected product groups, the choice of index type will
two indices attempt to be representative of all households affect the outcome. Each index will present different results concerning the
but in different ways. impact of the drought on households, particularly subsistence households.
261
include production for own consumption. If the primary use rural prices might provide a reliable estimate, particularly if
of the index is to index wages and government payments, the selection of rural outlets is approximately self-weighted.
production for own consumption should be excluded. In 11.205 Similar pricing considerations apply to the prices
both cases, only monetary transactions should be included used when compiling the results of a household survey used
in the index. It should be noted that the housing services to estimate subsistence consumption. CPI compilers can
produced by owner-occupiers are treated differently (as assist the household survey compilers with these estimates.
described in paragraphs 11.98–11.104). If the CPI compilers are producing a general consumption
11.201 A number of countries compile and dissemi- index, they should ensure that the prices used to value con-
nate a monetary transaction index as the headline measure sumption are consistent with the prices used in construction
of price change but also compile an alternative index that of the price index.
includes production for own consumption in the weights.
Such an index would be for analytical purposes and would Construction of a “Monetary” Transactions Index
meet the needs of poverty economists. 11.206 One of the main challenges in including sub-
sistence households in a monetary transactions index is to
Construction of a “General Consumption” Index construct appropriate expenditure weights. This is because
11.202 To include subsistence households and con- this index uses purchases of inputs used in producing sub-
sumption in a general consumption index it is necessary to sistence outputs as a proxy for subsistence consumption.
construct appropriate consumption weights and to decide Additional price information is only required for input items
how subsistence consumption will be priced. Both aspects that are unique to subsistence households. If items are pur-
are described in paragraphs 11.207–11.209. chased by nonsubsistence households and are already priced
11.203 In general, the available data source for confor CPI purposes, this price information can be used for the
sumption weights will be an HBS or a survey of subsistence subsistence inputs component of the monetary transactions
households. To measure consumption, households are asked index, such as spades, seeds, or fertilizer.
to record the quantities consumed of own-account pro- 11.207 In general, the available data source of weights
duced goods. To obtain a value weight a “price” is applied for inputs into subsistence production will be a household
to the product. Usually, this will be the price of the product survey, such as a general HBS or a survey of subsistence
in a relevant market. Most countries with significant sub- households. When preparing weights for a monetary trans-
sistence production will prepare estimates of the value of actions index these estimates of consumption should be
consumption classified by product. This will be prepared for combined with estimates of purchases of the same products
use in poverty analysis and possibly also for use in com- by nonsubsistence households to obtain the total expenditure
piling the national accounts. When preparing weights for weight for each product. It is important to exclude weights
a general consumption index these estimates of consump- for subsistence consumption that may have been computed
tion should be combined with estimates of purchases of the as part of the household survey. Subsistence consumption is
same products by nonsubsistence households to obtain the excluded because the weight for inputs used in subsistence
total consumption weight for each product. It is important production is included in this index as a proxy for subsis-
to exclude weights for any market purchases of items used tence consumption and to include that consumption would
as inputs into subsistence production, for example, fertilizer, involve double counting.
tools, and herbicides. These products are excluded but only
in respect of households for which subsistence production Construction of a “Compensation” Index
is estimated. For all other households, they are included in 11.208 If the NSO decides to compile a “compensation
the weights and classified within recreational expenditure index” it is unnecessary to construct subsistence consump-
(COICOP 09.3.1).26 tion weights and to decide how subsistence consumption
11.204 The price of basic food products may vary, par- should be priced because an index designed for the index-
ticularly between urban and rural areas. This reflects avail- ation of wages or benefits will generally exclude subsistence
ability, demand, transport costs, and other factors. The most households, or more precisely subsistence consumption,
appropriate price to apply to subsistence consumption would from its scope.
be the price in a market near to where the consumption takes
place. Typically, this would include many rural areas. If a Own-Account Household Services (Excluding
country collects basic foods products prices in both urban Owner-Occupied Housing Services)
and rural areas, the most appropriate price would be an aver- 11.209 Services produced by households include such
age of the rural prices weighted to reflect the distribution things as the preparation of meals, the care of children, the
of subsistence agriculture. In practice, a simple average of sick and the elderly, the cleaning and maintenance of dwell-
ings, and the transport of household members. The actual
26
outputs of the services are excluded from the scope of the
For nonsubsistence households, garden produce may be treated as a rec- CPI and no attempt should be made to impute expenditure
reational output rather than the result of a production process, depending
on the motivation and circumstances of production. Using this treatment, to them. Home-cooked meals are a service output produced
purchases such as tools or fertilizers are gardening or recreational expenses within the household, and it is recommended that there
and not inputs into the production of food. For some households the line should be no imputation of expenditure in respect of the
between subsistence production and recreational activity will be unclear. preparation of these meals. Similarly, it is recommended
NSOs should adopt practical rules based on national circumstances for
making this distinction. For example, urban food production for own con-
that there should be no imputation of expenditure to other
sumption may be treated as recreational and rural food production for own services produced within the household such as when par-
consumption may be treated as subsistence production. ents transport their children to school or care for a sick child.
262
These treatments are consistent with the SNA.27 Although Introduction

it may be conceptually more appropriate to regard the out- 11.212 Tariffs cover a large and diverse range of pric-
puts of household production as consumption rather than ing structures. For example, the EU definition states that a
the inputs, data in respect of the outputs of these areas of tariff is a list of preestablished prices and conditions for the
household production would require many assumptions and purchase and consumption of one and the same good or ser-
imputations and would be of little practical use and so are vice, or of similar goods and services, that has been centrally
excluded from the CPI and the SNA production boundary. fixed by the supplier, by the government, or by agreement
11.210 A long-term bias can result if households to exert influence on the consumption patterns by means of
increase their purchases of services and decrease their pro- appropriately differentiated prices and conditions according
duction for own consumption and this is not addressed when to characteristics of consumers, the level, the structure, or
updating the weights. For example, if households purchase the timing of the consumption. A tariff price can be defined
more takeaway or restaurant food instead of preparing food as a price within a tariff that applies to a component element
for their own consumption, over time the relative expendi- or unit of consumption of the good or service in question.
ture on takeaway and restaurant food will increase and the At its most basic, a tariff consists of a list of prices based
relative expenditure on food ingredients will decrease. If on detailed specifications of the goods or services that are
long-term labor costs increase more rapidly than basic food priced individually but can only be bought as part of the
prices, there will be a long-term downward bias to the index package.
(and conversely if they increase less rapidly) unless the 11.213 Added complications can exist where goods and
impact of this change in behavior is reflected in a revision services providers adopt a range of strategies to differentiate
of the weights for both own-account production associated their goods and services to attract and retain customers. For
purchases and other household purchases. This is because example, to appeal directly to different types of customers,
labor costs have a greater impact on takeaway and restau- suppliers of telecommunication services may bundle ser-
rant food prices than on food ingredient prices. Although vices in different ways or adopt tariff pricing. This is often
the household contributes labor when producing house- accompanied by regular changes in the contracts offered to
hold services, this is not included in the index because it potential customers to encourage the take-up of services.
is outside of the production boundary. Expenditure weights Such contracts are often fixed-term with prices fixed for the
should be reviewed on a regular basis (for more information period of the contract and a penalty clause if the contract is
on weights, see Chapter 3). terminated early. A more detailed discussion of tariff pricing
in the telecommunications sector is included in paragraphs
Tariffs 11.256–11.284. There may exist in the market many tariff
pricing programs. Some of the more common include:
11.211 A tariff is a list of prices for the purchase of a
particular kind of good or service under different terms and • Peak-load pricing (also known as “congestion” pricing).
conditions. For example, one price may be charged for elec- This occurs when producers charge higher fees during
tricity during the day and a lower price may be charged at periods of greater demand, usually because of the higher
night, or a higher price may be charged the more electricity production costs caused by capacity constraints. Peak-load
is consumed. Similarly, a telephone company may charge pricing thus helps in balancing capacity usage over a period
a lower price for a call on weekends than in the rest of the that decreases the need for firms to invest in costly infra-
week. Another example may be bus tickets sold at one price structure expansions. Such pricing practices are often found
to general passengers and at a lower price to children and in the areas of toll roads and bridges, ferry services, electric-
pensioners. Construction of the corresponding price index ity, long-distance phone calls, and home delivery of goods.
should adhere to the core principles that price collection • Two-part tariffs. These occur when consumers are charged
should be undertaken consistently over time and in a way both an entry (or lump-sum) fee and a per unit charge. A
that represents consumer purchasing patterns, and that the fee is charged up front for the right to use (or buy) the
selection of representative items, in this case the different product, and an additional fee is charged for each unit that
tariffs charged, should represent consumer behavior and the household consumes. Examples where two-part tariff
be weighted by consumer expenditure patterns. It follows pricing is often applied are the following:
that it is appropriate to assign weights to the different tar- • Amusement parks that charge an entry fee in addition
iffs or prices to calculate the price index for the elementary to a fee for each ride
aggregate. This section provides advice on the measurement
and inclusion of tariff prices in a CPI and gives illustrative • “Membership” discount programs or shopping clubs
examples of the more common types of tariffs. that require the purchase of a membership card to
access the point of sale and then the consumer pays a
lower price for the products purchased
27
• Landline telephone services where there is a fee to use
The SNA includes all production of goods for own use within the pro- the service (“line rental”) and also a fee per call (the
duction boundary, as the decision whether goods are to be sold or retained
for own use can only be made after they have been produced. However, line rental covers the cost of providing the service and
it excludes all production of services for own final consumption within the “per minute” or metered charge, covers the cost of
households, except for the services produced by employing paid domes- placing the call on the network)
tic staff and the own-account production of housing services by owner-
occupiers. The services are excluded because the decision to consume them • Taxi fares where a variable fare based on distance trav-
within the household is made even before the service is provided (2008 eled is added to the base “fixed” initial charge that is
SNA, paragraph 1.42). not dependent on mileage
263
• Block pricing. This occurs when prices vary according and in accounting for quality differences in the services
to the volume consumed. Electric utilities, for example, provided.
often charge a per kilowatt-hour (kW h) price up to X • Accounting for customers moving from one tariff to
units consumed and then charge a different price for the another under the same service provider might be dif-
remaining units. ficult. For instance, when a mobile telephone company
• Special group discounts. This occurs when members of offers several different pricing packages targeted at dif-
certain segments of the population such as students and ferent patterns of usage, current customers may undertake
seniors may be charged a lower price for the good or ser- periodic reviews during which they may decide to move
vice compared to the general population. Bus fares and to a more advantageous tariff for their level of usage
haircuts are good examples of services where these group while staying with the same provider.
discounts are applied.
11.214 In some countries, private and public firms Price Measurement Methods
have come to rely increasingly on tariff-type pricing as a 11.217 There are four basic approaches to measure tariff
tool for generating additional revenues or for public policy pricing in the CPI: matched samples, unit values, consumer
purposes. For example, publicly administered electric or profiles, and the sample of bills approach. The minimum
water utilities may charge a lower tariff for the first units of information required includes all tariff prices and weights
electricity consumed up to a certain level (the first block), that reflect the structure of the consumption of the product
followed by a higher tariff for subsequent units consumed according to the characteristics of the consumers, the level,
(the second block). It is argued that such an increasing block the structure, or the timing of the consumption.
tariff strategy is often used to promote better and greater
access by lower-income households to what is considered a Matched Samples
vital service. Similarly, other companies charge more for the 11.218 Matched samples are used when a full tariff list
first block than the second as an encouragement to increase or an element of the tariff structure is treated as a product
usage. More generally, private sellers can often increase specification and repriced in subsequent periods. This mir-
their profits by charging a higher price for a product to some rors the traditional CPI methodology of matched pairs. The
consumers and a lower price to others, but this depends on simplest form of traditional matching methodology would be
whether the customer base can be successfully segmented. to take the tariff price lists of some major companies and fol-
11.215 Tariff-type pricing programs can be applied to low the changes in the entire pricing program, preferably with
a wide range of products. The extent of the practice will accurate weighting information for the different elements of
vary according to the country. In cases such as electricity, the price list. This option may be feasible in markets of lim-
the expenditure weight can be relatively high, and so it is ited competition where the pricing structures are quite stable.
important to measure accurately the price change for such There are practical challenges associated with ensuring that
products. The CPI compiler will need to have available data the same product specification is priced each month (that is,
on the tariff prices and weights that reflect the structure of that the matched pairs methodology is adhered to). There is
the consumption of the product according to the characteris- usually limited data about the customer base. In practice, the
tics of the consumers, the level, the structure, or the timing product to be priced using a full price list usually has to be
of the consumption. priced in a rather simple manner without much differentiation
on the basis of customers’ characteristics, and for the sake of
simplicity, some variations in tariff are sometimes ignored.
Main Measurement Issues 11.219 Another version of matching methodology is to
11.216 In some cases, tariff-type pricing programs are select some elements of a tariff as representative items and
affected by price index problems associated with services. reprice them in subsequent periods. For each representative
These include: item, detailed specifications combined with information sup-
• Charges for services can be subject to such complex pric- plied by the provider are constructed. For instance, for air-
ing schedules that it is difficult to select the appropriate fares, this could be for each airline carrier, a nonrefundable and
prices for inclusion in the price index. nonchangeable airline fare from one prespecified location to
another, with predetermined outbound and inbound dates cho-
• Identification or specification of individual products or
sen by time of day and day of the week, including all surcharges.
units of consumption can be difficult, particularly when 11.220 The representative tariff element has the advan-
they are offered as bundled packages consisting of several tage that it can be applied to different service providers, at
types of microexpenditure that may be priced separately least in principle. It becomes possible to incorporate new
but are typically only available as a part of a package. service providers into the index without significant difficul-
• Services are often provided under long-term contrac- ties. However, the choice of only a few items as representa-
tual arrangements and these may include different types tive may limit the validity of the approach. Neither of the
of “customer loyalty” rebates, clauses for the minimum matching approaches is suitable for situations where there
duration of contracts, surcharges for the provision of are major changes in the structures of tariffs.
services not foreseen in the contract, and so on. Loyalty
rebates or coupons are usually ignored because of diffi- Unit Values
culties of measurement (see Chapter 5). 11.221 The unit value approach, in this case, the
• There are often difficulties in accounting for substitution overall average price of one unit, for example, of elec-
between different providers of the same type of service, tricity, for customers of a particular supplier is used for
264
monitoring prices instead of following individual ele- There can be an inherent time lag because of bills being
ments of the tariff. The unit value approach can be used issued after the index for the relevant time period has been
when contents of the tariff-based service are homoge- compiled.
neous (for example, kilowatt-hour of electricity) and the 11.224 In each of the previous approaches, the result-
method should be used only in such cases. The unit value ing calculation should be based on prices and weights that
is calculated using overall revenue and quantity data at reflect the structure of the consumption of the goods or ser-
a higher level that does not distinguish between differ- vices according to the predetermined characteristics of the
ent tariffs or customers. It attributes all the differences consumers and the level, the structure, or the timing of the
between different packages to price alone, assuming that consumption. The corresponding checklist of price-relevant
quality differences between different pricing approaches characteristics for tariff-based expenditure may include:
are zero or insignificant. This may not be the case where,
• The time-pattern of use of the service. The patterns of use
for example, the reliability of continuity of supply var-
may be measured over one day, one week, or even one
ies between suppliers. The advantages of this approach
year.
are that it can be easier to compute; more fully reflects
changes in customer profiles and usage; and better reflects • The volume of use of the service, for example, where
changes in discounts and promotions. pricing programs differentiate between small and large
customers.
Consumer Profiles • The past behavior of the customer. Particularly in insur-
11.222 The prespecified consumer profiles approach ance services where previous claims or medical histories
defines the product independently of a single producer’s may have an impact on the price paid.
or supplier’s tariff structure. Instead, a more general for- • The expected future behavior of a group of customers,
mulation of the consumer’s behavior is constructed based leading to price differentiation between, for example,
on information (for example, sales information provided males and females or for different socioeconomic groups.
by the industry as a whole) that can be used to define a This is particularly prevalent in insurance services where
range of typical consumers irrespective of supplier. Dif- companies analyze claims information by personal
ferent suppliers’ services are priced through these typical characteristics.
consumers. A unit value price is calculated for each con- • Income-dependent prices, most prevalent in public sector
sumer profile and then this unit value is recalculated over tariff structures.
time. The advantage of this approach is that it does not
require an actual sample of bills, only the details of the 11.225 The bundling of different services, where a sup-
different tariffs and some information on usage by typical plier might provide, for example, a mobile telephone plus
customers. Also, the CPI compiler could potentially define free calls at certain off-peak times, or telephone and digi-
consumer profiles by class of household, for example, tal services, needs to be considered in sampling procedures
prices for household with lower and higher income could where such practices are common.
be computed. 11.226 There are numerous combinations of the previ-
ous factors that can be used to create different tariff struc-
Sample of Bills Approach tures and the structure of tariffs can undergo regular change.
11.223 The sample of bills approach is a more refined The CPI compiler will need to thoroughly research the topic
version of the consumer profile approach where a level and monitor the market to ensure that all relevant price dif-
of service activity from an actual sample of customers is ferentials are picked up.
priced each month, rather than defining profiles representa-
tive of the average monthly activities of a range of custom-
ers. This can be done, for example, by selecting a sample
Examples of Price Measurement Methods
of customers from each category of customer, to reflect the 11.227 Examples of the different methods of incor-
structures of the tariffs of the corresponding service pro- porating tariff prices in a CPI are described in paragraphs
viders. For instance, a sample of customers and their bills 11.228–11.235. Quality adjustment is addressed in para-
might be drawn from low-, medium-, and high-volume graphs 11.236–11.247.
consumers of the product. The resulting index measures the
cost of the current billing period’s consumption (normally Matched Samples
over a month or a quarter) at prices charged in the index 11.228 The example given in Table 11.5 is based on
period compared with prices charged in the price reference telecommunications services (telecommunications services
period. The sample of bills approach has a number of differ- are addressed in more detail in paragraphs 11.249–11.275).
ences compared with the consumer profile approach, most 11.229 The example in Table 11.5 considers that there
particularly: is just one service provider for national calls and three for
international calls. It is assumed that in the price reference
• It takes into account seasonal variations in consumption,
period the overall unit price is 1.9200.
for example, a greater volume of international telephone 11.230 Table 11.5 illustrates several issues pertinent to
calls during public holidays and festivals. the different approaches to the measurement of tariff prices
• It reflects actual customer behavior by reference to bills. and the need to review and verify all calculations. For exam-
• It detects price changes not associated with tariff changes, ple, if the overall weighted average price is used to calculate
such as reduced unit charges when a minimum threshold the index, then the average fixed-weight price increase from
of consumption is reached. the price reference period to current period is 11.5 percent
265
Table 11.5 Matched Models: Landline Telephones

Expenditure Quantity Tariff (unit Tariff (unit Change in Unit
Weights Weights price $): price $): Tariff Price
Base Period Period t (percent in
brackets)
National Calls
Local 50.00
Peak 30.00 33.53 1.20 1.30 +0.10 (8.3)
Off-Peak 20.00 33.53 0.80 0.80 0.0 (0.0)
Long-Distance 20.00
Peak 5.00 3.19 2.10 2.50 +0.40 (19.0)
Off-Peak 15.00 22.37 0.90 1.10 +0.20 (22.2)
International Calls 30.00
To Cape Town (service provider A) 10.00 3.84 3.50 3.50 0.0 (0.0)
To Singapore (service provider B) 10.00 3.19 4.20 4.40 +0.20 (4.8)
To Washington (service provider C) 10.00 0.35 3.90 5.10 +1.20 (30.8)
Weighted Average Tariff Price (revenue weights) 100.00 100.00 1.92 2.14 +0.22 (11.5)
Weighted Average Tariff Price (quantity weights) 1.22 1.32 +0.10 (8.3)
if revenue weights are applied and 8.3 percent if quantity strategies. This type of information should generally be
weights are applied. The use of quantity weights is correct, more readily available as it is less sensitive than actual rev-
whereas the use of revenue weights should be avoided. The enue data. It can provide data for the unit value approach or
fixed-weight increase is 12.2 percent if the average of the for detailed weights in the matching methodology. National
price relatives is used (that is, Carli index), which is known regulatory authorities may also be able to provide detailed
for its upward bias. This provides a good example of why customer use profiles on a confidential basis.
a Carli index should not be used in the computation of the 11.233 The use of consumer profiles is quite straight-
CPI. If a geometric mean (that is, Jevons index) is applied, forward in principle. Once the profiles have been identified,
the increase is 11.6 percent. It also illustrates the importance each service plan selected for pricing is priced and repriced
of analyzing tariffs at a detailed level given the fact that dif- based on the (fictional) bills that these virtual customers
ferent elements of the tariff can be subject to very different would receive. Costs to each customer group can be esti-
price changes. Thus, the use of overall unit values, without mated with reference to several plans using sales informa-
some form of stratification, and where weights are contem- tion to determine the relative importance of the different
porary and not fixed could result in volatile average unit plans on offer and the corresponding consumer groups
value/price changes. being targeted. The overall index is then derived in the usual
way by weighting together the average unit costs for these
Unit Values user profiles according to the relative importance of each
11.231 For example, in telecommunications services, consumer profile (representing a category of consumer) in
the unit price for national long-distance calls is derived as expenditure terms.
the total revenue received from such calls, divided by the 11.234 A potential disadvantage with the consumer
number of call minutes. The advantage of the unit value profiles method is that if a new plan that is different and
approach is that, since there is no sampling, the coverage cheaper is introduced in the following year, but the old plan
of the services is complete and there is no need to specify continues to be priced, then the fall in unit value prices will
representative items or different user categories. On the not be reflected in the index. A new plan could offer higher
other hand, the compilation of unit values typically requires or lower average prices and the CPI would miss the price
close collaboration with the service providers. The CPI change going from year one to year two. This problem could
compiler will need to persuade the providers of the impor- be addressed by updating the weights used when introduc-
tance of providing this commercially sensitive information ing the new plan. Also, the consumer profile approach is
on a confidential basis. The NSO will need to assure that the sometimes based on the plan that would be cheapest for the
information will only be used for compiling the CPI, will be consumer. This assumes consumers minimize their costs,
kept secure, and will not be disclosed to a third party. In this have full knowledge of the options, and can move to the
case, stratification continues to be important, for instance by cheaper plan without any barriers such as a contract that ties
whether a call is local or long-distance and made off-peak or the customer into a particular tariff for a minimum length
at peak times. Effective stratification is particularly critical of time.
in the compilation process if the effect of mixing quality and
quantity is to be minimized.
The Sample of Bills Approach
Consumer Profiles 11.235 With this approach, a sample of actual bills is
11.232 For marketing purposes, companies often clas- selected and used to collect prices. Each month, the bill-
sify their customers based on their consumption habits. ing details are used to collect a price. This is in principle
Table 11.6 shows a typical range of representative consumer very similar to the consumer profiles approach but may be
profiles for mobile telephones. Information on different con- more difficult to apply in practice because it is more data-
sumer profiles may be obtained from service providers that intensive and requires access to personal and commercially
use consumer profiling when planning their product pricing sensitive information.
266
Table 11.6 Consumer Profiles: Mobile Telephones

Specification Unit Low Usage Medium Usage High Usage
Customer Customer Customer
Total Usage (per billing period) Excluding Text Minutes 14 24 59

Messages
Calls
Within Same Mobile Network
Peak Number* 5 (35) 10 (35) 15 (40)
Off-Peak Number* 10 (55) 10 (65) 20 (80)
To Different Mobile Network
Peak Number* 0 (0) 5 (20) 15 (25)
Off-Peak Number* 5 (20) 10 (30) 20 (40)
To Landline
Peak Number* 0 (0) 0 (0) 5 (20)
Off-Peak Number* 5 (3) 5 (8) 5 (13)
Other Services
Text Messages Number 5 10 30
Data Mb 500 Mb 2 GB 10 GB
* In brackets, average call length in seconds.
Mb, Megabyte; GB, Gigabyte.
Quality Adjustment close at 9:15 p.m. and they can now avoid waiting for a
11.236 The practical options to account for quality bus, in the cold, until 10:30 p.m. For them, it is a quality
change and new products are different for different tariff improvement. If possible, the CPI compiler should solicit
strategies. the views of users for an indication of whether the change
is considered to be for the better or for the worse. An indi-
Matched Samples cation of subsequent changes in passenger numbers may
11.237 The detection of quality changes depends on confirm whether the initial assessment was correct but
the degree of detail in the product specifications at the low- even this involves some judgment. The decision to use
est level of the index. This is because the matched-model direct comparison, without quality adjustment, is always to
method will not take account of price changes outside the a certain extent judgmental and should therefore be based
specifications, for example, those related to the total vol- on explicit lines of reasoning so that it is transparent and
ume of the service used. In the matched sample approach, can be justified.
the quality change situation is similar to the situation of dis-
appearing products (that is, some prespecified element of Unit Values
the tariff is no longer available for pricing). A replacement 11.239 The unit value approach regards quality differ-
should be selected for the disappearing element of the tar- ences related to different pricing programs as implicit price
iff and the elements compared directly (where applicable) differences. The detection of quality change depends on
or treated as entirely new products. An example follows for the degree of detail in the specification of the unit of con-
public transport bus fares. sumption. Unit values perform best when there is a high
degree of homogeneity among the items within a product
Public Bus Transport in City X category. To prevent quality differences from affecting the
price index, the specifications of the units of consumption
• Old tariff element. Bus fare from city center X to suburb should be as detailed as the available data permit to maxi-
Y, on a Saturday, 10:30 p.m. (night tariff, last connection) mize homogeneity.
• New tariff element. Bus fare from city center X to suburb 11.240 In Table 11.7, relating to bus fares for children
Y, on a Saturday, 9:30 p.m. (night tariff, last connection) the previous tariff categories have to be transformed into the
The basis for the direct comparison is the index compiler’s new ones. For customers aged from 12 to 16, there was a
judgment that from the consumers’ point of view the change price rise of 60 percent (from $5 to $8) while for 16-year-
in timing is not a significant change. olds there was a price decrease of 20 percent (from $10 to
11.238 Concerning the example in paragraph 11.236, $8). The proportion of customers falling in these categories
the CPI compiler could also reason that the quality of the (that is, the previous consumption patterns) should be used
service has changed and an allowance for that should be to calculate the average price change, or in the absence of
made in the index. But in drawing this conclusion and mak-
ing a quality adjustment the CPI compiler should not use
a subjective judgment or personal view. In this example, Table 11.7 Bus Fares: Old and New Tariffs
some travelers may consider this change in tariff and time
Current Ticket Prices ($) New Ticket Prices ($)
as a reduction in quality of service, if last cinema sessions
finish at 9:45 p.m. mean that a bus journey is replaced by Children 5–15 Years = $5 5–11 Years = $5
an expensive taxi journey home, but other users of the bus Teenagers 12–16 Years = $8
Adults 16 and above = $10 17 and above = $10
may welcome the earlier timing because most restaurants
267
such information, the CPI compiler might assume about the average price paid equals the unit price in both periods. This
proportions based on the available information, including assumption should be tested.
revenue information from the bus company. 11.243 The key point is that a unit price for the “volume
of use” is calculated for each class of consumer, that is, in
Consumer Profiles essence, a form of stratification. Implicit in the unit value
11.241 The key issue when dealing with quality changes approach is a specific form of quality adjustment—namely,
in the context of consumer profiles is the choice of which a quantity adjustment.
consumer profiles should be priced through the period of 11.244 Table 11.9 presents a simplified example where
tariff change and change in goods or services provider. An there are just two speeds (low and high) and where only the
example of a profile is based on a situation where broad- download speeds, expressed in megabits per second (Mbs),
band internet service providers increase the download speed change. Treating the two speeds as separate elementary
substantially (that is, an increase in quality), while keeping aggregates, the price per megabits per second both in the
prices constant. It can even be the case that an increase in price reference period and in the next period can be calcu-
download speed can be accompanied by a price reduction. lated. In column G, these unit values (prices) are expressed
In Table 11.8, it is assumed that the differentiating factors with regard to indices with the price reference period equal-
of an internet-provided broadband connection are limited to ing 100. Using standard fixed-based index methodology, an
the download speed and that this is reflected by different tar- expenditure weighted average of each elementary aggre-
iffs. It therefore follows that the increase in speed is a qual- gate index is taken to obtain the overall index for internet
ity enhancement, in the marketplace this feature is clearly a services. The expenditure weights are, respectively, 0.75
price-determining characteristic. As a result of the increased and 0.25 for low- and high-speed services. Although these
speed, page loading times will be faster and consumers will weights are likely not available from the HBS, they could
typically experience improved streaming services such as be obtained from administrative sources such as the regu-
videos and movies. Most times, only under more intensive lators, the service providers, or market-based intelligence.
applications (for example, downloading large files such as The quality-adjusted index so derived in the next period is
movies) will the difference in bandwidth be noticeable for 76.7, which corresponds to a price fall of 23.3 percent. If the
consumers. But this case may not hold in the future with average of the price relatives (the Carli index) was wrongly
improvements to the internet infrastructure. taken, the fall in the index would have been predictably
11.242 The assumption may be made that the product greater at 26.5 percent. If no quality (or quantity) adjust-
versions provided are also, in effect, descriptions of con- ment had been applied, the price index for internet services
sumer profiles, in this case essentially making the consumer in the next period would have been 105.4 that is obtained
profile approach equivalent to the unit value (or unit price) by using only the observed package prices in the calculation
approach. This applies as long as the usage is such that the of the index (in period 1: 55.95 and 74.95, and in period 2:
60.00 and 74.95). By using the unit value as the price when
computing the index, the improvement in the internet ser-
Table 11.8 Internet Services
vice that consumers now have access to is accounted for in
the index, which is the preferred treatment in this case.
Service Current New Speed Difference Current New
Level (slow, Speed in in kbit/s in Speed Price ($) Price ($)
medium, or kbit/s (down/ (down/up) (percent)
New Goods and Service Providers
fast speed) up) (down/up) and New Products
11.245 From the consumer’s point of view, many tariff-
Slow 256/64 384/96 50/50 34.90 34.90 priced goods or services are relatively uniform. Electricity
Medium 768/128 1024/160 33/25 52.25 52.25
Fast 1536/256 2048/320 33/25 86.10 79.95 delivered to the consumer is almost homogeneous, the main
kbit/s, Kilobit per second.
difference being the reliability of supply. The same may
hold true for other goods or services, like telephone services
Table 11.9 Changes in the Tariff for Internet Prices

Share of Download Transfer Price for Package in $ Price per Mbs in $ Index
Expenditure Speed in Allowance
Mbs in GB
(a) (b) (c) (d) (e) (f) (g)
Base Period
Low Speed 0.75 30 150 55.95 (100) 1.87 100.0
High Speed 0.25 120 Unlimited 74.95 (100) 0.62 100.0
Weighted Average Price 1.55 (100) 100.0
Next Period
Low Speed 0.75 45 150 60.00 (107.24) 1.33 71.5
High Speed 0.25 130 Unlimited 74.95 (100.00) 0.58 92.3
Weighted Average Price 1.14 (73.58)
Expenditure Weighted Average of the Two Elementary Aggregates 76.686
Mbs, Megabytes; GB, Gigabyte.
268
or internet services, despite the service provider’s attempts • Rapid changes in the contracts offered to consumers as an
to differentiate their products and make their pricing struc- effective means of encouraging the take-up of the ever-
tures less transparent. Therefore, different suppliers of tariff- increasing range of services
priced goods or services could be considered as providing
11.251 Increasingly, telecommunication companies
the same type of goods or services and treated as substitutes
offer services via plans that require customers to enter into
for each other.
longer-term contractual arrangements with the providers.
11.246 Changes in the market mix of a clearly homoge-
Two broad types of plans are typically offered. The first type
neous product from different tariff structures and suppliers
of plan has no fixed duration and makes allowance for the
should be incorporated in the index. Index construction will
provider to change pricing structures with advance notice to
require information about the market share of different pro-
the consumer. The second type of plan, increasingly more
ducers and the various pricing plans. As a general principle,
popular, provides a fixed-term contract (generally of one to
for homogeneous goods or services the price impact of a
two years) with prices fixed for the duration of the contract,
new good or service provider should be shown in the index,
and with scope for annual updates by an agreed-upon mea-
that is, the target price measure for homogeneous goods or
sure of inflation that should be accounted for in the price
services is the unit price in the overall market for the good
index. The type of contract can influence the approach to
or service, quality adjusted as necessary.
measurement. For instance, experience suggests that a con-
11.247 In some tariff-priced parts of the retail market,
sumer profile approach can work best for computing a price
new service elements are frequently introduced (for example,
index for customers on a prepaid tariff subject to the rel-
multimedia messages or email on mobile telephones). These
evant information being available, otherwise an alternative
should be introduced into the pricing programs or consumer
could be a matched sample. Similarly, where customers are
profiles by chaining, once they have significant market share.
on a fixed-term contract, other approaches may be adopted.
One method selects for pricing the appropriate package from
Classifications each service provider and assumes a rational consumer who
11.248 COICOP, as well as other national classification has perfect market information and uses this to purchase an
systems, is not constructed with a stratification structure that is appropriate package at the lowest price. Each user profile is
sufficiently detailed to reflect the various tariffs for those prod- assumed to switch freely between packages, picking up the
ucts that are subject to such pricing practices. For example, cheapest available each month but constrained by the real-
electricity under the COICOP hierarchy appears as a subclass ity that customers are generally unable to move service pro-
(five-digit COICOP) in its own right (COICOP 2018 04.5.1.0) viders without financial penalty. Thus, assuming an annual
and according to its description includes associated expenditure contract, each month only a twelfth of customers can move
such as hire of meters, reading of meters, or standing charges. service provider. In some cases, moving to a different tariff
A finer breakdown is not given in COICOP. However, when offered by the same service provider prior to the end of the
sellers of a product, such as electricity, use tariff pricing, the contract also is subject to a penalty.28 It is often desirable
accuracy of the CPI can be improved if the price index for this to compute separate indices for the different types of tariffs
product is constructed in a way that reflects as accurately as and combine these using sales weights.
possible the market realities. In other words, a subclass-level 11.252 If NSOs follow traditional sampling approaches
price index should be composed of a number of subindices, for fixed-term contracts and select price schedules according
each one corresponding to its particular tariff price. This may to some price reference period set of plans, and follow them
require the use of a specially designed classification for the pur- until they expire, no price changes will be observed (like-
pose of stratification within an elementary aggregate. wise if plans expire and replacements are linked to show no
change). The marketplace reality, by contrast, is that average
prices for telecommunication services have declined signifi-
Telecommunications cantly in many countries in recent years. Thus, perhaps the
biggest challenge is the quality adjustment when there is a
Introduction change of service provider. For instance, where one service
11.249 Telecommunications services represent a spe- provider is more reliable than another. This is generally
cific case of tariff pricing (detailed information on tariffs is ignored as being unmeasurable.
available in paragraphs 11.211–11.248), but the tariffs tend 11.253 The lack of a harmonized method across NSOs
to be less transparent, more complex, and dynamic with fre- demonstrates the complexities of this sector. It is recog-
quent updates of the tariff structures and prices. nized that current best practice approaches have difficulty
11.250 The global telecommunications sector is sub- in accounting for substitutions across providers and in ade-
ject to rapid change. Technological innovation has led to a quately accounting for changes in the quality of the services
proliferation of new services that has resulted in suppliers provided.
adopting a range of new strategies to differentiate their ser- 11.254 With the telecommunications sector under con-
vices to attract and retain customers. In this context, charac- tinuous change, statistical practices should be constantly
teristics of significance to CPI compilers include: reviewed. NSOs that are considering the construction of
• Fewer linear pricing schedules and the adoption of differ-
ent pricing structures across providers
28
For more information, see Adrian Ball and David Fenwick. 2004. “Cost-
• The increasing tendency to offer contracts that bundle ing Mobile Telephone Calls: The Use of Constrained User Profiles.” Paper
services in different ways to appeal to different types of presented at the Eighth Meeting of the International Working Group on
consumers Price Indices, Ottawa Group, Helsinki, August 23–25.
269
telecommunications indices for the first time, or are reviewing 11.257 For each representative item, detailed speci-
their current practices, are advised to seek out the most recent fications (for example, a telephone call from location A to
research in this field. Notwithstanding, this chapter seeks to location B, at time X, of duration Y minutes) are devel-
provide a general description of four approaches that are cur- oped to sufficiently represent the range of specific services
rently used by NSOs to measure changes in the prices of tele- purchased by consumers within each representative item
communication services. The approaches are the following: or tariff element. This specification is held constant from
period to period, and movements in the indices for represen-
• Matched samples
tative items are computed, based on the movements in the
• Unit values prices of this matched sample. Table 11.10 illustrates this
• Customer profiles approach. The detailed application will depend on the tariff
• Sample of bills structure. For instance, in some countries, there will be a
regional dimension.
11.255 The choice of approach will depend largely on 11.258 The list of representative items (that is, the low-
the market conditions prevailing in individual countries, the est level in the structure) should be sufficient to be repre-
sophistication of the index compilation system in use, and sentative of price behavior as a whole, taking account of
the extent of access to accurate and timely telecommunica- the published tariffs. As with other parts of the CPI basket,
tion services data. Depending on these factors, it may be expenditure on those services not selected for pricing should
appropriate to use different approaches for different tele- be distributed over the other services within that general
communication services, or even for the different services class for deriving weights. For example, the expenditure
of specific providers. on any fixed- or landline services not selected for pricing
should be distributed over those fixed-line services selected.
Matched Samples—Representative Items 11.259 Compared to suppliers of goods, service provid-
11.256 Matched samples are used when a full tariff list ers have a great capacity to tailor both the services and the
or an element of the tariff structure is treated as a product prices they charge, for example, based on the time at which
specification and repriced in subsequent periods. The use the service is provided. A five-minute telephone call at 9
of matched samples mirrors traditional techniques adopted a.m. can be regarded as a different product from an equiva-
elsewhere in the CPI. The total expenditure of reference lent call made at 9 p.m., and service providers are able to
group households on telecommunication services in the charge different prices for these calls. Representative items
weight reference period is derived from sources such as therefore need to be described in enough detail to capture all
HBSs. A sample of service providers is contacted to obtain the price-determining characteristics.
information on revenue by types of services (for example, 11.260 Given the ease with which providers can adjust
line rental, local calls, international calls, handset sales or the differential aspects of their pricing schedules (for
rentals, connection fees, voicemail services, or internet example, the time span designated as peak and the duration
charges) and a number of these are selected as representative of a call before a different rate applies), it is necessary to
items of the different tariff elements, or all are taken, with use a sufficient number of varied specifications to capture
weights derived from the revenue data. these aspects reliably. It is not enough to simply describe a
call as peak or off-peak, or from zone 1 to zone 2 without
defining the call in more detail. Examples of the types of
Table 11.10 An Illustrative Index Structure for specifications that may be applicable for two representative
Telecommunication Services (representative item
approach)
items—international calls (fixed line) and usage fees (inter-
net services)—are provided in Table 11.11.
Fixed-Line Services
Telephone Connection Costs
Telephone Line Rental
Local Calls Table 11.11 Examples of Specifications of
Peak Telecommunication Services
Off-Peak
Long-Distance National Calls Representative Examples of Specifications
Peak Item
Off-Peak International Peak: Plan A: Call to Athens at 8 a.m. on a
International Calls Calls (fixed line) Friday, Duration 10 Minutes
Peak Off-peak: Plan B: Call to London at 9 p.m. on a
Off-Peak Saturday, Duration 5 Minutes
Peak: Plan A: Call to New York at 11 a.m. on a
Mobile Telephones
Wednesday, Duration 20 Minutes
Connection Costs
Off-Peak: Plan B: Call to Paris at 7 p.m. on a
Handset Purchase or Rental
Sunday, Duration 15 Minutes
National Calls
Usage Fees Plan A: Broadband Connection, Unlimited
International Calls
(internet) Downloads, Speed 38 Mb
Payphones Plan B: Fiber Broadband, Unlimited Downloads,
Local Calls Speed 300 Mb
Internet Services Plan C: Mobile Broadband, 24-Month Contract,
Connection Fees 32 GB per Month
Usage Fees Note: Peak and off-peak are defined times. For example, peak may be between
the hours of 9 a.m. and 6 p.m. on weekdays and off-peak any other time.
Note: Peak and off-peak are defined times. For example, peak may be be-
tween 9 a.m. and 6 p.m. on weekdays and off-peak any other time. GB, Gigabyte.
270
11.261 In Table 11.11, it is assumed that the origin of inherent in some other methodologies, the compilation does
both the telephone calls and internet access is also identified. rest on analysis of aggregate company data and so is likely
All minutes consumed are domestic. It should also be noted to be less timely than methodologies based on prepublished
that the nature of internet access generally precludes pricing prices. However, care needs to be exercised to ensure that
on the basis of access, and hence the timing of access can- the measure is not affected by undesirable compositional
not be as tightly defined as for international telephone calls: changes. A unit value index should only be constructed for
instead, all specifications are for total monthly usage. truly homogeneous items. This points to a requirement for
11.262 The most challenging aspect of the matched defining the representative items at a relatively fine level of
samples approach is obtaining the data needed to establish disaggregation. For example, international calls may need
the representative items and to identify suitable specifica- to be further subdivided by destination to avoid changes
tions, as this will require detailed information from ser- in unit values arising purely from shifts in the numbers of
vice providers. Once implemented, most price information calls made to different destinations. Thus, unit values need
should be readily available from published fee schedules, to be calculated at a sufficient level of detail to minimize
minimizing the burden on respondents between reviews of the effect of changes in customer usage being reflected in
the specifications. However, much reliance will be placed the index.
on the service provider to supply the corresponding expen- 11.266 Although the unit value approach appears
diture information. to address at least some of the known deficiencies of the
11.263 The dynamic nature of the telecommunica- matched sample approach, it is likely to exhibit period-to-
tion sector and the common use of the pricing mechanism period volatility because of compositional shifts, for exam-
to change consumer behavior are likely to require that the ple, because of seasonal variations in usage patterns. There
specifications be updated frequently. A major practical are also several respondent and data quality aspects that need
challenge associated with the matched samples approach to be considered. The unit value approach imposes a greater
is ensuring that exactly the same product specification is data burden on service providers, who often regard revenue
priced each month and that the matched samples method- and quantity data as highly sensitive. To be effective, the
ology is adhered to. When a variety disappears (that is, a service providers also need to be able to furnish data relating
specific plan is no longer offered), all efforts must be made only to households (that is, they should be able to separate
to find a suitable comparison variety. Where varieties are out revenue and quantities relating to businesses) and the
replaced, it is possible to argue that, because different plans revenue information needs to conform to the requirements
involve different conditions of sale, they are fundamen- of the index. For example, some service providers may
tally different products. It is equally reasonable to question record certain discounts as a marketing expense, rather than
whether the entire price difference between plans is because a reduction in revenue as is required for the unit value index.
of quality differences, particularly in light of the evidence of
ever-increasing volumes and reductions in unit values. The
difficulty lies in quantifying the quality differences. Customer Profiles
11.267 For marketing purposes, telecommunication
companies often classify their customers according to their
Unit Values—Representative Items volume of service use. Although the number of categories
11.264 The unit value approach is used when the overall can vary, a common approach is to use a three-way clas-
price or, more precisely, the revenue generated by one unit sification: low-volume, medium-volume, and high-volume
(for example, a long-distance telephone call for customers customers. Service providers analyze customer usage pat-
of a specific supplier) is used to monitor prices. Unlike the terns by category when developing new plans targeted spe-
matched samples approach, individual specifications are not cifically at each group. National regulatory authorities may
priced. The unit value approach can be applied instead of the also be able to provide detailed customer use profiles on a
matched samples approach when the quality of the service confidential basis.
is homogeneous between providers with the advantage that 11.268 NSOs can take a similar approach to the tele-
it is less resource-intensive. The indicative price for each communications companies for the construction of price
representative item is calculated from revenue and quantity indices by devising profiles that reflect the average usage
data collected from the service provider. For example, the patterns for each category of consumer. Costs faced by these
price for national long-distance calls can be derived as the average consumers in each period can then be estimated by
total revenue received from such calls divided by the num- reference to the rates set out in the plan that is currently most
ber of call minutes. Similarly, in the case of monthly line commonly applicable to each customer category. Variations
rental fees, the price can be calculated as the total revenue on this general theme include estimation of costs based on
from line rental divided by the total number of subscribers. the plan that would deliver the cheapest overall cost, indi-
11.265 Compared to the matched samples approach, vidually, to each of the representative consumer profiles
the unit value approach attributes all the difference between (based on the simplifying assumption of cost-minimizing
plans, and time and duration of calls to price (that is, the consumer behavior with perfect knowledge). This has the
quality difference is assumed to be zero). The unit value advantage of providing a clear basis for choosing a com-
approach is also seen as providing a method for accounting parable replacement if an existing package ceases to be
for price change when the items are subject to a prolifera- available. Alternatively, costs to each customer group may
tion of discount programs or promotions (for example, $2 be estimated with reference to several plans, where sales
to call anywhere for any duration for the next week). While information indicates that this is a closer approximation to
the approach avoids some of the customer sampling choices reality. The overall index is derived by weighting together
271
Table 11.12 Example of a User Profile for Mobile rather than defining profiles representative of the average
Phone Services monthly activities of customers. A sample of customers
Specification Unit Low Usage Medium High Usage should be selected from each category of customer (that is,
Customer Usage Customer low-, medium-, and high-volume customers) and, ideally, the
Customer bills (or activity statements) should cover a full year’s activity.
Total Usage (per Minutes 14 24 59 11.273 The perceived advantages of this approach com-
billing period) pared to the customer profile approach are the following:
Excluding Text
Messages • It is able to take account of any within-year variations
in customer behavior (for example, a higher incidence of
Calls
international calls associated with religious or cultural
Within Same Mobile Network
Peak Number* 5 (35) 10 (35) 15 (40)
events of significance). But it can be argued that this con-
Off-Peak Number* 10 (55) 10 (65) 20 (80) travenes the principle of the fixed basket.
To Different Mobile Network • It better reflects the diversity of consumer behavior by
Peak Number* 0 (0) 5 (20) 15 (25) identifying actual activities (that is, calls actually made
Off-Peak Number* 5 (20) 10 (30) 20 (40)
by a sample of consumers) and potentially captures out
To Landline
Peak Number* 0 (0) 0 (0) 5 (20)
of contract activity charges (for example, exceeding
Off-Peak Number* 5 (3) 5 (8) 5 (13) allowances).
Other Services • It accommodates within each bill any instances of annual
Text Messages Number* 5 (20) 10 (25) 30 (35) charges.
*In brackets, average call length in seconds. • It allows for the detection and recording of other sources of
price change associated with customers’ overall relation-
ship with the service provider (for example, where overall
the results from these user profiles according to information discounts are provided when aggregate monthly spending
about the relative importance of each category of consumer. exceeds certain values, or where an aggregate discount is
11.269 In constructing the aggregate index, these cal- provided if customers acquire bundles of services from a
culations are likely to be made for a representative sample single provider, such as landline phone plus internet).
of service providers, using the information on their overall 11.274 Calculation of the index still requires monthly
market share for sampling or weighting purposes, if avail- information on the relative significance of various plans by
able. Information on the distribution of customer profiles by customer category. With the sample of bills repriced each
service provider may not be available or at least may be very period, the resulting index measures the cost of a full year’s
costly to obtain. Table 11.12 gives an example of a profile consumption at the prices prevailing in each index period
for mobile telephone calls. This approach can be extended to compared to the same cost at base prices. This assumes that
include internet usage. the quality difference between old and new plans is zero for
11.270 Consistent with the fixed-basket approach, the households changing plans. Because of the generally larger
activity of consumers (with regard to numbers and types of number of bills (compared with the number of available pro-
calls, numbers of texts, and amount of internet data used) files), price changes can be reflected more gradually, as the
is held constant between comparison periods. Prices may proportion of bills priced using each plan can better mirror
change when not fixed by contract or when plans are replaced. the changing population distribution.
CPI compilers may also allow tariffs to change in response 11.275 As with the consumer profiles approach, it is
to a changing mix of plans within customer categories. This important that the sample of bills is updated regularly to
approach assumes that tariff changes fundamentally represent reflect changes in consumption patterns and the take-up of
price change rather than quality change, but eliminates the new services such as call-waiting, voicemail, and text mes-
cruder compositional effects associated with the unit value saging. Although, with adequate sampling, the sample of bill
approach that does not take account of customer profiles. approach is likely to provide a better measure of the aggre-
11.271 The success in the use of the unit value approach gate rate of price change for telecommunication services as a
is determined by the degree to which the profiles truly reflect whole, it may not be best suited to the calculation of separate
consumer behavior and therefore a great effort needs to be indices for the components of those services (depending on
put into their development. The construction of the cus- whether overall or bottom-line discounts are offered). The
tomer profiles will require a high degree of cooperation approach is also data-intensive, requiring many calculations
from service providers or regulators and, given the known each period and thus a relatively advanced data processing
volume changes, they will require updating at regular inter- system capable of handling and manipulating large amounts
vals, possibly more frequently than other items in the CPI of data. It can also suffer from a lack of timeliness.
basket. Data on plan usage by customer category for each
index compilation period (a month or quarter) may also be Transport Services
required if compilers decide to allow for such effects.
Introduction
Sample of Bills 11.276 Transport services cover a wide range of modes
11.272 This method is a more refined application of the of transport from airplanes, trains, buses and ferries, taxis
customer profile approach. A fixed level of service activity and motorbike taxis, to private cars. The example given in
from an actual sample of customers is priced each month paragraph 11.282 focuses on air travel.
272
11.277 Transport, as defined by COICOP 2018 Division and class of travel should remain the same throughout the
07 covers the purchase of vehicles (COICOP 2018 7.1), the year. While some flexibility should be allowed to accom-
operation of personal transport equipment (COICOP 2018 modate changing schedules for specific routes, if specifica-
7.2), passenger transport services (COICOP 2018 7.3), and tions diverge too much from the price reference period, the
transport services of goods (COICOP 2018 7.4). replacement ticket should be considered noncomparable and
11.278 The current chapter does not cover the purchase a new base price imputed based on the movement of prices
of vehicles or the operation of personal transport equipment for routes in the same category. This is the standard form of
(COICOP 2018 7.1 and 7.2), rather it relates to transport implicit quality adjustment.
services provided by and purchased from third parties as 11.286 For sample selection of routes and providers
defined by COICOP 2018 7.3 and 7.4. The latter includes and for weighting between routes and travel subcatego-
bus and train fares, taxi fares, and the purchase of air and ries (for example, long-haul or domestic), expenditure data
ferry tickets. from appropriate surveys or administrative sources should
11.279 There are common elements between this sec- be used. Where this is unavailable, passenger numbers and
tion and the one on tariffs. Most particularly, the section on average prices can be used to calculate expenditure data.
tariffs uses examples of tariffs relating to public buses to Note that the coverage of routes depends on whether the
illustrate the application of the matched samples and unit CPI covers the national or domestic concept (as defined
value approaches. in Chapter 2) and, if purchased online, the location of the
11.280 Some public transport services may be paid for transaction and service delivery. For instance, if a ticket is
in full or in part by government or NPISH and are provided purchased online from a foreign carrier it can be argued
free or at a nominal price to households, and are considered that the purchase should be attributed to the country where
social transfers in kind in the SNA. The standard convention the consumption first commences. This is because the con-
is that the price entering the CPI should be the subsidized sumer needs to be present at the location of the service
price. provider for consumption to commence, and the place of
consumption normally determines where the sales tax, such
as VAT, is actually paid. If this reasoning is followed, then
Public Transportation such transactions for the booking of flights totally outside
11.281 Elementary aggregates should distinguish the country of residence will be excluded from a CPI fol-
between different modes of transport as reflected in the lowing the domestic concept. Paragraphs 11.65‒11.71 dis-
structure of COICOP (purchases of transport services are cuss the treatment of internet purchases and the domestic
generally classified by mode of transport in COICOP). concept.
The computation of prices should reflect the different tick- 11.287 Providers of transport services often practice
ets available for purchase from different channels. In the some form of price discrimination so that different groups
case of airlines, different prices may apply to online book- of travelers will be charged with different fares or prices
ings compared with bookings made through an agent, and for essentially the same service. Examples include spe-
to advanced nonrefundable purchases compared with, for cial discount fares for students or retired people, or trips
example, a fully flexible fare purchased close to or on the booked for weekend departure or return, and a 30-day
day of travel. advance booking. Since these can change frequently, CPIs
11.282 Let P be the total cost of providing a trip from A usually ignore minor changes in the requirements for a dis-
to B and let Q be the number of trips carried by the transport count fare.
system from A to B during the period under consideration. 11.288 The recommended approach to differential and
Suppose the household sector pays only sP of this price and complex fare structures that is most likely to ensure like-
the quantity is Q. Obtaining the prices for inclusion in the for-like comparisons is to price the cost of a journey on a
elementary aggregate normally requires a method of sam- specific day of the month (for example, the fourth Tuesday)
pling that takes account of complex fare structures. between two points, pricing a ticket for transportation pur-
11.283 The index for airline and other public trans- chased at a fixed time in advance, and with fixed terms and
port services should use the prices of a sample of specific conditions. This should be done for various fare classes,
trips rather than revenue per kilometer or per passenger- for example, a full economy fare (if purchased in signifi-
kilometer. If the CPI has strata for different geographic cant quantities for personal travel) and a typical discounted
locations that are typical in large countries, points of origin economy fare that may include travel restrictions such as
(for example, airports, train stations, or motor coach stops) for a stated date and time and not being refundable. Prices
should be chosen in each location, and trips selected within can be downloaded by staff at the head office from reserva-
origins or destinations in those locations. There are gener- tion systems available on the internet, although sometimes
ally multiple classes of service. In addition, transport fares the full cost is not shown without going through the proce-
may vary by day of the week, time of day, time of year in dures for purchase.
response to variations in demand, and how long in advance 11.289 When selecting the sample, the NSO should also
the ticket is purchased. The selected trips should reflect this consider how far in advance of the date of travel a ticket
variety but hold these variables constant. price is collected. In airfares, for example, ticket prices can
11.284 For airfares, the sample should be stratified by increase the closer the booking is made to the departure
domestic, short-haul, and long-haul flights as scheduling and time, because the number of available seats for a specific
service providers between these categories can vary widely. route becomes limited. However, in other cases, there can
11.285 Once chosen in the price reference period, the be last-minute discounts to fill empty seats. Further strati-
route, the departure and arrival times, and the ticket type fication of the sample should be considered to account for
273
this, for example, by collecting prices for the same depar- of household final consumption expenditure. In many coun-
ture time and date but collecting separate prices for book- tries, various government units or NPISH29 will finance and
ing the ticket so many months, weeks, or days in advance, pay for the full or partial provision of a significant propor-
depending on the nature of transport service being priced. tion of goods and services in the health, education, and
Taking airfares as an example, once the routes, ticket types, social protection sectors.
and departure date have been selected, price collection for 11.294 The preferred approach regarding the treatment
this sample could be repeated six, three, and one month in of health, education, and social protection services in the
advance to capture the changing price as the departure date CPI is as follows:
approaches.
• Only expenditure by households that is a direct result
11.290 Changes in transport schedules can cause price
of purchase of individual goods or services is within
changes. For example, an airline might cancel its noon flight,
the scope of CPI. These prices should be net of direct
forcing passengers to take the evening flight that is priced
reimbursements. Reimbursements refer to payments to
higher because that is a time of peak demand, unless a com-
households by government units, social security adminis-
petitor airline can be found that offers a noon flight. If the
trations, or NPISHs that are made as direct consequences
CPI was following the price of the noon flight, it must now
of purchases of individually specified goods and services,
price the evening flight if the customer has no choice but to
initially paid for by households.
use it, or price a noon flight available from an alternative
comparable airline if available. The replacement journey • Mandatory payments are excluded from the scope of the
should be treated as comparable and the full price change index (for example, employee’s contributions to social
should be reflected in the index if it is assumed that there has security programs). These are collected to finance social
been no change in the quality of the service being provided. security programs and these payments do not directly
But the latter might be a strong assumption, for instance, if relate to the provision of goods or services. Reimburse-
an evening flight is inconvenient because it means traveling ments that do not directly relate to goods or services are
the day before and staying in a hotel to attend a morning also not in the scope of the CPI.
appointment the following day. In the latter case, the noon 11.295 It should be noted that, in compiling a CPI, a
flight might best be treated as a missing item or the evening number of situations may be encountered. For example,
flight selected as a noncomparable replacement and a new with childcare systems where the government partially sub-
base price calculated for it. sidizes the full cost of providing the service, it is the net (or
11.291 Data challenges can be encountered when purchaser) price paid by the household that is covered by
selecting representative items, estimating weights, and pric- the CPI and should be included under the proper elementary
ing selected flights especially with the emergence of online aggregate (or elementary product group) entry in the clas-
booking. In the case of air transport, the main data sources sification system. This is regardless of whether the childcare
alternative to HBS for both sampling and weights estimation service is privately or publicly run. But if these services are
are airports, the airlines, national civil aviation authorities provided by the households themselves, they are regarded
that often collect detailed data covering the entire country, as own production and not included in the SNA produc-
or market research companies and trade organizations that tion boundary, as no monetary transaction is involved even
collect data. The data most likely to be available for the though the government might give the household money in
sampling of flights will generally relate to the total number the form of a social security payment. This approach is con-
of passenger flights taken and is likely to include business sistent with the conceptual basis of the national accounts.
travelers. Nonetheless, such data will be better than using a 11.296 Goods and services acquired by households and
subjective approach and assumptions can be made, such as that the government or NPISH provide a social benefit either
assuming that all business and first-class bookings are for through a full or partial reimbursement should be measured
business purposes. Normally, expenditure data on last-min- net of direct reimbursements. For example, individuals who
ute deals can be gathered either through the HBS or, more are part of a particular socioeconomic group may be eligible
likely, through an ad hoc survey of travel service providers, for a full or partial refund for dental care: if the refund cov-
if not available from other sources. ers the full cost of the dental service, then the expenditure
11.292 Transport ticket prices should feed into the index weight would be zero and no prices would need to be moni-
at the time of travel, not at the time they were booked. For tored for the purpose of the CPI.
example, a ticket price for a December flight should feed 11.297 Where individual goods and services are pro-
into the December index even if the price relates to a pur- vided at no charge to individual households by governments
chase in the previous October. or NPISH, these transfers are regarded as social transfers
in kind and costs are included in government or NPISH
Health, Education, and Social
Protection Services 29
NPISH make up an institutional sector in the context of national
accounts consisting of nonprofit institutions which are not mainly financed
Introduction and controlled by government and which provide goods or services
11.293 Health, education, and social protection cover a to households for free or at prices that are not economically significant
wide range of services. Transactions that involve an expen- (2008 SNA, paragraph 4.93). NPISH are private, nonmarket producers
which are separate legal entities. Their main resources, apart from those
diture by a household in the health, education, and social derived from occasional sales, are derived from voluntary contributions in
protection services fields are in scope and should be covered cash or in kind from households in their capacity as consumers, from pay-
by the CPI if it is to be consistent with the SNA concept ments made by general governments, and from property income.
274
expenditure.30 Examples include food stamps, social benefits Health Services (Doctors and Dentists)
to households prescription drugs, and job training programs. 11.303 The treatment of health services in the com-
These social transfers in kind can contribute substantially pilation of individual countries’ CPIs will depend on the
to the standard of living of the individual households that institutional arrangements for providing them. Government-
receive them. provided free medical care is out of scope, but many coun-
11.298 Depending on national circumstances, it could tries adhere to some form of two-tier health care system
be that the public authorities have decided to partly or fully whereby some health care services are provided, often for a
finance the provision of certain goods and services. Regard- fee, by a private system that coexists with the public system
less of the practice, those expenditures incurred by govern- or where medical services are financed partially by the gov-
ment or NPISH to finance social transfers in kind are outside ernment but the consumer pays part. In both cases, fees paid
the scope of a CPI (although it can be argued that it is desir- by the users are in scope. Employer-provided medical care
able to take them into account when estimating a compre- is beyond the scope of the CPI, because a monetary trans-
hensive COLI extending beyond purchases by the household action does not take place and it is treated in the national
sector). However, when the consumer pays part of the cost accounts as remuneration in kind.
associated with the provision of such goods and services, 11.304 Although health is a broad term, for the con-
this element is within the scope of the CPI: although there struction of an elementary aggregate in a CPI, it generally
is a social transfer, this expenditure may account for a large refers to medical care provided by medical professionals,
part of a household’s final consumption expenditure. For para-professionals, or medical institutions. COICOP divides
example, take a rental unit for which the monthly cost of the health care sector into classes according to the type
providing the rental service is $800 and where, after the of medical care provider (for example, doctors or hospi-
government transfer, the actual rent paid by the tenant is tals). This facilitates sampling, price collection, and index
$500. While the $300 transfer is excluded from the scope of construction.
the CPI, the rent of $500 paid by the tenant is a legitimate 11.305 The weights are the amounts consumers spend
consumption expenditure that should be included in the CPI. on each type of provider. In countries where the government
11.299 Users’ expenditure on all relevant goods and provides a substantial portion of medical care free of charge,
services should be included in the CPI to properly reflect the relative importance of medical care in the CPI is less
changes in prices and to support the addition of related ser- than its total expenditure share (by households, government,
vices in cases where governments and NPISH introduce and NPISH), as measured by the national accounts.
charges for services that were previously provided at no 11.306 The COICOP categories are based on purpose.
charge, or vice versa. Some critics state a preference for categorizing according to
11.300 In practice, there are many possible combina- types of medical condition and are of the view that the CPI
tions of payment and reimbursement systems and there should measure the cost of treating a disease or of obtain-
can be a complex mix of publicly financed social security ing a fixed outcome after treatment rather than the medical
programs, employee or worker financed social security service or treatment that, in their terms, is an input. But it
programs, and social security programs financed by house- should be noted that this medical condition approach, which
holds. The CPI compiler should investigate such programs would define elementary aggregates for categories of medi-
in sufficient depth to facilitate a well-informed decision on cal conditions, is not sufficiently developed to be recom-
their treatment in the CPI. Thus, the nature of the health, mended at this time.
education, and social protection sectors is a challenge for
price index compilation. The index methodologies used by Sampling
the NSO need to be built around the national circumstances. 11.307 The general CPI approach is to select samples of
11.301 The procedures used in the treatment of tariffs, providers within each COICOP category (for example, doc-
as described in paragraphs 11.211–11.248, are often appli- tors in Medical Services, or dentists in Dental Services), and
cable for pricing programs encountered in health, education, then choose one or more representative service items for each
and social protection services. sampled provider. The principle of using loose specification
11.302 Detailed in-depth research will be needed to when sampling and tight specification when pricing is fol-
identify and measure newly introduced significant charges. lowed. When first visiting a doctor or other medical provider,
Additionally, some of the goods and services included in the price collector should find what services are provided or
health, education, and social protection are difficult to mea- what medical procedures are performed and select a repre-
sure at constant quality. The social security systems can also sentative sample with guidance from the provider. The price
cause complications. The CPI compiler is advised to keep collector should describe them as completely as possible, and
in close contact with data users to obtain the information then continue pricing them for as long as possible or until
that is necessary to make informed decisions about which they are no longer part of the sample that constitutes the CPI
charges should be included in the index and how they should basket. One approach to the initial sample selection of ser-
be measured, and to gain access to relevant sources of data vices or medical procedures is for the price collector to ask
and advanced notice of any changes. the medical practitioner or dentist what was provided recently
to a typical patient with a representative medical condition
30
Social transfers in kind consist of individual goods and services provided or to ask the respondent to describe a simple service they
to individual households by government units (including social security recently performed. When returning to obtain prices in the
funds) and NPISHs (2008 SNA, paragraph 8.141), whether purchased following periods, the price collector should collect the price
on the market or produced as nonmarket output by government units or for that identical service, even if the doctor or dentist has
NPISHs.
275
the reduction to two visits to the doctor is associated with

Figure 11.2 Loose Specification Sampling and Tight
Specification Pricing
a shorter course of treatment and with the patient being in
less pain and suffering fewer adverse side effects, then the
Sampling Collection Replacement quality-adjusted price should see a steeper price reduction
Agree Service to be Collect Price of That Identify Replacement because of the better treatment. As quantifying this change
Priced with Medical Exact Service Across Procedure If Original in quality is generally not straightforward, implicit methods
Practitioner (for Months Service No Longer of quality adjustment are usually applied.
example, prosthetic Provided (for example,
replacement of hip prosthetic replacement
joint) of knee joint) Coverage of Medical Insurance
11.312 The existence of medical insurance adds com-
plexity to the situation and there is a lack of consensus
regarding measurement in a CPI. Some consumers buy
not performed it recently. However, if the respondent has not
medical care directly from medical care providers, while
performed it, for example, for a year or indicates they will no
others buy medical insurance that pays for some or all
longer perform that particular procedure, the price collector
their medical care. It is useful to review exactly what kind
should find a replacement procedure, preferably for the same
of product the consumer is purchasing. Medical insurance
medical purpose. The head office should then decide if it can
may be considered a way of prepaying for likely future
be treated as a comparable item (see Figure 11.2). There is
medical expenses, a way of reducing the risk of catastrophic
less need to price many types of medical services from the
expenses (that is, giving peace of mind), and a way of reduc-
same doctor, when it is known that price movements among
ing the total expected cost of medical care because insurance
various medical services move closely together.
companies have market power in determining prices that
individual consumers lack. The view taken about what is
Pricing
being purchased, which may be a mix of the previous ones,
11.308 There are two approaches to pricing that align
has implications for measurement. For example, the use
with the view taken on whether an input or medical con-
of net premiums for weights fits better, conceptually, with
dition approach is followed. The traditional input approach
insurance being purchased for peace of mind.
treats medical items as consumption items, without regard
11.313 In many countries, the government provides
to their effectiveness in preventing, curing, or ameliorating
health insurance either for free or through compulsory or
an illness or injury. The approach prices a selection of medi-
voluntary contributions. This is generally out of scope for
cal items, such as an annual check-up by a doctor or a par-
the CPI regardless of how the government funds it and is
ticular surgical procedure, such as an operation in a hospital
excluded from the CPI weights and pricing samples. Com-
or clinic, and follows the cost over time.
pulsory or voluntary contributions to public health insurance
11.309 The alternative approach, the treatment approach,
are not covered in household final consumption expenditure
is a partial response to increasing criticism that the input
as these are categorized as social contributions and treated
approach ignores medical advances and does not consider
as transfers. Health care that is covered by private insurance
that the patient generally looks to buy a cure to a medical
(or not covered by any health insurance at all) is part of the
condition rather than a specific course of treatment. Under
coverage of a CPI. Only the cost of insurance paid directly
the treatment approach, the CPI compiler or price collector
by the consumer is in the scope of the CPI.
first selects a specific medical condition by selecting a dis-
11.314 In constructing expenditure weights, care should
ease or injury that a patient recently received treatment and
be taken to avoid the double counting of expenditure. Con-
then follows the price of treating that condition regardless
sumers may pay insurance premiums and then either pay the
of medical procedures or medicines used. For example, if
costs (that is, the medical bills) that the insurance company
treating the disease or injury initially requires five visits to
reimburses in full or partially, or the insurance company
the doctor, the CPI price is the cost of five visits. However,
directly reimburses the providers (for example, the doctor
if subsequently the doctor reports a new way of treating the
or the hospital).
condition that requires only two visits, then from that point
11.315 In the paragraphs on property insurance ser-
the cost of two visits is included in the CPI.
vices (11.385–11.404) three alternative treatments of insur-
11.310 Only expenditure by households that is a direct
ance are identified that can similarly be applied to health
result of purchase of individual goods or services is within
insurance:
the scope of CPI. It follows that the prices entering the index
should be net of direct reimbursements. Mandatory pay- • Gross insurance premiums/net (medical) expenditure
ments, such as employees’ contributions to social security • Net insurance premiums/gross (medical) expenditure
programs, are out of the scope of a CPI. These are collected
• Gross insurance premiums/gross expenditure
to finance social security programs and do not directly relate
to the provision of goods or services. Reimbursements made 11.316 If applied to health insurance, the first alternative
where the recipient can spend the amount received on some- (that is, gross insurance premiums/net [medical] expendi-
thing unrelated to the treatment that is the subject of the ture) results in a relatively large weight for medical insur-
reimbursement are also excluded. ance and small weights for medical care categories. The
11.311 Both approaches, the input and the treatment second alternative (that is, net insurance premiums/gross
approach, may involve quality adjusting the prices where [medical] expenditure) reverses this. The third alternative
a change of treatment or course of medication results in a (that is, gross premiums/gross expenditure) double counts
change in outcome. For instance, if in the previous example consumer expenditure and should not be used.
276
11.317 The gross premiums method is not recom- plans and average utilization and patient risk are kept con-
mended for CPI compilation, but is supported by some stant over a long period, the change in the premiums of a
economists. Under this method, there is an elementary sample of insurance policies can be used as a proxy for a
aggregate for health insurance with a weight that is based measure of the change in the insurance companies’ over-
on consumers’ total expenditure on health insurance premi- heads and profit per capita (that is, the change in the cost of
ums during the reference period. The expenditure weights of providing the insurance service).
the other medical care elementary aggregates (for example, 11.321 The price relative for the net insurance premiums
for doctors or hospitals) must be reduced by the amount of aggregate can be developed in the same way as under the
insurance payments consumers received in the same period. gross premiums method, using a sample of insurance poli-
These elementary aggregates represent what consumers pay cies. The problem of holding the policies constant over time
directly for medical care. The total expenditure weight is the remains, but this is less troublesome because the weight is
same regardless of whether the insurer reimburses the pro- much smaller. Alternatively, a measure of the change in the
vider or if the consumer pays the provider and is later reim- insurance companies’ net earnings (premiums less payouts)
bursed. The CPI prices for these elementary aggregates are can be used. This may be volatile and net earnings could be
what providers receive from consumers, not from insurance negative in some years, and so NSOs using this approach
companies. It can be argued that, compared with the net pre- often take an average figure over several years. In addition,
mium approach, the use of gross premiums is more consis- such a measure requires data that insurance companies may
tent with the concept of prepaying for likely future medical be reluctant to make available and it takes no account of any
expenses and is therefore better able to cope with a situation change in the volume of insurance services provided. For
where some consumers purchase medical services directly example, earnings may rise because the insurance compa-
and others purchase the same medical services through an nies are taking new business and not because either indi-
insurance company. vidual gross or net premiums are rising.
11.318 The main difficulty with the gross premiums
method is constructing a price relative for gross premi- Weights
ums, which ideally would be a measure of the change in the 11.322 In the first row of Table 11.13, the HBS reports
premiums of a sample of constant-quality insurance poli- that consumers spent on average (from their own pockets—
cies. However, in practice, insurance companies change the government-provided benefits are not in scope) a total of
detail of their policies from year to year, with new rules and $1,800 on medical care, which includes $1,000 on insurance
requirements and changes in what they cover; some of this and the rest on direct payments to medical providers. Under
is in response to the appearance of new medical techniques. a gross premium approach, these are the weights for insur-
It is extremely difficult to track and adjust for these changes, ance and medical care categories, which is standard CPI
but without doing so a premium index is likely to be biased weights: expenditure reported in the HBS of what consum-
upward. ers paid directly.
11.319 For the reason stated previously, the “net” pre- 11.323 To obtain weights for a “net premiums” approach
miums method is the most common approach implemented more data are required. The second row of Table 11.13 pres-
in a CPI. In this case, the CPI weight of the insurance ele- ents the insurance industry’s spending on medical care: the
mentary aggregate is gross premiums paid minus reimburse- insurance companies spent $900 in benefits and retained
ments paid out by insurers. The CPI weight comprises the $100 to cover their costs and their profits. Using this indus-
insurance companies’ operating expenses and their profit, try data to construct CPI weights yields a small aggregate
which will be much smaller than the gross turnover. The for net insurance premiums, as shown in the third row.
net approach distributes part of the expenditure for health
insurance premiums that consumers report in the HBS to the Education
other medical care elementary aggregates such as doctors,
dentists, or pharmaceutical products. This is usually done 11.324 The scope of the CPI is limited to payments
using insurance industry data on the insurance companies’ actually made by consumers. Consequently, fully publicly
income received from premiums and the companies’ pay- funded education is out of the scope of the CPI. However,
ments to doctors, hospitals, clinics, drug stores, and any there can be some minor ancillary fees for materials or
other providers. To do this, the CPI compiler computes the services, such as pens and paper, and sports activities that
share of payments to each type of provider and allocates students must pay for, and these should be included in the
medical insurance premiums to other medical care elemen- index. Out-of-pocket educational expenses for tuition and
tary aggregates in proportion to these shares. Health insur-
ance companies retain a small portion of the premiums not
paid out as benefits to cover overheads and profits. The CPI Table 11.13 CPI Weights: Medical Insurance and
can treat this as a separate elementary aggregate for health Medical Care Elementary Aggregates
insurance to purchase “peace of mind” or just allocate it Total Health Hospitals Doctors Drugs
to the medical care categories using the same proportions. Insurance
Many consider this option to be conceptually more appro- Premiums
priate. Table 11.13 provides an example for the calculation HBS Data 1,800 1,000 300 300 200
of weights. Insurance 900 500 300 100
11.320 The CPI price movement over time can be the Industry Data
same as in the direct approach. Under the assumption that CPI Weights with 1,800 100 800 600 300
Net Insurance
there is proper control for changes in quality of insurance
277
related costs also are treated as standard consumption items. education costs or breaking university tuition fees down
The main issue is determining the main expenditure cate- into subcategories by type of course (for example, post-
gory for their inclusion. graduate or undergraduate). Therefore, a better source of
11.325 The goods and services included in the Educa- detailed weighting information may be the public institu-
tion Services category (COICOP 2018 division 10) cover tions in charge of the educational programs, for example,
educational services provided by the main channels of edu- administrative data from government departments or regu-
cation. These include: latory bodies.
• Education by radio or television broadcasting or over the
Pricing
internet
11.329 The typical item to be priced will be the cost of
• Educational programs, generally for adults, that do not a term or semester at a sample of schools. A school term is
require any special prior knowledge or instruction (these usually longer than one month, that is, the calculation fre-
can include vocational training and cultural development) quency for most CPIs. Standard practice in this case is to
• Literacy programs for students too old for primary school collect prices only in months when the terms begin, and to
including out-of-school secondary education for adults use that same price in the intervening months. The pricing
and young people and out-of-school postsecondary non- months can vary depending on the country or the school.
tertiary education for adults and young people For example, if a school has two semesters, one starting in
September and the other in January, then tuition fees should
11.326 COICOP 2018 division 10 does not include: be priced in these months only. In other months, the price is
• Expenditure on educational materials, such as books carried forward and the same price is used for the remaining
(COICOP 2018 09.7.1) and stationery (COICOP 2018 months of the term. The index and the associated inflation
09.7.4), or education support services, such as health care rate may display step changes, changing only in the months
services (COICOP 2018 06), transport services (COICOP when the terms begin.
2018 07.3), catering services (COICOP 2018 11.1.2), and 11.330 The recommended approach is to use student
accommodation services (COICOP 2018 11.2.0) profiles to price the cost of education for a selection of typi-
cal students who are chosen to reflect different levels of fees
• Driving lessons (COICOP 2018 07.2.4); recreational
and base-weighted to reflect their distribution in the particu-
training courses such as sport or bridge lessons given by
lar educational institute that is being priced. These profiles
independent teachers (COICOP 2018 09.4.6)
should aim to reflect the range of courses on offer, for exam-
11.327 In many countries, the government partially ple, covering full-time and part-time undergraduate courses
finances education services, particularly higher education, in arts, sciences, and medicine; postgraduate research; and
and students pay a portion of the cost of their education and teacher training courses. Where the fee structure is simple,
this portion is in the CPI’s scope. The tuition fees faced by it may be sufficient to simply collect prices for the posted
students are not always the full cost of education, and in or advertised tuition fees directly from the institution or
these cases, the CPI weight should not and does not repre- its website. In practice, any discounted tuition fees can
sent the full cost of providing education to the population, be ignored where it is known or can be assumed that they
just the cost to the individual of acquiring the education ser- change at the same rate as the full fees and that the propor-
vices. Changes to the amount of the government transfer can tion of students benefiting from them remains constant over
lead to changes in the tuition charged to students and the time. However, any assumptions would need to be checked
index should reflect any resulting change in the fees. If a on a regular basis as there is an inherent risk of introducing
student goes to a private school, where there is no govern- an element of bias in the index if differential changes in fee
ment support, the total advertised price P for fees should structures or changes in the student population are missed.
be reflected in the CPI weights and in the price collected, The approach should be used with care. Similarly, the use
as this is what it costs the household to educate the student. of average revenue (total tuition fees divided by the num-
In the case of government support, let s be the share of this ber of students) is generally not an appropriate way to price
total advertised price that is borne by the household (that is, tuition fees that include price differentials, as changes in the
the household is presented with a bill showing a net price composition of the students attending the college, and the
of sP), then the appropriate price for the CPI and for the courses they attend, can affect the average fee even when no
construction of weights is sP. In this way, the proportion of price change takes place.
the cost borne by the household will feed through into the 11.331 In some countries, it is common for stu-
calculation of the CPI; that is, the CPI will reflect the actual dents to travel abroad for their higher education. In such
fees paid by the household that, in this case, will be less than cases, expenditure paid by residents for tuition at for-
the basic advertised price. eign institutions may be deemed to be in scope for the
11.328 Weights for educational services can be diffi- national CPI depending on the geographical coverage
cult to obtain. Information from the HBS can be unreliable of the index and whether the expenditure was incurred
since sample sizes and relatively low response can result in the domestic territory of the country of residence or
in large sampling errors. Additionally, national accounts abroad (as discussed in Chapter 2). Where such costs
data are often not available at the level of detail needed are included, changes in the education or tuition index
to develop CPI weights. The level of detail in accounting from the destination countries can be used to construct
is often varied and may not allow the division of the costs the price relatives for the education component, adjusted
to the level of detail required for index compilation, for to the national currency using the exchange rate applying
example, separating out university tuition fees from other in the reference month.
278
Scholarships and Other Forms of Discounting Table 11.14 The Effects of a Subsidy and Tax Credit
11.332 Schools and universities often reduce the tuition on a CPI
fees for some students. Tuition fee reductions offered in Country A Country B
exchange for work (for example, teaching assistants or par- t−1 t t−1 t
ticipants in sports teams) should not be reflected in the index
as they are a form of income. Gross Price for Childcare 20.00 21.00 20.00 21.00
Services ($/hour)
11.333 The treatment in the CPI of scholarships awarded
Usage in Hours Per Month 50.00 50.00 50.00 50.00
to particularly talented students or tuition assistance in the
Income for Childcare 1,000.00 1,050.00 1,000.00 1,050.00
form of financial aid to students from certain socioeconomic Provider ($/month)
backgrounds, such as low-income households, depends on Government Subsidy 8.00 10.00 n/a n/a
the method of payment: ($/hour)
• Those that are paid directly to students to assist them in Income Tax Transfer Per n/a n/a 400.00 500.00
Month ($/month)
bearing the cost of the full fee but that the students can
Net Price Paid by 12.00 11.00 20.00 21.00
use, if they wish, to finance something else are transfer Households ($/hour)
payments that should not be reflected in the CPI. Net Expenditure 600.00 550.00 1,000.00 1,050.00
• Where the scholarship or subsidy results in students Determining Weight
being subject to reduced fees (that is, they are billed for ($/month)
a reduced amount), it is the reduced fee that should be Net Cost ($/month) 600.00 550.00 600.00 550.00
CPI (t − 1 = 100) 100.00 91.67 100.00 105.00
priced for the CPI.
n/a, Not applicable.
Social Protection
11.334 According to COICOP 2018, social protection
covers nonmedical assistance and support services provided whereas the corresponding index in country B increases by
to persons who are elderly, disabled, having occupational 5 percent. The reimbursement given in country B is clas-
injuries and diseases, survivors, unemployed, destitute, sified as a social transfer so there is no direct applicable
homeless, low-income earners, indigenous people, immi- price reduction. It should also be noted that the expenditure
grants, refugees, alcohol and substance abusers, and so on. weights differ according to the approach used for reducing
It also covers assistance and support services provided to the burden. In country A, this value is £600 in year t – 1,
families and children. and in country B, the equivalent value is 1,000 pesos. These
11.335 As with health and education, the purchaser are the amounts that would be reported by a household in
price of goods and services in the social protection sector is an HBS.
the amount to be paid by consumers, net of reimbursements.
11.336 Suppose that in country A, the entity offering Financial Services
childcare services charged a price of £20 per hour at period
t − 1; and at period t, the price of the services increases to Introduction
£21 per hour. Children typically spend 50 hours per month
in the kindergarten so that the generated revenue for the 11.339 The construction of price indices for financial
entity for the service goes from £1,000 per child to £1,050. services is by nature challenging, as there is no consensus
But in recognition of the greater financial burden on the about which financial services ought to be included in the
household, the hourly support offered by the public authori- CPI or how they should be measured. The discussion in this
ties is increased from £8 at period t − 1 to £10 at period t. section attempts to present what is feasible from a practical
The net price paid by the household per hour falls from £12 perspective.
to £11 (= £21 − £10). 11.340 Common examples of financial services acquired
11.337 In the case of country B, the exact same changes by households include: financial advice; currency exchange;
apply except that here the authorities provide an income services associated with deposit and loan facilities; ser-
tax transfer instead of a price subsidy. As a result of the vices provided by fund managers, life insurance offices, and
increased price, the household now receives an increase in superannuation funds; stockbroking services; and real estate
its childcare grant from 400 pesos at period t − 1 to 500 agency services. The range of financial services appropriate
pesos at period t. to be covered in a CPI, and the way to measure them, depend
11.338 Note that regardless of the approach used for on the principal use of the CPI and hence on whether an
reducing the burden from the higher price of the service, the acquisitions, use, or payments approach is employed.
net cost to the household of consuming the childcare service 11.341 Financial services covered in this section come
(a drop from $600 to $550) is the same for both countries.31 under COICOP 2018 12.2. This group covers actual charges
However, the impact on inflation as measured by the CPI for the financial services of banks, post offices, saving banks,
differs depending on the approach used for supporting child- money changers, and similar financial institutions; fees and
care services. The index in country A falls by 8.33 percent service charges of brokers, investment counselors, tax con-
sultants, and the like; and administrative charges of private
pension funds and the like. Also included in COICOP 2018
31
For country A, the immediate out-of-pocket expenditure is the net price 12.2 are financial intermediation services indirectly mea-
paid times the hours of the service which are consumed ($12 × 50 = $600 sured (FISIM) but in practice these are sometimes excluded
at period t − 1). from a CPI (see later discussion). Insurance is not covered
279
under financial services as it comes under COICOP 12.1. practice is more complicated, confounded by conceptual
This group includes service charges for insurance as classi- and measurement issues.
fied by type of insurance, namely, life insurance and nonlife 11.346 For example, the EU HICP and the CPIs com-
insurance (that is, insurance in connection with the dwell- piled in some countries exclude life insurance services, for
ing, health, transport, and so on).32 A discussion on medi- practical reasons and because some life insurance services
cal insurance is included in paragraphs 11.312–11.323, and (notably those that are fixed term and with profit) can be
mortgage arrangement and application fees are discussed in a form of tax-efficient saving. The premiums paid for life
paragraphs 11.81–11.146. insurance (COICOP 2018 12.1.1), including pension-
11.342 If a payments approach is used, the gross interest funding services, are partly to be regarded as savings. Life
payable on mortgages is often included as a cost of owner- insurance services are thus excluded from the HICP, as it
occupied housing services. For consistency, this might sug- is not feasible in practice to separate the implicit service
gest that the CPI should also cover consumer credit charges, charge from the investment component. In accordance with
measured in a similar way to mortgage interest charges, CPI conventions, nonlife insurance services are, however,
as well as gross outlays on direct fees and charges paid in included in the HICP. Public insurance connected with
respect of other financial services. In practice, and as noted health (ECOICOP33 12.5.3.1) is generally considered out-
in the section on housing services costs, the treatment of side the scope of a CPI, including the HICP, as compul-
housing services sometimes differs in concept from other sory contributions under social security programs are not
interest charges in national CPIs, partly reflecting mixed included in household final consumption expenditure.
objectives for the overall index combined with public per- 11.347 The HICP also excludes FISIM (COICOP
ceptions of what should be included in a CPI and the impor- 12.2.1), which comprises those parts of financial services
tance of housing services costs in household budgets. that are charged for by way of the interest margin of financial
11.343 On the assumption that households acquire all institutions. It is excluded because it is difficult to measure
their financial services from the private sector, and that the transaction. On the other hand, those financial services
these services are not generally supported by government that attract explicit charges, for example, annual charges
or provided by NPISH, the acquisitions and use approaches for credit cards (excluding interest charges), bank charges
take similar views of the measurement of financial services. for money transfers, and explicit currency exchange com-
With regard to coverage, however, some proponents of the missions, are included in the HICP. The HICP mentioned in
use approach take a more restrictive view of which services this section is one example of how financial services can be
should be included by limiting the scope to only those finan- treated in a CPI.
cial services that are acquired to directly facilitate current 11.348 One broad definition that could be adopted for
household final consumption. the coverage of financial services within the CPI is all those
11.344 Under the more restrictive view of coverage, services acquired by households in relation to the acqui-
it is argued that the use of some financial services is more sition, holding, and disposal of financial and real assets,
associated with capital or investment activity. This implies including advisory services, except those acquired for busi-
that such activities should be considered outside the scope ness purposes. This definition serves two purposes: first, it
of CPIs intended to measure changes in consumer prices. distinguishes between the services facilitating the transfer
Proponents of this view often draw upon national accounts and holding of assets and the assets themselves; second, it
practices as the starting point. For example, the 2008 SNA makes no distinction between whether the underlying asset
classifies expenses associated with the transfer of real estate is a tangible asset or a financial asset.
(for example, real estate agents’ commissions, legal fees, 11.349 The degree of complexity involved in placing a
and government taxes and charges) as part of gross fixed value on financial services acquired by households and in
capital formation (2008 SNA, paragraphs 10.48–10.52). constructing the companion price indices varies by service.
Although harmonization is desirable, in some countries The following examples illustrate the issues. It should be
some CPI concepts do not precisely follow the national noted in regard of the example on deposit and loan facili-
accounts concepts. ties that the complexity involved in measuring FISIM and
11.345 Practices relating to financial services vary. In an aversion to the overuse of imputed prices in a CPI have
principle a CPI should include the prices of all goods and led to many compilers to follow the HICP example by not
services included in household final consumption expendi- including these services in a CPI.
ture. Nonconsumption expenditure such as financial trans-
actions, transfers, and purchases of financial assets are Currency Exchange
excluded. According to the 2008 SNA all insurance services
are within the scope of household final consumption expen- 11.350 For weighting purposes, the estimation of the
diture and are to be included by the amount of the implicit weight reference period expenditure incurred by house-
service charge (2008 SNA, paragraph 9.64). However, the holds in exchanging domestic currency for currencies of
other countries is not entirely straightforward. In practice,
the amounts reported by HBSs can be biased downward
because of forgetful underreporting of currency exchange
32
Life insurance are generally excluded from the coverage of a CPI as the transactions by households. The national accounts can often
premiums paid for life insurance (COICOP 2018 12.1.1), including pen- provide a more accurate source for weighting data.
sion-funding services, are regarded as savings. Also, there are measurement
issues associated with the difficulty of separately identifying the implicit
33
service charge of the insurance element from the implicit service charge of ECOICOP refers to the European Classification of Individual Consump-
the investment component. tion According to Purpose.
280
11.351 Construction of the price index for currency change resulting from the value of an underlying transaction
exchange services is in general more complex. The ser- moving from one price band to another.
vice for which a price is required is that of facilitating
the exchange of domestic currency for other currency (the
acquisition of an asset—foreign currency). The price for Stockbroking Services
the service is often specified as some percentage of the 11.355 Stockbrokers buy or sell shares or other secu-
domestic currency value of the transaction plus a fixed per rities on behalf of clients (for example, the purchase of a
transaction charge. The percentage margins may change parcel of shares in a publicly listed company; in most coun-
only rarely, with service providers relying on the nominal tries this has to be arranged through a licensed stockbroker).
value of the transactions increasing over time to deliver The service provided consists of arranging for a transaction
increases in fee receipts. The price required for index con- to take place on the conditions specified by the client (that
struction purposes is the monetary value of the margin is, a certain block of shares that is being bought or sold).
(that is, the amount determined by applying the percent- The total amount paid by the purchaser generally comprises
age rate to the value of the currency transaction) plus the three elements: an amount for the shares (the asset), a fee
fixed charge. To measure price change over time, the CPI for the brokerage service, and some form of transaction tax
compiler needs to form a view about the quantity under- (stamp duty).
pinning the original transaction. Further feasible forms of 11.356 The tax should be considered part of the cost of
charging, other than percentage plus fixed fees, occur for acquiring the shares, as opposed to being part of the price of
currency exchange services, such as fees following a step the security. The tax should be included along with the bro-
function or a spread between service providers’ selling kerage cost in the CPI. This is consistent with both the inten-
and buying rates (that is, an implicit rather than an explicit tion of the tax and the more commonly accepted basis for
fee). The following description pertains to the issues in the valuation of the shares. It also aligns with the compara-
the case of percentage fees (plus fixed fee) for currency ble treatment of taxes on banking services. Allowing for cur-
exchange. rent tax schedules is feasible as they are publicly available.
11.352 The purchase of foreign currency can be viewed 11.357 Working from the premise that stockbrokers’
as facilitating the purchase of some desired quantity of for- fees can likely follow a step function, as opposed to a linear
eign goods and services (for example, purchased during function, a price measure would be constructed as follows.
traveling abroad, or for the direct import of a product). The First, a representative sample of transactions (domestic cur-
service price in comparison periods would be expressed as rency values) is selected and the fees and the tax payable
the amount payable on the conversion of a sum of domes- are calculated according to the respective schedules. A rep-
tic currency corresponding to that sum of foreign currency resentative unit of transaction should be the charge to be
required to purchase the same quantities of foreign goods paid by consumers in exchange for trading a set basket of
and services purchased in the price reference period. Thus, securities (defined in value terms) that are representative of
the fixed basket relates to the purchase of a fixed good or the base or reference period. The fees and taxes payable in
service in a foreign currency. subsequent periods are calculated by first indexing forward
11.353 To follow the principle of a fixed basket, the the values of the sample transactions and then applying cur-
original foreign currency amount should be indexed for- rent fee and tax schedules to the revalued transactions. This
ward using changes in foreign prices, and then converted to method raises two main questions: first, what is the most
domestic currency at the prevailing exchange rate, with the appropriate index for revaluing the transactions; and, sec-
prevailing percentage margin applied to this new amount, ond, how should the current schedule of fees be determined.
plus any fixed charge, to produce the current price. This cur- 11.358 The quantity underlying share transactions can
rent price would be compared to the base price to derive be regarded as forgone consumption, or the quantity of
the measure of price change. Although the ideal measure goods and services that could have been purchased instead.
for indexing forward the foreign currency amount would The value of a constant set of quantities of consumption
be an index specifically targeting those foreign goods and forgone in successive comparison periods thus will vary
services purchased by households wanting to exchange with consumer prices. In this case, one choice for a revalua-
currency, this is unlikely to be feasible. A practical alterna- tion index is the CPI, based on current month’s or quarter’s
tive is to use the published all-items CPI for the relevant preliminary estimates, or the previous month’s or quarter’s
foreign countries. A further approximate simplification for published CPI. However, it could be argued that the use of
the revaluation by indexing forward is to use the CPI of a single period’s movement in the CPI has the potential to
the country of residence; this can be defended as following show the prices of stockbroking services as moving in a way
potential expenditure for consumption forgone (see para- not reflecting reality. This could occur if the current or pre-
graphs 11.371–11.375). vious period’s CPI was influenced notably by some unusual
11.354 If a single margin (percentage rate) does not price change (for example, an oil price shock or change to
apply to all transactions (for example, different rates apply health care arrangements). As an alternative, a 12-month
to different size transactions), then the price measure should moving average CPI might be employed, consistent with a
be constructed by reference to a representative sample of price reference period comprising a full year’s activity.
price reference period transactions. The value margin for 11.359 Hypothetically, it might be argued from a con-
each transaction in the current period in the domestic cur- ceptual viewpoint that the set of shares could be revalued in
rency should be determined by the current domestic cur- subsequent periods in line with movements in equity prices.
rency value of each transaction and the current period According to this view, the price of equities would be seen as
percentage margin applying to each. This captures any price an important influence on the actual costs of storing forgone
281
consumption in much the same way as fee and tax schedules Deposit and Loan Facilities
specific to equity purchases are allowed to enter the calcula- 11.363 Accounting for the costs of services provided by
tions described previously. The strong argument against this financial intermediaries represents a step up in complexity
treatment is that it assumes that households have a desire to in CPI compilation. Even where a prior decision has been
acquire equities per se, rather than using them as an advanta- made to include such facilities within the scope of the CPI,
geous vehicle to store forgone consumption. Moreover, the the service being provided is challenging to visualize com-
introduction of equity prices within the price index is likely prehensively, and the prices comprise elements that are not
to impart additional short-term irrelevant volatility to the directly observable.
CPI, although not significantly so given the relatively small 11.364 The 2008 SNA (paragraphs 6.163–6.169 and
weight for brokerage fees in a CPI. A4.33) recommends that the value of financial intermedia-
11.360 Competition in the stockbroking industry means tion services output produced by an enterprise should be
that there is unlikely to be a common fee schedule. If indi- valued as the following sum:
vidual brokers adhere reasonably closely to an in-house fee
schedule, obtaining copies of these schedules should be rela- • For financial assets involved in financial intermediation,
tively simple. On the other hand, if no such fee schedules such as loans, the value of services provided by the enter-
exist, then a survey of stockbrokers may be required to col- prise to the borrower per monetary unit on account is the
lect information on a sample of trades (value of trade and margin between the rate payable by the borrower and a
fee charged) and this information used to derive a current reference rate; plus
period fee schedule. • For financial liabilities involved in financial intermedia-
11.361 In the case of sales of shares, the underlying tion, such as deposits, the value of services provided by
transaction represents the exchange of one asset for another the enterprise to the lender or depositor per monetary unit
(shares for cash). Quantities underlying sales can be viewed on account is the margin between the reference rate and
similarly to share purchases (that is, some current period the rate payable by the enterprise to the lender; plus
basket of consumption goods and services). In practice, • The value of actual or explicit financial intermediation
households presumably review their investment strategies service charges levied
occasionally to store their deferred consumption in asset
forms they may expect to offer security or prospect for 11.365 The Organisation for Economic Co-operation
growth. A symmetrical treatment of the purchase and sale of and Development has been instrumental in developments in
shares is a natural approach. Unless different fees or taxes the national account’s treatment in this area.34 In concept,
apply to sales, there is no need to distinguish between the the 2008 SNA describes the reference rate as the risk-free
two in constructing the index. or pure interest rate. The value of the service provided to
a borrower is the difference between the actual amount of
interest paid by the borrower and the lower amount that
Investment Funds would have been paid had the reference rate applied. The
11.362 Investment funds often apply an annual manage- converse applies for depositors. In practice, it is difficult to
ment charge defined as a given percentage of the current effectively identify the reference rate, and in particular to
asset value. For example, the charge can have the form that avoid volatility in or even negative measures of the value of
1.50 percent of the asset value is deducted annually from such services (as would occur if the reference rate lay above
the latter. This charge can be applied either instead of or in the lending rate or below the deposit rate). As a matter of
addition to charges applied to the buying or selling of fund practical expediency, an average of borrowing and lending
shares. In the CPI, the annual management charge propor- rates may be used, with preference for the midpoint. Given
tional to asset value can be treated in a similar way as a the complexities involved, expenditure on financial inter-
charge for currency exchange proportional to the transaction mediation required for index weighting purposes cannot be
value. Funds can be discontinued from time to time and in collected from households in expenditure surveys and thus
the price collection they should then be replaced with similar must be estimated by collecting data from financial institu-
funds to ensure that the index reflects actual price develop- tions or a regulatory authority.
ments. The underlying transaction for annual management 11.366 Concerns have been expressed about the use of
charges can be defined as the annual management of fund a midpoint reference rate as a measure of the risk-free rate
shares worth a given amount in monetary terms in the price of interest. There are, however, some doubts about whether
reference period. The underlying transaction is revalued
periodically by the CPI. The use of the CPI is motivated by 34
The SNA uses the concept of FISIM to put a value on financial services
the fact that the user functionality of the service deteriorates that are not explicitly priced. However, measurement is often narrowly
with inflation, which makes the monetary value of the assets defined around the traditional deposit/loan business, thereby leaving out
handled less useful to the consumer. The updating with the other financial instruments that may be carriers of financial services with
CPI adjusts for this change in user functionality. A stock implicit prices. An extension of the notion of FISIM relates the use of out-
put indirectly measured to obtain as complete as possible a measure of the
price index such the FTSE 100, DAX, or Dow Jones indices services produced by financial corporations. Not doing so may underes-
should not be used as these do not keep the user functional- timate value added. Nevertheless, considerations have recognized current
ity of the service constant. Namely, the more the investment practice of computing FISIM on deposits and loans only as a workable
is worth, the greater is the user functionality of the service and useful way of estimating the value of financial services, although these
calculations are not necessarily an exhaustive measure of the value of
managing it. Also, stock price indices follow asset price val- indirectly measured financial services. See, for example, the report of the
ues that can be volatile and asset price movements would OECD Task Force on Financial Services (Banking Services) in National
dominate the modeling of the service charge. Accounts, 2002, particularly Conclusion 5 of Section 14.2.
282
the conceptual ideal is for some “risk-free” interest rate, or homogeneous products that can then be weighted together to
whether a more appropriate concept might be the interest provide a measure for deposit and loan facilities, in aggre-
rate that would have been struck in the absence of finan- gate, and taking account of both the explicit and implicit
cial intermediaries (that is, the rate that would have been elements in total price. This represents a similar strategy to
struck by depositors dealing directly with borrowers). Such that adopted throughout the CPI. For example, the index for
a rate would have incorporated the lenders’ knowledge of motor vehicles is constructed by pricing a sample of indi-
risk. Taking the midpoint of the borrowing and lending rates vidual vehicles and weighting these price measures to derive
would appear to be a good means of estimating this market- an aggregate, instead of, for example, attempting to directly
clearing rate. construct an index for the supplier or producer of a range of
11.367 When planning the construction of the index vehicles.
number, it is useful to start by considering the case of a tra- 11.372 A basic process to treat the services of financial
ditional bank providing a single loan product and a single intermediaries in the CPI is, first, to select a sample of rep-
deposit product; the example will then be extended to a resentative products from each sampled institution; second,
typical bank. Much of a bank's income is derived through an to select a sample of customers for each product; and third,
interest margin on lending rates over deposit rates. to estimate the total base period value of the service for each
11.368 The base period weighting value of the finan- product by element (margin, direct fees, and taxes). These
cial service (and so household final consumption of such values can be viewed as being equivalent to prices for some
services) therefore is estimated by applying a margin (the quantity. Comparison period prices are derived by moving
absolute difference between the reference rate and the rate forward the base period value aggregates as follows:
of interest charged to borrowers or paid to depositors) to an
• Margin—index forward the base period balance and apply
aggregate balance (loan or deposit). In line with the sug-
the comparison period margin (the difference between the
gested treatment of other financial transactions, the con-
comparison period reference rate and the product yield).
struction of accompanying price measures should allow for
In practice, the price movement is given as the product of
the indexation forward of price reference period balances,
the indexation factor and the ratio of margins.
applying comparison period margins to calculate a money
value. The price index is then calculated as the ratio of com- • Fees—index forward the transaction values for each
parison period and price reference period money values. sampled account (or profile) and apply the comparison
11.369 Again, the issue of choosing an appropriate period fee structure. The ratio of new aggregate fees to
revaluation index needs to be addressed as with currency base fees is used to move the fee value aggregate. The
exchange. While the base period flows of deposits and with- aggregate fees in the base and comparison periods can be
drawals can readily be conceptualized as foregone consump- constructed as either arithmetic or geometric averages of
tion at base period prices, how should the balances (stocks) the fees calculated for the individual customers.
reflecting an accumulation of flows over a number of years, • Taxes—as for fees but use tax schedules instead of fee
be viewed? If an age profile for balances were available, schedules.
accumulated consumption forgone could be computed as a 11.373 Annex 11.3 contains an example of the calcula-
moving average of the CPI. The more practical alternative tion of a price index for a single deposit product.
is to view base period balances as representing some set of 11.374 Since step-function pricing and taxing schedules
quantities of final consumption goods and services at base (for example, fees that are only payable after some num-
period prices, in which case the 12-month moving average ber of transactions or if balances fall below some level) are
CPI can be used. This is consistent with the idea that house- prevalent in financial services, samples of detailed customer
holds review temporal consumption or investment decisions accounts with all the necessary charging variables identified
(and so accumulated financial balances) on a regular basis, will be required. These samples should cover a full year’s
in this case annually. activity. If it is not possible to sample actual accounts, cus-
11.370 The traditional bank is not currently so frequent tomer profiles may be developed as a fallback option.
in some countries and generally financial institutions now 11.375 To minimize problems associated with nonre-
derive income from a combination of implicit charges (mar- sponse and changing industry structures, a separate refer-
gins) and explicit fees, with the trend having been for a ence rate should be constructed for each sampled service
move toward more use of explicit fees and relatively less use provider. The reference rate should be calculated in respect
of margins. In this case a challenge is to construct measures of all loans and deposits (including those to businesses).
of price change that reflect the total price of the service and Furthermore, to avoid problems that may arise in the tim-
therefore capture any offsets between margins and direct ing of accounting entries (for example, revisions or interest
fees. As with stockbroking services, there may also be taxes income on credit cards), monthly yields, reference rates, and
levied on financial transactions or balances and these should margins should be constructed by reference to three-month
also be included in the price. Frost (2001), for example, pro- moving averages of the reported underlying balances and
vides a description of practical aspects of constructing price interest flows.
indices for deposit and loan facilities.
11.371 Given the options for financial intermediar-
ies to shift charges between the explicit or direct (fee) and Credit and Debit Card Fees When Abroad
implicit or indirect (margin) elements, there are risks in 11.376 The use of credit and debit cards while abroad to
choosing measures of margins (that is, FISIM in the SNA) purchase goods and services, or to withdraw cash, is usually
independently of direct fees and taxes. Rather, the approach subject to explicit fees and charges. The fees levied on the
should be to construct price measures for specific relatively use of the card abroad are not part of the price of the good
283
or service purchased as consumers have the option of pay- dwelling for a known price) can be computed by multiply-
ing in cash. However, ideally it should be considered that ing the value of the dwelling by the percentage margin, and
as the fees that consumers are charged for using their cards the index can be constructed on the basis of estimates of
abroad are levied by the service provider (bank or credit both components.
card company) in the country of residence and not in the 11.380 The method chosen for estimating the percent-
country where the purchase occurred, these charges should age margin will depend on an assessment of the variation in
be allocated to the country of residence. By contrast the situ- margins across and within individual agencies. In the most
ation is different for internet purchases as the credit or debit straightforward case, firms may operate with a single per-
card charges are unavoidable because paying by cash is not centage margin applicable to all transactions regardless of
an option, thus being part of the price for the good or service value. In other words, at any point in time the percentage
bought on the internet. margins charged may vary by agency, but not by value of
11.377 The underlying transactions for using credit and transaction within agency. In this case, what is required is
debit cards abroad typically include: an estimate, in each comparison period, of the average per-
centage margin charged by agencies. This can be achieved
• The service fee charged to the consumer’s account for
by collecting the percentage margins, exclusive of any taxes
using an automated teller machine (ATM) outside the
levied on agents’ fees such as VAT or GST, from a sample of
country of residence to withdraw an amount of local cur-
agencies and deriving an average.
rency equivalent of a given amount in the currency of the
11.381 Percentage margins charged by individual agen-
consumer’s residence, in the price reference period.
cies sometimes vary with transaction price (typically declin-
• The service fee charged to the consumer’s account when ing with higher prices of dwellings). Where tariffs do vary
purchasing abroad a good or service of a given value in within agencies, a more sophisticated estimation procedure
the currency of the consumer’s residence, in the price ref- may be needed. Using data from a sample of transactions
erence period. The value in real terms of the underlying from a sample of agents, the relationship between the value
transactions should be kept constant by monthly price of transaction and the percentage margin can be derived
updating using the CPI for revaluation, until the next through econometric analysis. Empirical analysis may be
annual basket update. It may be noted that keeping the useful to determine the functional form for this relationship.
real value of the reference transaction constant implies In one country, for example, research has shown that ordi-
that the value in nominal terms varies. nary least squares regression can be used to estimate this
• Prepaid cards for, for instance, toll charges for bridges relationship and that the following functional form can be
and motorways are allocated to the COICOP division adequate:
of the good or service being purchased. The transaction
price is the corresponding cost of, for example, one cross-
ing of the bridge. R = a + b1(1/p) + b2(1/p)2 (11.7)
where R denotes the commission rate, p denotes the house

Real Estate Agency Services price, a is a constant, and b1 and b2 are parameters to be
11.378 The services provided by real estate agencies in estimated.
the acquisition and disposal of properties can be treated in 11.382 Estimation of the current period value of trans-
various ways. The transfer costs involved in the acquisition actions to which the percentage margin applies depends on
of a dwelling (legal fees, real estate agency fees, and taxes) whether real estate agency fees are classified as (1) a cost of
can be included in both a payments and an acquisitions CPI. housing or as (2) a separate financial service. In the former
They can be classified as either a cost of homeownership or case, the value of the current period transaction, relative to
as a distinctly separate financial service. Although all trans- the value of the price reference period transaction, would
fer costs should be included in such measures, the following follow changes in house prices. In the latter case, the pur-
discussion focuses on real estate agents’ fees for simplicity. chase of a dwelling is regarded as forgone consumption, and
Price measures for the other elements are calculated using the current period value would follow changes in the CPI.
similar procedures. In all cases, the general approach is to 11.383 If a single percentage margin is assumed to
estimate the current cost of the various services relative to, operate, then only a single current period transaction value
and as they would apply to, some fixed basket of activity in is needed, namely, an estimate of the average value of price
the base period. Consistent with some of the areas already reference period transactions at comparison period prices.
discussed, this involves indexing forward the weight refer- For example, if real estate agency fees are classified as a
ence period expenditure on which the fees are charged (to housing cost, then the base period price is calculated by
preserve the underlying quantity) via some appropriate price applying the average price reference period percentage mar-
index, and then estimating the fees payable in the compari- gin to the average house price in the price reference period,
son period. with any VAT or GST then added. The comparison period
11.379 Real estate agents typically state their fees as price is calculated by revaluing the average base period
some percentage of the price settled for the dwelling. In house price by indexing forward, and then applying the
common with other services for which charges are deter- average comparison period percentage margin and adding
mined as a margin, this needs to be converted to a domes- GST or VAT.
tic currency price in cases where the sale price is stated in 11.384 If a single percentage margin is not assumed to
a foreign currency. If the percentage margin is known, the operate, then a sample of representative base period transac-
agents’ price for any given transaction (sale/purchase of a tions is needed. The monetary value of the margin on each
284
representative transaction is then calculated from published of purchases financed by insurance claims avoids double
tariffs or from an estimated functional relationship, such counting of that portion of gross premiums that funds the
as that described previously. Comparison period prices are claims. But there are some problems with this approach.
likewise estimated by first revaluing by indexing forward First, it is assumed that all proceeds from insurance claims
for each of the base period representative transactions, and are used to purchase replacement products or to repair
then applying the same model. damaged property. Second, in some cases, claims receiv-
able may be to compensate for damage to the property of
agents beyond the scope of the index (for example, busi-
Property Insurance Services nesses, government or even other households where the
11.385 The construction of reliable price indices for CPI reference group covers only some subset of house-
insurance can be challenging. This section is restricted to a holds). Households may also choose to use the proceeds for
discussion of property insurance, as this type of insurance entirely different purposes. Thus, the estimation of the net
can be assumed to operate in similar ways across countries. expenditure weights is likely to involve some partly arbi-
It provides only an illustration of the issues to be met, with trary choices. More generally, because money is fungible
each sector raising specific conceptual and measurement (that is, one sum of money can be replaced by another as
difficulties. The separation of the service charges relating to money can be used to purchase a variety of goods and ser-
the insurance and investment elements within a single pre- vices), attempts to restrict coverage only to that expendi-
mium poses particular problems. ture made from selected sources of funds is a questionable
11.386 For the purposes of this discussion, property practice. Finally, the potential distortion of weights for the
insurance is defined to include: products concerned might possibly reduce the usefulness of
• Dwelling insurance subindices for other purposes.
11.390 Considering the second option, net premiums,
• Household contents insurance
gross expenditure, within a payments index, this approach
• Motor vehicle insurance is based on the view that claims receivable should be
11.387 The common feature of these types of insur- regarded as negative expenditure on insurance. This may
ance policies is that for a fee (premium), households receive be viewed as an attempt to avoid a double counting of
financial compensation if a nominated event results in dam- expenditure on products financed by claims receivable
age to, or loss of, specified property. The alternative to pur- and already included in gross expenditure on other con-
chasing insurance is for the household to self-insure. For sumption products elsewhere in the index. The net premi-
households, the service received relieves the risk of financial ums approach is in some ways less problematic than the
loss. The appropriate treatment of property insurance with net expenditure approach, as the impact is restricted to
respect to scope, weighting structure, and pricing in the CPI the weights for insurance and does not affect weights for
partly depends on whether the CPI is constructed to follow replacement product expenditure financed out of claims. It
the acquisitions, use, or payments approach. may, however, be argued that the net premiums approach
is inconsistent with approaches adopted for other products
in a payments index—in particular, with mortgage interest
Payments Approach and consumer credit charges, where weights are based on
11.388 Under the payments approach, each of the pre- gross payments. Any allowance for interest receipts would
vious policy types is in the scope of the CPI. In thinking yield negative weights when households are net savers
about how property insurance should be included in the overall.
CPI, it is appropriate to consider both the gross premiums 11.391 The net premiums approach effectively mea-
payable and the claims receivable by households for inclu-
sures the value of the insurance service and can be appropri-
sion in weights. The definitions of gross premiums payable ate for indices constructed according to both the acquisitions
and claims receivable are straightforward. It is possible, and use approaches, as well as for a payments-based index
however, to treat claims receivable in various ways that considered here. It also tends to be favored for medical
will have an impact on either the weight assigned to insur- insurance although with such finely balanced arguments it is
ance or the weight assigned to consumption products difficult to be prescriptive.
insured. Spending on insurance can be weighted on either a 11.392 The use of gross premiums, gross expenditure is
gross basis (that is, valued using gross premiums payable) based on the view that the claims receivable by households
or on a net basis (that is, valued using gross premiums pay- represent one of the sources of funds from which expendi-
able less claims receivable). Similarly, damage to property ture are made. This is the main justification for using gross
insured may also be weighted gross or net (in the latter expenditure. The gross premiums, gross expenditure method
case, excluding purchases explicitly financed by insurance is the most appealing approach for a payments index, as
claims receivable). Taken together, this suggests three it recognizes the fungible nature of money for purchas-
basically plausible alternative treatments in the weighting ing goods and services and provides a consistent means of
structure: identifying both the product coverage of the index and the
• Gross premiums, net expenditure relative weights by reference only to the actual outlays of
• Net premiums, gross expenditure households.
• Gross premiums, gross expenditure
Use Approach
11.389 With respect to gross premiums, net expen- 11.393 Under the use approach, dwelling insurance
diture, it may be argued that calculating expenditure net can be seen as out of the scope of the CPI as conceptually
285
it is an input cost of the notional landlord, which is consumers’ decisions to insure are based on their assessment
reflected in the imputed rent. The weights should relate of the likelihood of suffering a loss compared to the pre-
to the value of the insurance service consumed by house- mium charged, the risk factors should be held constant. On
holds. This is defined as being equal to gross insurance the other hand, it may be argued that, once insured, the con-
premiums payable by households, plus premium supple- sumer expects to be compensated for any loss irrespective of
ments, less provisions for claims, and less changes in changes in risk. From the perspective of the consumer, any
actuarial reserves. increase in risk represents an increase in the insurer’s cost
11.394 It is not possible to estimate the nominal value base (which may or may not be passed on to the consumer
of the net insurance service from HBSs alone. For weighting by way of a price change). Obtaining reliable data to make
purposes, a feasible approach is to obtain data from a sam- quality adjustments in response to changes in risk is prob-
ple of insurance providers, or from a regulatory authority, lematic, so in practice most indices reflect changes in risk as
facilitating estimation of the ratio of net insurance services a price change.
to gross premiums, and to apply this ratio to the estimated 11.398 In pricing insurance policies, the approach
value of gross premiums obtained, for example, from the should be to select a sample of policies representative of
HBSs. However, it has not been possible to devise a cor- those policies held in the base period and to reprice these
responding price measure that is conceptually sound and in subsequent periods. Taking dwelling insurance as an
can be accurately observed frequently enough from insur- example, base period insurance policies would be taken out
ance companies’ accounts. For this reason, those countries to insure dwellings of various values and types (for exam-
that have adopted the net concept for weighting purposes are ple, detached versus terraced house; timber or brick-built)
using movements in gross insurance premiums as a proxy in different locations. The price samples should therefore
price measure. consist of specifications that aim to cover, in aggregate,
as many combinations of these variables as is reasonable.
Acquisitions Approach While the conditions of the policy, the dwelling type, and
11.395 Under the acquisitions approach, all three location should be held constant over time, the value of the
policy types (as listed earlier) are in the scope of the dwelling should be updated each period to reflect changes
CPI. Because the objective is to measure price inflation in house prices and in rebuilding costs (that is, the underly-
for the household sector, the expenditure required for ing real quantity needs to be preserved). It is important to
weighting purposes should reflect the insurance com- note that, as the premiums will be related in some way to
panies’ contribution to the inflation process, which the value of the insured property, the price index for insur-
equates to the value of the insurance service as per the ance can change without there being any change in premium
use approach. schedules.
11.399 Efforts should be made to identify any changes
in the conditions applying to selected policies in order
Pricing Gross Insurance Premiums to facilitate appropriate quality adjustments. Examples
11.396 The gross insurance premium payable by would include cessation of coverage for specific condi-
households in any one period is determined by the con- tions and changing the excess (or deductible) paid by the
ditions of the policy, the administration costs and profit consumer when a claim is made. Estimates of the value
objectives of the insurance provider, the risk of a claim of such changes may be based on the insurance compa-
being made, and any relevant taxes. For any single policy, ny’s own assessments of their likely impact on the value
the principal quality-determining characteristics (generally of total claims payable. If it is assumed that the change in
specified in the conditions of the policy) can be summa- the aggregate value of claims can be equated to the change
rized as being: in service to the consumer (compared to the service that
• The type of property being covered (for example, dwell- would have been provided prior to policy renewal), then an
ing, motor vehicle) appropriate adjustment can be made to the premium to pro-
• The type of cover provided (for example, physical dam- vide a (quality-adjusted) movement in price. For example,
age, liability) consider the case where the excess on a policy is doubled
and advice from the company is that this will result in a
• The nature of the compensation (for example, replace- 3 percent drop in the aggregate value of claims payable.
ment cost, current market value) This could be considered as equivalent to a 3 percent
• Any limits on the amount claimable increase in price.
• The location of the property 11.400 In some cases, clients reap the benefits of a “no
• Amount of any excess payable by the insured claims” bonus where the premium is lower if no insurance
claim has been made over a period of years. Measuring
• Risks (or events) covered
the cost of a fixed basket of goods and services implies
11.397 Although it is clear that pricing to constant qual- that the price relatives should follow the price evolution of
ity requires these conditions to be held fixed, there is also identical products; that is, the product specification should
a question about whether the risk of a claim being made remain constant. The price changes recorded should reflect
should be held constant. In other words, if the incidence of, pure price changes; that is, the specifications of the insur-
for example, vehicle theft increases, should an increase in ance premiums should be held constant while reflecting
the premium paid be regarded as a quality improvement or in the price reference period the presence of “no claims”
a price change? If, on the one hand, it is argued that as the bonuses.
286
11.401 A combination of, for example, motor insur- Table 11.15 Illustration of the Impact of Taxes on
ance and dwellings insurance or health and travel insur- Measures of Insurance Services ($)
ance may be offered as a package at a cheaper price Period Premiums Tax Gross Claims Insurance
than separate purchases. Bundled products are generally Before Tax Premiums Service
required, if feasible in practice, to be separated and clas- 1 100.00 5.00 105.00 60.00 45.00
sified under the appropriate COICOP subclasses or classes 2 100.00 20.00 120.00 60.00 60.00
within the same or different COICOP groups or divisions.
However, COICOP, while showing its awareness of the
problem of bundling, does not normally provide clear
11.403 For the sake of simplicity, the example in Table
guidance on their classification. Outlays covering two or
11.15 assumes that there are no premium supplements and
more purposes are dealt with on a case-by-case basis with
no actuarial reserves. Then the insurance service charge is
the aim of obtaining a purpose breakdown that is as precise
given by gross premiums less provisions for claims. Sup-
as possible and consistent with practical considerations of
pose the only change between two periods is a change in
data availability.
the tax rate, from 5 to 20 percent of gross premiums. Then
the values in the following table are likely to be observed.
Using Gross Premiums as a Proxy for the Net Under this scenario, it turns out that the insurance service
Insurance Service charge has increased from $45 to $60 (an increase of 33.3
11.402 The net insurance service charge captures the
percent); yet gross premiums have only increased by 14.3
administration costs and profits of the insurance provider
percent.
along with any taxes and represents the amount paid for
11.404 Given that changes in the tax rates on gross
“peace of mind.” A problem arises from the fact that taxes
insurance premiums are often subject to significant varia-
on insurance are normally levied on the gross premiums.
tion, this can result in a volatile index. It is not a trivial
Namely, if the gross insurance premiums are subject to a
problem. A practical solution is to decompose insurance
high tax rate, then the taxes will account for an even higher
service into two components: insurance service before tax
proportion of the net insurance service charge. Using the
(or net of tax) and tax on insurance services. The price
gross insurance premium inclusive of taxes as the price
measure for the first is constructed by reference to move-
measure understates the effect, in relation to the service
ments in gross premiums net of tax, and the price mea-
charge, of any increase in tax rates. This can be illustrated
sure for the second is given by changes in taxes on gross
by the example in Table 11.15.
premiums.
287
Annex 11.1
Example of Price Collection Checklist for Second-Hand
Clothing
Figure 11A.1 Price Collection Checklist for Second-Hand Clothing
289
Annex 11.2
Example Price Collection Letter to Retailer
Figure 11A.2 Price Collection Letter to Retailer
291
Annex 11.3
Calculation of a Price Index
for a Deposit Product
(1) Base period sample account. Only a single month’s many accounts would be sampled with each account con-
data is used in this example (Table 11A.1). In practice, taining data for a full year.
Table 11A.1 Calculation of a Price Index for a Deposit Product: Base Period Sample Account
Date Debit (D) or Credit (C) Transaction Transaction Value ($) Tax ($) Balance ($)
02-Jan. D Over-the-Counter Withdrawal 107.05 0.70 348.48

12-Jan. C Deposit 4,000.00 2.40 4,346.08
13-Jan. D EFTPOS1 Transaction 50.62 0.30 4,295.16
13-Jan. D Over-the-Counter Withdrawal 371.00 0.70 3,923.46
14-Jan. D Own ATM2 Cash 300.00 0.70 3,622.76
14-Jan. D Own ATM Cash 100.00 0.70 3,522.06
16-Jan. D Own ATM Cash 100.00 0.70 3,421.36
16-Jan. D Check 90.00 0.30 2,959.36
19-Jan. D Own ATM Cash 100.00 0.70 2,858.66
19-Jan. D Own ATM Cash 100.00 0.70 2,757.96
19-Jan. C Deposit 4,000.00 2.40 6,755.56
19-Jan. D Check 740.00 1.50 6,014.06
20-Jan. D EFTPOS Transaction 76.42 0.30 5,937.34
21-Jan. D Other ATM Cash 20.00 0.30 5,917.04
21-Jan. D Check 100.00 0.70 5,816.34
22-Jan. D Check 43.40 0.30 5,772.64
22-Jan. D Check 302.00 0.70 5,469.94
22-Jan. D Check 37.00 0.30 5,432.64
23-Jan. D Check 72.00 0.30 4,988.64
27-Jan. D Own ATM Cash 150.00 0.70 4,837.94
27-Jan. D Check 73.50 0.30 4,764.14
27-Jan. D Check 260.00 0.70 4,503.44
28-Jan. D Check 150.00 0.70 4,280.74
29-Jan. D Check 140.00 0.70 4,140.04
30-Jan. D Check 8.00 0.30 3,760.04
30-Jan. D Check 60.00 0.30 3,699.74
Total Taxes: 21.10
Fees
Activity Total Number Number Charged Amount ($)
Over-the-Counter Withdrawal 6 2 6.00

EFTPOS Transaction 3 0 0.00
Own ATM Cash 6 0 0.00
Check 13 3 3.00
Deposit 2 2 0.00
Total Fees: 10.20
1 EFTPOS, Electronic funds transfer point of sale.
2 ATM, Automatic teller machine. Table 11A.2 Calculation of a Price Index for a Deposit
Source: Woolford (2001). Product: Fee Schedule
Description Base Period Current Period
Number Free Charge ($) Number Free Charge ($)
Over-the-Counter 4 3.00 4 3.00

(2) Fee schedule. Table 11A.2 is a summary of the infor- Withdrawal
EFTPOS 10 0.50 9 0.50
mation typically available from financial institutions. For Transaction
each period, the table includes the number of free transac- Own ATM Cash 10 0.50 9 0.50
tions and the per transaction charge for additional transac- Other ATM Cash 0 1.20 0 1.20
tions. A zero number of free transactions indicates that no Check 10 1.00 9 1.00
Deposit 0 0.00 0 0.00
transactions are free and a zero charge indicates that all
transactions are free. Source: Woolford (2001).
293
(3) Tax schedule. Tables 11A.3 and 11A.4 present an Table 11A.3 Calculation of a Price Index for a Deposit
example of the types of taxes applied to a bank account. Product: Bank Accounts Debit Tax
For example, the debits tax is levied on all debit transac- Transaction Value ($) Tax ($)
tions to eligible accounts, with the amount charged being set
Min. Max. Base Period Current Period
for ranges of transaction values (that is, using a step func-
tion). Financial institutions’ duty is levied on all deposits, 0 1 0.00 0.00
the amount being determined as a percentage of the value 1 100 0.30 0.30
100 500 0.70 0.70
of the deposit. 500 5,000 1.50 1.50
(4) Interest data. Table 11A.5 presents, in summary form, 5,000 10,000 3.00 3.00
the balances and annualized interest flows derived by taking 10,000+ 4.00 4.00
moving averages of data reported by financial institutions. Source: Woolford (2001).
Interest rates and margins are calculated from the balances
and flows. Table 11A.4 Calculation of a Price Index for a Deposit
Product: Financial Institutions Duty (percent)
Base Period Current Period
0.06 0.06
Source: Woolford (2001).
Table 11A.5 Calculation of a Price Index for a Deposit Product: Interest Data
Base Period Current Period
Balance Interest Interest rate Margin Balance Interest Interest rate Margin
($ million) ($ million) (percent) (percent) ($ million) ($ million) (percent) (percent)
Deposit Products
Personal 22,000.00 740.00 3.364 2.494 23,600.00 775.00 3.284 2.397
Current Accounts 6,000.00 68.00 1.133 4.724 6,600.00 75.00 1.136 4.545
Other accounts 16,000.00 672.00 4.200 1.657 17,000.00 700.00 4.118 1.563
Business Accounts 25,000.00 920.00 3.680 2.177 28,000.00 1,000.00 3.571 2.110
Total Deposit Accounts 47,000.00 1,660.00 3.532 2.326 51,600.00 1,775.00 3.440 2.241
Loan Products
Personal 42,000.00 3,188.00 7.591 1.733 46,000.00 3,400.00 7.391 1.710
Business 28,000.00 2,540.00 9.071 3.214 31,000.00 2,700.00 8.710 3.029
Total Loan Accounts 70,000.00 5,728.00 8.183 2.326 77,000.00 6,100.00 7.922 2.241
Reference Rate 5.857 5.681
(5) CPI data. Table 11A.6 presents data required to derive Table 11A.6 Calculation of a Price Index for a Deposit
the indexation factor. This example follows the Australian Product: CPI Data
practice of a quarterly CPI. If a monthly CPI is produced, t−5 t−4 t−3 t−2 t−1
12-term moving averages would be required.
All Groups 117.5 121.2 123.4 127.6 129.1
(6) Projected current period sample account. Table 11A.7 Four-Term Moving 122.4 125.3
shows that the opening balance and transaction values are Average
derived by applying the indexation factor to the base period Indexation Factor 1.0237
amounts. The tax payable is determined by reference to the (movement)
data in Table 11A.3. Fees payable are determined by refer- Source: Woolford (2001).
ence to the data in Table 11A.2.
294
Table 11A.7 Calculation of a Price Index for a Deposit Product: Projected Current Period Sample Account
Date Debit (D) or Credit (C) Transaction Transaction Value ($) Tax ($) Balance ($)
02-Jan. D Over-the-Counter Withdrawal 109.59 0.70 356.75

12-Jan. C Deposit 4,094.75 2.46 4,449.05
14-Jan. D Own ATM Cash 307.11 0.70 3,708.63
14-Jan. D Own ATM Cash 102.37 0.70 3,605.56
16-Jan. D Own ATM Cash 102.37 0.70 3,502.50
16-Jan. D Check 92.13 0.30 3,029.57
19-Jan. D Own ATM Cash 102.37 0.70 2,926.51
19-Jan. D Own ATM Cash 102.37 0.70 2,823.44
19-Jan. C Deposit 4,094.75 2.46 6,915.73
19-Jan. D Check 757.53 1.50 6,156.70
21-Jan. D Other ATM Cash 20.47 0.30 6,057.40
21-Jan. D Check 102.37 0.70 5,954.33
22-Jan. D Check 44.43 0.30 5,909.60
22-Jan. D Check 309.15 0.70 5,599.75
22-Jan. D Check 37.88 0.30 5,561.57
23-Jan. D Check 73.71 0.30 5,107.08
27-Jan. D Own ATM Cash 153.55 0.70 4,952.83
27-Jan. D Check 75.24 0.30 4,877.28
27-Jan. D Check 266.16 0.70 4,610.43
28-Jan. D Check 153.55 0.70 4,382.48
29-Jan. D Check 143.32 0.70 4,238.46
30-Jan. D Check 8.19 0.30 3,849.49
30-Jan. D Check 61.42 0.30 3,787.77
Total Taxes: 21.21
Fees
Activity Total Number Number Amount ($)
Charged
Over-the-Counter Withdrawal 6 2 6.00

EFTPOS Transaction 3 0 0.00
Check 13 4 4.00
Deposit 2 2 0.00
Total Fees 11.20
(7) Indices for current accounts. Table 11A.8 brings the Table 11A.8 Calculation of a Price Index for a Deposit
results together. The current period value aggregates are Product: Indices for Current Accounts
derived as follows. For margins, the base period aggregate Component Base Period Current Period
is multiplied by the product of the indexation factor (5) and
Value Index Value Index
the ratio of the current and base period margins for current Aggregate ($) Aggregate ($)
accounts (4). For fees, the base period aggregate is multi-
plied by the ratio of total fees payable on the sample account Margins 28,344 100.00 27,913 98.50
Fees 11,904 100.00 13,071 109.80
in the current period (6) and the base period (1). For taxes, Taxes 14,739 100.00 14,818 100.50
the same procedure is followed as for fees. Total 54,987 100.00 55,803 101.50
295
ERRORS AND BIAS 12
Introduction • For each product group, a sample of items to calculate the
partial price index of the product group
12.1 The consumer price index (CPI), like all other sta- • For each item, a sample of outlets to calculate the ele-
tistics, may be subject to general error that may occur dur- mentary price index of the item from individual price
ing any stage of the estimation process but also errors that observations
are unique to the CPI (for example, substitution bias and
quality change bias). This chapter first describes the general • For each product group, a sample of a day or a time span
types of potential errors and the sources of sampling and of the month when the data collection has to be carried out
nonsampling error that arise in estimating a population CPI (concerning this issue, the introduction of scanner data,
from a sample of observed prices, and then reviews the argu- which in general cover more than one week of a month,
ments made in numerous studies that attribute bias to CPIs in CPI compilation could reduce the potential errors aris-
as a result of not properly addressing the treatment of quality ing in the traditional data collection for this dimension of
change, consumer substitution, and other factors. It should sampling)
be emphasized that many of the underlying issues discussed • A sample of households needed for the estimation of the
in this chapter are dealt with in much greater detail else- average expenditure shares of the item groups (some
where in the Manual. countries use alternative sources of data, such as national
12.2 The CPI is subject to various types of errors and accounts, instead of a household budget survey [HBS] to
biases that affect the precision and accuracy of the CPI esti- obtain the expenditure shares, as described in Chapter 3)
mates. Several potential sources of errors and bias have been Sampling error can be introduced at any of the stages of the
identified in the CPI and addressed, though the debate con- sample selection process. The potential for sampling error is
tinues over to what extent and in what direction bias may greater in the selection of outlets and even more so for prod-
still exist and the ways in which its accuracy can continue ucts because there is no comprehensive frame from which to
to be increased. select units for sampling.
12.5 The sampling error can be split into a selection
Types of Errors error and an estimation error. A selection error occurs when
the actual selection probabilities deviate from the selection
12.3 One of the main objectives of a sample survey is to probabilities as specified in the sample design. The estima-
compute estimates of population characteristics. Such esti- tion error denotes the effect caused by using a sample based
mates will never be exactly equal to the population charac- on a random selection procedure. Every new selection of a
teristics. There will always be some error, and the precision sample will result in different elements, and thus in a pos-
and accuracy of the estimate is affected by both sampling sibly different value of the estimator.
and nonsampling error. Table 12.1 gives a taxonomy of the
different types of errors.1 Two broad categories can be dis-
tinguished: sampling errors and nonsampling errors. Nonsampling Error
12.6 CPI surveys involve many operations, all of which
Sampling Error are potential sources of nonsampling error. The nonsam-
pling errors arise from the survey process, regardless of
12.4 Sampling errors are due to the fact that an estimated whether the data are collected from the entire universe or
CPI is based on samples and not on a complete enumera-
tion of the populations involved. Sampling errors vanish if
observations cover the complete population. As mentioned
in previous chapters, national statistical offices (NSOs) usu- Table 12.1 A Taxonomy of Errors in a CPI
ally adopt a fixed-weight price index as the object of estima- Total Error:
tion. A fixed-weight index is a weighted average of partial Sampling Error
indices of product groups, with weights being expenditure Selection Error
Estimation Error
shares. The estimation procedures that most NSOs apply to Nonsampling Error
a CPI involve different kinds of samples. The most impor- Observation Error
tant kinds are the following: Overcoverage
Response Error
Processing Error
1 Nonobservation Error
See also Balk and Kersten (1986) and Dalén (1995) for overviews of the Undercoverage
various sources of stochastic and nonstochastic errors experienced in cal- Nonresponse
culating a CPI.
297
from a sample of the population. They can be subdivided include new in-scope units or to exclude mail order firms
into observation errors and nonobservation errors. Observa- and nonfood market stalls from their outlet sampling frame.
tion errors are the errors made during the process of obtain- Undercoverage means that some outlets where relevant
ing and recording the basic observations or responses. The items are purchased cannot be contacted.
most general categories of observation errors are overcover- 12.11 Nonresponse is another category of nonobserva-
age, response error, and processing error. tion error. Nonresponse errors may arise from the failure
12.7 Overcoverage means that some elements are to obtain the required information in a timely manner from
included in the survey that do not belong to the target pop- some of the units selected in the sample. A distinction can
ulation or target universe. For outlets, NSOs usually have be drawn between total and partial (or item) nonresponse.
inadequate sampling frames. For example, in some coun- Total nonresponse occurs when selected outlets cannot
tries, a business register is used as the sampling frame for be contacted or refuse to participate in the price survey.
outlets, where outlets are classified according to the main Another instance of total nonresponse occurs when mail or
activity. The register thus usually exhibits extensive over- electronic questionnaires and collection forms are returned
coverage, because it contains numerous outlets which are by the respondent and the price collector, respectively, after
out of scope from the CPI perspective (for example, firms the deadline for processing has passed. Partial (item) nonre-
that sell to businesses rather than to households). In addi- sponse occurs when a responding unit does not complete the
tion, there is usually no detailed information on all the items information on an item or items on the survey questionnaire,
sold by an outlet, so it is possible that a sampled outlet may or the responses obtained are unusable. Mail or electronic
turn out not to sell a particular item at all. questionnaires and collection forms that are only partially
12.8 Response error results from the collection of incor- filled in, scanner data with missing information concerning
rect, inconsistent, or incomplete data. Response error may specific outlets or Global Trade Item Numbers (GTINs) in
arise because of the collection of data from inappropriate the sample, and web-scraped prices where some information
respondents, deliberate distortion of responses, interviewer is not downloaded from the internet are examples of partial
effects, misrecording of responses, pricing of wrong items, nonresponse. If the price changes of the nonresponding out-
misunderstanding or misapplication of data collection pro- lets differ from those of the responding outlets, the quality
cedures, misunderstanding of the questions or survey needs, of the price change estimates will be affected.
and lack of cooperation from respondents. In price surveys, 12.12 Another source of errors is the failure to mea-
where the main price collection method is by price collec- sure the price actually paid. This failure may be caused, for
tors who regularly visit outlets, they may collect prices of example, by the use of list prices (for example, for cars) and
unwanted items. by the presence of discounts, coupons, or bargaining, which
12.9 Processing error occurs after the survey data are are typically not accounted for or difficult to measure. In
collected, during the processes that convert reported data many countries, the discounting of prices is becoming more
to published estimates and consistent machine-readable common and the importance of discounted prices is increas-
information. Each of the processing steps, such as coding, ing. Another source of error is due to the tendency of price
data entry, transfer, and editing (control and correction), can collectors to choose an excessive proportion of regularly
generate errors. For example, at the outlets, the price col- priced varieties in the price reference period, whereas the
lectors write down the prices on paper forms or use dedi- proportion of sales prices increases later and becomes pro-
cated software on a tablet or handheld computer. In the first portional to their real share later in the year.
case (paper-and-pencil data collection), after the collectors
have returned home, a computer is used as the input and
means of transmission for the price information. This way
Measuring Error
of processing prices is susceptible to errors. The second case
(computer-assisted data collection) is less risky because it Estimation of Variance
has built-in validation checks, but it could be susceptible to 12.13 The variance estimator depends on both the cho-
sen estimator of a CPI and the sampling design. The 2012
errors for other reasons such as the lack of adequate controls
International Labor Organization Survey of country prac-
during the recording of prices. Processing error also fails to
tices2 gives an overview of the sampling methods that were
identify true errors during regular micro- and macro-edit-
applied in the compilation of CPIs by NSOs. It found that
ing. Even when errors are discovered, they can be corrected
only one out of three NSOs use some sort of probability
improperly because of inadequate imputation and quality-
techniques for location selection, 1 in 5 for outlet selec-
adjustment procedures. The occurrence of processing errors
tion, and only 1 in 10 uses probability sampling for item
is strongly influenced by survey planning, and to some
selection. In the absence of probability techniques, so-called
extent the survey’s resources (for example, staff and budget,
judgmental and cutoff selection methods are applied.
devices, and training) and constraints (for example, elapsed
12.14 In view of the complexity of the sample designs
time between data collection and publication).
in compiling a CPI (where the samples of locations, outlets,
12.10 Nonobservation errors are made when the
items, and varieties are just partially connected), an inte-
intended measurements cannot be carried out. The most
grated approach to variance estimation can be problematic.
general categories of nonobservation errors are undercov-
Therefore, it appears to be difficult to present a single for-
erage and nonresponse error. Undercoverage occurs when
mula for measuring the variance of a CPI, which captures all
elements in the target population are not included in the sam-
pling frame used for sample selection. The source of under-
coverage error is the sampling frame itself. For instance, it 2
https://www.ilo.org/stat/Areasofwork/WCMS_180241/lang--en/index.
is likely that there are delays in updating the outlet frame to htm.
298
Errors and Bias
sources of sampling error. It is, however, feasible to develop treated in the survey research literature. The occurrence of
partial (or conditional) measures, in which only the effect of processing errors can be reduced through survey process
a single source of variability is quantified. For instance, Balk improvements. Several categories of nonsampling errors
and Kersten (1986) calculated the variance of a CPI result- provide the bulk of the bias issues discussed in paragraphs
ing from the sampling variability of the HBS, conditional on 12.30–12.73.
the assumption that the partial price indices are known with
certainty. Ideally, all the conditional sampling errors should Procedures to Minimize Errors
be put together in a unified framework to assess the rela-
tive importance of the various sources of error. Under rather 12.18 The estimation error can be controlled and mini-
restrictive assumptions, Balk (1989a) derived an integrated mized by means of the sampling design. For example, by
framework for the overall sampling error of a CPI. increasing the sample size, or by taking selection probabili-
12.15 There are various procedures for estimating the ties proportional to some well-chosen auxiliary variable, the
sampling variance arising from a probability sampling error in the estimated CPI can be reduced. The choice of
design. For instance, assuming a cross-classified sampling an adequate sampling design for the CPI is an extremely
design in which samples of items and outlets are drawn complex matter (see Dorfman and others [2006]). The tar-
independently from a two-dimensional population, with get population is the set of all goods and services that are
probabilities proportional to size (PPS) in both dimensions, acquired, used, or paid for by households from outlets in a
a variance formula can be derived. Where an overall esti- particular period. A proper probability sampling procedure
mate of the sampling variance cannot be made, at the very selects a sample by a random mechanism in which each
least, a basic analysis should be conducted. good or service in the population has a known probabil-
12.16 The main problem with nonprobability sampling ity of selection. In combination with a Horvitz–Thompson
is that there is no theoretically acceptable way of knowing estimator,3 such a probability sampling design will produce
whether the dispersion in the sample data accurately reflects an index that is (approximately) unbiased and precise.
the dispersion in the population. It is then necessary to rely 12.19 The probability sampling designs used exten-
on approximation techniques for variance estimation. One sively in survey practice are simple random sampling and
such technique is quasi-randomization (see Särndal and oth- PPS sampling, with or without some form of stratification
ers [1992, 574]), in which assumptions are made about the (more details are provided in Chapter 4). The advantage
probabilities of sampling items and outlets. The problem of simple random sampling is its simplicity; it gives each
with this method is that it is difficult to find a probability population element the same probability of being included
model that adequately approximates the method actually in the sample. PPS sampling has the advantage that the more
used for outlet and item selection. Another possibility is important elements have a larger chance of being sampled
to use a replication method, such as the method of random than the less important ones. For instance, in one European
groups, balanced half-samples, jackknife, or bootstrap. This NSO, the outlets are selected with probabilities proportional
is a completely nonparametric class of methods to estimate to some proxy for size, namely their number of employ-
sampling distributions and standard errors. Each replication ees. Unequal probability designs can lead to a substantial
method works by drawing a large number of subsamples variance reduction in comparison with equal probability
from the given sample. From each subsample, the parameter designs. In stratified sampling, the population is divided into
of interest can be estimated. Under rather weak conditions, nonoverlapping subpopulations called strata. For instance,
it can be shown that the distribution of the resulting esti- in another European NSO, the population of outlets is split
mates approximates the sampling distribution of the original into two outlet types (chain and independent) to form differ-
estimator. For more details on the replication methods, see ent strata by region. In each stratum, a sample is selected by
Särndal and others (1992, 418–445). PPS sampling or simple random sampling. One of the rea-
sons why stratification is so popular is that, if strata are well
Qualitative Assessment of Nonsampling Errors constructed, it results in low variance of the price changes
12.17 As estimating the quantitative impact of the non- within a stratum. Stratification is a useful strategy to make
sampling errors is more difficult, a qualitative assessment the sample more efficient.
should be provided. For instance, the coverage of the sam- 12.20 Because appropriate sampling frames are gener-
pling frames as a proxy of the target populations can be ally not available, samples are frequently obtained by non-
described and provided (including gaps, duplications, and probability methods. Judgmental (or purposive) sampling
definitional problems). The percentage of the target outlet is one form of nonrandom selection. In this case, some-
samples from which responses or usable price data were one with knowledge on where households make expendi-
obtained (that is, the response rates) can be provided. Any ture (for example, data collector) selects certain “typical”
known difference in the prices of responding outlets and non- locations and outlets where data are to be collected. With
responding outlets can be described, as can an indication of their knowledge, a fairly good sample might result. A more
the method of imputation or estimation used to compensate sophisticated nonprobability method is quota sampling. In
for nonresponse. Other examples of qualitative measures quota sampling, the population is first divided into certain
of nonsampling errors are indicators such as implicit qual- strata. For each stratum, the number (quota) of locations and
ity indices, which compare indices with and without prices outlets to be included in the sample is fixed. Next, the price
adjusted for quality changes. Similarly, the effects of edit-
ing can be measured by comparing the CPI estimates based 3
D. G. Horvitz, and D. J. Thompson. 1952. “A Generalization of Sampling
on edited and nonedited data sets. As processing errors tend without Replacement from a Finite Universe.” Journal of the American
not to be well-reported or well-documented, they are seldom Statistical Association 47: 663–85. JSTOR 2280784
299
collector in the field simply fills the quotas, which means in outlets. Often, they include outlets that are no longer in busi-
the case of outlet sampling that the selection of the outlets ness or have changed the activity. Other sources, described
is ultimately based on the judgment of the price collectors. in more detail in Chapter 4, can be used as a sampling frame.
Another nonprobability method is cutoff sampling, which In the absence of any reliable source of data that can be used
means that a part of the target population is deliberately as a sampling frame, it is recommended to set up an appro-
excluded from the sample selection process. This procedure priate sampling frame by enumeration of the main outlets
is used when the distribution of the value of some auxiliary within each sampled location. Such enumeration yields a list
variable is highly skewed. For instance, a large part of the of all outlets in a location together with the item groups that
population may consist of small outlets whose contribution belong to their assortments. When formal sampling tech-
to total sales is modest. A decision may then be taken to niques cannot be applied, outlets can be selected using judg-
exclude from the sampling frame the outlets with the lowest mental methods. For example, a more judgmental approach
sales. Because the selection is nonrandom, nonprobability to organizing an outlet sampling frame is to ask the price
methods usually lead to biased estimates. Empirical results collectors—who may be assumed to know the local situ-
of research presented in de Haan and others (1997) never- ation well—to make a list of outlets where purchases are
theless show that nonprobability selection methods do not made by households. It is important that information about
necessarily perform worse, with regard to the mean square the quality of the sampling frame, with regard to overcover-
error, than probability sampling techniques. age or undercoverage, and its completeness for the target
12.21 Provided that the sampling design is given, the population is known.
sampling variance of an estimated (all-items) CPI can in 12.26 The populations of items (and varieties) and out-
general be lowered by: lets are continually changing through time. The composition
• Enlarging the samples of items and outlets
of most item groups is not constant over time, because items
disappear from the market and new ones appear. The outlet
• The application of suitable stratifications to the various population also changes over time: outlets close, temporar-
populations (for example, grouping items with respect to ily or permanently; new outlets open; and the importance of
similarity of price changes) some outlets diminishes or increases. The samples of items
12.22 It is important to allocate optimally the available (and varieties) and outlets should be reviewed and updated
resources both between and within the different CPI sam- periodically to maintain their representativity with respect to
ples, since badly allocated samples may lead to unnecessar- the current expenditure patterns of the households. In many
ily high sampling errors. Dalén and Ohlsson (1995) show countries, these are reviewed and updated every year.
that the error resulting from item sampling is relatively high 12.27 Measurement errors by price collectors can be
compared with the error resulting from outlet sampling. reduced by providing them with handheld computers or
In this case, it is worthwhile increasing the sample size of tablets for data entry that have integrated validation checks.
items and reducing the sample size of outlets. Beisteiner In this way, the validation and editing of observed prices
(2008) stresses the importance of allocating resources to can be executed at the point of price collection (that is, in
those areas where the effect on the quality of the all-items the outlet) by comparing the currently observed price quote
CPI is maximized, especially to goods and services with a with the previously observed one (by setting a limit on the
high relative expenditure weight and to goods and services percentage price change) and with the price quotes obtained
with high dispersion of prices. The paper presents a “ready- from other outlets (by setting suitable upper and lower lim-
to-use” formula, the Neyman formula, for the allocation of its). Details are provided in Chapter 5 on the use of handheld
the sample, which optimizes the precision of the CPI for computers and tablets for price collection. Although using
given resources, as discussed in Chapter 4. pricing forms that contain information on the previous peri-
12.23 A systematic analysis of sampling errors offers pos- od’s price can reduce response variance, it may also cause
sibilities for improving efficiency or reducing cost. The prob- response bias and delay in reporting price change. Before
lem of optimum sample allocation is usually formulated as the introducing handheld computers and tablets, proper usabil-
determination of the sizes of the samples of items and outlets, ity testing and training for price collectors are required to
and their distribution over the strata that minimizes the sam- avoid them being a source of error.
pling error of an all-items CPI, subject to the available budget. 12.28 It is useful to appoint data collection supervi-
12.24 The accuracy of the CPI could be improved by sors to conduct quality assurance checks on the price col-
making use of scanner data, which collect more prices for lectors. It is also a good idea to organize regular meetings
more varieties on more days of the month than traditional where price collectors and CPI compilers from the head
data collection methods. Bradley (1996) discusses the office can share their experiences. In this way, the compilers
potential for scanner data to reduce the sampling error of will keep in touch with the conditions in the field and may
the corresponding official CPI component index. The use of take the opportunity to provide more information about fre-
scanner data also has a positive effect on the time dimension quently made price collection errors and new representative
of sampling, covering a time span much longer than the one products.
covered by the traditional data collection. In Chapter 10, it 12.29 It is important to check the collected price data
is argued that scanner data should cover the whole period for for processing errors and, where possible, to correct these
which CPI is constructed, rather than a subperiod. In some errors. This activity is called data editing. The first stage
cases, the use of scanner data removes the need for sampling of editing includes the review and validation of individual
as a census of products can be used. observations. When the resources to spend on data editing
12.25 As already mentioned, a business register can be must be minimized, while at the same time maintaining a
subject to overcoverage when used as a sampling frame for high level of data quality, selective editing and a broad review
300
Errors and Bias
of the compiled data are possibilities. Selective editing is a implications for the measurement of recent trends in output
form of traditional micro-editing in which the number of and productivity, and second, that the elimination of upward
edits is kept to a minimum. Only those edits that have an bias could substantially improve the government budget
impact on the survey results are carried out. A review of the situation through reduced government expenditure and
compiled indices offers a top-down approach. The edits are increased tax revenue (see, for example, Eldridge [1999]
carried out on aggregated data (for example, the price index and Duggan and Gillingham [1999]). These findings led to a
numbers of an item group) instead of individual records (for series of papers and reports on CPI measurement problems,
example, price observations). A review of individual records often accompanied by point estimates of aggregate bias.
is then carried out only if the top-down review raises suspi- 12.33 One of the most prominent examples of these quan-
cion. Attention should particularly be paid to outliers among titative studies of bias is that by the Advisory Commission to
the observations (more information on data editing and the Study the CPI (US Senate 1996).4 Responses and estimates
use of algorithms is provided in Chapter 5; a comprehensive by statistical agencies include those provided by Abraham
description of statistical data editing procedures is given in and others (1998), US Bureau of Labor Statistics (1998),
De Waal and others [2011]). Ducharme (1997), Edwards (1997), Fenwick (1997), John-
12.30 Nonresponse reduces sample size, results in son and others (2006), Lequiller (1997), Moulton (1996b),
increased variance, and usually introduces selection bias. and Moulton and Moses (1997). Research undertaken has
Nonresponse rates, or missing observations, are often shown that it is difficult to both quantify and assess the direc-
viewed as a proxy for the quality of a survey. While nonre- tion of potential bias, and that the extent, the direction, and
sponse rates are important, imputation rates alone provide no even the existence of bias remain something that will depend
indication of nonresponse bias. There are three methods for upon the specific circumstances of each set of CPI estimates
the treatment of missing price observations. First, the cor- and cannot always be determined with certainty.
responding price can be excluded from the data set of previ- 12.34 Two points are worth making at the outset with
ous period prices, so that the set of previous period prices is respect to measuring bias in CPIs. First, the issue has usu-
“matched” with the set of current prices. Second, this match- ally been addressed in the context of the cost of living index
ing can be achieved by using an imputed (or artificial) price (COLI). That is, the CPI bias has been defined as the dif-
for the missing one. The imputed price can be calculated ference between the rate of increase in the CPI and the rate
by either carrying forward the previous price observation of increase in a true COLI. Many discussions on bias have
or by extrapolating the previous price observation using the taken as given that the COLI should be the CPI’s measure-
change of other price observations for the same item. Third, ment objective. Somewhat different conclusions might be
there is the possibility to reweight the sample to minimize reached if the index objective was taken to be a fixed-basket
the effect of nonresponse error. The objective of reweighting price index. Notably, the gains in consumer welfare from a
is to inflate the weight given to the prices of the responding widening array of new products, or the ability of consum-
outlets. This compensates for those prices that are lost by ers to substitute away from items with increasing relative
nonresponse (for details, including advantages and disad- prices, might be deemed irrelevant and an index that ignored
vantages of each approach, see Chapter 6). those factors might not be judged biased on that account.
12.35 The second point is that CPI bias is not amenable
to estimation with the same level of rigor as that used in
Types of Bias CPI variance estimation. Since the COLI or other ideal tar-
12.31 Bias is defined as a systematic tendency for the get index is unobserved, analysts have been forced to rely
calculated CPI to diverge from some ideal or preferred in part on conjectures and on generalizations from fragmen-
index, resulting from the method of data collection or pro- tary empirical evidence to quantify the extent of bias. The
cessing, or the index formula used. This section reviews notable exception is with respect to substitution bias, when
several categories of systematic error, either in pricing or indices using superlative formulas can be computed using
in index construction, that potentially can lead to bias in the the same underlying price and expenditure data and com-
all-items CPI. The emphasis here is on the categorization pared with historical CPI data to estimate the upward bias
of different types of bias, along with some consideration of from use of the traditional formulas.
their likely size, but also on methods to reduce or eliminate 12.36 Several different taxonomies of bias have
these categories of bias. The question might arise of why appeared in the literature mentioned previously. It is suf-
such a discussion is necessary, since such issues as quality ficient, however, to employ four categories roughly corre-
change, and the appropriate methods for handling them in sponding to those set forth in the best-known study, namely
the CPI, are dealt with at both a conceptual and operational the Final report of the Advisory Commission to Study the
level in other chapters (see Chapter 8 of the publication CPI (the Boskin Commission), established by the US Sen-
Consumer Price Indices Theory). ate Finance Committee in 1995. These categories are upper-
12.32 The reason why this chapter addresses the topic of level substitution bias; elementary aggregate bias; quality
CPI bias is the great surge in interest in price measurement change and new goods bias; and new outlet bias.
problems during the mid-1990s. Especially in the United 12.37 These categories can be further broken down into
States (US), the view became widespread that the CPI was two subgroups according to whether they refer to errors in
subject to systematic upward biases because of the failure
to deal adequately with product substitution by consumers, 4
Others include Congressional Budget Office (1994), Crawford (1998),
product quality improvements, and the introduction of new Cunningham (1996), Dalén (1999a), Diewert (1996c), Lebow and others
goods and services. Moreover, it was recognized, first, that (1994), Lebow and Rudd (2003), Shapiro and Wilcox (1997b), Shiratsuka
the existence of such upward bias would have fundamental (1999), White (1999), and Wynne and Sigalla (1994).
301
individual price measurements or errors in computing index estimated exact COLIs based on demand system estimation.
series. Quality change bias and new goods bias arise because A similar estimate for the Netherlands is provided by Balk
of failures to measure adequately the value to consumers of (1990). In these studies, and in the more recent analyses
individual goods and services that appear in (or disappear of US CPI data by Shapiro and Wilcox (1997a) and Cage
from) the marketplace. It should be recognized that discus- and others (2003), the existence of an upward bias from the
sions of “new goods” problems apply equally to all prod- Laspeyres formula is demonstrated consistently. The biases
ucts, whether goods or services. At a conceptual level, it can in the annual index changes in individual years are relatively
be difficult to distinguish these two biases from each other. small, typically 0.3 percentage points or less, and depend
Operationally, however, quality change bias pertains to the empirically on such factors as the distance from the Laspey-
procedures for comparing new products or models with the res base period, the level of index detail at which the alterna-
older products they replace in the CPI samples. In general, tive formulas are applied, and whether the superlative index
new goods bias can be thought of as applying to entirely is of the fixed base or chained variety.
new types of products, or products that would not enter sam- 12.41 The major differences between Laspeyres and
ples routinely through forced replacement. New outlet bias, superlative indices rise from the variation in the rela-
sometimes referred to as outlet substitution bias, is similar tive prices over the period being compared, and from the
to new goods bias but is focused on the appearance of new shift in quantities consumed toward those index catego-
types of outlets or marketing methods that offer goods and ries that have fallen in relative price. This leads to several
services at lower prices or higher quality. conclusions:
12.38 The other categories of bias refer to the proce- • If index movements are characterized by continuing, uni-
dures for constructing index values from component series. form drift in relative prices over time, with accompanying
As noted throughout this Manual, CPI compilation can be drifts in consumption, the size of the annual Laspeyres
thought of as taking place in two steps, or at two levels. At bias will tend to increase with the distance from the base
the lower level, individual price quotations are combined; period. The estimates of the upper-level substitution bias
at the upper level, these elementary indices are aggregated presented in Australian Bureau of Statistics (2017) show
together. Corresponding to these two levels are two forms that average annual substitution bias is 0.11 one year after
of potential bias. Elementary aggregate bias involves the a reweight of the CPI, increasing to 0.20 in the sixth year.
averaging formulas used to combine price quotations into Greenlees (1997) notes that there is little evidence for this
elementary indices. Upper-level substitution bias applies to phenomenon in the United States; see also Szulc (1983).
the formulas used to combine those elementary aggregates
• Under the same circumstances, reducing the expenditure
into higher-level indices. These components of potential
bias, and the means used to measure them, are discussed in weight chaining interval will work to reduce the upper-
more detail in paragraphs 12.38–12.72. level substitution bias in the Laspeyres-type CPI. The
more frequent chaining will increase the weight given to
indices that are falling in relative price, thereby reducing
Components of Bias the rate of CPI growth. Conversely, if there is “bouncing”
in relative index movements, frequent chaining can lead
Upper-Level Substitution Bias to an upward “chain drift” in a Laspeyres index.
12.39 Upper-level substitution bias is perhaps the most
widely accepted source of CPI bias, and the kind with which • Upper-level substitution bias will tend to be larger dur-
economists are most familiar from literature on price index ing periods of higher inflation, if these periods also have
theory and practice. Simply stated, it arises when CPIs greater relative price variation. However, little empirical
employ the Laspeyres formula, which is well known to pro- evidence exists on this point.
vide an upper bound on a COLI under certain assumptions 12.42 The concept of upper-level substitution bias has
about consumer behavior (see Chapter 4 of the publication been derived and discussed in the context of COLI theory,
Consumer Price Index Theory), or a similar method that but an equivalent bias may be defined from the perspec-
uses a fixed-base or -basket index, like the Lowe and Young tive of the fixed-basket price index. If the Fisher ideal or
formulas. The assumption in the definition of the Laspeyres- other superlative index is judged preferable based on its
type price index is that substitution among goods is zero, symmetric treatment of base period and current period
which is against one of the cornerstones in the theory of con- expenditure patterns, then the difference between that index
sumer demand. Quantitative measures of upper-level substi- and a Laspeyres could be interpreted as a measure of rep-
tution bias can be generated by comparing Laspeyres-type resentativity bias. A similar argument could be applied with
price indices to Fisher ideal, Törnqvist, or other superlative respect to lower-level substitution bias within elementary
indices. Under certain assumptions about, for example, con- index aggregates.
stant preferences, these will stand as relatively precise bias 12.43 Lebow and Rudd (2003) have defined and esti-
estimates. mated another category of bias related to upper-level aggre-
12.40 Genereux (1983) and Aizcorbe and Jackman gation. They concluded that the HBS weights used in the
(1993) provide such index comparisons and estimates of United States CPI were subject to error because of, for
upper-level substitution bias using actual CPI series for example, underreporting of alcohol and tobacco expendi-
Canada and the United States, respectively. Other early stud- tures. This will lead to a weighting bias if the errors in rela-
ies by Braithwait (1980) and Manser and McDonald (1988) tive weight are correlated with component index changes
estimate the substitution bias in US national account indi- (sources for and problems with expenditure weight estima-
ces. Instead of superlative indices, the Braithwait study uses tion are discussed in detail in Chapter 3).
302
Errors and Bias
Elementary Aggregate Bias when bilateral superlative price indices are used. The high-
12.44 Elementary aggregate bias arises from the use of frequency chaining, used to handle the high attrition rate of
an inappropriate method for aggregating price quotations items, can create drift in the index series when prices and
at the lowest level of aggregation. An elementary index in quantities change or bounce arising from sales. Therefore,
the CPI is biased if its expectation differs from its measure- new methods have been developed for price measurement
ment objective. This bias can take two forms: formula bias based on scanner data. The approach proposed in Ivancic
or lower-level substitution bias. The index suffers from for- and others (2009) that provides drift-free, superlative-type
mula bias if, as a result of the properties of the formula, the indices through adapting multilateral index number theory
result produced is biased relative to what would have been seems to provide a solution to this problem. The meth-
the result if a price change of a fixed basket could have been ods proposed, however, pose some practical challenges
estimated. The index suffers from lower-level substitution and require some more evidence before becoming widely
bias if it does not reflect product substitution by consum- accepted. For an overview of methods for price measure-
ers among the items contained in that elementary aggregate. ment using scanner data, see Chessa and others (2017) and
Lower-level substitution error is only relevant where the Chapter 10 on scanner data.
measurement objective is a COLI. Thus, given any elemen- 12.49 The method used by the NSO for sampling items
tary index formula, the two forms of bias can be distin- within a category will determine the effectiveness of for-
guished according to the objective of the elementary index. mula choice in dealing with lower-level substitution bias.
12.45 Chapter 8 of this Manual and Chapter 6 of the For example, if only a single representative item is chosen to
publication Consumer Price Index Theory discuss the char- represent the category, the index formula will fail to reflect
acteristics and provide detail, an illustration, and the relative the consumer response to any relative price change in the
merits of the use of different elementary index number for- universe of items. Therefore, a larger sample of represen-
mulas. A key finding is that the Carli formula or the arithme- tative items should yield a smaller sampling variance for
tic average of price ratios is unsuitable for a CPI because it a given elementary index. More generally, the geometric
is liable to lead to substantial drift in the results, especially mean formula index suffers from an upward bias in very
when used in its chained form. Therefore, the recommen- small samples (fewer than five observations), so lower-level
dation is that the Carli formula should not be used, espe- substitution bias may be underestimated in empirical com-
cially in its chained form. The problems with elementary parisons of the geometric mean to other index formulas.
aggregate bias and the methods chosen to address them are White (1999) discusses the relationship between sampling
discussed, for example, by Reinsdorf (1998), Reinsdorf and error and bias estimates. McClelland and Reinsdorf (1999)
Moulton (1997), and Moulton (1996b). also study the impact of small sample sizes on the index and
12.46 The ratio of arithmetic mean (Dutot) and geo- conclude that it has the effect of raising the expected values
metric mean (Jevons) formulas eliminate formula bias of an index based on nonlinear formulas, especially the geo-
as defined here. Their expectations differ, however, when metric mean formulas. More extensive use of scanner data
item prices do not change at a uniform rate. The differences may mitigate the problem of small sample given that sample
provide one way of evaluating the potential importance of sizes in a typical scanner data set are large. In some cases,
lower-level substitution bias. The geometric mean formula the use of scanner data may remove the need for sampling.
is exact for a COLI if consumers follow the Cobb–Douglas 12.50 The impact of formula choice can be estimated
behavioral model (that is, assuming that consumers adjust with some degree of precision over a given historical period.
the relative quantities they consume inversely in proportion Any corresponding bias, however, can be estimated only
to the changes in relative prices so that expenditure shares by assuming that the geometric mean or other functional
remain constant), whereas the formula based on the ratio of form successfully approximates the index’s measurement
arithmetic means corresponds to zero-substitution behavior. objective.
Thus, if the goal is to approximate a COLI, the geometric 12.51 As implied by the previous discussion, the impor-
mean formula is judged preferable. tance of elementary aggregate bias will vary by country,
12.47 Scanner data provide new opportunities for mea- depending on the particular index formulas used, the degree
suring and addressing elementary aggregate bias. The avail- of heterogeneity within index strata, and the sampling meth-
ability of both prices and quantity information in scanner ods employed. Also, as with upper-level substitution bias,
data remove the need for an unweighted index formula, at elementary aggregate bias will vary with the overall level
least for those items where unit values are available, and of inflation in the economy if absolute and relative price
allows the calculation of elementary indices by employing changes are correlated.
superlative formulas. By using scanner data, Gabor and Ver- 12.52 The performance of any formula for elementary
meulen (2015) compute product category level elementary aggregate calculation will also be affected by the meth-
price indices using nine different index formulas (Carli, ods used by the NSO to handle special situations, such as
Dutot, Jevons, Laspeyres, Paasche, Fisher, Lowe, Geomet- seasonal products and other products that are temporarily
ric Lowe, and expenditure weighted Jevons) and compare unavailable. Armknecht and Maitland-Smith (1999) discuss
the resulting indices with the Fisher index. The main find- how the failure to impute missing prices can lead to bias in
ings are that across product groups mean levels of annual the modified Laspeyres and other index formulas.
elementary index bias vary between −0.53 and 0.55 percent-
age points depending on the index. Quality Change and New Goods Bias
12.48 Haan and Heymerik (2009) have identified a 12.53 Discussion of potential CPI biases arising from
problem associated with the use of scanner data, especially inadequate quality adjustment has a long history. For
303
example, the Stigler Committee report on US price statistics products; the treatment of new digital products that replace
(Price Statistics Review Committee, 1961) indicated that if existing nondigital products; and improved variety selection
a poll were taken of professional economists and statisti- of digital and nondigital products yield overestimation of
cians, in all probability they would designate (and by a wide inflation. By using the weights in an average consumption
majority) the failure of the price indices to take full account basket for the Organisation of Economic Co-operation and
of quality changes as the most important defect of these indi- Development (OECD) member countries from the OECD
ces. In most studies of bias, unmeasured or mismeasured purchasing power parities program,5 the inflation was over-
quality change is also the largest contributor to the total esti- estimated by 0.28 percentage points because of possible
mated bias. Just as quality adjustment is widely recognized underadjustment for quality changes in digital products such
as an extremely difficult process, it is correspondingly dif- as computers, information and communication technology
ficult to measure any quality change bias. equipment, and telecommunication services.
12.54 Unlike substitution bias, which can be estimated 12.59 New goods bias, like elementary aggregate bias,
by comparison of alternative formulas, quality change bias can be divided conceptually into two components. The first
must be analyzed on a product-by-product basis. Products concerns the failure to bring new products into the CPI sam-
and their associated index components will experience ple with sufficient speed. This can lead to upward bias if
widely varying rates of quality change over time. More- those new products later experience large price reductions
over, the methods used for quality adjustment will also vary. that are not reflected in the index. The second component
Whereas the linking method (link to show no change) may is the welfare gain that consumers experience when a new
dominate with regard to frequency of use, important index product appears; however, this may not be viewed as a bias
components may employ production cost, hedonic adjust- if the CPI measurement objective is a cost of goods index
ment, or the other methods described in Chapter 6 of this and not a COLI.
Manual and Chapter 8 of the publication Consumer Price 12.60 As discussed in Chapter 6, “new goods” can be
Index Theory. replacements for disappearing items, for example, cloud
12.55 A crucial point to recognize is that the direction storage areas replacing physical storage devices; new variet-
of overall quality change does not imply the direction of ies of an existing product that widen the range of consumer
any quality change bias. Nonexperts sometimes assume that choice, such as nonalcoholic beers and ethnic restaurants; or
little or no quality adjustment is carried out in the CPI, and products that represent entirely new categories of consump-
that it therefore must overestimate price change in view of tion, such as multitask robots for cooking or smartphones.
the many demonstrable improvements over time in the qual- 12.61 Like quality change bias, new goods bias has
ity of goods and services. Rather, for any component index, sometimes been estimated primarily by generalization from
the issue is whether the direct or indirect method chosen individual product evidence. A frequent approach has been
for quality adjustment overestimates or underestimates the to measure the price change for a product or category dur-
relative quality of replacement items in the CPI sample. The ing a period prior to its entry into the CPI sample. Studies
resulting bias can be either positive or negative. by Hausman (1997, 1999) on breakfast cereals and mobile
12.56 Empirical evidence on quality change bias has phones provided quantitative measures of the consumer sur-
been based largely on extrapolation from individual studies plus gain from the new products, but this complex econo-
of particular products. These individual studies may involve, metric approach has not been applied widely. Some of the
for example, comparisons of hedonic regression indices to Boskin Commission’s estimates of new product bias, nota-
the corresponding CPI series or estimates of the value of bly those for food, were necessarily based on conjecture.
some product improvement that is ignored in CPI calcula- 12.62 Also, like quality change bias, new goods bias
tions. Although the majority of such studies have suggested could be negative if the range of products decreases, if
upward rather than downward bias, the reliance on fragmen- valuable consumer goods disappear from the market, or if
tary evidence has led to criticism by observers who point to the index fails to capture phases of rapid price increase for
evidence of quality declines that have not been subjected to items. Most observers, however, seem to agree on the direc-
systematic analysis. tion of bias as upward, and that the uncertainty concerns the
12.57 Overall quality trends can also be a matter of magnitude. The extent of the new goods bias depends on
subjective valuation, especially for services. New technol- the importance of the new products with regard to the pro-
ogy has led to unambiguous improvements in the quality portion of consumer expenditure spent on new products not
of many consumer durables and other goods. By contrast, yet introduced in the CPI basket, and the extent of the price
in service sectors such as mail delivery, public transport, decline from the initial price.
and medical care, it can be difficult to evaluate changes in 12.63 One of the risks of downward bias in the CPI is
quality. Airline travel, for example, has become safer and associated with producers that reduce the package size of
faster but perhaps less comfortable and reliable in recent household goods keeping the price stable (“shrinkflation”)
decades, and the lack of cross-sectional variation in these or repacking the old product. This phenomenon is strictly
characteristics makes the use of hedonic quality adjustment linked with the minor changes in product packaging or prod-
problematic. uct characteristics (so-called product relaunches). These
12.58 Digitalization of the economy, if not properly should be properly handled, through the use of a unit value
captured, could also be a source of bias. Reinsdorf and approach, in particular in scanner data, because the product
Schreyer (2017) identify three possible sources of distortion relaunched in most of the cases presents a new Global Trade
that the digital economy can cause, one of which is incom-
plete adjustment for quality change, that is, the treatment
of new, and typically improved, varieties of existing digital 5
https://www.oecd.org/sdd/prices-ppp/.
304
Errors and Bias
Item Number but it is directly comparable with the product substantial decline in average prices. Hausman (2004) also
before the relaunch. When scanner data are used in CPI and documents the growing role of discount outlets and provides
large amounts of data are processed on a weekly basis, it is a specific example of outlet bias.
not possible to observe and report all changes incurred in 12.69 Like with new goods bias, most studies seem to
the size or in the characteristics of the product and assess agree on the direction of bias as upward. The extent of the
the comparability of the previous and the replacing prod- new outlet bias depends on (1) the components of the CPI
uct. Therefore, automatic procedures have to be carefully basket that are likely to be affected, (2) the change in market
implemented to link different Global Trade Item Numbers in share of new outlets for these items, and (3) the percentage
contiguous months and correctly estimate the price change, difference in quality-adjusted prices between new outlets
avoiding the risk of bias that usually is downward. and old outlets. Estimates of the size of outlet substitution
bias must take into account the fact that the market price of
New Outlet Bias an item depends on both the quality of the item and the qual-
12.64 Conceptually, new outlet bias is identical to new ity of the outlet where it is purchased, based on such factors
goods bias. It arises because of the failure to reflect either as the level of service and the convenience of the location.
price changes in new outlets not yet sampled, or the wel-
fare gain to consumers when the new outlets appear. The
explanation for its existence as a separate bias category is Summary of Bias Estimates
twofold. The first reason is historical: new outlet bias was 12.70 The 1996 Boskin Commission report gave a range
identified by Reinsdorf (1993) as a potentially major expla- of estimates for the total upward US CPI bias of 0.8–1.6 per-
nation for anomalous movements in the US CPI. Second, centage points, with the point estimate being 1.1 percentage
the methods used to sample and compare outlets differ from points. This total reflects the straightforward summation of
those used with products, and the problems in controlling the component bias estimates. As reported in US General
new outlet bias are somewhat different. Accounting Office (2000), however, changes in CPI methods
12.65 A failure to maintain a current outlet sample can subsequent to 1996 led the Boskin Commission members to
introduce bias because the new outlets are distinctive in reduce their estimates of total bias. Lacking evidence to the
their pricing or service policy. Reinsdorf (1993) and, more contrary, additivity of biases has been assumed in most such
recently, Hausman and Leibtag (2004, 2005) focus on the studies. Shapiro and Wilcox (1997b) provide probability
growth of discount stores. It should be noted, however, distributions and correlations of their component bias esti-
that the problem could also be geographical in nature; it is mates, yielding an overall confidence interval for the total
important to employ outlet sampling frames that reflect new bias. Most detailed studies of bias also conclude that the
as well as traditional shopping locations, although the wide- CPI bias is in an upward direction, although there have been
spread and increasing weight of online outlets have changed numerous criticisms of that conclusion. For example, Brown
the dimension of this issue. and Stockburger (2006) estimated that the hedonic quality-
12.66 One way that new products enter the CPI sample adjustment methods in apparel have had both upward and
is through forced replacement, when exiting or less success- downward impacts at different points in time and for differ-
ful products disappear from shelves. Outlet disappearance ent categories of clothing in the United States.
is less frequent, and NSO procedures may not provide for 12.71 In general, NSOs cannot compute or publish CPI
automatic replacement. Moreover, when a new outlet enters bias estimates on a regular basis. Many of the same obsta-
the sample there are no standard procedures for compar- cles that prevent the elimination of bias also stand in the
ing data at the new and old outlets. Thus, the index will not way of estimating bias. These include the lack of complete
incorporate any effects of, for example, lower price or infe- data on product-level consumer preferences and expenditure
rior service quality at the new outlet. behavior, and the inability to observe and value all differ-
12.67 Reinsdorf (1993) estimated the degree of new ences in quality among items in the market. Without such
outlet bias by comparing average prices at outlets entering information, it is impossible to calculate a true COLI, and
and disappearing from US CPI samples. There has been similarly impossible to measure the divergence between its
little or no empirical work, however, on the measurement rate of growth and the growth rate of the CPI.
or consumer valuation of outlet quality such as product 12.72 NSOs have been reluctant to provide their own
variety, location, car parking, and customer services. As a estimates of CPI bias. In some cases, they have accepted the
consequence, there is little evidence on how to evaluate the existence of substitution bias, recognizing that the use of a
accuracy of new outlet bias estimates. Laspeyres formula implies that the CPI usually will over-
12.68 Greenlees and McClelland (2012) confirm the state price change relative to a COLI estimated by a super-
potential importance of new outlets bias in the CPI. This lative index such as the Fisher. NSOs have, however, been
study provides new and detailed evidence on the impact of reluctant to draw even qualitative conclusions from the frag-
the appearance and growth of new types of outlets on the mentary and speculative evidence on quality change, new
CPI. Using price data collected by the US Bureau of Labor goods, and new outlet bias.
Statistics for 2002–2007, the authors observed a continuous 12.73 The CPI bias may appear to a different extent in
increase in the market share of discount department stores different countries. Hanousek and Filer (2001) show that
and warehouse club stores, and significantly lower prices bias was especially high in countries in transition and during
in these stores than at large grocery stores. The increasing the period of high inflation. They argue that the substitution
shares of lower-priced store categories reduced the average bias increases with the increase of the variance of relative
prices collected by the Bureau of Labor Statistics. Changes price changes. They argue that bias arising from substitution
in the distribution of outlets within categories also led to a increases in line with the level of inflation, because, as the
305
rate of inflation increases, so does the variance of relative ability to track market trends and detect the emergence
price changes. of new products on the market, helping to reduce the
lag in introducing new goods into the CPI basket. The
Procedures to Minimize Bias use of scanner data also makes it possible to compile
superlative price indices at detailed aggregation levels
12.74 Although it is almost impossible to elimi- since prices and quantities are available.
nate sources of bias, measures can be taken to minimize
them. These include:
Key Recommendations
i. Use appropriate formulas in compiling elementary ag-
gregate indices, in particular use of the geometric mean • In order to ensure public confidence in a CPI, a detailed
(Jevons) formula where appropriate or the ratio of arith- and up-to-date description of the methods and data
metic prices (Dutot) formula. sources should be published. The document should
ii. Review and update weights and CPI baskets frequently, include, among other things, the objectives and scope of
but at least once every five years. Given that a sig- the index, details of the weights, and a discussion of the
nificant part of the total measurement bias in the CPI accuracy of the index.
may be caused by the fixed nature of the CPI basket, • A description of the sources and magnitude of the sam-
the item-substitution bias and some of the new prod- pling and nonsampling errors (for example, coverage or
ucts bias could be reduced by increasing the frequency nonresponse rates) in a CPI should be published to pro-
at which weights are updated. For some categories, it vide users with valuable information on the limitations
may be necessary to update the weights more frequently that might apply to their uses of the index.6
as such weights are likely to become out of date more
quickly than higher-level weights. In periods of high • Resources should be allocated to those areas where the
inflation, the weights should be updated even more fre- effect on the quality of the all-items CPI is maximized,
quently. Scanner data may help in this, at least for some especially to goods and services with a high relative
areas such as food. expenditure weight and to those with high dispersion of
prices.
iii. Use a superlative index formula rather than the Laspey-
res, if current period weighting data can be obtained • To reduce the index’s potential for giving a misleading
on time. Where Lowe or Young indices are used, the picture, it is in general essential:
upper-level substitution bias can be reduced by more • To update weights and baskets regularly
frequent updating of expenditure weights, implement-
• To employ unbiased elementary aggregate formulae
ing them with minimal time lag. Other options might be
to use formulas that allow substitution or assumptions • To make appropriate adjustments for quality change
about substitution between elementary aggregates to be • To allow adequately and correctly for new products
entered. • To take proper account of substitution issues
iv. Closely monitor and update outlet samples to reflect • To undertake quality control of the entire compilation
changes in the outlets from which households pur- process
chase. For example, there is clearly a need to plan for
the inclusion in CPIs of purchases from outlets operat- 12.75 Improving precision and accuracy of the CPI will
ing exclusively online, but also from discount outlets, take both time and resources. Resources should be allocated
factory outlets, or others whose importance has been to those areas where the effect on the quality of the all-
increasing. items CPI is maximized, especially to goods and services
with a high relative expenditure weight and to those with
v. Include new goods in the CPI as soon as possible. For
a fixed-weight index such as Laspeyres, there would high dispersion of prices. Further uses of scanner data can
help NSOs deal with the quality changes, outlet substitution,
also be a need to update the fixed weights to allow for
and new goods bias problems. An investigation into factors
the inclusion of new goods if they are substituting for
affecting consumer choice and an expanded HBS would
all goods in general, or to adjust the weights within an
help identify consumer preferences for different outlet types
item group if the new goods are substituting for specific
and to improve accurate price measurement in areas where
items. For example, one could argue that MP3 play-
quality change is rapid. Opportunities presented by technol-
ers were a new good, but as they were substituting for
ogy, such as the use of computer-assisted techniques, scan-
portable cassette and CD players, they could be intro-
ning, and web scraping techniques, can minimize processing
duced into the item grouping for portable cassette and
errors.
CD players, and weights between these items adjusted
accordingly.
vi. Ensure that the most appropriate quality-adjustment
methods are applied.
vii. Make greater use of the scanner data to deal with qual-
ity change, substitution, and new products. Scanner 6
Examples of handbooks of CPI methods are those published by the US
data contain detailed and timely information on the Bureau of Labor Statistics (2015) and Australian Bureau of Statistics (2018,
prices and quantities of all consumer transactions. The Chapter 11), which devote a section to the varieties and sources of possible
role for scanner data cannot be understated, given its error in the index.
306
DATA QUALITY MANAGEMENT AND REPORTING 13
Introduction organized around five dimensions and a set of prerequisites
for the assessment of data quality. The five dimensions of
13.1 Consumer price indices (CPI) are one of the most data quality include: (1) assurances of integrity (institutional
important statistical indicators produced on a regular basis integrity, transparency, and ethical standards); (2) method-
by national statistical offices (NSOs). Besides informing ological soundness (concepts and definitions, scope, and
economic policy, they are used for indexation of social secu- classifications); (3) accuracy and reliability (adequate data
rity benefits, pensions, salaries and wages, and also for esca- sources and statistical techniques); (4) serviceability (peri-
lation clauses in private contracts, as mentioned in Chapter odicity, consistency, and revisions policy); and (5) acces-
2. Given the considerable financial consequences that any sibility (data and metadata accessibility, and assistance to
errors in the CPI can have on the government budget over users). Each dimension comprises three to five elements
the long term, as well as other financial implications related that are associated with a set of good practices and several
to wages and other uses as an escalator, accuracy and reli- relevant indicators. The focus of this chapter is on meth-
ability are particularly paramount for a CPI. odological soundness and, more particularly, accuracy and
13.2 This chapter addresses the issue of quality man- reliability. The DQAF for a CPI is described in the section
agement and reporting. It gives an overview of the pro- that follows.
cesses and procedures that can be used for quality control
of the CPI production process and the quality indicators that
can be used to measure the extent to which the computed Data Quality Assessment
index meets the prescribed concepts and methodologies that Framework (DQAF) for a
underlie and define the target index. It begins by providing Consumer Price Index
an overview in the context of the International Monetary
Fund (IMF) Data Quality Assessment Framework (DQAF) 13.6 A conceptually based and systematically executed
and then describes various quality management systems and approach to data quality assessment is essential to achieve
key aspects and processes of quality management in the a high-quality CPI. The IMF DQAF for a CPI4 provides a
ongoing production of a CPI. flexible structure specifically for the qualitative assessment
13.3 The CPI should be produced in accordance with of a CPI in a country context. The DQAF for CPI covers the
the United Nations Fundamental Principles of Official Sta- various quality aspects of data collection, processing, and
tistics.1 The International Labour Organization (ILO) guide- dissemination.
lines concerning dissemination practices for labor statistics 13.7 The Framework is organized in a cascading struc-
should also be respected.2 ture that progresses from abstract principles to more con-
13.4 The IMF DQAF identifies quality-related fea- crete details.
tures of governance of statistical systems, processes, and 13.8 The methodological soundness of a CPI is assessed
products. It is rooted in the United Nations Fundamental against the guidelines outlined in the System of National
Principles of Official Statistics and grew out of the IMF’s Accounts 2008 (2008 SNA), this Manual, and the 2003
Data Standards Initiatives3 on data dissemination, including resolution concerning CPIs adopted by the International
the Special Data Dissemination Standard (SDDS) and the Conference of Labor Statisticians (ILO 2003). ILO 2003
Enhanced General Data Dissemination System (e-GDDS). outlines the basic principles for the compilation of CPIs.
The DQAF provides a structure for assessing existing prac- The concepts and definitions from the 2008 SNA are used as
tices against internationally accepted standards, guidelines, guidelines for coverage and valuation, and the methods and
or good practices. It has proved to be useful to NSOs in procedures from this Manual are used as guidelines for com-
undertaking self-assessments of the quality of their CPIs piling the CPI. The DQAF lists four elements to be assessed:
that can be the basis for internal planning, justification of concepts and definition, scope, classification/sectorization,
additional resources, and evaluation of whether they are ful- and basis for recording.
filling their obligations to compute a fit-for-purpose CPI, as 13.9 The accuracy and reliability of a CPI are assessed
well as in guiding data users in evaluating data for policy against whether the source data and statistical techniques
analysis, forecasts, and economic performance. are sound and statistical outputs sufficiently portray real-
13.5 The DQAF covers various quality aspects of data ity. This dimension covers five elements to be assessed:
governance, collection, processing, and dissemination. It is (1) source data, (2) assessment of source data, (3) statisti-
cal techniques, (4) assessment and validation of intermedi-
ate data and statistical outputs, and (5) revision studies. The
1
http://unstats.un.org/unsd/dnss/gp/fundprinciples.aspx.
2
Sixteenth International Conference of Labour Statisticians, 1998.
3 4
https://dsbb.imf.org/. https://dsbb.imf.org/content/pdfs/dqrs_cpi.pdf.
307
considerations are wide-ranging. For example, with respect objectives underpinning user engagement and communica-
to source data it is assessed if the sample and the estimation are to identify and where possible address user needs
tion procedures are soundly designed to represent the survey (for example, the publication of a family of indices or CPIs
universe. This covers both (1) sampling and whether scien- at a subnational level), and to build user understanding and
tific random sampling techniques and cutoff sampling are confidence in the CPI.
used to select geographic areas, items, outlets, and varieties, 13.13 In addition, quality management should include
and (2) where sampling frames are not adequate to support effective customer education on the use of CPI data. In these
statistical sampling, judgmental sampling is used as a sec- terms, success can be measured not only by the achievement
ond-best procedure with well-defined, published criteria for of a high level of satisfaction among well-informed users
selection. Similarly, for statistical techniques the assessment but also by their proper use of statistics. The accessibility to
is based on a wide range of indicators relating to whether users of relevant metadata has an important role.
the statistical techniques used conform to sound statistical 13.14 In many countries, issues relating to the governance
procedures. These procedures include specific issues that of the NSO are set down in a framework or statistical law.
arise in compiling the CPI, such as the treatment of qual- This defines the functions and responsibilities of the NSO, or
ity change for matched samples, the alternative methods of other related agencies with a role in official statistics, and gen-
index construction, and of measurement of owner-occupier erally guides and directs the work of the office. For instance,
housing costs. Concerning index construction, sound statis- an objective that may be stated in the framework document “to
tical procedures require that the arithmetic mean of price rel- improve the quality and relevance of service to customers—in
atives (Carli) is not used to calculate the elementary (item) both government and the wider user community”—provides a
level indices because of its bias that the ratio of arithmetic powerful statement to guide and support NSOs.
mean prices (Dutot) is used only for homogeneous items, 13.15 This recognition of the importance of quality can be
and that the geometric mean of price relatives (equivalent further endorsed by a published vision statement of the NSO as
to the ratio of geometric means) is adopted as the preferred a key supplier of official, timely, and high-quality information.
measure. Further explanation is given in Annex 13.1. Such a vision statement can be encapsulated by publishing
13.10 The DQAF for the CPI is very comprehen- objectives in an annual report. These objectives can include
sive with regard to its inclusions. For example, under the improving quality and relevance, thereby increasing public
subheading Sound Statistical Techniques, it also requires confidence in the integrity and validity of statistical outputs.
assessment of the statistical methods used to handle missing
prices and the introduction of new products that are within
the scope of the CPI. Thus, the compiler is asked to confirm Quality Management Systems
that prices for temporarily missing products are appropri- 13.16 A quality management system is a formalized
ately handled (for example, a price is imputed based on the system that documents processes, procedures, and respon-
month-on-month price changes of a higher group, or a more sibilities for achieving quality products, policies, and objec-
targeted subgroup when judged, and prior data show that it tives. A quality management system helps coordinate and
is more suitable); the imputed price is posted in the database; direct the NSOs’ activities to meet customer and organiza-
and a limit is established and adhered to as to how long (for tion requirements, to comply with regulatory requirements,
example, three months) prices for temporarily missing items and to improve its effectiveness and efficiency on a continu-
are imputed or otherwise handled. For seasonal products the ous basis.
compiler is asked to ensure that prices are imputed using 13.17 Several internationally recognized systems are
the higher group price change or a more targeted subgroup available to help NSOs improve quality management:
when judged (based on prior data) that it is more suitable,
sample sizes permitting. • Total quality management. Total quality management
(TQM) is a management philosophy that promotes an
effective culture of quality in an organization to fulfill oper-
Quality Management ational objectives efficiently and effectively, including:
13.11 For most NSOs, data production will be an area • Clearly defined organizational goals
that represents a high risk, given the complexity of the pro- • Strong customer focus
cess from price collection to index computation and the
• Strategic quality planning
financial implications of an error in the index. This is so
regardless of the institutional arrangements and formal pro- • Process orientation
cesses in place for auditing. It follows that a priority area in • Employee empowerment
the quality management of a CPI is quality control of the • Information sharing
production process.
• Continuous quality improvement
13.12 A general theme in statistical production is cus-
tomer focus and the effective dissemination of relevant, TQM requires a commitment by all relevant parts of a pro-
accurate, and timely statistics. Thus, a high level of under- duction system to define their starting points, procedures,
standing of customer needs and a coherent statistical and and final results. Revilla (2004) describes four dimensions
quality framework is required. User needs can be inter- of TQM: consumer satisfaction; constant improvement;
viewed either formally through negotiation of contractual fact-based management; and people-based management.
obligations for the provision and dissemination of data that 13.18 TQM has a broad focus ranging from an individual
may be legally binding or less formally through contact- statistical product and its production to the entire system of sta-
ing customers or through customer surveys. The overall tistics production and to other core processes, and from there
308
Data Quality Management and Reporting
on to the entire management system, personnel, partnerships, (the results: people results; customer results; society results;
and resources. By systematic management of these aspects, and key performance results). Evidence based on feedback
one can create an operating system that emphasizes quality from focus groups, questionnaires, and personal interviews
and thereby also improves the quality of end products. is used to score performance, and a resulting action plan for
improvement is introduced, which is then included in the
• Benchmarking. Benchmarking is a process of comparing
business plan.
with others and learning and improving with their experi-
ence. Areas that can be considered when benchmarking a 13.19 Underlying the EFQM Excellence Model is the
CPI collection may include: realization that business excellence—as measured through
customer satisfaction—is achieved through effective lead-
• Timelines, accuracy, and coverage of collection ership that drives policy and strategy, allocates resources
• Benefits of index methodologies for various items (for compatible with that policy, and manages human resources
example, seasonal items) in such a way as to enable them to manage the processes.
• Frequency of collection and publication 13.20 In the case of an NSO, where some procedures
• Cost of collection per unit of item are governed by statute or regulation, the use of the EFQM
Excellence Model enables continuous improvement to be
A number of general observations can be made about
taken forward across a range of processes and functions.
benchmarking:
To work effectively and to contribute to the production of a
• Initial discussions between benchmarking partners, high-quality CPI, the Excellence Model needs the commit-
prior to the process, can provide a useful tool for the ment of senior managers, who must be responsible for lead-
identification of potential issues through informal ing any self-assessment. However, unlike ISO 9001, where
self-evaluation. assessment is carried out by qualified auditors often from
• The benchmarking is not restricted to performance outside the work area (see Section E), the EFQM Excellence
indicators that may be available. In addition to review- Model relies on input from internal staff.
ing issues that are directly measurable, the discus-
• Generic Statistical Business Process Model. The Generic
sion can extend to topics such as why different NSOs
adopt different approaches to some aspects of index Statistical Business Process Model (GSBPM) describes
construction. and defines a set of business processes needed to produce
official statistics. It provides a standard framework and
• Benefits often continue to accrue beyond the benchmark- harmonized terminology to help statistical organizations
ing exercise from, for example, the work on follow-up to modernize their statistical production processes, as well
action points. Correspondingly, the further explora- as to share methods and components. The GSBPM can
tion in greater detail of issues raised during the initial also be used for integrating data and metadata standards,
benchmarking can result in further improvement. The as a template for processing documentation, for harmoniz-
general experience is that work becomes more focused ing statistical computing infrastructures, and to providing
on specific issues as the benchmarking progresses and a framework for quality assessment and improvement.
the issues of concern become more apparent. The GSBPM is intended to apply to all activities under-
• Longer-term benefits also include the subsequent taken by producers of official statistics, at both the national
opportunity for networking. and international levels, that result in data outputs. It is
• Financial and management information compiled spe- designed to be independent of the data source, so it can
cifically for benchmarking can also be useful manage- be used for the description and quality assessment of pro-
ment information in its own right. cesses based on surveys, censuses, administrative records,
• Performance indicators are a necessary ingredient of and other nonstatistical or mixed sources.
the process of continuous improvement and are not just While the GSBPM includes several overarching statisti-
short-term management tools. cal processes, quality and metadata management are two
• A number of general factors can be identified that of the key elements of the model. The data quality man-
contribute to successful benchmarking. In particular, agement process includes quality assessment and control
benchmarking is dependent on trust and mutual respect mechanisms. It recognizes the importance of evaluation
between the parties involved. and feedback throughout the statistical business process.
• European Foundation for Quality Management Excel- The GSBPM processes and generates metadata within
lence Model. The Excellence Model constructed by the each phase, and therefore there is a strong requirement for
European Foundation for Quality Management (EFQM) a metadata management system to ensure that the appro-
is a diagnostic tool for self-assessment. The model has priate metadata retain their links to the data throughout
been widely used by governmental organizations across the GSBPM. Both processes guide NSOs to improve data
Europe to improve quality and performance. It may be quality management and the dissemination of metadata to
described as a tool that drives the philosophy of TQM. enhance user confidence in statistical outputs.
The EFQM Excellence Model focuses on general business • ISO 9001. ISO 9001 sets out the criteria for a quality man-
areas and assesses performance against two sets of criteria. agement system that can be applied to any field of activity
The first consists of five criteria covering what the business including the computation of a CPI. This standard is based
area does (the enablers: leadership, people, policy and strat- on a number of quality management principles including
egy, partnership and resources, and process), and the second a strong customer focus, the motivation and implications
consists of four criteria on what the business area achieves with respect to management, the process approach, and
309
continuous improvement. The International Standard ISO validation of the production cycle, and an annual review
9001 is an international quality standard for manage- process that focuses on strategic and longer-term issues.
ment systems. ISO 9001 notes that a quality system is a This review is strongly encouraged because it allows to
common sense, well-documented business management learn from past experiences and to identify and take forward
system that is applicable to all business sectors. It helps actions that will improve the future quality of the CPI.
to ensure consistency and improvement of working prac- 13.23 A conscious decision needs to be made on whether
tices, including the goods and services produced. Users to include in the main quality management system periodic
of ISO 9001 add value to their activities and improve review processes (for example, chain linking, or the updat-
their performance continually by focusing on the major ing of the CPI basket), and technical development work
processes within the organization. There is a closer align- (for example, introduction of better sampling techniques, or
ment of the quality management system with the needs methods of quality adjustment for replacement goods).
of the organization and the process reflects the way an 13.24 Each aspect of a quality management system
organization runs its business activities. By meeting the should be seen as interdependent and an integral part of the
ISO 9001 standard, an organization will come more into whole. For instance, a good quality CPI depends both on
line with TQM and the EFQM Excellence Model. the accuracy and reliability of the source data and on the
methodological soundness of the index computation, and
13.21 A coordinated use of these and other quality man-
the computation of the index relies on the delivery of an
agement techniques at a strategic level in fields of statistics
accurate, accessible, and timely database. The precise out-
supports the dissemination of better data to meet user needs.
line of such a system will depend on the detailed arrange-
ments and approach to price collection. For instance, Figure
Prototype of a Quality 13.1 allows for the possibility that some prices data are
Management System collected directly from the headquarters of large supermar-
kets or other chain stores (sometimes referred to as central
13.22 A prototype of a quality management system for price collection) and that some price indices are calculated
the monthly collection of prices and compilation of the CPI using price and sales information provided by a central
is given in Figure 13.1. It covers all aspects of CPI data col- authority, such as for energy prices or telecommunications,
lection and compilation including the auditing of prices, or are based on special methodologies requiring tailor-made
Figure 13.1 An Example of a Quality Management System for CPI Data Collection
310
index calculations such as for owner-occupier housing costs processing cycle and outline the responsibilities of the
(sometimes referred, generically, as centrally calculated staff involved.
indices). • Level 3—Work Instructions. The work instructions give
full details on exactly how a task should be carried out.
Documentation • Level 4—Reference Documents. These include all CPI
metadata, such as a technical manual. The CPI Technical
Overview Manual describes the procedures used to produce the CPI
13.25 Good documentation is crucial. Documents are and the price indices derived from it. It is aimed at users
needed to explain what, when, how, and why the different of the CPI who want to know how the data are collected
CPI tasks should take place. Preparing such documents pro- and analyzed and what formulas are used in the calcula-
vides a useful opportunity to assure the quality of current tion, together with other methodological detail.
procedures used to collect prices and compile the index. It 13.26 The first three documents are intended for internal
also provides an opportunity to review and improve these use only as they refer to internal processes and procedures.
procedures. Once in place, documentation serves two pur- The CPI Technical Manual and other reference documents
poses in the context of producing the index. First, it enables should be made available to all users on the NSO website
another person to take over the work if the responsible per- and in hardcopy upon request.
son is not available or leaves. Second, it provides a quality
check to ensure that the procedures that should be carried
out are indeed being carried out in practice. More gener- Documentation Control
ally, documentation can provide a useful reference for CPI
users. Figure 13.2 shows a typical structure for documenta- 13.27 All documents in the quality management system
tion relating to a CPI. should be subject to document control. Document control
procedures should ensure that all staff have access to the
• Level 1—The Quality Manual. This document defines the most recent version when carrying out their work. In some
quality policy and gives a general description of the sys- NSOs, this is done by storing documents such as the Quality
tem. It also describes the organization of staff involved in Manual, Procedures, and Work Instructions electronically in
producing the CPI, the division of responsibilities for the a database managed by a document control custodian and
management of all aspects of the production cycle, and using numbered and dated versions to identify the latest copy.
the general structure of the lower levels of documentation. 13.28 An electronic system of documentation storage
• Level 2—Procedures. These are a set of mandatory proce- and control is recommended where the necessary technical
dures, covering all aspects of the production cycle. They infrastructure exists. Electronic systems have four benefits
explain in broad detail the different parts of the monthly over a manual system:
Figure 13.2 Documentation
Policy and
Quality Manual System
Description
Procedures Who and

What
Work Instructions How
Reference Documents Reference

(e.g., Technical Manual) Documents
311
• More efficient production of documentation as it helps involvement in the production and compilation of the CPI.
with initial compilation and reduces the need to print and Too much familiarity can lead to unwarranted assumptions
circulate paper copies about the appropriateness of the procedures and the extent to
• Better informed staff because they have immediate elec- which they are followed.
tronic access to the latest documentation 13.33 To achieve external accreditation, such as ISO
• Better quality control to allow authors, with the involve-
9001, requires the organization to undergo an external audit
ment of the document control custodian, to readily periodically by the appropriate accreditation body. The
amend, date stamp, and reference number updates, and advantages of obtaining external accreditation are that:
limit access of nonauthors to “read only” • It provides a routine and regular review of production
• Better search facility, for example, if staff are looking for procedures and ensures proper documentation and quality
cross-references to a particular subject such as chain link- control.
ing or weights • It provides the efficiencies associated with a ready-made
13.29 Where an NSO does not have the necessary standard for documentation and quality control and the
information technology infrastructure or capacity to oper- added confidence associated with a well-tested system.
ate an electronic system, it is still important that a docu- It minimizes risks of errors by adding effective controls
ment control custodian is appointed with the task and including a framework for the initiation, evaluation, and
authority to keep a record of the most up-to-date paper implementation of change.
documentation. The same principles of good documenta- • It increases public confidence in the CPI, for instance as
tion control illustrated in Figure 13.2 apply whether the an index produced in accordance with internationally rec-
documents are stored electronically or are kept in paper ognized standards for quality management.
folders. A form for a document control template is given • It provides a basis for more effective training of experi-
in Annex 13.2. enced staff and the induction of new staff.
13.34 Where external accreditation is not an option,
Internal and External Audits of internal audits can provide the same level of objectivity and
Production Processes discipline. Another option is an external audit carried out by
CPI compilers from another NSO.
Overview
13.30 In the context of a CPI, an audit is a systematic
and independent examination of the agreed-upon processes The Role of an Audit Team and the
undertaken to compile the index. An audit evaluates perfor- Responsibilities of Its Members
mance against the objective of producing a reliable, accu- 13.35 It is recommended that an audit team be assembled
rate, and timely CPI that adheres to the defined scope and that consists of a quality manager and an internal auditor,
definition. Quality audits are performed to verify confor- each with a distinct, well-defined, and complementary role.
mance to standards and best practice through a review of Where this is not possible because of resource constraints,
objective evidence but they can also be used to verify the the two roles can be combined, or the roles can be undertaken
effectiveness of the quality management system. Recom- in conjunction with other responsibilities. In some NSOs, the
mendations with action points are given in the audit report role of auditing is undertaken by a methodology branch or
and should systematically be followed up. The advantage by an external government body. Whatever arrangement is
of internal and external audits compared with a less for- adopted, it is important that sufficient staff qualified in audit-
mal approach is that they are standardized, systematic, and ing are available to carry out these audit functions.
transparent. 13.36 The quality manager should be responsible for:
13.31 The auditing function is represented by the
left-hand column of Figure 13.1. Note that the auditing of • Producing an audit schedule, managing it, and ensuring it
price collection is specifically addressed in Chapter 5. It is is updated as necessary
strongly advised that internal audits of the entire produc- • Agreeing and specifying the objectives and scope of the
tion process should be carried out regularly according to audit
a systematic timetable. Audits should cover all aspects of • Managing implementation of the audits specified in the
the monthly and annual processing cycle to ensure that the schedule
management systems are fully implemented. The purpose of • Ensuring the auditor is properly trained
each audit should be to verify that operational procedures
• Ensuring that, where possible, the auditor is independent
and controls comply with the documented procedures and
to determine their effectiveness in delivering a CPI that is of the function being audited
fit for purpose. Thus, an audit should aim not only to ensure • Ensuring review reports are written
that the index compilers adhere to the agreed-upon proce- • Ensuring audit action points are implemented
dures but also that the procedures are improved where found
13.37 The quality manager should prepare a schedule of
lacking or insufficient.
audits covering all aspects of the CPI production processes.
13.32 Internal audits should be conducted by staff hav-
This schedule should take account of:
ing appropriate training and experience, ideally by people
who are sufficiently detached from routine operations to • The importance and complexity of the different stages
take an independent and objective view unhindered by close involved in the compilation of the CPI
312
• The results and concerns arising from previous audits and • Provide compilers with the opportunity to comment
any issues that have arisen since and then agree on an action plan
• The time since the previous audit • Try and resolve any disagreements and, where appro-
priate, include the index compilers comments in the
13.38 The internal auditor should be responsible for:
report
• Undertaking any necessary preaudit investigation that • Note opportunities for improvement to current proce-
will help determine the audit schedule and the focus and dures as well as noncompliance with them
scope of assessment
• Conducting the audit • Annex 13.3 provides an example of a pro forma audit
schedule.
• Producing and circulating the audit reports
• Updating documentation where necessary
Audit Report
• Taking responsibility for monitoring and following up
13.42 It is recommended that similar issues identified
actions from the audits
in several areas should be grouped under the same heading
and documented as one audit observation. Clear and suc-
Objectives of an Audit cinct headings should be used for each issue identified and a
13.39 The objectives of an audit need to be clearly short description should be given of what was found. Obser-
defined and agreed on before it starts. The precise objec- vations can either be made in order of the most significant
tives will depend on local circumstances but may be stated ones first or they can follow a logical sequence, generally
in general terms as follows: the order in which the work activities underlying the compi-
lation of the index are carried out.
• To assess compliance with documented procedures 13.43 For each audit observation, suitable corrective
• To provide assurance to senior management that the action should be recommended together with a statement of
agreed-upon quality management system is being imple- who is responsible for carrying out the corrective action and
mented and is effective and relevant the date when the action should be complete. It is important
• To identify improvements required and any necessary to check that action points are carried through and that this
corrective actions and preventive measures is recorded with any further issues that arise.
• To ensure the procedures are adequate
Risk Assessment
Auditing Procedures and Techniques 13.44 In order to prioritize follow-up actions, it is rec-
13.40 The success of an audit relies not only on clear ommended practice for the NSO to attribute an overall score
objective setting and well-trained auditors but also on the to the risk associated with each issue arising from the audit.
use of effective auditing procedures and techniques, includ- 13.45 The issues arising from an audit may be catego-
ing a review of documentation and the carrying out of struc- rized as follows:
tured interviews with index compilers. • Low Risk if the issues that have been identified are
13.41 It is recommended that audits should incorporate unlikely to arise and, if they did, that they would not
the following procedures as standard: result in major difficulties for the published CPI
• Before the audit: • High Risk if there is a significant chance that, unless
• Review documentation for completeness and ensure addressed, the issue could lead to an error in the index;
that is it up to date and indexed properly (where appli- that is, the error will be of high impact and there is a sig-
cable, check that updates are signed off correctly, and nificant chance of it happening
that minutes, action points, and other documentation • The results of all other audits should be considered of
are correctly filed) Medium Risk
• Trace action points from previous audits and any other 13.46 If judged appropriate, the quality manager may
form of review that has taken place (for example, com- issue an instruction for a follow-up audit. The proposed date
pilation procedures) for this audit should be added to the schedule for follow-up
• Prepare checklists to help facilitate the structured audits and progress checks, and the actual date of the audit
interviews should be added to the schedule of internal audits. When-
ever possible, the follow-up audit should be carried out by
• During the audit: the same auditor who undertook the initial audit. Annex 13.4
• Conduct structured interviews gives an example of a template for an audit report and the
• Ask to be shown documentary evidence (for example, recording of follow-up actions.
spreadsheets or signatures on spreadsheets) to support
staff responses to questions Review Systems
• Identify issues as they arise and advise the index 13.47 The review system is represented by the right-
compiler hand side of Figure 13.1. A review system not only provides
a check on current operational procedures but also helps to
• After the audit: inform decisions on the introduction of longer-term improve-
• Produce a report, with recommendations ments so that the quality management system continues to
313
be up to date and relevant, and to ensure that business risks relevant members of staff with a view to acting to minimize
are kept to the necessary minimum. For this reason, NSOs such errors in the future.
are encouraged to put in place monitoring arrangements to 13.53 The monthly review meeting should also include
track performance, supplemented by both short-term and a forward look at issues arising over the next cycle, so that
longer-term review procedures. appropriate working arrangements and solutions can be put
in place. Thus, the focus should be both on learning from
Monitoring Performance experience, for example, to avoid repeating problems, and
on anticipating future issues for forward planning. Action
13.48 The process of producing a CPI should rely on
points should be recorded, and individuals identified to
an agreed-upon set of objectives that, where possible, are
follow them up. Recommendations to senior management
supplemented by measurable targets.
should also be recorded as appropriate.
13.49 Targets for the delivery of a CPI may cover both
13.54 The monthly review should also be used as a basis
quality (data and statistical methodology) and timeliness,
for continuous improvement processes. As an example, data
and may encompass both the data collection and compila-
collectors can be encouraged to analyze the root causes of
tion processes and the quality of the final output. The NSO
pricing errors and to develop individual development plans
will need to decide which are the most relevant targets for
to correct these.
the CPI. Possible targets for monitoring monthly perfor-
mance with regard to the process of compiling the CPI and
maintaining its relevance can include: Longer-Term Annual Reviews
• Timeliness. The targets ensure that process delivery meets 13.55 A longer-term element of a quality review system
the agreed-upon schedule. For example, the targets may takes a higher-level strategic look at objectives and should
monitor whether the prices data was entered on to the be conducted through the annual planning cycle, where such
computer and edited to the agreed-upon timetable, or a cycle exists, and should address both the quality of outputs
whether index compilation took place on time so as not to assessed against user needs and the processes by which qual-
potentially compromise publication. ity is achieved. Ideally, the quality review system should be
laid down in an annual Quality Management Action Plan
• Accuracy. The targets may include the proportion of
annexed to the main CPI work program.
prices that are found to be incorrect, the number of prices 13.56 The objectives of an annual review and the activi-
collected compared with the target sample, or errors in
ties to be covered need to be stated with clarity. For example:
the compilation of elementary aggregates. Accuracy and
timeliness are interconnected. It is important that the qual- • The overall aim may be to successfully complete the fol-
ity of the index is not compromised by, for example, not lowing processes:
allowing sufficient time for data editing and for checking • The reweight of the CPI
of the index calculation prior to publication. • The update of the basket of goods and services
• Delivery. The targets may include the delivery of planned
• The compilation of the item list
reviews of specific subindices and methodological reviews.
• The update of the new items list on the computer
13.50 Quality measures should be set for each stage of system
the compilation process. These measures, which should be • The rewrite of existing computer programs to incorpo-
as quantitative as possible, should be evaluated against pre- rate the new items
determined targets on an ongoing basis. Problems should be
flagged during the production process allowing immediate • The objective would be to:
corrective action to be taken if necessary. • Improve the quality and accuracy of the CPI
• Ensure that the CPI reflects consumer expenditure
Short-Term Monthly Reviews patterns
13.51 It is good practice to hold an internal meeting of • Enable data collection to be more effective and efficient
the CPI production team at the end of each month focused
• Ensure a standardized approach for introducing
on quality and operational issues that have arisen during the
improvements
most recent production cycle. The meetings can be an infor-
mal gathering of the team to exchange experiences and raise • To achieve this, the review may cover three main areas:
issues for resolution, or a more structured general session
where management presents a monthly performance report • Locally collected prices
and team members have the opportunity to react or raise • Centrally collected prices
particular concerns. The format will reflect local circum- • Weights
stances, including the size of the team producing the CPI
and the management arrangements. Depending on the issues
that arise, it may be useful to follow up these meetings by Quality Reporting and Improving
smaller ad hoc groups of staff brought together to tackle spe- the Consumer Price Index:
cific issues. Seminars and presentations may also be given. Frameworks
13.52 Monthly reports on errors observed at data col-
lection, data entry, editing, coding, and data cleaning stages, 13.57 This section discusses quality reporting and
together with any compilation issues, should be circulated to reviews different quality frameworks that provide guidance
314
on reporting mechanisms; they can assist NSOs to determine • Disseminating the data required by the SDDS punctu-
whether the quality of the CPIs produced meets the needs of ally and with the prescribed periodicity and timeliness on
users and how to develop a program for improvement. Qual- a national webpage, the National Summary Data Page,
ity reports should cover not only metadata about the basic which is hyperlinked to the DSBB (provide to the IMF an
characteristics of the CPI, but also wider issues such as its advance release calendar containing release dates for the
quality, statistical integrity, accessibility, and dissemination. current month and at least the following three months for
13.58 Examples of quality frameworks include the IMF each prescribed category of data for posting on the DSBB)
e-GDDS, SDDS,5 and the IMF DQAF. All frameworks, • Provide detailed information about their statistical prac-
apart from Eurostat's quality reporting framework for the tices, or metadata, for dissemination on the DSBB (the
European Union Harmonised Index of Consumer Prices and metadata follow the rigorous format of the DQAF)
the IMF’s DQAF (that covers different areas of statistics), • Certify the accuracy of the metadata on an annual basis
are generic in nature and not specific to CPIs.
13.59 The IMF DQAF (discussed in paragraphs • Use standardized electronic reporting procedures to mon-
13.6–13.10) provides a flexible structure for the qualitative itor more effectively their observance of the SDDS
assessment of a CPI that can be used in a variety of con- 13.62 Using one of the internationally recognized qual-
texts, including self-assessments performed by data produc- ity frameworks for reporting on the CPI is recommended
ing agencies. Reporting arrangements will depend on the because it has several advantages. These include:
governance structure that is in place and this can vary across
countries. In general, the expectation should be that qual- • It has the authority associated with an internationally
ity reports should be made public with opportunity given to agreed-upon framework and benefits from the experience
users to react. of different countries.
13.60 The IMF’s e-GDDS provides a framework for • It allows comparisons on a like-for-like basis with CPIs
developing a clear roadmap to achieving higher data dissem- produced by other countries.
ination standards at a pace consistent with evolving statisti- • It is readily available.
cal capacity. It focuses on the publication of data through a • It fulfills the reporting obligations to international
standardized platform to improve efficiency in data sharing, organizations.
while identifying critical gaps to help prioritize technical
• It provides the basis for reporting to users.
assistance and donor support. IMF member countries that
participate in the e-GDDS agree to: • It provides a benchmark for future developments, particu-
larly when carried out in conjunction with a checklist.
• Commit to using the e-GDDS as a framework for statisti-
cal development 13.63 Annex 13.5 gives an example of a model quality
• Designate a country coordinator
report document based on the reporting framework for the
Harmonised Index of Consumer Prices. Quality is defined
• Prepare descriptions of current statistical production and as “fitness for use” with regard to user needs and extends
dissemination practices and plans for their improvement beyond the statistical accuracy of the index to its definition
for posting on the Dissemination Standards Bulletin and coverage, effective dissemination, and the transparency
Board (DSBB)6 of the statistical system.
13.61 The IMF’s SDDS is a global benchmark for dis- 13.64 The general principles underlying the publication of
seminating macroeconomic statistics to the public. Countries quality reports on the CPI, as with all official statistics, are that:
that subscribe to the SDDS agree to follow good practices • The reports should be easy to access and use by all inter-
in four areas: coverage, periodicity, and timeliness of data; ested parties.
public access to those data; data integrity; and data quality.
• The contents should be sufficiently detailed to allow
Subscribing countries commit to:
users to assess fitness for particular purposes. Qualita-
tive (and where possible quantitative) measures of quality
should be included to help users to understand better the
strengths and the limitations of the CPI and associated
5
The IMF has taken steps to enhance member country transparency and series, and the corresponding implications for interpreta-
openness, including setting voluntary standards for dissemination of eco- tion and appropriate use.
nomic and financial data. The SDDS was established in 1996 to guide IMF
members that have, or might seek, access to international capital markets in • Quality measures and detailed technical information
providing their economic and financial data to the public. The GDDS was should be supplemented by guidance on interpretation to
established in 1997 for member countries with less developed statistical help users assess fitness for purpose.
systems as a framework for evaluating their needs for data improvement
and setting priorities. In 2012 the SDDS Plus was created as an upper tier
• Clear statements should be given on the degree of com-
of the IMF’s Data Standards Initiatives to help address data gaps identified pliance with agreed-upon definitions, methods, and prac-
during the global financial crisis. In 2015 the enhanced GDDS (e-GDDS) tice, including both those determined nationally and those
replaced the GDDS. Details can be found on https://www.imf.org/en/About/ laid down in the 2003 ILO Resolution on CPIs,7 and any
Factsheets/Sheets/2016/07/27/15/45/Standards-for-Data-Dissemination. known reasons for deviations.
6
The IMF’s Dissemination Standards Bulletin Board (DSBB) provides
access to SDDS Plus, SDDS, e-GDDS, and the Data Quality Reference
Sites. The Data Quality Reference Sites have been created to foster a
7
common understanding of data quality. It provides access to contributions http://www.ilo.org/global/statistics-and-databases/standards-and-guide-
in the field and includes a selection of articles and other sources related to lines/resolutions-adopted-by-international-conferences-of-labour-statisti-
data quality issues. cians/lang--en/index.htm.
315
• Where possible, the presentation of information on qual- • The setting of protocols relating to the scope and defini-
ity will be tailored to meet the needs of different types of tion of the CPI and the methodological detail that sup-
users, with more comprehensive information being pre- ports the previous point
pared for expert users. This may indicate separate quality • Putting in place and managing the routine operational
reports directed at different user groups. arrangements
• Producers should systematically review at regular inter-
13.67 Transparency is built on the free flow of relevant
vals the documentation relating to the CPI and update it
nonconfidential information directly accessible to users of
to reflect up-to-date methods and processes.
the CPI—enough information for users to understand, inter-
13.65 For operational purposes relating to internal work pret, and properly use the index. Transparency generates
programs, a quality report can be usefully supplemented by trust.
a more detailed checklist of issues arising and the corre- 13.68 The operational arrangements should be consis-
sponding corrective actions that need to be taken. tent with the governance arrangements and should:
• Incorporate an effective process for consulting with users
Work Programs: Programming, • Provide a mechanism for regularly reporting (for exam-
Planning, and Reporting ple, annually), to users and other relevant parties on the
answers to three questions:
13.66 The general principles underlying a program-
ming, planning, and reporting system include clear and • What has been done to maintain the integrity of the CPI
transparent governance arrangements relating to: over the past year?
• The allocation of responsibilities for monitoring and • What are the outstanding shortcomings and issues?
reporting on the production and dissemination of the CPI • What does the NSO intend to do during the next year to
and on its development address these questions?
316
Annex 13.1 • Classification/sectorization

• Basis for recording
Data Quality Assessment 3. Accuracy and reliability. This dimension covers the idea
that statistical outputs sufficiently portray the reality of
Framework (DQAF) for the the economy. This dimension is also data specific, reflect-
Consumer Price Index ing the sources used and their processing. The five ele-
ments of this dimension cover the following:
• Source data
The DQAF dimensions and respective elements are the • Assessment of source data
following:
• Statistical techniques
0. Prerequisites of quality. Although not itself a dimension • Assessment and validation of intermediate data and
of quality, this group of “pointers to quality” includes el- statistical outputs
ements and indicators that have an overarching role as
• Revision studies
prerequisites, or institutional preconditions, for quality of
statistics. Note that the focus is on the agency, such as an 4. Serviceability. This dimension relates to the need that sta-
NSO, central bank, or a ministry/department. These pre- tistics are disseminated with an appropriate periodicity in
requisites cover the following elements: a timely fashion, are consistent internally and with other
major data sets, and follow a regular revision policy. The
• Legal and institutional environment
three elements for this dimension are as follows:
• Resources available for the statistical program
• Periodicity and timeliness
• Relevance
• Consistency
• Other quality management
• Revision policy and practice
1. Assurances of integrity. This dimension relates to the ad-
5. Accessibility. This dimension relates to the need for data
herence to the principle of objectivity in the collection,
and metadata to be presented in a clear and understand-
compilation, and dissemination of statistics. The dimen-
able manner on an easily available and impartial basis that
sion encompasses institutional arrangements that ensure
metadata are up to date and pertinent, and that a prompt
professionalism in statistical policies and practices, trans-
and knowledgeable support service is available. This di-
parency, and ethical standards. The three elements for this
mension has three elements, namely:
dimension of quality are the following:
• Data accessibility
• Institutional integrity (statistical policies and practices
are guided by professional principles) • Metadata accessibility
• Transparency (the terms and conditions under which sta- • Assistance to users
tistics are collected, processed, and disseminated are avail- The methodological soundness dimension of quality has
able to the public and meet international best practice) four elements—concepts and definitions, scope, classifi-
• Ethical standards (guidelines supporting appropriate cations and sectorization, and basis for recording—each
staff behavior to sustain a strong culture for maintain- of which is associated with best practice and indicators.
ing ethical standards that discourage political interfer- For instance, for concepts and definitions, the good prac-
ence, are in place and are well known to the staff) tice is that the concepts and definitions used are in accord
2. Methodological soundness. This dimension covers the with internationally accepted statistical frameworks.
idea that the methodological basis for the production of Similarly, for scope the good practice is that the scope
statistics should be sound and can be attained by fol- is in accord with internationally accepted standards,
lowing internationally accepted standards, guidelines, or guidelines, or good practice. The indicators for the lat-
good practices. This dimension is necessarily data set- ter include that the household transactions included in the
specific, reflecting different methodologies for different CPI are selected components of the following 2008 SNA
data sets. This dimension has four elements, namely: aggregates: household final consumption expenditure,
• Concepts and definitions
and fixed capital formation in residential structures, as
applicable.
• Scope
317
Annex 13.2
Documentation Control Template
Figure 13A.1 Documentation Control Template

Date Issued Documentation Reference Details of Change Reason for Change Name of Issuer
Day/ Calculating CPI 2.1 Change in Process CPI Technical Board Has Agreed L. Smith,
Month/ Food Item Weights with Effect from. . . That in Future Weights Should Be Statistician, CPI
Year (nonseasonal) (date) Taken from National Accounts Program
XX/XX/XX Calculating 2.5 Change in Process Methodology Changes in Pricing L. Smith,
and Updating with Effect from. . . Structures for Mobile Telephones— Statistician, CPI
Price Index for New Methodology Agreed by CPI Program
(date)
Telecommunication Technical Board. Reflects Changing
Services Market
XX/XX/XX Desk Instructions 3.1 Additional Checks to Last Audit Indicated Current Checks C. Brown,
for Checking and Be Carried out Based Inadequate Resulting in Incorrect Operations
Editing of Prices on Month-on-Month Prices Entering the CPI Manager, CPI
Price Change Program
319
Annex 13.3
Pro Forma for an Audit Schedule
Figure 13A.2 Pro Forma for an Audit Schedule

Process and Document to Be Ref Scheduled Audit Last Audit
Audited
Date Audit Audit Last Auditor Date of Risk Assessment Follow-Up Target Date
Due Allocated To Audit (High, Medium, or Required?
Low) (Y/N)
Calculating CPI Food Item 2.1 January J. Graham B. Jones XX/XX/XX High N N/A
Weights (nonseasonal)
Calculating and 2.5 January L. Smith J. Graham XX/XX/XX Medium Y April
Updating Price Index for
Telecommunication Services
Desk Instructions for 3.1 January B. Jones L. Smith XX/XX/XX Low Y April
Checking and Editing of
Prices
321
Annex 13.4
Audit Report Template
Figure 13A.3 Audit Report Template
323
Annex 13.5
Model Quality Report Document for the Consumer Price
Index
Figure 13A.4 Model Quality Report Document for the CPI
325
PUBLICATION, DISSEMINATION, AND USER
RELATIONS 14
Introduction previously. The illustrative presentation in Figure 14.1 pro-
vides an example of this.
14.1 The consumer price index (CPI) represents a key 14.5 The data release should also focus on the month-
indicator of economic performance in most countries, as on-previous-month change or highlight quarter-on-quarter
described in Chapter 2. Where statistics are categorized changes. These provide an indication of price change over
according to their potential impact, the CPI is always pri- a short time frame, which would highlight those products
oritized. It follows therefore that it must be published, and with volatile prices, such as fuels. To avoid any confusion
otherwise disseminated, according to the policies, codes of in interpreting the results, it is very important to precisely
practice, and standards set for such data. In addition to hav- specify to which period the published inflation relates.
ing information on price movement at the total level, users 14.6 The index reference period (a month or a longer
often require information on the weights, methodology, and period, as described in Chapter 8) refers to a period in the past
price movements at a more disaggregated level. where the index equals 100. All subsequent months/periods
14.2 The CPI should therefore be: then have index numbers that reflect the relative change over
• Released as soon as possible time since the index reference period. Indeed, it is that index
that is used as the basic figure from which the other changes
• Made available to all users at the same time (exceptions
are calculated. The index reference period generally coincides
must be communicated in a transparent way; for example, with a routine update, but some countries choose to retain
if the central bank receives the results a few days before the old index reference period when updating weights. When
they are published on account of its monetary policy implementing changes to compilation methods, re-referencing
tasks, this should be mentioned in the press release) (setting the index back to 100) signals to users that a new index
• Released according to preannounced release calendars has been published. The reference period should be defined in
• Released separately from ministerial comment all publications and in the methodological explanation.
• Made available in convenient form to users 14.7 Indices and rates of change are usually shown only
to one decimal place in the press release, and in this case,
• Accompanied by methodological explanation
figures would be rounded. Rounding may lead to incon-
• Backed up by CPI compilers and economists who can sistencies. For example, if the unrounded index of t − 1 is
answer questions and provide further information 101.1459, and the unrounded index of the following month
14.3 Above all, the CPI should comply with the Fun- t is 102.7591, the rate of change compiled with unrounded
damental Principles of Official Statistics1 (United Nations indices will be 1.6 percent while the rate of change compiled
[UN], A/RES/68/261 from 29 January 2014). These prin- with rounded indices will be 1.7 percent. As long as this
ciples are published in several languages on the websites inconsistency can be explained (use full precision to calcu-
of the United Nations and the United Nations Economic late the rate of change), it is not a problem. An option could
Commission for Europe. They refer to dissemination and to be to make available to users both rounded and unrounded
all aspects of statistical work. These and other standards are figures: in the press release, the figures are rounded to one
discussed in paragraphs 14.40–14.42. decimal, but national statistical offices (NSOs) often publish
data on the website at or near full precision for analytical and
research purposes. To avoid this inconsistency, it is also pos-
Time-Series Presentation of Level sible to compile the published rate of change with rounded
and Change indices to one or two decimals. This is the case for the Euro-
pean Union (EU) Harmonised Indices of Consumer Prices
14.4 The presentation of the CPI data to users (for exam-
(HICPs) where the rate of change is compiled with two deci-
ple, on press releases disseminated by the national statistical
mals rounded indices.
office [NSO]) commonly focuses on the percentage change
14.8 Care must be taken to differentiate between per-
over 12 months (price movement between the current month
centages and index points. If in one month the index is,
and the same month one year earlier). The 12-month com-
for example, 200 and in the following month 201, then
parison provides an indication of price changes over a rea-
the change can be described as one index point (above the
sonably long time frame, by reference to periods that may
period when the index was set at 100) or as a half percentage
be expected to be similar year to year. Thus, seasonal fac-
point (where the previous month is taken as 100 percent).
tors are unlikely to be influential. It is also usual to compare
Both are valid, even if it is more common to refer to change
this annual change with the annual change shown one month
as a percentage.
14.9 The CPI is, by definition, an index and therefore not
1
https://unstats.un.org/unsd/dnss/gp/fundprinciples.aspx. a level or a series of absolute changes in prices. Nevertheless,
327
in the process of compiling the CPI, average prices can be cal- unadjusted series. There are other ways of smoothing monthly
culated for categories of products. It is thus possible to publish series, for example using three-month moving averages.
some average prices for groups of goods or services, and to 14.16 NSOs do not usually smooth the CPI series in their
show the upper and lower bands of the prices from which the published presentations. In general, consumer price changes
averages have been calculated. These averages may be use- are not so erratic from month to month as to disguise price
ful for some users, such as researchers. Average prices should trends. If there is an erratic change, the producers of the index
only be published for tightly defined, homogeneous item can usually explain the reasons for it. In any case, if season-
groups that are relatively similar (in quality) and for which the ally adjusted or smoothed series are published, it is important
variation coefficient is acceptable. It is also important to make to also publish the unadjusted series, so that the effect of the
it clear to users that average prices are a by-product of CPI adjustment process is clear to users who may wish to know
compilation and are not used to calculate price changes. what has happened to prices and whether or not the changes
14.10 The previous discussion refers not only to the can be ascribed to seasonal factors. Similarly, a full explana-
all-items CPI, but also to a more disaggregated level such tion should be given for the reasons why a particular seasonal
as regions of a country, population subgroups (for example, adjustment procedure has been followed.
pensioners), or to related or alternative measures of price
change. Related or alternative measures, and subaggregate Analysis of Contributions to
indices, are discussed in paragraphs 14.23–14.34. Change
14.17 Contributions to change help explain what groups
Seasonal Adjustment and of goods and services contributes most to inflation. These
Smoothing of the Index data are useful to better understand the sources of inflation
and can increase transparency.
14.11 The treatment of seasonal products and the esti- 14.18 The CPI is an aggregate of many different goods and
mation of seasonal effects are discussed in Chapter 11 of this services with prices changing at different rates, some of which
Manual and Chapter 4 of the publication Consumer Price may be going up while others are going down. The weights of
Index Theory. This chapter discusses the dissemination of these products or groups of products are different, resulting
adjusted or smoothed series. in a varying impact on the all-items index. If the weight of a
14.12 Many series of economic statistics are dissemi- group of products is high and its price trend strong, the impact
nated seasonally adjusted, as well as unadjusted. CPIs, how- on the overall inflation rate will therefore also be high. Many
ever, are not normally seasonally adjusted, although some users of the index want to know which goods or services have
countries do produce a seasonally adjusted CPI. Seasonally contributed most to changes in the total, and which prices may
adjusted CPIs are purely analytical and do not replace the be out of step with general price trends.
headline unadjusted index. Seasonal factors, for any series, 14.19 The CPI compilers are well placed to provide
are frequently recalculated using the most recent data, so analyses of the contributions to the overall price change,
seasonally adjusted series can be changed retrospectively, at the same time as the index is published. Sufficient detail
but unadjusted CPIs are not normally revised. should be made available so that users can understand what
14.13 In comparing one month with the same month a has happened to various groups of prices. In addition, to
year earlier, it is assumed that seasonal patterns are much the assist journalists and others working under time constraints,
same from one year to the next. There may be, however, excep- the CPI compiler should indicate the goods and services
tional months when the usual seasonal pattern is advanced or or group of products whose changes in price are the main
delayed. Such exceptional circumstances should be noted in contributors to the all-items CPI, and also goods and ser-
the monthly release as one of the likely causes of a change in vices whose changes in price are the most different from
the CPI or in one of its components. Even if seasonal patterns the aggregate. The statistics can be presented in the form
are much the same from one year to the next, there may be of tables and charts so that the trends may be compared.
months when relevant calendar effects differ from one year to Similarly, compilers should indicate any reasons for price
the next due to moving holidays such as Ramadan or Easter. changes that may not be immediately obvious but are never-
Seasonal expenditure patterns can be observed around these theless discernible from the published figures. For example,
holidays which could be considered seasonal in nature. if there has been a sharp price rise or fall one year earlier,
14.14 Changes over periods of less than a year are subject then it will affect the current year-on-year change, regard-
to seasonality. To differentiate seasonal patterns from other less of what is currently happening to prices.
factors, it is necessary to make estimates of seasonal effects 14.20 The formulas used to calculate the contributions
and to note them as factors that have contributed to changes to change are described in Chapter 9.
in the index. For this purpose, it is necessary to clearly iden-
tify seasonal products. The NSO also may possibly calculate Economic Commentary and
complementary indices, for example, a CPI that only contains
seasonal products and a CPI without seasonal products.
Interpretation of the Index
14.15 Although the CPI is not usually seasonally adjusted, 14.21 In undertaking an analysis of the CPI results, CPI
some variants of the CPI (for analytical purposes) may be sea- compilers must be objective so that users of the data may
sonally adjusted. These variants should be explained to the differentiate clearly between the figures and their interpre-
users and can be revised in retrospect if necessary. Seasonal tation. It is therefore essential that care is taken to avoid
adjustment usually leads to smoother series than the original expressing any judgment of the impact of current policy on
328
Publication, Dissemination, and User Relations
price changes or the possible implications of price changes clearly explain the methods used and define the purpose of
for future policies. Whether the figures should be seen as compiling these alternative indices. It must be clear to users
good or bad news is for the users to decide for themselves. how these indices can be used and why the NSO has pub-
The NSO’s role is to provide objective information so that lished this alternative aggregation.
users can form their own judgment from the perspective of 14.27 Countries commonly calculate price indices for
their own economic or political views. hundreds of products (for example, bread or footwear),
14.22 There are several ways of avoiding any apparent based on thousands of individual price records. The number
or real lapses in the analysis. The first, and perhaps the most of possible subaggregates is therefore very large. The choice
important, is to publish the figures independently of any of disseminated subaggregates is left to NSOs, according to
ministerial or other political comment. Another is to be con- the users’ needs.
sistent in the way in which the figures are presented; that is, 14.28 One kind of subaggregation is the grouping of
the data should be presented in much the same format every items or products that, when taken together, comprise the
month (see paragraphs 14.35–14.39). For example, tables all-items index. An important consideration here is the rela-
and charts should cover the same periods every month and tionship of the products within the subgroups. For example,
use the same baselines. an index may be presented for food and, under the head-
ing of food, indices may be presented for subgroups such as
cereals or vegetables.
Presentation of Related or 14.29 Subaggregates from different divisions of the
Alternative Measures Classification of Individual Consumption According to Pur-
pose (COICOP) can be combined to compile special aggre-
Core Inflation gates. For example, a special index for education can be
14.23 For the purpose of economic analysis, it is desirable compiled using weights and indices from different groups.
to construct measures of “core” or “underlying” inflation that Tuition and fees are part of Education Services (Division 10
exclude movements in the inflation rate that are attributable to of COICOP 2018), while school uniforms are part of Cloth-
transient factors. In other words, measures of core or underlying and Footwear (Division 03); textbooks and school sup-
ing inflation seek to measure the persistent or generalized trend plies are part of Recreation, Sport, and Culture (Division 09);
of inflation. For example, central banks use measures of the and school transport is under Transport (Division 07). An
general trend of inflation when setting monetary policy, and alternative index for education gives users a more complete
for this reason, economists and statisticians are increasingly picture of price change for education. Other examples would
interested in developing measures of underlying inflation. be to compile the CPI with and without production for own
14.24 Several methods can be used to derive a measure consumption in the weights. This analytical series meets the
of core or underlying inflation. Most measures of underly- needs of poverty economists and analysts. Other examples
ing inflation focus on reducing or eliminating the influence include CPIs compiled by income group or for the elderly.
of exceptionally volatile prices, or of exceptionally large 14.30 Other forms of subaggregate indices include
individual price changes. The most traditional approach is the dissemination of regional indices. For those countries
to exclude particular components of the CPI on a discretion- that compile a national index based on regional indices, the
ary basis. The items to be excluded would be based on the detailed regional indices should be disseminated. As with the
compiler’s knowledge of the volatility of particular items, national index, the monthly release should include data at a
depending on the country’s economic conditions. Items more aggregate level with detailed indices published on the
commonly excluded under this approach are fresh meat, NSO website. Data should be disseminated to the lowest level
fruit, and vegetables, and fuels and other energy products. possible, ideally down to the elementary aggregate level.
Many countries also exclude imported goods, government 14.31 One of the first considerations in presenting such
charges, and government-controlled prices. In some coun- subaggregate data for related products is consistency. There
tries, a calculation is made to exclude the effect of indirect should be a set of subaggregates for which indices are cal-
taxes such as the value-added tax. Of course, care must culated and presented each period. Users commonly attach
be taken not to exclude so many items that the remainder great importance to being able to continue their analysis for
becomes only a small and unrepresentative component of the most recent period.
the total. The chosen method for producing underlying infla- 14.32 Another consideration is international standard-
tion should be described in the metadata and publication. ization of the division of the index into groups of goods and
14.25 Other methods of deriving an underlying mea- services, which enables comparison between countries. Some
sure of inflation include smoothing techniques, for example, countries also have their own subaggregate groupings which
annualizing three-month average inflation. A more complex may precede the current international standard. The generally
method is to exclude outliers (that is, those items with the accepted international standard for the presentation of subag-
highest or lowest increases). gregates is COICOP, as discussed in Chapter 2. It is used, for
example, in the EU HICPs. Because COICOP defines groups
of items by the general purpose for which they are used (for
Alternative and Subaggregate Indices example, “transport” or “housing and household services”),
14.26 While publishing alternative aggregations of the it combines goods and services within the same subgroups.
CPI meets data user needs, this can also create confusion Where the national CPI is subaggregated by divisions other
for other data users. Users can be confused over what is the than the international standard, it is advisable either to pres-
headline, or official, rate of inflation and how these alterna- ent a breakdown also by COICOP or at least to show how the
tive measures compare to the headline index. NSOs must national classification compares to the international standard.
329
14.33 Another common type of subaggregate index is include a seasonally adjusted index for analytical purposes.
an index that excludes certain items. The core or underlying As indicated in the model, the presentation should contain
inflation index discussed in paragraphs 14.23–14.25 is an the following information:
example. Some countries publish, in addition to the all-items
CPI, an index or indices that exclude certain expenditures • Details of issuing office
(for example, a CPI without fuels) or merge the products • Date and time of release
differently (for example, a CPI for durable goods or a CPI • Percentage change in current month over the same month
for public services). one year earlier
14.34 In the presentation of all related or alternative
• Comparison with change in the previous month
measures, their definitions (for example, methodology or
differences with the all-items CPI) should be made clear. • Information on the product groups which contributed to
It is also advisable to give the reasons for their publication. the change and on any significant component price
Most importantly, it should not be suggested that the subag- • Reference to where more information (for example,
gregate index is more meaningful than the CPI. detailed results or metadata) can be found
• Date for next release and link to advance release calendar
Press Release, Bulletin, and Note that no judgments are offered on policy or economic
Methodological Statement reasons for the price change, and no judgment is given on
14.35 The model presentation of a CPI in Figure 14.1 is whether the change is good or bad.
an example of a press release for a fictitious country. Other 14.36 The format of the press release should be the
formats are possible. For example, the presentation might same from month to month. Using a consistent format is
Figure 14.1 Example of an Illustrative CPI Press Release
330
Figure 14.2 Model Note on Methodology—To Be International Standards Concerning

Included in Press Releases on CPI or on the Official the Dissemination of Consumer
Website
Price Indices
14.40 There are many international standards that
What is the consumer price index (CPI) measuring
apply, in general terms or specifically, to the CPI. The intro-
and how is it done?
duction to this chapter lists some of the broad principles that
The CPI measures inflation, the average change in the
are reflected in many of the international standards in some
prices of goods and services consumed by households.
form. One very general standard, but by its nature a funda-
Prices are collected each month from shops and
mental one, is the UN Fundamental Principles of Official
other suppliers of goods and services. The pattern of
Statistics. It refers not just to dissemination but to all aspects
household expenditure on these goods and services (the
of statistical work.
weights) is derived from a regular household budget
14.41 The International Monetary Fund (IMF) stan-
(or expenditure) survey. The prices and spending pat-
dards are particularly pertinent in regard to dissemination.
terns are then combined to calculate the price indices
As discussed in Chapter 13, two standards refer to statistics
for groups of goods and services and for the all-items
including the CPI, the Enhanced General Data Dissemina-
index.
tion System (e-GDDS) and the Special Data Dissemination
The all-items index and its component indices are
Standard (SDDS and SDDS Plus). The e-GDDS provides a
published each month in our CPI Bulletin. The Bul-
general framework, with some specific indicators defined as
letin also contains more information on the methodol-
“core” and others defined as “encouraged.” The Special Data
ogy used in calculating the CPI. A small booklet is also
Dissemination Standard is based on the GDDS framework
available. For a detailed account of the methodology
but is more demanding and applies to countries that choose
used in calculating the CPI, please see the CPI technical
to subscribe to it in writing to the IMF Board. Detailed infor-
manual. For more information on these publications,
mation on both standards is available on the IMF website.2
and how they may be obtained, please refer to our web-
14.42 Under the heading of quality, the e-GDDS refers
site at www.ous.gov or contact the telephone numbers
to the need to provide information on sources and methods,
given on the front of this press release.
as well as on component details and checking procedures.
Under the heading of integrity, it refers to declared stan-
dards of confidentiality, internal government access before
data release, identification of ministerial commentary, and
information on revision and advance notice of changes in
important to avoid appearing to choose a different format methodology. Under the heading access by the public, it
to indicate a preferred trend. Using the same format also refers to the need for preannounced release dates and simul-
allows for rationalization. taneous access for all users. In the tables of data categories,
14.37 Additional content (in the paper or digital version it refers to the CPI as a core indicator that should be issued
available on the NSO website) of the press release should monthly, within one to two months of the reference period.
provide information on the indices from which the percent- All these standards are reflected in the present Manual. The
age changes are calculated. Similar indices should also be International Labour Organization has also published Guide-
published for major groups of goods and services. Charts lines concerning dissemination practices for labor statistics
may also be used to illustrate, for example, which prices (ILO 1998), which are available on the International Labour
have contributed most or least to the all-items CPI. Organization website.3
14.38 If any other consumer price variant is also
being published, then the differences between the indi-
ces should be briefly explained, including any method- Timing of Dissemination of the
ological differences. Variants that require explanation Consumer Price Index
include, for example, a national index based on the EU
HICPs methodology, any regional indices, or versions of 14.43 The CPI should be released as soon as possible
the CPI that exclude particular components of consumer following the reference month, but it is equally important
expenditure. The press release should include a short note to release the index following a strict timetable. This time-
on methodology, similar to that given in Figure 14.2 or table of release dates should be published as far in advance
a link to the official website where the methodology is as possible. Having a fixed release date, published well
described. A more detailed explanation could be given in in advance, is important for three main reasons. First, it
a handbook. reduces the scope for manipulation of the release date for
14.39 Whether released in paper or digitally, the format political expediency. Second, it instills confidence in users
of the press release remains the same. The only difference is that the release date is as soon as possible and has not been
the mode of dissemination. Countries continue to move to a delayed (or brought forward) for purely political reasons. A
digital format and disseminate the monthly press release in
a digital format on the NSO website and via email to those
requesting or subscribing to the monthly release. Some 2
https://dsbb.imf.org/.
countries release the data electronically on the website, but 3
https://www.ilo.org/global/statistics-and-databases/standards-and-
also continue to officially release the data via a press confer- guidelines/guidelines-adopted-by-international-conferences-of-labour-
ence using the traditional paper format. statisticians/WCMS_087614/lang--en/index.htm.
331
third advantage is that users know when to expect the data on the NSO website costs little more than disseminating a
and can be prepared to use them. smaller amount.
14.50 Ideally, the CPI, accompanied by any essential
metadata, should be released simultaneously to the press
Timeliness of Release versus Data and other users. One way in which some NSOs are doing
Accuracy this is by making the press release available confidentially
14.44 The IMF’s e-GDDS, discussed in paragraphs to the journalists shortly before the official release time
14.41 and 14.42, recommends that the CPI be released each (maybe half an hour), providing them with the printed press
month within one to two months of the reference period. It release. Then, when the data are released, the journalists
is usual, in practice, for most countries to release the CPI in are permitted to release their reports or stories to the public.
the middle of the month after the month to which the index Care must be taken to ensure that the CPI is available to
refers. This is possible because, in many cases, the data are all users at the same time, regardless of the dissemination
collected mainly over a limited period in the middle of the medium used.
month to which the latest data refer. Thus, the CPI compilers 14.51 With the CPI as with other statistics, users should
have some time to check and analyze the data and to prepare be allowed access to as much data as possible for two main
the tables and charts in which the data will be disseminated. reasons. First, some users find the detailed data very useful
in their analysis. Second, access to the data inspires confi-
14.45 The accuracy of the index is particularly relevant
dence in the data.
because of the importance of the CPI, as discussed in more
14.52 Data should be disseminated to the lowest level
detail in Chapter 2. Also, partly because data are collected
possible, ideally the elementary aggregate level. Whether to
according to a strict schedule by price collectors it is rare
publish a particular elementary aggregate may depend upon
for data to be reported after the CPI is published, and partly
confidentiality issues, as addressed in paragraph 14.55.
because of the way in which the index is used in contracts,
14.53 In general, the CPI and its major components are
it is very rarely revised. This represents a major difference
deemed to be of such wide importance that they are made
between the CPI and other economic or socioeconomic available for free through press releases and on the NSO
statistics. website. While the goal should be to meet data user needs,
14.46 It follows that, although timeliness is important, special analyses made at the request of particular users may
the dissemination timetable must allow time for the data incur costs outside of the normal monthly production and
to be properly prepared and thoroughly checked. After the processing budget. Some countries will charge for special
release date, in most cases, a revision to the CPI would not analysis requests to defray the additional cost of preparing
be permissible, except in the case of a seasonally adjusted the data.
CPI. The HICPs of the EU are an exception as they are
revised from time to time. If any series is revised, then of
course the changes must be fully described and explained Confidentiality
when the new data are released. If there is any methodologi- 14.54 Although, in general, as much data as possible
cal change, this is usually known in advance. Users should should be made available to users, there are reasons why
be informed before any such change occurs. confidentiality is important in some instances. First, some
14.47 Best practice suggests that NSOs develop a revi- data are supplied by retailers and others on the understand-
sion policy for the CPI. Should an error be discovered that ing that the data will be used only for the purpose of aggre-
exceeds a defined threshold, the CPI would be revised. gation with other data and will not be released in any other
This revision policy allows for the correction of errors and form. This can be especially important where the data are
enhances transparency. Users should be made aware of the given voluntarily, as they often are. For example, in the case
revision policy in the metadata. when a single respondent provides data for a given elemen-
14.48 A possible compromise between accuracy and tary aggregate, publishing data at the elementary aggregate
timeliness can be the publication of flash estimates. A flash level would identify the data provider. To avoid any issues,
estimate is an early estimate of inflation computed from pre- some countries will obtain written permission from the
liminary data at a given time, released at the end of the cur- respondent authorizing the dissemination of the elemen-
rent month or at the very beginning of the following month, tary aggregate even though such publication would reveal
giving users a provisional figure very quickly. A flash esti- the identity of the data provider. Second, some elementary
mate is always followed by the official publication of the aggregates may be compiled based on a small number of
results, once the data are complete and all controls and anal- prices and could be deemed as not being sufficiently rep-
yses have been carried out. For example, Eurostat publishes resentative for publication purposes. Samples should be
flash estimates for the euro area. selected in such a way as to support the dissemination of all
elementary aggregates.
Access to Data
14.49 For a number of countries, the internet has become
Presentation of Methodology
the main dissemination medium, usually via the NSO web- 14.55 When the CPI is published each month, users
site. For the data producer, distribution costs are relatively want to see the main figures and to use them. Users do not
small. No printing or mailing costs are involved. As soon generally want to be burdened with explanations concerning
as the data are disseminated online, they are available to all the methodology underlying the data. Nevertheless, method-
users simultaneously. Disseminating a large amount of data ological explanations must be accessible to those who may
332
want them, and in forms which are comprehensible to users User Consultation
with different levels of expertise and interest. Any significant
changes in methodology must be fully explained and notified Different Uses of Consumer Price Indices
as far in advance as possible of the change being made.
14.61 It is important to explain the different uses of the
14.56 In addition to a brief statement in press releases
CPI to potential data users (CPI uses are discussed in Chap-
(see paragraphs 14.35–14.39), methodological explana-
ter 2). To this end, it is important to explain how the CPI is
tions should be available on at least two levels. Nonexperts
constructed and to provide details of its sources and meth-
should be able to refer to a booklet that explains the history,
ods. It is also important to make readily available explana-
principles, and practice underlying the CPI and any alterna-
tions of alternative indices or subindices, indicating how
tive measures which may also be available. A more thorough
their uses differ from the uses of the CPI.
explanation of sources and methods should also be readily
14.62 If there are different uses for CPIs, there are also
available for those users who are sufficiently interested and,
different users. It is useful to identify the different users to
for example, for compilers who may be working on the pro-
provide them the relevant information. The basic user would
duction of the CPI for the first time. The information must
be interested in knowing general results on an occasional
also be kept up to date despite the pressures to devote time
basis, while the central bank or an academic would be inter-
to the output at the expense of documentation. The ready
ested in detailed results over a longer period. The identi-
availability of a full explanation of sources and methods is
fication and classification of CPI users are useful to better
essential to confidence and trust in the CPI.
respond to their expectations.
14.57 Developing frequently asked questions (FAQs)
provides another level of information on the methodology
for users. FAQs provide concise responses to explain key Role of Advisory Committees
methodological questions and define important concepts. 14.63 For a statistical series as important as the CPI, it is
Overly technical language and explanations should be essential to organize an advisory committee, or a set of com-
avoided. FAQs should be written in plain language under- mittees, representing users and producers. There are many
standable by all users. FAQs should be posted on the NSO contentious issues in the construction of the CPI. In many
website and can be used to summarize index compilation countries, there have been fierce arguments about, for exam-
methods at the end of a press release. ple, which components should be included and excluded. The
role of an advisory committee is to consider and advise on best
Explaining Index Quality practice and methodologies. An equally important role of an
advisory committee is to enhance the credibility of the CPI.
14.58 Some users may regard the CPI results with sus- 14.64 In those countries where advisory committees
picion, as noted in Chapter 2. Metadata usually refers to have not been the norm, there may be a fear on the part of the
the “average consumer” or “average household,” but each CPI compiler that including nongovernmental participants
consumer and household has different expenditure patterns may raise expectations beyond what the NSO can deliver,
from the expenditure patterns of others and may notice thereby increasing dissatisfaction among the general public.
changes in one set of prices but not in others. More impor- In fact, the inclusion of nongovernmental users can lead to a
tantly, perhaps, there is criticism of the index because of greater understanding of the realities and the practical con-
suspicion that it does not keep track of newer types of goods straints to meeting theoretical needs. This is the usual experi-
and services, changes in the quality of products, or newer ence of NSOs that already have advisory bodies that include
types of retail outlets. See Chapter 2, paragraphs 2.47–2.49, representatives of all the major constituencies, both inside
for more details on inflation perceptions. and outside government. It is therefore important that the
14.59 It is important for CPI compilers to be willing advisory committee should include academics, employers,
to discuss these issues and to explain how the compilation trade union representatives, and others who have an interest
methods used address these issues. The producers of the in the index from differing points of view. It is also important
index must be open about their methods and the extent to that the reports of the advisory committee are made available
which they can overcome the potential or real problems that to the public in full and without undue delay.
have been identified. It follows that the NSO should publish
explanations concerning the quality aspects, whether or not
the quality of the index is currently being questioned.
Key Recommendations
14.60 Some countries develop personal inflation calcu- • NSOs should disseminate two key measures—the
lators that make data more relevant to individual users. A 12-month (current month to the same month in the previ-
personal inflation calculator allows a user to define their own ous year) and the month-on-month change (current to the
individual basket. Users are asked either to input expendi- previous month).
ture details for a group of items (monthly, annually, etc.) or
• Disseminate contributions to change so users better
to define specific items and the proportion of expenditure
understand what items have contributed most to inflation
made on each item. Based on these inputs, a personal infla-
in a given period.
tion rate is compiled and presented.4
• Written analysis or commentary should be neutral and
focus only on describing important or unusual move-
4
ments in a given period.
https://www.cbs.nl/en-gb/visualisaties/personal-inflation-calculator; https://
www.geostat.ge/personalinflation/; https://service.destatis.de/inflationsrechner/ • Provide alternative aggregations of CPI data to better
InflationCalculator.svg. meet data user needs.
333
• Develop and disseminate detailed metadata describing • Data should be published down to the elementary aggre-
CPI compilation methods. The release should include a gate level. Every effort should be made to design samples
short methodology note, and a more detailed description that support the detailed dissemination of data.
should be available on the NSO website. • To enhance transparency and increase user confi-
• A calendar of release dates should be disseminated at dence, NSOs should explain data quality issues and
least one year in advance. Once the release date has been provide details on the methods used to minimize
established, data should be released without delay. these biases.
• CPI data should be released simultaneously to all users. • User groups should be consulted and informed when
• Detailed data should be disseminated on the NSO web- making updates or revisions to the CPI. Advisory groups
site. These include detailed weights, indices, and changes can be a useful means of maintaining contact with key
(monthly and annual). data users.
334
A GLOSSARY OF MAIN TERMS
Acquisitions An approach to consumer price indices (CPIs) in which consumption is identified with the consumption
approach of goods and services acquired by a household in some period (as distinct from those wholly or partially
used up for purposes of consumption). Depending on the intended scope of the CPI, acquisitions may
include not only goods and services purchased but also those acquired by own-account production or as
social transfers in kind from government or nonprofit institutions.
Additivity At current prices, the value of an aggregate is identical to the sum of the values of its components.
Additivity requires this identity to be preserved for the extrapolated values of the aggregate and its com-
ponents when their current values in some period are extrapolated using a set of interrelated quantity
indices; or, alternatively, when the current values of an aggregate and its components in some period are
deflated using a set of interrelated price indices.
Aggregate A set of transactions relating to a specified flow of goods and services, such as the total purchases made
by resident households on consumer goods and services in some period. The term “aggregate” is also
used to mean the value of the designated set of transactions.
Aggregation The process of combining, or adding, different sets of transactions to obtain larger sets of transactions. The
larger set is described as having a higher level of aggregation than the sets of which it is composed. The term
“aggregation” is also used to mean the process of adding the values of the lower-level aggregates to obtain
higher-level aggregates. In the case of price indices, it means the process by which price indices for lower-
level aggregates are averaged, or otherwise combined, to obtain price indices for higher-level aggregates.
Axiomatic, or test The approach to index number theory that determines the choice of index number formula on
approach the basis of its mathematical properties. A list of tests is drawn up, each test requiring an index to possess
a certain property or satisfy a certain axiom. An index number may then be chosen based on the number
of tests satisfied. Not all tests may be considered to be equally important and the failure to satisfy one or
two key tests may be considered sufficient grounds for rejecting an index.
Average of price See Carli index.
relatives
Base period The base period is usually understood to mean the period with which all the other periods are compared.
The term, however, has different meanings in different contexts. Three types of base periods may be
distinguished:
• The price reference period—the period that provides the prices to which the prices in other periods
are compared. The prices of the price reference period appear in the denominators of the price rela-
tives, or price ratios, used to calculate the index. The price reference period is typically designated as
period 0.
• The weight reference period—the period, usually one or more years, of which the expenditures serve
as weights for the index. When the expenditures are hybrid (that is, the quantities of one period are
valued at the prices of some other period), the weight reference period is the period to which the
quantities refer. The weight reference period is typically designated as period b in this Manual.
• The index reference period—the period for which the value of the index is set equal to 100.
It should be noted that, in practice, the duration of the weight reference period for a CPI is typically a
year, or even two or more years, whereas the CPI is calculated monthly or quarterly, the duration of the
price reference period being a month, quarter, or year. The weight and price reference periods seldom
coincide in practice, at least when a CPI is first calculated, although the price and index reference periods
frequently coincide.
Basket A specified set of goods and services and their quantities. In a CPI context, the set may comprise the
actual quantities of consumption goods or services acquired or used by households in some period, or
may be made up of hypothetical quantities.
Basket price index A price index that measures the proportionate change between periods 0 and t in the total value of a
specified basket of goods and services: that is, ∑ p t q / ∑ p 0 q, where the term q is the quantities of the
specified goods and services. See Lowe index.
Bias A systematic tendency for the calculated CPI to diverge from some ideal or preferred index, resulting
from the method of data collection or processing, or the index formula used. See Cost of living bias and
Representativity bias.
Bilateral indices A type of index that measures the aggregate price change between two periods based on prices observed
in these two periods only. Depending on the index formula, the underlying quantities purchased or
expenditures can be from the price reference period (for example, Laspeyres), the current period (for
example, Paasche), the price reference and the current periods (for example, Fisher, Walsh, Törnqvist),
or any other past period (for example, Lowe, Young).
335
A Glossary of Main Terms
Bouncing A situation in which the set of prices for the second period is simply a reordering of the set of prices for
the first period, the price relatives thus being obtained by matching each price in the first period with
another price from the same set of prices.
Carli price index An elementary price index defined as a simple, or unweighted, arithmetic average of the sample price
relatives.
Carryforward A situation in which a missing price in some period is imputed as being equal to the last price observed
for that item.
Central product An internationally agreed-upon classification of goods and services based on the physical characteristics
classification of goods or on the nature of the services rendered.
Chain index An index number series for a long sequence of periods obtained by linking together index numbers
spanning shorter sequences of periods. See Linking; see also equation 6 of Appendix 6 to the Manual.
Chaining/chain The construction of a continuous price series by multiplying together price indices that have been con-
linking structed using different weight reference periods. The resulting index is referred to as a “chain index.”
Characteristics The physical and economic attributes of a good or service that serve to identify it and enable it to be clas-
sified. Some characteristics will help determine price and are commonly referred to as price-determining
characteristics.
Circularity An index number property such that, if j I k denotes a particular kind of price index that measures the
(transitivity) change between periods j and k, then j I l ≡ j I k . k I l where the indices j I l and k I l are of the same type.
When an index is transitive, the index that compares periods j and l indirectly through period k is identi-
cal with the index that compares j and l directly. One test that might be required under the axiomatic
approach is that the index number should be transitive.
COICOP The Classification of Individual Consumption by Purpose. It is the internationally preferred classifica-
tion for CPIs, household budget surveys, and the International Comparison Program.
Commensurability See Invariance to changes in the units of measurement test.
Commodity A test that might be used under the axiomatic approach that requires that, for a given set of products, the
reversal test price index should remain unchanged when the ordering of the products is changed.
Component A subset of the goods and services that make up some defined aggregate.
Conditional cost of A conditional cost of living index measures the change in the cost of maintaining a given utility level,
living index or standard of living, on the assumption that all the factors, except the prices covered by the index, that
influence the consumer’s utility or welfare (for example, the state of the physical environment) remain
constant. See Cost of living index.
Consistency in An index is said to be consistent in aggregation when the index for some aggregate has the same value
aggregation whether it is calculated directly in a single operation, without distinguishing its components, or whether
it is calculated in two or more steps by first calculating separate indices, or subindices, for its compo-
nents, or subcomponents, and then aggregating them, the same formula being used at each step.
Consumer price An official indicator constructed to measure the changes over time in the general level of prices for
index (CPI) consumer goods and services that households acquire, use, or pay for consumption. Its exact definition
may vary from country to country.
Consumption Consumption of goods and services is the act of completely using up the goods and services in a pro-
cess of production or for the direct satisfaction of human needs or wants. The activity of consumption
consists of the use of goods and services for the satisfaction of individual or collective human needs or
wants. Additionally,
• Intermediate consumption consists of the value of the goods and services consumed as inputs by a
process of production, excluding fixed assets whose consumption is recorded as consumption of fixed
capital; it is excluded from CPIs.
• Final consumption consists of goods and services used by individual households or the community to
satisfy their individual or collective needs or wants.
The System of National Accounts 2008 (2008 SNA) also defines individual and collective consumption
goods and services:
• An individual consumption good or service is one that is acquired by a household and used to satisfy
the needs or wants of members of that household.
• A collective consumption service is a service provided simultaneously to all members of the com-
munity or to all members of a particular section of the community, such as all households living in a
particular region; it is excluded from CPIs.
See Household final consumption expenditure.
Consumption of Goods or services that are consumed by the same household that produces them. The housing services con-
own production sumed by owner-occupiers fall within this category. If goods and services produced and consumed within
the same household are to be included in CPIs, prices must be estimated for them. Their inclusion or exclu-
sion depends on the intended scope of the CPI. It should be noted that activities undertaken by households
that produce services for their own use are outside of the scope of the System of National Accounts (SNA)
production boundary, except for services provided by owner-occupied dwellings and services produced by
employing paid domestic staff. The own-account production of all goods that are retained by their produc-
ers for their own final consumption or gross capital formation is included in the SNA production boundary.
336
Continuity The property whereby the price index is a continuous function of its price and quantity vectors.
Cost of living bias An alternative term used to describe Substitution bias.
Cost of living index An index that measures the change between two periods in the minimum expenditures that would be
required to achieve a constant standard of living (that is level of utility or economic well-being). As
consumers may be expected to change the quantities they consume in response to changes in relative
prices (see Substitution effect), the cost of living index is not a basket index. The quantities and expen-
ditures made in one or other, or possibly both, periods cannot usually be observed in a timely manner.
Cost of living indices cannot be directly calculated but may be approximated by superlative indices. See
Conditional cost of living index.
Coverage The set of goods and services of which the prices are actually included in the index. For practical rea-
sons, coverage may have to be less than the ideal scope of the index, that is, the set of goods and services
that the compilers of the index would prefer to include if it were feasible.
Current period, or In principle, the current period should refer to the most recent period for which the index has been com-
comparison period piled or is being compiled. The term is widely used, however, to mean the comparison period; that is,
the period that is compared with the base period, usually the price reference or index reference period.
It is also widely used simply to mean the later of the two periods being compared. The exact meaning is
usually clear in the context.
Current prices The actual prices prevailing in the period in question.
Current value The actual value of some aggregate in the period in question: the quantities in the period multiplied by
the prices of the same period.
Cutoff sampling A sampling procedure in which a predetermined threshold is established with all units in the universe at
or above the threshold being included in the sample, and all units below the threshold being excluded.
The threshold is usually specified with regard to the size of some relevant variable, the largest sampling
units being included and the rest given a zero chance of inclusion.
Deflation The division of the current value of some aggregate by a price index (described as a deflator) in order to
revalue its quantities at the prices of the price reference period.
Democratic index A form of CPI in which the expenditure proportions of each household are given equal weight in the
calculation of the index, irrespectively of the size of its expenditures. See Plutocratic index.
Discount A deduction from the list or advertised price of a good or a service that is available to specific custom-
ers under specific conditions. Examples include cash discounts, prompt payment discounts, volume
discounts, trade discounts, and advertising discounts.
Divisia index A price or quantity index that treats both prices and quantities as continuous functions of time. By dif-
ferentiating with respect to time, the rate of change in the value of the aggregate in question is partitioned
into two components, one of which is the price index and the other the quantity index. In practice, the
indices cannot be calculated directly, but it may be possible to approximate them by chain indices in
which indices measuring the changes between consecutive periods are linked together.
Domain An alternative term for the scope of an index. See Scope.
Domestic concept The use of weights covering all consumption expenditure on the national territory, regardless of the
nationality or normal residence of the consumer. See National concept.
Drift A chain index is said to drift if it does not return to unity when prices in the current period return to their
levels in the base period. Chain indices are liable to drift when prices fluctuate over the periods they cover.
Drobisch price The arithmetic average of the Laspeyres price index and the Paasche price index.
index
Durable consumer A consumer good that can be used repeatedly or continuously for purposes of consumption over a period
good of one year or more.
Dutot index An elementary price index defined as the ratio of the unweighted arithmetic averages of the prices in the
two periods compared.
Economic approach The economic approach to index number theory assumes that the quantities are functions of the
prices, the observed data being generated as solutions to various economic optimization problems.
In the CPI context, the economic approach usually requires the CPI to be some kind of cost of living
index.
Edgeworth price A basket price index in which the quantities in the basket are simple arithmetic averages of the quantities
index consumed in the two periods.
Editing The process of scrutinizing and checking the prices reported by price collectors. Some checks may be
carried out by computers using statistical programs written for the purpose.
Elasticity of A measure of the extent to which one product is substituted for another in response to relative price
substitution changes. A zero elasticity of substitution means that there is no substitution.
Elementary The lowest level of groups of goods or services for which expenditure weights are assigned and held
aggregate constant for a period of one year or more. An elementary aggregate should consist of relatively homo-
geneous set of goods or services, with similar end-uses and similar expected price movements. More
detailed weights to reflect the relative importance of individual price observations within elementary
aggregates may be applied and updated more frequently. The elementary aggregates are the building
blocks for the calculation of the higher-level indices.
337
Elementary price In general, an elementary index is a price index for an elementary aggregate, but it can also refer to a
index price index for a more detailed level below the elementary aggregate. Expenditure weights cannot usu-
ally be assigned to the price relatives for the sampled products within an elementary aggregate, although
other kinds of weighting may be explicitly or implicitly introduced into the calculation of elementary
indices. For example, scanner data could be used to develop a more detailed weighting program within
an elementary aggregate. Three examples of unweighted elementary index number formulae are the
Carli, the Dutot, and the Jevons.
Expenditure weights See Weights.
Explicit (or direct) A direct estimate of the value of the quality difference between the old and new product used to adjust
quality adjustment one of the prices accordingly. Pure price change is then estimated as the difference in the adjusted prices.
See Implicit quality adjustment.
Factor reversal test Suppose the prices and quantities in a price index are interchanged to yield a quantity index of exactly
the same functional form as the price index. Under the axiomatic approach, the factor reversal test
requires that the product of this quantity index and the original price index should be identical to the
proportionate change in the value of the aggregate in question.
Fisher price index The geometric average of the Laspeyres price index and the Paasche price index. It is a symmetric index
and a superlative index.
Fixed-basket A time series of basket indices that all use the same basket; see equation 4 of Appendix 6 to the Manual.
indices In a CPI context, the fixed basket usually consists of the total quantities of goods and services consumed
by the designated set of households over a period of a year or more.
Fixed-weight An abbreviated description for a series of weighted averages of price relatives that all use the same
indices weights; see equation 13 of Appendix 6 to the Manual. The weights are usually either actual or hybrid
expenditure shares.
Geometric A weighted geometric average of the price relatives using the expenditure shares of the price reference
Laspeyres index period as weights. Also called Logarithmic Laspeyres index.
Goods Physical, produced objects for which a demand exists, over which ownership rights can be established
and for which ownership can be transferred between institutional units by engaging in transactions on
the market.
Hedonic method A regression model in which the market prices of different products are expressed as a function of
their characteristics. Nonnumerical characteristics are represented by dummy variables. Each regression
coefficient is treated as an estimate of the marginal contribution of that characteristic to the total price.
The estimates may be used to predict the price of a new product for which the mix of characteristics is
different from that of any product already on the market. The hedonic method can therefore be used to
estimate the effects of quality changes on prices.
Higher-level index An aggregate index as distinct from an elementary index.
Household budget Sample surveys of households in which the households are asked to provide data on, or estimates of, the
surveys value of the goods and services acquired, paid, and used for consumption as well as other purposes over
a given period of time. Also referred to household income and expenditure surveys.
Household final It consists of the expenditure, including expenditure whose value must be estimated indirectly, incurred
consumption by resident households on individual consumption goods and services, including those sold at prices that
expenditure are not economically significant and including consumption goods and services acquired abroad. They
exclude expenditures incurred by governments or nonprofit institutions on goods or services provided to
households as free social transfers in kind.
Households A household is a group of persons who share the same living accommodation, who pool some, or all, of
their income and wealth and who consume certain types of goods and services collectively, mainly hous-
ing and food. Most countries choose to exclude groups of persons living in large institutional households
(barracks, retirement homes, and so on) from the scope of their CPIs.
Hybrid values or Hypothetical values, or expenditures, in which the quantities are valued at a different set of prices from
expenditures those at which they were actually bought or sold: for example, when the quantities purchased in an ear-
lier period, such as b, are valued at the prices prevailing in a later period, such as 0.
Hybrid weights Weights defined as hybrid value, or hybrid expenditure, shares.
Identity test A test under the axiomatic approach that requires that, if each and every price remains unchanged
between the two periods, the price index must equal unity.
Implicit (or Estimating the pure price change component of the price difference between the old and new products
indirect) quality based on the price changes observed for similar products. The difference between the estimate of pure
adjustment price change and the observed price change is considered as change due to quality difference. See
Explicit quality adjustment.
Imputed price1 The price assigned to a variety for which the price is missing in a particular period. The term “imputed
price” may also refer to the price assigned to an variety that is not sold on the market, such as a good
or service produced for own consumption, including housing services produced by owner-occupiers, or
one received as payment in kind or as a free transfer from a government or nonprofit institution.
1
This definition differs from that used by the 2008 SNA.
338
Indexation The periodic adjustment of the money values of some regular scheduled payments based on the move-
ment of the CPI or some other price index. The payments may be wages or salaries, social security or
other pensions, other social security benefits, rents, interest payments, and so on.
Index reference The period for which the value of the index is set at 100.
period
Institutional unit An economic entity that is capable, in its own right, of owning assets, incurring liabilities, and engag-
ing in economic activities and transactions with other entities. Households are institutional units. Other
kinds of units include enterprises and governments.
Institutional A group of individuals staying in an institution for long periods of time, such as retirement homes, mili-
households tary accommodation, and boarding schools, and sharing resources. Such people are treated as belonging
to private “institutional” households.
Invariance to A test under the axiomatic approach that requires that the price index does not change when the units of
changes in the units quantity to which the prices refer are changed: for example, when the price of some drink is quoted per
of measurement liter rather than per pint. This test is also described as the commensurability test.
test
Invariance to A test under the axiomatic approach that requires that the price index remains unchanged when all the
proportional change base period quantities, or all the current period quantities, are multiplied by a positive scalar.
in current or base
quantities test
Inverse A test that may be invoked under the axiomatic approach that requires that, if all the base period prices
proportionality in are multiplied by the positive scalar l, the new price index is 1/l times the old price index.
base year prices test
Item or product The deliberate replacement of a sampled item, or product, for which prices are being collected, by
rotation another product before the replaced product has disappeared from the market or individual outlet. It is
designed to keep the sample of products up to date and reduce the need for forced replacements caused
by the disappearance of products.
Jevons price index An elementary price index defined as the unweighted geometric average of the sample price relatives.
Laspeyres price A basket index in which the basket is composed of the actual quantities of goods and services in the
index earlier of the two periods compared, the price reference period; see equation 3 of Appendix 6 to the
Manual. It can also be expressed as a weighted arithmetic average of the price relatives that uses the
expenditure shares in the earlier period as weights; see equations 7–10 of Appendix 6 to the Manual.
The earlier period serves as both the weight reference period and the price reference period.
Linking Splicing together two consecutive sequences of price observations, or price indices, that overlap in one
or more periods. When the two sequences overlap by a single period, the usual procedure is simply to
rescale one or the other sequence so that the value in the overlap period is the same in both sequences
and the spliced sequences form one continuous series. See equation 6 of Appendix 6 to the Manual.
Lowe index A price index that measures the proportionate change between periods 0 and t in the total value of
a specified basket of goods and services; that is, ∑ p t q / ∑ p 0 q , where the term q is the specified
quantities. The basket does not necessarily have to consist of the actual quantities in some period. See
Appendix 6 to the Manual. This type of index is described in the Manual as a Lowe index after the
index number pioneer who first proposed this general type of index. The class of indices covered by
this definition is very broad and includes, by appropriate specification of the terms q, the Laspeyres,
Paasche, Edgeworth, and Walsh indices. Lowe indices are widely used for CPI purposes, the quanti-
ties in the basket typically being those of some weight reference period b, which precedes the price
reference period 0.
Lower-level index An elementary index as distinct from an aggregate index.
Matched products The practice of pricing exactly the same product in two or more consecutive periods. It is designed
or models to ensure that the observed price changes are not affected by the quality change. The change in price
between two perfectly matched products is described as a pure price change.
Model A specific variety whose characteristics are regularly updated. See Variety.
Mean value test for A test under the axiomatic approach, which requires that the price index should lie between the smallest
prices price relative and the largest price relative.
Modified Lowe A version of the Lowe index that compiles the index based on short-term price changes rather than long-
index term changes. This approach makes it easier for national statistical offices to introduce replacement vari-
eties in the sample when a sampled variety disappears. The short-term approach also facilitates quality
adjustments because only the current and previous period prices are needed in order to introduce a new
variety into the index. See Lowe index.
Modified Young A version of the Young index that compiles the index based on short-term price changes rather than
index long-term changes. This approach makes it easier for national statistical offices to introduce replacement
varieties in the sample when a sample variety disappears. The short-term approach also facilitates qual-
ity adjustments because only the current and previous period prices are needed in order to introduce a
new variety in the index. See Young index.
339
Multilateral indices A type of index that measures the aggregate price change between two periods based on prices observed
in multiple periods including the two comparison periods. Multilateral indices were developed for price
comparisons across countries (purchasing power parities) and have been adapted to compare prices
over time. For price comparisons over time, multilateral index formulas are mainly used with scanner
data. In this context, their main advantage is to avoid chain drift. The most common multilateral index
formulas in CPIs are the Gini–Eltetö–Köves–Szulc, the Geary–Khamis, and the time product dummy.
See Chapter 10 for further details.
National concept The use of weights covering the expenditure of residents of a country, regardless of whether the expen-
diture is made within or outside of the country. See Domestic concept.
Nonprobability The deliberate (that is, nonrandom) selection of a sample of outlets and products on the basis of the knowl-
sampling edge or judgment of the person responsible. Also known as purposive sampling and judgmental sampling.
Outlier A term that is generally used to describe any extreme value in a set of survey data. In a CPI context, it
is used for an extreme value of price or price relative that requires further investigation or that has been
verified as being correct.
Owner-occupied Dwellings owned by the households that live in them. The dwellings are fixed assets that their owners
housing use to produce housing services for their own consumption, these services being usually included within
the scope of the CPI. The value of the services provided may be imputed by the rents payable on the mar-
ket for equivalent accommodation or by user costs. See Rental equivalence and User cost. In the System
of National Accounts (SNA) framework, the production of housing services for their own final consump-
tion by owner-occupiers has always been included within the production boundary in national accounts,
although it constitutes an exception to the general exclusion of own-account service production.
Paasche price index A basket index in which the basket is composed of the actual quantities of goods and services in the
later of the two periods compared. The later period serves as the weight reference period and the earlier
period as the price reference period. The Paasche index can also be expressed as a weighted harmonic
average of the price relatives that uses the actual expenditure shares in the later period as weights. See
equations 7–11 of Appendix 6 to the Manual.
Plutocratic index A form of CPI in which the weights are based on total aggregated expenditure values rather than average
household expenditure proportions.
Price reference The period of which the prices appear in the denominators of the price relatives. See Base period.
period
Price relative The ratio of the price of a variety in one period to the price of that same variety in some other period.
Price updating A procedure whereby the quantities in an earlier period are revalued at the prices of a later period. The
resulting expenditures are hybrid. In practice, the price-updated expenditures may be obtained by mul-
tiplying the original expenditures by price relatives or price indices.
Probability A sampling procedure whereby each unit in the universe has a probability of selection proportional to
proportional to size the size of some known variable, such as the value of the sales of an outlet.
sampling
Probability The random selection of a sample of units, such as outlets or products, in such a way that each unit in the
sampling universe has a known nonzero probability of selection.
Proportionality in A test under the axiomatic approach that requires that, if all current period prices are multiplied by the
current prices test positive scalar λ, the new price index is λ times the old price index.
Purchaser’s price The purchaser’s price is the amount paid by the purchaser, excluding any value-added tax or similar tax
deductible by the purchaser, in order to take delivery of a unit of a good or service at the time and place
required by the purchaser. The purchaser’s price of a good includes any transport charges paid separately
by the purchaser to take delivery at the required time and place.
Pure price change The change in the price of the same variety; or the change in the price after adjusting for any change in
quality.
Quality adjustment An adjustment to the price of a variety of which the characteristics have changed over time. Quality
adjustments are designed to remove the part of the observed price that is due to differences in the price-
determining characteristics. In a CPI context, the adjustment is needed when the price of a replacement
variety must be compared with the price of the variety it replaces. In practice, the required adjustment
can only be estimated. Different methods of estimation, including hedonic methods, may be used in dif-
ferent circumstances. See Explicit quality adjustment and Implicit quality adjustment.
Quantity relative The ratio of the quantity of a product in one period to the quantity of that same product in some other period.
Quantity weights A term sometimes used to describe the quantities in the basket. However, expenditures rather than quan-
tities act as weights for price relatives. See Weights.
Quota sampling Sample selection using judgmental procedures with respect to known characteristics such as product
group or outlet type. The sample is drawn so as to contain the same proportions as in the total population
of products, items, or outlets.
Ratio of averages See Dutot index.
Rebasing It refers to changing the weight reference period, price reference period, or index reference period. The
weight reference period, price reference period, and index reference period may be changed separately
or at the same time.
340
Reference The set of households included within the scope of the index.
population
Rental equivalence The estimation of the imputed rents payable by owner-occupiers on the basis of the rents payable on the
market for accommodation of the same type.
Replacement A variety chosen to replace a variety for which prices have been collected previously, either because the
variety previous variety has disappeared altogether or because it accounts for a diminishing share of the sales of
the outlet, or the expenditures within the elementary aggregate.
Representative A variety, or category of varieties, that accounts for a significant proportion of the total expenditures
variety within an elementary aggregate, and for which the average price change is expected to be close to the
average for all varieties within the aggregate.
Representativity Bias in a basket index that results from the use of quantities that are not representative of the two
bias periods compared; that is, that systematically diverge from the average quantities consumed in
the two periods. For example, representativity bias may result from the use of an old, out-of-date
basket which deviates systematically from the baskets in both the periods compared. In practice,
representativity bias tends to be similar to substitution bias, as it is attributable to the same eco-
nomic factors.
Reweighting Replacing the weights used in an index by a new set of weights.
Revision Often refers to changing index weights and implementing new calculation or compilation methodologies.
Sample Maintaining and adding to the sample of outlets, items, and varieties in the survey panel to ensure
augmentation that they continue to be representative of the population of outlets. A fixed sample of outlets tends to
be depleted over time, as outlets cease trading or stop responding. Including new outlets also tends to
facilitate the inclusion of new products in the CPI.
Sampled price The price collected for a sampled product in a specific outlet at a specific time, sometimes described as
a price quote.
Sample rotation Limiting the length of time that outlets and products are included in the price surveys by dropping a
proportion of them, or possibly all of them, after a certain period of time and selecting a new sample of
outlets and products. Rotation is designed to keep the sample up to date.
Sampling frame A list of the units in the universe from which a sample of units can be selected. The list may contain
information about the units, which may be used for probability proportional to size sampling. Examples
of lists that may be used for retail outlets are business registers, telephone directories (“yellow pages”),
local authority records, or trade directories. Such lists may not cover all the units in the designated uni-
verse and may also include units that do not form part of that universe.
Scanner data Detailed data on sales of products obtained by scanning the bar codes for individual varieties at elec-
tronic points of sale in retail outlets. The data can provide detailed information about quantities, charac-
teristics, and values of varieties sold, as well as their prices. Scanner data constitute a rapidly expanding
source of data with considerable potential for CPI purposes.
Scope The set of products for which the index is intended to measure the price changes. The scope of a CPI
will generally be defined with regard to a designated set of consumption goods and services purchased
by a designated set of households. In practice, certain goods and services or households may have to
be excluded because it is too difficult, time-consuming, or costly to collect the relevant data on expen-
ditures or prices: for example, illegal expenditures. The coverage of an index denotes the actual set of
products included, as distinct from the intended scope of the index.
Seasonal products They are products that either are not available on the market during certain seasons or periods of the year
or are available throughout the year but with regular fluctuations in their quantities and prices that are
linked to the season or time of the year.
Services They are the result of a production activity that changes the conditions of the consuming units or facili-
tates the exchange of products or financial assets.
Specification A description or list of the characteristics that can be used to identify an individual sampled vari-
ety to be priced. A tight specification is a fairly precise description of an item intended to narrow
the range of varieties from which a price collector might choose, possibly reducing it to a unique
variety, such as a particular brand of television set identified by a specific code number. A loose
specification is a generic description of a range of items that allows the price collector some discre-
tion as to which particular variety or model to select for pricing, such as color television sets of a
particular size.
Stochastic approach The approach to index number theory that treats the observed price relatives as if they were a random
sample drawn from a defined universe for which the mean can be interpreted as the general rate of infla-
tion. The sample mean provides an estimate of the rate of inflation.
Substitute A product of which the characteristics are similar to those of another product and that can be used to
meet the same kinds of consumer needs or wants.
Substitution The replacement of products by substitutes, typically in response to changes in relative prices. Rational
utility-maximizing consumers, as price takers, typically react to changes in relative prices by reducing,
at least marginally, their consumption of goods and services that have become relatively dearer and
increasing their consumption of substitutes that have become relatively cheaper. Substitution results in
a negative correlation between the quantity and price relatives.
341
Substitution bias This is generally understood to be the bias that results when a basket index is used to estimate a cost of
living index, because a basket index cannot take account of the effects on the cost of living of the substi-
tutions made by consumers in response to changes in relative prices. In general, the earlier the period of
which the basket is used, the greater the upward bias in the index; see Representativity bias.
Substitution effect The effect of substitution on the value of an index.
Superlative index A type of index formula that can be expected to approximate to the cost of living index. An index is said
to be exact when it equals the true cost of living index for consumers whose preferences can be repre-
sented by a particular functional form. A superlative index is then defined as an index that is exact for
a flexible functional form that can provide a second-order approximation to other twice-differentiable
functions around the same point. The Fisher, the Törnqvist, and the Walsh price indices are examples of
superlative indices. Superlative indices are generally symmetric indices.
Symmetric index An index that treats both periods symmetrically by attaching equal importance to the price and expen-
diture data in both periods. The price and expenditure data for both periods enter into the index formula
in a symmetric way.
System of National A coherent, consistent, and integrated set of macroeconomic accounts, balance sheets, and tables based on
Accounts (SNA) a set of internationally agreed-upon concepts, definitions, classifications, and accounting rules. Distribu-
tion and use of income accounts and household final consumption expenditure form part of the SNA. The
expenditure data are some of the sources that are used to estimate expenditure weights for CPI purposes.
Time reversal An index number property such that, if j I k denotes a particular kind of price index formula that mea-
sures the change from period j to period k, then j I k ≡ 1 / k I j where k I j measures the change from period
k to period j. When an index has this property, the change is the same whether it is measured forward
from the first to the second period or backward from the second to the first period. An index may be
required to satisfy the time reversal test under the axiomatic approach.
Törnqvist price A symmetric index defined as the weighted geometric average of the price relatives in which the weights
index are simple arithmetic averages of the expenditure shares in the two periods. It is a superlative index. Also
known as the Törnqvist–Theil price index.
Transitivity See Circularity.
Unit value or The unit value of a set of homogeneous products is the total value of the purchases/sales divided
average value by the sum of the quantities. It is therefore a quantity-weighted average of the different prices at which
the product is purchased/sold. Unit values may change over time as a result of a change in the mix of the
products sold at different prices, even if the prices do not change.
Updating Changing the index weights. See Revision and Rebasing.
User cost The cost incurred over a period of time by the owner of a fixed asset or consumer durable as a consequence
of using it to provide a flow of capital or consumption services. User cost consists mainly of the depreciation
of the asset or durable (measured at current prices and not at historic cost) plus the capital, or interest, cost.
Uses approach An approach to CPIs in which the consumption in some period is identified with the consumption goods
and services actually used up by a household to satisfy their needs and wants (as distinct from the con-
sumption goods and services acquired). In this approach, the consumption of consumer durables in a
given period is measured by the values of the flows of services provided by the stocks of durables owned
by households. These values may be estimated by the user costs.
Value Price times quantity. The value of the expenditures on a set of homogeneous products can be factored uniquely
into its price and quantity components. Similarly, the change over time in the value of a set of homogeneous
products can be factored uniquely into the change in the unit value and the change in the total quantities. There
are, however, many different ways of factoring the change over time in the value of a set of heterogeneous
products into its price and quantity components, a phenomenon that gives rise to the index number problem.
Variety The individual product for which prices are collected during the collection period. It includes the detailed
specification of the product or item observed.
Walsh price index A basket index in which the quantities are geometric averages of the quantities in the two periods; see
Appendix 6 to the Manual. It is a symmetric index and a superlative index.
Weight reference The period of which the expenditure shares serve as the weights for a Young index, or of which the
period quantities make up the basket for a Lowe index. There may be no weight reference period when the
expenditure shares for the two periods are averaged, as in the Törnqvist index, or when the quantities are
averaged, as in the Walsh index. See Base period.
Weighted An index defined as a weighted arithmetic average of the price relatives: namely, ∑w( p t / p 0 ), where
arithmetic average the weights w sum to unity.
index
Weights A set of numbers summing to unity that are used to calculate averages. In a CPI context, the weights are
generally actual or hybrid expenditure shares that sum to unity by definition. They are used to average
price relatives or elementary price indices; see Appendix 6 to the Manual. Quantities of different kinds
of products are not commensurate and not additive. They cannot be used to average elementary price
indices. The quantities that make up a basket should therefore not be described as quantity weights.
Young index A Young index is a weighted arithmetic average of the price relatives, ∑w( p t / p 0 ), in which
the terms w refer to the actual expenditure shares of period b, the weight reference period; that is,
w = s b = p b q b / ∑ p b q b . It is a weighted version of the Carli index.
342
Appendix 1 transactions only. This household final monetary consump-
tion expenditure (HFMCE) is thus a narrower concept than
The Harmonised Index household final consumption expenditure, which includes
both monetary and nonmonetary transactions. The HFMCE
of Consumer Prices is a fundamental concept for the HICP. It delineates the
(European Union) scope of the HICP, namely that it includes only expenditure
on consumption products by the household sector. And that
it excludes all capital expenditure, such as the purchase of
land or financial assets like stocks and shares, as well as all
imputed transactions, own production, and barter.
Introduction
The Harmonised Index of Consumer Prices (HICP) is a spe- Coverage
cific inflation measure that is developed within the European The domestic concept is used to define the geographic
Union (EU) to result in indices that can be directly compared coverage of the HICP. This takes into account all HFMCE
and aggregated across countries. The HICP is the outcome within the economic territory of a country, whether made by
of major harmonization work by the statistical office of the resident or nonresident households. Consumption expendi-
EU (Eurostat) and the EU countries. ture incurred by residents when they are outside the country
The HICP is a cost of goods index; that is, it measures the of residence is excluded from the HICP, while expenditure
changing cost of a fixed basket of products over time. The incurred by visitors from other countries is included. Bear-
production of the HICP and its methodology are governed by ing in mind the principal use of the HICP as an indicator for
EU law, which gives the common definitions and concepts monetary policy purposes, there are two main reasons for
to follow. The key HICP aggregates are the euro area index, using the domestic concept for the coverage of the HICP:
covering the countries whose currency is the euro, the EU
index, and the national HICP for each of the EU countries. • By confining expenditure to that incurred within an eco-
Eurostat publishes also HICPs for countries that are not part nomic territory, the resulting HICPs cover only those
of the EU, which produce the data on voluntary basis. The price changes which national/euro area monetary policies
national statistical offices produce the national HICP, while can directly influence.
Eurostat produces the country-group aggregates. • Consistent aggregation of national HICPs: if a European
The HICP serves two main purposes: aggregate HICP (for example, for the euro area) is com-
pared with another country or economic bloc, it must be
• For quantifying the price stability in the European Cen- certain to represent the whole of consumer price inflation
tral Bank’s monetary policy strategy. Maintaining price within the euro area and none of it must be double-counted.
stability is the primary objective of the European Central
Bank and the national central banks of the euro area, as Using the domestic concept ensures that these conditions are
set out in the Treaty on the Functioning of the European met. An alternative to the domestic concept is the national
Union. The ECB’s Governing Council has defined price concept, where all expenditure incurred by residents of a
stability as a year-on-year increase in the HICP for the country, whether nationals or nonnationals, is measured,
euro area of below 2 percent. The Governing Council has regardless of whether it is incurred inside or outside the eco-
clarified that, in the pursuit of price stability, it aims to nomic territory. It should be noted that both the domestic
maintain inflation rates below, but close to, 2 percent over and national concepts at the world level would produce the-
the medium term. oretically the same results with regard to aggregate expen-
• For assessing the price stability criterion, which is one of diture. However, at the EU and euro area levels they would
the convergence criteria used to evaluate if a country can not, due to the expenditure of EU/euro area residents out-
join the euro area. side the EU and the expenditure of non-EU residents within
the EU. In practical terms, this makes the domestic concept
In addition to these specific EU uses, the HICP may be used, the easier one to implement, as the national concept would
like other consumer price indices, for economic analysis and require detailed information on residents’ expenditure and
for indexing for example contracts and wages. The HICP prices paid outside the economic territory.
has been produced and published since March 1997. The HICP product coverage is based on the concept of
HFMCE and the 1999 version of the Classification of Indi-
vidual Consumption According to Purpose (COICOP) to
Concepts and Definitions which an additional level of (five-digit) subclasses has been
added. Some categories of COICOP are excluded either in
Household Final Monetary Consumption principle or on practical grounds from the HICP coverage:
Expenditure
In general, the concepts, definitions, and conventions adopted 02.3 Narcotics
in the HICP are as far as possible consistent with those used 04.2 Imputed rentals for housing
in the United Nations System of National Accounts 2008 09.4.3 Games of chance
(2008 SNA) and its EU version, the European System of 12.2 Prostitution
Accounts 2010 (ESA 2010). Among these concepts is that of 12.5.1 Life insurance
household final consumption expenditure. For the HICP this 12.6.1 Financial intermediation services indirectly
has been adapted to refer to the part that occurs in monetary measured
343
APPENDIX
In principle, “narcotics” and “prostitution” belong to not, be price updated between year t ‒ 2 and year t ‒ 1. The
HFMCE but they are not covered in the HICP for practical objective is to obtain the best possible estimate of the expen-
reasons. “Games of chance” also falls within the scope of diture shares for the year t ‒ 1. If goods and services are per-
the HFMCE. However, they are excluded from the HICP fect complements (that is, there is no substitution between
because no harmonized method for their treatment has yet them and they are consumed in fixed proportions), the best
been agreed upon. approximation would be the price-updated weights. If goods
“Owner-occupied housing” costs are excluded from the and services are substitutes at such a rate that expenditure on
HICP product coverage because no available method is both one product relative to another is independent of the relative
conceptually consistent with the HICP and able to provide prices, the preferred approach would be not to price update.
timely data with monthly frequency. The rental equivalence EU countries must carry out an annual review of weights to
approach is currently considered noncompatible with the determine if there have been any important and sustained
HICP on conceptual grounds as that method uses imputa- market developments, for example, the appearance of new
tions and not actual monetary transactions. Eurostat has significant products, and adjust the weights if necessary.
instead established a separate index of owner-occupied These representative annual weights need to be adjusted
housing costs using the net acquisition approach. This index with an appropriate price change to reflect the prices of
meets the requirement of monetary transactions, but can be December of the year t − 1.
criticized for including assets (land and dwellings). This is
one of the reasons for not including this index into the HICP. Sampling
In addition, it is apparent that net acquisition-based owner-
occupied housing price indices cannot currently be produced The HICP is a sample statistic that represents the change in
monthly and timely in the EU countries. Eurostat is continu- prices, on average, over the target universe, which consists
ing the methodological work required for the integration of of all the transactions that fall within the scope of the HICP
an owner-occupied housing price index into the HICP. in the two periods being compared. Given the differences in
The HICP also excludes financial intermediation services the national markets and populations of the EU countries, it
indirectly measured consumed by households (that is, those would not be possible to impose a uniform sampling struc-
parts of financial services for which an explicit (monetary) ture across all countries. Nevertheless, certain minimum
charge cannot be identified) because it is considered as an standards must be followed. Countries are required to ensure
imputed transaction. Note that financial services that have that:
explicit charges, for example, annual charges for credit • Each COICOP subclass contains a sufficient number of
cards (excluding interest charges), bank charges for money elementary aggregates to represent the diversity of prod-
transfers, or currency exchange commissions, are included ucts within the category.
in the HICP. • The number of prices recorded within each elementary
Insurance services are within the scope of the HFMCE. aggregate is sufficient to represent the price movements
However, “life insurance” is excluded from the HICP cover- in the population.
age. Premiums paid for life insurance, including pension-
• All product categories that form a significant part of total
funding services, are regarded as savings and thus are not
part of the HICP. “Life insurance” is excluded because it is consumption expenditure (at least one part per thousand)
not possible to separate out the implicit service charge for must be covered in the sample.
the insurance component from the implicit service charge The HICP sample must be designed in such a way as to
for the investment component. Nonlife insurance services ensure that it is representative of all transactions: the result
are, however, included in the HICP. Although for the HICP is a target sample, which must be maintained to ensure that it
weights nonlife insurance services are measured with regard remains valid in the face of continual market changes.
to implicit service charge, it is actually the gross insurance
premiums that enter the index compilations, for practical Prices
reasons.
The prices used in the HICP should be purchasers’ prices,
which are the prices actually paid by households. Product-
Weights related taxes such as the value-added tax and other sales
The HICP weights for published subindices are updated taxes and excise duties are included and any subsidies
every year. The weights should represent the expenditure in deducted. Discounts such as sales prices should be taken
the previous year and take account of preliminary national into account. Interest payments or service charges added
accounts data on the HFMCE. In practice, this means that under credit arrangements, and any extra charges incurred
compilers should estimate new subindex weights using the as a result of failing to pay within the period specified at
latest available data from the national accounts, normally the time of purchase, are disregarded. Purchaser prices also
preliminary estimates relating to the calendar year t − 2 include all unavoidable additional costs such as booking and
(where t is the current year). delivery charges, which are typically associated with inter-
To obtain weights for lower levels of aggregation down net purchases. However, it is not always possible to observe
to the most detailed product level, when national accounts the actual transaction price; so, in practice, it is usual to
do not provide data, other data sources such as household record the offer price or shelf price.
budget surveys, administrative data, retail sales, market The HICP follows the acquisition approach to the record-
research data, and data derived from scanner data can also ing of prices. In this approach, the acquisition is deemed
be used. The observed year t ‒ 2 expenditure may, or may to take place when the purchaser incurs a liability to the
344
APPENDIX
seller. In practical terms, for goods it is generally assumed In the HICP, the weight reference period is defined as the
that they are consumed at the time they are purchased, so year t − 1 and the price reference period is a month (Decem-
prices for goods enter the HICP in the month that they are ber month of the previous year). Thus, in the HICP it is
observed. However, many services (for example, flights or necessary to revalue the weights at the prices of the price
package holidays) are either purchased in advance of when reference period. This procedure is applied in order to take
they are consumed or they are consumed over a period of into account the relative price changes, which have occurred
time (for example, season tickets for transport and sporting between the weight reference period and the price reference
services). The prices of services are therefore recorded in period.
the HICP in the first month that consumption of the service
can commence. Release, Timeliness, and Revisions
Quality Adjustment The HICP is produced and published each month accord-
ing to a preannounced schedule—in general between 16 and
The HICP is designed to be a measure of pure price change 18 days after the end of the reference month. Each month,
for goods and services falling within its scope. This implies Eurostat also publishes an HICP flash estimate for the euro
that the index should be calculated from identical or almost area and countries within that area. The flash estimate gives
identical goods and services; that is, the priced products an early indication of what the HICP inflation is likely to be
should remain of similar specification or quality over time. in the month in question. It is released on the last working
When a product offer in the sample is no longer available day of the reference month, or shortly after.
or no longer popular, it needs to be replaced, with possi- The current index reference period of the HICP is 2015,
bly a quality adjustment, to keep the sample representa- and it is changed every 10 years unless there is a major
tive. Replacement product offers should be either the same methodological change that requires re-referencing.
or similar enough with regard to their characteristics that HICPs are in principle revisable, but only in limited
consumers perceive them to be equivalent. Where, from the circumstances. In general, revisions have to be made if
consumer perspective, a replacement product offer is not errors are found after the initial publication, and may also
seen as equivalent (that is, its characteristics differ from the be made if the HICP is published as provisional and if new
replaced product offer), it is necessary to apply some type of or improved basic information becomes available that is
quality adjustment to the price comparison. needed to improve its accuracy.
Index Formulas HICP Methodological Manual

The HICP uses different formulas for aggregating prices and
price indices. In order to compile an elementary price index, The HICP methodological manual explains the agreed and
either unweighted or weighted index formulas (for example, preferred methods to be applied for the compilation of the
multilateral methods) can be used depending on whether HICP in detail. It also gives examples of good practice, pro-
detailed weights are available. At the level of the elemen- vides index compilers with a practical guide, and serves as
tary aggregate and above weights are available. At those compendium of the HICP.1
levels, the HICP uses a Laspeyres-type formula. The HICP
is a Laspeyres-type index obtained by annual chain linking 1
Eurostat. 2018. Harmonised Index of Consumer Prices (HICP) Meth-
13-month (December to December) indices. December of odological Manual, 2018 Edition. Luxembourg: European Commission.
each year is the overlap or link month in which the new and https://ec.europa.eu/eurostat/en/web/products-manuals-and-guidelines/-/
old basket of products are priced. KS-GQ-17-015.
345
APPENDIX
Appendix 2 Cereal preparations (cornflakes, oatflakes, and so on), and

other cereal products (malt, malt flour, malt extract,
Classification of Individual potato starch, tapioca, sago, and other starches)
Consumption According to Includes: farinaceous-based products prepared with meat,

fish, seafood, cheese, vegetables, or fruit.
Purpose 1999 (COICOP Excludes: meat pies (01.1.2); fish pies (01.1.3); and sweet
corn (01.1.7).
1999)
01.1.2 Meat (ND)
Fresh, chilled, or frozen meat of:
Bovine animals, swine, sheep, and goat
Introduction Horse, mule, donkey, camel, and the like
Classification of Individual Consumption According to Poultry (chicken, duck, goose, turkey, and guinea fowl)
Purpose (COICOP) 2018 has been endorsed by the United Hare, rabbit, and game (antelope, deer, boar, pheasant,
Nations Statistical Commission; however, many countries grouse, pigeon, quail, and so on)
continue to use COICOP 1999. Countries should begin mak- Fresh, chilled, or frozen edible offal
ing plans to implement COICOP 2018 as part of their next Dried, salted, or smoked meat, and edible offal (sausages,
consumer price index update and revision. While for some salami, bacon, ham, pâté, and so on)
countries the next update may be within the next couple of Other preserved or processed meat and meat-based prepa-
years, for others, it will be longer. This appendix provides rations (canned meat, meat extracts, meat juices, meat
detailed explanations and breakdowns for COICOP 1999. pies, and so on)
The objectives of this appendix are: (1) to maintain this Includes: meat and edible offal of marine mammals (seals,
reference for those countries who have not yet adopted walruses, whales, and so on) and exotic animals (kangaroo,
COICOP 2018; (2) to allow countries the opportunity to bet- ostrich, alligator, and so on); animals and poultry purchased
ter understand the differences between COICOP 2018 and live for consumption as food.
COICOP 1999; and (3) to facilitate planning for the imple- Excludes: land and sea snails (01.1.3); lard and other
mentation of the new classification standard. edible animal fats (01.1.5); and soups, broths, and stocks
containing meat (01.1.9).
COICOP: Breakdown of Individual
Consumption Expenditure of Households 01.1.3 Fish and Seafood (ND)
by Division and Group Fresh, chilled, or frozen fish
Fresh, chilled, or frozen seafood (crustaceans, molluscs
01 Food and Nonalcoholic Beverages and other shellfish, and sea snails)
01.1 Food Dried, smoked, or salted fish, and seafood
The food products classified here are those purchased for Other preserved or processed fish and seafood and fish
consumption at home. The group excludes food products and seafood-based preparations (canned fish and sea-
sold for immediate consumption away from the home by food, caviar, and other hard roes, fish pies, and so on)
hotels, restaurants, cafés, bars, kiosks, street vendors, auto- Includes: land crabs, land snails, and frogs; fish and seafood
matic vending machines, and so on (11.1.1); cooked dishes purchased live for consumption as food.
prepared by restaurants for consumption off their premises Excludes: soups, broths, and stocks containing fish and
(11.1.1); cooked dishes prepared by catering contractors seafood (01.1.9).
whether collected by the customer or delivered to the cus-
tomer’s home (11.1.1); and products sold specifically as pet
01.1.4 Milk, Cheese, and Eggs (ND)
foods (09.3.4).
Raw milk; pasteurized or sterilized milk
01.1.1 Bread and Cereals (ND)1 Condensed, evaporated, or powdered milk
Rice in all forms Yogurt, cream, milk-based desserts, milk-based bever-
Maize, wheat, barley, oats, rye, and other cereals in the ages, and other similar milk-based products
form of grain, flour, or meal Cheese and curd
Bread and other bakery products (crispbread, rusks, Eggs and egg products made wholly from eggs
toasted bread, biscuits, gingerbread, wafers, waf- Includes: milk, cream, and yogurt containing sugar, cocoa,
fles, crumpets, muffins, croissants, cakes, tarts, pies, fruit, or flavorings; dairy products not based on milk such
quiches, pizzas, and so on) as soya milk.
Mixes and dough for the preparation of bakery products Excludes: butter and butter products (01.1.5).
Pasta products in all forms; couscous
01.1.5 Oils and Fats (ND)
Butter and butter products (butter oil, ghee, and so on)
1
Margarine (including “diet” margarine) and other veg-
Note: ND, SD, D, and S denote nondurable goods, semidurable goods,
durable goods, and services, respectively.
etable fats (including peanut butter)
347
APPENDIX
Edible oils (olive oil, corn oil, sunflower seed oil, cot- Prepared baking powders, baker’s yeast, dessert prepara-
tonseed oil, soybean oil, groundnut oil, walnut oil, and tions, soups, broths, stocks, culinary ingredients, and
so on) so on
Edible animal fats (lard and so on) Homogenized baby food and dietary preparations irre-
spective of the composition
Excludes: cod or halibut liver oil (06.1.1).
Excludes: milk-based desserts (01.1.4); soya milk (01.1.4);
01.1.6 Fruit (ND) artificial sugar substitutes (01.1.8); and cocoa-based des-
Fresh, chilled, or frozen fruit sert preparations (01.1.8).
Dried fruit, fruit peel, fruit kernels, nuts, and edible seeds
Preserved fruit and fruit-based products 01.2 Nonalcoholic Beverages
The nonalcoholic beverages classified here are those pur-
Includes: melons and watermelons. chased for consumption at home. The group excludes non-
Excludes: vegetables cultivated for their fruit such as alcoholic beverages sold for immediate consumption away
aubergines, cucumbers, and tomatoes (01.1.7); jams, mar- from the home by hotels, restaurants, cafés, bars, kiosks,
malades, compotes, jellies, fruit purées, and pastes (01.1.8); street vendors, automatic vending machines, and so on
parts of plants preserved in sugar (01.1.8); and fruit juices (11.1.1).
and syrups (01.2.2).
01.2.1 Coffee, Tea, and Cocoa (ND)
01.1.7 Vegetables (ND) Coffee, whether or not decaffeinated, roasted or ground,
Fresh, chilled, frozen, or dried vegetables cultivated for including instant coffee
their leaves or stalks (asparagus, broccoli, cauliflower, Tea, maté, and other plant products for infusions
endives, fennel, spinach, and so on), for their fruit Cocoa, whether or not sweetened, and chocolate-based
(aubergines, cucumbers, courgettes, green peppers, powder
pumpkins, tomatoes, and so on), and for their roots
Includes: cocoa-based beverage preparations; coffee and tea
(beetroots, carrots, onions, parsnips, radishes, turnips,
substitutes; and extracts and essences of coffee and tea.
and so on)
Excludes: chocolate in bars or slabs (01.1.8); cocoa-based
Fresh or chilled potatoes and other tuber vegetables (man-
food; and cocoa-based dessert preparations (01.1.8).
ioc, arrowroot, cassava, sweet potatoes, and so on)
Preserved or processed vegetables and vegetable-based 01.2.2 Mineral Waters, Soft Drinks, and Fruit and
products Vegetable Juices (ND)
Products of tuber vegetables (flours, meals, flakes, purées,
chips, and crisps) including frozen preparations such as Mineral or spring waters; all drinking water sold in
chipped potatoes containers
Includes: olives; garlic; pulses; sweet corn; sea fennel and Soft drinks such as sodas, lemonades, and colas
other edible seaweed; mushrooms and other edible fungi. Fruit and vegetable juices
Excludes: potato starch, tapioca, sago, and other starches Syrups and concentrates for the preparation of beverages
(01.1.1); soups, broths, and stocks containing vegetables Excludes: nonalcoholic beverages which are generally
(01.1.9); culinary herbs (parsley, rosemary, thyme, and so alcoholic such as nonalcoholic beer (02.1).
on) and spices (pepper, pimento, ginger, and so on) (01.1.9);
and vegetable juices (01.2.2). 02 Alcoholic Beverages, Tobacco, and Narcotics
01.1.8 Sugar, Jam, Honey, Chocolate, 02.1 Alcoholic Beverages
and Confectionery (ND) The alcoholic beverages classified here are those purchased
Cane or beet sugar, unrefined or refined, powdered, crys- for consumption at home. The group excludes alcoholic
tallized, or in lumps beverages sold for immediate consumption away from the
Jams, marmalades, compotes, jellies, fruit purées, and home by hotels, restaurants, cafés, bars, kiosks, street ven-
pastes, natural, and artificial honey, maple syrup, dors, automatic vending machines, and so on (11.1.1).
molasses, and parts of plants preserved in sugar The beverages classified here include low or nonal-
Chocolate in bars or slabs, chewing gum, sweets, toffees, coholic beverages which are generally alcoholic such as
pastilles, and other confectionery products nonalcoholic beer.
Cocoa-based foods and cocoa-based dessert preparations
Edible ice, ice cream, and sorbet 02.1.1 Spirits (ND)
Eaux-de-vie, liqueurs, and other spirits
Includes: artificial sugar substitutes.
Excludes: cocoa and chocolate-based powder (01.2.1). Includes: mead; aperitifs other than wine-based aperitifs
(02.1.2).
01.1.9 Food Products N.E.C. (ND)
Salt, spices (pepper, pimento, ginger, and so on), culi- 02.1.2 Wine (ND)
nary herbs (parsley, rosemary, thyme, and so on), Wine, cider, and perry, including sake
sauces, condiments, seasonings (mustard, mayonnaise, Wine-based aperitifs, fortified wines, champagne, and
ketchup, soy sauce, and so on), vinegar other sparkling wines
348
APPENDIX
02.1.3 Beer (ND) Sewing threads, knitting yarns, and accessories for mak-
All kinds of beer such as ale, lager, and porter ing clothing such as buckles, buttons, press studs, zip
fasteners, ribbons, laces, trimmings, and so on
Includes: low-alcoholic beer and nonalcoholic beer; shandy.
Includes: gardening gloves and working gloves; crash hel-
02.2 Tobacco mets for motorcycles and bicycles.
This group covers all purchases of tobacco by households, Excludes: gloves and other articles made of rubber
including purchases of tobacco in restaurants, cafés, bars, (05.6.1); pins, safety pins, sewing needles, knitting needles,
service stations, and so on. thimbles (05.6.1); protective headgear for sports (09.3.2);
other protective gear for sports such as life jackets, boxing
02.2.0 Tobacco (ND) gloves, body padding, belts, supports, and so on (09.3.2);
paper handkerchiefs (12.1.3); watches, jewelry, cuff links,
Cigarettes; cigarette tobacco and cigarette papers
tiepins (12.3.1); walking sticks and canes, umbrellas and
Cigars, pipe tobacco, chewing tobacco, or snuff
parasols, fans, and keyrings (12.3.2).
Excludes: other smokers’ articles (12.3.2).
03.1.4 Cleaning, Repair, and Hire of Clothing (S)
02.3 Narcotics Dry-cleaning, laundering, and dyeing of garments
Darning, mending, repair, and altering of garments
02.3.0 Narcotics (ND) Hire of garments
Marijuana, opium, cocaine, and their derivatives Includes: total value of the repair service (that is, both the
Other vegetable-based narcotics such as cola nuts, betel cost of labor and the cost of materials are covered).
leaves, and betel nuts Excludes: materials, threads, accessories, and so on pur-
Other narcotics including chemicals and man-made drugs chased by households with the intention of undertaking the
repairs themselves (03.1.1) or (03.1.3); repair of household
03 Clothing and Footwear linen and other household textiles (05.2.0); dry-cleaning,
03.1 Clothing laundering, dyeing, and hiring of household linen, and other
household textiles (05.6.2).
03.1.1 Clothing Materials (SD)
Clothing materials of natural fibers, of man-made fibers, 03.2 Footwear
and of their mixtures 03.2.1 Shoes and Other Footwear (SD)
Excludes: furnishing fabrics (05.2.0). All footwear for men, women, children (3–13 years), and
infants (0–2 years) including sports footwear suitable
03.1.2 Garments (SD) for everyday or leisure wear (shoes for jogging, cross-
Garments for men, women, children (3–13 years) and training, tennis, basketball, boating, and so on)
infants (0–2 years), either ready-to-wear or made-to-
Includes: gaiters, leggings, and similar articles; shoelaces;
measure, in all materials (including leather, furs, plas-
parts of footwear, such as heels, soles, and so on, purchased
tics, and rubber), for everyday wear, for sport or for
by households with the intention of repairing footwear
work:
themselves.
Capes, overcoats, raincoats, anoraks, parkas, blousons, Excludes: babies’ booties made of fabric (03.1.2); shoe
jackets, trousers, waistcoats, suits, costumes, dresses, trees, shoehorns and polishes, creams, and other shoe-
skirts, and so on cleaning articles (05.6.1); orthopedic footwear (06.1.3);
Shirts, blouses, pullovers, sweaters, cardigans, shorts, game-specific footwear (ski boots, football boots, golfing
swimsuits, tracksuits, jogging suits, sweatshirts, T-shirts, shoes, and other such footwear fitted with ice skates, rollers,
leotards, and so on spikes, studs, and so on) (09.3.2); and shin guards, cricket
Vests, underpants, socks, stockpiles, tights, petticoats, bras- pads, and other such protective apparel for sport (09.3.2).
sières, knickers, slips, girdles, corsets, body stockpiles,
and so on 03.2.2 Repair and Hire of Footwear (S)
Pajamas, nightshirts, nightdresses, housecoats, dressing Repair of footwear; shoe-cleaning services
gowns, bathrobes, and so on Hire of footwear
Baby clothes and babies’ booties made of fabric.
Includes: total value of the repair service (that is, both the
Excludes: articles of medical hosiery such as elasticated cost of labor and the cost of materials are covered).
stockpiles (06.1.2); babies’ napkins (12.1.3). Excludes: parts of footwear, such as heels, soles, and so
on, purchased by households with the intention of under-
03.1.3 Other Articles of Clothing and Clothing taking the repair themselves (03.2.1); polishes, creams, and
Accessories (SD) other shoe-cleaning articles (05.6.1); and repair (09.3.2) or
Ties, handkerchiefs, scarves, squares, gloves, mittens, hire (09.4.1) of game-specific footwear (ski boots, football
muffs, belts, braces, aprons, smocks, bibs, sleeve pro- boots, golfing shoes, and other such footwear fitted with ice
tectors, hats, caps, berets, bonnets, and so on skates, rollers, spikes, studs, and so on).
349
APPENDIX
04 Housing, Water, Electricity, Gas, and Other replastering walls or repairing roofs, and which are carried
Fuels out by owners only.
Only expenditures which tenants and owner-occupiers
04.1 Actual Rentals for Housing incur on materials and services for minor maintenance and
Rentals normally include payment for the use of the land on repair are part of individual consumption expenditure of
which the property stands, the dwelling occupied, the fix- households. Expenditures which owner-occupiers incur on
tures and fittings for heating, plumbing, lighting, and so on, materials and services for major maintenance and repair
and, in the case of a dwelling let furnished, the furniture. are not part of individual consumption expenditure of
Rentals also include payment for the use of a garage to households.
provide parking in connection with the dwelling. The garage Purchases of materials made by tenants or owner-occu-
does not have to be physically contiguous to the dwelling, piers with the intention of undertaking the maintenance or
nor does it have to be leased from the same landlord. repair themselves should be shown under (04.3.1). If tenants
Rentals do not include payment for the use of garages or or owner-occupiers pay an enterprise to carry out the main-
parking spaces not providing parking in connection with the tenance or repair, the total value of the service, including the
dwelling (07.2.4). Nor do they include charges for water costs of the materials used, should be shown under (04.3.2).
supply (04.4.1); refuse collection (04.4.2) and sewage col-
lection (04.4.3); coproprietor charges for caretaking, garden- 04.3.1 Materials for the Maintenance and Repair of the
ing, stairwell cleaning, heating and lighting, maintenance of Dwelling (ND)
lifts and refuse disposal chutes, and so on in multi-occupied Products and materials, such as paints and varnishes,
buildings (04.4.4); charges for electricity (04.5.1) and gas renderings, wallpapers, fabric wall coverings, window
(04.5.2); and charges for heating and hot water supplied by panes, plaster, cement, putty, wallpaper pastes, and so
district heating plants (04.5.5). on, purchased for minor maintenance and repair of the
dwelling
04.1.1 Actual Rentals Paid by Tenants (S)
Includes: small plumbing items (pipes, taps, joints, and so
Rentals actually paid by tenants or subtenants occupy- on), surfacing materials (floorboards, ceramic tiles, and
ing unfurnished or furnished premises as their main so on), and brushes and scrapers for paint, varnish, and
residence wallpaper.
Includes: payments by households occupying a room in a Excludes: fitted carpets and linoleum (05.1.2); hand
hotel or boarding house as their main residence. tools, door fittings, power sockets, wiring flex, and lamp
Excludes: accommodation services of educational estab- bulbs (05.5.2); brooms, scrubbing brushes, dusting brushes,
lishments and hostels (11.2.0) and of retirement homes for and cleaning products (05.6.1); products, materials, and fix-
elderly persons (12.4.0). tures used for major maintenance and repair (intermediate
consumption) or for extension and conversion of the dwell-
04.1.2 Other Actual Rentals (S) ing (capital formation).
Rentals actually paid for secondary residences
04.3.2 Services for the Maintenance and Repair of the
Excludes: accommodation services of holiday villages and Dwelling (S)
holiday centers (11.2.0). Services of plumbers, electricians, carpenters, glaziers,
painters, decorators, floor polishers, and so on engaged
04.2 Imputed Rentals for Housing for minor maintenance and repair of the dwelling
For coverage see note to (04.1).
Includes: total value of the service (that is, both the cost of
04.2.1 Imputed Rentals of Owner-Occupiers (S) labor and the cost of materials are covered).
Imputed rentals of owners occupying their main residence Excludes: separate purchases of materials made by
households with the intention of undertaking the mainte-
04.2.2 Other Imputed Rentals (S) nance or repair themselves (04.3.1); services engaged for
major maintenance and repair (intermediate consumption)
Imputed rentals for secondary residences
or for extension and conversion of the dwelling (capital
Imputed rentals of households paying a reduced rental or
formation).
housed free
04.3 Maintenance and Repair of the Dwelling 04.4 Water Supply and Miscellaneous Services Relating
Maintenance and repair of dwellings are distinguished by to the Dwelling
two features: first, they are activities that have to be under- 04.4.1 Water Supply (ND)
taken regularly in order to maintain the dwelling in good
working order; second, they do not change the dwelling’s Water supply
performance, capacity or expected service life. Includes: associated expenditure such as hire of meters,
There are two types of maintenance and repair of dwell- reading of meters, standing charges, and so on.
ings: those which are minor, such as interior decoration and Excludes: drinking water sold in bottles or containers
repairs to fittings, and which are commonly carried out by (01.2.2); hot water or steam purchased from district heating
both tenants and owners; and those which are major, such as plants (04.5.5).
350
APPENDIX
04.4.2 Refuse Collection (S) Includes: delivery and installation when applicable; base
Refuse collection and disposal mattresses, mattresses, tatamis; bathroom cabinets; baby
furniture such as cradles, high chairs, and playpens; blinds;
04.4.3 Sewage Collection (S) camping and garden furniture; and mirrors, candleholders,
Sewage collection and disposal and candlesticks.
Excludes: bedding and sunshades (05.2.0); safes (05.3.1);
04.4.4 Other Services Relating to the ornamental glass and ceramic articles (05.4.0); clocks
Dwelling N.E.C. (S) (12.3.1); wall thermometers and barometers (12.3.2); carry-
Coproprietor charges for caretaking, gardening, stairwell cots and pushchairs (12.3.2); and works of art and antique fur-
niture acquired primarily as stores of value (capital formation).
cleaning, heating and lighting, maintenance of lifts and
refuse disposal chutes, and so on in multi-occupied 05.1.2 Carpets and Other Floor Coverings (D)
buildings
Security services Loose carpets, fitted carpets, linoleum, and other such floor
Snow removal and chimney sweeping coverings
Excludes: household services such as window cleaning, Includes: laying of floor coverings
disinfecting, fumigation, and pest extermination (05.6.2); Excludes: bathroom mats, rush mats, and doormats (05.2.0);
bodyguards (12.7.0). antique floor coverings acquired primarily as stores of value
(capital formation).
04.5 Electricity, Gas, and Other Fuels 05.1.3 Repair of Furniture, Furnishings, and Floor
Coverings (S)
04.5.1 Electricity (ND)
Electricity Repair of furniture, furnishings, and floor coverings
Includes: associated expenditure such as hire of meters,
labor and the cost of materials are covered); restoration of
reading of meters, standing charges, and so on.
works of art, antique furniture, and antique floor coverings
04.5.2 Gas (ND) other than those acquired primarily as stores of value (capi-
tal formation).
Town gas and natural gas Excludes: separate purchases of materials made by house-
Liquefied hydrocarbons (butane, propane, and so on) holds with the intention of undertaking the repair themselves
Includes: associated expenditure such as hire of meters, read- (05.1.1) or (05.1.2); dry-cleaning of carpets (05.6.2).
ing of meters, storage containers, standing charges, and so on.
05.2 Household Textiles
04.5.3 Liquid Fuels (ND)
Domestic heating and lighting oils 05.2.0 Household Textiles (SD)
Furnishing fabrics, curtain material, curtains, double cur-
04.5.4 Solid Fuels (ND) tains, awnings, door curtains, and fabric blinds
Coal, coke, briquettes, firewood, charcoal, peat, and the like Bedding such as futons, pillows, bolsters, and hammocks
Bed linen such as sheets, pillowcases, blankets, traveling
04.5.5 Heat Energy (ND) rugs, plaids, eiderdowns, counterpanes, and mosquito nets
Hot water and steam purchased from district heating Table linen and bathroom linen such as tablecloths, table
plants napkins, towels, and face cloths
Other household textiles such as shopping bags, laundry
Includes: associated expenditure such as hire of meters, bags, shoe bags, covers for clothes and furniture, flags,
reading of meters, standing charges, and so on; ice used for sunshades, and so on
cooling and refrigeration purposes. Repair of such articles
05 Furnishings, Household Equipment, and Includes: cloth bought by the piece; oilcloth; bathroom
Routine Household Maintenance mats, rush mats, and doormats.
Excludes: fabric wall coverings (04.3.1); tapestries
05.1 Furniture and Furnishings, Carpets, and Other (05.1.1); floor coverings such as carpets and fitted carpets
Floor Coverings (05.1.2); electric blankets (05.3.2); covers for motor cars,
motorcycles, and so on (07.2.1); and air mattresses and
05.1.1 Furniture and Furnishings (D) sleeping bags (09.3.2).
Beds, sofas, couches, tables, chairs, cupboards, chests of
drawers, and bookshelves 05.3 Household Appliances
Lighting equipment such as ceiling lights, standard lamps,
globe lights, and bedside lamps 05.3.1 Major Household Appliances Whether Electric or
Pictures, sculptures, engravings, tapestries, and other art Not (D)
objects including reproductions of works of art and Refrigerators, freezers, and fridge freezers
other ornaments Washing machines, dryers, drying cabinets, dishwashers,
Screens, folding partitions, and other furniture and fixtures ironing, and pressing machines
351
APPENDIX
Cookers, spit roasters, hobs, ranges, ovens, and micro- 05.5 Tools and Equipment for House and Garden
wave ovens
Air-conditioners, humidifiers, space heaters, water heat- 05.5.1 Major Tools and Equipment (D)
ers, ventilators, and extractor hoods Motorized tools and equipment such as electric drills,
Vacuum cleaners, steam-cleaning machines, carpet sham- saws, sanders and hedge cutters, garden tractors, lawn-
pooing machines, and machines for scrubbing, waxing, mowers, cultivators, chainsaws, and water pumps
and polishing floors Repair of such articles
Other major household appliances such as safes, sewing Includes: charges for the leasing or rental of do-it-yourself
machines, knitting machines, water softeners, and so machinery and equipment.
on
05.5.2 Small Tools and Miscellaneous Accessories (SD)
Includes: delivery and installation of the appliances when
applicable. Hand tools such as saws, hammers, screwdrivers, wrenches,
Excludes: such appliances that are built into the structure spanners, pliers, trimming knives, rasps, and files
Garden tools such as wheelbarrows, watering cans, hoses,
of the building (capital formation).
spades, shovels, rakes, forks, scythes, sickles, and secateurs
Ladders and steps
05.3.2 Small Electric Household Appliances (SD)
Door fittings (hinges, handles, and locks), fittings for
Coffee mills, coffee makers, juice extractors, can openers, radiators and fireplaces, and other metal articles for
food mixers, deep fryers, meat grills, knives, toasters, the house (curtain rails, carpet rods, hooks, and so on)
ice cream makers, sorbet makers, yogurt makers, hot- or for the garden (chains, grids, stakes, and hoop seg-
plates, irons, kettles, fans, electric blankets, and so on ments for fencing and bordering)
Small electric accessories such as power sockets,
Excludes: small nonelectric household articles and kitchen switches, wiring flex, electric bulbs, fluorescent light-
utensils (05.4.0); household scales (05.4.0); personal weighing tubes, torches, flashlights, hand lamps, electric bat-
ing machines and baby scales (12.1.3). teries for general use, bells, and alarms
Repair of such articles
05.3.3 Repair of Household Appliances (S)
Repair of household appliances 05.6 Goods and Services for Routine Household
Maintenance
labor and the cost of materials are covered); charges for the 05.6.1 Nondurable Household Goods (ND)
leasing or rental of major household appliances. Cleaning and maintenance products such as soaps, wash-
Excludes: separate purchases of materials made by house- ing powders, washing liquids, scouring powders, deter-
holds with the intention of undertaking the repair themselves gents, disinfectant bleaches, softeners, conditioners,
(05.3.1) or (05.3.2). window-cleaning products, waxes, polishes, dyes,
unblocking agents, disinfectants, insecticides, pesti-
05.4 Glassware, Tableware, and Household Utensils cides, fungicides, and distilled water
Articles for cleaning such as brooms, scrubbing brushes,
05.4.0 Glassware, Tableware, and Household Utensils dustpans and dust brushes, dusters, tea towels, floor-
(SD) cloths, household sponges, scourers, steel wool, and
Glassware, crystal ware, ceramic ware, and chinaware chamois leathers
of the kind used for table, kitchen, bathroom, toilet, Paper products such as filters, tablecloths and table nap-
office, and indoor decoration kins, kitchen paper, vacuum cleaner bags, and card-
Cutlery, flatware, and silverware board tableware, including aluminum foil and plastic
Nonelectric kitchen utensils of all materials such as bin liners
saucepans, stewpots, pressure cookers, frying pans, Other nondurable household articles such as matches,
candles, lamp wicks, methylated spirits, clothes-pegs,
coffee mills, purée makers, mincers, hotplates, house-
clothes hangers, pins, safety pins, sewing needles, knit-
hold scales, and other such mechanical devices
ting needles, thimbles, nails, screws, nuts and bolts,
Nonelectric household articles of all materials such as tacks, washers, glues and adhesive tapes for household
containers for bread, coffee, spices, and so on, waste use, string, twine, and rubber gloves
bins, waste-paper baskets, laundry baskets, portable
money boxes and strongboxes, towel rails, bottle racks, Includes: polishes, creams, and other shoe-cleaning articles;
irons and ironing boards, letter boxes, feeding bottles, fire extinguishers for households.
thermos flasks, and iceboxes Excludes: brushes and scrapers for paint, varnish, and
Repair of such articles wallpaper (04.3.1); fire extinguishers for transport equip-
ment (07.2.1); products specifically for the cleaning and
Excludes: lighting equipment (05.1.1); electric house- maintenance of transport equipment such as paints, chrome
hold appliances (05.3.1) or (05.3.2); cardboard tableware cleaners, sealing compounds, and bodywork polishes
(05.6.1); personal weighing machines and baby scales (07.2.1); horticultural products for the upkeep of orna-
(12.1.3); ashtrays (12.3.2). mental gardens (09.3.3); paper handkerchiefs, toilet paper,
352
APPENDIX
toilet soaps, toilet sponges, and other products for personal medical massage equipment and health lamps, pow-
hygiene (12.1.3); cigarette, cigar, and pipe lighters, and ered and unpowered wheelchairs and invalid carriages,
lighter fuel (12.3.2). “special” beds, crutches, electronic and other devices
for monitoring blood pressure, and so on
05.6.2 Domestic Services and Household Services (S) Repair of such articles
Domestic services supplied by paid staff employed in Includes: dentures but not fitting costs.
private services such as butlers, cooks, maids, drivers, Excludes: hire of therapeutic equipment (06.2.3); protec-
gardeners, governesses, secretaries, tutors, and au pairs tive goggles, belts, and supports for sport (09.3.2); and sun-
Similar services, including babysitting and housework, glasses not fitted with corrective lenses (12.3.2).
supplied by enterprises or self-employed persons
Household services such as window cleaning, disinfect- 06.2 Outpatient Services
ing, fumigation, and pest extermination This group covers medical, dental, and paramedical services
Dry-cleaning, laundering, and dyeing of household linen, delivered to outpatients by medical, dental, and paramedical
household textiles, and carpets practitioners and auxiliaries. The services may be delivered
Hire of furniture, furnishings, carpets, household equip- at home, in individual or group consulting facilities, dispen-
ment, and household linen saries or the outpatient clinics of hospitals, and the like.
Excludes: dry-cleaning, laundering, and dyeing of garments Outpatient services include the medicaments, prostheses,
(03.1.4); refuse collection (04.4.2); sewerage collection medical appliances and equipment, and other health-related
(04.4.3); coproprietor charges for caretaking, gardening, stair- products supplied directly to outpatients by medical, dental,
well cleaning, heating and lighting, maintenance of lifts and and paramedical practitioners and auxiliaries.
refuse disposal chutes, and so on in multi-occupied buildings Medical, dental, and paramedical services provided to
(04.4.4); security services (04.4.4); snow removal and chimney inpatients by hospitals and the like are included in hospital
sweeping (04.4.4); removal and storage services (07.3.6); ser- services (06.3).
vices of wet-nurses, crèches, day care centers, and other child-
minding facilities (12.4.0); and bodyguards (12.7.0). 06.2.1 Medical Services (S)
Consultations of physicians in general or specialist practice
06 Health
Includes: services of orthodontic specialists.
This division also includes health services purchased from
Excludes: services of medical analysis laboratories and
school and university health centers.
X-ray centers (06.2.3); services of practitioners of tradi-
06.1 Medical Products, Appliances, and Equipment
tional medicine (06.2.3).
This group covers medicaments, prostheses, medical appliances 06.2.2 Dental Services (S)
and equipment, and other health-related products purchased by
individuals or households, either with or without a prescrip- Services of dentists, oral hygienists, and other dental
tion, usually from dispensing chemists, pharmacists, or medical auxiliaries
equipment suppliers. They are intended for consumption or use Includes: fitting costs of dentures.
outside a health facility or institution. Such products supplied Excludes: dentures (06.1.3); services of orthodontic spe-
directly to outpatients by medical, dental, and paramedical prac- cialists (06.2.1); and services of medical analysis laborato-
titioners or to inpatients by hospitals and the like are included in ries and X-ray centers (06.2.3).
outpatient services (06.2) or hospital services (06.3).
06.2.3 Paramedical Services (S)
06.1.1 Pharmaceutical Products (ND)
Services of medical analysis laboratories and X-ray
Medicinal preparations, medicinal drugs, patent medi- centers
cines, serums and vaccines, vitamins and minerals, cod Services of freelance nurses and midwives
liver oil and halibut liver oil, and oral contraceptives Services of freelance acupuncturists, chiropractors,
Excludes: veterinary products (09.3.4); articles for personal optometrists, physiotherapists, speech therapists, and
hygiene such as medicinal soaps (12.1.3). so on
Medically prescribed corrective-gymnastic therapy
06.1.2 Other Medical Products (ND) Outpatient thermal bath or seawater treatments
Clinical thermometers, adhesive and nonadhesive ban- Ambulance services
dages, hypodermic syringes, first aid kits, hot-water Hire of therapeutic equipment
bottles and ice bags, medical hosiery items such as elas- Includes: services of practitioners of traditional medicine.
ticated stockpiles and knee supports, pregnancy tests,
condoms, and other mechanical contraceptive devices 06.3 Hospital Services
Hospitalization is defined as occurring when a patient is
06.1.3 Therapeutic Appliances and Equipment (D) accommodated in a hospital for the duration of the treat-
Corrective eyeglasses and contact lenses, hearing aids, ment. Hospital day care and home-based hospital treatment
glass eyes, artificial limbs and other prosthetic devices, are included as are hospices for terminally ill persons.
orthopedic braces and supports, orthopedic footwear, This group covers the services of general and specialist
surgical belts, trusses and supports, neck braces, hospitals, the services of medical centers, maternity centers,
353
APPENDIX
nursing homes, and convalescent homes which chiefly pro- 07.1.4 Animal-Drawn Vehicles (D)
vide inpatient health care, the services of institutions serv- Animal-drawn vehicles
ing old people in which medical monitoring is an essential
component and the services of rehabilitation centers provid- Includes: animals required to draw the vehicles and related
ing inpatient health care and rehabilitative therapy where the equipment (yokes, collars, harnesses, bridles, reins, and so
objective is to treat the patient rather than to provide long- on).
Excludes: horses and ponies, horse- or pony-drawn vehi-
term support.
cles, and related equipment purchased for recreational pur-
Hospitals are defined as institutions which offer inpatient
poses (09.2.1).
care under the direct supervision of qualified medical doc-
tors. Medical centers, maternity centers, nursing homes, and 07.2 Operation of Personal Transport Equipment
convalescent homes also provide inpatient care but their Purchases of spare parts, accessories, or lubricants made by
services are supervised and frequently delivered by staff of households with the intention of undertaking the mainte-
lower qualification than medical doctors. nance, repair, or intervention themselves should be shown
This group does not cover the services of facilities, such under (07.2.1) or (07.2.2). If households pay an enterprise to
as surgeries, clinics, and dispensaries, devoted exclusively carry out the maintenance, repair, or fitting, the total value of
to outpatient care (06.2). Nor does it include the services the service, including the costs of the materials used, should
of retirement homes for elderly persons, institutions for dis- be shown under (07.2.3).
abled persons, and rehabilitation centers providing primarily
long-term support (12.4). 07.2.1 Spare Parts and Accessories for Personal
Transport Equipment (SD)
06.3.0 Hospital Services (S) Tires (new, used, or retreaded), inner tubes, spark plugs,
Hospital services comprise the provision of the following batteries, shock absorbers, filters, pumps, and other spare
services to hospital inpatients: parts or accessories for personal transport equipment
Basic services: administration; accommodation; food
Includes: fire extinguishers for transport equipment; prod-
and drink; supervision and care by nonspecialist
ucts specifically for the cleaning and maintenance of trans-
staff (nursing auxiliaries); first aid and resuscita- port equipment such as paints, chrome cleaners, sealing
tion; ambulance transport; provision of medicines compounds, and bodywork polishes; and covers for motor
and other pharmaceutical products; and provision of cars, motorcycles, and so on.
therapeutic appliances and equipment Excludes: crash helmets for motorcycles and bicycles
Medical services: services of physicians in general or (03.1.3); nonspecific products for cleaning and maintenance
specialist practice, of surgeons, and of dentists; med- such as distilled water, household sponges, chamois leathers,
ical analyses and X-rays; and paramedical services detergents, and so on (05.6.1); charges for the fitting of spare
such as those of nurses, midwives, chiropractors, parts and accessories and for the painting, washing, and pol-
optometrists, physiotherapists, speech therapists, ishing of bodywork (07.2.3); radiotelephones (08.2.0); car
and so on. radios (09.1.1); and baby seats for cars (12.3.2).
07 Transport 07.2.2 Fuels and Lubricants for Personal Transport

Equipment (ND)
07.1 Purchase of Vehicles
Purchases of recreational vehicles such as camper vans, car- Petrol and other fuels such as diesel, liquid petroleum gas,
avans, trailers, airplanes, and boats are covered by (09.2.1). alcohol, and two-stroke mixtures
Lubricants, brake and transmission fluids, coolants, and
07.1.1 Motor Cars (D) additives
Motor cars, passenger vans, station wagons, estate cars, Includes: fuel for major tools and equipment covered under
and the like with either two-wheel drive or four-wheel (05.5.1) and recreational vehicles covered under (09.2.1).
drive Excludes: charges for oil changes and greasing (07.2.3).
Excludes: invalid carriages (06.1.3); camper vans (09.2.1); 07.2.3 Maintenance and Repair of Personal Transport
and golf carts (09.2.1). Equipment (S)
Services purchased for the maintenance and repair of
07.1.2 Motor Cycles (D)
personal transport equipment such as fitting of parts
Motorcycles of all types, scooters, and powered bicycles and accessories, wheel balancing, technical inspec-
Includes: sidecars; snowmobiles. tion, breakdown services, oil changes, greasing, and
Excludes: invalid carriages (06.1.3); golf carts (09.2.1). washing
Includes: total value of the service (that is both the cost of
07.1.3 Bicycles (D) labor and the cost of materials are covered).
Bicycles and tricycles of all types Excludes: separate purchases of spare parts, accesso-
ries, or lubricants made by households with the intention of
Includes: rickshaws. undertaking the maintenance or repair themselves (07.2.1)
Excludes: toy bicycles and tricycles (09.3.1). or (07.2.2); roadworthiness tests (07.2.4).
354
APPENDIX
07.2.4 Other Services in Respect of Personal Transport Services of porters and left-luggage and luggage-for-
Equipment (S) warding offices
Hire of garages or parking spaces not providing parking Travel agents’ commissions, if separately priced
in connection with the dwelling Excludes: cable car and chairlift transport at ski resorts and
Toll facilities (bridges, tunnels, shuttle ferries, motor- holiday centers (09.4.1).
ways) and parking meters
Driving lessons, driving tests, and driving licenses 08 Communication
Roadworthiness tests
Hire of personal transport equipment without drivers 08.1 Postal Services
Excludes: hire of a car with driver (07.3.2); service charges 08.1.0 Postal Services (S)
for insurance in respect of personal transport equipment
(12.5.4). Payments for the delivery of letters, postcards, and parcels
Private mail and parcel delivery
07.3 Transport Services Includes: all purchases of new postage stamps, prefranked
Purchases of transport services are generally classified by postcards, and aerograms.
mode of transport. When a ticket covers two or more modes Excludes: purchase of used or canceled postage stamps
of transport—for example, intraurban bus and underground (09.3.1); financial services of post offices (12.6.2).
or interurban train and ferry—and the expenditure cannot be
apportioned between them, then such purchases should be 08.2 Telephone and Telefax Equipment
classified in 07.3.5.
Costs of meals, snacks, drinks, refreshments, or accom- 08.2.0 Telephone and Telefax Equipment (D)
modation services have to be included if covered by the fare
Purchases of telephones, radiotelephones, telefax machines,
and not separately priced. If separately priced, these costs telephone-answering machines, and telephone loud-
have to be classified in Division 11. speakers
School transport services are included, but ambulance Repair of such equipment
services are excluded (06.2.3).
Excludes: telefax and telephone-answering facilities pro-
07.3.1 Passenger Transport by Railway (S) vided by personal computers (09.1.3).
Transport of individuals and groups of persons and lug-
gage by train, tram, and underground 08.3 Telephone and Telefax Services
Includes: transport of private vehicles. 08.3.0 Telephone and Telefax Services (S)
Excludes: funicular transport (07.3.6). Installation and subscription costs of personal telephone
equipment
07.3.2 Passenger Transport by Road (S) Telephone calls from a private line or from a public line
Transport of individuals and groups of persons and lug- (public telephone box, post office cabin, and so on);
gage by bus, coach, taxi, and hired car with driver telephone calls from hotels, cafés, restaurants, and the
like
07.3.3 Passenger Transport by Air (S) Telegraphy, telex, and telefax services
Transport of individuals and groups of persons and lug- Information transmission services; internet connection
gage by airplane and helicopter services
Hire of telephones, telefax machines, telephone-answer-
07.3.4 Passenger Transport by Sea and Inland ing machines, and telephone loudspeakers
Waterway (S) Includes: radiotelephony, radiotelegraphy, and radiotelex
Transport of individuals and groups of persons and lug- services.
gage by ship, boat, ferry, hovercraft, and hydrofoil
09 Recreation and Culture
Includes: transport of private vehicles.
09.1 Audiovisual, Photographic, and Information
07.3.5 Combined Passenger Transport (S) Processing Equipment
Transport of individuals and groups of persons and lug-
09.1.1 Equipment for the Reception, Recording, and
gage by two or more modes of transport when the
Reproduction of Sound and Pictures (D)
expenditure cannot be apportioned between them
Includes: transport of private vehicles. Television sets, video cassette players and recorders, and
Excludes: package holidays (09.6.0). television aerials of all types
Radio sets, car radios, radio clocks, two-way radios, ama-
07.3.6 Other Purchased Transport Services (S) teur radio receivers, and transmitters
Gramophones, tape players and recorders, cassette play-
Funicular, cable car, and chairlift transport ers and recorders, CD players, personal stereos, stereo
Removal and storage services systems and their constituent units (turntables, tuners,
355
APPENDIX
amplifiers, speakers, and so on), microphones, and Includes: total value of the service (that is, both the cost of
earphones labor and the cost of materials are covered).
Excludes: separate purchases of materials made by
Excludes: video cameras, camcorders, and sound-recording households with the intention of undertaking the repair
cameras (09.1.2). themselves (09.1.1), (09.1.2), or (09.1.3).
09.1.2 Photographic and Cinematographic Equipment
09.2 Other Major Durables for Recreation and Culture
and Optical Instruments (D)
Still cameras, movie cameras and sound-recording cam- 09.2.1 Major Durables for Outdoor Recreation (D)
eras, video cameras and camcorders, film and slide Camper vans, caravans, and trailers
projectors, enlargers and film processing equipment, Airplanes, microlight aircraft, gliders, hang gliders, and
and accessories (screens, viewers, lenses, flash attach- hot-air balloons
ments, filters, exposure meters, and so on) Boats, outboard motors, sails, rigging, and superstructures
Binoculars, microscopes, telescopes, and compasses Horses and ponies, horse- or pony-drawn vehicles, and
related equipment (harnesses, bridles, reins, saddles,
09.1.3 Information Processing Equipment (D) and so on)
Major items for games and sport such as canoes, kayaks,
Personal computers, visual display units, printers, and
windsurfing boards, sea-diving equipment, and golf
miscellaneous accessories accompanying them; com-
carts
puter software packages such as operating systems,
applications, languages, and so on Includes: fitting out of boats, camper vans, caravans, and
Calculators, including pocket calculators so on.
Typewriters and word processors Excludes: horses and ponies, horse- or pony-drawn vehi-
cles, and related equipment purchased for personal trans-
Includes: telefax and telephone-answering facilities pro- port (07.1.4); inflatable boats, rafts, and swimming pools
vided by personal computers. for children and the beach (09.3.2).
Excludes: prerecorded diskettes and CD-ROMs contain-
ing books, dictionaries, encyclopedias, foreign language 09.2.2 Musical Instruments and Major Durables for
trainers, multimedia presentations, and so on in the form of Indoor Recreation (D)
software (09.1.4); video game software (09.3.1); video game Musical instruments of all sizes, including electronic
computers that plug into a television set (09.3.1); typewriter musical instruments, such as pianos, organs, violins,
ribbons (09.5.4); toner and ink cartridges (09.5.4); and guitars, drums, trumpets, clarinets, flutes, recorders,
slide rules (09.5.4). harmonicas, and so on
Billiard tables, ping-pong tables, pinball machines, gam-
09.1.4 Recording Media (SD) ing machines, and so on
Records and compact discs
Prerecorded tapes, cassettes, video cassettes, diskettes Excludes: toys (09.3.1).
and CD-ROMs for tape recorders, cassette recorders, 09.2.3 Maintenance and Repair of Other Major Durables
video recorders, and personal computers for Recreation and Culture (S)
Unrecorded tapes, cassettes, video cassettes, diskettes and
Maintenance and repair of other major durables for recre-
CD-ROMs for tape recorders, cassette recorders, video
ation and culture
recorders, and personal computers
Unexposed films, cartridges, and disks for photographic Includes: total value of the service (that is, both the cost of
and cinematographic use labor and the cost of materials are covered); laying up for
winter of boats, camper vans, caravans, and so on; hangar
Includes: prerecorded tapes and compact discs of novels, services for private planes; marina services for boats; and vet-
plays, poetry, and so on; prerecorded diskettes and CD- erinary and other services (stabling, feeding, farriery, and so
ROMs containing books, dictionaries, encyclopedias, foreign on) for horses and ponies purchased for recreational purposes.
language trainers, multimedia presentations, and so on in the Excludes: fuel for recreational vehicles (07.2.2); separate
form of software; photographic supplies such as paper and purchases of materials made by households with the inten-
flashbulbs; and unexposed film the price of which includes tion of undertaking the maintenance or repair themselves
the cost of processing without separately identifying it. (09.2.1) or (09.2.2); and veterinary and other services for
Excludes: batteries (05.5.2); computer software packages pets (09.3.5).
such as operating systems, applications, languages, and so
on (09.1.3); video game software, video game cassettes, and 09.3 Other Recreational Items and Equipment, Gardens,
video game CD-ROMs (09.3.1); and development of films and Pets
and printing of photographs (09.4.2).
09.3.1 Games, Toys, and Hobbies (SD)
09.1.5 Repair of Audiovisual, Photographic, and Card games, parlor games, chess sets, and the like
Information Processing Equipment (S) Toys of all kinds including dolls, soft toys, toy cars and
Repair of audiovisual, photographic, and information trains, toy bicycles and tricycles, toy construction sets,
processing equipment puzzles, plasticine, electronic games, masks, disguises,
356
APPENDIX
jokes, novelties, fireworks and rockets, festoons, and 09.3.5 Veterinary and Other Services for Pets (S)
Christmas tree decorations Veterinary and other services for pets such as grooming,
Stamp-collecting requisites (used or canceled postage boarding, tattooing, and training
stamps, stamp albums, and so on), other items for
collections (coins, medals, minerals, zoological and Excludes: veterinary and other services (stabling, farriery,
botanical specimens, and so on), and other tools and and so on) for horses and ponies purchased for recreational
articles n.e.c. for hobbies purposes (09.2.3).
Includes: video game software; video game computers that 09.4 Recreational and Cultural Services
plug into a television set; video game cassettes, and video
game CD-ROMs. 09.4.1 Recreational and Sporting Services (S)
Excludes: collectors’ items falling into the category of Services provided by:
works of art or antiques (05.1.1); unused postage stamps
(08.1.0); Christmas trees (09.3.3); and children’s scrap- Sports stadiums, horse-racing courses, motor-racing cir-
books (09.5.1). cuits, velodromes, and so on
Skating rinks, swimming pools, golf courses, gymna-
09.3.2 Equipment for Sport, Camping, and Open-Air sia, fitness centers, tennis courts, squash courts, and
Recreation (SD) bowling alleys
Gymnastic, physical education, and sport equipment such Fairgrounds and amusement parks
as balls, shuttlecocks, nets, rackets, bats, skis, golf Roundabouts, seesaws, and other playground facilities
clubs, foils, sabers, poles, weights, discuses, javelins, for children
dumbbells, chest expanders, and other body-building Pinball machines and other games for adults other than
equipment games of chance
Parachutes and other skydiving equipment Ski slopes, ski lifts, and the like
Firearms and ammunition for hunting, sport, and personal Hire of equipment and accessories for sport and recre-
protection ation, such as airplanes, boats, horses, skiing, and
Fishing rods and other equipment for fishing camping equipment
Equipment for beach and open-air games, such as bowls, Out-of-school individual or group lessons in bridge,
croquet, frisbee, volleyball, and inflatable boats, rafts, chess, aerobics, dancing, music, skating, skiing, swim-
and swimming pools ming, or other pastimes
Camping equipment such as tents and accessories, sleep- Services of mountain guides, tour guides, and so on
ing bags, backpacks, air mattresses and inflating Navigational aid services for boating
pumps, camping stoves, and barbecues Includes: hire of game-specific footwear (ski boots, football
Repair of such articles boots, golfing shoes, and other such footwear fitted with ice
Includes: game-specific footwear (ski boots, football boots, skates, rollers, spikes, studs, and so on).
golfing shoes, and other such footwear fitted with ice-skates, Excludes: cable car and chairlift transport not at ski resorts
rollers, spikes, studs, and so on); protective headgear for or holiday centers (07.3.6).
sports; other protective gear for sports such as life jackets,
09.4.2 Cultural Services (S)
boxing gloves, body padding, shin guards, goggles, belts,
supports, and so on. Services provided by:
Excludes: crash helmets for motorcycles and bicycles Cinemas, theatres, opera houses, concert halls, music
(03.1.3); camping and garden furniture (05.1.1). halls, circuses, and sound and light shows
09.3.3 Gardens, Plants, and Flowers (ND) Museums, libraries, art galleries, and exhibitions
Historic monuments, national parks, zoological and
Natural or artificial flowers and foliage, plants, shrubs, botanical gardens, and aquaria
bulbs, tubers, seeds, fertilizers, composts, garden peat, Hire of equipment and accessories for culture, such as
turf for lawns, specially treated soils for ornamental gar- television sets, video cassettes, and so on
dens, horticultural preparations, pots, and pot holders Television and radio broadcasting, in particular license
Includes: natural and artificial Christmas trees; delivery fees for television equipment and subscriptions to tele-
charges for flowers and plants. vision networks
Excludes: gardening gloves (03.1.3); gardening services Services of photographers such as film developing, print
(04.4.4) or (05.6.2); gardening equipment (05.5.1); garden- processing, enlarging, portrait photography, wedding
ing tools (05.5.2); and insecticides and pesticides for house- photography, and so on
hold use (05.6.1). Includes: services of musicians, clowns, performers for pri-
09.3.4 Pets and Related Products (ND) vate entertainments.
Pets, pet foods, veterinary and grooming products for 09.4.3 Games of Chance (S)
pets, collars, leashes, kennels, birdcages, fish tanks, cat Service charges for lotteries, bookmakers, totalizators,
litter, and so on casinos, and other gambling establishments, gaming
Excludes: horses and ponies (07.1.4) or (09.2.1); veterinary machines, bingo halls, scratch cards, sweepstakes,
services (09.3.5). and so on (service charge is defined as the difference
357
APPENDIX
between the amounts paid for lottery tickets or placed It includes education by radio or television broadcasting.
in bets and the amounts paid out to winners) The breakdown of educational services is based upon the
level categories of the 1997 International Standard Classifi-
09.5 Newspapers, Books, and Stationery cation of Education (ISCED97) of the United Nations Edu-
cational, Scientific and Cultural Organization.
09.5.1 Books (SD)
Books, including atlases, dictionaries, encyclopedias, text- 10.1 Preprimary and Primary Education
books, guidebooks, and musical scores
10.1.0 Preprimary and Primary Education (S)
Includes: scrapbooks and albums for children; bookbinding.
Levels 0 and 1 of ISCED97: preprimary and primary
Excludes: prerecorded tapes and compact discs of novels,
education
plays, poetry, and so on (09.1.4); prerecorded diskettes and
CD-ROMs containing books, dictionaries, encyclopedias, Includes: literacy programs for students too old for primary
foreign language trainers, and so on in the form of software school.
(09.1.4); and stamp albums (09.3.1).
10.2 Secondary Education
09.5.2 Newspapers and Periodicals (ND)
10.2.0 Secondary Education (S)
Newspapers, magazines, and other periodicals
Levels 2 and 3 of ISCED97: lower-secondary and upper-
09.5.3 Miscellaneous Printed Matter (ND) secondary education
Catalogs and advertising material Includes: out-of-school secondary education for adults and
Posters, plain or picture postcards, calendars young people.
Greeting cards and visiting cards, announcement, and
message cards 10.3 Postsecondary Nontertiary Education
Maps and globes 10.3.0 Postsecondary Nontertiary Education (S)
Excludes: prefranked postcards and aerograms (08.1.0); Level 4 of ISCED97: postsecondary nontertiary education
stamp albums (09.3.1).
Includes: out-of-school postsecondary nontertiary education
09.5.4 Stationery and Drawing Materials (ND) for adults and young people.
Writing pads, envelopes, account books, notebooks, dia- 10.4 Tertiary Education
ries, and so on
Pens, pencils, fountain pens, ballpoint pens, felt-tip pens, 10.4.0 Tertiary Education (S)
inks, erasers, pencil sharpeners, and so on Levels 5 and 6 of ISCED97: first stage and second stage
Stencils, carbon paper, typewriter ribbons, inking pads, of tertiary education
correcting fluids, and so on
Paper punches, paper cutters, paper scissors, office glues 10.5 Education Not Definable by Level
and adhesives, staplers and staples, paper clips, draw-
ing pins, and so on 10.5.0 Education Not Definable by Level (S)
Drawing and painting materials such as canvas, paper, Educational programs, generally for adults, which do not
card, paints, crayons, pastels, and brushes require any special prior instruction, in particular voca-
Includes: toner and ink cartridges; educational materials tional training and cultural development
such as exercise books, slide rules, geometry instruments, Excludes: driving lessons (07.2.4); recreational training
slates, chalks, and pencil boxes. courses such as sport or bridge lessons given by indepen-
Excludes: pocket calculators (09.1.3). dent teachers (09.4.1).
09.6 Package Holidays
11 Restaurants and Hotels
09.6.0 Package Holidays (S) 11.1 Catering Services
All-inclusive holidays or tours which provide for travel,
food, accommodation, guides, and so on 11.1.1 Restaurants, Cafés, and the Like (S)
Catering services (meals, snacks, drinks, and refreshments)
Includes: half-day and one-day excursion tours; pilgrimages.
provided by restaurants, cafés, buffets, bars, tearooms, and
so on, including those provided:
10 Education
This division covers educational services only. It does not In places providing recreational, cultural, sporting or
include expenditures on educational materials, such as entertainment services: theatres, cinemas, sports stadi-
books (09.5.1) and stationery (09.5.4), or education support ums, swimming pools, sports complexes, museums, art
services, such as health care services (06), transport services galleries, nightclubs, dancing establishments, and so on
(07.3), catering services (11.1.2), and accommodation ser- On public transport (coaches, trains, boats, airplanes, and
vices (11.2.0). so on) when priced separately
358
APPENDIX
Also included are the following: vibrators, electric toothbrushes and other electric appli-
ances for dental hygiene, and so on
The sale of food products and beverages for immediate
Repair of such appliances
consumption by kiosks, street vendors, and the like,
including food products and beverages dispensed ready 12.1.3 Other Appliances, Articles, and Products for
for consumption by automatic vending machines Personal Care (ND)
The sale of cooked dishes by restaurants for consumption
off their premises Nonelectric appliances: razors and hair trimmers and
The sale of cooked dishes by catering contractors whether blades therefor, scissors, nail files, combs, shaving
collected by the customer or delivered to the customer’s brushes, hairbrushes, toothbrushes, nail brushes, hair-
home. pins, curlers, personal weighing machines, baby scales,
and so on
Includes: tips. Articles for personal hygiene: toilet soap, medicinal soap,
Excludes: tobacco purchases (02.2.0); telephone calls (08.3.0). cleansing oil and milk, shaving soap, shaving cream
and foam, toothpaste, and so on
11.1.2 Canteens (S)
Beauty products: lipstick, nail varnish, makeup and
Catering services of works canteens, office canteens and makeup removal products (including powder com-
canteens in schools, universities, and other educational pacts, brushes, and powder puffs), hair lacquers and
establishments lotions, pre-shave and after-shave products, sunbathing
Includes: university refectories, military messes, and wardrooms. products, hair removers, perfumes and toilet waters,
Excludes: food and drink provided to hospital inpatients personal deodorants, bath products, and so on
(06.3.0). Other products: toilet paper, paper handkerchiefs, paper
towels, sanitary towels, cotton wool, cotton tops,
11.2 Accommodation Services babies’ napkins, toilet sponges, and so on
Excludes: handkerchiefs made of fabric (03.1.3).
11.2.0 Accommodation Services (S)
Accommodation services of: 12.2 Prostitution

Hotels, boarding houses, motels, inns, and establish-
12.2.0 Prostitution (S)
ments offering “bed and breakfast”
Holiday villages and holiday centers, camping and car- Services provided by prostitutes and the like
avan sites, youth hostels, and mountain chalets
Boarding schools, universities, and other educational 12.3 Personal Effects N.E.C.
establishments
12.3.1 Jewelry, Clocks, and Watches (D)
Public transport (trains, boats, and so on) when priced
separately Precious stones and metals and jewelry fashioned out of
Hostels for young workers or immigrants. such stones and metals
Costume jewelry, cuff links, and tiepins
Includes: tips, porters. Clocks, watches, stopwatches, alarm clocks, travel clocks
Excludes: payments of households occupying a room in Repair of such articles
a hotel or boarding house as their main residence (04.1.1);
rentals paid by households for a secondary residence for Excludes: ornaments (05.1.1) or (05.4.0); radio clocks
the duration of a holiday (04.1.2); telephone calls (08.3.0); (09.1.1); precious stones and metals and jewelry fashioned
catering services in such establishments except for breakfast out of such stones and metals acquired primarily as stores of
or other meals included in the price of the accommodation value (capital formation).
(11.1.1); and housing in orphanages, homes for disabled or
maladjusted persons (12.4.0). 12.3.2 Other Personal Effects (SD)
Travel goods and other carriers of personal effects: suit-
12 Miscellaneous Goods and Services cases, trunks, travel bags, attaché cases, satchels, hand-
bags, wallets, purses, and so on
12.1 Personal Care Articles for babies: baby carriages, pushchairs, carrycots,
recliners, car beds and seats, back carriers, front carri-
12.1.1 Hairdressing Salons and Personal Grooming ers, reins and harnesses, and so on
Establishments (S) Articles for smokers: pipes, lighters, cigarette cases, cigar
Services of hairdressing salons, barbers, beauty shops, cutters, ashtrays, and so on
manicures, pedicures, Turkish baths, saunas, solari- Miscellaneous personal articles: sunglasses, walking
ums, nonmedical massages, and so on sticks and canes, umbrellas and parasols, fans, key-
rings, and so on
Includes: bodycare, depilation, and the like. Funerary articles: coffins, gravestones, urns, and so on
Excludes: spas (06.2.3) or (06.3.0); fitness centers (09.4.1). Repair of such articles
12.1.2 Electric Appliances for Personal Care (SD) Includes: lighter fuel; wall thermometers and barometers.
Electric razors and hair trimmers, handheld and hood Excludes: baby furniture (05.1.1); shopping bags (05.2.0);
hairdryers, curling tongs and styling combs, sunlamps, and feeding bottles (05.4.0).
359
APPENDIX
12.4 Social Protection 12.5.3 Insurance Connected with Health (S)

Social protection as defined here covers assistance and sup- Service charges for private sickness and accident insurance
port services provided to persons who are elderly, disabled,
having occupational injuries and diseases, survivors, unem- 12.5.4 Insurance Connected with Transport (S)
ployed, destitute, homeless, low-income earners, indigenous Service charges for insurance in respect of personal trans-
people, immigrants, refugees, alcohol and substance abus- port equipment
ers, and so on. It also covers assistance and support services Service charges for travel insurance and luggage
provided to families and children. insurance
12.4.0 Social Protection (S) 12.5.5 Other Insurance (S)
Such services include residential care, home help, day Service charges for other insurance such as civil liability
care, and rehabilitation. More specifically, this class for injury or damage to third parties or their property
covers payments by households for: Excludes: civil liability or damage to third parties or their
Retirement homes for elderly persons, residences for dis- property arising from the operation of personal transport
abled persons, rehabilitation centers providing long-term equipment (12.5.4).
support for patients rather than health care and rehabilita-
tive therapy, and schools for disabled persons where the 12.6 Financial Services N.E.C.
main aim is to help students overcome their disability
Help to maintain elderly and disabled persons at home 12.6.1 FISIM (S)
(home-cleaning services, meal programs, day care Financial intermediation services indirectly measured
centers, day care services, and holiday care services)
Wet-nurses, crèches, play schools, and other child- 12.6.2 Other Financial Services N.E.C. (S)
minding facilities
Actual charges for the financial services of banks, post
Counseling, guidance, arbitration, fostering, and adop-
offices, saving banks, money changers, and similar
tion services for families.
financial institutions
Fees and service charges of brokers, investment counsel-
12.5 Insurance
ors, tax consultants, and the like
Service charges for insurance are classified by type of insur-
Administrative charges of private pension funds and the
ance, namely, life insurance and nonlife insurance (that is,
like
insurance in connection with the dwelling, health, transport,
and so on). Service charges for multirisk insurance covering
12.7 Other Services N.E.C.
several risks should be classified on the basis of the cost of
the principal risk if it is not possible to allocate the service 12.7.0 Other Services N.E.C. (S)
charges to the various risks covered.
Service charge is defined as the difference between claims Fees for legal services, employment agencies, and so on
due and premiums earned and premium supplement. Charges for undertaking and other funeral services
Payment for the services of estate agents, housing
12.5.1 Life Insurance (S) agents, auctioneers, salesroom operators, and other
intermediaries
Service charges for life assurance, death benefit assur-
Payment for photocopies and other reproductions of
ance, education assurance, and so on
documents
12.5.2 Insurance Connected with the Dwelling (S) Fees for the issue of birth, marriage, and death certifi-
cates, and other administrative documents
Service charges paid by owner-occupiers and by tenants Payment for newspaper notices and advertisements
for the kinds of insurance typically taken out by tenants Payment for the services of graphologists, astrologers,
against fire, theft, water damage, and so on private detectives, bodyguards, matrimonial agencies
Excludes: service charges paid by owner-occupiers for the and marriage guidance counselors, public writers, mis-
kinds of insurance typically taken out by landlords (interme- cellaneous concessions (seats, toilets, and cloakrooms),
diate consumption). and so on
360
APPENDIX
Appendix 3 01.1 Food

Food purchased by the household mainly for consumption
Classification of Individual or preparation at home. It excludes food that is provided as
part of a food serving service.
Consumption According to Food is composed of all edible goods that are purchased
Purpose 2018 (COICOP and consumed by the household with the purpose of nourish-
ing. It includes cereals and cereal products; meat; fish and
2018)1 other seafood; milk, other dairy products, and eggs; oils and
fats; fruit and nuts; vegetables, tubers, plantains, cooking
bananas, and pulses; sugar, confectionery, and desserts; salt,
sauces and condiments, spices and culinary herbs, and seeds.
Introduction Includes:
• Products that need to be cooked and further prepared as
The Classification of Individual Consumption According
well as ready-made food
to Purpose (COICOP) serves as the international reference
classification of household expenditure. COICOP provides Excludes:
a framework of homogeneous categories of goods and ser- • Service of local delivery of food or drinks when sepa-
vices, which are considered a function or purpose of house- rately invoiced (07.4.9.2)
hold consumption expenditure. COICOP functions as an
• Products for animal feeding (09.3.2.2)
integral part of the System of National Accounts (SNA), but
it is also used in several other statistical areas, such as the • Food provided by food serving services (Division 11)
household budget survey or the consumer price index.
This appendix presents descriptions and explanations of 01.1.1 Cereals and Cereal Products (ND)
the COICOP revision endorsed in 2018 (COICOP 2018) by
the 49th Session of the United Nations Statistical Commis- 01.1.1.1 Cereals (ND)
sion. COICOP 2018 provides more detail than the previous Dried grains of cereals, whether or not precooked, but not
version (COICOP 1999), responding to users’ need for more further processed.
detail, and addresses several other issues that prompted the Includes:
revision of the classification. COICOP 2018 reflects the • Wheat
significant changes in goods and services in some areas,
improves the links of COICOP to other classifications, and • Rice, including parboiled rice
addresses emerging statistical and policy needs of several • Sorghum
international organizations. • Barley
• Millet
COICOP: Breakdown of Individual • Quinoa
Consumption Expenditure of • Maize
Households by Division and Group • Other cereals
Includes also:
01 Food and Nonalcoholic Beverages
Division 01 covers food and nonalcoholic beverages pur- • Teff
chased by the household mainly for consumption or prepa- • Rye
ration at home. It excludes food and nonalcoholic beverages • Oats
that are provided as part of a food serving service (Division
• Triticale
11).
Services for food processing for own consumption are • Buckwheat
also included in this division. • Canary seed
Food is composed of all edible goods that are purchased • Quihuicha or Inca wheat
and consumed by the household with the purpose of nour-
• Canagua or coaihua
ishing. It includes: cereals and cereal products; meat; fish
and other seafood; milk, other dairy products, and eggs; oils • Adlay or Job’s tears
and fats; fruit and nuts; vegetables, tubers, plantains, cook- • Mixed cereal grains, in the form of dried grains, but not
ing bananas, and pulses; sugar, confectionery, and desserts; further processed
salt, sauces and condiments, spices and culinary herbs, and Excludes:
seeds.
Division 01 also includes ready-made food that can be • Flour of cereals (01.1.1.2)
eaten as it is or requires heating. Nonalcoholic beverages • Breakfast cereals (01.1.1.4)
include drinks that do not contain any alcohol. • Ready-made preparations based on cereals, such as
ready-made soups based on cereals (01.1.9.1)
1 • Baby rice cereals and other baby cereals (01.1.9.2)
As of this draft, COICOP 2018 has not yet been finalized. While only
minor edits are expected, additional changes could be identified. • Seeds for planting (09.3.1.2)
361
APPENDIX
01.1.1.2 Flour of Cereals (ND) Excludes:

Includes: • Soy meat and burgers, veggie burgers, tofu, tempeh,
• Flour of cereals mentioned in subclass 01.1.1.1 and other meat substitutes made from vegetables and
Excludes: nuts; potato starch, tapioca, sago, and other starches
(01.1.7.9)
• Flour as baby food, baby rice cereals, and other baby
• Ready-made preparations based on cereals, such as
cereals (01.1.9.2)
ready-made soups based on cereals (01.1.9.1)
01.1.1.3 Bread and Bakery Products (ND)
01.1.2 Live Animals, Meat, and Other Parts of
Includes:
Slaughtered Land Animals (ND)
• Bread and bread rolls
01.1.2.1 Live Land Animals (ND)
• Crispbread, rusks, toasted bread, crackers
• Tortillas Live land animals, both domestic and wild, for food purpose.
Includes:
• Injera
• Pizza bases without topping, whether precooked or not • Cattle (cow, veal, common zebu or humped ox, watussi
ox, gaur, gayal, banteng, Tibetan yak)
• Gingerbread and the like
• Buffaloes (Indian or water buffaloes, Asiatic buffalo or
• Sweet biscuits (cookies)
arni, Celebese anoa or pigmy buffaloes, African buffa-
• Waffles and wafers loes, such as the dwarf buffaloes and the large Caffrarian
• Ice cream cones buffaloes, the American bison or “buffalo” and the Euro-
• Crumpets, muffins, croissants, cakes, sweet tarts, sweet pean bison, the “beefalo” [a cross between a bison and a
pies, and other pastry goods and cakes domestic beef animal])
• Pigs
Excludes:
• Goats, lambs, and sheep
• Pizza (with topping), quiche, meat, or fish pies (01.1.9.1)
• Poultry (chicken, ducks, geese, turkeys, and guinea fowl)
01.1.1.4 Breakfast Cereals (ND) • Horses, donkeys, and other equines
Includes: • Camels, dromedaries, alpaca, llama, guanaco, vicuña, and
other camelids
• Cornflakes, oatmeal and oat flakes, muesli, granola, and
puffed cereals including puffed rice cakes • Seals, walruses, whales, and other marine mammals
• Breakfast cereals with nuts or dried fruit • Antelopes, deer, boars, kangaroos
• Ostriches, emus, rhea, pheasant, grouse, pigeon, quail,
Excludes: and other birds
• Popcorn (1.1.1.9) • Snakes, alligators, and other reptiles
• Spiders, scorpions, and other insects and worms
01.1.1.5 Macaroni, Noodles, Couscous, and Similar
Pasta Products (ND) • Terrestrial snails
Includes: • Frogs
• Pasta uncooked, whether stuffed or not, and couscous Excludes:

• Uncooked dumplings, ravioli, and similar • Live fish (01.1.3.1)
• Bulgur • Other live seafood (01.1.3.4)
Excludes: • Live animals for transport (07.1.4.0)
• Prepared dishes containing stuffed pasta; prepared cous- 01.1.2.2 Meat, Fresh, Chilled, or Frozen (ND)
cous dishes (01.1.9.1) Meat of all animals, both domestic and wild, fresh, chilled,
01.1.1.9 Other Cereal and Grain Mill Products (ND) or frozen.
Includes:
Includes:
• Meat of cattle (cow, veal, common zebu or humped ox,
• Mixes and doughs for the preparation of bakery watussi ox, gaur, gayal, banteng, Tibetan yak)
products
• Meat of buffaloes (Indian or water buffaloes, Asiatic buf-
• Popcorn falo or arni, Celebese anoa or pigmy buffaloes, African
• Granola bars buffaloes, such as the dwarf buffaloes and the large Caf-
• Malt; malt extract frarian buffaloes, the American bison or “buffalo” and the
• Chips and crisps of cereals European bison, the “beefalo” [a cross between a bison
and a domestic beef animal])
Includes also: • Meat of pigs
• Seitan and other meat substitutes made from cereals • Meat of goats, lambs, and sheep
362
APPENDIX
• Meat of poultry (chicken, ducks, geese, turkeys, and Excludes:

guinea fowl)
• Marine mammals, frogs, and terrestrial snails (01.1.2.1,
• Meat of horses, donkeys, and other equines 01.1.2.2, 01.1.2.3)
• Meat of camels, dromedaries, alpaca, llama, guanaco, • Seaweeds and other aquatic plants (01.1.7.4)
vicuña, and other camelids
• Meat of seals, walruses, whales, and other marine mammals 01.1.3.1 Fish, Live, Fresh, Chilled, or Frozen (ND)
• Meat of antelopes, deer, boars, and kangaroos Fish, live, fresh, chilled, or frozen.
• Meat of ostriches, emus, rhea, pheasant, grouse, pigeon, Includes:
quail, and other birds • Fish live for food purpose
• Meat of snakes, alligators, and other reptiles • Fish fresh, chilled, or frozen
• Meat of spiders, scorpions, and other insects and worms • Fish fillets and meat, minced or not, fresh, chilled, or frozen
• Meat of terrestrial snails
Excludes:
• Meat of frogs
• Livers, roes, fins, and other offal (01.1.3.7)
Includes also:
• Minced meat of animals mentioned previously 01.1.3.2 Fish, Dried, Salted, in Brine or Smoked (ND)
Fish, dried, salted or in brine; smoked.
01.1.2.3 Meat, Dried, Salted, in Brine or Smoked (ND) Includes:
Meat of all animals dried, salted, in brine or smoked. • Fish fillets and meat dried, salted or in brine; smoked
Includes:
Excludes:
• Bacon, ham, salami
• Livers, roes, fins, and other offal (01.1.3.7)
Excludes:
01.1.3.3 Fish Preparations (ND)
• Pâté (01.1.2.5)
Prepared foodstuffs made of fish whether frozen or not.
01.1.2.4 Offal, Blood, and Other Parts of Slaughtered Includes:
Animals, Fresh, Chilled, or Frozen, Dried, Salted, in • Fillets merely covered with batter or bread crumbs,
Brine or Smoked (ND) whether or not frozen
Offal and other parts of slaughtered animals fresh, chilled or • Crabmeat and surimi
frozen, dried, salted, in brine or smoked.
Includes: Excludes:
• Bones • Caviar and caviar substitutes and prepared and preserved
• Pork heads, tails, and ears; chicken feet
shark fins; livers, roes, fins, and other offal (01.1.3.7)
Excludes: 01.1.3.4 Other Seafood, Live, Fresh, Chilled,

or Frozen (ND)
• Offal, blood, and other parts of slaughtered animals'
preparations (01.1.2.5) Includes:
• Crabs, lobsters, crayfish, krill, sea spider, shrimps and
01.1.2.5 Meat, Offal, Blood, and Other Parts of prawns, and other crustaceans, live, fresh, chilled, or
Slaughtered Animals Preparations (ND) frozen
Meat, offal or blood preparations, whether frozen or not. • Cuttlefish, squid, octopus, sea snails, abalone and
Includes: bivalves, such as oysters, scallops, mussels, clams, cock-
• Sausages and similar products of meat, offal, or blood les and ark shells, and other molluscs, live, fresh, chilled,
or frozen
• Marinated meat
• Sea urchins, sea cucumbers (bêches-de-mer) and jelly-
• Canned meat, meat extracts, meat juices fish, and other aquatic invertebrates, live, fresh, chilled,
• Minced meat, if mixed meat from more than one kind of or frozen
animal’s minced meat
Excludes:
• All kind of pâté, including liver pâté
• Breaded meat • Terrestrial molluscs, including terrestrial snails, live,
fresh, chilled, or frozen (01.1.2.1, 01.1.2.2, 01.1.2.3)
• Other preparations of meat, meat offal, or blood
• Seaweeds and other edible aquatic plants (01.1.7.4)
01.1.3 Fish and Other Seafood (ND)
Fish and other seafood, such as crustaceans, molluscs, and 01.1.3.5 Other Seafood, Dried, Salted, in Brine or
other aquatic invertebrates, as whole or part of (that is cuts, Smoked (ND)
fillets, meat, minced or not), including livers, roes, fins, Other seafood as defined in 01.1.3.4, dried, salted or in
meal, and other offal. brine; smoked.
363
APPENDIX
01.1.3.6 Other Seafood Preparations (ND) Excludes:

Prepared foodstuffs made of other seafood as defined in • Tofu (01.1.7.9)
01.1.3.4 whether frozen or not.
01.1.4.6 Yogurt and Similar Products (ND)
01.1.3.7 Livers, Roes, and Offal of Fish and of Other
Includes:
Seafood in All Forms (ND)
Includes: • Yogurt
• Buttermilk
• Livers, roes, fins, maws, meal, and other offal in all prod-
uct forms, that is, fresh, chilled, frozen, dried, salted, in • Curdled milk and cream
brine or smoked, prepared or preserved • Kefir and other fermented or acidified milk and cream
whether or not concentrated or containing added sugar or
Excludes: other sweetening matter or flavored or containing added
• Seaweeds and other aquatic plants (01.1.7.4) fruit, nuts, or cocoa
• Yogurt from nonanimal milk (for example, coconut
01.1.4 Milk, Other Dairy Products, and Eggs (ND) yogurt, soy yogurt)
01.1.4.1 Raw and Whole Milk (ND) Excludes:
Raw and whole milk of cattle and buffalo, sheep and goat, • Frozen yogurt (01.1.8.6)
camels and other animals, fresh, pasteurized, sterilized
(UHT), and reconstituted. 01.1.4.7 Milk-Based Dessert and Beverages (ND)
Excludes: Includes:
• Curdled, fermented, or acidified milk and cream (01.1.4.6) • Milk-based desserts from the milk of animal and nonani-
mal origin
01.1.4.2 Skimmed Milk (ND)
• Beverages based on milk of animal and nonanimal origin
Skimmed and semiskimmed milk of cattle and buffalo, flavored with cocoa, coffee, or other substances
sheep and goat, camels, and other animals.
Includes also:
01.1.4.3 Other Milk and Cream (ND) • Puddings based on milk of animal and nonanimal origin;
Includes: panna cotta; crème brûlée; and crema catalana
• Powdered, whole or skimmed milk Excludes:
• Evaporated and condensed milk • Lemon curd and similar (01.1.8.3)
• Fresh, clotted, thickened, and whipped cream
01.1.4.8 Eggs (ND)
Includes also:
Includes:
• Baked milk
• Eggs of hen and other birds in shell, fresh
Excludes: • Eggs of other animals, including turtle eggs, in shell, fresh
• Beverages flavored with cocoa, coffee, or other substances • Eggs in shell or not, preserved or cooked
(01.1.4.7) • Eggs yolks, fresh or preserved
• Condensed, evaporated or powdered milk as baby food • Eggs albumin
(01.1.9.2)
Excludes:
01.1.4.4 Nonanimal Milk (ND)
• Fish roes (01.1.3.7)
Milk from nonanimal origin.
Includes: • Omelets, crepes, and other ready-made food products
based on eggs (01.1.9.1)
• Almond milk
• Coconut milk drink 01.1.4.9 Other Dairy Products (ND)
• Oat milk Includes:
• Rice milk • Whey and casein
• Soy milk • Cream concentrated or containing added sugar or other
• Other milk from vegetables and nuts sweetening matter; cream powder
• Other products consisting of milk constituents, milk
Excludes: protein concentrates, and products obtained from
• Coconut milk for cooking (01.1.9.3) whey
01.1.4.5 Cheese (ND) Excludes:

All kinds of cheese (fresh hard, semihard, blue cheese, cot- • Butter and other fats and oils derived from milk (01.1.5.2)
tage cheese) and curd both from animal and nonanimal milk. • Ice cream (01.1.8.6)
364
APPENDIX
01.1.5 Oils and Fats (ND) • Pineapples

• Coconut
01.1.5.1 Vegetable Oils (ND)
• Breadfruits
All oil of vegetable origin, including oil extracted from fruit,
nuts, and other crops. • Other tropical and subtropical fruit, fresh
Includes: Excludes:
• Sunflower-seed and safflower oil • Plantains and cooking bananas (01.1.7.5)
• Palm oil; olive oil
01.1.6.2 Citrus Fruits, Fresh (ND)
• Soya bean oil
Includes:
• Groundnut oil
• Rapeseed, colza, and mustard oil • Oranges
• Corn oil • Pomelos and grapefruit
• Cotton oil • Lemons and limes
• Other oils of vegetable origin • Tangerines
• Mandarins and clementines
Includes also:
• Other citrus fruits, fresh
• Coconut oil
01.1.6.3 Stone Fruits and Pome Fruits, Fresh (ND)
• Avocado oil
Includes:
• Rice bran oil
• Apples
01.1.5.2 Butter and Other Fats and Oils Derived from
• Pears and quinces
Milk (ND)
• Apricots
Includes:
• Cherries and sour cherries
• Butter • Peaches and nectarines
• Butter oil • Plums and sloes
• Ghee • Other pome fruits and stone fruits, fresh
01.1.5.3 Margarine and Similar Preparations (ND) 01.1.6.4 Berries, Fresh (ND)
Includes: Includes:
• Margarine • Currants and gooseberries
• Diet margarine • Raspberries
Excludes: • Blackberries
• Peanut butter (01.1.8.4) • Mulberries and loganberries
• Strawberries
01.1.5.9 Other Animal Oils and Fats (ND)
• Other berries fresh
Includes:
01.1.6.5 Other Fruits, Fresh (ND)
• Pig fat and lard, also including leaves or other spices
• Fat from cattle, buffalos, sheep, goats, and poultry Includes:
• Greaves fat and oleo stock • Grapes
• Fats and oils of fish • Kiwi fruit
Excludes: • Cantaloupes and other melons
• Cod or halibut liver oil (06.1.1.1) • Watermelons
• Pomegranates
01.1.6 Fruits and Nuts (ND) • Other fruits, fresh, n.e.c.
01.1.6.1 Dates, Figs, and Tropical Fruits, Fresh (ND) 01.1.6.6 Frozen Fruit (ND)
Includes: Fruit, uncooked or cooked, frozen, whether or not contain-
ing added sugar or other sweetening matter.
• Dates
• Figs 01.1.6.7 Fruit, Dried, and Dehydrated (ND)
• Avocados Includes:
• Bananas • Raisins
• Mangoes, guavas, and mangosteens • Prunes
• Papayas • Dried apricots
365
APPENDIX
• Dried coconut • Squash and gourds

• Other dried fruits • Other fruit-bearing vegetables, fresh or chilled
Excludes: 01.1.7.3 Green Leguminous Vegetables, Fresh or
• Fruit flour (01.1.6.9) Chilled (ND)
Includes:
01.1.6.8 Nuts, in Shell or Shelled (ND)
• Beans
Includes:
• Peas
• Almonds
• Broad beans and horse beans green
• Cashew nuts
• Soya beans
• Chestnuts
• Other green leguminous vegetables, fresh or chilled
• Hazelnuts
• Pistachios 01.1.7.4 Other Vegetables, Fresh or Chilled (ND)
• Walnuts Includes:
• Brazil nuts • Carrots and turnips
• Groundnuts • Garlic
• Other nuts in shell or shelled • Onions
• Leeks and other alliaceous vegetables
01.1.6.9 Fruit and Nuts Ground and Other Preparations (ND)
• Other root, bulb, and tuberous vegetables, fresh or chilled
Fruit and nuts preparations whether frozen or not.
Includes: • Mushrooms and truffles
• Seaweeds and other aquatic plants
• Fruit flour
• Olives
• Nuts, groundnuts, and other seeds, roasted, salted, or oth-
erwise prepared • Other vegetables, fresh or chilled n.e.c.
• Canned fruit • Mixtures of vegetables, fresh or chilled
• Homogenized fruit preparations 01.1.7.5 Tubers, Plantains, and Cooking Bananas (ND)
• Fruit pickles Includes:
Excludes: • All types of potatoes including sweet potatoes
• Jams, fruit jellies, marmalades, and fruit purée and pastes • Cassava; manioc and yucca
(01.1.8.3) • Yams
• Nut purée, nut butter, and nut pastes (01.1.8.4) • Taro
• Homogenized preparations as baby food (01.1.9.2) • Yautia, also known as malanga, new cocoyam, ocumo,
tannia
01.1.7 Vegetables, Tubers, Plantains, Cooking Bananas,
• Plantains and cooking bananas
and Pulses (ND)
Includes also:
01.1.7.1 Leafy or Stem Vegetables, Fresh or Chilled (ND) • Arrowroots, lotus roots, salep, Jerusalem artichokes,
Includes: topinambur, and Tacca
• Asparagus Excludes:
• Cabbages • Dessert bananas (01.1.6.1)
• Cauliflowers and broccoli
01.1.7.6 Pulses (ND)
• Lettuce and chicory
Dried leguminous vegetables.
• Spinach
Includes:
• Artichokes
• Common beans and other beans
• Other leafy or stem vegetables, fresh or chilled
• Broad beans and horse beans, also known as fava beans
01.1.7.2 Fruit-Bearing Vegetables, Fresh or Chilled (ND) • Chickpeas
Includes: • Lentil
• Chilies and peppers • Peas
• Cucumbers and gherkins • Cowpeas
• Eggplants (aubergines) • Pigeon peas
• Tomatoes • Bambara beans
• Pumpkins • Mixtures of pulses
366
APPENDIX
01.1.7.7 Other Vegetables, Tubers, Plantains, and 01.1.8.2 Other Sugar and Sugar Substitutes (ND)
Cooking Bananas, Dried and Dehydrated (ND) Includes:
Vegetables, other than leguminous, tubers, plantains, and • Sugar other than cane and beet (for example, coconut sugar)
cooking bananas dried and dehydrated.
• Stevia
Includes:
• Glucose and glucose syrup; fructose and fructose syrup
• Dried soybeans
• Lactose and lactose syrup
• Dried potatoes
• Invert sugar
• Dehydrated garlic and onions
• Artificial honey, saccharin, and other artificial sweeteners
Excludes: • Refined cane or beet sugar, in solid form, containing
• Vegetable flours (01.1.7.9) added flavoring or coloring matter
01.1.7.8 Vegetables, Tubers, Plantains, and Cooking • Maple sugar and maple syrup; caramel; molasses; sugars
Bananas, Frozen (ND) and sugar syrups n.e.c.
Includes: 01.1.8.3 Jams, Fruit Jellies, Marmalades, Fruit Purée
• Frozen vegetables in 01.1.7.1–01.1.7.4 and Pastes, and Honey (ND)
• Frozen tubers Includes:
• Frozen plantains and cooking bananas • Honey, jams, marmalades, compotes, jellies, and fruit
Excludes: purées and pastes
Includes also:
• Frozen preparations, such as frozen, chipped potatoes (01.1.7.9)
• Lemon curd and other fruit curd
01.1.7.9 Vegetables, Tubers, Plantains, Cooking Bananas
and Pulses Ground, and Other Preparations (ND) 01.1.8.4 Nut Purée, Nut Butter, and Nut Pastes (ND)
Vegetables, tubers, plantains, cooking bananas, and pulses Includes:
preparations, whether frozen or not. • Peanut butter
Includes:
• Other nut butter, such as almond butter, cashew butter,
• Flours of vegetables, pulses, tubers, plantains, and cook- hazelnut butter, macadamia nut butter, pecan butter, pista-
ing bananas chio butter, and walnut butter
• Canned vegetables
01.1.8.5 Chocolate, Cocoa, and Cocoa-Based Food
• Preserved olives, vegetable flakes, vegetable purée, veg- Products (ND)
etable chips, and crisps
Includes:
• Frozen, chipped potatoes
• Cocoa (including cocoa beans) and cocoa powder for all
• Vegetable concentrates
purposes
• Homogenized preparations
• Chocolate in bars or slabs, including white chocolate
• Vegetable pickles
• Chocolate and cocoa-based foods and cocoa-based dessert
Includes also: • Chocolate and cocoa-based creams or spreads
• Soy meat and burgers, veggie burgers, tofu, tempeh, and • Chocolate-covered marshmallows and chocolate-cov-
other meat substitutes made from vegetables and nuts ered jelly if the product is composed of a majority of
• Kocho (flatbread made of plant stem) chocolate
• Potato starch, tapioca, sago, and other starches Excludes:
Excludes: • Cocoa and chocolate-based drinks (01.2.4.0)
• Ready-made vegetables prepared, frozen or not, includ- 01.1.8.6 Ice, Ice Cream, and Sorbet (ND)
ing other ingredients, such as cheese or meat/fish; ready- Includes:
made soups (01.1.9.1)
• Homogenized preparations as baby food (01.1.9.2) • Ice cubes for drinks
• Culinary herbs and spices (01.1.9.4) • Ice cream and kulfi
• Broths and stocks containing vegetables (01.1.9.9) • Sorbet
• Vegetable juices (01.2.1.0) • Frozen yogurt
• Ice pop
01.1.8 Sugar, Confectionery, and Desserts (ND)
Includes also:
01.1.8.1 Cane and Beet Sugar (ND)
• Tofu ice cream
Includes:
Excludes:
• Cane or beet sugar, raw or refined, powdered, crystallized
or in lumps • Ice for cooling (04.5.5.0)
367
APPENDIX
01.1.8.9 Other Sugar Confectionery and Desserts • Homogenized baby food

N.E.C. (ND)
Excludes:
Includes:
• Cereals and flour not intended for baby use exclusively
• Desserts n.e.c. (01.1.1.1, 01.1.1.2)
• Vegetables, fruit, nuts, fruit peel, and other parts of plants, • Powdered milk not intended for baby use exclusively
preserved by sugar (01.1.4.3)
• Chewing gum, toffees, lollies, candies, and pastilles • Yogurt for children (01.1.4.6)
• Other confectionary products • Homogenized fruit and vegetable preparations not
Excludes: intended for baby use exclusively (01.1.6.9; 0.1.1.7.9)
• Sugar confectionery based on cocoa and chocolate 01.1.9.3 Salt, Condiments, and Sauces (ND)
(01.1.8.5) Includes:
01.1.9 Ready-Made Food and Other Food Products • Salt, sauces, condiments and seasonings (mustard, may-
N.E.C. (ND) onnaise, ketchup, soy sauce, and so on), and vinegar
01.1.9.1 Ready-Made Food (ND) Includes also:
Prepared food and meals that can be eaten as is or after heat- • Coconut milk for cooking
ing but that do not require cooking. They contain mixed
Excludes:
ingredients and can be fresh, frozen, or dehydrated/instant.
They can include sauces and dressing which, especially • Coconut milk drink (01.1.4.4)
when food is fresh, can be provided in separate bags. Dis- • Fruit and vegetable pickles (01.1.7.9)
posable fork, knife, spoon, or chopsticks are sometimes also
included in the package. 01.1.9.4 Spices, Culinary Herbs, and Seeds (ND)
Includes: Includes:
• Precooked dishes containing stuffed pasta, rice, and other • Pepper, pimento, ginger, and other spices
cereals, such as couscous dishes, including vegetables, • Parsley, rosemary, thyme, and other culinary herbs
meat, fish, cheese, or other ingredients; ready to eat dump-
• Poppy seed, sesame seed, linseed, and other seeds
lings, ravioli, noodles, and similar with sauce
• Ready-made meals and dishes based on precooked meat, Excludes:
meat substitutes, and fish • Vegetable oils (01.1.5.1)
• TV dinners • Seeds for plating (09.3.1.2)
• Composed salads and other prepared dishes and meals
based on vegetables, pulses, and potatoes also including 01.1.9.9 Other Food Products N.E.C. (ND)
other ingredients, such as meat, fish, and cheese Includes:
• Sandwiches, pizzas, quiches, meat, or fish pies, frozen or not
• Sugar cane consumed for extracting juice or as a snack
• Omelets, crepes, and other food products based on eggs
• Prepared baking powders and yeasts, broth, stocks, bouillon
when precooked and served as a main dish cubes, soup bases, agar-agar, instant dessert preparations
• Ready-made soups including dehydrated and instant
• Nutritional supplements and fortified food products
soups and stews
• Other prepared ready-made dishes and meals n.e.c. 01.2 Nonalcoholic Beverages
Nonalcoholic beverages purchased by the household,
Excludes: regardless of where these are consumed excluding bever-
• Bread and bakery products (01.1.1.3) ages that are provided as part of a food and beverage serving
• Macaroni, noodles, couscous, and similar pasta products service. (Division 11)
uncooked, whether stuffed or not (01.1.1.5) Includes:
• Cheese (01.1.4.5) and yogurt (01.1.4.6) • Beverages that do not contain any alcohol
• Cakes (0.1.1.1.3), ice cream (0.1.1.8.6), and other des- Excludes:
serts n.e.c. (01.1.8.9) • Nonalcoholic beverages for immediate consumption pro-
• Frozen, chipped potatoes (01.1.7.9) vided by a serving service (11.1.1)
• Milk (01.1.4.1, 01.1.4.2, 01.1.4.3, 01.1.4.4)
01.1.9.2 Baby Food (ND)
Food that is for baby use exclusively. 01.2.1 Fruit and Vegetable Juices (ND)
Includes: Includes:
• Baby formula (powdered, condensed, and evaporated • Fruit and vegetable juices unfermented and not contain-
milk for baby use) ing added alcohol, whether or not containing added sugar
• Baby rice cereals and flour for baby meals or other sweetening matter
368
APPENDIX
• Concentrated juices and frozen juices 01.2.3.0 Tea, Maté, and Other Plant Products for
• Powdered juices Infusion (ND)
Includes:
Excludes:
• Green tea (not fermented), black tea (fermented) and partly
• Sparkling juices (01.2.6.0)
fermented tea, maté, and other plant products for infusion
01.2.1.0 Fruit and Vegetable Juices (ND) • Tea substitutes and extracts and essences of tea
Includes: Includes also:
• Fruit and vegetable juices unfermented and not contain- • Fruit and herbal tea
ing added alcohol, whether or not containing added sugar
• Rooibos tea
or other sweetening matter
• Instant tea
• Concentrated juices and frozen juices
• Iced tea
• Powdered juices
Excludes: 01.2.4 Cocoa Drinks (ND)
Includes:
• Sparkling juices (01.2.6.0)
• Cocoa and chocolate-based drinks
01.2.2 Coffee and Coffee Substitutes (ND) Excludes:
Includes:
• Milk flavored with chocolate or cacao (01.1.4.7)
• Coffee, whether or not decaffeinated, roasted or ground,
• Cocoa powder for all purposes; chocolate in bars or
including instant coffee
slabs; cocoa-based food and cocoa-based dessert prepa-
• Coffee substitutes rations (01.1.8.5)
• Extracts, essences, and concentrates of coffee
01.2.4.0 Cocoa Drinks (ND)
• Coffee-based beverage preparations
Includes:
Includes also: • Cocoa and chocolate-based drinks
• Roasted chicory and other roasted coffee substitutes, and
Excludes:
extracts, essences, and concentrates thereof
• Milk flavored with chocolate or cacao (01.1.4.7)
Excludes:
• Cocoa powder for all purposes; chocolate in bars or
• Milk flavored with coffee (01.1.4.7) slabs; cocoa-based food and cocoa-based dessert prepa-
01.2.2.0 Coffee and Coffee Substitutes (ND)
rations (01.1.8.5)
Includes: 01.2.5 Water (ND)
Mineral or spring waters, still or sparkling, not added with
• Coffee, whether or not decaffeinated, roasted or ground,
other ingredients.
including instant coffee Excludes:
• Coffee substitutes
• Flavored water (01.2.9.0)
• Extracts, essences, and concentrates of coffee
• Coffee-based beverage preparations 01.2.5.0 Water (ND)
Includes also: Mineral or spring waters, still or sparkling, not added with
other ingredients.
• Roasted chicory and other roasted coffee substitutes, and Excludes:
extracts, essences, and concentrates thereof
Excludes:
01.2.6 Soft Drinks (ND)
• Milk flavored with coffee (01.1.4.7)
Includes:
01.2.3 Tea, Maté, and Other Plant Products for Infusion (ND) • Soft drinks, such as sodas, lemonades, and colas
Includes:
• Sparkling juices
• Green tea (not fermented), black tea (fermented) and partly
Excludes:
fermented tea, maté, and other plant products for infusion
• Tea substitutes and extracts and essences of tea • Sparkling water (01.2.5.0)
Includes also:
01.2.6.0 Soft Drinks (ND)
• Fruit and herbal tea
• Rooibos tea Includes:
• Instant tea • Soft drinks, such as sodas, lemonades, and colas
• Iced tea • Sparkling juices
369
APPENDIX
Excludes: Excludes:
• Sparkling water (01.2.5.0) • Food processing for business purpose (not in the scope of
• Flavored water (01.2.9.0) COICOP)
• Fruit/vegetable crushing and pressing services for the
01.2.9 Other Nonalcoholic Beverages (ND) production of alcoholic beverages (02.2.1.0)
Includes:
• Flavored water 02 Alcoholic Beverages, Tobacco, and
Narcotics
• Energy drinks, energy supplement, and protein powder
for drink preparation Division 02 covers the purchase of alcoholic beverages and
of tobacco products and narcotics, regardless of where these
• Birch juice and sap, aloe vera juice, and drinks
are consumed but not provided as part of a food and bever-
• Syrups and concentrates for the preparation of beverages age serving service.
• Other nonalcoholic beverages It also includes low and nonalcoholic beverages, which
are generally alcoholic, such as nonalcoholic beer. Services
01.2.9.0 Other Nonalcoholic Beverages (ND) for the production of alcohol for own consumption are also
Includes: included in this division.
Division 02 excludes alcoholic beverages purchased for
• Flavored water
immediate consumption in hotels, restaurants, cafes, bars, kiosks,
• Energy drinks, energy supplement, and protein powder street vendors, automatic vending machines, and so on (11.1.1).
for drink preparation
• Birch juice and sap, aloe vera juice, and drinks 02.1 Alcoholic Beverages
The beverages classified here include low or nonalcoholic
• Syrups and concentrates for the preparation of beverages
beverages which generally contain some alcohol.
• Other nonalcoholic beverages
02.1.1 Spirits and Liquors (ND)
01.3 Services for Processing Primary Goods for Includes:
Food and Nonalcoholic Beverages • Eaux-de-vie, liqueurs, and other spirits with high alcohol content
Services purchased for the processing of primary prod-
ucts provided by households to produce food and non- • Mead
alcoholic beverages for own final consumption by • Pomace brandy, such as pisco, grappa, marc, and so on
households. • Aperitifs other than wine-based aperitifs
01.3.0 Services for Processing Primary Goods for Food Excludes:

and Nonalcoholic Beverages (S) • Wine-based aperitifs (02.1.2.1, 02.1.2.2)
Services purchased for the processing of primary products
provided by households to produce food and nonalcoholic 02.1.1.0 Spirits and Liquors (ND)
beverages for own final consumption by households. Includes:
Includes:
• Eaux-de-vie, liqueurs, and other spirits with high alcohol content
• Grinding of cereals for flour production
• Mead
• Oil pressing
• Pomace brandy, such as pisco, grappa, marc, and so on
• Fruit/vegetable crushing and pressing services for the
• Aperitifs other than wine-based aperitifs
production of juices
Excludes:
Excludes:
• Wine-based aperitifs (02.1.2.1, 02.1.2.2)
• Food processing for business purpose (not in the scope of
COICOP) 02.1.2 Wine (ND)
• Fruit/vegetable crushing and pressing services for the Includes:
production of alcoholic beverages (02.2.1.0)
• Wine, cider, and perry, including sake
01.3.0.0 Services for Processing Primary Goods for • Wine-based aperitifs, fortified wines, champagne, and
Food and Nonalcoholic Beverages (S) other sparkling wines
Services purchased for the processing of primary products • Ice wine
provided by households to produce food and nonalcoholic • Low and nonalcoholic wine
beverages for own final consumption by households. • Wine-based aperitifs
Includes:
02.1.2.1 Wine from Grapes (ND)
• Grinding of cereals for flour production
• Oil pressing Includes:
• Fruit/vegetable crushing and pressing services for the • Wine from grapes
production of juices • Fortified wines, such as vermouth, sherry, port wine
370
APPENDIX
• Champagne and other sparkling wines from grapes • Brewing services

• Ice wine • Aging and bottling services
• Low and nonalcoholic wine
• Wine-based aperitifs 02.3 Tobacco
This group covers all purchases of tobacco and tobacco
02.1.2.2 Wine from Other Sources (ND) products by households, including purchases of tobacco in
restaurants, cafés, bars, service stations, and so on due to
Includes: these venues not adding value, or a service, to the tobacco
• Cider and perry, including sake products sold. The consumption of tobacco through shisha
or hookah pipes in these venues, such as restaurants, cafés,
02.1.3 Beer (ND) or bars, is included in Division 11 as the venue does pro-
Includes: vide a service. Electronic cigarette refills are included in this
• All kinds of beer, such as ale, lager, stout, and porter group—even though they do not contain tobacco—as they
are a substitute for smoking tobacco products.
• Low-alcoholic beer and nonalcoholic beer
02.1.3.0 Beer (ND) 02.3.0 Tobacco (ND)

Includes:
Includes:
• Cigarettes, cigarette tobacco, and tobacco leaf
• All kinds of beer, such as ale, lager, stout, and porter
• Cigarette papers and single-use filters that are consumed
• Low-alcoholic beer and nonalcoholic beer with the cigarette
02.1.9 Other Alcoholic Beverages (ND) • Cigars, pipe tobacco, chewing tobacco or snuff
Includes: • Refills for electronic cigarettes with or without nicotine
• Mixed alcoholic-based drinks, such as soda water • Tobacco that is consumed with a shisha or hookah pipes if
or mineral water-based mixed alcoholic drinks consumed at home
(alcopops) • Tobacco that is purchased in bars and restaurants, pro-
• Shandy, cola beer, radler vided that a service charge is not applied
Excludes:
02.1.9.0 Other Alcoholic Beverages (ND)
• Tobacco that is consumed with a shisha or hookah pipes
Includes:
in restaurants, cafés, shisha lounges (11.1.1)
• Mixed alcoholic-based drinks, such as soda water or min- • Other smokers’ articles; electronic cigarette devices
eral water-based mixed alcoholic drinks (alcopops) (13.2.9.1)
• Shandy, cola beer, Radler
02.3.0.1 Cigarettes (ND)
02.2 Alcohol Production Services Includes:
provided by households to produce alcohol for own final • Cigarettes
consumption by households. • Cigarettes that are purchased in bars and restaurants, pro-
vided that a service charge is not applied
02.2.1 Alcohol Production Services (S)
Services purchased for the processing of primary products 02.3.0.2 Cigars (ND)
provided by households to produce alcohol for own final Includes:
consumption by households.
• Cigars
Includes:
• Fruit/vegetable crushing and pressing services for the 02.3.0.9 Other Tobacco Products (ND)
production of alcoholic beverages Includes:
• Distilling and fermentation services • Pipe tobacco, chewing tobacco, hookah blends, snus or
• Brewing services snuff
• Aging and bottling services • Cigarette tobacco and tobacco leaf
• Cigarette papers and single-use filters that are consumed
02.2.1.0 Alcohol Production Services (S)
with the cigarette
• Refills for electronic cigarettes with or without nicotine
provided by households to produce alcohol for own final
consumption by households. • Tobacco that is consumed with a shisha or hookah pipes if
Includes: consumed at home
• Fruit/vegetable crushing and pressing services for the Excludes:
production of alcoholic beverages • Tobacco that is consumed with a shisha or hookah pipes
• Distilling and fermentation services in restaurants, cafés, shisha lounges (11.1.1)
371
APPENDIX
• Other smokers’ articles; electronic cigarette devices 03.1.1 Clothing Materials (SD)
• Clothing materials of natural fibers, of man-made fibers
02.4 Narcotics and of their mixtures
This group covers all narcotics purchased by households,
both legal and illegal. • Leather, fur, fusible webbing, wadding, and felt filling for
making wearing apparel
02.4.0 Narcotics (ND) Excludes:
Includes:
• Furnishing fabrics (05.2.1.1)
• Marijuana, opium, cocaine, and their derivatives
• Other vegetable-based narcotics, such as cola nuts, kava, 03.1.1.0 Clothing Materials (SD)
chat, betel leaves, psilocybin mushroom, and betel nuts Includes:
• Other narcotics including chemicals and man-made drugs • Clothing materials of natural fibers, of man-made fibers,
Excludes: and of their mixtures
• Prepared joints and pipes containing marijuana, hashish • Leather, fur, fusible webbing, wadding, and felt filling for
or similar purchased in coffee shops for immediate con- making wearing apparel
sumption (11.1.1.1) Excludes:
• Narcotics for medicinal purpose (06.1.1) • Furnishing fabrics (05.2.1.1)
02.4.0.0 Narcotics (ND)
03.1.2 Garments (SD)
Includes: Includes:
• Marijuana, opium, cocaine, and their derivatives • Garments for men or boys, women or girls, and infants,
• Other vegetable-based narcotics, such as cola nuts, either ready-to-wear or made-to-measure, in all materials
kava, chat, betel leaves, psilocybin mushroom, and (including leather, furs, plastics, and rubber), for every-
betel nuts day wear, for sport, or for work
• Other narcotics including chemicals and man-made • Capes, overcoats, raincoats, anoraks, parkas, blousons,
drugs jackets, trousers, waistcoats, suits, costumes, dresses,
Excludes: skirts, and so on
• Shirts, blouses, pullovers, sweaters, cardigans, shorts,
• Prepared joints and pipes containing marijuana, hashish swimsuits, tracksuits, jogging suits, sweatshirts, T-shirts,
or similar purchased in coffee shops for immediate con- leotards, and so on
sumption (11.1.1)
• Vests, underpants, socks, stockpiles, tights, petticoats,
• Narcotics for medicinal purpose (06.1.1.1) brassières, knickers, slips, girdles, corsets, body stock-
piles, and so on
03 Clothing and Footwear • Pajamas, nightshirts, nightdresses, housecoats, dressing
Division 03 covers all clothing materials, garments, articles gowns, bathrobes, and so on
and accessories, footwear, and related services including • Traditional garments
cleaning, repair, and hire of clothing and footwear. The
purchase of second-hand clothing and footwear should be Excludes:
included in the same classes as the new articles as the pur- • Tailoring services when the customer supplies the mate-
pose is the same. Unisex garments and footwear should be rial (03.1.4.2)
classified according to the gender of the person wearing
• Articles of medical hosiery, such as elasticated stockpiles
them.
(06.1.2.2)
Division 03 excludes sport—and game-specific sports—
and footwear (09.2.2.1). • Babies’ napkins (13.2.9.1)
03.1.2.1 Garments for Men or Boys (SD)

03.1 Clothing
The clothing classified in this group covers materials pur- Includes:
chased to be transformed into clothing, garments and acces- • Garments for men or boys, either ready-to-wear or made-
sories, and services related to clothing. to-measure, in all materials (including leather, furs, plas-
Made-to-measure refers to the service of providing cus- tics, and rubber), for everyday wear, for sport, or for work
tom-fitted clothing when the retailer supplies all of the mate-
• Capes, overcoats, raincoats, anoraks, parkas, jackets,
rials and is included in 03.1.2 Garments as the cost of the
trousers, waistcoats, suits, costumes, and so on
garment usually outweighs the cost of the service. Tailoring
refers to creating clothing garments where the main mate- • Shirts, pullovers, sweaters, cardigans, shorts, swimsuits,
rial is supplied by the customer and is classified in 03.1.4. tracksuits, jogging suits, sweatshirts, T-shirts, leotards,
Cleaning, repair, tailoring, and hire of clothing as the service and so on
is the higher expenditure. • Traditional garments
372
APPENDIX
• Vests, underpants, socks, and so on • Sewing threads, knitting yarns, and accessories for mak-
• Pajamas, dressing gowns, bathrobes, and so on ing clothing, such as buckles, buttons, press studs, zip fas-
teners, ribbons, laces, trimmings, and so on
Excludes:
Includes also:
• Garments for infants (0 to under 2 years) (03.1.2.3)
• Working gloves
• Tailoring services when the customer supplies the mate-
rial (03.1.4.2) Excludes:
03.1.2.2 Garments for Women or Girls (SD) • Pins, safety pins, sewing needles, knitting needles, thim-
bles; rubber gloves and other articles made of rubber; gar-
Includes:
dening gloves (05.6.1.9)
• Garments for women or girls, either ready-to-wear or • Protective headgear for sports; other protective gear for
made-to-measure, in all materials (including leather, furs, sports, such as life jackets, boxing and other sporting
plastics, and rubber), for everyday wear, for sport, or for gloves, body padding, belts, supports, and so on (09.2.2.1)
work • Paper handkerchiefs (13.1.2.0)
• Capes, overcoats, raincoats, anoraks, parkas, blousons,
• Watches, jewelry, cuff links, tiepins (13.2.1.1)
jackets, trousers, waistcoats, suits, costumes, dresses,
skirts, and so on • Walking sticks and canes, umbrellas and parasols, fans,
and keyrings (13.2.9.1)
• Shirts, blouses, pullovers, sweaters, cardigans, shorts,
swimsuits, tracksuits, jogging suits, sweatshirts, T-shirts, 03.1.3.1 Other Articles of Clothing (SD)
leotards, and so on Includes:
• Traditional garments
• Ties, handkerchiefs, scarves, gloves, mittens, muffs,
• Vests, underpants, socks, stockpiles, tights, petticoats, belts, braces, aprons, smocks, bibs, sleeve protectors,
brassières, knickers, slips, girdles, corsets, body stock- hats, caps, berets, bonnets, and so on
piles, and so on
Includes also:
• Pajamas, nightshirts, nightdresses, housecoats, dressing
gowns, bathrobes, and so on • Working gloves
Excludes: Excludes:
• Garments for infants (0 to under 2 years) (03.1.2.3) • Pins, safety pins, sewing needles, knitting needles, thim-
• Tailoring services when the customer supplies the mate- bles; rubber gloves and other articles made of rubber;
rial (03.1.4.2) and gardening gloves (05.6.1.9)
• Protective headgear for sports; other protective gear for
03.1.2.3 Garments for Infants (0 to under 2 Years) (SD) sports, such as life jackets, boxing and other sporting
Includes: gloves, body padding, belts, supports, and so on (09.2.2.1)
• Garments for infants and babies (0 to under 2 years), • Paper handkerchiefs (13.1.2.0)
either ready-to-wear or made-to-measure, in all • Watches, jewelry, cuff links, tiepins (13.2.1.1)
materials • Walking sticks and canes, umbrellas and parasols, fans,
• Raincoats, anoraks, parkas, blousons, jackets, trousers, keyrings (13.2.9.1)
waistcoats, suits, costumes, dresses, skirts, and so on 03.1.3.2 Clothing Accessories (SD)
• Vests, underpants, socks, stockpiles, tights, and so on
Includes:
• Pajamas, nightshirts, nightdresses, dressing gowns, bath-
robes, and so on • Sewing threads, knitting yarns, and accessories for mak-
ing clothing, such as buckles, buttons, press studs, zip fas-
Excludes: teners, ribbons, laces, trimmings, and so on
• Tailoring services when the customer supplies the mate- 03.1.4 Cleaning, Repair, Tailoring, and Hire of Clothing (S)
rial (03.1.4.2) The cost of materials is included only if the materials are not
separately invoiced.
03.1.2.4 School Uniforms (SD)
Includes:
Includes:
• Dry-cleaning, laundering, and dyeing of garments
• School uniforms • Darning, mending, repair, and altering of garments
• Tailoring services when the customer supplies the material
03.1.3 Other Articles of Clothing and Clothing
Accessories (SD) • Hire of garments
Includes: Excludes:
• Ties, handkerchiefs, scarves, gloves, mittens, muffs, • Materials, threads, and other accessories purchased by
belts, braces, aprons, smocks, bibs, sleeve protectors, households with the intention of undertaking the repairs
hats, caps, berets, bonnets, and so on themselves (03.1.1.0, 03.1.3.2)
373
APPENDIX
• Made-to-measure clothing (03.1.2.1, 03.1.2.2, 03.1.2.3, 03.2.1.1 Footwear for Men (SD)
03.1.2.4) Includes:
• Repair of household linen and other household textiles
• All footwear for men either ready-to-wear or
(05.2.2.0)
made-to-measure
• Dry-cleaning, laundering, dyeing and hire of household
linen, and other household textiles (05.6.2.9) Includes also:
• Gaiters and similar articles
03.1.4.1 Cleaning of Clothing (S)
• Shoelaces
Includes:
• Parts of footwear, such as heels, soles, and so on, pur-
• Dry-cleaning, laundering, and dyeing of garments chased by households with the intention of repairing foot-
Excludes: wear themselves
• Sports footwear suitable for everyday or leisure wear
• Dry-cleaning, laundering, dyeing, and hire of household
(shoes for jogging, cross-training, tennis, basketball,
linen, and other household textiles (05.6.2.9) boating, and so on)
03.1.4.2 Repair, Tailoring, and Hire of Clothing (S) Excludes:
The cost of materials is included only if the materials are not • Cleaning, repair, and hire of footwear (03.2.2.0)
separately invoiced. • Polishes, creams, and other shoe-cleaning articles (05.6.1.1)
Includes:
• Orthopedic footwear (06.1.3.3)
• Darning, mending, repair, and altering of garments • Game-specific footwear (ski boots, football boots, golfing
• Tailoring services when the customer supplies the material shoes, and other such footwear fitted with ice skates, roll-
• Hire of garments ers, spikes, studs, and so on); shin guards, cricket pads,
and other such protective apparel for sport (09.2.2.1)
Excludes:
• Materials, threads, and other accessories purchased by 03.2.1.2 Footwear for Women (SD)
households with the intention of undertaking the repairs Includes:
themselves (03.1.1.0, 03.1.3.2)
• All footwear for women either ready-to-wear or
• Made-to measure clothing (03.1.2.1, 03.1.2.2, 03.1.2.3, made-to-measure
03.1.2.4)
• Repair of household linen and other household textiles
Includes also:
(05.2.2.0) • Gaiters and similar articles
• Shoelaces
03.2 Footwear • Parts of footwear, such as heels, soles, and so on, pur-
This group covers all general footwear, split by footwear for
chased by households with the intention of repairing foot-
men, footwear for women, and footwear for infants and chil-
wear themselves
dren, and footwear-related services. Sport-specific footwear
is classified in Division 09 Recreation and Culture. • Sports footwear suitable for everyday or leisure wear
03.2.1 Shoes and Other Footwear (SD) boating, and so on)
Includes: Excludes:
• All footwear for men, women, infants, and children • Cleaning, repair, and hire of footwear (03.2.2.0)
either ready-to-wear or made-to-measure including • Polishes, creams, and other shoe-cleaning articles
sports footwear suitable for everyday or leisure wear (05.6.1.1)
• Orthopedic footwear (06.1.3.3)
boating, and so on)
• Game-specific footwear (ski boots, football boots, golfing
• Gaiters and similar articles; shoelaces; parts of footwear,
shoes, and other such footwear fitted with ice skates, roll-
such as heels, soles, and so on, purchased by households
ers, spikes, studs, and so on); shin guards, cricket pads,
with the intention of repairing footwear themselves
and other such protective apparel for sport (09.2.2.1)
Excludes:
03.2.1.3 Footwear for Infants and Children (SD)
• Polishes, creams, and other shoe-cleaning articles
• Orthopedic footwear (06.1.3.3) • All footwear for infants and children (under 13 years)
• Game-specific footwear (ski boots, football boots, either ready-to-wear or made-to-measure
golfing shoes, and other such footwear fitted with ice Includes also:
skates, rollers, spikes, studs, and so on); shin guards,
cricket pads, and other such protective apparel for • Gaiters and similar articles
sport (09.2.2.1) • Shoelaces
374
APPENDIX
• Parts of footwear, such as heels, soles, and so on, pur- 04 Housing, Water, Electricity, Gas, and
chased by households with the intention of repairing foot- Other Fuels
wear themselves
The “Housing, water, electricity, gas, and other fuels” divi-
• Sports footwear suitable for everyday or leisure wear sion comprises goods and services for the use of the house or
(shoes for jogging cross-training, tennis, basketball, boat- dwelling, its maintenance and repair, the supply of water and
ing, and so on) miscellaneous services related to the dwelling, and energy
• Baby’s booties made of fabric or sheepskin used for heating or cooling. Actual rentals and imputed rent-
Excludes: als (according to national accounts [SNA 2008]) for main or
secondary residences are classified in groups 04.1 and 04.2,
• Cleaning, repair, and hire of footwear (03.2.2.0) respectively.
• Polishes, creams, and other shoe-cleaning articles Maintenance, repair, and security of the dwelling include
(05.6.1.1) materials for repair purchased with the intention of under-
• Orthopedic footwear (06.1.3.3) taking the repair and maintenance themselves, as well as
the repair services purchased from enterprises. It should be
• Game-specific footwear (ski boots, football boots, golfing noted that only expenditures on materials and services for
shoes, and other such footwear fitted with ice skates, roll- minor repairs are covered by 04.3. Expenditures on materi-
ers, spikes, studs, and so on); shin guards, cricket pads, als and services for major maintenance and repair are not
and other such protective apparel for sport (09.2.2.1) part of individual consumption expenditures of households
and are thus out of the scope of the classification. This refers
03.2.2 Cleaning, Repair, and Hire of Footwear (S) especially to owner-occupiers; tenants will not make such
The cost of materials is included only if the materials are not expenditures on major maintenance and repair at all as they
separately invoiced. are not the owner of the dwelling.
Includes:
• Repair of footwear 04.1 Actual Rentals for Housing
Rentals normally include payment for the use of the land on
• Shoe-cleaning services
which the property stands, the dwelling occupied, the fix-
• Dyeing of shoes tures and fittings for heating, plumbing, lighting, and so on,
• Hire of footwear and, in the case of a dwelling let furnished, the furniture.
Rentals also include payment for the use of a garage to
Excludes:
provide parking in connection with the dwelling.
• Parts of footwear, such as heels, soles, and so on, pur- The garage does not have to be physically contiguous to
chased by households with the intention of undertaking the dwelling, nor does it have to be leased from the same
the repair themselves (03.2.1) landlord.
• Polishes, creams, and other shoe-cleaning articles Rentals do not include payment for the use of garages or
(05.6.1.1) parking spaces not providing parking in connection with the
• Hire and repair of game-specific footwear (ski boots,
dwelling (07.2.4). Nor do they include charges for water
football boots, golfing shoes, and other such footwear supply (04.4.1), refuse collection (04.4.2), and sewage col-
fitted with ice skates, rollers, spikes, studs, and so on) lection (04.4.3); coproprietor charges for caretaking, garden-
(09.4.4.0) ing, stairwell cleaning, heating and lighting, maintenance of
lifts and refuse disposal chutes, and so on in multi-occupied
buildings (04.4.4); charges for electricity (04.5.1) and gas
03.2.2.0 Cleaning, Repair, and Hire of Footwear (S) (04.5.2); and charges for heating and hot water supplied by
The cost of materials is included only if the materials are not district heating plants (04.5.5).
separately invoiced. Each household has a principal dwelling (sometimes
Includes: also designated as main or primary home), usually defined
• Repair of footwear with reference to time spent there, whose location defines
the country of residence and place of usual residence of
• Shoe-cleaning services this household and of all its members. All other dwellings
• Dyeing of shoes (owned or leased by the household) are considered second-
• Hire of footwear ary dwellings.
Includes:
Excludes:
• Rentals actually paid by tenants or subtenants occupying
• Parts of footwear, such as heels, soles, and so on, pur-
unfurnished or furnished premises as their main residence
chased by households with the intention of undertaking
the repair themselves (03.2.1) Includes also:
• Polishes, creams, and other shoe-cleaning articles • Payments by households occupying a room in a hotel or
(05.6.1.1) boarding house as their main residence
• Hire and repair of game-specific footwear (ski boots,
Excludes:
football boots, golfing shoes, and other such footwear
fitted with ice skates, rollers, spikes, studs, and so on) • Garage rentals to provide parking and storage in connec-
(09.4.4.0) tion with the dwelling (04.1.2.2)
375
APPENDIX
• Accommodation services of educational establishments 04.1.2.2 Garage Rentals and Other Rentals Paid by
and hostels (11.2.0.9) Tenants (S)
• Retirement homes for elderly persons (13.3.0.2) Rentals also include payment for the use of a garage to pro-
vide parking and storage in connection with the dwelling.
04.1.1 Actual Rentals Paid by Tenants for Main The garage or storage does not have to be physically con-
Residence (S) tiguous to the dwelling, nor does it have to be leased from
Includes: the same landlord.
Includes:
• Rentals actually paid by tenants or subtenants occupy-
ing unfurnished or furnished premises as their main • Garage rentals in connection with the dwelling
residence • Rentals of self-storage units
Includes also: Excludes:
• Payments by households occupying a room in a hotel or • Payment for the use of garages or parking spaces not pro-
boarding house as their main residence viding parking in connection with the dwelling (07.2.4.1)
Excludes: • (Long-term) storage of furniture and other personal
effects (07.4.9.1)
• Garage rentals to provide parking and storage in connec-
tion with the dwelling (04.1.2.2) 04.2 Imputed Rentals for Housing
• Accommodation services of educational establishments Persons who own the dwellings in which they live are treated
and hostels (11.2.0.9) as owning unincorporated enterprises that produce housing
• Retirement homes for elderly persons (13.3.0.2) services that are consumed by the household to which the
owner belongs. The housing services produced are deemed
04.1.1.0 Actual Rentals Paid by Tenants for Main to be equal in value to the rentals that would be paid on
Residence (S) the market for accommodation of the same size, quality,
Includes: and type. The imputed values of the housing services are
recorded as final consumption expenditures of the owners.
• Rentals actually paid by tenants or subtenants occupy- Imputed rentals normally include value for the use of the
ing unfurnished or furnished premises as their main land on which the property stands, the dwelling occupied,
residence and the fixtures and fittings for heating, plumbing, lighting,
Includes also: and so on.
Imputed rentals also include the use of a garage to provide
• Payments by households occupying a room in a hotel or parking in connection with the dwelling. The garage does
boarding house as their main residence not have to be physically contiguous to the dwelling.
Excludes: Imputed rentals do not include payment for the use of
garages or parking spaces not providing parking in connec-
• Garage rentals to provide parking and storage in connection with the dwelling (07.2.4). Nor do they include charges
tion with the dwelling (04.1.2.2) for water supply (04.4.1), refuse collection (04.4.2), and
• Accommodation services of educational establishments sewage collection (04.4.3); coproprietor charges for care-
and hostels (11.2.0.9) taking, gardening, stairwell cleaning, heating and lighting,
• Retirement homes for elderly persons (13.3.0.2) maintenance of lifts and refuse disposal chutes, and so on
in multi-occupied buildings (04.4.4); charges for electricity
04.1.2 Other Actual Rentals (S) (04.5.1) and gas (04.5.2); and charges for heating and hot
Includes: water supplied by district heating plants (04.5.5).
• Rentals actually paid for secondary residences 04.2.1 Imputed Rentals of Owner-Occupiers for Main
• Rentals of self-storage units Residence (S)
• Garage rentals Includes:
Excludes: • Imputed rentals of owners occupying their main residence
• Accommodation services of holiday villages and holiday 04.2.1.0 Imputed Rentals of Owner-Occupiers for Main
centers (11.2.0.2) Residence (S)
Includes:
04.1.2.1 Actual Rentals Paid by Tenants for Secondary
Residences (S) • Imputed rentals of owners occupying their main residence
Includes:
04.2.2 Other Imputed Rentals (S)
• Rentals actually paid for secondary residences Includes:
Excludes: • Imputed rentals for secondary residences
• Accommodation services of holiday villages and holiday • Imputed garage rentals in connection with the dwelling
centers (11.2.0.2) • Imputed rentals of storage units
376
APPENDIX
04.2.2.0 Other Imputed Rentals (S) • Products, materials, and fixtures used for major main-
Includes: tenance and repair (intermediate consumption) or
for extension and conversion of the dwelling (capital
• Imputed rentals for secondary residences formation)
• Imputed garage rentals in connection with the dwelling
• Imputed rentals of storage units 04.3.1.1 Materials for the Maintenance Repair of the
Dwelling (ND)
Includes:
04.3 Maintenance, Repair, and Security of the
Dwelling • Products and materials, such as paints and varnishes,
Maintenance and repair of dwellings are distinguished renderings, wallpapers, fabric wall coverings, window
by two features: first, they are activities that have to be panes, plaster, cement, putty, wallpaper pastes, and so
undertaken regularly in order to maintain the dwelling on, purchased for minor maintenance and repair of the
in good working order; second, they do not change the dwelling
dwelling’s performance, capacity, or expected service • Small plumbing items (pipes, taps, joints, and so on), sur-
life. facing materials (floorboards, ceramic tiles, and so on)
There are two types of maintenance and repair of dwell- and brushes and scrapers for paint, varnish, and wallpaper
ings: those which are minor, such as interior decoration and • Fitted carpets and linoleum
repairs to fittings, and which are commonly carried out by
• Door fittings, power sockets, wiring flex
both tenants and owners; and those which are major, such as
replastering walls or repairing roofs, and which are carried Excludes:
out by owners only.
• Hand tools (05.5.2.1)
Only expenditures which tenants and owner-occupiers
incur on materials and services for minor maintenance and • Lamp bulbs (05.5.2.2)
repair are part of individual consumption expenditure of • Brooms, scrubbing brushes, dusting brushes, and clean-
households. ing products (05.6.1.1)
Expenditures which owner-occupiers incur on materials • Products, materials, and fixtures used for major mainte-
and services for major maintenance and repair are not part of nance and repair (intermediate consumption) or for exten-
individual consumption expenditure of households. sion and conversion of the dwelling (capital formation)
Purchases of materials made by tenants or owner-occu-
piers with the intention of undertaking the maintenance or 04.3.1.2 Security Equipment (SD)
repair themselves should be shown under (04.3.1). If tenants Includes:
or owner-occupiers pay an enterprise to carry out the main-
tenance or repair, the total value of the service, including the • Small equipment for surveillance/security for the individ-
costs of the materials used, should be shown under (04.3.2) ual dwelling (smoke detector, security alarms, security/
unless the materials are separately invoiced. surveillance cameras)
• Door phone for dwelling
04.3.1 Materials for the Maintenance and Repair of the
• Fire extinguishers
Dwelling (ND)
Includes:
04.3.2 Services for the Maintenance, Repair, and
• Products and materials, such as paints and varnishes, Security of the Dwelling (S)
renderings, wallpapers, fabric wall coverings, window The cost of materials is included only if the materials are not
panes, plaster, cement, putty, wallpaper pastes, and so separately invoiced.
on, purchased for minor maintenance and repair of the Includes:
dwelling
• Services of plumbers, electricians, carpenters, glaziers,
• Small plumbing items (pipes, taps, joints, and so on), sur- painters, decorators, floor polishers, and so on engaged
facing materials (floorboards, ceramic tiles, and so on) for minor maintenance and repair of the dwelling
and brushes and scrapers for paint, varnish, and wallpaper
• Locksmith services
• Fitted carpets and linoleum
• Service of laying fitted carpets and linoleum
• Door fittings, power sockets, wiring flex
• Security services
• Small equipment for surveillance/security for the indi-
vidual dwelling Excludes:
• Door phone for dwelling • Separate purchases of materials made by households with
• Fire extinguishers the intention of undertaking the maintenance or repair
themselves (04.3.1.1)
Excludes:
• Locksmith services for cars (07.2.3.0)
• Hand tools (05.5.2.1) • Bodyguards (13.9.0.9)
• Lamp bulbs (05.5.2.2) • Services engaged for major maintenance and repair
• Brooms, scrubbing brushes, dusting brushes, and clean- (intermediate consumption) or for extension and conver-
ing products (05.6.1.1) sion of the dwelling (capital formation)
377
APPENDIX
04.3.2.0 Services for the Maintenance, Repair, and Excludes:

Security of the Dwelling (S)
• Drinking water sold in bottles or containers (01.2.5.0)
The cost of materials is included only if the materials are not
separately invoiced. 04.4.2 Refuse Collection (S)
Includes: Includes:
• Services of plumbers, electricians, carpenters, glaziers, • Refuse collection and disposal
painters, decorators, floor polishers, and so on engaged • Fees for recycling paid by households
for minor maintenance and repair of the dwelling
• Locksmith services 04.4.2.0 Refuse Collection (S)
• Service of laying fitted carpets and linoleum Includes:
• Security services • Refuse collection and disposal
Excludes: • Fees for recycling paid by households
• Separate purchases of materials made by households with
04.4.3 Sewage Collection (S)
the intention of undertaking the maintenance or repair
themselves (04.3.1.1) Sewer systems, also known as sanitary sewer systems, are
(most often) an underground carriage system (most usu-
• Locksmith services for cars (07.2.3.0) ally waterborne) specifically for transporting sewage from
• Bodyguards (13.9.0.9) houses and commercial buildings and industries, through
• Services engaged for major maintenance and repair pipes or other conduits to treatment facilities or disposal
(intermediate consumption) or for extension and conver- sites. They are part of an overall system called a sewerage
sion of the dwelling (capital formation) or sewage system.
Sewage may be treated to reduce water pollution before
04.4 Water Supply and Miscellaneous Services discharge to surface waters. Sanitary sewers serving mixed
Relating to the Dwelling urban agglomerations (including commercial and industrial
areas) carry “municipal wastewater” which comprises a mix
04.4.1 Water Supply (ND) from all sources including in some cases, surface runoff, and
Includes: stormwater.
• Water supply
“Separate” sewer systems are designed to transport sew-
age alone. In municipalities served by sewers, separate storm
Includes also: drains may convey surface runoff directly to surface waters.
• Associated expenditure, such as rental of meters, reading
“Separate” sewers are distinguished from “combined sew-
of meters, standing charges, and so on ers,” which combine sewage with stormwater runoff in the
same conduit. Sanitary sewer systems are often preferred
Excludes: because they avoid the production of large volumes of com-
bined wastewater flows. However, in certain circumstances,
they may be preferred and reduce costs.
• Hot water or steam purchased from district heating plants Basic sanitation systems are improved sanitation facili-
(04.5.5.0) ties where excreta are contained or disposed of onsite.
These are generally where fecal (and other) material is col-
04.4.1.1 Water Supply through Network Systems (ND)
lected in a pit or septic tanks or are composting toilets.
Includes: Their purpose is to hygienically separate human excreta
• All charges usually included in the bills paid by house- from human contact.
holds, including meters installations and rentals, volumet- Includes:
ric or fix charge for water consumption through mains, • Sewage collection, emptying cesspools and disposal
except steam, and hot water (on a fee and contract basis)
04.4.3.1 Sewage Collection through Sewer Systems (S)
Includes also:
Includes:
of meters, standing charges, and so on • Services paid to the sanitation or water provider or
municipality for the collection, transport, and disposal of
Excludes: sewage through sewer systems
04.4.3.2 Sewage Collection through Basic Sanitation
• Hot water or steam purchased from district heating plants
Systems (S)
(04.5.5.0)
Includes:
04.4.1.2 Water Supply through Basic Systems (ND)
• Services paid to empty and evacuate liquid waste (excreta
Includes: and wastewater) from onsite
• Services paid at a public stand post/fountain and to a • Sanitation systems (pit latrines, septic tanks, and soak
water vendor (for example, by tanker truck, cart) pits) and clean them
378
APPENDIX
• Payments for using communal/commercial collective 04.5.1.0 Electricity (ND)

toilets Includes:
04.4.4 Other Services Relating to the Dwelling N.E.C. (S) • Electricity from all sources (coal, solar, hydro, and so on)
• Coproprietor charges for caretaking, gardening, stairwell • Associated expenditure, such as rental of meters, reading
cleaning, heating and lighting, maintenance of lifts and of meters, standing charges, and so on
refuse disposal chutes, pool cleaning, and so on in multi-
• Charges for self-produced energy (in some countries,
occupied buildings
households producing more electricity than what
• Road and sidewalk cleaning and chimney sweeping they consume are charged storage costs if they feed
• Measuring background radiation and content of harmful the surplus electricity back into the electricity supply
substances in dwellings grid)
• Landscaping and cleaning of grounds surrounding the
dwelling 04.5.2 Gas (ND)
• Snow removal Includes:
Excludes: • Town gas and natural gas

• Liquefied hydrocarbons (butane, propane, and so on)
• Household services, such as window cleaning, disinfect-
ing, fumigation, and pest extermination (05.6.2.9) Includes also:
• Bodyguards (13.9.0.9) • Associated expenditure, such as rental of meters, reading
of meters, rental or purchase of storage containers, stand-
04.4.4.1 Maintenance Charges in Multi-Occupied ing charges, and so on
Buildings (S)
04.5.2.1 Natural Gas through Networks (ND)
Includes:
Includes:
• Coproprietor charges for caretaking, gardening, stairwell
cleaning, heating and lighting, maintenance of lifts and • Natural gas and town gas delivered through gas networks
refuse disposal chutes, pool cleaning, and so on, in multi- • Associated expenditure, such as rental of meters, reading
occupied buildings of meters, standing charges, and so on
Excludes: 04.5.2.2 Liquefied Hydrocarbons (ND)
• Household services, such as window cleaning, disinfect- Includes:
ing, fumigation, and pest extermination (05.6.2.9)
• Liquefied hydrocarbons (butane, propane, and so on)
• Bodyguards (13.9.0.9)
delivered in storage containers
04.4.4.9 Other Services Related to Dwelling (S) • Associated expenditure, such as rental or purchase of
Includes: storage containers, standing charges, and so on
• Road and sidewalk cleaning and chimney sweeping

Excludes:
• Measuring background radiation and content of harmful • Delivery fees of liquefied hydrocarbons (butane, propane,
substances in dwellings and so on) when charged separately (07.4.9.2)
• Landscaping and cleaning of grounds surrounding the • Camping gas in cylinder less than 50 kilograms (09.2.2.2)
dwelling
• Snow removal 04.5.3 Liquid Fuels (ND)
Includes:
04.5 Electricity, Gas, and Other Fuels • Domestic heating, lighting, and cooking fuel oils
• Biofuels for domestic use
04.5.1 Electricity (ND)
Includes: • Alcohol for fireplaces
• Electricity from all sources (coal, solar, hydro, and so on) Excludes:
Includes also: • Liquid fuels for transportation (07.2.2.1, 07.2.2.2,

07.2.2.3)
of meters, standing charges, and so on 04.5.3.0 Liquid Fuels (ND)
• Charges for self-produced energy (in some countries, Includes:
households producing more electricity than what
they consume are charged storage costs if they feed • Domestic heating, lighting, and cooking fuel oils
the surplus electricity back into the electricity supply • Biofuels for domestic use
grid) • Alcohol for fireplaces
379
APPENDIX
Excludes: Excludes:
• Liquid fuels for transportation (07.2.2.1, 07.2.2.2, • Ice cubes (01.1.8.6)
07.2.2.3)
04.5.5.0 Other Energy for Heating and Cooling (ND)
04.5.4 Solid Fuels (ND) Includes:
Includes:
• Hot water and steam purchased from district heating
• Coal, coke, briquettes, firewood, charcoal, peat, and the plants
like, biomass (wheat, nutshell, and so on) and dry animal
• Ice used for cooling and refrigeration purposes
dung
Includes also:
04.5.4.1 Coal, Coal Briquettes, and Peat (ND)
Includes: of meters, standing charges, and so on
• Coal Excludes:
• Coal briquettes
• Ice cubes (01.1.8.6)
• Peat
• Peat briquettes
05 Furnishings, Household Equipment,
04.5.4.2 Wood Fuel, Including Pellets and Briquettes and Routine Household Maintenance
(ND) Division 05 covers a wide range of products for the equipment
Includes: of the house or dwelling and the household durables, semi-
durables, and nondurables as well as some kind of household
• Fuelwood, in logs, in billets, in twigs, in faggots, or in services. Division 05 includes all kinds of furniture, including
similar forms lighting equipment, household textiles, glassware, tableware,
• Wood in chips or particles and household utensils, major and smaller electric household
• Nonagglomerated sawdust and wood waste and scrap appliances, tools and equipment for house and garden, and
• Sawdust and wood waste and scrap agglomerated in bri- goods for the routine household maintenance.
quettes, pellets, or similar forms Division 05 also includes repair, installation, and rental
services of the goods classified in Division 05.
04.5.4.3 Charcoal (ND) Domestic services by paid staff in private service, sup-
plied by enterprises or self-employed persons are also
Whether or not agglomerated, in the form of blocks, sticks
included. Furthermore, window cleaning and disinfecting
or in granules or powder, or agglomerated with tar or other
services, as well as dry-cleaning and laundering of house-
substances in briquettes, tablets, balls, and so on.
hold textiles and carpets, are classified in Division 05.
Includes:
• Wood and bamboo charcoal 05.1 Furniture, Furnishings, and Loose Carpets
• Shell or nut charcoal
05.1.1 Furniture, Furnishings, and Loose Carpets (D)
• Charcoal briquettes
Includes:
Includes also:
• Sofas, couches, tables, chairs, cupboards, chests of draw-
• Charcoal briquettes for barbecue ers and bookshelves, hanger stands, and coat stands
• Bunk bed, baby furniture, such as cradles, highchairs, and
04.5.4.9 Other Solid Fuels (ND)
playpens
Includes: • Beds, mattresses, base mattresses (tatamis), wardrobes,
• Coke and bedside tables
• Other briquettes • Kitchen tables and chairs, cupboards, and surfaces
• Other biomass n.e.c., such as waste from agricultural pro- • Furniture primarily for bathroom use
duction (for example, wheat and nutshells) and dry ani- • Furniture primarily for garden use
mal dung • Wrought iron benches and tables, arbors
04.5.5 Other Energy for Heating and Cooling (ND) • Small garden houses to store garden tools and machines
Includes: • Camping furniture
• Hot water and steam purchased from district heating • Lighting equipment, such as ceiling lights, standard lamps,
plants globe lights and bedside lamps, light, and LED strings
• Ice used for cooling and refrigeration purposes Includes also:
Includes also: • Inflatable sofas, armchairs, and beds
• Associated expenditure, such as rental of meters, reading • Pieces of furniture specifically made for the customer
of meters, standing charges, and so on • Rugs (loose) carpets
380
APPENDIX
• Pictures, sculptures, engravings, tapestries, and other art Includes also:

objects including reproductions of works of art and other
• Pieces of furniture specifically made for the customer
ornaments
• Statuettes and other decorative articles of porcelain and Excludes:
crystal glass • Decorative materials for gardens (05.1.1.4)
• Wall clocks, alarm clocks, travel clocks • Repair and hire of garden and camping furniture (05.1.2.0)
• Screens, folding partitions, nontextile blinds, mirrors, • Camping equipment (09.2.2.2)
candleholders, and candlesticks
• Decorative materials for gardens 05.1.1.3 Lighting Equipment (D)
• Leather and fur for upholstery and room decorating Includes:
Excludes: • Lighting equipment, such as ceiling lights, standard
lamps, globe lights and bedside lamps, light, and LED
• Decorative materials for gardens (05.1.1.4)
strings
• Repair and hire of household, garden, and camping furni-
ture and lighting equipment (05.1.2.0) Excludes:
• Bedding (bed linen) and sunshades (05.2.1.2, 05.2.1.9) • Repair and hire of lighting equipment (05.1.2.0)
• Safes (05.3.1.9) • Light bulbs, tubes, LEDs (05.5.2.2)
• Ornamental glass and ceramic articles (05.4.0.1) • Light strings for Christmas tree (09.2.1.3)
• Light bulbs, tubes, LEDs (05.5.2.2)
05.1.1.4 Furnishings, Loose Carpets, and Rugs (D)
• Delivery and installation of loose carpets and other furni-
ture and furnishings if charged separately (07.4.9.2) Includes:
• Light strings for Christmas tree (09.2.1.3) • Rugs (loose) carpets
• Camping equipment (09.2.2.2) • Pictures, sculptures, engravings, tapestries, and other art
• Carrycots and pushchairs/strollers; wall thermometers objects including reproductions of works of art and other
and barometers (13.2.9.1) ornaments
• Works of art and antique furniture acquired primarily as • Statuettes and other decorative articles of porcelain and
stores of value (capital formation) crystal glass
• Wall clocks, alarm clocks, and travel clocks
05.1.1.1 Household Furniture (D) • Screens, folding partitions, nontextile blinds, mirrors,
Includes: candleholders, and candlesticks
• Decorative materials for gardens
• Sofas, couches, tables, chairs, cupboards, chests of
drawers and bookshelves, hanger stands, and coat • Leather and fur for upholstery and room decorating
stands Excludes:
• Bunk bed, baby furniture, such as cradles, highchairs, and
• Fitted carpets and linoleum (04.3.1.1)
playpens
• Repair and hire of loose carpets and other furniture and
• Beds, mattresses, base-mattresses (tatamis), wardrobes,
and bedside tables furnishings (05.1.2.0)
• Kitchen tables and chairs, cupboards, and surfaces • Bedding (bed linen) and sunshades (05.2.1.2, 05.2.1.9)
• Furniture primarily for bathroom use • Safes (05.3.1.9)
• Ornamental glass and ceramic articles (05.4.0.1)
Includes also:
• Charges for the delivery of furniture (when charged sepa-
• Inflatable sofas, armchairs, and beds rately) (07.4.9.2)
• Pieces of furniture specifically made for the customer • Carrycots and pushchairs/strollers; wall thermometers
and barometers (13.2.9.1)
Excludes:
• Works of art acquired primarily as stores of value (capital
• Repair and hire of household furniture (05.1.2.0) formation)
• Antique furniture acquired primarily as stores of value
(capital formation) 05.1.2 Repair, Installation, and Hire of Furniture;
Furnishings; and Loose Carpets (S)
05.1.1.2 Garden and Camping Furniture (D) The cost of materials is included only if the materials are not
Includes: Includes:
• Furniture primarily for gardens use • Repair of furniture, furnishings, and loose carpets
• Wrought iron benches and tables, arbors • Restoration of works of art, antique furniture, and antique
• Small garden houses to store garden tools and machines floor coverings other than those acquired primarily as
• Camping furniture stores of value (capital formation)
381
APPENDIX
• Charges for the hire of furniture, furnishings, and loose • Covers for motor cars, motorcycles, and so on (07.2.1.3)
carpets • Air mattresses and sleeping bags (09.2.2.2)
• Charges for the installation of furniture (when charged
separately) 05.2.1.1 Furnishing Fabrics and Curtains (SD)
Includes:
Excludes:
• Furnishing fabrics, curtain material, curtains, double cur-
• Laying and repair of fitted carpets, linoleum, and other
tains, awnings, door curtains, and fabric blinds
such floor coverings (04.3.2.0)
• Separate purchases of materials made by households with Excludes:
the intention of undertaking the repair themselves (05.1.1) • Repair and sewing services of curtains (05.2.2.0)
• Dry-cleaning and shampooing of carpets (05.6.2.9)
05.2.1.2 Bed Linen (SD)
05.1.2.0 Repair, Installation, and Hire of Furniture;
Includes:
Furnishings; and Loose Carpets (S)
The cost of materials is included only if the materials are not • Bed linen, such as sheets, pillowcases, blankets, traveling
separately invoiced. rugs, plaids, eiderdowns, counterpanes, and mosquito nets
Includes: Excludes:
• Repair of furniture, furnishings, and loose carpets • Repair and sewing services of bed linen (05.2.2.0)
• Restoration of works of art, antique furniture, and antique • Electric blankets (05.3.2.9)
floor coverings other than those acquired primarily as • Air mattresses and sleeping bags (09.2.2.2)
stores of value (capital formation)
• Charges for the hire of furniture, furnishings, and loose 05.2.1.3 Table Linen and Bathroom Linen (SD)
carpets Includes:
• Charges for the installation of furniture (when charged
• Table linen and bathroom linen, such as tablecloths, table
separately)
napkins, towels, and face cloths
Excludes:
Excludes:
• Laying and repair of fitted carpets, linoleum, and other
• Repair and sewing services of table linen and bathroom
such floor coverings (04.3.2.0)
linen (05.2.2.0)
the intention of undertaking the repair themselves (05.1.1) 05.2.1.9 Other Household Textiles (SD)
• Dry-cleaning and shampooing of carpets (05.6.2.9) Includes:
05.2 Household Textiles • Other household textiles, such as shopping bags, laundry
bags, shoe bags, covers for clothes and furniture, flags,
05.2.1 Household Textiles (SD) sunshades, and so on
Includes:
• Bedding, such as futons, pillows, bolsters, and hammocks
• Furnishing fabrics, curtain material, curtains, double cur- • Oilcloth
tains, awnings, door curtains, and fabric blinds • Bathroom mats, rush mats, and doormats
• Bedding, such as futons, pillows, bolsters, and hammocks
Includes also:
• Bed linen, such as sheets, pillowcases, blankets, traveling
rugs, plaids, eiderdowns, counterpanes, and mosquito nets • Feather and other fillers for pillows
• Table linen and bathroom linen, such as tablecloths, table Excludes:
napkins, towels, and face cloths
• Fabric wall coverings (04.3.1.1)
• Other household textiles, such as shopping bags, laundry
bags, shoe bags, covers for clothes and furniture, flags, • Floor coverings, such as loose carpets; tapestries (05.1.1.4)
sunshades, and so on • Repair and sewing services of other household textiles
• Oilcloth; bathroom mats, rush mats, and doormats, mate- (05.2.2.0)
rial costs of made-to-measure household textiles • Covers for motor cars, motorcycles, and so on (07.2.1.3)
Includes also: 05.2.2 Repair, Hire, and Sewing Services of Household
• Feather and other fillers for pillows Textiles (S)
Excludes: separately invoiced.
• Fabric wall coverings (04.3.1.1) Includes:
• Floor coverings, such as loose carpets; tapestries (05.1.1.4) • Repair of household textiles
• Repair and sewing services of household textiles (05.2.2.0) • Sewing services of household textiles
• Electric blankets (05.3.2.9) • Charges for the hire of household textiles
382
APPENDIX
05.2.2.0 Repair, Hire, and Sewing Services of Household • Ironing machines and electric mangles
Textiles (S)
Excludes:
separately invoiced. • Repair or hire of laundry appliances (05.3.3.0)
Includes: 05.3.1.3 Heaters, Air-Conditioners (D)
• Repair of household textiles Includes:
• Sewing services of household textiles
• Air-conditioners, humidifiers, space heaters, water heat-
• Charges for the hire of household textiles ers, and ventilators
05.3 Household Appliances Excludes:
05.3.1 Major Household Appliances, Whether Electric or • Repair or hire of these appliances (05.3.3.0)
Not (D)
Includes: 05.3.1.4 Cleaning Equipment (D)
• Refrigerators and dual temperature refrigerators Includes:
• Dishwashers • Vacuum cleaners, steam-cleaning machines, carpet sham-
• Electric/gas/oil/ceramic/induction panels, hobs, spit pooing machines, and machines for scrubbing, waxing,
roasters, electric/gas/convection ovens, combined cook- and polishing floors
ers, and microwave ovens Excludes:
• Extractor hoods
• Repair or hire of cleaning equipment (05.3.3.0)
Includes also:
05.3.1.9 Other Major Household Appliances (D)
• Delivery and installation of the appliances when appli-
Includes:
cable and when not charged separately
• Separate purchases of parts/materials made by households • Other major household appliances, such as safes, sew-
with the intention of undertaking the repair themselves ing machines, knitting machines, water softeners, and
so on
• Washing machines, dryers, drum dryers, drying cabinets,
drying radiators Excludes:
• Ironing machines and electric mangles • Repair or hire of major household appliances (05.3.3.0)
• Air-conditioners, humidifiers, space heaters, water heat-
ers, and ventilators 05.3.2 Small Electric Household Appliances (SD)
Includes:
• Vacuum cleaners, steam-cleaning machines, carpet sham-
pooing machines, and machines for scrubbing, waxing, • Multifunction machine, food mixers, blenders, and blend-
and polishing floors ers with heating elements
• Other major household appliances, such as safes, sewing • Slicing machines
machines, knitting machines, water softeners, and so on • Rice cookers, slow cookers
Excludes: • Toasters
• Repair or hire of major household appliances (05.3.3.0) • Sandwich grills
• Such appliances that are built into the structure of the • Meat and fish grills
building (capital formation) • Deep fryers
• Ice cream makers
05.3.1.1 Major Kitchen Appliances (D)
• Sorbet makers
Includes:
• Yogurt makers
• Refrigerators and dual temperature refrigerators
• Hotplates
• Dishwashers
• Can openers
• Electric/gas/oil/ceramic/induction panels, hobs, spit
• Electric knives
roasters, electric/gas/convection ovens, combined cook-
ers, and microwave ovens • Coffee machines
• Extractor hoods • Tea makers
• Water boilers
Excludes:
• Kettles
• Repair or hire of major kitchen appliances (05.3.3.0)
• Coffee mills
05.3.1.2 Major Laundry Appliances (D) • Juice extractors
Includes: • Electric irons
• Washing machines, dryers, drum dryers, drying cabinets, • Fans
and drying radiators • Electric blankets
383
APPENDIX
Excludes: • Small nonelectric household articles and kitchen utensils

(05.4.0.3)
• Ironing machines (05.3.1.2)
• Household scales (05.4.0.3)
• Repair or hire of small electric household appliances
(05.3.3.0) • Personal weighing machines (13.1.2.0)
• Small nonelectric household articles and kitchen utensils; • Baby scales (06.1.2.1)
household scales (05.4.0.3) 05.3.3 Repair, Installation, and Hire of Household
• Baby scales (06.1.2.1) Appliances (S)
• Personal weighing machines (13.1.2.0) The cost of materials is included only if the materials are not
05.3.2.1 Small Electric Appliances for Cooking and Includes:
Processing of Food (SD)
• Repair of household appliances
Includes: • Charges for the hire of major household appliance
• Multifunction machine, food mixers, blenders, and blend- • Charges for the installation of household appliances (if
ers with heating elements charged separately)
• Slicing machines
Excludes:
• Rice cookers, slow cookers
• Separate purchases of materials made by households
• Toasters
with the intention of undertaking the repair themselves
• Sandwich grills (05.3.1)
• Meat and fish grills
• Deep fryers 05.3.3.0 Repair, Installation, and Hire of Household
Appliances (S)
• Ice cream makers
• Sorbet makers
• Yogurt makers Includes:
• Hotplates
• Repair of household appliances
• Can openers
• Charges for the hire of major household appliance
• Electric knives
• Charges for the installation of household appliances (if
Excludes: charged separately)
• Repair or hire of food processing appliances (05.3.3.0) Excludes:
• Separate purchases of materials made by households
05.3.2.2 Small Electric Appliances for Preparing
Beverages (SD)
with the intention of undertaking the repair themselves
(05.3.1)
Includes:
• Coffee machines 05.4 Glassware, Tableware, and Household
Utensils
• Tea makers
• Water boilers 05.4.0 Glassware, Tableware, and Household
Utensils (SD)
• Kettles
Includes:
• Coffee mills
• Sparkling water makers • Glassware, crystal ware, ceramic ware, and chinaware of
the kind used for table, kitchen, bathroom, toilet, office,
• Juice extractors and indoor decoration
Excludes: • Cutlery, flatware, and silverware
• Repair or hire of coffee machines, tea makers, and similar • Nonelectric kitchen utensils of all materials, such as
appliances (05.3.3.0) saucepans, stew pots, pressure cookers, frying pans, cof-
fee mills, sparkling water makers, purée makers, mincers,
05.3.2.9 Other Small Electric Household Appliances (SD) hotplates, household scales, and other such mechanical
Includes: devices
• Nonelectric household articles of all materials, such as
• Electric irons
containers for bread, coffee, spices, and so on, waste bins,
• Fans waste-paper baskets, laundry baskets, portable money
• Electric blankets boxes and strongboxes, towel rails, bottle racks, irons
and ironing boards, letter boxes, feeding bottles, thermos
Excludes:
flasks, and iceboxes
• Ironing machines (05.3.1.2) • Repair and hire of glassware, tableware, and household
• Repair or hire of irons (05.3.3.0) utensils
384
APPENDIX
Excludes: 05.5 Tools and Equipment for House and Garden

• Lighting equipment (05.1.1.3) 05.5.1 Motorized Tools and Equipment (D)
• Cardboard tableware (05.6.1.9) Includes:
• Personal weighing machines (13.1.2.0) • Electric drills, percussion drill, electric saws, electric sanders
• Baby scales (06.1.2.1) • Garden tractors, chain saws, lawn mowers, clipper for
• Ashtrays (13.2.9.1) lawn, hedge cutters, and cultivators
05.4.0.1 Glassware, Crystal Ware, Ceramic Ware, and • Water pumps
Chinaware (SD) • Electric screwdrivers
Includes: Excludes:
• Glassware, crystal ware, ceramic ware, and chinaware of • Repair or hire of motorized major tools and equipment
the kind used for table, kitchen, bathroom, toilet, office, (05.5.3.0)
and indoor decoration
05.5.1.0 Motorized Tools and Equipment (D)
Excludes:
Includes:
• Repair or hire of glassware, crystal ware, ceramic ware,
and chinaware (05.4.0.4) • Electric drills, percussion drill, electric saws, electric sanders
• Garden tractors, chain saws, lawn mowers, clipper for
05.4.0.2 Cutlery, Flatware, and Silverware (SD) lawn, hedge cutters, and cultivators
Includes: • Water pumps
• Cutlery, flatware, and silverware • Electric screwdrivers
Excludes: Excludes:
• Repair or hire of such cutlery, flatware, and silverware • Repair or hire of motorized major tools and equipment
(05.4.0.4) (05.5.3.0)
05.4.0.3 Nonelectric Kitchen Utensils and
Articles (SD) 05.5.2 Nonmotorized Tools and Miscellaneous
Accessories (SD)
Includes: Includes:
• Nonelectric kitchen utensils of all materials, such as • Hand tools, such as saws, hammers, screwdrivers, wrenches,
saucepans, stew pots, pressure cookers, frying pans, cof- spanners, pliers, trimming knives, rasps, and files
fee mills, sparkling water makers, purée makers, mincers,
hotplates, household scales, and other such mechanical • Garden tools, such as wheelbarrows, watering cans, hoses,
devices spades, shovels, rakes, forks, scythes, sickles, and secateurs
• Nonelectric household articles of all materials, such as • Ladders and steps
containers for bread, coffee, spices, and so on, waste bins, • Fittings for radiators and fireplaces, other metal articles
waste-paper baskets, laundry baskets, portable money for the house (curtain rails, curtain rods of wood or plas-
boxes and strongboxes, towel rails, bottle racks, irons tics, string curtain rods, carpet rods, hooks, and so on) or
and ironing boards, letter boxes, feeding bottles, thermos for the garden (chains, grids, stakes, and hoop segments
flasks, and iceboxes for fencing and bordering)
• Small electric accessories, such as switches, electric
Excludes:
bulbs, fluorescent lighting tubes, torches, flashlights,
• Lighting equipment (05.1.1.3) hand lamps, and electric batteries for general use
• Repair or hire of nonelectric kitchen utensils and articles Excludes:
(05.4.0.4)
• Cardboard tableware (05.6.1.9) • Door fittings, power sockets, switches, and wiring flex
(04.3.1.1)
• Baby scales (06.1.2.1)
• Repair or hire of miscellaneous small tool accessories
• Personal weighing machines (13.1.2.0) (05.5.3.0)
• Ashtrays (13.2.9.1) • Batteries for information and communication appliances
05.4.0.4 Repair and Hire of Glassware, Tableware, and (08.1.9.2)
Household Utensils (S) • Batteries for photographic and cinematographic equip-
The cost of materials is included only if the materials are not ment (09.1.1.2)
Includes: 05.5.2.1 Nonmotorized Tools (SD)
Includes:
• Repair and hire of glassware, crystal ware, ceramic ware
and chinaware, cutlery, flatware and silverware, and non- • Saws, hammers, screwdriver, wrenches, spanners, pliers,
electric kitchen utensils and articles trimming knives, rasps, and files
385
APPENDIX
• Power shears, wheelbarrows, watering cans, hoses, • Polishes, creams, and other shoe-cleaning articles (shoe
spades, shovels, rakes, forks, scythes, sickles, and brush)
secateurs • Insecticides, fungicides, and distilled water
• Ladders and steps • Dustpans and dust brushes, dusters
Excludes: • Cloths, floor cloths, and chamois leathers
• Repair or hire of nonmotorized small tools (05.5.3.0) • Dish brush, household sponges, scourers, and steel wool
• Filters, tablecloths and table napkins, kitchen paper, bak-
05.5.2.2 Miscellaneous Accessories (SD) ing parchment roll, kitchen film, aluminum foil, and doily
Includes: • Disposable plates, cups, and cutlery
• Fittings for radiators and fireplaces, other metal articles • Vacuum cleaner bags
for the house (curtain rails, curtain rods of wood or plas- • Candles, lamp wicks, methylated spirits, plastic bag, and
tics, string curtain rods, carpet rods, hooks, and so on) or garbage bag
for the garden (chains, grids, stakes, and hoop segments • Matches, clothes-pegs, clothes hangers, pins, safety pins,
for fencing and bordering) sewing needles, knitting needles, thimbles, nails, screws,
• Small electric accessories, such as electric bulbs, fluores- nuts and bolts, tacks, washers, glues and adhesive tapes
cent lighting tubes, torches, flashlights, hand lamps, and for household use, string, twine, and rubber gloves
electric batteries for general use • Pool cleaning chemicals and water treatment chemicals
Excludes: Excludes:
• Door fittings, power sockets, switches, and wiring flex • Brushes and scrapers for paint, varnish, and wallpaper
(04.3.1.1) (04.3.1.1)
• Repair or hire of miscellaneous small tool accessories • Products specifically for the cleaning and maintenance
(05.5.3.0) of transport equipment, such as paints, chrome cleaners,
• Batteries for information and communication appliances sealing compounds, and bodywork polishes (07.2.1.3)
(08.1.9.2) • Horticultural products for the upkeep of ornamental gar-
• Batteries for photographic and cinematographic equip- dens (09.3.1.1)
ment (09.1.1.2) • Paper handkerchiefs, toilet paper, toilet soaps, toi-
let sponges, and other products for personal hygiene
05.5.3 Repair and Hire of Motorized and Nonmotorized (13.1.2.0)
Tools and Equipment (S) • Cigarette, cigar, and pipe lighters and lighter fuel
The cost of materials is included only if the materials are not (13.2.9.1)
Includes: 05.6.1.1 Cleaning and Maintenance Products (ND)
• Repair and hire of motorized tools and equipment Includes:
• Repair and hire of nonmotorized tools and miscellaneous • Detergents, hand and machine dishwashing detergent,
accessories scouring powders, disinfectant bleaches, softeners, con-
ditioners, and stain remover
05.5.3.0 Repair and Hire of Motorized and
• General-purpose cleanser, window-cleaning products,
Nonmotorized Tools and Equipment (S)
unblocking agents, and disinfectants
The cost of materials is included only if the materials are not • Floor wax, polishes
Includes: • Polishes, creams, and other shoe-cleaning articles (shoe
brush)
• Repair and hire of motorized tools and equipment • Dustpans and dust brushes, dusters
• Repair and hire of nonmotorized small tools and miscel- • Cloths, floor cloths, and chamois leathers
laneous accessories
• Dish brush, household sponges, scourers, and steel wool
05.6 Goods and Services for Routine Household • Vacuum cleaner bags
Maintenance • Pool cleaning chemicals and water treatment chemicals
05.6.1 Nondurable Household Goods (ND) Excludes:
Includes:
• Brushes and scrapers for paint, varnish, and wallpaper
• Detergents, hand and machine dishwashing detergent, (04.3.1.1)
scouring powders, disinfectant bleaches, softeners, con- • Products specifically for the cleaning and maintenance
ditioners, and stain remover of transport equipment, such as paints, chrome cleaners,
• General-purpose cleanser, window-cleaning products, sealing compounds, and bodywork polishes (07.2.1.3)
unblocking agents, and disinfectants • Horticultural products for the upkeep of ornamental gar-
• Floor wax, polishes dens (09.3.1.1)
386
APPENDIX
05.6.1.9 Other Nondurable Household Goods (ND) • Coproprietor charges for caretaking, gardening, stair-
Includes: well cleaning, heating and lighting, maintenance of lifts
and refuse disposal chutes, and so on in multi-occupied
• Insecticides, fungicides, and distilled water buildings (04.4.4.1)
• Filters, tablecloths and table napkins, kitchen paper, bak- • Snow removal and chimney sweeping (04.4.4.9)
ing parchment roll, kitchen film, aluminum foil, and doily
• Removal and storage services (07.4.9.1)
• Disposable plates, cups, and cutlery
• Child-minding with an educational component (10.1.0.1)
• Candles, lamp wicks, methylated spirits, plastic bags, and
• Child-minding outside home (13.3.0.1)
garbage bags
• Bodyguards (13.9.0.9)
• Matches, clothes-pegs, clothes hangers, pins, safety pins,
sewing needles, knitting needles, thimbles, nails, screws,
nuts and bolts, tacks, washers, glues and adhesive tapes 05.6.2.1 Domestic Services by Paid Staff (S)
for household use, string, twine and rubber gloves, and Includes:
gardening gloves
• Domestic services supplied by paid staff employed in
Excludes: private services, such as butlers, cooks, maids, drivers,
gardeners, governesses, and au pairs or nannies
• Paper handkerchiefs, toilet paper, toilet soaps, toi-
let sponges, and other products for personal hygiene Includes also:
(13.1.2.0)
• Housemaids that iron household linen and clothes in the
• Cigarette, cigar, and pipe lighters and lighter fuel family residence
(13.2.9.1)
Excludes:
05.6.2 Domestic Services and Household Services (S) • Child-minding with an educational component (10.1.0.1)
Domestic services (05.6.2.1) are services provided by per- • Child-minding outside home (13.3.0.1)
sonnel employed by households (such as butlers, maids,
cooks, nannies, governesses, cleaners, au pairs, and garden- 05.6.2.9 Other Household Services (S)
ers) who receive a compensation or wage for their services.
Domestic services also include services such as babysit- Includes:
ting, gardening, and cleaning supplied by enterprises and • Dry-cleaning of household linen and textiles
self-employed persons. Other household services (05.6.2.9) • Carpet cleaning
are services provided by enterprises and self-employed per-
• Laundering and dyeing of household textiles
sons that are not routinely provided by staff employed by
households, such as carpet cleaning, pest extermination, and • Other services supplied by enterprises or self-employed
disinfection. persons
Includes: • Pest extermination, disinfection
• Domestic services supplied by paid staff employed in • Maintaining flat in the absence of owner
private services, such as butlers, cooks, maids, drivers, Excludes:
gardeners, governesses, and au pairs or nannies
• Dry-cleaning, laundering, and dyeing of garments
• Similar services, including babysitting and housework,
(03.1.4.1)
supplied by enterprises or self-employed persons
• Refuse collection (04.4.2.0)
• Household services, such as window cleaning, disinfect-
ing, fumigation, and pest extermination • Security services (04.3.2.0)
• Dry-cleaning, laundering, and dyeing of household linen, • Sewerage collection (04.4.3.0)
household textiles, and carpets • Coproprietor charges for caretaking, gardening, stair-
• Shampooing of carpets well cleaning, heating and lighting, maintenance of lifts
and refuse disposal chutes, and so on in multi-occupied
• Pest extermination, disinfection
buildings (04.4.4.1)
• Maintaining flat in the absence of owner
• Snow removal and chimney sweeping (04.4.4.9)
Includes also: • Removal and storage services (07.4.9.1)
• Housemaids that iron household linen and clothes in the • Bodyguards (13.9.0.9)
family residence
Excludes: 06 Health
• Dry-cleaning, laundering, and dyeing of garments Division 06 comprises four main categories, one of which
(03.1.4.1) concerns health products, while the three other concern
health services. Specifically, health services provided dur-
• Refuse collection (04.4.2.0)
ing an overnight stay (06.3); services that do not require an
• Security services (04.3.2.0) overnight stay (06.2); and diagnostic imaging services, med-
• Sewerage collection (04.4.3.0) ical laboratory services, patient emergency transportation
387
APPENDIX
and emergency rescue services (06.4). Medicines and other pharmaceutical products used to prevent, to diag-
health products (06.1) cover all products that are separately nose or to treat the diseases; vitamins and minerals; fluids
invoiced from health services except when administered required for dialysis, as well as gases used in health care,
under the direct supervision of a health care professional such as oxygen, when the patient purchases them directly)
during an overnight stay (06.3). • Herbal medicines and homeopathic products
Services and products included in Division 06 are those
that can only be used in response to a health need. These Includes also:
services and products aim at preventing the occurrence of • Service fees to dispense medicines charged by the
illnesses and diseases (for example, through vaccination); pharmacists
cure from illnesses, diseases, and injuries; relieving symp-
toms of illnesses, diseases, and injuries; reduce the severity Excludes:
of illnesses, diseases, and injuries; protect against exacer- • Nutritional supplements and fortified food products
bation and complication of illnesses, diseases, and injuries; (01.1.9.9)
and restoring the health status or maintaining/preventing the • Medicines and health products supplied directly by a
deterioration of a health condition. health care provider to outpatients (no overnight stay)
It also includes services provided by health care pro- or inpatients (overnight stay) for consumption outside a
fessional purely for aesthetic beautification purposes (for health facility or institution are to be classified as out-
example, cosmetic surgery; dental work to whiten teeth). patient services (06.2) or inpatient services (06.3) if not
Division 06 excludes food believed to be highly beneficial separately identifiable from the service
to health, especially food grown organically as well as foods
or food ingredients that have been shown to affect specific • Service fees to dispense medicines charged by a health
functions or system of the body (Division 01). practitioner during an outpatient service (06.2) or an
inpatient service (06.3) if not separately identifiable from
06.1 Medicines and Health Products the service
Includes: • Veterinary products (09.4.5.0)
• Medicines, vaccines, pharmaceutical preparations, medi- • Products for personal hygiene or personal care includ-
cal devices, assistive products, and other health-related ing beauty products and creams (13.1) (for example, aro-
products used for the prevention, diagnosis, and treat- matherapy products and slimming and beauty enhancing
ment of illnesses, diseases, and injuries, and purchased by creams)
individuals or households, either with or without a pre-
06.1.1.1 Medicines, Vaccines, and Other
scription, usually from pharmacies, health facilities, or
Pharmaceutical Preparations (ND)
medical/assistive equipment suppliers and from reliable
internet sources. They are intended for consumption or Includes:
use outside a health facility or institution. • All medicines, including branded and generic products to
Excludes: treat illnesses, diseases, and injuries
• Pharmaceutical preparations used to treat illnesses, dis-
• Nutritional supplements and fortified food products
(01.1.9.9) eases, and injuries (for example, extemporaneous oint-
ments, syrups, capsules, or other galenical substances
• Medicines and health products supplied directly by a prepared on prescription)
health care provider to outpatients (no overnight stay)
• Vaccines, hormones, oral contraceptives, and other phar-
or inpatients (overnight stay) for consumption outside a
health facility or institution are to be classified as out- maceutical products used to prevent, diagnose, or treat
patient services (06.2) or inpatient services (06.3) if not illnesses and diseases
separately identifiable from the service • Vitamins and minerals
• Services fees to dispense medicines charged by a health • Fluids required for dialysis, as well as gases used in health
practitioner during an outpatient service (06.2) or an care, such as oxygen, when the patient purchases them
inpatient service (06.3) if not separately identifiable from directly
the service Includes also:
• Veterinary products (09.4.5.0)
• Service fees to dispense medicines charged by the
• Products for personal hygiene or personal care includ- pharmacists
ing beauty products and creams (13.1) (for example, aro-
matherapy products and slimming and beauty enhancing Excludes:
creams) • Nutritional supplements and fortified food products
(01.1.9.9)
06.1.1 Medicines (ND)
• Medicines and health products supplied directly by a
Includes:
health care provider to outpatients (no overnight stay)
• Medicines, vaccines, and other pharmaceutical prepa- or inpatients (overnight stay) for consumption outside a
rations (all medicines, including branded and generic health facility or institution are to be classified as out-
products, and pharmaceutical preparations used to treat patient services (06.2) or inpatient services (06.3) if not
diseases; vaccines, hormones, oral contraceptives, and separately identifiable from the service
388
APPENDIX
• Service fees to dispense medicines charged by a health • Pregnancy tests; thermometers, glucose meters, blood
practitioner during an outpatient service (06.2) or an pressure meters, and other point of care tests, baby scales,
inpatient service (06.3) if not separately identifiable from and so on
the service
Includes also:
• Veterinary products (09.4.5.0)
• All internet purchases of medical products for personal
• Products for personal hygiene or personal care includ-
use
ing beauty products and creams (13.1) (for example, aro-
matherapy products and slimming and beauty enhancing Excludes:
creams)
• Diagnostic products for use inside a health facility or
06.1.1.2 Herbal Medicines and Homeopathic Products institution (06.3)
(ND) • Scales (13.1.2.0)
Includes:
06.1.2.2 Prevention and Protective Devices (ND)
• Herbal medicines include herbs, herbal materials, herbal Includes:
preparations, and finished herbal products, which contain
as active ingredients parts of plants, or other plant materi- • Condoms and other mechanical contraceptive devices
als, or combinations generally used in traditional medi- (does not include oral contraceptives), masks, medicinal
cine or complementary medicine stockpiles (for example, compression stockpiles), medici-
• Homeopathic products include any medicine prepared in nal gloves, insecticide-treated mosquito nets, and so on
accordance with a homeopathic manufacturing procedure
described by a pharmacopeia in official use or other offi- 06.1.2.3 Treatment Devices for Personal Use (ND)
cially recognized documents (a homeopathic medicine Includes:
may contain a number of homeopathic preparations) • Inhalers, syringes, humidifiers, nebulizers, hot bags, ice
Excludes: packs, first aid kits, bandages, and so on
• Herbal medicines and homeopathic products supplied
06.1.3 Assistive Products (D)
directly by a health care provider to outpatients (no over- Includes:
night stay) or inpatients (overnight stay) for consumption
or use outside a health facility are to be classified with • Assistive products for vision, hearing, and communica-
outpatient services (06.2) or inpatient services (06.3) if tion; mobility and daily living, such as spectacles (correc-
not separately identifiable from the service. tive eyeglasses and contact lenses), wheelchairs, hearing
aids, walking frames, and artificial legs
06.1.2 Medical Products (ND)
Includes: Excludes:
• Diagnostic equipment for self-test or over-the-counter • Sunglasses; earplugs (13.2.9.1)

sale for personal use outside a health facility or institution • Glasses for the protection against potential eye damage
• Pregnancy tests; thermometers, glucose meters, blood due to the practice of a sport (09.2.2.1)
pressure meters, and other point of care tests, baby scales,
06.1.3.1 Assistive Products for Vision (D)
and so on
• Condoms and other mechanical contraceptive devices This comprises all external products whose primary purpose
(does not include oral contraceptives), masks, medicinal is to maintain or improve an individual’s vision, to com-
stockpiles (for example, compression stockpiles), medici- pensate for an impairment/a loss of intrinsic visual, and
nal gloves, insecticide-treated mosquito nets, and so on to reduce the consequences of gradual functional visual
decline.
• Inhalers, syringes, humidifiers, nebulizers, hot bags, ice Includes:
packs, first aid kits, bandages, and so on
• Corrective eyeglasses (spectacles for low vision, short
Includes also: distance, long distance)
• All internet purchases of medical products for personal • White canes
use • Ocular prosthesis (for example, glass eyes) or contact
Excludes: lenses
• Diagnostic products for use inside a health facility or Excludes:
institution (06.3) • Dental implants and dentures are to be included in
• Scales (13.1.2.0) 06.3.1.0 when an overnight stay is required and in 06.2.2.
when not
06.1.2.1 Medical Diagnostic Products (ND)
• Walking sticks and canes for recreational purposes (hik-
Includes: ing/tracking); glasses for the protection against potential
• Diagnostic equipment for self-test or over-the-counter eye damage due to the practice of a sport (09.2.2.1)
sale for personal use outside a health facility or institution • Sunglasses (13.2.9.1)
389
APPENDIX
06.1.3.2 Assistive Products for Hearing and vision, hearing, mobility, and daily living (for example,
Communication (D) rental of medical alarms for in-home use)
This comprises all external products whose primary purpose
is to maintain or improve an individual’s hearing and commu- 06.2 Outpatient Care Services
nication, to compensate for an impairment/a loss of intrinsic This group covers all preventive, dental, curative, reha-
hearing and communication capacity, and to reduce the conse- bilitative, and long-term care services not provided during
quences of gradual functional hearing/communication decline. an overnight stay. The admission criteria are irrelevant as
Includes: well as the setting where the outpatient care occurs or
the type of provider. As such, outpatient services may be
• Digital hearing aids delivered in a hospital setting or outside a hospital setting
Includes also: including at home or even on the street by any type of
health care provider. Outpatient services include all med-
• Cleaning, adjustment, and batteries if not separately iden- icines and health products supplied directly by a health
tifiable from the product care provider for consumption or use outside a health
facility/institution if not separately identifiable from the
06.1.3.3 Assistive Products for Mobility and Daily Living (D)
service.
This comprises all assistive products to maintain or improve It also includes respite care as well as services to maintain
an individual’s mobility and daily living; to compensate for an people in their private homes that are integrated into a pack-
impairment/a loss of intrinsic mobility or the inability to perform age of medical outpatient care services.
daily activities; and to reduce the consequences of gradual func- Excludes:
tional mobility and decline in the ability to perform daily activities.
Includes: • The vaccine itself when separately invoiced from the pre-
ventive service (06.1.1.1)
• Therapeutic footwear (diabetic/neuropathic/orthopedic);
• Dental, curative, rehabilitative, and long-term care ser-
trusses and supports
vices provided overnight (06.3)
• Orthoses (brace, splint, or other artificial external device
• Control and follow-up services after symptoms appeared
serving to support the leg, spine, neck, hand)
with overnight care are to be included in 06.3
• Prostheses (leg/hand including implants); spinal belts/
• Laboratory and imaging services separately identifiable
braces including neck braces or cervical collars
from the preventive service (06.4)
• Crutches
• Nonmedical services to maintain people in their private
• Rollators/walking/standing frames homes that are not integrated into a package of medical
• Wheelchairs with or without cushions (powered/manual) care (13.3.0.2)
• Walkers; walking sticks and canes for mobility
06.2.1 Preventive Care Services (S)
• Chairs for shower/bath/toilet; handrails/grab bars Preventive services aim at avoiding illnesses and diseases,
• Incontinence products, absorbent including diapers for and detecting diseases (for example, via screening). The
the aging population main distinction between preventive services and other out-
• Pressure relief mattresses and special beds patient services is the criterion “before symptoms appear.”
• Portable ramps
Includes:
• Immunization/vaccination services
Excludes:
• Family planning and counseling
• Dental implants and dentures are to be included in
• Healthy condition monitoring services (prenatal care and
06.3.1.0 when an overnight stay is required and in 06.2.2.
postnatal care services)
when not
• General and routine check-ups other than dental
• Walking sticks and canes for recreational purposes (hik-
ing/tracking) (09.2.2.1) • Child growth and development check-ups
• Early disease detection services, before symptoms appear
06.1.4 Repair, Rental, and Maintenance of Medical and (including screening, diagnostic tests, and medical exam-
Assistive Products (S) inations directed to detect communicable and noncom-
Includes: municable diseases before symptoms appear)
• Cleaning repair, rental, and maintenance of medical diag- Includes also:
nostic products for personal use and assistive products for • Laboratory and imaging services needed to provide pre-
vision, hearing, mobility, and daily living (for example, ventive services when jointly invoiced with the time and
rental of medical alarms for in-home use) skills of the personnel
06.1.4.0 Repair, Rental, and Maintenance of Medical Excludes:
and Assistive Products (S)
• The vaccine itself when separately invoiced from the pre-
Includes: ventive service (06.1.1.1)
• Cleaning repair, rental, and maintenance of medical diag- • Dental, curative, rehabilitative, and long-term care ser-
nostic products for personal use and assistive products for vices provided overnight (06.3)
390
APPENDIX
• Control and follow-up services after symptoms appeared • Control and follow-up services after symptoms appeared
with overnight care are to be included in 06.3 with overnight care (06.3)
• Laboratory and imaging services separately identifiable • Laboratory and imaging services separately identifiable
from the preventive service (06.4) from the preventive service (06.4)
• Nonmedical services to maintain people in their private
homes that are not integrated into a package of medical 06.2.2 Outpatient Dental Services (S)
care (13.3.0.2) Includes:
• Services of dentists, dental practitioners, endodontist,
06.2.1.1 Immunization Services (S)
dental surgeons; oral and maxillofacial surgeons; oral
The expenditure involved in the consultation, for both the pathologists; orthodontists; pedodontists; periodontists;
time and skills of the personnel and the purchase of the vac- prosthodontists; oral hygienists and other dental auxilia-
cine itself when jointly invoiced with the service should be ries that do not require an overnight stay
accounted for.
Illustrative examples are immunization for: Includes also:
Polio, rabies, rubella, tetanus, varicella (chicken pox), • Services of dentists for aesthetic reasons
and yellow fever Excludes:
Influenza
Measles • All health products (pharmaceutical, medical, assistive,
Meningococcal infections, mumps, pertussis (whooping therapeutic) needed to deliver outpatient dental services
cough), and pneumococcal infections separately invoiced from the provider’s service fee (06.1)
Diphtheria, hepatitis, herpes zoster, and HPV • Services of medical analysis laboratories and imaging
centers separately identifiable from the dental services
Includes: received (06.4.1.0)
• Immunization/vaccination services for maternal and
childcare 06.2.2.1 Dental Preventive Services (S)
• Travel and tourism vaccination as well as any other com- Includes:
pulsory or voluntary immunization/vaccination service • Routine preventive dental check-ups
Excludes:
• The vaccine itself when separately invoiced from the ser- 06.2.2.9 Other Outpatient Dental Services (S)
vice (06.1.1.1) Includes:
06.2.1.9 Other Preventive Services (S) • All other dental services that do not require an overnight
stay (excluding preventive dental services)
Includes:
• All cost concerning dentures (including the fitting costs)
• Family planning and counseling (including genetic
counseling) Includes also:
• Prenatal and postnatal care services • Services of dentists for aesthetic reasons
• General routine check-up services including child growth Excludes:
and development
• All health products (pharmaceutical, medical, assistive,
• Identification of genetic abnormalities
therapeutic) needed to deliver outpatient dental services
• Screening, diagnostic tests, and medical examinations separately invoiced from the provider’s service fee (06.1)
directed to detect communicable and noncommunica- • Services of medical analysis laboratories and imaging
ble diseases (for example, malaria, tuberculosis, breast centers separately identifiable from the dental services
cancer, prostate cancer, cervical cancer, colorectal received (06.4.1.0)
cancer, diabetes, HIV/AIDS, and any other communi-
cable or noncommunicable disease) before symptoms 06.2.3 Other Outpatient Care Services (Excluding
appear Preventive and Dental) (S)
• Any other medical service provided before symptoms Include all medical services other than preventive and dental
appear that do not require an overnight stay aimed at relieving symp-
Includes also: toms of illnesses, diseases, and injuries (06.2.3.1); reduce the
severity of illnesses, diseases, and injuries (06.2.3.1); protect
• Laboratory and imaging services needed to provide pre- against exacerbation and complication of illnesses, diseases,
ventive services jointly invoiced with the time and skills and injuries (06.2.3.1); restoring health status (06.2.3.1); and
of the personnel (for example, mammogram) maintaining the level of health available or preventing the
Excludes: deterioration of a health condition (06.2.3.2).
The admission criteria, the type of provider, as well as the
• Control and follow-up services after symptoms appeared setting where these curative, rehabilitative, long-term care
without overnight care (06.2.2 or 06.2.3) outpatient care services occur are irrelevant as long as no
• Dental routine preventive check-up (06.2.2.1) overnight care is involved.
391
APPENDIX
Includes: Rehabilitation aims to achieve and maintain optimal func-

tioning. In some cases, after rehabilitation, a patient can be
• Services at hospitals or in a hospital setting without an
better than before or only avoid deterioration.
overnight stay (excluding preventive 06.2.1 and dental
Includes:
06.2.2)
• Medical day care services, that is, services delivered in a • All components of the curative care of illness or the treat-
hospital setting (sometimes in medical day care centers) ment of injury; the surgery performed; diagnostic and
to a patient who is formally admitted as long as it does not therapeutic procedures; and obstetric services as long as
involve an overnight stay it does not involve an overnight stay.
• Home-based hospital treatment (for example, dialysis • Curative and rehabilitative care regardless of the type of
06.2.3.2); medical services delivered at home (for exam- provider: (specialized) physician and other health profes-
ple, palliative care 06.2.3.2); medical and nursing services sionals (for example, nurses and midwives).
to maintain people in their private home (including the • Curative and rehabilitative services provided in any set-
elderly and people with disabilities 06.2.3.2); and nursing ting: in hospitals without an overnight stay; in individual
care delivered at home including care aimed at retarding (for example, private office) or group consulting facili-
or reducing deterioration or maintaining functionality (for ties; at home or any other setting outside the hospital
example, nasogastric feeding) or for the management of including on the street.
chronic diseases (06.2.3.2) • Physical, psychological, and speech therapy. As such,
• Services delivered in individual (for example, private it includes services of chiropractors; physiotherapists
office) or group consulting facilities (excluding preven- and physical therapists; speech therapists; audiologists;
tive services 06.2.1 and dental 06.2.2) by any type of pro- inhalation or respiratory therapists, and so on. All health
vider (doctors, specialists, nurses, and other health care products (pharmaceutical, medical, assistive, therapeutic)
practitioners) needed to deliver curative outpatient services not sepa-
• Medical services delivered on the street (for example, rately invoiced from the provider’s service fee.
injections) as well as any other outpatient service deliv- • All health products (pharmaceutical, medical, assistive,
ered outside a hospital setting therapeutic) needed to deliver curative/rehabilitative out-
patient services invoiced from the provider’s service fee
Includes also:
(doctor/specialist/nurse/other health care practitioner).
• All health products (pharmaceutical, medical, assis-
Excludes:
tive, and therapeutic), diagnostic imaging services, and
medical laboratory services needed to deliver outpatient • Medicines and health products (medical, assistive, thera-
services not separately identifiable from the provider’s peutic) separately invoiced from the provider’s service fee
service fee (doctor/specialist/nurse/other health care (doctor/specialist/nurse and other health care practitio-
practitioner) ner) (06.1)
Excludes: • Preventive services (06.2.1)
• Outpatient dental services (06.2.2)
• Medicines and health products (medical, assistive, thera-
peutic) separately invoiced from the provider’s service fee • Diagnostic imaging services and medical laboratory ser-
(doctor/specialist/nurse and other health care practitio- vices separately identifiable from the provider’s service
ner) (06.1) fee (doctor/specialist/nurse and other health care practi-
tioner) (06.4.1.0)
• Preventive services (06.2.1)
• Patient emergency transportation services and emer-
• Outpatient dental services (06.2.2) gency rescue (06.4.2.0)
• Diagnostic imaging services and medical laboratory ser-
vices separately identifiable from the provider’s service
06.2.3.2 Outpatient Long-Term Care Services (S)
fee (doctor/specialist/nurse and other health care practi-
tioner) (06.4.1.0) Outpatient long-term care services do not aim at curing an
• Patient emergency transportation services and emer- illness or rehabilitating an individual, but to prevent deterio-
gency rescue (06.4.2.0) ration of a health condition and remain at the level of health
available. Some medical activities and nursing are inher-
ent parts of this status quo. Individuals getting such type
06.2.3.1 Outpatient Curative and Rehabilitative Services of outpatient long-term care might include, for example,
(Excluding Dental Services) (S) the elderly with limited capacity to perform daily activities,
Comprises curative and rehabilitative services that do not physically disabled members of the households, and those
require an overnight stay. with chronic conditions.
Curative care comprises health care contacts during In some cases, outpatient long-term care services are part
which the principal intent is to relieve symptoms of illness of a package of services including social assistance and
or injury, to reduce the severity of an illness or injury, or to social transfer. Where possible only the medical component
protect against exacerbation and complication of an illness should go into Division 06 (the social protection part should
and injury that could threaten the life or normal function. go to Division 13). If it is not possible to distinguish both
Curative services aim at bringing the person back to their the choice of the division should be based upon the extent
original status before the cure was needed. to which the package of services predominantly includes a
392
APPENDIX
social protection component (Division 13) or medical com- • Medicines and health products (medical, assistive, ther-
ponent (Division 06). apeutic) needed to deliver inpatient services during the
Includes: overnight stay even if separately invoiced
• Services of medical day care centers; medical day care Includes also:
services for the elderly and people with disabilities
• Expenditures related to the “accommodation costs” for
• Home-based long-term care hospital treatment without an the patient (for example, cooking, cleaning, accommoda-
overnight stay (for example, dialysis) tion) and associated with the hosting of patients’ relatives
• Medical and nursing services to maintain people in their pri- if indispensable and associated with the overnight stay
vate home (including the elderly and people with disabilities) (both types of costs are to be added if separately invoiced)
• Nursing care delivered at home including care aimed at • Beauty treatments carried out in hospitals (for example, cos-
retarding or reducing deterioration or maintaining func- metic surgery for other purposes than reconstructive surgery)
tionality (for example, nasogastric feeding) or for the
Excludes:
management of chronic diseases (for example, provision
of the prescribed psychiatric medication) • Hospital day care (with or without admission but no over-
• Nonmedical services to maintain people in their private night stay) and home-based hospital treatment (06.2)
homes that are integrated into a package of care and • Services received in hospitals or hospital setting with-
jointly invoiced out an overnight stay (06.2) including services of facili-
• All health products (pharmaceutical, medical, assistive, ties, such as surgeries, clinics, and dispensaries, devoted
therapeutic), diagnostic imaging services, and medical exclusively to outpatient care (06.2)
laboratory services needed to deliver outpatient services • Emergency patient transportation and rescue services
jointly invoiced with the provider’s service fee (doctor/ (06.4.2.0)
specialist/nurse/other health care practitioner) • Services of nonmedical retirement homes for elderly per-
Excludes: sons, nonmedical institutions for disabled persons and
nonmedical rehabilitation centers providing primarily
• Medicines and health products (medical, assistive, thera- long-term social support; retirement villages without
peutic) separately invoiced from the provider’s service fee inpatient medical services (13.3.0.2)
(doctor/specialist/nurse and other health care practitio-
ner) (06.1) 06.3.1 Inpatient Curative and Rehabilitative Services (S)
• Diagnostic imaging services and medical laboratory ser- Comprises curative, dental, and rehabilitative care that
vices separately invoiced from the provider’s service fee requires an overnight stay. Inpatient curative care comprises
(doctor/specialist/nurse and other health care practitio- health care contacts that require an overnight stay during
ner) (06.4.1.0) which the principal intent is to relieve symptoms of illness or
• Patient emergency transportation services and emer- injury, to reduce the severity of an illness or injury, or to pro-
gency rescue (06.4.2.0) tect against exacerbation and complication of an illness and
• Social assistance services relate to care that enables a per- injury that could threaten the life or normal function. Inpa-
son to live independently in a house or apartment, com- tient rehabilitation services that require an overnight stay
munity activities and occupational support given on a aiming at achieving and maintaining optimal functioning.
continuing or recurrent basis to individuals, such as activi- Includes:
ties whose primary purpose is social and leisure (13.3.0.3) • Curative/rehabilitative inpatient care services comprise
treatment and care (including dental) that requires an
06.3 Inpatient Care Services overnight stay from any type of provider (for example,
The overnight stay criterion is the main distinction between a hospital, nursing care facility, or facilities classified as
outpatient care services (06.2) and inpatient care services ambulatory care providers but which perform occasional
(06.3). The type of provider is irrelevant. This may be a hos- procedures requiring overnight accommodation; services
pital, nursing care facility, or facilities classified as ambula- of alcoholism or drug addiction rehabilitation facilities
tory care providers but which perform occasional procedures [other than licensed hospitals]; services of mental health
requiring overnight accommodation. convalescent homes or hospitals and any other health
It can also include health facilities within any type of facility within any type of establishment that accommo-
establishment that accommodates patients justifying an dates patients justifying an overnight stay)
overnight stay. Tuberculosis hospitals and sanatoriums are
• All medical services needed to deliver inpatient care ser-
often organized to include accommodation along with medi-
vices during the overnight stay even if separately invoiced
cal treatment, which is the predominant purpose during a
(for example, lab tests, diagnostic imaging services)
stay in such facilities.
Includes: • Medicines and health products (medical, assistive, ther-
apeutic) needed to deliver inpatient services during the
• Inpatient care services comprises treatment and care overnight stay even if separately invoiced
(including dental) that requires an overnight stay
Includes also:
• All medical services needed to deliver inpatient care ser-
vices during the overnight stay even if separately invoiced • Expenditures related to the “accommodation costs” for the
(for example, lab tests, diagnostic imaging services) patient (for example, cooking, cleaning, accommodation)
393
APPENDIX
and associated with the hosting of patients’ relatives if Excludes:

indispensable and associated with the overnight stay
• Hospital day care (with or without admission but no over-
(both types of costs are to be added if separately invoiced)
night stay) and home-based hospital treatment (06.2)
• Beauty treatments carried out in hospitals (for example,
• Services received in hospitals or hospital setting with-
cosmetic surgery for other purposes than reconstructive
out an overnight stay (06.2) including services of facili-
surgery)
ties, such as surgeries, clinics, and dispensaries, devoted
Excludes: exclusively to outpatient care (06.2)
• Hospital day care (with or without admission but no over- • Emergency patient transportation and rescue services
night stay) and home-based hospital treatment (06.2) (06.4.2.0)
• Services received in hospitals or hospital setting with- • Services of nonmedical retirement homes for elderly per-
out an overnight stay (06.2) including services of facili- sons, nonmedical institutions for disabled persons, and
ties, such as surgeries, clinics, and dispensaries, devoted nonmedical rehabilitation centers providing primarily
exclusively to outpatient care (06.2) long-term social support; retirement villages without
inpatient medical services (13.3.0.2)
• Emergency patient transportation and rescue services
(06.4.2.0)
• Services of nonmedical retirement homes for elderly per- 06.3.2 Inpatient Long-Term Care Services (S)
sons, nonmedical institutions for disabled persons, and Comprises long-term care services that require an overnight
nonmedical rehabilitation centers providing primarily stay.
long-term social support; retirement villages without Inpatient long-term care services do not aim at curing
inpatient medical services (13.3.0.2) an illness or rehabilitating an individual, but to prevent
deterioration of a health condition and remain at the level
06.3.1.0 Inpatient Curative and Rehabilitative Services (S) of health available. Some medical and nursing activities
Comprises curative, dental, and rehabilitative care that that require an overnight stay are inherent parts of this
requires an overnight stay. Inpatient curative care comprises status quo.
health care contacts that require an overnight stay during Individuals getting such type of inpatient long-term care
which the principal intent is to relieve symptoms of illness or might include, for example, the elderly with limited capacity
injury, to reduce the severity of an illness or injury, or to pro- to perform daily activities, physically disabled members of
tect against exacerbation and complication of an illness and the households, and those with chronic conditions.
injury that could threaten the life or normal function. Inpa- Includes:
tient rehabilitation services that require an overnight stay • Services of medical convalescent homes or convalescent
aiming at achieving and maintaining optimal functioning. hospitals; services of homes for the elderly with nursing
Includes: care; inpatient care hospices; services of palliative care
• Curative/rehabilitative inpatient care services comprise
establishments for the terminally ill; services of nursing
treatment and care (including dental) that requires an homes; rest homes with nursing care; services of skilled
overnight stay from any type of provider (for example, nursing facilities; services of teaching nursing homes;
a hospital, nursing care facility, or facilities classified as services of residential mental retardation facilities; and
ambulatory care providers but which perform occasional mental health and substance abuse facilities for chronic
procedures requiring overnight accommodation; services patient (for example, those with dementia)
of alcoholism or drug addiction rehabilitation facilities • Services of medical retirement homes for the elderly and
[other than licensed hospitals]; services of mental health medical residence for people with disabilities
convalescent homes or hospitals and any other health • All medical services needed to deliver inpatient care ser-
facility within any type of establishment that accommo- vices during the overnight stay (for example, lab tests,
dates patients justifying an overnight stay) diagnostic imaging services)
• All medical services needed to deliver inpatient care ser- • Medicines and health products (medical, assistive, ther-
vices during the overnight stay even if separately invoiced apeutic) needed to deliver inpatient services during the
(for example, lab tests, diagnostic imaging services) overnight stay
• Medicines and health products (medical, assistive, ther- Includes also:
apeutic) needed to deliver inpatient services during the
overnight stay even if separately invoiced • Expenditures related to the “hotel costs” for the patient
(for example, cooking, cleaning, accommodation)
Includes also: and associated with the hosting of patients’ relatives
• Expenditures related to the “accommodation costs” for if indispensable and associated with the overnight
the patient (for example, cooking, cleaning, accommoda- stay (both types of costs are to be added if separately
tion) and associated with the hosting of patients’ relatives invoiced)
if indispensable and associated with the overnight stay Excludes:
(both types of costs are to be added if separately invoiced)
• Social protection services (13.3)
• Beauty treatments carried out in hospitals (for example,
cosmetic surgery for other purposes than reconstructive • Nonmedical retirement homes for elderly persons and
surgery) nonmedical residences for disabled persons (13.3.0.2)
394
APPENDIX
06.3.2.0 Inpatient Long-Term Care Services (S) imaging technology, such as X-rays and radiation for the
Comprises long-term care services that require an overnight diagnosis and monitoring of patients
stay. Excludes:
Inpatient long-term care services do not aim at curing an
illness or rehabilitating an individual, but to prevent deterio- • Diagnostic imaging services and medical laboratory ser-
ration of a health condition and remain at the level of health vices not separately identifiable from outpatient preven-
available. Some medical and nursing activities that require tive services (06.2.1)
an overnight stay are inherent parts of this status quo. • Dental services (06.2.2)
Individuals getting such type of inpatient long-term care
might include, for example, the elderly with limited capacity 06.4.1.0 Diagnostic Imaging Services and Medical
to perform daily activities, physically disabled members of Laboratory Services (S)
the households, and those with chronic conditions. Includes:
Includes:
• Services of medical analysis laboratories (for example,
• Services of medical convalescent homes or convalescent urine/blood tests)
hospitals; services of homes for the elderly with nursing • Diagnostic imaging services including all diagnostic
care; inpatient care hospices; services of palliative care imaging methods (that is computerized tomography, mag-
establishments for the terminally ill; services of nursing netic resonance imaging, sonography); imaging diagnosis
homes; rest homes with nursing care; services of skilled comprises a variety of services that employ imaging tech-
nursing facilities; services of teaching nursing homes; nology, such as X-rays and radiation for the diagnosis and
services of residential mental retardation facilities; and monitoring of patients
mental health and substance abuse facilities for chronic
patient (for example, those with dementia) Excludes:
• Services of medical retirement homes for the elderly and • Diagnostic imaging services and medical laboratory ser-
medical residence for people with disabilities vices not separately identifiable from outpatient preven-
• All medical services needed to deliver inpatient care ser- tive services (06.2.1)
vices during the overnight stay (for example, lab tests, • Dental services (06.2.2)
diagnostic imaging services)
• Medicines and health products (medical, assistive, ther- 06.4.2 Patient Emergency Transportation Services and
apeutic) needed to deliver inpatient services during the Emergency Rescue (S)
overnight stay Includes:
Includes also: • Ambulance services for individuals with or without emer-
gency rescue
• Expenditures related to the “hotel costs” for the patient
• Individual’s transportation by airplane and vehicles for
(for example, cooking, cleaning, and accommodation)
medical emergency reasons whether or not they have
and associated with the hosting of patients’ relatives if
been specially adjusted for a medical purpose
indispensable and associated with the overnight stay
(both types of costs are to be added if separately invoiced) Includes also:
Excludes: • Memberships for emergency transport services
• Social protection services (13.3) 06.4.2.0 Patient Emergency Transportation Services
• Nonmedical retirement homes for elderly persons and and Emergency Rescue (S)
nonmedical residences for disabled persons (13.3.0.2) Includes:
06.4 Other Health Services • Ambulance services for individuals with or without emer-
Includes: gency rescue
• Individual’s transportation by airplane and vehicles for
• Diagnostic imaging services and medical laboratory
medical emergency reasons whether or not they have
services
been specially adjusted for a medical purpose
• Patient emergency transportation services and emergency
rescue Includes also:
• Memberships for emergency transport services
06.4.1 Diagnostic Imaging Services and Medical
Laboratory Services (S)
Includes: 07 Transport
Division 07 distinguishes four main categories of goods and
• Services of medical analysis laboratories (for example, services for the purpose of transportation: purchase of vehi-
urine/blood tests) cles, goods and services for the operation of the personal
• Diagnostic imaging services including all diagnostic transport equipment, the transport services provided by the
imaging methods (that is computerized tomography, market, and the transport services of goods.
magnetic resonance imaging, and sonography); imag- The purchase of vehicles covers motor cars, motorcycles,
ing diagnosis comprises a variety of services that employ bicycles, and animal-drawn vehicles. Due to the high share
395
APPENDIX
of second-hand motor vehicles in private consumption 07.1.1.2 Second-Hand Motor Cars (D)
expenditures, separate subclasses for new and for second- Includes:
hand motor cars are distinguished.
The second group concerning the operation of personal • Used or second-hand motor cars, passenger vans, station
transport equipment cover parts and accessories for personal wagons, estate cars, and the like with either two-wheel
transport equipment, fuels and lubricants, and the repair and drive or four-wheel drive
maintenance of personal transport equipment. This group Includes also:
also includes expenditures for parking spaces in garages or
in public places, expenditures for tolls, and expenditures to • Second-hand racing motor vehicles and vehicles for shows
acquire a driving certificate. Excludes:
Passenger transport services are structured by the
mode of transport. However, due to the fact that trans- • Invalid carriages (06.1.3.3)
port tickets may cover more than one mode of transport, • Camper vans (09.1.2.1)
a class for the combined passenger transport was cre- • Golf carts (09.1.2.9)
ated (07.3.6).
The fourth group covers postal and courier services, 07.1.2 Motorcycles (D)
removal and storage services, and the delivery services of Includes:
any kind of goods when charged separately. • Motorcycles of all types, motor scooters, and motorized
Division 07 does not cover purchases of recreational bicycles with combustion engines
vehicles, such as camper vans, caravans, trailers, airplanes,
and boats (09.1.2). Includes also:
• Sidecars
07.1 Purchase of Vehicles • Used or second-hand motorcycles
Includes:
07.1.2.0 Motorcycles (D)
• Purchase of vehicles used for transport Includes:
Excludes: • Motorcycles of all types, scooters, and motorized bicycles
• Purchases of recreational vehicles, such as camper with combustion engines
vans, caravans, trailers, airplanes, and boats (09.1.2.1, Includes also:
09.1.2.2, 09.1.2.3)
• Sidecars
• Used or second-hand motorcycles
07.1.1 Motor Cars (D)
Includes: 07.1.3 Bicycles (D)
• Motor cars, passenger vans, station wagons, estate cars, Includes:
and the like with either two-wheel drive or four-wheel • Bicycles and tricycles of all types
drive
• Rickshaws
Includes also: • E-bikes, pedelecs
• Racing motor vehicles and vehicles for shows Excludes:
Excludes: • Motorized bicycles with combustion engines (07.1.2.0)
• Invalid carriages (06.1.3.3) • Toy bicycles and tricycles (09.2.1.2)
• Camper vans (09.1.2.1)
07.1.3.0 Bicycles (D)
• Golf carts (09.1.2.9)
Includes:
07.1.1.1 New Motor Cars (D) • Bicycles and tricycles of all types
Includes: • Rickshaws
• New motor cars, passenger vans, station wagons, estate • E-bikes, pedelecs
cars, and the like with either two-wheel drive or four-wheel Excludes:
drive
• Motorized bicycles with combustion engines (07.1.2.0)
Includes also: • Toy bicycles and tricycles (09.2.1.2)
• Racing motor vehicles and vehicles for shows
07.1.4 Animal-Drawn Vehicles (D)
Excludes: Includes:
• Invalid carriages (06.1.3.3) • Animal-drawn vehicles
• Camper vans (09.1.2.1) • Animals required to draw the vehicles and related equip-
• Golf carts (09.1.2.9) ment (yokes, collars, harnesses, bridles, reins, and so on)
396
APPENDIX
Excludes: 07.2.1.2 Parts for Personal Transport Equipment (SD)

• Horses and ponies, horse- or pony-drawn vehicles, and Includes:
related equipment purchased for recreational purposes • Spark plugs, batteries, shock absorbers, filters, pumps,
(09.1.2.4) and other parts for personal transport equipment
• Rims
07.1.4.0 Animal-Drawn Vehicles (D)
Includes: Excludes:
• Animal-drawn vehicles • Charges for the fitting of parts and accessories and
for the painting, washing, and polishing of bodywork
• Animals required to draw the vehicles and related
(07.2.3.0)
equipment (yokes, collars, harnesses, bridles, reins, and
so on)
07.2.1.3 Accessories for Personal Transport
Excludes: Equipment (SD)
• Horses and ponies, horse- or pony-drawn vehicles, and Includes:
related equipment purchased for recreational purposes • GPS (satellite positioning) equipment for personal
(09.1.2.4) transport
07.2 Operation of Personal Transport Equipment • Accessories for personal transport equipment but bought
Purchases of parts, accessories, or lubricants made by separately
households with the intention of undertaking the mainte- • Fire extinguishers for transport equipment
nance; repair or intervention themselves should be shown • Products specifically for the cleaning and maintenance
under (07.2.1) or (07.2.2). If households pay an enterprise to of transport equipment, such as paints, chrome cleaners,
carry out the maintenance, repair, or fitting, the total value of sealing compounds, and bodywork polishes; covers for
the service, including the costs of the materials used, should motor cars, motorcycles, and so on
be shown under (07.2.3). • Hubcaps, if bought separately
• Crash helmets for motorcycles and bicycles
07.2.1 Parts and Accessories for Personal Transport
Equipment (SD) • Baby and child seats for cars, motorcycles, and bicycles
Includes: • Dash cameras
• Tires (new, used, or retreaded), inner tubes, spark plugs, Includes also:
batteries, shock absorbers, filters, pumps, and other
• Crash helmets with cameras incorporated
replacement parts or accessories for personal transport
equipment, crash helmets for motorcycles and bicycles Excludes:
• Hubcaps, if bought separately • Nonspecific products for cleaning and maintenance, such
• Baby and children seats for cars as distilled water, household sponges, chamois leathers,
• Fire extinguishers for transport equipment; products spe- detergents, and so on (05.6.1.1)
cifically for the cleaning and maintenance of transport • Car radios (08.1.4.0)
equipment, such as paints, chrome cleaners, sealing com-
pounds, and bodywork polishes; covers for motor cars, 07.2.2 Fuels and Lubricants for Personal Transport
motorcycles, and so on Equipment (ND)
• Dash cameras Includes:
Includes also: • Petrol and other fuels, such as diesel, liquid petroleum
gas, alcohol, and two-stroke mixtures
• Crash helmets with cameras incorporated
• Lubricants, brake and transmission fluids, coolants, and
Excludes: additives
• Nonspecific products for cleaning and maintenance, such • Electricity as fuel for cars when separately priced from
as distilled water, household sponges, chamois leathers, other electricity
detergents, and so on (05.6.1.1) • Hydrogen
• Charges for the fitting of parts and accessories and • Fuel for major tools and equipment covered under 05.5.1
for the painting, washing, and polishing of bodywork and recreational vehicles covered under 09.1.2
(07.2.3.0)
Excludes:
• Car radios (08.1.4.0)
• Charges for service of oil changes and greasing (07.2.3.0)
07.2.1.1 Tires (SD)
Includes: 07.2.2.1 Diesel (ND)
Includes:
• New, used, or retreaded, including inner tubes for car,
bicycles, motorcycles, and so on • Diesel
397
APPENDIX
07.2.2.2 Petrol (ND) Excludes:

Includes: • Separate purchases of parts, accessories, or lubri-
• Petrol/gasoline cants made by households with the intention of under-
taking the maintenance or repair themselves (07.2.1,
Includes also: 07.2.2.4)
• Petrol blends, such as petrol with 10 percent ethanol • Roadworthiness tests (07.2.4.3)
Excludes:
07.2.4 Other Services in Respect of Personal Transport
• Two-stroke mixtures (07.2.2.3)
Equipment (S)
07.2.2.3 Other Fuels for Personal Transport Includes:
Equipment (ND) • Rental of garages or parking spaces not providing parking
Includes: in connection with the dwelling
• Rental of municipal street parking
• Liquid petroleum gas, natural gas (compressed natural
gas, liquefied natural gas), alcohol, biofuels (ethanol, • Parking meters
methanol), methane, and two-stroke mixtures • Charges for parking places in parking garages, such as
• Electricity as fuel for cars when separately priced from those in shopping centers hired for a few hours or less
other electricity (including valet service)
• Hydrogen Includes also:
07.2.2.4 Lubricants (ND) • Parking permits for designated areas (for example, resi-
dents parking permits)
Includes:
• Tolls (bridges, tunnels, shuttle ferries, and motorways)
• Lubricants, brake and transmission fluids, coolants, and and parking meters
additives • Charges for hire or purchase of electronic tags and toll
Excludes: devices
• Charges for oil change and greasing service (07.2.3.0) • Driving lessons, driving tests, and driving licenses
• Roadworthiness tests
07.2.3 Maintenance and Repair of Personal Transport • Hire of personal transport equipment without drivers
Equipment (S)
The cost of materials is included only if the materials are not Excludes:
separately invoiced. • Rental of a garage or parking space to provide parking in
Includes: connection with the dwelling (04.1.2.2)
• Services purchased for the maintenance and repair of • Hire of a car with driver; payments for private arrange-
personal transport equipment, such as fitting of parts and ments of sharing a means of transport (carpooling,
accessories, wheel balancing, breakdown services, oil dynamic ridesharing) (07.3.2.2)
changes, greasing, and washing • Service charges for insurance in respect of personal
• Installation of car cameras transport equipment (12.1.4.1)
• Locksmith services for cars
Excludes: 07.2.4.1 Services for Parking (S)
• Separate purchases of parts, accessories, or lubricants Includes:
made by households with the intention of undertaking the • Rental of garages or parking spaces not providing parking
maintenance or repair themselves (07.2.1, 07.2.2.4) in connection with the dwelling
• Roadworthiness tests (07.2.4.3) • Rental of municipal street parking
07.2.3.0 Maintenance and Repair of Personal Transport • Parking meters
Equipment (S) • Charges for parking places in parking garages, such as
The cost of materials is included only if the materials are not those in shopping centers hired for a few hours or less
separately invoiced. (including valet service)
• Services purchased for the maintenance and repair of • Parking permits for designated areas (for example, resi-
personal transport equipment, such as fitting of parts and dents parking permits)
accessories, wheel balancing, breakdown services, oil
changes, greasing, and washing Excludes:
• Installation of car cameras • Rental of a garage or parking space to provide parking in
• Locksmith services for cars connection with the dwelling (04.1.2.2)
398
APPENDIX
07.2.4.2 Toll Facilities (S) Excludes:

Includes: • Funicular transport (07.3.6.0)
• Tolls (bridges, tunnels, shuttle ferries, and motorways)
• Charges for hire or purchase of electronic tags and toll devices 07.3.1.2 Passenger Transport by Rapid Transit and Tram (S)
Includes:
07.2.4.3 Driving Lessons, Tests, Licenses, and Road
Worthiness Tests (S) • Transport of individuals and groups of persons and
Includes: accompanied luggage by rapid transit (electric railways
that operate on an exclusive right-of-way, which can-
• Driving lessons, tests, licenses not be accessed by pedestrians or other vehicles of any
• Roadworthiness tests and pollution control tests sort, and which is often grade separated in tunnels or on
elevated railways), light-rail, underground, rubber-tired
Includes also: metros, and people movers
• Ownership transfer charges for vehicles • Transport of individuals and groups of persons and
• Driving lessons, test, licenses for recreation transport accompanied luggage by tram
vehicles (mobile homes, boats, and planes)
07.3.2 Passenger Transport by Road (S)
07.2.4.4 Hire of Personal Transport Equipment Without Includes:
Driver (S)
• Transport of individuals and groups of persons and
Includes:
accompanied luggage by bus, trolleybus, and coach
• Hire of personal transport equipment without driver (for • Transport of individuals and groups of persons and
example, car hire) accompanied luggage by share taxis (mode of transport
• Fees for the participation in “car-sharing” or “bike-sharing” which falls between both taxicabs and buses; these vehi-
programs cles for hire are typically smaller than buses and usually
take passengers on a fixed or semifixed route without
Excludes:
timetables; they may stop anywhere to pick up or drop off
• Payments for private arrangements of sharing a means of their passengers; often found in developing countries, the
transport (carpooling, dynamic ridesharing) (07.3.2.2) vehicles used as share taxis range from four-seat cars to
• Service charges for insurance in respect of personal minibuses; and they are often owner-operated)
transport equipment (12.1.4.1) • Transport of individuals and groups of persons and accom-
panied luggage by taxi and hired vehicle with driver
07.3 Passenger Transport Services • Transport of individuals and groups of persons and
Purchases of transport services are generally classified by accompanied luggage through private arrangements, such
mode of transport. When a ticket covers two or more modes as carpooling and ridesharing
of transport—for example, intraurban bus and underground
or interurban train and ferry—and the expenditure cannot be • Local school bus charter services, with driver
apportioned between them, then such purchases should be • Scheduled interurban or intercity school bus charter
classified in 07.3.5.0. services
Costs of meals, snacks, drinks, refreshments, or accom- • Transportation services of pupils by school bus between
modation services have to be included if covered by the fare their homes and school and between schools, including in
and not separately priced. If separately priced, these costs rural areas
have to be classified in Division 11. • Transport of individuals and groups of persons and
School transport services are included, but ambulance accompanied luggage by animal-drawn vehicles with
services are excluded (06.4.2.0). driver
07.3.1 Passenger Transport by Railway (S) Includes also:
Includes:
• Tuk-tuks, rickshaws, and motorcycles as taxis
accompanied luggage by train, high-speed trains, mag- 07.3.2.1 Passenger Transport by Bus and Coach (S)
levs, light-rail, tram, and underground Includes:
• Accompanied transport of private vehicles • Transport of individuals and groups of persons and
Excludes: accompanied luggage by bus, trolleybus, and coach
• Funicular transport (07.3.6.0) • Transport of individuals and groups of persons and
accompanied luggage by share taxis (mode of transport
07.3.1.1 Passenger Transport by Train (S) which falls between both taxicabs and buses; these vehi-
cles for hire are typically smaller than buses and usually
Includes:
take passengers on a fixed or semifixed route without
• Transport of individuals and groups of persons and timetables; they may stop anywhere to pick up or drop off
accompanied luggage by train, high-speed trains, maglevs their passengers; often found in developing countries, the
399
APPENDIX
vehicles used as share taxis range from four-seat cars to • Accompanied transport of private vehicles
minibuses; and they are often owner-operated) • Water taxis
07.3.2.2 Passenger Transport by Taxi and Hired Car 07.3.4.0 Passenger Transport by Sea and Inland
with Driver (S) Waterway (S)
Includes: Includes:
• Transport of individuals and groups of persons and accom- • Transport of individuals and groups of persons and
panied luggage by taxi and hired vehicle with driver accompanied luggage by ship, boat, ferry, hovercraft, and
• Transport of individuals and groups of persons and hydrofoil
accompanied luggage through private arrangements, such • Accompanied transport of private vehicles
as carpooling and ridesharing • Water taxis
07.3.2.3 Passenger Transport for Students to and from
School (S) 07.3.5 Combined Passenger Transport (S)
Includes:
Includes:
• Local school bus charter services, with driver
accompanied luggage by two or more modes of transport
• Scheduled interurban or intercity school bus charter when the expenditure cannot be apportioned between
services them
• Transportation services of pupils by school bus between • Multiple modes of transport (for example, bus, tram, sub-
their homes and school and between schools, including in way, and ferry) available on the one ticket
rural areas
Includes also:
07.3.2.9 Other Passenger Transport by Road (S) • Accompanied transport of private vehicles (for exam-
Includes: ple, train ticket that includes transport of private
• Transport of individuals and groups of persons and accom- vehicles)
panied luggage by animal-drawn vehicles with driver Excludes:
Includes also: • Package holidays (09.8.0.0)
• Tuk-tuks, rickshaws, and motorcycles as taxis
07.3.5.0 Combined Passenger Transport (S)
Excludes: Includes:
• Bus with driver (07.3.2.1) • Transport of individuals and groups of persons and
accompanied luggage by two or more modes of transport
07.3.3 Passenger Transport by Air (S) when the expenditure cannot be apportioned between
Includes: them
• Transport of individuals and groups of persons and lug- • Multiple modes of transport (for example, bus, tram, sub-
gage by airplane and helicopter. It includes also passenger way, and ferry) available on the one ticket
drones and multicopters Includes also:
07.3.3.1 Passenger Transport by Air, Domestic (S) • Accompanied transport of private vehicles (for exam-
Includes: ple, train ticket that includes transport of private
vehicles)
• Domestic scheduled and chartered air passenger transpor-
tation services Excludes:
• Domestic transport of individuals and groups of persons • Package holidays (09.8.0.0)
and accompanied luggage by airplane and helicopter
• Domestic air passenger transport by drones and multicopters 07.3.6 Other Purchased Transport Services (S)
Includes:
07.3.3.2 Passenger Transport by Air, International (S)
• Funicular, elevator, cable car, and chairlift transport
Includes:
• Services of porters and left-luggage (services to store
• Transportation of passengers by air on an international travelers’ luggage for limited amount of time) and lug-
scheduled and chartered flight gage-forwarding offices
• Travel agents’ commissions, if separately priced
07.3.4 Passenger Transport by Sea and Inland Waterway (S)
• Transporter bridges, elevators, incline elevators
Includes:
Excludes:
accompanied luggage by ship, boat, ferry, hovercraft, and • Cable car and chairlift transport at ski resorts and holi-
hydrofoil day centers (09.4.6.2)
400
APPENDIX
07.3.6.0 Other Purchased Transport Services (S) 07.4.1.2 Courier and Parcel Delivery Services (S)
Includes: Includes:
• Funicular, elevator, cable car, and chairlift transport • Parcel delivery services (incoming and outgoing—par-
• Services of porters and left-luggage (services to store cels sent and delivery of parcels to a home)
travelers’ luggage for limited amount of time) and lug- • Courier services for small parcels
gage-forwarding offices • Parcel delivery services of goods purchased online
• Travel agents’ commissions, if separately priced
Excludes:
• Transporter bridges, elevators, incline elevators
• Services of delivery of goods, such as furniture, super-
Excludes: market shopping when charged separately (07.4.9.2)
• Cable car and chairlift transport at ski resorts and holi-
day centers (09.4.6.2) 07.4.9 Other Transport of Goods (S)
Includes:
07.4 Transport Services of Goods • Removal services of furniture
• Storage services of furniture
07.4.1 Postal and Courier Services (S)
Includes: • Services of delivery of goods, such as furniture, super-
market shopping when charged separately
• New postage stamps and other prefranked postal mat-
• Local delivery of purchased items, such as take-out meals,
ter (for example, prefranked postcards, envelopes, and
groceries, and prescription drugs
so on)
• Delivery services generally made immediately after the
• Courier services for letters
item is purchased or scheduled within a short time
• Parcels delivery services (incoming and outcoming—
parcels sent and delivery of parcels at home) Excludes:
• Courier services for small parcels • Self-storage units (04.1.2.2)
• Parcel delivery services of goods purchased online • Installation of goods (furniture, electric devices, and so
• Removal services of furniture on) when separately charged (05.1.2.0 or 05.3.3.0)
• Storage services of furniture • Services of porters and left-luggage and luggage-
• Services of delivery of goods, such as furniture, super- forwarding offices (07.3.6.0)
market shopping when charged separately
07.4.9.1 Removal and Storage Services (S)
• Local delivery of purchased items, such as take-out meals,
groceries, and prescription drugs Includes:
• Delivery services generally made immediately after the • Removal services of furniture
item is purchased or scheduled within a short time • Storage services of furniture
Excludes: Excludes:
• Self-storage units (04.1.2.2) • Self-storage units (04.1.2.2)
• Installation of goods (furniture, electric devices, and so • Services of porters and left-luggage and luggage-for-
on) when separately charged (05.1.2.0 or 05.3.3.0) warding offices (07.3.6.0)
• Services of porters and left-luggage and luggage-forwarding
07.4.9.2 Delivery of Goods (S)
offices (07.3.6.0)
• Services of delivery of goods, such as furniture, Includes:
supermarket shopping when charged separately • Services of delivery of goods, such as furniture, super-
(07.4.9.2) market shopping when charged separately
• Not prefranked postcards, envelopes, and other not pre- • Local delivery of purchased items, such as take-out meals,
franked postal matter (09.7.3.0) groceries, and prescription drugs
• Delivery services generally made immediately after the
07.4.1.1 Letter Handling Services (S) item is purchased or scheduled within a short time
Includes:
Excludes:
• New postage stamps and other prefranked postal mat-
• Installation of goods (furniture, electric devices, and so
ter (for example, prefranked postcards, envelopes, and
on) when separately charged (05.1.2.0 or 05.3.3.0)
so on)
• Courier services for letters
08 Information and Communication
Excludes:
Division 08 covers three main groups of goods and services:
• Not prefranked postcards, envelopes, and other not pre- information and communication equipment, including equip-
franked postal matter (09.7.3.0) ment for the reception, recording, and reproduction of sound
401
APPENDIX
and vision; software; and information and communication • Video game software; video game computers and con-
services. Information and communication services include soles (09.2.1.1)
telephone other communication services, internet access ser- • Removable media containing books, dictionaries, ency-
vices, TV and radio licenses, and fee and subscription ser- clopedias, foreign language trainers, multimedia presen-
vices, including streaming services of film and music. tations, and so on in the form of software (09.7.1)
Division 08 also includes repair, maintenance, and hire of
information and communication equipment. 08.1.3.1 Computers, Laptops, and Tablets (D)
Includes:
08.1 Information and Communication Equipment
• Desktop computers and laptops
08.1.1 Fixed Telephone Equipment (D)
• Tablets
Includes:
Includes also:
• Telephones, radiotelephones, telefax machines, tele-
phone-answering machines, and telephone loudspeakers • Telefax and telephone-answering facilities provided by
personal computers
Excludes:
Excludes:
• Telefax and telephone-answering facilities provided by
personal computers (08.1.3.1) • Computer software packages, such as operating systems,
applications, languages, and so on (08.2.0.0)
08.1.1.0 Fixed Telephone Equipment (D) • Video game software; video game computers and con-
Includes: soles (09.2.1.1)
• Telephones, radiotelephones, telefax machines, tele- • Removable media containing books, dictionaries, ency-
phone-answering machines, and telephone loudspeakers clopedias, foreign language trainers, multimedia presen-
tations, and so on in the form of software (09.7.1)
Excludes:
• Telefax and telephone-answering facilities provided by 08.1.3.2 Peripheral Equipment and Its Consumable
personal computers (08.1.3.1) Components (D)
Includes:
08.1.2 Mobile Telephone Equipment (D)
• Printers, scanners, monitors, projectors, virtual reality
Includes:
head mounts, modems, routers, network switches, and the
• Mobile telephone handsets, including devices with sev- like, keyboards, mice, digitizers
eral functions • Typewriters and word processors (device)
• Smartphones • Toner and ink cartridges, laser printer drums, and type-
writer ribbons
08.1.2.0 Mobile Telephone Equipment (D)
• Calculators, including pocket calculators
Includes:
• Web cameras
• Mobile telephone handsets, including devices with sev-
eral functions 08.1.4 Equipment for the Reception, Recording, and
• Smartphones Reproduction of Sound and Vision (D)
Includes:
08.1.3 Information Processing Equipment (D) • Television sets, video cassette players and recorders, digi-
Includes: tal video recorders, DVD players, Blu-ray players, Ultra
• Personal computers, printers, scanners, monitors, projec- HD Blu-ray players, streaming boxes, television aerials of
tors, virtual reality head mounts, modems, routers, net- all types
work switches, and the like, keyboards, mice, digitizers • Radio receivers (radio sets, digital radio sets, internet radio
• Tablets sets, satellite radio sets, car radios, radio clocks, two-way
• Calculators, including pocket calculators radios, amateur radio receivers, and transmitters)
• Typewriters and word processors (device) • Portable and nonportable CD players
• Toner and ink cartridges, laser printer drums, and type- • Portable and nonportable sound players
writer ribbons • Stereo equipment and CD radio cassette recorder
Includes also: • Turntables, tuners, amplifiers, cassette decks, microphones
and speakers, disc jockey equipment, and karaoke systems
• Telefax and telephone-answering facilities provided by • Car stereos, video systems for cars
personal computers
• Set-top boxes, satellite receivers, IPTV receivers, and
Excludes: television converter boxes
• Computer software packages, such as operating systems, • Digital media players
applications, languages, and so on (08.2.0.0) • Headphone, earplugs, and wireless/Bluetooth headsets
402
APPENDIX
08.1.4.0 Equipment for the Reception, Recording, • Other magnetic recording media
and Reproduction of Sound and Vision (D) • Other optical recording media
Includes: • Other recording media (phase-change recording media,
• Television sets, video cassette players and recorders, digi- holographic recording media, molecular recording media)
tal video recorders, DVD players, Blu-ray players, Ultra Excludes:
HD Blu-ray players, streaming boxes, and television aeri-
als of all types • Recorded recording media (09.5.2.0, 09.7.1)
• Radio receivers (radio sets, digital radio sets, internet radio
08.1.9 Other Information and Communication
sets, satellite radio sets, car radios, radio clocks, two-way
Equipment and Accessories (D)
radios, amateur radio receivers, and transmitters)
Includes:
• Portable and nonportable CD players
• Walkie-talkies
• Portable and nonportable sound players
• Baby monitors
• Stereo equipment and CD radio cassette recorder
• Smartwatches
• Turntables, tuners, amplifiers, cassette decks, micro-
phones and speakers, disc jockey equipment, and karaoke • Fitness trackers, and other wearable devices, such as, for
systems example, smart glasses that do not work without a smart-
phone or tablet
• Car stereos, video systems for cars
• E-book readers
• Set-top boxes, satellite receivers, IPTV receivers, and
television converter boxes • Chargers, batteries, cables, power banks, docking sta-
tions, covers, cases, cradles, and mounts
• Digital media players
• Disk drives, processors, main boards, and hard drives
• Headphone, earplugs, and wireless/Bluetooth headsets
08.1.9.1 Other Information and Communication
08.1.5 Unrecorded Recording Media (SD) Equipment (D)
Includes:
Includes:
• CDs (R and RW)
• Walkie-talkies
• DVDs (R and RW)
• Baby monitors
• Blu-ray discs (R and RE)
• Smartwatches
• Video cassettes
• Fitness trackers, and other wearable devices, such as
• Audio tapes, cassettes, DAT
smart glasses that do not work without a smartphone or
• External hard drives and solid state disks, network tablet
attached storage • E-book readers
• USB keys/flash drives
• SD cards, compact flash, and so on 08.1.9.2 Other Information and Communication
Accessories (SD)
• Magnetic data tapes
• Other magnetic recording media Includes:
• Other optical recording media • Chargers, batteries for information and communication
• Other recording media (phase-change recording media, equipment, cables, power banks, docking stations, cov-
holographic recording media, molecular recording media) ers, cases, cradles, and mounts
• Disk drives, processors, main boards, and hard drives
Excludes:
• Recorded recording media (09.5.2.0, 09.7.1) 08.2 Software Excluding Games
08.1.5.0 Unrecorded Recording Media (SD) 08.2.0 Software (S)
Includes:
Includes:
• Computer software packages, such as operating systems,
• CDs (R and RW) applications, programming languages, and so on
• DVDs (R and RW)
Includes also:
• Blu-ray discs (R and RE)
• Video cassettes • Software subscriptions and use of online software
• Audio tapes, cassettes, DAT • Apps
• External hard drives and solid state disks, network Excludes:
attached storage
• Video game software (09.2.1.1)
• USB keys/flash drives
• Removable media containing books, dictionaries, ency-
• SD cards, compact flash, and so on clopedias, foreign language trainers, multimedia presen-
• Magnetic data tapes tations, and so on in the form of software (09.7.1.1)
403
APPENDIX
08.2.0.0 Software (S) 08.3.2.0 Mobile Communication Services (S)

Includes: Includes:
• Computer software packages, such as operating systems, • National calls, including voice and video calls
applications, programming languages, and so on • International calls, including voice and video calls
Includes also: • Messages, including voice, written (short message ser-
vice), and image (multimedia message service) messages,
• Software subscriptions and use of online software
and subscription fees for other messengers
• Apps
• Additional calling features, such as voicemail and call
Excludes: display, whether sold separately or bundled with the
mobile local service plan
• Video game software (09.2.1.1)
• Voice and messaging cell phone plans that also include
limited data
clopedias, foreign language trainers, multimedia presen-
tations, and so on in the form of software (09.7.1.1) • Mobile phone voice, text, and data plans
• Other mobile telephone services
08.3 Information and Communication Services Includes also:
08.3.1 Fixed Communication Services (S) • Costs of telephone equipment if included in subscription costs
Includes:
• Mobile phones included in a package, that is, prepaid or
• Installation and subscription costs of personal telephone postpaid packages, generally tied to a specific operator for
equipment a certain period of time if not separately priced
• Telephone calls from a private line or from a public line
(public telephone box, post office cabin, and so on) 08.3.3 Internet Access Provision Services and Net
Storage Services (S)
• Local, regional, national, and international calls Includes:
• Telephone calls from hotels, cafés, restaurants, and the like
• Internet access services provided by operators of wired,
wireless, or satellite infrastructure
08.3.1.0 Fixed Communication Services (S)
• Cloud storage, file hosting, and web hosting services
Includes:
• Subscriptions for email services
• Installation and subscription costs of personal telephone
Includes also:
equipment
• Telephone calls from a private line or from a public line • Activation and installation fees and monthly rate
(public telephone box, post office cabin, and so on) 08.3.3.0 Internet Access Provision Services and Net
• Local, regional, national, and international calls Storage Services (S)
• Telephone calls from hotels, cafés, restaurants, and the like Includes:
• Internet access services provided by operators of wired,
08.3.2 Mobile Communication Services (S)
Includes: wireless, or satellite infrastructure
• Cloud storage, file hosting, and web hosting services
• National calls, including voice and video calls
• Subscriptions for email services
• International calls, including voice and video calls
• Messages, including voice, written (short message ser- Includes also:
vice), and image (multimedia message service) messages, • Activation and installation fees and monthly rate
and subscription fees for other messengers
08.3.4 Bundled Telecommunication Services (S)
• Additional calling features, such as voicemail and call
display, whether sold separately or bundled with the Includes:
mobile local service plan • Telephony/internet/television packages
• Voice and messaging cell phone plans that also include • Any combination of telecommunication package
limited data
• Mobile phone voice, text, and data plans 08.3.4.0 Bundled Telecommunication Services (S)
• Other mobile telephone services Includes:
Includes also: • Telephony/internet/television packages
• Any combination of telecommunication package
• Costs of telephone equipment if included in subscription
costs 08.3.5 Repair and Rental of Information and
• Mobile phones included in a package, that is, prepaid or Communication Equipment (S)
postpaid packages, generally tied to a specific operator for The cost of materials is included only if the materials are not
a certain period of time if not separately priced separately invoiced.
404
APPENDIX
Includes: • Subscription to cable TV, satellite TV, IPTV, and Pay-TV

• Repair of all information and communication equipment • Online videorecorder services (web-based DVR services)
• Rental of telephones, telefax machines, telephone- • Video-on-demand services
answering machines, and telephone loudspeakers • Subscription to TV via decoder and rental of decoders
• Rental of wireless telephone equipment Excludes:
• Rental of internet access provision equipment
• Rental or subscription of video game software and online
• Rental of telegraphy, telex, telefax, radiotelephony, radio- games (09.2.1.1)
telegraphy, and radiotelex equipment
• Audiovisual content purchased online for immediate
08.3.5.0 Repair and Rental of Information and downloading (09.5.2.0)
Communication Equipment (S)
08.3.9.9 Other Information and Communication Services (S)
Includes:
Includes: • Telegraphy, telex, and telefax services
• Repair of all information and communication equipment • Radiotelephony, radiotelegraphy, and radiotelex services
• Rental of telephones, telefax machines, telephone- • Voice over Internet Protocol provision (nomadic use)
answering machines, and telephone loudspeakers • Rental/lease fees for a decoder
• Rental of wireless telephone equipment • Software installation services
• Rental of internet access provision equipment
• Rental of telegraphy, telex, telefax, radiotelephony, radio- 09 Recreation, Sport, and Culture
telegraphy, and radiotelex equipment Division 09 covers a wide range of goods and services for
recreation, sport, and culture, and it is structured into eight
08.3.9 Other Information and Communication Services (S) groups. The first group covers recreation durables: photo-
Includes: graphic equipment, other major durables for recreation, such
• Telegraphy, telex, and telefax services
as camper vans, boats, yachts, airplanes, and the like. Divi-
sion 09.2 covers nonmajor durable recreational goods, such
• Voice over Internet Protocol provision (nomadic use) as games and toys, including video game computers, cel-
• TV and radio licenses ebration articles, equipment for sport, camping, and open-
• Subscription to cable TV, satellite TV, IPTV, and Pay-TV air recreation. The third group covers garden products and
• Streaming services plants and flowers and purchases of pets and expenditures
for pets, excluding veterinary services. Recreational ser-
• Online videorecorder services (web-based DVR services)
vices cover rental, maintenance, and repair of goods clas-
• Video-on-demand services sified in Division 09, veterinary and other services for pets,
• Rental or subscription of CDs, videotapes, DVDs, Blu- recreational and leisure services, such as amusement parks,
ray discs, and software (excluding game software) games of chance and expenditures for sporting services,
both expenditures for practising sports as well as expendi-
Includes also:
tures for attendance of sports events.
• Radiotelephony, radiotelegraphy, and radiotelex services Division 09.5 covers cultural goods, such as musical
• Software installation services instruments and audiovisual media, followed by cultural
services (cinema, theatre, concert, museum, and other cul-
• Rental/lease fees for a decoder, TV set-top boxes, and so on
tural sites, and photographic services). Whether books or
Excludes: newspapers are in printed or electronic form is irrelevant for
classifying them under 09.7.
• Rental or subscription of video game software and online
The last group (09.8) covers package holidays, which as
games (09.4.3.1)
a bundle include all kinds of services, such as transportation,
• Audiovisual content purchased online for immediate accommodation, food provision, tour guide, and so on. Expen-
downloading (09.5.2.0) ditures for package holidays cannot be split into its components
and thus a separate classification category was created. The
08.3.9.1 TV and Radio Licenses and Fees (S)
length of the package holiday and whether the holiday destina-
Includes: tion is within the home country or outside does not matter.
• TV and radio licenses
09.1 Recreational Durables
08.3.9.2 Subscription to Audiovisual Content, Streaming
09.1.1 Photographic and Cinematographic Equipment
Services, and Rentals of Audiovisual Content (S)
and Optical Instruments (D)
Includes: Includes:
• Streaming services (film and music) • Still cameras, movie cameras and sound-recording cam-
• Rental, download or subscription of CDs, videotapes, DVDs, eras, film and slide projectors, enlargers, and film pro-
Blu-ray discs, and software (excluding game software) cessing equipment
405
APPENDIX
• Video cameras, including camcorders, action cameras • Sea-diving equipment

• Screens, viewers, lenses (including zoom lenses), lenses, • Horses and ponies, horse- or pony-drawn vehicles, cam-
flash attachments, filters, exposure meters, and so on els, and dromedaries, purchased for recreational purposes
• Photographic developer and photographic paper and related equipment (harnesses, bridles, reins, saddles,
and so on)
• Binoculars, microscopes, telescopes, and compasses
• Golf carts
Includes also:
• Large steel-frame swimming pools for the garden
• Separate material purchased by households with the • Billiard tables, ping-pong tables, pinball machines, gam-
intention of undertaking the repairs themselves ing machines, and so on
• Batteries and chargers for cameras and other photographic • Bicycles with four wheels
and cinematographic equipment • Other major durables for recreation n.e.c.
Excludes: Includes also:
• Dash cameras (07.2.1.3)
• Separate material purchased by households with the
• Webcams (08.1.3.2) intention of undertaking the repairs themselves
09.1.1.1 Cameras (D) Excludes:
Includes: • Horses and ponies, horse- or pony-drawn vehicles, cam-
• Still cameras, movie cameras and sound-recording camels and dromedaries, and related equipment purchased
eras, film and slide projectors, enlargers, and film pro- for personal transport (07.1.4.0)
cessing equipment • Inflatable boats, rafts, and swimming pools for children
• Video cameras, including camcorders, action cameras and the beach (09.2.2.2)
Includes also: 09.1.2.1 Camper Vans, Caravans, and Trailers (D)
• Separate material purchased by households with the Includes:

intention of undertaking the repairs themselves • Camper vans, caravans, and trailers
Excludes: 09.1.2.2 Airplanes, Microlight Aircraft, Gliders, Hang
• Dash cameras (07.2.1.3) Gliders, and Hot-Air Balloons (D)
• Webcams (08.1.3.2) Includes:
09.1.1.2 Accessories for Photographic and • Airplanes, microlight aircraft, gliders, hang gliders, and
Cinematographic Equipment (D) hot-air balloons
Includes: 09.1.2.3 Boats, Yachts, Outboard Motors, and Other
• Screens, viewers, lenses (including zoom lenses), Water Sports Equipment (D)
lenses, flash attachments, filters, exposure meters, and Includes:
so on • Boats, yachts, outboard motors, sails, jet sky, rigging, and
• Photographic developer and photographic paper superstructures
Includes also: • Vessels for recreation, sailboats, sailboards, and water
sport boards
• Separate material purchased by households with the
intention of undertaking the repairs themselves • Canoes, kayaks, and windsurfing boards
• Camera-specific batteries and chargers • Sea-diving equipment
Excludes:
09.1.1.3 Optical Instruments (D)
• Inflatable boats, rafts, and swimming pools for children
Includes: (09.2.2.2)
• Binoculars, microscopes, telescopes, and compasses
09.1.2.4 Horses, Ponies, Camel and Dromedaries, and
09.1.2 Major Durables for Recreation (D) Accessories (D)
Includes: Includes:
• Camper vans, caravans, and trailers • Horses and ponies, horse- or pony-drawn vehicles, camels and
• Airplanes, microlight aircraft, gliders, hang gliders, and dromedaries, purchased for recreational purposes and related
hot-air balloons equipment (harnesses, bridles, reins, saddles, and so on)
• Boats, yachts, outboard motors, sails, rigging, and superstructures Excludes:
• Vessels for recreation, sailboats, sailboards, and water • Horses and ponies, horse- or pony-drawn vehicles, cam-
sport boards els and dromedaries, and related equipment purchased
• Canoes, kayaks, and windsurfing boards for personal transport (07.1.4.0)
406
APPENDIX
09.1.2.9 Other Major Durables for Recreation (D) • Electronic games

Includes: • Video game software (for game consoles, computers, tab-
lets, smartphones, download, and on any media, including
• Golf carts
CD-ROMs, cartridges, DVDs, Blu-ray discs, flash drives,
• Snowmobiles and so on)
• Large steel-frame swimming pools for the garden • Game apps
• Billiard tables, ping-pong tables, pinball machines, gam-
Excludes:
ing machines, and so on
• Bicycles with four wheels • Video game subscriptions and rentals (09.4.3.1)
• Other major durables for recreation n.e.c.
09.2.1.2 Other Games, Toys, and Hobbies (SD)
Excludes:
Includes:
and the beach (09.2.2.2) • Traditional games, for example, card games, parlor
games, board games, chess sets
09.2 Other Recreational Goods • Dolls
• Toy cars, including toy trains, toy bicycles, and tricycles
09.2.1 Games, Toys, and Hobbies (SD)
Includes: • Soft toys, teddy bears, and so on
• Toy construction sets
• Card games, board games, parlor games, chess sets, and
the like • Puzzles
• Video game software, video game computers that plug • Modeling clay
into a television set, video game cassettes, video game • Masks
CD-ROMs, and video game downloads • Disguises
• Game apps • Jokes
• Gamepads, joysticks, racing wheels, and other accesso- • Models/replicas of planes, boats, trains, and so on
ries for video gaming • Hobby-stamp-collecting requisites (used or canceled
• Electronic games postage stamps, stamp albums, and so on)
• Toys of all kinds including dolls, soft toys, toy cars and • Other items for collections (coins, medals, minerals, zoo-
trains, toy bicycles and tricycles, toy construction sets, logical and botanical specimens, and so on) and other
puzzles, modeling clay, electronic games, masks, dis- tools and articles n.e.c. for hobby purposes
guises, jokes, novelties, fireworks and rockets, festoons,
and holiday decorations Excludes:
• Stamp-collecting requisites (used or canceled postage • Collectors’ items falling into the category of works of art
stamps, stamp albums, and so on), other items for collec- or antiques (05.1.1.1 or capital formation if acquired pri-
tions (coins, medals, minerals, zoological and botanical marily as stores of value)
specimens, and so on), and other tools and articles n.e.c. • Children’s scrapbooks (09.7.1.9)
for hobbies
09.2.1.3 Celebration Articles (ND)
Includes also:
Includes:
• Separate purchases of materials made by households,
with the intention of undertaking the maintenance and • Fireworks and rockets
repair themselves • Festoons
Excludes: • Christmas trees
• Holiday decorations (for Christmas, Easter, Hanukkah,
• Collectors’ items falling into the category of works of art
Eid, Diwali, and similar)
or antiques (05.1.1.1 or capital formation if acquired pri-
marily as stores of value)
09.2.2 Equipment for Sport, Camping, and Open-Air
• Video game subscriptions and rentals (09.4.3.1)
Recreation (SD)
• Children’s scrapbooks (09.7.1.9) Includes:
09.2.1.1 Video Game Computers, Game Consoles, • Gymnastic, physical education, and sport equipment,
Game Apps, and Software (SD) such as balls, shuttlecocks, nets, rackets, bats, skis, golf
clubs, foils, sabers, poles, weights, discuses, javelins,
Includes:
dumbbells, chest expanders, and other body-building
• Video game computers equipment
• Video game consoles • Parachutes, paragliders, and other skydiving equipment
• Gamepads, joysticks, racing wheels, and other accesso- • Firearms and ammunition for hunting, sport, and personal
ries for video gaming protection
407
APPENDIX
• Fishing rods and other equipment for fishing • Tents, sleeping bags, backpacks, air mattresses and inflat-
• Equipment for beach and open-air games, such as bowls, ing pumps, camping stoves, barbecues, and other acces-
croquet, flying disks, volleyball, and inflatable boats, sories related to camping
rafts, and swimming pools Excludes:
• Camping furniture (05.1.1.2)
• Skateboards, kickboards, smart balance wheels, and
• Repair of equipment for camping and open-air recreation
hoverboards
(09.4.4.0)
• Camping equipment, such as tents and accessories, sleep-
ing bags, backpacks, air mattresses and inflating pumps, 09.3 Garden Products and Pets
camping stoves, and barbecues
09.3.1 Garden Products, Plants, and Flowers (ND)
Includes also: Includes:
• Game-specific footwear (ski boots, football boots, golfing • Soil, peat and fertilizers, pesticides, and composts
shoes, and other such footwear fitted with ice skates, rollers, • Turf for lawns, specially treated soils for ornamental gar-
spikes, studs, and so on); protective headgear for sports; other dens, horticultural preparations
protective gear for sports, such as life jackets, boxing gloves, • Pots and pot holders
body padding, shin guards, goggles, belts, supports, and so on
• Decorations and ornaments for gardens (which are not
• Separate purchases of materials made by households with the
plants)
intention of undertaking the maintenance or repair themselves
• Indoor plants, including natural or artificial
• GPS (satellite positioning) equipment for boats or hiking
• Outdoor plants
Excludes: • Seeds, bulbs, and tubers for planting
• Garden and camping furniture (05.1.1.2) • Shrubs
• Crash helmets for motorcycles and bicycles (07.2.1.3) • Indoor flowers (natural or artificial in vase or not)
• Repair of equipment for sport (09.4.4.0) • Outdoor flowers
• Flower seeds and bulbs
09.2.2.1 Equipment for Sport (SD)
• Natural or artificial flowers and wreaths for decoration of
Includes: burial places
• Gymnastic, physical education, and sport equipment, • Cut flowers
such as balls, shuttlecocks, nets, rackets, bats, skis, golf • Delivery charges for flowers and plants if not separately
clubs, discuses, and javelins priced
• Parachutes, paragliders, and other skydiving equipment
Excludes:
• Firearms and ammunition for hunting, sport, and personal
protection • Gardening services (04.4.4.9, 05.6.2.1)
• Fishing rods and other equipment for fishing • Gardening equipment (05.5.1.0)
• Skateboards, kickboards, smart balance wheels, and hoverboards • Gardening tools (05.5.2.1)
• Insecticides and pesticides for household use; gardening
Includes also:
gloves (05.6.1.9)
• Game-specific footwear (ski boots, football boots, golfing • Delivery charges for flowers and plants if separately
shoes, and other such footwear fitted with ice skates, roll- priced (07.4.9.2)
ers, spikes, studs, and so on) • Christmas trees (09.2.1.3)
• Game-specific sportswear (ski suits and so on)
• Protective headgear for sports 09.3.1.1 Garden Products (ND)
• Other protective gear for sports, such as life jackets, box- Includes:
ing gloves, sport gloves, body padding, shin guards, pads, • Soil, peat and fertilizers, pesticides, and composts
goggles, belts, supports, and helmets, for example, for
skateboarding, inline skating, and ice hockey • Turf for lawns, specially treated soils for ornamental gar-
dens, horticultural preparations
Excludes: • Pots and pot holders
• Repair of equipment for sport (09.4.4.0) • Decorations and ornaments for gardens (which are not plants)
09.2.2.2 Equipment for Camping and Open-Air Excludes:

Recreation (SD) • Gardening services (04.4.4.9, 05.6.2.1)
Includes: • Gardening equipment (05.5.1.0)
• Equipment for beach and open-air games, such as bowls, • Gardening tools (05.5.2.1)
croquet, flying disks, volleyball and inflatable boats, rafts, • Insecticides and pesticides for household use; gardening
and swimming pools gloves (05.6.1.9)
408
APPENDIX
09.3.1.2 Plants, Seeds, and Flowers (ND) 09.4 Recreational Services

Includes: 09.4.1 Hire and Repair of Photographic and
• Indoor plants, including natural or artificial Cinematographic Equipment and Optical
• Outdoor plants Instruments (S)
• Seeds, bulbs, and tubers for planting separately invoiced.
• Shrubs Includes:
• Indoor flowers (natural or artificial in vase or not)
• Hire of photographic and cinematographic equipment and
• Outdoor flowers optical instruments
• Flower seeds and bulbs • Repair of photographic and cinematographic equipment
• Natural or artificial flowers and wreaths for decoration of and optical instruments
burial places
Excludes:
• Cut flowers
Excludes: the intention of undertaking the repair themselves (09.1.1.2)
• Soil, peat, and fertilizers (09.3.1.1)
09.4.1.0 Hire and Repair of Photographic and
• Christmas trees (09.2.1.3) Cinematographic Equipment and Optical
Instruments (S)
09.3.2 Pets and Related Products (ND)
Includes: The cost of materials is included only if the materials are not
• Pets, pet foods, veterinary and grooming products for Includes:
pets, collars, leashes, kennels, birdcages, fish tanks, cat
litter, and so on • Hire of photographic and cinematographic equipment and
optical instruments
Includes also: • Repair of photographic and cinematographic equipment
• Feed and veterinary products for animals used for trans- and optical instruments
portation, own consumption or recreation Excludes:
Excludes: • Separate purchases of materials made by households with
• Horses and ponies, camels, and dromedaries, for trans- the intention of undertaking the repair themselves (09.1.1.2)
port (07.1.4.0)
09.4.2 Hire, Maintenance, and Repair of Major Durables
• Horses and ponies, camels, and dromedaries, for recre-
for Recreation (S)
ation (09.1.2.4)
• Veterinary and other services for pets (09.4.5.0) separately invoiced.
Includes:
09.3.2.1 Purchase of Pets (ND)
Includes: • Maintenance and repair of major durables for recreation
as described in 09.1.2
• Purchase of pets • Laying up for winter of boats, yachts, camper vans, cara-
Excludes: vans, and so on; hangar services for private planes; and
marina services for boats
• Horses and ponies, camels, and dromedaries, for trans-
port (07.1.4.0) • Hire of major durables for recreation as described in
09.1.2.2, 09.1.2.3, and 09.1.2.9
• Horses and ponies, camels, and dromedaries, for recre-
ation (09.1.2.4) Excludes:
• Fuel for recreational vehicles (07.2.2.1, 07.2.2.2, 07.2.2.3)
09.3.2.2 Products for Pets and Other Household
Animals (ND) • Separate purchases of materials made by households with
Includes:
• Pet foods, veterinary and grooming products for pets, collars,
leashes, kennels, birdcages, fish tanks, cat litter, and so on 09.4.2.1 Hire, Maintenance, and Repair of Camper Vans
and Caravans (S)
Includes also:
• Feed and veterinary products for animals used for trans- separately invoiced.
portation, own consumption or recreation Includes:
Excludes: • Hire of camper vans and caravans
• Veterinary and other services for pets and other house- • Maintenance and repair of camper vans and caravans
hold animals (09.4.5.0) • Laying up for winter of camper vans and caravans
409
APPENDIX
Excludes: 09.4.4.0 Hire and Repair of Equipment for Sport,

Camping, and Open-Air Recreation (S)
Includes:
• Hire and repair of equipment for sport, camping, and
09.4.2.2 Hire, Maintenance, and Repair of Other Major open-air recreation
Durables for Recreation (S) • Hire of beach umbrellas and deckchairs
separately invoiced. 09.4.5 Veterinary and Other Services for Pets (S)
Includes: Includes:
• Hire of major durables for recreation as described in • Veterinary and other services for pets, such as grooming,
09.1.2.2, 09.1.2.3, and 09.1.2.9 boarding, tattooing, and training
• Laying up for winter of boats, yachts, and so on; han- • Pet boarding services or pet day care services
gar services for private planes; and marina services for
boats Includes also:
Excludes: • Veterinary and hosting services for animals used for

transportation
Excludes:
the intention of undertaking the maintenance or repair • Products for pets (09.3.2.2)
09.4.5.0 Veterinary and Other Services for Pets (S)
09.4.3 Hire and Repair of Games, Toys, and Hobbies (S) Includes:
separately invoiced. • Veterinary and other services for pets, such as grooming,
Includes: tattooing, and training
• Pet boarding services or pet day care services
• Rental and subscriptions of video game consoles and apps
or software Includes also:
• Repair of video game consoles, toys, and hobby articles • Veterinary and hosting services for animals used for
transportation
09.4.3.1 Rental of Game Software and Subscription to
Excludes:
Online Games (S)
Includes: • Products for pets (09.3.2.2)
• Rental of game software (games on CDs, DVDs, Blu-ray
discs, and so on) 09.4.6 Recreational and Sporting Services (S)
Includes:
• Subscription to play online games (or streaming)
Services provided by:
09.4.3.2 Rental and Repair of Games, Toys, • Sports stadiums, horse-racing courses, motor-racing cir-
and Hobbies (S) cuits, velodromes, and so on
The cost of materials is included only if the materials are not • Skating rinks, swimming pools, golf courses, gymnasia,
separately invoiced. fitness centers, tennis courts, squash courts, and bowling
Includes: alleys
• Fairgrounds and amusement parks
• Hire and repair of video game consoles and other equip-
ment to play games • Roundabouts, seesaws, and other playground facilities for
children
• Hire and repair of toys and hobby articles
• Pinball machines and other games for adults other than
• Hire of toys and games
games of chance
• Arcade games
09.4.4 Hire and Repair of Equipment for Sport,
Camping, and Open-Air Recreation (S) • Ski slopes, ski lifts, and the like
The cost of materials is included only if the materials are not • Out-of-school individual or group lessons in bridge,
separately invoiced. chess, aerobics, skating, skiing, swimming, or other
Includes: pastimes
• Hire and repair of equipment for sport, camping, and • Membership fees for sports clubs and fitness centers
open-air recreation • Services of mountain guides, tour guides, and so on
• Hire of beach umbrellas and deckchairs • Navigational aid services for boating
410
APPENDIX
• Water parks Excludes:

• Arcade games • Cable car and chairlift transport not at ski resorts or holi-
• Services of mountain guides, tour guides, and so on day centers (07.3.6.0)
• Entrance fees for dancing establishments, nightclubs
09.4.6.3 Sporting Services—Attendance (S)
• Fees for taking part in sports competitions
Includes:
• Fees for sports title and category certificates
• Admissions (tickets) to attend live sporting events like
Includes also:
football games, hockey games, ice skating competitions,
• Hire of game-specific footwear (ski boots, football boots, ski competitions, soccer games, tennis matches, horse-rac-
golfing shoes, and other such footwear fitted with ice ing courses, motor-racing circuits, velodromes, and so on
skates, rollers, spikes, studs, and so on)
• Paid fishing 09.4.7 Games of Chance (S)
Includes:
• Membership fees of fishermen’s and hunters’ clubs
• Service charges for lotteries, bookmakers, totalizators, casi-
Excludes:
nos, and other gambling establishments, gaming machines,
• Cable car and chairlift transport not at ski resorts or holi- bingo halls, scratch cards, sweepstakes, and so on
day centers (07.3.6.0) • Online games of chance
09.4.6.1 Recreational and Leisure Services (S) 09.4.7.0 Games of Chance (S)
Includes: Includes:
Services provided by:
• Service charges for lotteries, bookmakers, totalizators, casi-
• Fairgrounds and amusement parks nos, and other gambling establishments, gaming machines,
• Roundabouts, seesaws, and other playground facilities for bingo halls, scratch cards, sweepstakes, and so on
children • Online games of chance
• Out-of-school individual or group lessons in bridge,
chess, sewing, cooking, and so on 09.5 Cultural Goods
• Pin-ball machines and other games for adults other than 09.5.1 Musical Instruments (D)
games of chance Includes:
• Water parks • Musical instruments of all sizes, including electronic musical
• Arcade games instruments, such as pianos, organs, violins, guitars, drums,
• Services of mountain guides, tour guides, and so on trumpets, clarinets, flutes, recorders, harmonicas, and so on
• Entrance fees for dancing establishments, nightclubs Includes also:
09.4.6.2 Sporting Services—Practice (S) • Spare parts of musical instruments
Includes: Excludes:
• Skating rinks, swimming pools, golf courses, gymnasia, • Toy instruments (09.2.1.2)
fitness centers, tennis courts, squash courts, and bowling
alleys 09.5.1.0 Musical Instruments (D)
• Ski slopes, ski lifts, and the like Includes:
• Cable car and chairlift transport at ski resorts and holiday • Musical instruments of all sizes, including electronic
centers musical instruments, such as pianos, organs, violins, gui-
• Out-of-school individual or group lessons in aerobics, tars, drums, trumpets, clarinets, flutes, recorders, harmon-
skating, skiing, swimming, or other sports icas, and so on
• Memberships fees for sports clubs and fitness centers Includes also:
• Navigational aid services for boating • Spare parts of musical instruments
• Fees for taking part in sports competitions
Excludes:
• Fees for sports title and category certificates
• Toy instruments (09.2.1.2)
Includes also:
• Hire of game-specific footwear (ski boots, football boots, 09.5.2 Audiovisual Media (SD)
golfing shoes, and other such footwear fitted with ice Includes:
skates, rollers, spikes, studs, and so on) • Recorded tapes, CD-ROMs, DVDs, Blu-ray discs, gramo-
• Hunting licenses phone records, and flash drives, for reproduction of sound
• Paid fishing and picture material
• Membership fees of fishermen’s and hunters’ clubs • Downloads of music and films
411
APPENDIX
Excludes: 09.6.2.0 Museums, Libraries, and Cultural Sites (S)

• Software (08.2.0.0) Includes:
• Video games and game apps and software (09.2.1.1) • Museums, art galleries, exhibitions, including historical
• Recorded tapes and CDs, DVDs, Blu-ray, flash drives of monuments and archeological sites
books, novels, plays, poetry, dictionaries, encyclopedias, • Libraries
and so on (09.7.1) • Historic monuments, national parks, zoological and
botanical gardens, and aquaria
09.5.2.0 Audiovisual Media (SD)
Includes: 09.6.3 Photographic Services (S)
• Recorded tapes, CD-ROMs, DVDs, Blu-ray discs, gramo- Includes:
phone records, and flash drives, for reproduction of sound • Services of photographers, such as film developing, print
and picture material processing, enlarging, portrait photography, event pho-
• Downloads of music and films tography and video (for example, for weddings), and so
on
Excludes:
Includes also:
• Software (08.2.0.0)
• Video games and game apps and software (09.2.1.1) • Photographic services provided by nonspecialized shops
• Recorded tapes and CDs, DVDs, Blu-ray discs, flash
(for example, supermarkets, consumer electronic stores,
drives of books, novels, plays, poetry, dictionaries, ency- and so on) and purchased by internet
clopedias, and so on (09.7.1)
09.6.3.0 Photographic Services (S)
09.6 Cultural Services Includes:
09.6.1 Cinemas, Theatres, and Concerts (S) • Services of photographers, such as film developing,
Includes: print processing, enlarging, portrait photography, event
photography and video (for example, for weddings), and
• Cinemas so on
• Theatres, opera houses
Includes also:
• Concert and music venues
• Photographic services provided by nonspecialized shops
• Circuses, sound and light shows, and other
(for example, supermarkets, consumer electronic stores,
Includes also: and so on) and purchased by internet
• Services of musicians, clowns, performers for private
entertainments 09.6.9 Other Cultural Services (S)
Includes:
• Music, dancing, and artistic performance
• Art and music festivals • Services of rental and repair of musical instruments
• Binding services
09.6.1.0 Cinemas, Theatres, and Concerts (S)
• Photography, music, dancing, and artistic classes (in
Includes: classroom or via e-learning)
• Cinemas Includes also:
• Theatres, opera houses
• Rental of premises for rehearsals of amateur music
• Concert and music venues groups, weddings, and other celebrations
• Circuses, sound and light shows, and other
Excludes:
Includes also:
• Formal education of music, dancing, and art (Division 10
• Services of musicians, clowns, performers for private according to the level)
entertainments
• Music, dancing, and artistic performance 09.6.9.0 Other Cultural Services (S)
• Art and music festivals Includes:
• Services of rental and repair of musical instruments
09.6.2 Museums, Libraries, and Cultural Sites (S)
Includes: • Binding services
• Photography, music, dancing, and artistic classes (in
• Museums, art galleries, and exhibitions, including histori-
classroom or via e-learning)
cal monuments and archeological sites
• Libraries Includes also:
• Historic monuments, national parks, zoological and • Rental of premises for rehearsals of amateur music
botanical gardens, and aquaria groups, weddings, and other celebrations
412
APPENDIX
Excludes: 09.7.2 Newspapers and Periodicals (ND)

Includes:
• Formal education of music, dancing, and art (Division 10
according to the level) • Newspapers, magazines, and other periodicals
Includes also:
09.7 Newspapers, Books, and Stationery
• All electronic forms of newspapers and periodicals
09.7.1 Books (SD)
Includes: 09.7.2.1 Newspapers (ND)
• Books educational or not, including atlases, dictionaries, Includes:
encyclopedias, guidebooks, and musical scores • Newspaper bought in kiosks
• Recorded tapes and CDs, DVDs, Blu-ray discs, flash drives • Subscription for newspapers (home delivery)
of educational books novels, plays, poetry, and so on
• Internet subscription for newspapers
• All electronic forms of books (e-books and audio books)
Includes also:
clopedias, foreign language trainers, multimedia presen- • All electronic forms of newspapers
tations, and so on in the form of software
• Scrapbooks and albums for children 09.7.2.2 Magazines and Periodicals (ND)
Includes:
Includes also:
• Lifestyle magazines
• All electronic forms of educational books (e-books and
audio books) • Children magazines
• Hobbies, leisure magazines
Excludes:
• Business, political magazines
• Stamp albums (09.2.1.2)
• TV magazines
09.7.1.1 Educational and Text Books (SD) • Subscription for magazines and periodicals (home
delivery)
Includes:
• Internet subscription for magazines and periodicals
• Formal education text books (school/academic manuals,
and so on) Includes also:
• Recorded tapes and CDs, DVDs, Blu-ray discs, flash • All electronic forms of magazines and periodicals
drives of educational books
• Download of educational books 09.7.3 Miscellaneous Printed Matter (ND)
• Removable media containing books, dictionaries, ency- Includes:
clopedias, foreign language trainers, in the form of • Catalogs and advertising material
software • Posters, plain or picture postcards, calendars
Includes also: • Greeting cards and visiting cards, announcement, and
• All electronic forms of educational books (e-books and
message cards
audio books) • Maps and globes
Includes also:
09.7.1.9 Other Books (SD)
• GPS maps bought separately
Includes:
Excludes:
• Fiction and nonfiction books
• Children’s books, scrapbooks and albums for children, • Prefranked postcards and aerogrammes (07.4.1.1)
and coloring books for children • Stamp albums (09.2.1.2)
• Art books
• Travel guides 09.7.3.0 Miscellaneous Printed Matter (ND)
• Recorded tapes and CDs, DVDs, Blu-ray discs, flash Includes:
drives of books, novels, plays, poetry, and so on
• Catalogs and advertising material
• Download of noneducational books
• Posters, plain or picture postcards, calendars
Includes also: • Greeting cards and visiting cards, announcement, and
• All electronic forms of books (e-books and audio books); message cards
scrapbooks and albums for children • Maps and globes
Excludes: Includes also:
• Stamp albums (09.2.1.2) • GPS maps bought separately
413
APPENDIX
Excludes: 09.8.0.0 Package Holidays (S)

• Prefranked postcards and aerogrammes (07.4.1.1) Includes:
• Stamp albums (09.2.1.2) • All-inclusive holidays or tours which provide for travel,
food, accommodation, guides, and so on
09.7.4 Stationery and Drawing Materials (ND) • Excursion tours including transport and guide
Includes:
• Writing pads, envelopes, account books, diaries, and so on 10 Education Services
• Drawing paper Division 10 covers educational services only. The focus of
• Educational materials, such as exercise books groups 10.1–10.4 is on formal education (that leads to cer-
tificate or diploma). It includes education by radio or televi-
• Wrapping paper sion broadcasting as well as e-learning and correspondence
• Pens, pencils, fountain pens, ballpoint pens, felt-tip pens, courses. It includes admission and registration fees as well
inks, erasers, pencil sharpeners, and so on as tuition fees.
• Stencils, carbon paper, inking pads, correcting fluids, and so on It also includes other education-related fees like camps/
• Paper punches, paper cutters, paper scissors, office glues field trips, course fees, diploma fees, examination fees,
and adhesives, staplers and staples, paper clips, drawing graduation fees, laboratory fees, physical education fee, and
pins, and so on so on.
The breakdown of educational services is based upon
• Drawing and painting materials, such as canvas, card,
the level categories of the 2011 revision of the Interna-
paints, crayons, pastels, and brushes
tional Standard Classification of Education (ISCED 2011)
Includes also: of the United Nations Educational, Scientific and Cultural
Organization.
• Slide rules, geometry instruments, slates, chalks, and pen-
This division does not include expenditures on other edu-
cil boxes
cation-related goods and services, such as:
Excludes:
• School uniforms (03.1.2.3)
• Pocket calculators (08.1.3.2) • Education support services, such as health care ser-
vices (06)
09.7.4.0 Stationery and Drawing Materials (ND) • Transport services except in the case of excursions which
Includes: are part of the normal school program (07.3.2.3)
• Writing pads, envelopes, account books, diaries, and so on • Textbooks and academic journals (09.7.1.1)
• Drawing paper • Stationery (09.7.4.0)
• Educational materials, such as exercise books • Catering services (11.1.2.1)
• Wrapping paper • Accommodation services (11.2.0.3)
• Pens, pencils, fountain pens, ballpoint pens, felt-tip pens,
inks, erasers, pencil sharpeners, and so on 10.1 Early Childhood and Primary Education
• Stencils, carbon paper, inking pads, correcting fluids, and 10.1.0 Early Childhood and Primary Education (S)
so on Includes:
• Paper punches, paper cutters, paper scissors, office glues
• Levels 0 and 1 of ISCED 2011 Early childhood and pri-
and adhesives, staplers and staples, paper clips, drawing mary education
pins, and so on
• Literacy programs for students too old for primary school
• Drawing and painting materials, such as canvas, card,
paints, crayons, pastels, and brushes • Education services for children with special educational
needs
Includes also:
Includes also:
• Slide rules, geometry instruments, slates, chalks, and pen-
cil boxes • Excursions which are part of the normal school program
(travel and accommodation costs)
Excludes:
Excludes:
• Pocket calculators (08.1.3.2)
• Childcare services without educational program (13.3.0.1)
09.8 Package Holidays 10.1.0.1 Early Childhood Education (S)
09.8.0 Package Holidays (S) Includes:
Includes:
• ISCED 2011 level 0: early childhood education is
• All-inclusive holidays or tours which provide for travel, designed primarily to support children’s early cognitive,
food, accommodation, guides, and so on physical, social, and emotional development and intro-
• Excursion tours including transport and guide duce very young children to organized instruction outside
414
APPENDIX
of the family context. Programs are designed for children 10.2.0.0 Secondary Education (S)
below the entry age to primary education and are typi- Includes:
cally center- or school-based. This education may also be
provided in hospitals or in special schools or training cen- • Levels 2 and 3 of ISCED 2011: lower-secondary and
ters; special education services for children with special upper-secondary education. Secondary education is
educational needs. designed to lay the foundation for lifelong learning and
human development and to provide the skills and knowl-
• Special education services for children with special edu-
edge needed either for further studies at postsecondary and
cational needs.
tertiary levels or for entry to the labor market, or both. Pro-
Includes also: grams are increasingly subject-oriented and specialized.
Different study options or streams are offered, including
• Excursions which are part of the normal school program
vocational education and training. Organized instruction
(travel and accommodation costs)
for young people with special needs is also covered.
Excludes: • Secondary education for adults and young people includ-
• Childcare services without educational program (13.3.0.1) ing second chance or reintegration; programs.
• Out-of-school secondary education for adults and young
10.1.0.2 Primary Education (S) people.
Includes: • Education services for adolescents with special educa-
• ISCED 2011 level 1: primary education usually begins
tional needs.
at age five, six, or seven and generally lasts for four to Includes also:
seven years. Programs are normally designed to provide
• Excursions or student exchanges which are part of the
students with fundamental skills in reading, writing, and
mathematics and establish a solid foundation for learning normal school program (travel and accommodation costs)
and understanding core areas of knowledge and personal
and social development. Organized instruction for chil- 10.3 Postsecondary Nontertiary Education
dren with special needs should also be included and lit- 10.3.0 Postsecondary Nontertiary Education (S)
eracy or basic skills programs within or outside the school Includes:
system which are similar in content to program in primary
education. • Level 4 of ISCED 2011: postsecondary nontertiary edu-
cation provides learning experiences building on second-
• Literacy programs for students too old for primary school.
ary education, preparing for labor market entry as well
• Excursions which are part of the normal school program as tertiary education. It aims at the individual acquisition
(travel and accommodation costs). of knowledge, skills, and competencies lower than the
• Education services for children with special educational level of complexity characteristic of tertiary education.
needs. Postsecondary nontertiary education is typically designed
to provide individuals who completed upper-secondary
10.2 Secondary Education education without qualifications required for progression
to tertiary education and for individuals seeking specific
10.2.0 Secondary Education (S)
types of employment when their secondary qualification
Includes:
does not grant such access. For example, graduates from
• Levels 2 and 3 of ISCED 2011: lower-secondary and general secondary programs may choose to complete a
upper-secondary education. Secondary education is nontertiary vocational qualification; or graduates from
designed to lay the foundation for lifelong learning and vocational secondary programs may choose to increase
human development and to provide the skills and knowl- their level of qualifications or specialize further. Usually,
edge needed either for further studies at postsecondary programs at this level are designed for direct labor mar-
and tertiary levels or for entry to the labor market, or both. ket entry. General programs designed to give access to or
improved access to tertiary education can also exist.
Programs are increasingly subject-oriented and specialized.
Different study options or streams are offered, including • Out-of-school postsecondary nontertiary education for
vocational education and training. Organized instruction for adults and young people.
young people with special needs is also covered: • Education services for young people and adults with spe-
cial educational needs.
• Secondary education for adults and young people includ-
ing second chance or reintegration; programs Includes also:
• Out-of-school secondary education for adults and young • Excursions or student exchanges which are part of the
people normal school program (travel and accommodation costs)
• Education services for adolescents with special educa-
tional needs 10.3.0.0 Postsecondary Nontertiary Education (S)
• Excursions or student exchanges which are part of the • Level 4 of ISCED 2011: postsecondary nontertiary educa-
normal school program (travel and accommodation costs) tion provides learning experiences building on secondary
415
APPENDIX
education, preparing for labor market entry as well as • University admission tests
tertiary education. It aims at the individual acquisition • E-learning courses
of knowledge, skills, and competencies lower than the
level of complexity characteristic of tertiary education. 10.5 Education Not Defined by Level
Postsecondary nontertiary education is typically designed
to provide individuals who completed upper-secondary 10.5.0 Education Not Defined by Level (S)
education without qualifications required for progression Includes:
to tertiary education and for individuals seeking specific • Independent tutors (private lessons), tutor centers, home-
types of employment when their secondary qualification work help centers, and the like
does not grant such access. For example, graduates from
• Short educational courses, generally for adults, which
general secondary programs may choose to complete a
do not require any special prior instruction, in particu-
nontertiary vocational qualification; or graduates from
lar cultural development or some types of vocational
vocational secondary programs may choose to increase
training
their level of qualifications or specialize further. Usually,
programs at this level are designed for direct labor mar- • Language immersion courses and international travel for
ket entry. General programs designed to give access to or learning languages
improved access to tertiary education can also exist. • Languages courses in classroom or online
• Out-of-school postsecondary nontertiary education for • Language proficiency tests
adults and young people. • Courses in the use of computers and specific software
• Education services for young people and adults with spe- • Exam preparation courses
cial educational needs. • Online tutoring
Includes also: Excludes:
• Excursions or student exchanges which are part of the • Driving lessons (07.2.4.3)
normal school program (travel and accommodation costs)
• Recreational training courses, such as sport or bridge
10.4 Tertiary Education lessons given by independent teachers (09.4.6.1)
• Removable media containing foreign language trainers
10.4.0 Tertiary Education (S) (09.7.1.1)
Includes:
10.5.0.1 Tutoring (S)
• Levels 5, 6, 7, and 8 of ISCED 2011: tertiary education
builds on secondary education and provides learning Includes:
activities in specialized fields of education. It aims at a • Independent tutors (private lessons to support formal
high level of complexity and specialization. It includes education), tutor centers, homework help centers, and
both academic education and advanced vocational or pro- the like
fessional education. At the highest levels, programs lead • Exam preparation courses
to an advanced research qualification based on advanced
• Online tutoring
study and original research.
Includes also: 10.5.0.9 Other Education Not Defined by
• Excursions or student exchanges which are part of the Level (S)
normal school program (travel and accommodation costs) Includes:
• University admission tests
• Educational programs, generally for adults, which do not
• E-learning courses require any special prior instruction, in particular voca-
tional training or cultural development
10.4.0.0 Tertiary Education (S)
• Language immersion courses and international travels
Includes: with educational purposes (for example, languages)
• Levels 5, 6, 7, and 8 of ISCED 2011: tertiary education • Languages courses in classroom, online, in form of soft-
builds on secondary education and provides learning ware or audio tapes
activities in specialized fields of education. It aims at a • Language proficiency tests
high level of complexity and specialization. It includes
• Information technology courses (for example, learning
both academic education and advanced vocational or pro-
how to use a specific software)
fessional education. At the highest levels, programs lead
to an advanced research qualification based on advanced Excludes:
study and original research.
• Driving lessons (07.2.4.3)
Includes also: • Recreational training courses, such as sport or bridge
• Excursions or student exchanges which are part of the lessons given by independent teachers (09.4.6.1)
normal school program (travel and accommodation • Removable media containing foreign language trainers
costs) (09.7.1.1)
416
APPENDIX
11 Restaurants and Accommodation 11.1.1.1 Restaurants, Cafés, and the Like—With Full
Services Service (S)
Division 11 covers food and beverage serving services Food and beverages services provided by restaurants, cafés,
provided by restaurants, cafés, and similar facilities, either and similar eating facilities providing full service consisting
with full, limited, or self-service, or by canteens, cafete- of waiter service to individual customers seated at tables,
rias, or refectories at work or school, and other educational with or without entertainment.
establishment’s premises. The distinction between full ser- Includes:
vice and limited service relates to the range of the food • Food and beverages provided by full-service facilities
and beverage serving services: a service by waiters to the mainly serving drinks: cafés, buffets, bars, tearooms, and
individual customer seated at tables will be qualified to be similar
a full service. • Food and beverages provided by full-service restaurants,
Accommodation in this division includes services for cafés, and the like and consumed off their premises; food
visitors and other travelers away from their principal or and beverages provided by full-service restaurants, cafés,
secondary residence. When not separately invoiced, it also and the like and home-delivered
includes food and beverage services and other serving ser-
vices, such as housekeeping, parking, laundry, swimming • Food and beverage full services in hotels or other
pools and exercise rooms, recreational facilities, and confer- accommodation places, if separately invoiced from
ence and convention facilities. accommodation
Excludes: • Food and beverage full services in places providing recre-
ational, cultural, sporting, or entertainment services (the-
• Permanent principal or secondary residence accommo- atres, cinemas, sports stadiums, swimming pools, sports
dation (Division 04) complexes, museums, art galleries, nightclubs, danc-
11.1 Food and Beverage Serving Services ing establishments, and similar facilities) if separately
invoiced
11.1.1 Restaurants, Cafés, and the Like (S) • Tobacco that is consumed with a shisha or hookah pipes
Food and beverages services provided by restaurants, cafés, in restaurants, cafés, or shisha lounges
and similar eating facilities either providing full, limited, or
self-service. • Narcotics purchased in coffee shops, if service charge is
Includes also: applied
• Tips
in restaurants, cafés, or shisha lounges Excludes:
• Narcotics purchased in coffee shops, if service charge is • Tobacco purchases (02.3.0)
applied • Food and beverage services on public transport, if not
• Tips separately invoiced (07.3)
Excludes: • Food and beverage provided by supermarkets and mar-
kets and consumed at their premises (related products in
• Food products and beverages dispensed by automatic Division 01)
vending machines not as a delivery service provided res-
taurants, cafés, buffets, bars, tearooms, and similar facili- • Food and beverages delivery, if separately invoiced
ties (01.1.9.1) (07.4.9.2)
• Tobacco purchases (02.3.0) • Food and beverage services in places providing recre-
ational, cultural, sporting, or entertainment services,
• Food and beverage services on public transport, if not if not separately invoiced (09.4.6, 09.6.1.0, 09.6.2.0,
separately invoiced (07.3) 09.6.9.0)
• Food and beverage provided by supermarkets and mar- • Entrance fees for dancing establishments, nightclubs
kets and consumed at their premises (related products in (09.4.6.1)
Division 01)
• Food and beverage services provided in package holi-
• Food and beverages delivery services, if separately days, if not separately invoiced (09.8.0.0)
invoiced (07.4.9.2)
• Food and beverage services in hotels or other lodging
• Dancing establishments and nightclubs entrance fees places, if not separately invoiced from accommodation
(09.4.6.1) (11.2)
• Food and beverage services in places providing rec-
reational, cultural, sporting or entertainment services, 11.1.1.2 Restaurants, Cafés, and the Like—With Limited
if not separately invoiced (09.4.6, 09.6.1.0, 09.6.2.0, Service (S)
09.6.9.0) Food and beverages serving services provided by limited
• Food and beverage services provided in package holi- and self-service facilities that are without waiter service and
days, if not separately invoiced (09.8.0.0) with or without seating.
Includes:
• Food and beverage services provided by hotels or other
lodging places, if not separately invoiced from accommo- • Food and beverage services provided by self-service
dation (11.2) restaurants
417
APPENDIX
• Food and beverages provided by refreshment stands, fish • Food and beverages provided to hospital inpatients if not
and chip stands, and fast-food outlets without seating and separately invoiced (06.3)
takeaway facilities
11.1.2.1 Canteens, Cafeterias, and Refectories of
• Food products and beverages prepared on the premises
Universities, Schools, and Kindergartens (S)
and dispensed through automatic vending machines as a
delivery service in a self-service restaurant or similar (for Food and beverages provided by canteens, cafeterias, or
example, automat restaurants) refectories, that is, restaurants, cafés, and the like at in
• Food and beverage for immediate consumption provided schools, universities, and other educational establishments.
by kiosks, food stalls, street vendors, and the like Includes:
Includes also: • Kindergarten canteens

• School canteens
• Ice cream parlors and cake serving places
• University refectories
• Food and beverages provided by catering contractors
whether collected by the customer or delivered to the cus- Excludes:
tomer’s home • Food and beverages provided by schools, university,
• Food and beverage services on public transport (coaches, and other educational establishments, if not separately
trains, boats, airplanes, and so on), if separately invoiced invoiced (10.1–10.5)
in restaurants, cafés, or shisha lounges 11.1.2.9 Other Canteens, Cafeterias, and Refectories (S)
• Narcotics purchased in coffee shops, if service charge is Food and beverages provided by canteens, cafeterias,
applied or refectories, that is, restaurants, cafés, and the like at
work/office premises, to hospital inpatients, if separately
• Tips
invoiced.
Excludes: Includes:
• Food products and beverages dispensed by automatic • Military messes and wardrooms
vending machines not as a delivery service provided res-
Excludes:
taurants, cafés, buffets, bars, tearooms, and similar facili-
ties (01.1.9.1) • Food and beverages provided to hospital inpatients if not
• Tobacco purchases (02.3.0) separately invoiced (06.3)
• Food and beverage services on public transport, if not
separately invoiced (07.3) 11.2 Accommodation Services
• Food and beverage provided by supermarkets and mar- 11.2.0 Accommodation Services (S)
kets and consumed at their premises (related products in Accommodation services for visitors and other travelers
Division 01) away from their permanent principal or secondary residence.
• Food and beverages delivery, if separately invoiced It includes other services, when not separately invoiced,
(07.4.9.2) such as food and beverage services, housekeeping, parking,
laundry, swimming pools and exercise rooms, recreational
• Food and beverage services in places providing recre- facilities, and conference and convention facilities.
ational, cultural, sporting, or entertainment services, if not Includes:
separately invoiced (09.4.6, 09.6.1.0, 09.6.2.0, 09.6.9.0)
• Entrance fees for dancing establishments, nightclubs • Resorts, hotels, hotels letting rooms by the hour, motels,
(09.4.6.1) inns, and pensions
• Food and beverage services provided in package holi- • Accommodation services provided by private, holiday
days, if not separately invoiced (09.8.0.0) homes, guesthouses and boarding houses as secondary
residence, and other “bed and breakfast” units
• Food and beverage services in hotels or other lodging places,
if not separately invoiced from accommodation (11.2) • Time-share units
• Holiday villages and holiday centers, camping, and onsite
11.1.2 Canteens, Cafeterias, and Refectories (S) (nonmobile) caravans and boats
Food and beverages provided by canteens, cafeterias, or refec- • Youth hostels and mountain refuges
tories, that is, restaurants, cafés, and the like at work/office • Bungalows, chalets, housekeeping cottages, and cabins
premises, in schools, universities, and other educational estab-
• Student residences, school, and other educational estab-
lishments, to hospital inpatients, if separately invoiced.
lishments dormitories, when separately invoiced; hostels
Includes:
and other accommodations for workers
• University refectories, military messes, and wardrooms • Railway sleeping cars and other public transport, when
Excludes: separately invoiced
• Website fees for people to list, find, and rent lodging
• Food and beverages provided by schools, university,
and other educational establishments, if not separately • Travel agency fees for accommodation if priced separately
invoiced (10.1–10.5) • Tips for bellmen, bellhops, hotel porters, and maids
418
APPENDIX
Excludes: 11.2.0.3 Accommodation Services of Boarding

Schools, Universities, and Other Educational
• Payments of households occupying a room in a hotel or
Establishments (S)
boarding house as their main residence (04.1.1.0)
Includes:
• Rentals paid by households for a secondary residence or
time-share residence (04.1.2) • Student residences, school, and other educational estab-
• Telephone calls (08.3.1.0) lishments dormitories, when separately invoiced
• Package holidays (09.8.0.0) Excludes:
• Breakfast, meals, and other food and beverages provided • Housing in orphanages, homes for disabled or malad-
in accommodation and similar establishments, when sep- justed persons (13.3.0.1, 13.3.0.2)
arately invoiced (11.1.1)
• Housing in orphanages, homes for disabled or malad- 11.2.0.9 Other Accommodation Services (S)
justed persons (13.3.0.2) Includes:
• Hostels and other accommodations for workers
11.2.0.1 Hotels, Motels, Inns, and Similar
Accommodation Services (S) • Railway sleeping cars and other public transport, when
separately invoiced
Includes:
• Website fees for people to list, find, and rent lodging
• Accommodation services in resorts, hotels, and motels
• Travel agency fees for accommodation if priced
• Accommodation services in inns, pensions, and similar separately
establishments
Excludes:
• Short-term accommodation services provided by private,
holiday homes, guesthouses and boarding houses as sec- • Telephone calls (08.3.1.0)
ondary residence, and other “bed and breakfast” units • Breakfast, meals, and other food and beverages provided
• Time-share units in accommodation and similar establishments, when sep-
arately invoiced (11.1.1)
Includes also:
• Housing in orphanages, homes for disabled or malad-
• Hotels letting rooms by the hour justed persons (13.3.0.3)
Excludes:
12 Insurance and Financial Services
• Payments of households occupying a room in a hotel or
boarding house as their main residence (04.1.1.0) Division 12 covers insurance and financial services. Insur-
ance and financial services are provided by financial cor-
• Rentals paid by households for a secondary residence or
porations. These may be paid for explicitly or implicitly.
time-share residence (04.1.2)
Some transactions in financial assets may also involve both
• Telephone calls (08.3.1.0) explicit and implicit charges. Implicit charges are not always
• Package holidays (09.8.0.0) as evident as the way in which charges are made for most
• Breakfast, meals, and other food and beverages provided goods and services and are typically calculated using a num-
in accommodation and similar establishments, when sep- ber of observable transactions. Examples of implicit charges
arately invoiced (11.1.1) arise from the financial services provided in association
with interest charges on loans and deposits, the acquisition
11.2.0.2 Holiday Centers, Camping Sites, Youth Hostels, and disposal of financial assets and liabilities in financial
and Similar Accommodation Services (S) markets, and insurance and pension programs. The imputed
values of expenditure on insurance and financial services are
Includes:
not directly observable from household income and expen-
• Holiday villages and holiday centers diture surveys.
• Camping sites, onsite (nonmobile) caravans and boats, Insurance services are subdivided by type of insurance.
camping grounds, recreational vehicle, and trailer parks Financial services are subdivided by financial intermedia-
• Youth hostels and mountain refuges
tion services indirectly measured and by the other forms of
actual and indirect charges and remittance fees for finan-
• Bungalows, chalets, housekeeping cottages, and cabins cial services. For the definition and the measurement of the
Excludes: insurance and financial services, specific national accounts
concepts (2008 SNA) do apply.
• Payments of households occupying a room in holi-
day centers, camping sites, youth hostels, and similar 12.1 Insurance
accommodation as their main or secondary residence Service charges for insurance are classified by type
(04.1.1.0) of insurance, namely, life insurance and nonlife insur-
• Telephone calls (08.3.1.0) ance (that is, insurance in connection with the dwelling,
• Breakfast, meals, and other food and beverages provided health, transport, and so on). Service charges for multi-
in accommodation and similar establishments, when sep- risk insurance covering several risks should be classi-
arately invoiced (11.1.1) fied on the basis of the cost of the principal risk if it is
419
APPENDIX
not possible to allocate the service charges to the various 12.1.3 Insurance Connected with the Dwelling (S)
risks covered. Includes:
The basic method for measuring the service charge for
• Service charges paid by owner-occupiers and by tenants
nonlife insurance is total premiums earned plus investment
for the kinds of insurance typically taken out by tenants
income earned from investing the premiums less adjusted
against fire, theft, water damage, and so on
claims incurred.
The basic method for measuring the service charge for life • Service charges for household contents insurance
insurance is premiums earned plus investment income earned Excludes:
from investing the premiums less benefits due less increases
(plus decreases) in life insurance technical reserves. • Service charges paid by owner-occupiers for the kinds of
insurance typically taken out by landlords (intermediate
12.1.1 Life and Accident Insurance (S) consumption)
Includes:
12.1.3.0 Insurance Connected with the Dwelling (S)
• Service charges for life insurance, annuities, death benefit Includes:
insurance, education insurance, and so on
12.1.1.0 Life and Accident Insurance (S) for the kinds of insurance typically taken out by tenants
Includes: against fire, theft, water damage, and so on
• Service charges for life insurance, annuities, death benefit
for household contents insurance
insurance, education insurance, and so on
Excludes:
Includes also:
• Service charges paid by owner-occupiers for the kinds of
• Service charges for funeral insurance
insurance typically taken out by landlords (intermediate
• Service charges for accident insurance consumption)
12.1.2 Insurance Connected with Health (S) 12.1.4 Insurance Connected with Transport (S)
Includes: Includes:
• Service charges for direct sickness insurance • Service charges for insurance in respect of personal trans-
• Service charges for social health insurance port equipment
• Service charges for disability insurance • Service charges for travel insurance (for change of travel
• Service charges for critical illness insurance plans, travel cancellation, and so on) and luggage insurance
• Service charges for long-term care insurance 12.1.4.1 Personal Transport Insurance (S)
• Service charges for supplementary health insurance Includes:
• Service charges for supplementary prescription drug
• Service charges for car insurance
insurance
• Service charges for insurance in respect of personal trans-
• Service charges for travel health insurance
port equipment
Includes also: • Service charges for insurances covering civil liability or
• Service charges for medical transport (for example, damage to third parties or their property arising from the
ambulance) insurance operation of personal transport equipment
Includes also:
12.1.2.0 Insurance Connected with Health (S)
Includes: • Service charges for insurance in respect of bikes, motor-
cycles, boats, yachts, sailboats, airplanes, and so on
• Service charges for direct sickness insurance
• Service charges for social health insurance 12.1.4.2 Travel Insurance (S)
• Service charges for disability insurance Includes:
• Service charges for critical illness insurance • Service charges for travel insurance (for change of travel
• Service charges for long-term care insurance plans, travel cancellation, and so on) and luggage insurance
• Service charges for supplementary health insurance Excludes:
• Service charges for supplementary prescription drug
insurance • Travel health insurance (12.1.2.0)
• Service charges for travel health insurance
12.1.9 Other Insurance (S)
• Service charges for medical transport (for example, • Service charges for other insurance, such as civil liability
ambulance) insurance for injury or damage to third parties or their property
420
APPENDIX
• Service charges for standardized guarantees 12.2.2 Explicit Charges by Deposit-Taking Corporations (S)
• Service charges for legal insurance Includes:
• Service charges for pet medical insurance • Explicit charges for the financial services of deposit-
taking corporations, such as commercial banks, credit
Excludes:
unions, cooperative banks, savings banks, post banks, and
• Civil liability or damage to third parties or their property postal savings banks
arising from the operation of personal transport equip-
Excludes:
ment (12.1.4.1)
• Remittances fees (12.2.9.1)
12.1.9.0 Other Insurance (S)
Includes: 12.2.2.0 Explicit Charges by Deposit-Taking
Corporations (S)
• Service charges for other insurance, such as civil liability
Includes:
for injury or damage to third parties or their property
• Service charges for standardized guarantees • Actual charges for the financial services of deposit-taking
• Service charges for legal insurance
corporations, such as commercial banks, credit unions,
cooperative banks, savings banks, post banks, and postal
• Service charges for pet medical insurance savings banks
Excludes: Excludes:
• Civil liability or damage to third parties or their property • Remittances fees (12.2.9.1)
arising from the operation of personal transport equip-
ment (12.1.4.1) 12.2.9 Other Financial Services N.E.C. (S)
Includes:
12.2 Financial Services
Financial services are subdivided by financial intermedia- • Actual and implicit charges for the financial services of
tion services indirectly measured and by the other forms of money changers and other financial institutions
actual and implicit charges and remittance fees for financial • Fees and service charges and implicit charges of money
services. market funds, nonmoney market investment funds, bro-
kers, and the like
12.2.1 Financial Intermediation Services Indirectly
• Administrative charges and implicit charges of pension
Measured (S)
Financial intermediation services indirectly measured is cal- funds and the like
culated as the difference between the actual bank interest • Remittance fees
receivable and the SNA interest receivable by financial cor-
porations on the loans they issue, and the difference between 12.2.9.1 Remittances Fees (S)
SNA interest payable and the actual bank interest payable Includes:
by financial corporations on the deposits of their customers. • Remittances fees for financial services provided by finan-
Includes: cial auxiliaries, such as money transfer agents and the like
• Financial intermediation services indirectly measured of
deposit-taking corporations 12.2.9.9 Other Financial Services N.E.C. (S)
Includes:
Includes also:
• Actual and implicit charges for the financial services of
• Implicit financial services provided by moneylenders,
money changers and other financial institutions
credit card issuers, finance associates of retailers who
may be responsible for providing loans, pawnshops and • Fees and service charges and implicit charges of money
corporations involved in lending (for example, in provid- market funds, nonmoney market investment funds, bro-
ing student loans and import/export loans) kers, and the like
• Administrative and implicit charges of pension funds and
12.2.1.0 Financial Intermediation Services Indirectly the like
Measured (S) • Other financial services n.e.c.
Includes:
• Financial intermediation services indirectly measured of 13 Personal Care, Social Protection, and
deposit-taking corporations Miscellaneous Goods
Division 13 covers goods and services for personal care, jewelry
Includes also:
and watches, services of social protection, and all other services
• Implicit financial services provided by moneylenders, for households, which are not classified elsewhere. Goods and
credit card issuers, finance associates of retailers who services for personal care cover electric and nonelectric appli-
may be responsible for providing loans, pawnshops and ances for personal care as well as hairdressing services. Goods
corporations involved in lending (for example, in provid- of personal effects cover jewelry and watches, celebration and
ing student loans and import/export loans) devotional articles, and travel goods and articles.
421
APPENDIX
Social protection services cover childcare, nonmedical Excludes:

retirement homes for elderly persons and disabled persons,
• Handkerchiefs made of fabric (03.1.3.1)
services to maintain people in their private homes, and
related service. Group 13.9 covers all the other services for 13.1.2.0 Other Appliances, Articles, and Products for
households, which are not classified elsewhere, such as fees Personal Care (ND)
for legal and administrative services, fees for real estate
Includes:
agencies, charges for undertaking, and payment for various
personal services (for example, graphologists, bodyguards, • Nonelectric appliances: shavers, razors and hair trim-
matrimonial agencies, and so on). mers and blades therefor, scissors, nail files, combs,
This group also includes religious services. shaving brushes, hairbrushes, toothbrushes, nail brushes,
hairpins, curlers, personal weighing machines, scales,
13.1 Personal Care and so on
13.1.1 Electric Appliances for Personal Care (SD) • Articles for personal hygiene: toilet soap, medicinal soap,
Includes: cleansing oil and milk, shaving soap, shaving cream and
foam, toothpaste, epilation wax, paper handkerchiefs, and
• Electric razors, hair trimmers and epilators, hand-held so on
and hood hairdryers, straightening irons, curling tongs
• Beauty products: lipstick, nail varnish, makeup and
and styling combs, sunlamps, vibrators, electric tooth-
makeup removal products (including powder compacts,
brushes, other electric appliances for dental hygiene, and
brushes, and powder puffs), hair lacquers and lotions, pre-
so on
shave and after-shave products, sunbathing products and
• Repair of such appliances sunscreens, hair removers, perfumes and toilet waters,
personal deodorants, bath products, and so on
13.1.1.1 Electric Appliances for Personal Care (SD)
Excludes:
Includes:
• Handkerchiefs made of fabric (03.1.3.1)
• Electric razors, hair trimmers and epilators, hand-held
and hood hairdryers, straightening irons, curling tongs 13.1.3 Hairdressing Salons and Personal Grooming
and styling combs, sunlamps, vibrators, electric tooth- Establishments (S)
brushes and other electric appliances for dental hygiene, Includes:
and so on
• Services of hairdressing salons, barbers, beauty shops,
Excludes: manicures, pedicures, Turkish baths, saunas, solariums,
• Repair of electric appliances for personal care (13.1.1.2) nonmedical massages, and so on
• Body care, depilation, and the like, diet clubs, tattoo, and
13.1.1.2 Repair of Electric Appliances for Personal piercing services
Care (S) • Cosmetic surgery for other purposes than reconstructive
The cost of materials is included only if the materials are not surgery
Includes: Excludes:
• Repair of electric appliances for personal care • Spas for medical purpose (06.2, 06.3)
• Animals tattooing (09.4.5.0)
13.1.2 Other Appliances, Articles, and Products for • Fitness centers (09.4.6.2)
Personal Care (ND)
Includes: 13.1.3.1 Hairdressing (S)
• Nonelectric appliances: shavers, razors and hair trim- Includes:
mers and blades therefor, scissors, nail files, combs, • Services of hairdressing salons or barbers for women,
shaving brushes, hairbrushes, toothbrushes, nail men, and children
brushes, hairpins, curlers, personal weighing machines,
scales, and so on 13.1.3.2 Personal Grooming Treatments (S)
• Articles for personal hygiene: toilet soap, medicinal soap, Includes:
cleansing oil and milk, shaving soap, shaving cream and
• Facial beauty treatments, depilation, solarium, pedicure,
foam, toothpaste, epilation wax, paper handkerchiefs, and
body care, manicure, thalassotherapy, Turkish baths, sau-
so on
nas, nonmedical massages, and so on
• Beauty products: lipstick, nail varnish, makeup and
• Diet clubs, tattoo, and piercing services
makeup removal products (including powder com-
pacts, brushes, and powder puffs), hair lacquers and Excludes:
lotions, pre-shave and after-shave products, sunbathing
• Spas for medical purpose (06.2, 06.3)
products and sunscreens, hair removers, perfumes and
toilet waters, personal deodorants, bath products, and • Animals tattooing (09.4.5.0)
so on • Fitness centers (09.4.6.2)
422
APPENDIX
13.2 Personal Effects N.E.C. 13.2.2.0 Devotional Articles and Articles for Religious
and Ritual Celebrations (SD)
13.2.1 Jewelry and Watches (D)
Includes: Includes:
• Precious stones and metals and jewelry fashioned out of • Religious and ritual articles like crucifixes and rosaries,
such stones and metals figurines, pictures, votive candles, amulets, strips of paper
with prayers, menorah chandeliers, advent wreaths, and
• Costume jewelry, cuff links, and tiepins others
• Watches, stopwatches • Articles to be used in religious and ritual celebrations
• Repair, remodeling, and hire of jewelry and watches
Excludes:
Excludes:
• Christmas trees, holiday decorations (for Christmas, Eas-
• Wall clocks, alarm clocks, travel clocks (05.1.1.4) ter, Hanukkah, Eid, Diwali, and similar) (09.2.1.3)
• Ornaments (05.1.1.4, 05.4.0.1) • Religious books (09.7.1.9)
• Radio clocks (08.1.4.0)
• Smartwatches (08.1.9.1) 13.2.9 Other Personal Effects (SD)
Includes:
• Precious stones and metals and jewelry fashioned out of
such stones and metals acquired primarily as stores of • Travel goods and other carriers of personal effects: suit-
value (capital formation) cases, trunks, travel bags, attaché cases, satchels, hand-
bags, wallets, purses, and so on
13.2.1.1 Jewelry and Watches (D) • Articles for babies: baby carriages, pushchairs, carry-
Includes: cots, recliners, back carriers, front carriers, harnesses for
• Precious stones and metals and jewelry fashioned out of babies, and so on
such stones and metals • Articles for smokers: pipes, lighters, cigarette cases, cigar
• Costume jewelry, cuff links, and tiepins cutters, ashtrays, electronic cigarettes devices, and so on
• Watches, stopwatches • Miscellaneous personal articles: sunglasses, protective
glasses, walking sticks and canes, umbrellas and parasols,
Excludes: fans, keyrings, pill organizers, earplugs, and so on
• Wall clocks, alarm clocks, travel clocks (05.1.1.4) • Funerary articles: coffins, gravestones, urns, and so on
• Ornaments (05.1.1.4, 05.4.0.1) • Lighter fuel; wall thermometers and barometers
• Radio clocks (08.1.4.0) • Repair and hire of other personal effects
• Smartwatches (08.1.9.1) Excludes:
• Precious stones and metals and jewelry fashioned out of • Baby furniture (05.1.1.1)
such stones and metals acquired primarily as stores of
value (capital formation) • Shopping bags (05.2.1.9)
• Feeding bottles (05.4.0.3)
13.2.1.2 Repair and Hire of Jewelry, Clocks, and • Walking sticks and canes used for medical reasons
Watches (S) (06.1.3.3)
The cost of materials is included only if the materials are not • Car seats for babies (07.2.1.3)
Includes: 13.2.9.1 Travel Goods and Articles for Babies and Other
Personal Effects N.E.C. (SD)
• Repair of jewelry, clocks, and watches
Includes:
• Remodeling of jewelry
• Hire of jewelry, clocks, and watches • Travel goods and other carriers of personal effects: suit-
cases, trunks, travel bags, attaché cases, satchels, hand-
13.2.2 Devotional Articles and Articles for Religious and bags, wallets, purses, and so on
Ritual Celebrations (SD) • Articles for babies: baby carriages, pushchairs, carry-
Includes: cots, recliners, back carriers, front carriers, harnesses for
babies, and so on
• Religious and ritual articles like crucifixes and rosaries, figu-
rines, pictures, votive candles, amulets, strips of paper with • Articles for smokers: pipes, lighters, cigarette cases,
prayers, menorah chandeliers, advent wreaths, and others cigar cutters, ashtrays, electronic cigarettes devices,
and so on
• Articles to be used in religious celebrations and rituals
• Miscellaneous personal articles: sunglasses, protective
Excludes: glasses, walking sticks and canes, umbrellas and parasols,
• Christmas trees, holiday decorations (for Christmas, Eas- fans, keyrings, pill organizers, earplugs, and so on
ter, Hanukkah, Eid, Diwali, and similar) (09.2.1.3) • Funerary articles: coffins, gravestones, urns, and so on
• Religious books (09.7.1.9) • Lighter fuel, wall thermometers, and barometers
423
APPENDIX
Excludes: chronic patient (for example, those with dementia); and

services of alcoholism or drug addiction rehabilitation
• Baby furniture (05.1.1.1)
facilities (other than licensed hospitals).
• Shopping bags (05.2.1.9)
• Services of mental health convalescent homes or hospi-
• Feeding bottles (05.4.0.3) tals (06.3.2.0).
• Walking sticks and canes used for medical reasons • Babysitters and so on (05.6.2.1).
(06.1.3.3)
• Educational kindergartens (10.1.0.1).
• Car seats for babies (07.2.1.3)
13.3.0.1 Childcare Services (S)
13.2.9.2 Repair or Hire of Other Personal Effects (S) Includes:
• Child-minding outside the home
Includes: • Nurseries, day care facilities, wet-nurses, crèches, and
other child-minding facilities for babies
• Repair of other personal effects
• After school centers
• Hire of other personal effects
Excludes:
13.3 Social Protection • Babysitters and so on (05.6.2.1)
Social protection as defined here covers nonmedical assis- • Educational kindergartens (10.1.0.1)
tance and support services provided to persons who are
elderly, disabled, having occupational injuries and diseases,
survivors, unemployed, destitute, homeless, low-income 13.3.0.2 Nonmedical Retirement Homes for Elderly
earners, indigenous people, immigrants, refugees, alcohol Persons and Residences for Disabled Persons (S)
and substance abusers, and so on. Includes:
It also covers assistance and support services provided to • Nonmedical retirement homes for elderly persons
families and children.
• Nonmedical residences for disabled persons
13.3.0 Social Protection (S) Excludes:
Social protection services include residential care, home
• Social assistance and other social care services that are
help, and day care. More specifically, this class covers pay-
integrated into a package of care along with medical
ments by households for:
services are to be included in 06.2.3 if medical services
• Nonmedical retirement homes for elderly persons, non- do not require an overnight stay and 06.3.2 if they do.
medical residences for disabled persons, rehabilitation For example, services of medical convalescent homes or
centers providing nonmedical long-term support for indi- convalescent hospitals; services of homes for the elderly
viduals rather than health care and rehabilitative therapy, with nursing care; inpatient care hospices; services of
schools for disabled persons where the main aim is to help palliative care establishments for the terminally ill; ser-
students overcome their disability vices of nursing homes; rest homes with nursing care;
• Nonmedical help to maintain elderly and disabled persons services of skilled nursing facilities; services of teaching
at home (home-cleaning services, meal programs, non- nursing homes; services of residential mental retardation
medical day care centers, day care services, and holiday facilities; mental health and substance abuse facilities for
care services) chronic patient (for example, those with dementia); and
• Child-minding outside the home, nurseries, day care facil- services of alcoholism or drug addiction rehabilitation
ities wet-nurses, crèches, kindergartens (other than educa- facilities (other than licensed hospitals).
tional), play schools, and other child-minding facilities • Services of mental health convalescent homes or hospi-
• Counseling, guidance, arbitration, fostering, and adoption tals (06.3.2.0).
services for families 13.3.0.3 Services to Maintain People in Their Private
Excludes: Homes (S)
• Social assistance and other social care services that are Includes:
integrated into a package of care along with medical • Help to maintain elderly and disabled persons at home
services are to be included in 06.2.3 if medical services (home-cleaning services, meal programs, day care cen-
do not require an overnight stay and 06.3.2 if they do. ters, day care services, and holiday care services)
For example, services of medical convalescent homes or
convalescent hospitals; services of homes for the elderly 13.3.0.9 Other Social Protection Services (S)
with nursing care; inpatient care hospices; services of Includes:
palliative care establishments for the terminally ill; ser-
vices of nursing homes; rest homes with nursing care; • Schools for disabled persons where the main aim is to
services of skilled nursing facilities; services of teaching help students overcome their disability
nursing homes; services of residential mental retardation • Guidance, arbitration, fostering, and adoption services for
facilities; mental health and substance abuse facilities for families
424
APPENDIX
13.9 Other Services N.E.C. Excludes:

13.9.0 Other Services N.E.C. (S) • Food or beverages consumed in brothels if charged sepa-
Includes: rately (11.1.1, 11.1.2)
• Fees for legal services, employment agencies, and so on
13.9.0.2 Religious Services (S)
• Payment for the services of lawyers, notaries, accoun-
Includes:
tants, and so on
• Charges for undertaking and other funeral services • Expenditures for religious services, for example, requi-
• Payment for the services of estate agents, housing agents,
ems, baptizing, marriage services
auctioneers, salesroom operators, and other intermediaries
13.9.0.9 Other Services N.E.C. (S)
• Payment for photocopies and other reproductions of
Includes:
documents
• Fees for the issue of birth, marriage, and death certifi- • Fees for legal services, employment agencies, and
cates, and other administrative documents so on
• Expenditures for religious services, for example, requi- • Payment for the services of lawyers, accountants, and
ems, baptizing, marriage services so on
• Expenditure for nonreligious services and events, such as • Expenditure for nonreligious services and events, such as
coming of age celebrations in Latin American “Quince” coming of age celebrations in Latin American “quince” or
or debutante balls debutante balls
• Payment for newspaper notices and advertisements • Charges for undertaking and other funeral services
• Payment for the services of graphologists, astrologers, palm- • Payment for the services of estate agents, housing
ists, private detectives, bodyguards, matrimonial agencies agents, auctioneers, salesroom operators, and other
and marriage guidance counselors, public writers, miscella- intermediaries
neous concessions (seats, toilets, cloakrooms), and so on • Payment for photocopies, printing services, and other
• Services provided by prostitutes and the like reproductions of documents
• Firearms licenses • Fees for the issue of birth, marriage, and death certifi-
cates, and other administrative documents
Excludes:
• Payment for newspaper notices and advertisements
• Food or beverages consumed in brothels if charged sepa- • Payment for the services of graphologists, astrologers,
rately (11.1.1, 11.1.2) palmists, private detectives, bodyguards, matrimonial
agencies and marriage guidance counselors, public writ-
13.9.0.1 Prostitution (S)
ers, miscellaneous concessions (seats, toilets, cloak-
Includes: rooms), and so on
• Services provided by prostitutes or sex workers and the like • Firearms licenses
425
APPENDIX
Appendix 4 The Nature and Meaning of a

Consumer Price Index
Resolution Concerning
Consumer Price Indices 1. The CPI is a current social and economic indicator that
is constructed to measure changes over time in the gen-
Adopted by the Seventeenth eral level of prices of consumer goods and services that
households acquire, use, or pay for consumption.
International Conference of 2. The index aims to measure the change in consumer
Labour Statisticians, 2003 prices over time. This may be done by measuring the
cost of purchasing a fixed basket of consumer goods and
services of constant quality and similar characteristics,
with the products in the basket being selected to be rep-
Preamble1 resentative of households’ expenditure during a year or
other specified period. Such an index is called a fixed-
The Seventeenth International Conference of Labour basket price index.
Statisticians 3. The index may also aim to measure the effects of price
Having been convened at Geneva by the Governing Body changes on the cost of achieving a constant standard of
of the International Labour Organization and having living (that is level of utility or welfare). This concept
met from 24 November to 3 December 2003 is called a cost of living index (COLI). A fixed-basket
Recalling the resolution adopted by the Fourteenth Inter- price index, or another appropriate design, may be em-
national Conference of Labour Statisticians concerning ployed as an approximation to a COLI.
consumer price indices and recognizing the continuing
validity of the basic principles recommended therein The Uses of a Consumer Price
and, in particular, the fact that the consumer price index Index
(CPI) is designed primarily to measure the changes
over time in the general level of prices of goods and 4. The CPI is used for a wide variety of purposes, the two
services that a reference population acquires, uses, or most common ones being (1) to adjust wages as well as
pays for social security and other benefits to compensate, partly
Recognizing the need to modify and broaden the exist- or completely, for changes in the cost of living or in
ing standards in the light of recent methodological and consumer prices; and (2) to provide an average measure
computational developments to enhance the usefulness of price inflation for the household sector as a whole,
of the international standards in the provision of techni- for use as a macroeconomic indicator. CPI subindices
cal guidelines to all countries are also used to deflate components of household final
Recognizing the usefulness of such standards in enhanc- consumption expenditure in the national accounts and
ing the international comparability of the statistics the value of retail sales to obtain estimates of changes in
Recognizing that the CPI is used for a wide variety of their volume.
purposes and that governments should be encouraged 5. CPIs are also used for other purposes, such as moni-
to identify the (priority) purposes a CPI is to serve, toring the overall rate of price inflation for all sectors
to provide adequate resources for its compilation, of the economy, the adjustment of government fees and
and to guarantee the professional independence of its charges, the adjustment of payments in commercial
compilers contracts, and for formulating and assessing fiscal and
Recognizing that the (priority) objectives and uses of a monetary policies and trade and exchange rate policies.
CPI differ among countries and that, therefore, a single In these types of cases, the CPI is used as more appro-
standard could not be applied universally priate measures do not exist at present, or because other
Recognizing that the CPI needs to be credible to observ- characteristics of the CPI (for example, high profile,
ers and users, both national and international, and that wide acceptance, predictable publication schedule, and
better understanding of the principles and procedures so on) are seen to outweigh any conceptual or technical
used to compile the index will enhance the users’ con- deficiencies.
fidence in the index
Agrees that the principles and methods used in construct- 6. Given that the CPI may be used for many purposes, it
ing a CPI should be based on the guidelines and meth- is unlikely that one index can perform equally satisfac-
ods that are generally accepted as constituting good torily in all applications. It may therefore be appropri-
statistical practices ate to construct a number of alternative price indices for
Adopts, this third day of December 2003, the following specific purposes, if the requirements of the users justify
resolution which replaces the previous one adopted in the extra expense. Each index should be properly de-
1987 fined and named to avoid confusion and a “headline.”
CPI measure should be explicitly identified.
7. Where only one index is compiled, it is the main use that
should determine the type of index compiled, the range
of goods and services covered, its geographic coverage,
1
All annexes referred to in the footnotes are the annexes to the resolution. the households it relates to, as well as to the concept of
427
APPENDIX
price and the formula used. If there are several major payments of income taxes, social security contributions,
uses, it is likely that compromises may have to be made and fines are not considered to be consumer goods or
with regard to how the CPI is constructed. Users should services and should be excluded from the coverage of
be informed of the compromises made and of the limita- the index. Some countries regard expenditures on the
tions of such an index. purchase of houses entirely as a capital investment and,
as such, exclude them from the index.
Scope of the Index Acquisition, Use, or Payment

8. The scope of the index depends on the main use for
which it is intended and should be defined with regard 15. In determining the scope of the index, the time of re-
to the type of households, geographic areas, and the cat- cording, and valuation of consumption, it is important
egories of consumer goods and services acquired, used, to consider whether the purposes for which the index
or paid for by the reference population. is used are best satisfied by defining consumption with
regard to “acquisition,” “use,” or “payment.”2 The “ac-
9. If the primary use of the CPI is for adjusting money quisition” approach is often used when the primary
incomes, a relevant group of households, such as wage purpose of the index is to serve as a macroeconomic
and salary earners, may be the appropriate target popula- indicator. The “payment” approach is often used when
tion. For this use, all consumption expenditures by these the primary purpose of the index is for the adjustment
households, at home and abroad, may be covered. If the of compensation or income. Where the aim of the index
primary use of the CPI is to measure inflation in the is to measure changes in the cost of living, the “use”
domestic economy, it may be appropriate to cover con- approach may be most suitable. The decision regarding
sumption expenditures made within the country, rather the approach to follow for a particular group of prod-
than the expenditures of households resident within the ucts should in principle be based on the purpose of the
country. index, as well as on the costs and the acceptability of the
10. In general, the reference population for a national index decision to the users who should be informed of the ap-
should be defined very widely. If any income groups, proach followed for different products. Because of the
types of households or particular geographic areas are practical difficulties in uniformly defining consumption
excluded, for example, for cost or practical consider- and estimating the flow of services provided by other
ations, then this should be explicitly stated. durable goods with regard to “use,” it may be neces-
11. The geographic scope refers to the geographic cover- sary to adopt a mixed approach, for example, “use” for
age of price collection and of consumption expenditures owner-occupied housing and “acquisition” or “pay-
of the reference population and both should be defined ment” basis for other consumer durables.
as widely as possible, and preferably consistently. If 16. The differences between the three approaches are most
price collection is restricted to particular areas due to pronounced in dealing with products for which the
resource constraints, then this should be specified. The times of acquisition, use, and payment do not coincide,
geographic coverage of the consumption expenditure such as owner-occupied housing, durable goods, and
may be defined either as covering consumption expen- products acquired on credit.
diture of the resident population (resident consumption) 17. The most complex and important of the products men-
or consumption expenditure within the country (domes- tioned previously is owner-occupied housing. In most
tic consumption). countries, a significant proportion of households are
12. Significant differences in the expenditure patterns and owner-occupiers of their housing, with the housing
price movements between specific population groups being characterized by a long useful life and a high pur-
or regions may exist, and care should be taken to en- chase outlay (price). Under the “acquisition” approach,
sure that they are represented in the index. Separate the value of the new dwellings acquired in the weight
indices for these population groups or regions may be reference period may be used for deriving the weight
computed if there is sufficient demand to justify the ad- (and the full price of the dwelling is included in the
ditional cost. CPI at the time of acquisition, regardless of when the
13. In accordance with its main purpose, the CPI should consumption is taking place). Under the “payment” ap-
conceptually cover all types of consumer goods and proach, the weights reflect the amounts actually paid out
services of significance to the reference population, for housing (and the prices enter the CPI in the period(s)
without any omission of those that may not be legally when the prices are paid). Under the “use” approach the
available or may be considered socially undesirable. weights are based on the value of the flow of housing
Where appropriate, special aggregates may be con- services consumed during the weight reference period
structed to assist those users who may wish to exclude estimated using an implicit or notional cost (and prices
certain categories of goods or services for particular ap- or estimated opportunity costs enter the CPI when the
plications or for analysis. Whenever certain goods or consumption is taking place).
services have been excluded from the index, this should 18. Own-account consumption, remuneration in kind, and
be clearly documented. goods and services provided without charge or subsi-
14. Goods and services purchased for business purposes,
expenditures on assets such as works of art, financial
investment (as distinct from financial services), and 2
See Annex 1.
428
APPENDIX
dized by governments and nonprofit institutions serv- expenditure pattern. Sources of such information that
ing households may be important in some countries can be used for disaggregating the expenditures are sur-
where the purpose of the index is best satisfied by defin- veys of sales in retail outlets, point-of-purchase surveys,
ing consumption with regard to “use” or “acquisition” surveys of production, export and import data, and ad-
(under the payment approach these are out of scope). ministrative sources. Based on these data the weights
The inclusion of these products will require special val- for certain products may be further disaggregated by
uation and pricing techniques. region and type of outlet. Where the data obtained from
different sources relate to different periods, it is impor-
Basket and Weights tant to ensure, before weights are allocated, that expen-
ditures are adjusted so that they have the same reference
19. Decisions on the composition of the basket and the period.
weights follow directly from the scope, as well as from 25. Where the weight reference period differs significantly
the choice between the “acquisition,” “use,” or “pay- from the price reference period, the weights should be
ment” approaches. price updated to take account of price changes between
20. Once defined, the expenditures that fall within the scope the weight reference period and price reference period.
of the index should be grouped into similar categories Where it is likely that price-updated weights are less
in a hierarchical classification system, for example, di- representative of the consumption pattern in the price
visions/groups/classes, for compilation as well as ana- reference period this procedure may be omitted.
lytical purposes. There should be consistency between 26. Weights should be reviewed and if appropriate revised
the classification used for index compilation and the one as often as accurate and reliable data are available for
used for household expenditure statistics. The CPI clas- this to be done, but at least once every five years. Revi-
sification should meet the needs of users for special sub- sions are important to reduce the impact on the index of
indices. For the purposes of international comparisons, product substitutions and to ensure the basket of goods
the classification should also be reconcilable with the and services and their weights remain representative.4
most recent version of the United Nations Classifica- For some categories, it may be necessary to update
tion of Individual Consumption According to Purpose, the weights more frequently as such weights are likely
at least at its division level.3 to become out of date more quickly than higher-level
21. In order to facilitate the analysis and interpretation of weights. In periods of high inflation, the weights should
the results of the index, it may be desirable to classify be updated frequently.
goods and services according to various supplementary 27. When a new basket (structure or weights) replaces
classifications, for example, source of origin, durability, the old, a continuous CPI series should be created by
and seasonality. Calculation of the CPI by using various linking5 together the index numbers based on the new
classifications should generate the same overall results basket of goods and services to those based on the ear-
as the original index. lier basket. The particular procedure used to link index
22. The classification should also provide a framework for number series will depend on the particular index com-
the allocation of expenditure weights. Expenditures at pilation technique used. The objective is to ensure that
the lowest level of the classification system, expressed the technique used to introduce a new basket does not,
as a proportion of the total expenditure, determine the of itself, alter the level of the index.
weights to be used at this level. When the weights are to 28. Completely new types of goods and services (that is
remain fixed for several years, the objective should be goods and services that cannot be classified to any of
to adopt weights that are representative of the contem- the existing elementary aggregates) should normally be
porary household behavior. considered for inclusion only during one of the periodic
23. The two main sources for deriving the weights are the review and reweighting exercises. A new model or va-
results from household expenditure surveys and na- riety of an existing product that can be fitted within an
tional accounts estimates on household consumption existing elementary aggregate should be included at the
expenditure. The results from a household expenditure time it is assessed as having a significant and sustain-
survey are appropriate for an index defined to cover able market share. If a quality change is detected an ap-
the consumption expenditures of reference population propriate quality adjustment should be made.6
groups resident within the country, while national ac- 29. Some products such as seasonal products, insurance,
count estimates are suitable for an index defined to second-hand goods, expenditure abroad, interest, own
cover consumption expenditures within the country. production, expenditures on purchase and construction
The decision about what source or sources to use and of dwellings, and so on, may need special treatment
how they should be used depends on the main purpose when constructing their weights. The way these prod-
of the index and on the availability and quality of ap- ucts are dealt with should be determined by the main
propriate data. purpose of the index, national circumstances, and the
24. The information from the main source (household ex- practicalities of compilation.
penditure surveys or national accounts) should be sup-
plemented with all other available information on the
4
See Annex 1.
5
See Annex 2.
3 6
See Annex 4. See Annex 2.
429
APPENDIX
30. Seasonal products should be included in the basket. It is A mixture of probability and nonprobability sampling
possible to use (1) a fixed-weight approach which uses techniques may be used.
the same weight for the seasonal product in all months 37. Efficient and representative sampling, whether random
using an imputed price in the out-of-season months; or purposive, requires comprehensive and up-to-date
or (2) a variable weights approach where a changing sampling frames for outlets and products. Sample se-
weight is attached to the product in various months. The lection can be done by the head office from centrally
decision on the approach should be based on national held sampling frames, in the field by price collectors,
circumstances. or by a mixture of the two. In the first case, price col-
31. The expenditure weights for second-hand goods should lectors should be given precise instructions on which
be based either on the net expenditure of the reference outlets to visit and which products to price. In the
population on such goods or the gross expenditure, de- second case, price collectors should be given detailed
pending on the purpose of the index. and unambiguous guidelines on the local sampling
32. When consumption from own production is within the procedures to be adopted. Statistical business regis-
scope of the index, the weights should be based on the ters, business telephone directories, results from the
value of quantities consumed from own production. point-of-purchase surveys or from surveys of sales in
Valuation of consumption from own production should different types of outlets, and lists of internet sellers
be made on the basis of prices prevailing on the mar- may be used as sampling frames for the central selec-
ket, unless there is some reason to conclude that market tion of outlets. Catalogs or other product lists drawn up
prices are not relevant or cannot be reliably observed, by major manufacturers, wholesalers, or trade associa-
or there is no interest in using hypothetically imputed tions, or lists of products that are specific to individual
prices. In this case, the expenditures and prices for the outlets such as large supermarkets might be used as the
inputs into the production of these goods and services sampling frame for selection of products. Data scanned
could be used instead. The third option is to valuate it by barcode readers at the cashier’s desk (electronic da-
by using quality-adjusted market prices. tabases) can be particularly helpful in the selection of
goods and services.
38. The sample of outlets and of goods and services should
Sampling for Price Collection be reviewed periodically and updated where necessary
33. A CPI is an estimate based on a sample of households to to maintain its representativeness.
estimate weights, and a sample of zones within regions,
a sample of outlets, a sample of goods and services, and
a sample of time periods for price observation.
Index Calculation
34. The sample size and sample selection methods for 39. The compilation of a CPI consists of collecting and pro-
both outlets and the goods and services for which price cessing price and expenditure data according to speci-
movements over time are to be observed should ensure fied concepts, definitions, methods, and practices. The
that the prices collected are representative and sufficient detailed procedures that are applied will depend on par-
to meet the requirements for the accuracy of the index, ticular circumstances.
but also that the collection process is cost-effective. The 40. CPIs are calculated in steps. In the first step, the elemen-
sample of prices should reflect the importance, with re- tary aggregate indices are calculated. In the subsequent
gard to relative expenditures, of the goods and services steps, higher-level indices are calculated by aggregating
available for purchase by consumers in the reference the elementary aggregate indices.
period, the number, types, and geographic spread of out-
lets that are relevant for each good and service, and the
dispersion of prices and price changes across outlets. Elementary Aggregate Indices
35. Probability sampling techniques are the preferred meth-
ods, in principle, as they permit sound statistical infer- 41. The elementary aggregate is the smallest and rela-
ence and control over the representativity of the sample. tively homogeneous set of goods or services for which
In addition, they permit estimation of sampling varia- expenditure data are defined (used) for CPI purposes.
tion (errors). However, they may be costly to implement It is the only aggregate for which an index number is
and can result in the selection of products that are very constructed without any explicit expenditure weights,
difficult to price to constant quality. although other kinds of weights might be explicitly or
implicitly introduced into the calculation. The set of
36. In cases where appropriate sampling frames are lack- goods or services covered by an elementary aggregate
ing and it is too costly to obtain them, samples of out- should be similar in their end-uses and are expected to
lets and products have to be obtained by nonprobability have similar price movements. They may be defined not
methods. Statisticians should use available information only with regard to their characteristics but also with re-
and apply their best judgment to ensure that represen- gard to the type of location and outlet in which they are
tative samples are selected. The possibility of apply- sold. The degree of homogeneity achieved in practice
ing cutoff or detailed quota sampling7 strategy may be will depend on the availability of corresponding expen-
considered, especially where the sample size is small. diture data.
42. An elementary index is a price index for an ele-
7
See Annex 1. mentary aggregate. As expenditure weights usually
430
APPENDIX
cannot be attached to the prices or price relatives for period t − 1. Failure to do so may result in a biased
the sampled products within the elementary aggre- index.
gate, an elementary index is usually calculated as an
unweighted average of the prices or price relatives.
When some information on weights is available, this Price Observations
should be taken into account when compiling the el-
ementary indices. 48. The number and quality of the prices collected are criti-
cal determinants of the reliability of the index, along
43. There are several ways in which the prices, or the price
with the specifications of the products priced. Standard
relatives, might be averaged. The three most com-
methods for collecting and processing price information
monly used formulae are the ratio of arithmetic mean
should be developed and procedures put in place for
prices, the geometric mean, and the arithmetic mean of
collecting them systematically and accurately at regu-
price relatives. The choice of formula depends on the
lar intervals. Price collectors should be well trained and
purpose of the index, the sample design, and the math-
well supervised, and should be provided with a compre-
ematical properties of the formula. It is possible to use
hensive manual explaining the procedures they have to
different formulae for different elementary aggregates
follow.
within the same CPI. It is recommended that the geo-
metric mean formula be used, particularly where there
is a need to reflect substitution within the elementary Collection
aggregate or where the dispersion in prices or price
changes within the elementary aggregate is large. The 49. An important consideration is whether the index or
geometric mean has many advantages because of its parts of the index should relate to monthly (or quar-
mathematical properties. The ratio of arithmetic mean terly) average prices or to prices for a specific period
prices may be used for elementary aggregates that are of time (for example, a single day or week in a month).
homogeneous and where consumers have only limited This decision is related to a number of issues, which
opportunity to substitute or where substitution is not include the use of an index, the practicalities of carry-
to be reflected in the index. The arithmetic mean of ing out price collection, and the pattern of price move-
price relative formula should be avoided in its chained ments. When point-in-time pricing is adopted, prices
form, as it is known to result in biased estimates of the should be collected over a very small number of days
elementary indices. each month (or quarter). The interval between price ob-
44. The elementary index may be computed by using either servations should be uniform for each product. Since
a chained or direct form of the formula chosen. The use the length of the month (or quarter) varies, this uni-
of a chained form may make the estimation of miss- formity needs to be defined carefully. When the aim is
ing prices and the introduction of replacement products monthly (or quarterly) average prices, the prices col-
easier. lected should be representative of the period to which
they refer.
Upper-Level Indices 50. Attention should also be paid to the time of day selected
for price observation. For example, in the case of per-
45. These price indices are constructed as weighted aver- ishable goods, price observations may need to be col-
ages of elementary aggregate indices. Several types lected at the same time on the same day of the week and
of formulae can be used to average the elementary ag- not just before closing time, when stocks may be low or
gregate indices. In order to compile a timely index, the sold cheaply to minimize wastage.
practical option is to use a formula that relies on the 51. Price collection should be carried out in such a way as
weights relating to some past period. One such formula to be representative of all geographical areas within the
is the Laspeyres-type index, the formula used by most scope of the index. Special care should be taken where
national statistical agencies. significant differences in price movements between
46. For some purposes, it may be appropriate to calculate areas may be expected.
the index retrospectively by using an index number for- 52. Prices should be collected in all types of outlets that are
mula that employs both base period weights and current important, including internet sellers, open-air markets
period weights, such as the Fisher, Törnqvist, or Walsh and informal markets, and in free markets as well as
index. Comparing the difference between the index of price-controlled markets. Where more than one type of
this type and the Laspeyres-type index can give some outlet is important for a particular type of product, this
indication of the combined impact of income changes, should be reflected in the initial sample design and an
preference changes, and substitution effects over the appropriately weighted average should be used in the
period in question, providing important information for calculation of the index.
producers and users of the CPI. 53. Specifications should be provided detailing the vari-
47. Where the change in an upper-level index between two ety and size of the products for which price informa-
consecutive periods such as t − 1 and t is calculated tion is to be collected. These should be precise enough
as the weighted average of the individual indices be- to identify all the price-determining characteristics
tween t − 1 and t, care should be taken to ensure that the that are necessary to ensure that, as far as possible,
weights are updated to take account of the price changes the same goods and services are priced in successive
between the price reference period 0 and the preceding periods in the same outlet. The specifications should
431
APPENDIX
include, for example, make, model, size, terms of consumers are observed, for example, those linked to
payment, conditions of delivery, type of guarantees, the age of the purchaser or to memberships of particular
and type of outlet. This information could be used in associations.
the procedures used for replacement and for quality 59. The collected price information should be reviewed for
adjustment. comparability and consistency with previous observa-
54. Prices to be collected are actual transaction prices, in- tions, the presence of replacements, unusual or large
cluding indirect taxes and nonconditional discounts, price changes, and to ensure that price conversions of
which would be paid, agreed, or costed (accepted) by goods priced in multiple units or varying quantities are
the reference population. Where prices are not dis- properly calculated. Extremely large or unusual price
played or have to be negotiated, where quantity units changes should be examined to determine whether they
are poorly defined or where actual purchase prices may are genuine price changes or are due to changes in qual-
deviate from listed or fixed prices, it may be necessary ity. Procedures should be put in place for checking the
for the price collectors to purchase products in order reliability of all price observations. This could include
to determine the transaction prices. A budget may be a program of direct pricing and selective repricing of
provided for any such purchases. When this is not pos- some products shortly after the initial observation was
sible, consideration may be given to interviewing cus- made.
tomers about the prices actually paid. Tips for services, 60. Consistent procedures should be established for deal-
where compulsory, should be treated as part of the price ing with missing price observations because of, for
paid. example, inability to contact the seller, nonresponse,
55. Exceptional prices charged for stale, shop-soiled, dam- observation rejected as unreliable, or products tem-
aged, or otherwise imperfect goods sold at clearance porarily unavailable. Prices of nonseasonal products
prices should be excluded, unless the sale of such prod- that are temporarily unavailable should be estimated
ucts is a permanent and widespread phenomenon. Sale until they reappear or are replaced, by using appropri-
prices, discounts, cut prices, and special offers should ate estimation procedures, for example, imputation
be included when applicable to all customers without on the basis of price changes of similar nonmissing
there being significant limits to the quantities that can products. Carrying forward the last observed price
be purchased by each customer. should be avoided, especially in periods of high
56. In periods of price control or rationing, where limited inflation.
supplies are available at prices which are held at a low
level by measures such as subsidies to the sellers, gov- Replacements
ernment procurement, price control, and so on, such
prices as well as those charged on any significant unre- 61. Replacement of a product will be necessary when it
stricted markets should be collected. The different price disappears permanently. Replacement should be made
observations should be combined in a way that uses the within the first three months (quarter) of the product
best information available with respect to the actual becoming unavailable. It may also be necessary when
prices paid and the relative importance of the different the product is no longer available or sold in significant
types of sales. quantities or under normal sale conditions. Clear and
57. For each type of product, different alternatives for col- precise rules should be developed for selecting the re-
lecting prices should be carefully investigated, to ensure placement product. Depending on the frequency of sam-
that the price observations could be made reliably and pling and the potential for accurate quality adjustment,
effectively. Means of collection could include visits to the most commonly used alternatives are to select (1)
outlets with paper forms or hand-held devices, inter- the most similar to the replaced variety, (2) the most
views with customers, computer-assisted telephone in- popular variety among those that belong to the same el-
terviews, mail-out questionnaires, brochures, price lists ementary aggregate, and (3) the variety most likely to
provided by large or monopoly suppliers of services, be available in the future. Precise procedures should be
scanner data, and prices posted on the internet. For each laid down for price adjustments with respect to the dif-
alternative, the possible cost advantages need to be bal- ference in characteristics when replacements are neces-
anced against an assessment of the reliability and time- sary, so that the impact of changes in quality is excluded
liness of each of the alternatives. from the observed price.
58. Where centrally regulated or centrally fixed prices are 62. Replacement of an outlet may be motivated if prices
collected from the regulatory authorities, checks should cannot be obtained, for example, because it has closed
be made to ascertain whether the goods and services in permanently, because of a decline in representative-
question are actually sold and whether these prices are ness or because the outlet no longer cooperates. Clear
in fact paid. For goods and services where the prices rules should be established on when to discontinue price
paid are determined by combinations of subscription observations from a selected outlet, on the criteria for
fees and piece rates (for example, for newspapers, jour- selecting a replacement, as well as on the adjustments
nals, public transport, electricity, and telecommunica- that may be required to price observations or weights.
tions) care must be taken to ensure that a representative Such rules should be consistent with the objectives of
range of price offers is observed. Care must also be the index and with the way in which the outlet sample
taken to ensure that prices charged to different types of has been determined.
432
APPENDIX
63. Deletion of an entire elementary aggregate will be nec- of the price difference between the old and new
essary if all products in that elementary aggregate dis- products based on the price changes observed for
appear from most or all outlets and it is not possible similar products. The difference between the esti-
to locate a sufficient number of price observations to mate of pure price change and the observed price
continue to compile a reliable index for this elementary change is considered as change due to quality
aggregate. In such situations, it is necessary to redis- difference.
tribute the weight assigned to the elementary aggregate Some of these methods are complex, costly, and diffi-
among the other elementary aggregates included in the cult to apply. The methods used should as far as possible
next level of aggregation. be based on objective criteria.
Quality Changes Accuracy

64. The same product should be priced in each period as 68. As with all statistics, CPI estimates are subject to er-
long as it is representative. However, in practice, prod- rors that may arise from a variety of sources.9 Compil-
ucts that can be observed at different time periods may ers of CPIs need to be aware of the possible sources of
differ with respect to package sizes, weights, volumes, error and to take steps during the design of the index,
features, and terms of sale, as well as other characteris- its construction, and compilation processes to minimize
tics. Thus, it is necessary to monitor the characteristics their impact, for which adequate resources should be
of the products being priced to ensure that the impact allocated.
of any differences in price-relevant or utility-relevant 69. The following are some well-known sources of poten-
characteristics can be excluded from the estimated price tial error, either in pricing or in index construction, that
change. over time can lead to errors in the overall CPI: incorrect
65. Identifying changes in quality or utility is relatively selection of products and incorrect observation and re-
more difficult for complex durable goods and services. cording of their prices, incorrect selection of outlets and
It is necessary, therefore, to collect a considerable timing of price collection, failure to observe and adjust
amount of information on the relevant characteristics of correctly for quality changes, appearance of new goods
the products for which prices are collected. The most and outlets, failure to adjust for product and outlet sub-
important information can be obtained in the course stitution or loss of representativity, and the use of inap-
of collecting prices. Other sources of information on propriate formulae for computing elementary aggregate
price-relevant or utility-relevant characteristics can be and upper-level indices.
producers, importers, or wholesalers of the goods in- 70. To reduce the index’s potential for giving a misleading
cluded and the study of articles and advertisements in picture, it is in general essential to update weights and
trade publications. baskets regularly, to employ unbiased elementary ag-
66. When a quality change is detected, an adjustment must gregate formulae, to make appropriate adjustments for
be made to the price, so that the index reflects as nearly quality change, to allow adequately and correctly for new
as possible the pure price change. If this is not done, products, and to take proper account of substitution is-
the index will either record a price change that has not sues as well as quality control of the entire compilation
taken place or fail to record a price change that did process.
happen. The choice of method for such adjustments
will depend on the particular goods and services in-
volved. Great care needs to be exercised because the Dissemination
accuracy of the resulting index depends on the qual- 71. The CPI estimate should be computed and publicly re-
ity of this process. To assume automatically that all leased as quickly as possible after the end of the pe-
price change is a reflection of the change in quality riod to which it refers and according to a preannounced
should be avoided, as should the automatic assumption timetable. It should be made available to all users at the
that products with different qualities are essentially same time, in a convenient form, and should be accom-
equivalent. panied by a short methodological explanation. Rules
67. The methods for estimating quality-adjusted prices8 relating to its release should be made publicly available
may be: and strictly observed. In particular, they should include
(1) Explicit (or direct) quality-adjustment methods that details of who has prerelease access to the results, why,
directly estimate the value of the quality difference under what conditions, and how long before the official
between the old and new product and adjust one of release time.
the prices accordingly. Pure price change is then im- 72. The general CPI should be compiled and released
plicitly estimated as the difference in the adjusted monthly. Where there is no strong user demand for a
prices. monthly series or countries do not have the necessary
(2) Implicit (or indirect) quality-adjustment methods resources, the CPI may be prepared and released quar-
which estimate the pure price change component terly. Depending on national circumstances, subindices
8 9
See Annex 2. See Annex 3.
433
APPENDIX
may be released with a frequency that corresponds to of housing (actual rents and either imputed rents or
users’ needs. acquisition of new houses, and maintenance and repair
73. When it is found that published index estimates have of dwellings) and financial services from the all-items
been seriously distorted because of errors or mistakes index will make the resulting estimates of price change
made in their compilation, corrections should be made for the remaining products more comparable across
and published. Such corrections should be made as countries. Therefore, in addition to the all-items index,
soon as possible after detection according to publicly countries should, if possible, compile and provide for
available policy for correction. Where the CPI is widely dissemination to the international community an index
used for adjustment purposes for wages and contracts, that excludes housing and financial services. It should
retrospective revisions should be avoided to the extent be emphasized, though, that even for the remaining
possible. products in scope, there can still be difficulties when
making international comparisons of changes in con-
74. The publication of the CPI results should show the
sumer prices.
index level from the index reference period. It is also
useful to present derived indices, such as the one that
shows changes in the major aggregates between (1) Consultations and Integrity
the current month and the previous month, (2) the
current month and the same month of the previous 81. The compiling agency should have the professional in-
year, and (3) the average of the latest 12 months and dependence, competence, and resources necessary to
the average of the previous 12 months. The indices support a high-quality CPI program. The United Na-
should be presented in both seasonally adjusted and tions Fundamental Principles of Official Statistics11 and
unadjusted terms, if seasonally adjusted data are the International Labour Organization Guidelines Con-
available. cerning Dissemination Practices for Labour Statistics12
should be respected.
75. Comments and interpretation of the index should ac-
company its publication to assist users. An analysis 82. The agency responsible for the index should consult
of the contributions of various products or group of representatives of users on issues of importance for the
products to the overall change and an explanation of CPI, particularly during preparations for any changes to
any unusual factors affecting the price changes of the the methodology used in compiling the CPI. One way of
major contributors to the overall change should be organizing such consultations is through the establish-
included. ment of advisory committee(s) on which social partners,
as well as other users and independent experts, might be
76. Indices for the major expenditure groups should also
represented.
be compiled and released. Consideration should be
given to compiling indices for the divisions and groups 83. In order to ensure public confidence in the index, a
of the Classification of Individual Consumption Ac- full description of the data collection procedures and
cording to Purpose.10 Nine subindices for different the index methodology should be prepared and made
regions or population groups, and alternative indices widely available. Reference to this description should
designed for analytical purposes, may be compiled be made when the CPI is published. The documentation
and publicly released if there is a demand from users, should include an explanation of the main objectives
they are judged to be reliable, and their preparation is of the index, details of the weights, the index number
cost-effective. formulae used, and a discussion of the accuracy of the
index estimates. The precise identities of the outlets and
77. The index reference period may be chosen to coincide
goods and services used for price collection should not
with the latest weight reference period or it could be
be revealed.
established to coincide with the base period of other
statistical series. It should be changed as frequently as 84. Users should be informed in advance of any changes
necessary to ensure that the index numbers remain easy that are going to be made to the scope, weights, or meth-
to present and understand. odology used to estimate the CPI.
78. Average prices and price ranges for important and rea- 85. Technical guidance on the compilation of consumer
sonably homogeneous products may be estimated and price indices is provided in the Consumer Price Index
published in order to support the research and analytical Manual: Theory and Practice.13 This manual should
needs of users. be updated periodically in order to reflect current best
practices.
79. Countries should report national CPI results and
methodological information to the International La-
bour Office as soon as possible after their national
release.
80. Comparing national CPI movements across countries
is difficult because of the different measurement ap-
proaches used by countries for certain products, par-
ticularly housing and financial services. The exclusion 11
United Nations Economic and Social Council (1994).
12
Sixteenth International Conference of Labour Statisticians (1998).
13
International Labour Organization, International Monetary Fund,
Organisation for Economic Co-operation and Development, Eurostat, UN
10
See Annex 4. Economic Commission for Europe, and World Bank (2004).
434
APPENDIX
Annex 1 (10) The “weight reference period” is the period, usually a

year, whose estimates of the volume of consumption
Terminology and Definitions and its components are used to calculate the weights.
(11) “Probability sampling” is the selection of a sample of units,
such as outlets or products, in such a way that each unit in
(1) “Consumer goods” are goods or services that are used by the universe has a known nonzero probability of selection.
households for the satisfaction of individual needs or wants. (12) “Cutoff sampling” is a sampling procedure in which a
(2) “Consumption expenditures” are expenditure on con- predetermined threshold is established with all units in
sumer goods and services and can be defined with re- the relevant population at or above the threshold being
gard to “acquisition,”1 “use,” or “payment”: eligible for inclusion in the sample and all units below
• “Acquisition” indicates that it is the total value of the the threshold being excluded. The threshold is usually
goods and services acquired during a given period specified with regard to the size of some relevant vari-
that should be taken into account, irrespective of able (such as some percentage of total sales), the largest
whether they were wholly paid for or used during the sampling units being included and the rest excluded.
period. This approach could be extended to include (13) “Quota sampling” is a nonprobability method where the
the estimated values of own-account production and population is divided into certain strata. For each stra-
social transfers in kind received from government tum, the number (“quota”) of elements to be included in
or nonprofit institutions. The prices enter the CPI in the sample is specified. The price collector simply “fills
the period when consumers accept or agree prices, as the quotas,” which means, in the case of outlet sampling,
distinct from the time payment is made. that the selection of the outlets is based on the judgment
• “Use” indicates that it is the total value of all goods of the price collectors and the specified criteria.
and services actually consumed during a given period (14) “Imputed expenditures” are the expenditures assigned to
that should be taken into account; for durable goods a product that has not been purchased, such as a product
this approach requires valuing the services provided that has been produced by the household for its own con-
by these goods during the period. The prices (opportu- sumption (including housing services produced by owner-
nity costs) enter the CPI in the period of consumption. occupiers) or a product received as payment in kind or as
• “Payment” indicates that it is the total payment made for a free transfer from government or nonprofit institutions.
goods and services during a given period that should be (15) “Imputed price” refers to the estimated price of a prod-
taken into account, without regard to whether they were uct whose price during a particular period has not been
delivered or used during the period. The prices enter the observed and is therefore missing. It is also the price
CPI in the period or periods when the payment is made. assigned to a product for which the expenditures have
(3) “Scope of the index” refers to the population groups, been imputed, see (14).
geographic areas, products, and outlets for which the (16) “Outlet” indicates a shop, market stall, service establish-
index is constructed. ment, internet seller, or other place where goods and services
(4) “Coverage” of the index is the set of goods and ser- are sold or provided to consumers for nonbusiness use.
vices represented in the index. For practical reasons, (17) “Linking” means joining together two consecutive se-
coverage may have to be less than what corresponds to quences of price observations, or price indices, that overlap
the defined scope of the index. in one or more periods, by rescaling one of them so that the
(5) “Reference population” refers to that specific popula- value in the overlap period is the same in both sequences,
tion group for which the index has been constructed. thus combining them into a single continuous series.
(6) “Weights” are the aggregate consumption expenditures on (18) “Price” is defined as the value of one unit of a product,
any set of goods and services expressed as a proportion of for which the quantities are perfectly homogeneous not
the total consumption expenditures on all goods and ser- only in a physical sense but also in respect of a number
vices within the scope of the index in the weight reference of other characteristics.
period. They are a set of numbers summing up to unity. (19) “Pure price change” is that change in the price of a good
(7) “Price updating of weights” is a procedure that is used or service which is not due to any change in its qual-
to bring the expenditure weights in line with the index ity. When the quality does change, the pure price change
or price reference period. The price-updated weights is the price change remaining after eliminating the es-
are calculated by multiplying the weights from the timated contribution of the change in quality to the ob-
weight reference period by elementary indices measur- served price change.
ing the price changes between weight reference and (20) “Quality adjustment” refers to the process of adjusting
price reference period and rescaling to sum to unity. the observed prices of a product to remove the effect of
(8) “Index reference period” is the period for which the any changes in the quality of that product over time so
value of the index is set at 100. that pure price change may be identified.
(9) “Price reference period” is the period whose prices are (21) “Consumer substitution” occurs when, faced with
compared with the prices in the current period, the pe- changes in relative price, consumers buy more of the
riod whose prices appear in the denominators of the good that has become relatively cheaper and less of
price relatives. the good that has become relatively more expensive.
It may occur between varieties of the same product or
1
This definition differs from the one adopted by the Fourteenth ICLS (1987). between different expenditure categories.
435
APPENDIX
Annex 2 price statisticians or price collectors on the value of any

quality difference between the old and replacement prod-
Quality-Adjustment Methods uct. None, some, or all of the price difference may be at-
tributed to the improved quality.
5. The “differences in production costs” approach relies on
Implicit Quality-Adjustment Methods the information provided by the manufacturers on the pro-
duction costs of new features of the replacements (new
1. The “overlap” method assumes that the entire price differ- models), to which retail markups and associated indirect
ence at a common point in time between the disappearing taxes are then added. This approach is most practicable
product and its replacement is due to a difference in quality. in markets with a relatively small number of producers,
2. The “overall mean imputation” method first calculates the with infrequent and predictable model updates. However,
average price change for an aggregate without the disappear- it should be used with caution as it is possible for new
ing product and its replacement, and then uses that rate of production techniques to reduce costs while simultane-
price change to impute a price change for the disappearing ously improving quality.
product. It assumes that the pure price difference between 6. The “quantity adjustment” method is applicable to prod-
the disappearing product and its replacement is equal to the ucts for which the replacement product is of a different
average price changes for continuing (nonmissing) products. size to the previously available one. It should only be
3. The “class mean imputation” method is a variant of the used if the differences in quantities do not have an impact
overall mean imputation method. The only difference is in on the quality of the good.
the source of the imputed rate of price change to period t + 1 7. The “option cost” method adjusts the price of the replace-
for the disappearing product. Rather than using the average ments for the value of the new observable characteristics.
index change for all the nonmissing products in the aggre- An example of this is the addition of a feature that earlier
gate, the imputed rate of price change is estimated using has been a priced option as standard to a new automobile
only those price changes of the products that were judged model.
essentially equivalent or were directly quality adjusted. 8. A “hedonic” regression method estimates the price of a
product as a function of the characteristics it possesses.
Explicit Quality-Adjustment Methods The relationship between the prices and all relevant and
observable price-determining characteristics is first esti-
4. The “expert’s adjustment” method relies on the judgment mated and then results are used in the estimation of the
of one or more industry experts, commodity specialists, index.
437
APPENDIX
Annex 3 categories of consumption owing to the use of an inap-

propriate method for aggregating elementary aggregates
Types of Errors in the construction of the overall index value. Only rel-
evant to a COLI, although an equivalent (representativity
error) may be defined from the perspective of the pure
price index.
• “Quality change error” is the error that can occur as a
• “Elementary index error” arises from the use of an
result of the index’s failure to make proper allowance for
changes in the quality of goods and services. inappropriate method for aggregating price quotations
at the very lowest level of aggregation. The elemen-
• “New goods error” is the failure to reflect either price tary index error can take two forms: formula error and
changes in new products not yet sampled, or given a lower-level substitution error. The index suffers from
COLI objective, the welfare gain to consumers when formula error if, as a result of the properties of the
those products appear. formula, the result produced is biased relative to what
• “Outlet substitution error” can occur when consumers would have been the result if a pure price change could
shift their purchases among outlets for the same product have been estimated. The index suffers from lower-
without proper reflection of this shift in the data collec- level substitution error if it does not reflect consumer
tion for the index. substitution among the products contained in the ele-
• “New outlets error” is conceptually identical to new goods mentary aggregate.
error. It arises because of the failure to reflect either price • “Selection error” arises when the sample of price obser-
changes in new outlets not yet sampled, or the welfare vations is not fully representative of the intended popula-
gain to consumers when the new outlets appear. tion of outlets or products. The first four types of errors
• “Upper-level substitution error” arises when the index listed previously can be seen as special cases of this type
does not reflect consumer substitution among the basic of error.
439
APPENDIX
Annex 4 05.6 Goods and services for routine household

maintenance
Classification of Individual 06 Health
Consumption According to 06.1 Medical products, appliances, and equipment
06.2 Outpatient services
Purpose1 06.3 Hospital services
07 Transport
07.1 Purchase of vehicles
07.2 Operation of personal transport equipment
(BREAKDOWN OF INDIVIDUAL 07.3 Transport services
CONSUMPTION EXPENDITURE OF
HOUSEHOLDS BY DIVISION AND GROUP) 08 Communication
08.1 Postal services
01 Food and nonalcoholic beverages 08.2 Telephone and telefax equipment
01.1 Food 08.3 Telephone and telefax services
01.2 Nonalcoholic beverages 09 Recreation and Culture
02 Alcoholic beverages, tobacco, and narcotics 09.1 Audiovisual, photographic, and information pro-
02.1 Alcoholic beverages cessing equipment
02.2 Tobacco 09.2 Other major durables for recreation and culture
02.3 Narcotics 09.3 Other recreational products and equipment, gar-
dens, and pets
03 Clothing and footwear 09.4 Recreational and cultural services
03.1 Clothing 09.5 Newspapers, books, and stationery
03.2 Footwear 09.6 Package holidays
04 Housing, water, electricity, gas, and other fuels 10 Education
04.1 Actual rentals for housing 10.1 Preprimary and primary education
04.2 Imputed rentals for housing 10.2 Secondary education
04.3 Maintenance and repair of the dwelling 10.3 Postsecondary nontertiary education
04.4 Water supply and miscellaneous services related 10.4 Tertiary education
to the dwelling 10.5 Education not definable by level
04.5 Electricity, gas, and other fuels
11 Restaurants and hotels
05 Furnishings, household equipment, and routine 11.1 Catering services
household maintenance 11.2 Accommodation services
05.1 Furniture and furnishings, carpets, and other floor
coverings 12 Miscellaneous goods and services
05.2 Household textiles 12.1 Personal care
05.3 Household appliances 12.2 Prostitution
05.4 Glassware, tableware, and household utensils 12.3 Personal effects n.e.c.
05.5 Tools and equipment for house and garden 12.4 Social protection
12.5 Insurance
12.6 Financial services n.e.c.
1
See Appendix 2 to the Manual for explanatory notes. 12.7 Other services n.e.c.
441
APPENDIX
Appendix 5 While not providing an exhaustive account of the related

problems and relevant aggregation methods, this annex aims
Spatial Comparisons of to achieve a degree of completeness in the coverage of the
problem of consumer price comparisons in the Manual by
Consumer Prices, Purchasing adding the spatial and international dimensions to the tempo-
Power Parities, and the ral comparisons dealt with in various chapters. The annex also
attempts to identify possible avenues for a closer integration
International Comparison between spatial and temporal comparisons of consumer prices.
The main objectives of the annex are: (1) to provide a brief
Program summary of the index number problems encountered in the
process of international and interarea price comparisons and
to highlight the need for the development and use of special-
Introduction ized aggregation methods; (2) to describe selected aggrega-
This annex deals with the topic of comparing price levels tion methods used in deriving PPPs and spatial measures
across different areas or regions within a country, as well of price levels; (3) to examine the relationship between the
as across countries. Even though international price com- ICP and PPPs, for cross-country comparisons, with the con-
parisons are required to handle differences in currencies in sumer price index (CPI); and (4) to explore the feasibility of
different countries, the index number problems involved in integrating the ICP activities with the streamlined activities
price comparisons across countries mirror those encoun- of national statistical offices for the compilation of the CPI.
tered in comparisons over time. There is a large body of The annex is also designed to provide an introduction for
literature on cross-country comparisons of prices and real the CPI compilers in various national statistical offices to the
income generated through the International Comparison issues and methods involved in spatial comparisons of con-
Program (ICP). sumer prices. The annex outlines some of the principal dif-
The ICP is a worldwide statistical initiative led by the ferences in the approaches to spatial comparisons. Countries
World Bank under the auspices of the United Nations Sta- embarking on interarea or regional consumer price compari-
tistical Commission. The ICP estimates purchasing power sons, as well as those countries that may participate in the ICP
parities (PPPs) of countries through a partnership with inter- in the near future, may find the contents of the annex useful.
national, regional, and national agencies.
PPPs are price indices that serve both as currency con- Differences between Temporal and
vertors and spatial price deflators. PPPs convert different Spatial Comparisons
currencies to a common currency and, in the process of con- There are several major qualitative differences in the nature
version, equalize their purchasing power by eliminating the of price comparisons involved in the standard CPI compari-
differences in price levels between countries. In their sim- sons over time and price comparisons over space involving
plest form, PPPs are price relatives that show the ratio of the regions or countries. These differences highlight the need
prices in national currencies of the same good or service in for specialized methods for aggregating price data in deriv-
different countries. ing summary measures of price levels, as well as specific
PPPs are used to generate the price and volume indi- types of data requirements associated with cross-country
ces needed for economic research and policy analysis that and interarea comparisons.
involves intercountry comparisons of gross domestic prod- The foremost difference is the absence of a natural order-
uct and its expenditure components. The volume indices can ing of price and quantity observations in the context of
be used to compare the size of economies and their levels cross-country or interarea comparisons. The CPI framework
of material well-being, consumption, investment, govern- and methods are devised to measure changes over time.
ment expenditure, and overall productivity. The price indi- Therefore, the price observations appear in a chronological
ces can be used to compare price levels, price structures, order. The presence of a natural ordering over time of price
price convergence, and competitiveness. In addition, PPPs observations makes it possible to examine the feasibility and
are employed for several administrative purposes. relative merits of the fixed and chain index numbers. In con-
The ICP was launched in the late 1960s as a joint venture trast, in the context of constructing price comparisons across
of the United Nations Statistical Division and the Interna- countries, or across regions within a country, it is impossible
tional Comparisons Unit of the University of Pennsylvania. to arrive at any kind of natural ordering.
The first experimental comparison was conducted in 1970, The multilateral nature of spatial comparisons is a dis-
covering 10 countries. Since then, several rounds of com- tinguishing feature of price comparisons across regions and
parisons have been conducted covering a varying number countries. When price levels of goods and services across
of countries, combined with significant advancements in the different countries are compared, it is essential that such
PPP estimation methodology. comparisons are undertaken for every pair of countries
Recognizing the needs for more frequent and reliable being considered. This multilateral nature of comparisons
PPP estimates, the United Nations Statistical Commission creates several problems. First, the number of comparisons
agreed, at its 47th Session in 2016, that the ICP should (one for each pair) can be quite large, and presentation and
become a permanent element of the global statistical pro- use of such results may be quite unwieldy. For example, if
gram, conducted more frequently with shorter intervals a particular comparison exercise involves 100 countries,
between successive rounds and further aligned with regular n n! 
national statistical programs. then it requires 100 × 99/2 = 4,950  = , (where
k k !(n − k )!
443
APPENDIX
n is the number of countries and k = 2) separate binary Spatial comparisons pose several problems with regard
comparisons involving distinct pairs of countries. Second, to identifying products that are to be priced from different
results from such a large tableau of binary comparisons areas, regions, or countries involved in a comparison exer-
require a degree of consistency. This requirement trans- cise. This problem is less severe when areas with similar
lates into the “transitivity” condition described in the fol- economic structures or homogeneous consumption pat-
lowing text. terns are being compared. However, when comparisons
The uses and applications of interarea price compari- are made between two heterogeneous countries, say the
sons may differ significantly from general consumer price United States and India, there will be major differences in
indices. The CPI is probably the most significant economic the consumption baskets at the detailed item level, even
statistic produced in any country. It is not only used as a though the expenditure categories may be identical. This
general measure of price changes over time, but often in problem is somewhat similar to the treatment of disap-
assessing and calibrating monetary policy. Despite the pearing and new goods in the context of the CPI, but is
conceptual similarities in price comparisons over time and more serious when cross-country comparisons are being
across space, spatial price comparisons are useful in com- attempted.
parisons of standards of living and well-being in different Quality mismatches are likely to be smaller in the case
regions in a country or across countries. Such comparisons of temporal comparisons, but can be a serious problem
are essential in assessing development and in ensuring more when comparisons across countries are attempted. The ICP
balanced growth in different regions. There is considerable follows the principle of identity (that is, comparing like
demand for measures of CPI across different cities and dif- with like) in dealing with the problem of quality differ-
ferent states and regions (rural versus urban) within coun- ences across countries. A comprehensive list of products
tries. However, there are very few countries where interarea with detailed product specifications is developed at the
price level comparisons are readily available. planning stages of any cross-country comparison exercise.
International comparisons of prices, in the form of PPPs These items are priced in different countries from various
from the ICP, are used by international organizations and outlets distributed across the country, a procedure very
individual researchers in assessing the growth and produc- similar to that used in the CPI. Development of the prod-
tivity performance of countries, and also in making mean- uct list is, however, a difficult step, with the degree of dif-
ingful comparisons of various national income aggregates ficulty depending upon the size and heterogeneity of the
across different countries. PPPs are also essential in assess- group of countries involved. The use of a product listing,
ing the nature and extent of global poverty and its distribu- based on the identity principle, can have serious implica-
tion across countries and regions of the world. tions for the representativeness of the product list of the
In recognition of the major analytical differences between consumption baskets in different countries. At the same
the standard CPI comparisons over time and spatial (between time, the requirement of comparing like with like leads to
countries or between regions in a country) comparisons of multiple gaps in the price matrix, as many countries cannot
consumer prices and PPPs, considerable research efforts price all the items on the list. A more detailed account of
have focused on the development of the data and methods the problems and recommended solutions can be found in
necessary for spatial comparisons of prices. A brief sum- Measuring the Real Size of the World Economy (2013) and
mary of the results of this research is given in the following ICP Operational Guide (2013).
text. For purposes of making interarea comparisons of
consumer prices, it is necessary to obtain expenditure
Data Requirements for Spatial data specific to each area or country included in the
(Interarea) Comparisons comparisons.
Thus, spatial comparisons of consumer prices pose spe-
The basic data requirements for spatial comparisons are cific problems because of the nonoverlapping nature of
very similar to the data required for standard CPI calcu- the consumption baskets, potential major differences in
lation. The main components are the data on prices of a the quality of items priced in different regions and coun-
large range of products representative of the consumption tries, and the difficulties to fill data gaps in expenditure
baskets of households and information on weights associ- patterns.
ated with various product categories reflecting the impor-
tance attached to different products. Within the CPI, it is
common practice to collect price quotations from differ- Aggregation Methods for Spatial
ent outlets scattered throughout the country. The selection Comparisons
of the outlets and areas from which prices are collected This section briefly describes the types of aggregation
is based on complex multistage sampling designs. The methods that are commonly used in spatial comparisons of
expenditure weights are based on a classification of goods prices. Though most of these methods have been developed
and services using a standard system such as the Clas- in the context of the ICP, and are equally valid for inter-
sification of Individual Consumption According to Pur- area or regional comparisons, the following discussion uses
pose or a similar national classification. The lowest level countries as spatial entities. This section is further divided
of product classification at which expenditure weights into three parts. The first deals with the notation and con-
are available is used in identifying the elementary indi- ceptual framework necessary to deal with multilateral spa-
ces and higher-level indices at progressively higher levels tial comparisons. The second describes the construction of
of aggregation, leading ultimately to the total household elementary indices for aggregation of prices when no quan-
expenditure level. tity or expenditure information is available. Finally, a small
444
APPENDIX
selection of index number methods used in spatial price the microeconomic theoretical and test approaches to cross-
comparisons is presented. country comparisons. Thus, it is possible to apply Fisher,
Törnqvist, Walsh, or other index number formulae described
in the Manual.
Notation and Conceptual To ensure meaningful interpretation of the results from
Framework multilateral cross-country comparisons, the index number
Consider the case involving comparisons across M coun- methods applied need to satisfy a number of basic require-
tries, and price and quantity data on N products. These prod- ments, only the most important of which are discussed in
ucts refer to goods and services that are priced in all the the following text. Diewert (2013) provides a complete list
countries. If the products refer to items below the elemen- of these requirements.
tary level at which no quantity or expenditure share data are Transitivity. An index number formula Ijk is said to satisfy
available, only the price data is used. At this stage, all the the transitivity property if and only if for all choices of j,k
problems relating to nonoverlapping product lists are set and ( j,k, = 1, 2, . . ., M), the index satisfies
aside so that the main focus is just on the aggregation issues.
Let pj = [ p1j, … pNj ] and qj = [ q1j, … qNj ] represent Ijk = Ijl × Ilk (A5.3)
the price and quantity vectors from country j ( j = 1,2, …,
M). All the prices are expressed in the respective national Equation A5.3 requires that the application of a formula
currency units. As in the case of the CPI computation, the to make a direct comparison, Ijk, should result in the same
problem is one of decomposing the differences in the value numerical measure as an indirect comparison between j and
aggregates k through a link country . Note that the transitivity prop-
N
erty ensures internal consistency of index numbers in the
V j = ∑ pi j qi j (A5.1) matrix given in equation A5.2. It guarantees that the PPP
i =1
for two currencies, say A and B, is the same whether it is
into measures of price and real expenditure components. derived through a direct comparison of A and B or through
Since there are M sets of price and quantity vectors and, an indirect comparison that compares A with C and C with
therefore, M(M − 1)/2 binary comparisons between all dis- B, which are then combined to provide an indirect PPP for
tinct pairs of countries, a simpler notation is used in this A and B. This requirement arises mainly from the spatial
annex in the place of the notation generally used in the nature of the comparisons where no natural ordering of the
Manual. Let Ijk denote the price index number for country countries involved could be imposed without a value judg-
k with country j as the base. If j and k are, respectively, the ment. Most of the commonly used index number procedures
United States and India, and if Ijk = 22.50, then the index is do not satisfy this requirement.
interpreted to mean that 22.50 Indian rupees have the same Base invariance. An index number formula is said to be
purchasing power as one US dollar for the goods and ser- base invariant if a comparison between a given pair of coun-
vices covered in computing the index. Thus, the index can tries ( j,k) is invariant to the order in which the countries are
be interpreted as the PPP between the currencies of j and k. listed. This implies that multilateral comparisons should be
This interpretation is consistent with the meaning accorded invariant to all possible permutations of the data set. For
to the CPI. Since currency denominations are involved here, example, consider a set of transitive comparisons derived
a proper measure of relative price level differences can be using a particular country (say the United States) as a base
obtained if the PPP is compared to the exchange rate prevail- country: under this program, price comparison between any
ing at the time when comparisons are made. In the case of pair of countries, say A and B, is effected through the United
subnational comparisons, the PPPs would be the price level States which serves as a link country. Therefore,
differences.
Because of the multilateral nature of spatial comparisons, PPPA, B = PPPA, USA × PPPUSA , B
(A5.4)
when M countries are involved, it is necessary to provide
comparisons between all pairs of countries. Thus, it becomes
necessary to compute each and every entry in the following This program is inadmissible under the base invariance cri-
matrix of binary comparisons: terion since the choice of the base country clearly affects the
PPP of currencies of countries A and B. Further, the United
I I12 . I1M  States is accorded a special status, in the form of a link coun-
 11  try, in deriving transitive multilateral comparisons.
 I 21 I 22 . I2M 
I =   (A5.2) Characteristicity. This is a requirement outlined in
 . . . .  Drechsler (1973). This property requires that any set of
I IM 2 . I MM  multilateral comparisons satisfying the transitivity prop-
 M 1
erty should retain the essential features of the binary com-
Several points concerning the matrix, I, are worth noting. parisons constructed without the transitivity requirement.
First, the matrix can be large if the number of countries (or Since condition (A5.3) implies that a transitive comparison
areas) involved is large. Second, the results recorded in the between a pair of countries j and k is necessarily influenced
matrix need to be internally consistent. All the index num- by the price and quantity data for all the other countries, the
ber issues and various approaches discussed in the Manual characteristicity property requires that distortions resulting
apply directly to each binary comparison involving two from adherence to the transitivity property should be kept at
countries. Diewert (1986, 1999b) provides a summary of a minimum. The Gini–Eltetö–Köves–Szulc (GEKS) method
445
APPENDIX
for multilateral comparisons, discussed in the following text, Obviously, these binary indices are not transitive, since each
is specifically designed with the characheristicity in mind. index is based on prices of a different set of products. The
Additivity or matrix consistency. Another desirable prop- GEKS procedure is then applied to derive a transitive set of
erty in spatial comparisons is additivity or matrix consistency. indices. The resulting multilateral index is given by
This property ensures that the additive nature of national
accounts within a country, expressed in national currency I jk EKS = Π M=1  I j I k 
1/ M
(A5.6)
units, is maintained within the international comparison
framework. Basically, additivity means that for any coun-
try the real expenditures components should add up to the The elementary index number formula in A5.5 is similar to
total real expenditures. Additivity would enable researchers the formula used in the construction of the CPI. The princi-
to examine the structure of the real expenditures. However, pal difference results from the fact there is a need for tran-
additivity imposes certain theoretical restrictions and there- sitivity at all stages of aggregation. The properties of these
fore is not always preferred as a property to be maintained in indices are discussed in Chapter 7 of the publication Con-
international comparisons. Indices such as the Geary–Kha- sumer Price Index Theory.
mis (GK) or Iklé–Dikhanov–Balk (IDB) maintain additivity. The Eurostat and the Organisation for Economic Co-
In intracountry comparisons, those indices seem less prob- operation and Development used a somewhat modified
lematic as the price and expenditure structures vary much variant of the binary indices shown in equation A5.5,
less within a country than between countries. referred to as Jevons–GEKS*. Although this elementary
index still does not use expenditure weights below the basic
heading level, it incorporates information about importance
Index Number Methods for Spatial of products in one or both of the countries. The procedure
Comparisons takes explicit account of those products which are “starred,”
Spatial price comparisons, similar to temporal indices, use indicating that the item is important or representative in a
index number methods for aggregating price and quantity given country. The Jevons–GEKS* method uses the same
data at two different levels. The first is the basic heading formula (A5.6), but the binary elementary index (A5.5) is
level (the elementary aggregation). This is normally the replaced by
lowest level of aggregation at which expenditure data and
1
weights are available. The basic headings usually consist  1 2
 pi s  n ( j ) 
1
of a fairly homogeneous group of items that are priced in   pi s  n ( s )
= ∏  j    (A5.7)
 i∈M ( s ) pi  i∈∏
different outlets in the countries. The subsequent levels of I js
 j 
M ( j )  pi 
aggregation lead to indices for higher expenditure catego- 
 
ries, up to the level of total expenditures.
where n(s) and n( j) are, respectively, the number of “starred”
items in countries s and j; M(s) and M( j) are, respectively,
Aggregation at the Basic Heading Level the sets of products that are “starred” (considered represen-
(Elementary Index) tative) in countries s and j.
Two commonly used index number methods and their ver- In theory, equation A5.7 can have an advantage over
sions are described in the following text. These procedures (A5.5) due to its considering “representativity” of the
explicitly allow for the possibility that price data may not products priced in different countries. However, much will
be available for all items in the product list constructed depend on how consistent the definition of representativity
for a given international comparison exercise. Such a situ- is across various countries and within basic headings (see
ation occurs also in the case of temporal comparisons, more in a discussion on the country–product–representativ-
though it is usually limited to either disappearing or new ity–dummy method in the following text).
goods. Today the Eurostat–Organisation for Economic Co-oper-
The Jevons–Gini–Eltetö–Köves–Szulc (Jevons–GEKS) ation and Development program uses a modification of the
Method. The original GEKS method proposed in Eltetö and method, called Jevons–GEKS*(S), due to Sergeev (2003).
Köves (1964) and Szulc (1964) is generally used in aggre- The method uses an elementary binary index that has three
gating price data above the basic heading level. However, components which are based on (1) items representative in
the GEKS procedure can be used as an elementary aggrega- country s and not in country j, (2) items representative in
tor as well (that is, at the basic heading level). At this stage, country j and not in country s, and (3) items representative
binary comparisons have to be constructed using price rel- in both countries. Then the binary elementary index is com-
atives of those products for which prices are available in puted as a weighted geometric mean with the weights pro-
both countries, without applying expenditure weights. The portional to the number of items in different components,
Jevons–GEKS method uses the Jevons index as its basic ele- with double weight assigned for the third component based
ment. The Jevons index can be described as follows. If S( j) on items representative in both countries:
and S(k) are the sets of products priced in countries j and k,
respectively, and njk is the number of products that are priced  s
w1
 s
w2 w
 pi s  3
in both countries, then the binary elementary index uses the I js = ∏  pi  ∏  pi  ∏   (A5.7a)
 j
i ∈ M ( s *) pi 
 j
i ∈ M ( j *) pi 
 j
i ∈ M ( s , j )  pi 
following formula:
pk
1/ n jk
where M(s*) and M( j*) are the sets of priced products that
I jk = Πi∈S ( j )∩ S ( k )  i j  (A5.5) are “starred” (considered representative) only in countries s
p 
 i  and j, respectively, and not in other countries, and M(s,j) is
446
APPENDIX
the set of products that is representative in both countries s pseudoweights to account for higher sales of more typical
and j. items.
Finally, equation A5.6 is applied to achieve transitivity. Matrix X becomes in this case:
In both methods, the Jevons–GEKS* and Jevons–
GEKS*(S), products that are unrepresentative in both coun- xcp′ = Dc2 ... DcNc Dp1 ... DpNp Rcp 
tries do not influence the respective binary index. At the T
same time, both methods’ binary indices may be distorted β ′ = α2′ ... αNc ′ ρ
′ γ1′ ... γ Np
 (A5.13a)
because they rely on products that are representative in one
country but not in the other. The Eurostat-OECD addresses y = X ′β ′ + ε (A5.13b)
this issue by endeavoring to equalize the number of repre-
sentative items in each country. where Rcp stands for representativity dummy and ρ is its
The Country–Product–Dummy (CPD) Method. The CPD respective regression coefficient.
method was originally proposed by Summers (1973) as a However, using actual ICP data, it was found that the effi-
tool to deal with missing price observations. The method is ciency of the index depended significantly on the consis-
a simple statistical device that can be used in deriving the tency of representativity definitions across countries (just as
PPPs for a particular basic heading by simply regressing the in the case of Jevons–GEKS* and Jevons–GEKS*(S)). This
logarithm of observed prices against a set of dummy vari- led to the random nature of the representativity coefficients
ables, defined with respect to products and countries. in many cases, instead of strongly negative coefficients.
The CPD procedure can be written as This does not mean that the index is less efficient than the
regular CPD, just that the representativity data quality has to
lnpcp = ycp = xcp β + εcp (A5.8) be increased in order to achieve higher efficiency.
The Country–Product–Dummy Weighted Method. Diew-
ert (2004a) suggested using product weights in formula
where pcp is the price of product p in country c; Dc j and Dpi (A5.10). Essentially, this variant uses weights 3 for repre-
are country and product dummy variables; Np and Nc are the sentative and 1 for nonrepresentative products. The weights
numbers of products and countries, respectively; and were somewhat arbitrary, and, again the index efficiency
in this case will depend on the quality of representativity/
xcp =  Dc2 … DcNc Dp1 Dp2 … DpNp  importance data. Other CPD versions are possible too, for
T example, the country–product–representativity–dummy
β = α2 ... αNc γ1γ 2 ... γ Np  (A5.8a) with weights, CPD with data frequency adjustments, and so on
It is worth noting that in the case of a full price matrix,
In matrix notation, by stacking individual observations, this both the CPD and Jevons–GEKS indices degenerate into
can be written as the simple geometric mean of price relatives. An important
consideration is the efficiency of the elementary indices.
y = Xβ +ε (A5.9) Dikhanov (2005) considered multiple Jevons–GEKS and
CPD variants and using Monte Carlo simulations had two
Note that the first country dummy is dropped from the sys- findings. One is that Jevons–GEKS variants were less stable
tem because matrix X is of rank (Np + Nc − 1). In fact, any and had higher failure rate, especially, when used as defined
variable can be dropped from the system; dropping the first previously, without any modifications—the modifications
country’s dummy simply makes it the base country. that were employed to decrease failure rate could introduce
The solution is given (under the conditions of indepen- various biases. The other is that the CPD variants used with
dently and identically distributed random disturbances) by weights/representativity were the most efficient indices.
βˆ = ( X T X ) X T y
−1
(A5.10) Aggregation above the Basic Heading Level
This section presents a small selection of the range of aggrega-
Once this regression equation is estimated, the PPP for tion methods used in the context of spatial comparisons. For
the currency of country j with country 1 as base can be a more comprehensive overview see, for example, Measuring
obtained by the Real Size of the World Economy (World Bank 2013).
This level of aggregation in ICP (above the basic heading
PPPj = exp (α j ) (A5.11) level) is similar to the stage where elementary indices are
aggregated to derive the all-items CPI (above CPI items). It
Then the desired index at the basic heading level is given by is important to note that as the basic heading level is the low-
est level with expenditure weights; it corresponds to the CPI
PPPk item level, the lowest level with expenditure weights in CPI.
I jk = . (A5.12)
PPPj In addition, the multilateral nature of spatial comparisons
necessitates a different approach to index numbers than in
The Country–Product–Representativity–Dummy Method. temporal comparisons.
The CPD model offers a number of generalizations that A number of index number methods for aggregation
can explicitly account for a number of data-related prob- above the basic heading level have been developed but, in
lems. One of them was due to Cuthbert (1988), who the interest of brevity, only the principal methods used in the
suggested incorporating the representativity dummy vari- context of international comparisons are discussed here.
able into equation A5.8. The rationale was to introduce The Geary–Khamis (GK), Iklé–Dikhanov–Balk (IDB),
447
APPENDIX
and GEKS methods were all used in the ICP, at one time or the following system of interrelated equations, defined for
another. In addition, the weighted CPD and Minimum Span- each country j and each product i:
ning Tree approach are discussed as well.
The GEKS Method. The GEKS system is a simple
∑
M
∑
N
p j qi j / PPPj p j qi j
method of generating transitive multilateral index numbers P=
j =1 i
and PPPj = i =1 i
(A5.15)
from a system of binary index numbers, with the prop- i
∑
N
∑
M j
qj Pq
i =1 i i
j =1 i
erty that the resulting multilateral indices deviate the least
(according to a specific criterion) from the binary indi-
These simultaneous equations are then solved to yield
ces. Since the seminal paper by Drechsler (1973), it has
numerical values of PPPs and Ps, after selecting one of the
been well recognized that (transitive) multilateral systems
currencies as a numeraire. Once the PPPs are solved, the
necessarily deviate from their binary counterparts and
spatial price index numbers are simply defined as
therefore result in a loss of “characteristicity.” The GEKS
system is designed to minimize such loss of characteris-
PPPk
ticity. The original GEKS system uses the Fisher binary I jk = (A5.16)
indices; however, any other binary index can be used in PPPj
conjunction with the GEKS method of multilateralization.
Thus, for any pair of countries j and k, if Fjk represents the One of the main reasons for the continued use of the GK
Fisher binary index, then method is “additivity.” Additivity requires that aggre-
gates, such as real domestic product, derived by converting
M
1/ M national aggregates using PPPs, should be equal to aggre-
GEKS jk = ∏  Fj .Fk  (A5.14) gates derived through valuation of quantities at international
=1
prices. Thus, additivity requires
provides the GEKS index.
n n
There are several notable features of the GEKS tech-
nique. First, it is based on the premise that direct binary
∑pi =1
i
j
qi j / PPPj = ∑Pq
i i
j
i =1
(A5.17)
comparisons, derived using any chosen formula, provide
the best comparison between pairs of countries. Second, This requirement is satisfied automatically by the PPPs and
even though the GEKS index in equation A5.11 is defined Ps derived from the GK system defined in equation A5.12.
using the Fisher index, this approach can be applied in The GK system is also useful in analyzing the structure of
conjunction with any other index number formula. For real gross domestic product and shares of different com-
example, the Fisher index in equation A5.11 may be ponents across different countries. This system provides a
replaced by another superlative index, such as the Törn- framework within which internationally comparable national
qvist index (see Caves and others 1982b). Third, the accounts could be constructed. When country outputs are
GEKS index in equation A5.11 is the multilateral index valued at the international reference prices, values are addi-
that deviates the least from the matrix of nontransitive tive across both countries and products. However, additive
binary indices, when the deviations are measured using multilateral methods are not consistent with economic com-
a logarithmic distance function. Finally, the GEKS index parisons of utility across countries if the number of countries
can be presented as a simple geometric mean of all (direct in the comparison is greater than two (see Diewert in Chap-
and indirect) comparisons between j and k through all pos- ter 5 of Measuring the Size of the World Economy, 2013).
sible link countries. In addition, larger countries will have a larger effect on the
The simple unweighted nature of the GEKS index has international prices than the smaller ones which in practice
attracted attention in recent years. Since different binary leads to an overestimation of poorer countries’ gross domes-
comparisons have different levels of reliability, measured tic product (the so-called Gerschenkron effect).
using various criteria, it is necessary to reflect these differ- The Iklé–Dikhanov–Balk (IDB) Method. The IDB method
ences in defining weighted GEKS index numbers. Rao and was originally intimated by Iklé (1972) but in a very indi-
Timmer (2000), Rao and others (2000), and Rao (2001b) rect way, so its properties remained unknown. It was subse-
provide illustrations of how weighted GEKS indices can quently further developed into its current form by Dikhanov
be generated in order to account for various data-related (1994), and Balk (1996a) provided the first existence proof.
problems. The IDB method is defined through the following system
The Geary–Khamis (GK) Method. The GK method was of interrelated equations, defined for each country j and each
originally proposed by Geary (1958) and subsequently product i:
developed by Khamis (1970, 1972, and 1984). The GK −1
method has been the principal aggregation method in earlier  M PPP / p j × s j 
 ∑ j =1 ∑
N
j i i  p j qi j
ICP phases. Pi =   and PPP = i =1 i
(A5.18a)
  j
∑ ∑
M N
The GK method provides a way of calculating PPPs of  j =1
si
j

 i =1
Pq
i i
j
currencies of different countries from the observed price

and quantity data (applied at the basic heading level). The pi j qi j
concept of PPP is applicable even when the currency unit where sij = .
∑
N
is the same in several areas of a country. The GK method p j qi j
i =1 i
simultaneously determines the international average prices
of different countries. Let Pi denote the international aver- These equations are similar to the GK equations but
age price of ith product. The GK method is defined through now the weights involve country expenditure shares, and
448
APPENDIX
countries have more equal influence on forming the inter- measure, with weights being the country’s expenditure
national prices. In effect, the index becomes more “demo- shares. Using a matrix of distances calculated for all pairs
cratic” in nature, not unlike the GEKS. Deaton and Heston of countries, Hill (1999c, 1999d) suggests that a minimum
(2010) show that the IDB-generated PPPs are much closer spanning tree (MST) be extracted and used in constructing
to the GEKS than the GK PPPs. Dikhanov (1994) shows chained links between all pairs of countries. The MST has
that the GK becomes the IDB when all the countries have the property that a chained comparison between any pair of
identical volumes. countries has the least distance, and therefore can be con-
Weighted Country–Product–Dummy (CPD) Method. It sidered as the most reliable. It also has the property that the
is possible to generalize the CPD method discussed pre- sum of the distances between all the links, in the MST, is
viously in the context of aggregation at the basic head- the least when compared to all possible tree configurations.
ing level. Rao (1995) has generalized the CPD method by Once the MST is identified, a transitive comparison
incorporating quantity and value data directly into the CPD between a given pair of countries in a particular exercise is
method described in equation A5.8. The method minimizes constructed using binary indices calculated using a chosen
a weighted residual sum of squares, with each observation formula, such as the Fisher or Törnqvist index, and the links
weighted according to the expenditure share of the product indicated in the MST. Since the MST provides a unique
in each country. The resulting equation becomes chain of links between any two countries, comparisons are
uniquely defined. The spanning trees are, however, sensitive
wij lnpij = π1 wij D1 + π2 wij D2 + ... + πM wij DM to the countries included, and the types of measures used in
assessing the degree of reliability or comparability of any
+ η1 wij D1* + ... + ηn wij Dn* + vij (A5.19) two countries. In addition, the path of bilateral links between
countries generated by the method could be rather unstable
pij qij and could change drastically from one comparison to next.
where wij = is the value share of the ith basic
∑
N
p q
i =1 ij ij
heading in the jth country. Fixity and Interregional Linking in International
Rao (1995) has shown that the weighted CPD method Comparisons
may be considered as a bridge between the GK approach Any discussion on the methods presented previously, both
to international comparisons and the standard stochastic at and above the basic heading level, would be incomplete,
approach to index numbers. however, without addressing fixity and interregional link-
ing. The need for fixity arises from the regional nature of
Minimum Spanning Tree Method for Making actual ICP exercises, where individual ICP regions conduct
Multilateral Comparisons their own comparisons which are later linked into the global
This approach advocates spatial chaining of binary compari- comparison, as well as from the regions’ desire to preserve
sons where links are identified using a procedure based on intraregional volume relatives within the global compari-
some measure of distance of binary comparisons involved. son. Thus, fixity ensures that regional results get replicated
Recall that the standard GEKS treats each country’s asso- exactly within the global comparison. The fixity principle
ciated set of Fisher indices as equally robust, and thus an also allows for a certain independence of the regional com-
averaging of the parities would be appropriate. putations, in both timing and methods.
Using the graphical theoretical concept of minimum span- The fixity principle necessitates interregional linking to
ning trees, Hill (1999c, 1999d) proposed a method of deriv- bind the regional results, both at and above the basic head-
ing a system of transitive multilateral comparisons from a ing level. This issue is discussed here only briefly. A fuller
matrix of binary comparisons. The Hill approach is based treatment is given in Chapters 4 and 6 of Measuring the Real
on the fact that direct binary comparisons may not always Size of the World Economy (2013).
be the best. Obviously, only the items common across regions can be
For any pair of countries j and k, Hill suggests a measure used in linking. To this end, every region was required to
of distance (indicating the reliability of the binary compari- collect a certain number of common “core” items.
son) using the Laspeyres–Paasche spread defined as Thus, as the first step, the regional basic heading PPPs are
estimated independently for each region. The basic principle
 L( j , k  of linking the regions at the basic heading level is to convert
D ( j , k ) = ln   (A5.20) countries’ average item prices into regional currencies using
 P ( j , k ) the countries’ respective regional basic heading PPPs; then
use one of the standard CPD models described previously
where L( j,k) and P( j,k) are, respectively, binary Laspeyres to estimate interregional linking factors, treating each coun-
and Paasche price index numbers. Note that the same dis- try’s average price as an observation in the CPD regression.
tance function emerges if price index numbers are replaced Upon linking, the countries’ regional PPPs were multiplied
by quantity index numbers. D( j,k) is equal to zero if the by the interregional linking factors to arrive at the global
price structures or quantity structures are identical in coun- linked set of basic heading PPPs with fixity.
tries, j and k. Thus, this distance function serves as an indi- The linking above the basic heading level is carried out
cator of similarity of price and quantity structures in these using the country aggregation with redistribution method.
countries. In simple terms, the linking procedure consists of four steps:
Other distance (similarity) measures are possible. Diew- (1) at every level above basic heading, regional country
ert (2009), for example, suggests the weighted log quadratic volumes are estimated independently for each region; thus
449
APPENDIX
volume for country j within region r is denoted Vr j ; (2) a ICP concepts and practices for national purposes. For
full unrestricted GEKS is carried out with every country example, if analyzing price level differences within
included in the global comparison, and country volumes a country is of interest, the ICP may provide useful
within the global comparison are obtained; thus volume for insights into the dynamics of spatial price comparisons.
country j in the global comparison is written as V j ; (3) each Similarly, ICP concepts such as the structured product
region’s total is estimated by adding up that region’s coun- description, which add rigor to item definitions by list-
try volumes from the global comparison Vr = ∑ V j , where ing price-determining characteristics, may provide a
j ∈C (r ) valuable basis for designing more efficient price col-
C(r) is the set of countries in region r; (4) these regional lection forms.
totals are redistributed among individual countries within Scaling opportunities. Increased synergies between ICP
the regions using the regional country volumes from step and CPI product and outlet frames may allow countries
(1), the resulting country volumes in the global comparison to more easily extend their CPI’s geographical and out-
with fixity are written as: Vworld
j
= Vr *Vr j / ∑ Vr j . let coverage. Since the ICP normally requires coverage
j ∈C (r ) which reflects the average level of prices in the entire
country, ICP information on product availability and
outlets may serve as building blocks for the design of a
Integration of the Consumer Price more extensive CPI. For instance, it may help expand
Index and International Comparison the CPI to cover all urban areas in cases where the CPI
Program only covers the capital city.
This section provides a brief overview of potential synergies It is important to note that the extent to which coun-
and benefits that could result from the integration of CPI tries will benefit from CPI and ICP integration will vary
and ICP activities. In addition, it introduces the concept of depending on national circumstances. It will also depend
subnational PPPs and their relation to CPI and ICP. on the depth of the level of the integration process, with
increased harmonization being a common precursor to
Benefits from Consumer Price Index and integration. Nonetheless, the breadth and depth of the
International Comparison Program Integration potential benefits of bringing CPI and ICP further together
PPPs and CPIs serve different analytical purposes and are even greater in the current context of a permanent and
address different dimensions of prices: spatial and tempo- more frequent ICP.
ral. Because of these conceptual differences, PPPs and CPIs
have generally been constructed under different production
infrastructures. Subnational PPPs
Despite these differences, harmonization and integra- Subnational PPPs are PPPs compiled within a country. They
tion of ICP and CPI activities may provide countries with are similar to ICP PPPs in that they both compare price lev-
some potential benefits. One of the most obvious com- els between geographical areas. In the case of subnational
monalities between CPIs and PPPs is that both require PPPs, comparability over time is an occasional feature. The
basic price data for household goods and services. This link between CPI and ICP is perhaps most prominent and
common feature opens the door to potential cost savings well established within the framework of subnational PPPs.
by countries from having a joint CPI-ICP data collection The general approach to constructing subnational PPPs
process. Savings in data collection are only one of many can be summarized in three steps:
positive prospects from greater CPI and ICP integration.
For instance, cost savings may also result from using a (1) Organizing the required price and expenditure data
shared data curation and validation process, to name a few (2) Determining overlapping items across geographical
possibilities. regions
A brief compilation of potential benefits from CPI and (3) Estimating subnational PPPs through interregional
ICP integration is provided in the following text. These ben- aggregation
efits are seen from the point of view of a national statistical
Because steps (1) and (2) generally present the biggest
office already carrying out CPI-related work. While going
challenges, they will be the focus of interest. Step (3) on
through them, note that potential benefits do not exist in iso-
estimation involves extending the aggregation procedures
lation, but result from the interaction, dependencies, or rela-
for international spatial comparisons (presented earlier) to
tionships they form with other elements that influence the
a subnational setting, so it presents no additional difficul-
overall quality of a statistical system. This leads to cost sav-
ties. The aggregation problem becomes more intractable
ings, accompanied by improvements in efficiency, among
when there is a need to account for the temporal dimension.
other areas.
This occurs when a set of subnational PPPs compares not
Cost savings. Shared CPI-ICP data collection and pro- only price differences across geographical regions of the
cessing operations may result in cost savings in areas same country, but also over time. This space–time analysis
such as price collection, data curation, quality assur- requires a temporally consistent spatial index for aggrega-
ance procedures, information and technology, and staff tion, which can be achieved, for example, by treating each
training. region in each time period under consideration as a separate
Procedural and methodological spillovers. Knowledge entity in standard multilateral aggregation (Dikhanov and
of ICP methods can stimulate countries into adopting others 2011). Thus estimated indices would provide regional
450
APPENDIX
PPPs already linked over time and space. Other approaches price observations for identical item specifications (variet-
are possible, for example, by aggregating regions in each ies) across geographical regions. The idea is to follow the
time period independently, and then linking each time peri- ICP principle of comparing like with like. It is where ICP
od’s total using national price measures such as CPI. This concepts come into play as a means to add rigor and facili-
two-stage approach to producing time and space consistent tate the subnational PPP approach. Concepts such as struc-
estimates would be equivalent to applying fixity not unlike tured product descriptions, for defining item characteristics,
the approach in the current ICP where different regions are or the “core items” approach, for ensuring the pricing of
linked without changes to their respective regional results. common items across regions, help ensure that the required
Constructing subnational PPPs requires price and expen- item variety overlap exists across geographical regions. In
diture data. Prices are required for a set of goods and ser- this sense, ICP concepts help improve the consistency of the
vices that are comparable across the geographical areas subnational PPP procedure.
of the country being studied. Expenditures are needed at Because subnational PPPs have multiple uses at the
the product group level and disaggregated by geographic country level, many countries, especially large ones, have
location. expressed interest in them. Indeed, subnational PPPs can
Expenditure values can be sourced from representa- serve as inputs for welfare studies that compare well-being
tive national household budget surveys, or from the coun- across geographical regions within a country, or in fiscal
try’s regional economic accounts. Since regional accounts studies of public transfers between a country’s geographi-
are uncommon, expenditure values are generally sourced cal regions. Subnational PPPs may also be used for regional
from the most recent nationally representative household cost of living adjustments when determining wages of pub-
budget survey. Price data, in turn, may be sourced from lic officials working in different locations across a given
the CPI collected prices, ICP collected prices, or from spe- country.
cially designed intracountry regional price surveys. The To meet the growing demand for information on this
most attractive would be to utilize CPI price information, topic, research on subnational PPPs will also be carried out
since this option typically requires no additional price data under the umbrella of the ICP Research Agenda. The task
collection. force working on CPI-ICP integration will be studying sub-
The challenge of compiling subnational PPPs is in obtain- national PPPs in detail and will formulate guidelines on the
ing price data at the level of disaggregation and detail use of CPI or ICP products list and prices to compute sub-
required. Price data for subnational PPPs require obtaining national PPPs.
451
APPENDIX
Appendix 6 is termed a series of fixed-basket price indices. In particular,

when qn = qn0 , a series of Laspeyres indices are found.
Some Basic Index Number 4. At period T, one could change to a new quantity basket
q ′ and calculate from this period onward:
Formulas and Terminology

∑p T +1
n nq′
,
∑p T +2
n nq′
,
∑p T +3
n nq′
(A6.5)
1. Throughout this appendix, the sums are understood to be ∑p T
n qn′ ∑p T
n qn′ ∑p T
n qn′
running over all varieties n. To relate the prices of periods T + 1, T + 2, T + 3, … to those
A basket price index is an index of the form of period 0, chain linking can be used to transform the series
(5) into a series of the form:
∑p q t
n n
(A6.1)
∑p q 0
n n ∑p q T
n n ∑p q′ , ∑ p q ∑ p
T +1
n n
T
n n
T +2
n n q′
,
which compares the prices of period t with those of (an ear-
∑p q 0
n n ∑p q′ ∑ p q ∑ p
T
n n
0
n n
T
n qn′

lier) price reference period 0, using a certain specified quan- ∑p q T
n n ∑ p q ′ ,…T +3
n n
(A6.6)
tity basket. The basket does not have to consist of the actual
quantities in any particular period. This general type of index ∑p q 0
n n ∑p q′ T
n n
is called a Lowe price index after the index number pioneer This could be termed a series of chain-linked fixed-basket price
who first proposed this general type of index. The family of indices. In particular, when qn = qn0 and qn′ = qn , a series of
T
Lowe indices includes some well-known indices as special chain-linked Laspeyres indices are found. Since the basket is
cases: changed at period T, the adjective fixed applies literally only
0
• When qn = qn , one gets the Laspeyres index. over a certain number of time intervals. The basket is fixed from
t
• When qn = qn , one gets the Paasche index. period 1 to period T and is again fixed from period T + 1 onward.
When the time intervals during which the basket is kept fixed
• When qn = (qn0 + q1n ) / 2 , one gets the Marshall–Edge- are of the same length, such as one, two, or five years, this fea-
worth index. ture can be made explicit by describing the index as an annual,
• When qn = (qn0 qnt ) , one gets the Walsh index. In prac-
1/ 2
biannual, or five-yearly chain-linked fixed-basket price index.
tice, national statistical offices frequently work with 5. A weighted arithmetic-average price index (so-called to
b
a Lowe index in which qn = qn , where b denotes some distinguish it from a geometric or other kind of average)
weight reference period that is typically prior to 0. is an index of the form:
2. A useful feature of a Lowe index for period t relative to
period 0 is that it can be decomposed, or factored, into ∑w ( p n
t
n / pn0 ) (A6.7)
the product of two or more indices of the same type: for
instance, as the product of an index for period t − 1 rela- which compares the prices of period t with those of period
tive to period 0 and an index for period t relative to period 0, using a certain set of expenditure weights adding up to 1.
t − 1. Formally, In particular, when the weights are the period b value shares
wn = snb ≡ pnb qnb / ∑ pnb qnb (A6.8)

∑p q t
n n
=
∑p q ∑p q
t −1
n n
t
n n
(A6.2)
∑p q 0
n n ∑p q ∑p q
0
n n
t −1
n n
the Young index is obtained.
Note that any basket price index (1) can be expressed in
In particular, when qn = q , expression (2) turns into
0
n the form (7), since

∑p q t 0
n n
=
∑p q ∑p q
t −1 0
n n
t 0
n n
(A6.3) ∑p q t
pn0 qn pnt
=∑
n n
(A6.9)
∑p q 0 0
n n ∑p q ∑p q
0 0
n n
t −1 0
n n ∑p q 0
n n ∑ pn0 qn pn0
The left-hand side of expression (3) is a direct Laspeyres When the weights are the period 0 value shares,
index. Note that only the first of the indices that make up
the right-hand side is itself a Laspeyres index, the sec- wn = sn0 ≡ pn0 qn0 / ∑ pn0 qn0 (A6.10)
ond being a Lowe index for period t relative to period
t − 1 that uses the quantity basket of period 0 (not t − 1). expression (7) turns into the Laspeyres index. When
Some national statistical offices describe the index on the
right-hand side of expression (3) as a modified Laspeyres wn = pn0 qnt / ∑ pn0 qnt (A6.11)
index.
3. In a time-series context, say when t runs from 1 to that is, hybrid period (0, t) value shares, one can get the
T, the series Paasche index.
One could also think of setting

∑p q , ∑p q
1
n n
2
n n
,…,
∑p q
T
n n
(A6.4)
wn = snb ( pn0 / pnb ) / ∑snb ( pn0 / pnb ) = pn0 qnb / ∑ pn0 qnb (A6.12)
∑p q ∑p q
0
n n
0
n n ∑p q
0
n n
453
APPENDIX
that is, price-updated period b value shares. or, using chain linking to relate the prices of periods
Note that hybrid value shares, such as given by expres- T + 1,T + 2,T + 3,… to those of period 0,
sions (11) or (12), are not observable but must be constructed.
6. In a time-series context, when t runs from 1 to T,
the series ∑w ( p n
T
n / pn0 ) ∑wn′ ( pnT +1 / pnT ) ,

∑w ( p n
T
n / pn0 ) ∑wn′ ( pnT +2 / pnT ) (A6.15)
∑w ( pn
1
n / p ) , ∑wn ( p / p ) ,…, ∑wn ( p / p ) (A6.13)
0
n
2
n
0
n
T
n
0
n
This could be termed a series of chain-linked fixed-weighted
arithmetic-average price indices. In particular, when
is termed a series of fixed-weighted arithmetic-average wn = sn0 and wn′ = snT , a series of chain-linked Laspeyres indi-
price indices. In particular, when the weights are equal
to the period 0 expenditure shares, a series of Laspeyres ces are found. When wn = snb ( pn0 / pnb ) / ∑snb ( pn0 / pnb ) and
indices are found, and when the weights are equal to the wn′ = snb ( pnT / pnb ) / ∑snb ( pnT / pnb ) for some later period b′,
′ ′ ′ ′
price-updated period b expenditure shares, a series of Lowe a series of chain-linked Lowe indices are found.
indices are found in which the quantities in the basket are
those of period b. 8. Again, since the weights are changed at period T, the
adjective fixed applies literally only over a certain
7. In period T, one could change to a new set of weights number of time intervals. The weights are fixed from
w′, and calculate from this period onward period 1 to period T and are again fixed from period
T + 1 onward. When the time intervals during which
∑w′ ( p
n
T +1
n / pnT ) , ∑ wn′ ( pnT + 2 / pnT ) , the weights are kept fixed are of the same length, this
feature can be made explicit by adding a temporal ad-
∑w′ ( p
n
T +3
n / pnT ) ,… (A6.14) jective, such as annual, biannual, or five-yearly.
454
APPENDIX
Appendix 7 Consumer Price Index Research

Topics
Consumer Price Index
Research Agenda Data Sources
Scanner Data
Scanner data may lead to more and better official price sta-
The consumer price index (CPI) research agenda lists topics
tistics, efficiency gains, and reduced response burden. Scan-
where more work and research are needed to develop guid-
ner data can complement and, in some instances, replace
ance and good practices for CPI compilation and meeting
traditional approaches to collecting data for compiling the
existing and emerging user needs.
CPI. The use of scanner data challenges traditional interpre-
The research agenda includes conceptual and method-
tations and applications of concepts and raises both method-
ological issues as well as practical measurement and imple-
ological and practical issues related to CPI production. Areas
mentation challenges. Different concepts and methods may
where more work is needed include: improving calculation
be applied when developing the CPI, including decisions on
methods for price indices based on scanner data; combin-
index coverage, incorporating different goods and services,
ing scanner data and data obtained from traditional methods;
and measuring their price development over time. There is a
aggregating scanner data into higher-level indices; using
significant need for analytical work and sharing of practical
scanner data for sampling purposes and estimating weights;
experiences to define and develop best practices for those
classifying scanner data; dealing with quality changes (for
compilation areas where there is no consensus on which
example, implicit versus explicit quality adjustments); clari-
method should be recommended. For example, there is a
fying the treatment of seasonal products; and comparing
need to develop clear guidance on the methods used to make
price indices based on scanner data with price indices based
quality adjustments and how to use scanner data to compile
on traditional sources. A growing number of national statis-
the CPI.
tical offices (NSOs) have implemented scanner data in the
Countries face ongoing measurement and implementation
routine production of the CPI, and the practical experience
challenges. These are associated with practical problems
gained can be used to assist those NSOs who want to begin
implementing best practices due to a lack of data sources,
using scanner data. Practical experience can be shared on
inappropriate methods, or resource constraints. The emerg-
assessing the quality of scanner data, information technol-
ing and rapidly evolving technology and the availability of
ogy and software issues, costs and benefits of using scanner
new electronic data sources such as scanner data and web
data, and organizational and legal issues (such as access to
data also pose particular challenges to countries.
data, or cooperation with scanner data providers). Research
The CPI research topics are listed in the following sec-
and testing are required on expanding the use of scanner
tion grouped under data sources, expenditure weights, index
data to include other goods, such as electronics and other
compilation methods, services, and synergies between CPIs,
items with high model turnover, and services. The develop-
purchasing power parities (PPPs), and the national accounts.
ment of methodologies and frameworks around scanner data
The organization of the topics does not reflect their impor-
is needed, for instance, to ensure consistency in coverage or
tance. Many of the CPI research topics are linked to the three
in the treatment of replacements. There is also no generally
priority areas of work of the research agenda of the System
agreed quality assurance framework developed, particularly
of National Accounts (SNA): digitalization, economic well-
for scanner data. Scanner data offer opportunities to both
being and sustainability, and globalization. These provide
countries with developed statistical systems and countries
major challenges to official statistics, including the CPI, and
with less developed statistical systems.
are therefore listed at the end of the CPI research agenda,
highlighting possible challenges involved for the CPI. Web Prices
Countries, organizations, and researchers are encouraged Internet purchases continue to grow in importance. Tradi-
to contribute to developing further guidance by undertak- tional outlets have established an online presence, and there
ing work on the research topics and sharing experiences and are a growing number of internet-based outlets. There is
good practices in these areas. The research topics also aim a need to expand price collection and the outlet sample to
to guide future work of the Expert Group on Consumer Price include these new types of outlets in the CPI. First, for those
Indices, jointly organized by the United Nations Economic outlets that have both an online and physical presence, there
Commission for Europe and the International Labour Orga- is a need to determine whether the prices charged and the
nization and the Ottawa Group on Price Indices. The Inter- price changes are the same or different. Second, there is a
secretariat Working Group on Price Statistics1 has agreed to need to identify and sample the web-based outlets and the
take the lead to ensure these issues are addressed. goods and services available. Weights consistent with the
scope of the CPI need to be estimated for the web-based
outlets. A particular important issue emerges when products
1
The Intersecretariat Working Group on Price Statistics works to develop from physical outlets are replaced by products purchased on
international standards and recommendations in price statistics, document the internet. In such cases, there is a need to decide if price
best practice guidelines, and support their implementation. The member differences should be treated as genuine price changes, and
organizations of the Intersecretariat Working Group on Price Statistics are included in the CPI, or as a change in quality, and hence
the Statistical Office of the European Union (Eurostat), the International
Labour Organization, the International Monetary Fund, the Organisation not included as price changes in the CPI. Experiences with
for Economic Co-operation and Development, the United Nations Eco- the collection of data from internet-based outlets should be
nomic Commission for Europe, and the World Bank. shared to help develop guidance.
455
APPENDIX
Issues to be addressed when collecting prices on the Expenditure Weights

web include calculation formulas for the CPI; analysis of
the performance over time of price indices using prices Plutocratic versus Democratic Weighting
from the internet and comparing with indices based on Some NSOs have initiated projects to produce democratic
traditional sources; maintenance of the sample; treatment indices. Other NSOs have attempted to compare the evo-
of replacements and quality changes; classification of the lution of prices as they affect the average household and
data; and combination with other data sources used in the average consumption. More clarity regarding the con-
the CPI. There have been discussions around using and ceptual basis of such an index and calculation methods,
weighting various data sources. Obtaining expenditure but also interpretation and use of democratic indices, is
weights for web-based prices continues to be a challenge, needed.
and there is no obvious way of obtaining this information.
More research and practical experience in this area would Price Updating of Expenditure Weights
be useful. It is left to the NSO to decide whether to price update the
expenditure weights. Since price updating may have a sig-
Web Scraping nificant influence on the overall CPI this may negatively
Web scraping has huge potential for CPI compilation with affect the international comparability of CPIs. It would be
regard to real-time access to a large amount of informa- useful to discuss empirical studies that assess the magni-
tion, but also raises technical and potential legal issues and tude of the effect of price updating and to identify whether
challenges. The drivers for the change toward web scrap- clearer guidance could be developed.
ing include competition from other providers of alterna-
tive measures of price changes, the possibility of producing Index Compilation
timelier and more frequent CPIs, more efficient production,
and reduction of costs. There is a need for exchanging soft- Calculating Elementary Price Indices
ware experiences associated with web scraping and other The expenditure weights of the elementary aggregate and
methods for collecting or harvesting data. While there are the elementary price indices are the building blocks of the
advantages to web scraping, it also poses a number of issues CPI. In the absence of weighting information, most coun-
that NSOs should be aware of, including the practical and tries compile the elementary price indices as unweighted
legal issues concerning access to internet price data; how to averages of the individual prices (or price ratios) that
ensure good cooperation with the owners of these data; and belong to the elementary aggregate. However, with the
vulnerability, including becoming too dependent on one or growing availability of data sources, it may be possible
a few owners of data and ways to deal with this and reduce to obtain information that can be used when aggregating
possible risks. Countries may develop in-house software or individual price observations into elementary price indices.
buy this from a provider of software for web scraping. Both Weighting information may be derived from a variety of
ways have their advantages and disadvantages that countries sources and may be used for groups of observations. For
must consider. example, weights may be available for particular product
groups within an elementary index, for certain outlets or
Administrative Data outlet chains, or for individual observations. Applying such
In addition to scanner data and web prices, future work on detailed weights has the potential to improve the statistical
data sources should also address the use of administrative quality of the elementary price indices. It would be useful
data sources. Depending on national circumstances, data on to share practical experiences on applying weights for the
different goods and services may be available from admin- calculation of elementary indices, including on data sources
istrative sources, for example in the areas of transport, and index calculation formulas.
housing, social protection, health, and education. Sharing
of experiences and good practices will be helpful to further Quality Adjustment
investigate the potentials of administrative data for compila- Quality adjustment is a cross-cutting issue that continues to
tion of the CPI. pose challenges. Quality-adjustment issues have been noted
previously, but there is a general need to provide better guid-
Credit Card and Bank Data ance on the treatment of quality changes. In particular, NSOs
As economies move away from cash transactions, it may continue to struggle with measuring quality changes for
be possible in the future to access household transaction clothing, cars, telecommunication equipment, multipurpose
records from banks, credit card companies, or providers of information technology devices, computers, and, in general,
payment apps for mobile telephones (often referred to as products with high churn. Sharing of practical experiences
mobile money). For this information to be useful for statisti- implementing methods and best practices is needed.
cal purposes, each transaction would need to have an identi-
fying product code associated with it. At present, only rough Seasonal Products
descriptions of the purchases are available, but this more The treatment of seasonal products continues to confound
limited information would still be useful to construct house- compilers. There is a need for research and discussion to
hold expenditure data by major category of expenditure. improve how seasonal items are treated in the CPI. It may
While the use of credit card and other payment information also be useful to further discuss and compare different ver-
involves issues concerning access to the data and confiden- sions of fixed weights and variable weights approaches with
tiality, this is a promising future source for household data the objective to develop clearer guidelines on the treatment
on purchases. of seasonal products.
456
APPENDIX
The Use of Target Price Indices for the issues and user relations: How to identify and meet user
Consumer Price Index needs, and how to deal with reactions from users?
It is useful to have a measurable target index for the CPI. The
target index can be used as guidance for deciding calculation Classification of Individual Consumption
methods and practices for the regular CPI and for measuring According to Purpose (COICOP) 2018
potential bias. Empirical research can address issues includ- The update of COICOP has significant implications for
ing identifying potential target indices (for example, Walsh, compilers. Guidance on how to proceed with implementing
Fisher, Törnqvist, or Constant Elasticity of Substitution the COICOP 2018 while preserving a historic time series of
[CES]) and how to apply these formulas in practice. data would be helpful.
Formula for Calculating Higher-Level Price Uncertainty Measures

Indices It is important to inform users about the accuracy and the
Arithmetic aggregation is used by almost all countries for reliability of a CPI and to guide CPI compilers in allocat-
calculation of higher-level price indices. Is this the best ing resources for CPI compilation in the most efficient way.
solution? What are the alternatives, with regard to geomet- Often complex and nonprobability sampling programs are
ric aggregation or aggregation by other types of averages or used to obtain prices from multiple data sources. In such a
by use of indices that apply explicit estimates of substitution context, there is no generally agreed approach for estimating
elasticities, such as the CES/Lloyd–Moulton price index? sampling errors. Further studies and practical approaches to
measure uncertainty and to assess errors in a CPI are there-
The Use of Long-Term and Short-Term Links fore needed.
The long-term/short-term link approach has been used in
Sweden for many years and has now also been adopted by Services
the United States. This method facilitates calculating the
long-term links of the CPI by the use of superlative index Measuring Services in the Consumer Price Index
number formulas. The use of long-term and short-term Defining and pricing services continue to create challenges
links was mentioned in the paper Estimating the Benefits for CPI compilers. Additional research and sharing of prac-
and Costs of New and Disappearing Products (Diewert and tical experiences on how to define and price various services
Feenstra 2017)2 as the best way to get around the problem are needed to better identify best practices and develop
of dated weighting information. A growing number of coun- guidance. Key issues include how to identify suitable units
tries will begin to explore this approach. Research, discus- of services that can be priced over time, use of input or out-
sion, and practical experiences can be shared to identify the put measures (for example, using the hourly salary of the
advantages and disadvantages of this method. service provider or following the prices of a basket of repre-
sentative services provided), and how to adjust for changes
Retrospective Calculations of Superlative Price in the quality of services. Digital platform services also have
Indices grown significantly and raise questions on how these should
Retrospective calculations of superlative price indices are be incorporated into the CPI. Timeliness is also an issue.
very useful for analytical purposes and to serve as a bench- Digital platform services have emerged quickly and often
mark to assess the quality of the CPI and quantify poten- NSOs incorporate these services into the CPI with a consid-
tial bias. A limited number of NSOs have begun to compile erable delay. The treatment of telecommunication services
superlative price indices. A sharing of experience can be in the CPI continues to create issues for compilers. All these
used to develop best practices. issues raise user concerns that the CPI becomes less rep-
resentative and reliable. Where relevant, work on services
Consumer Price Indices for Different Groups should be coordinated with the Voorburg Group on Service
and Geographic Areas Statistics.
CPIs for different population groups or geographic areas
may be required to meet specific user needs, for instance, Insurance and Financial Services
indices used for compensation purposes for specific popu- Insurance and financial services continue to pose mea-
lation groups (for example, wage earners in urban areas, surement issues for the CPI. With the update of the Clas-
receivers of retirement benefits, and so on) or for other sification of Individual Consumption According to Purpose
political or analytical purposes. Since such indices may be (COICOP) to the 2018 version (COICOP 2018), a separate
produced as part of the regular CPI, this would increase the division (Division 12) was created for insurance and finan-
efficiency of the resources used to produce CPIs. Issues in cial services. The net versus gross approaches have not
the compilation of CPIs for specific income and population been fully reconciled. There are also problems of choosing
groups and geographic areas include sources and methods appropriate deflators for the payment of premiums. More
for developing weights, sample of outlets and items/varieties discussion and research are needed to guide compilers on
collecting prices, or defining the scope of the index. Discus- the appropriate measurement of these services.
sions should also address communication and dissemination
Owner-Occupied Housing (OOH)
The measurement of OOH and its inclusion in the CPI con-
tinues to present a major challenge that requires discus-
2
https://www.unece.org/fileadmin/DAM/stats/documents/ece/ces/ge.22/ sion to formulate clearer guidance. The inclusion of OOH,
2018/University_of_British_Columbia.pdf. representing a large share of household consumption, also
457
APPENDIX
involves conceptual and measurement issues. Country- Co-operation and Development Statistics Working Paper,
specific circumstances can create problems with different “Measuring Consumer Inflation in a Digital Economy” (Rein-
approaches for the measurement of OOH (for example, sdorf and Schreyer 2019),3 discusses in detail the problems
small and concentrated rental markets). Informal housing is in measuring the welfare effects of the digital economy,
another issue. Further empirical work representing a wide including the effects of services provided for free (or at least
range of countries with different housing market types and without any direct payment) on the internet. According to
access to different data sources would be very beneficial. this working paper, there is a significant need to identify
how CPIs can better reflect and incorporate digital goods
Synergies Between Consumer Price and services, and work is needed to clarify the conceptual
Indices, Purchasing Power Parities issues and develop methods that better measure the digi-
tal economy in the CPI. These issues include, but are not
(PPPs), and National Accounts limited to, defining and identifying the goods and services,
CPIs and PPPs serve different purposes. CPIs measure price including different types of internet purchases, services for
changes over time, generally month to month, primarily at free, and shared economy services.
a national level. PPPs, on the other hand, measure price dif-
ferences across space, generally countries, at a given point Well-Being and Sustainability
in time. The temporal and spatial focus of CPIs and PPPs Measures of well-being attract much interest from poli-
has resulted in different production infrastructures for the cymakers, media, and the public. Ongoing research has
two indices. However, there are important synergies to be focused on how time allocation information is needed to
captured through the integration of data collection and other better measure household welfare change (Stiglitz and oth-
processing activities, especially given the implementation of ers, Report on the Measurement of Economic Performance
the rolling survey approach in the International Comparison and Social Progress, 2009).4 As described previously, the
Programme. Both CPIs and PPPs require basic price data Reinsdorf and Schreyer working paper not only focuses on
for final consumption goods and services, so economies of the welfare effects of the digital economy but also outlines
scale may be realized through a joint data collection process, key problems in using the CPI in measuring economic well-
as well as common data processing and quality assurance being in general. The paper identifies three reasons why the
procedures. Examples of relevant topics include but are not CPI will overestimate the cost of living and hence under-
limited to: country experiences with establishing synergies estimate progress in real welfare: (1) insufficient adjust-
between PPP and CPI data collection activities; harmoni- ment for quality changes; (2) delayed inclusion of truly
zation of classification programs for PPPs and CPIs; estab- new products; and (3) disregarding the appearance and use
lishing overlap in PPP and CPI product lists; establishing of free products. Solving these issues involves addressing
common data processing and quality assurance processes; both conceptual and practical measurement issues. In a cost
and applications in subnational price comparisons. of living index context, the theoretically correct way of
Moreover, CPI compilers, PPP compilers, and national including truly new products and products offered for free
accountants often face similar challenges when it comes would be the use of estimated “reservation” or “shadow”
to distinguishing pure price differences from differences prices. This can be done in theory and in research studies.
in quantity and quality. These problems have been exacer- However, for the regular production of the monthly CPI,
bated by recent developments related to the digitalization this is usually not feasible and other approaches must be
of our economies, such as the rapid evolution in the design implemented. According to Reinsdorf and Schreyer, the
of new products, the combination of goods and services, welfare effects would be difficult to measure, and estab-
new business models, and changes from high street shop- lished measures of gross domestic product and CPI should
ping to online purchases. It is important for CPI compilers, not be extended to include estimates of welfare effects. The
PPP compilers, and national accountants to join forces and authors conclude that if the aim is to measure welfare, it
develop consistent methodologies on all these aspects. would be better to develop special measures for this pur-
pose. The issue of a CPI for measuring economic well-
being is not restricted to the effects of digitalization, but
Links to the SNA Research Agenda also includes further discussion on the coverage of the CPI
On a broader level, the CPI research topics are linked to and the treatment of different types of goods and services
the SNA research agenda, which includes three priority provided for free, potentially including public goods and
areas of work: digitalization, well-being and sustainability, services such as education, health, safety, or parks. The
and globalization. These are briefly described in the fol- issue relates to the discussion of cost of goods indices ver-
lowing text, and possible challenges involved for the CPI sus cost of living indices, and conditional versus uncondi-
are mentioned. tional cost of living indices, where additional experiences
and guidance would be useful. It may be useful to invite
Digitalization experts from other areas of official statistics to discuss mea-
Digitalization influences the way households can acquire suring welfare and economic well-being.
and consume goods and services. Digital goods or services
may replace existing ones or constitute new goods and ser-
vices that were not available earlier. Digitalization, there-
fore, raises challenges on how to identify and incorporate 3
https://www.oecd-ilibrary.org/docserver/1d002364-en.pdf.
digital goods or services in the CPI and how to price such 4
https://ec.europa.eu/eurostat/documents/118025/118123/Fitoussi+
goods or services over time. The Organisation for Economic Commission+report.
458
APPENDIX
Globalization could include conceptual clarifications as well as practical

In the discussions about the problems in measuring glo- studies and development of guidelines on measurement.
balization, the issue of having suitable price indices (pro- There will be a need to coordinate and align with the SNA
ducer price indices, export and import price indices, or and the Balance of Payments standards, for instance on
CPIs) for the deflation of international flows of goods and how to classify different types of internet purchases as
services is often mentioned as a major challenge. Work goods or services.
459
BIBLIOGRAPHY
Abe, N., and K. Shinozaki. 2018. “Compilation of Experimental Astin, J. 1999. “The European Union Harmonized Indices of Con-
Price Indexes Using Big Data and Machine Learning: A Com- sumer Prices (HICP).” In Proceedings of the Ottawa Group
parative Analysis and Validity Verification.” Paper presented at Fifth Meeting, Reykjavik, Iceland, August 25–27, edited by R.
the UNECE/ILO Meeting of the Group of Experts on Consumer Gudnason and T. Gylfadottir. Reykjavík: Statistics Iceland; also
Price Indices, Geneva, May 7–9. https://www.unece.org. published in Statistical Journal of the United Nations ECE 16:
Abraham, K. G., J. S. Greenlees, and B. R. Moulton. 1998. “Work- 123–35. http://www.statcan.ca/secure/english/ottawagroup/.
ing to Improve the Consumer Price Index.” Journal of Eco- Astin, J. 2001. “New Uses of PPPs within the European Union.”
nomic Perspectives 12 (1): 27–36. Paper presented at the Joint World Bank–OECD Seminar on
Aczél, J. 1987. A Short Course on Functional Equations. Dor- Purchasing Power Parities: Recent Advances in Methods and
drecht: Reidel Publishing Co. Applications, Washington, DC, January 30–February 2.
Advisory Commission to Study the Consumer Price Index. 1995. Aten, B., R. Summers, and A. Heston. 2001. “An Explanation of
Toward a More Accurate Measure of the Cost of Living. Interim Stability in Country Price Structures: Implications for Spatial–
Report to the Senate Finance Committee, September 15. Wash- Temporal Comparisons.” Paper presented at the Joint World
ington, DC. Bank–OECD Seminar on Purchasing Power Parities: Recent
Aizcorbe, A. M., and P. C. Jackman. 1993. “The Commodity Sub- Advances in Methods and Applications, Washington, DC, Janu-
stitution Effect in CPI Data, 1982–91.” Monthly Labor Review ary 30–February 2.
116 (12): 25–33. Australian Bureau of Statistics. 1997. “An Analytical Framework
Aizcorbe, A., C. Corrado, and M. Doms. 2001. Constructing Price for Price Indexes in Australia.” Information Paper, Catalogue
and Quantity Indexes for High Technology Goods. Industrial No. 6421.0. http://www.abs.gov.au.
Output Section, Division of Research and Statistics. Washing- Australian Bureau of Statistics. 2003. “Australian Consumer Price
ton, DC: Board of Governors of the Federal Reserve System. Index: Concepts, Sources and Methods.” Catalogue No. 6461.0.
Aizcorbe, A., C. Corrado, and M. Doms. 2003. “When Do http://www.abs.gov.au/.
Matched-Model and Hedonic Techniques Yield Similar Price Australian Bureau of Statistics. 2009. “Consumer Price Index:
Measures?” Working Paper 2003–14, Federal Reserve Bank of Concepts, Sources and Methods.” Catalogue No. 6461.0, ABS,
San Francisco, San Francisco, CA. Canberra. http://www.abs.gov.au/ausstats/[email protected]/mf/6461.0.
Alterman, W. F., W. E. Diewert, and R. C. Feenstra. 1999. Interna- Australian Bureau of Statistics. 2013. “Feature Article: The Use of
tional Trade Price Indexes and Seasonal Commodities. Wash- Transactions Data to Compile the Australian Consumer Price
ington, DC: Bureau of Labor Statistics. Index.” Catalogue No. 6401.0, ABS, Canberra. http://www.abs.
Anderson, R. G., B. E. Jones, and T. Nesmith. 1997. “Building gov.au/.
New Monetary Services Indexes: Concepts, Data and Meth- Australian Bureau of Statistics. 2015. “Enhancing the Australian
ods.” Federal Reserve Bank of St. Louis Review 79 (1): 53–83. CPI: A Roadmap.” Catalogue No. 6401.0.60.001, ABS, Can-
Arguea, N. M., C. Haseo, and G. A. Taylor. 1994. “Estimating Con- berra. http://www.abs.gov.au/.
sumer Preferences Using Market Data: An Application to U.S. Australian Bureau of Statistics. 2016a. “Making Greater Use of
Automobile Demand.” Journal of Applied Econometrics 9: 1–18. Transactions Data to Compile the Consumer Price Index.”
Armknecht, P. A. 1996. “Improving the Efficiency of the U.S. Information Paper 6401.0.60.003, ABS, Canberra. http://www.
CPI.” Working Paper No. 96/103, International Monetary Fund, abs.gov.au/.
Washington, DC. Australian Bureau of Statistics. 2016b. “Measuring Price Change
Armknecht, P. A. 2016. “Fixed Basket Methods for Compiling of Attached Dwellings in the CPI.” Feature Article, ABS, Can-
Consumer Price Indexes.” Meeting of the Group of Experts on berra. http://www.abs.gov.au.
Consumer Price Indices, Geneva, Switzerland, May 2–4, 2016. Australian Bureau of Statistics. 2017a. “An Implementation Plan to
http://www.unece.org/index.php. Maximise the Use of Transactions Data in the CPI.” Information
Armknecht, P. A., and D. Weyback. 1989. “Adjustments for Qual- Paper 6401.0.60.004, ABS, Canberra. http://www.abs.gov.au/.
ity Change in the U.S. Consumer Price Index.” Journal of Offi- Australian Bureau of Statistics. 2017b. “Introduction of the 17th
cial Statistics 5 (2): 107–23. Series Australian Consumer Price Index.” Information Paper
Armknecht, P. A., and F. Maitland-Smith. 1999. “Price Imputation 6470.0.55.001, ABS, Canberra. http://www.abs.gov.au/AUS-
and Other Techniques for Dealing with Missing Observations, STATS/[email protected]/DetailsPage/6470.0.55.0012017.
Seasonality and Quality Change in Price Indices.” Working Bäckström, Peter, and Muhanad Sammar. 2012. “The Use of
Paper No. 99/78, International Monetary Fund, Washington, Superlative Index Links in the Swedish CPI.” Paper presented
DC. http://www.imf.org/external/pubs/ft/wp/1999/wp9978.pdf at the meeting of the Group of Experts on CPI, Geneva. https://
Armknecht, P. A., and M. Silver. 2014. “Post-Laspeyres: The Case www.unece.org/index.php.
for a New Formula for Compiling Consumer Price Indexes.” Bailey, M. J., R. F. Muth, and H. O. Nourse. 1963. “A Regres-
The Review of Income and Wealth 60 (2): 225–44. sion Method for Real Estate Price Construction.” Journal of the
Armknecht, P. A., W. F. Lane, and K. J. Stewart. (1997) 2003. “New American Statistical Association 58: 933–42.
Products and the U.S. Consumer Price Index.” In Scanner Data Baker, D. 1998. “Does the CPI Overstate Inflation? An Analysis of the
and Price Indexes, NBER Studies in Income and Wealth, edited Boskin Commission Report.” In Getting Prices Right, edited by
by R. C. Feenstra and M. D. Shapiro. Chicago, IL: University of D. Baker. Washington, DC: Economic Policy Institute, pp. 79–155.
Chicago Press, pp. 375–91. Baldwin, A. 1990. “Seasonal Baskets in Consumer Price Indexes.”
Arrow, K. J., H. B. Chenery, B. S. Minhas, and R. M. Solow. 1961. Journal of Official Statistics 6 (3): 251–73.
“Capital–Labor Substitution and Economic Efficiency.” Review Baldwin, A., A. Nakamura, and M. Prudhomme. 2010. “An Empiri-
of Economics and Statistics 63: 225–50. cal Analysis of the Different Concepts for Owned Accommodation
461
Bibliography
in the Canadian CPI: The Case of Ottawa, 1996–2005.” Paper Budget Surveys.” Journal of Economic and Social Measure-
presented at the UNECE/ILO Meeting of the Group of Experts ment 14: 19–35.
on Consumer Price Indices, Geneva, May 10–12. https://www. Balk, B. M., and W. E. Diewert. 2001. “A Characterization of the
unece.org. Törnqvist Price Index.” Economics Letters 73: 279–81.
Balk, B. M. 1980a. “Seasonal Products in Agriculture and Hor- Ball, A., and D. Fenwick. 2004. “Costing Mobile Telephone Calls:
ticulture and Methods for Computing Price Indices.” Statisti- The Use of Constrained User Profiles.” Paper presented at the
cal Studies No. 24. The Hague: Netherlands Central Bureau of Eighth Meeting of the International Working Group on Price
Statistics. Indices, Helsinki, August 23–25. http://www.ottawagroup.org.
Balk, B. M. 1980b. “Seasonal Commodities and the Construction of Bartik, T. J. 1988. “Measuring the Benefits of Land Improvements
Annual and Monthly Price Indexes.” Statistische Hefte 21: 110–6. in Hedonic Models.” Land Economics 64 (2): 172–83.
Balk, B. M. 1980c. “A Method for Constructing Price Indices for Bas, C. 2018. “Different Approaches and Methods for Strong Sea-
Seasonal Commodities.” The Journal of the Royal Statistical sonal Items in the CPI.” Paper presented at the UNECE/ILO
Society Series A 143: 68–75. Meeting of the Group of Experts on Consumer Price Indices,
Balk, B. M. 1981. “A Simple Method for Constructing Price Indi- Geneva, May 7–9. https://www.unece.org.
ces for Seasonal Commodities.” Statistische Hefte 22: 72–8. Bascher, J., and T. Lacroix. 1999. “Dishwashers and PCs in the
Balk, B. M. 1983. “Does There Exist a Relation between Inflation French CPI: Hedonic Modeling, from Design to Practice.”
and Relative Price Change Variability? The Effect of the Aggre- Paper presented at the Fifth Meeting of the International Work-
gation Level.” Economic Letters 13: 173–180. ing Group on Price Indices, Reykjavik, August 25–27. http://
Balk, B. M. 1985. “A Simple Characterization of Fisher’s Price www.ottawagroup.org.
Index.” Statistische Hefte 26: 59–63. Baxter, M. (ed.). 1998. The Retail Prices Index. Technical Manual.
Balk, B. M. 1989a. “Changing Consumer Preferences and the London: Office for National Statistics, UK.
Cost of Living Index: Theory and Nonparametric Expressions.” Bean, L. H., and O. C. Stine. 1924. “Four Types of Index Numbers
Zeitschrift für Nationalökonomie 50 (2): 157–69. of Farm Prices.” Journal of the American Statistical Association
Balk, B. M. 1989b. “On Calculating the Precision of Consumer 19: 30–5.
Price Indices.” Contributed Papers 47th Session of the ISI. Becker, G. S. 1965. “A Theory of the Allocation of Time.” Eco-
Paris: International Statistical Institute. nomic Journal 75: 493–517.
Balk, B. M. 1990. “On Calculating Cost-of-Living Index Numbers Beck-MacNeil, R. 2018. “Pricing the Internet—The Internet
for Arbitrary Income Levels.” Econometrica 58 (1): 75–92. Access Services Component of the Canadian CPI.” Statistics
Balk, B. M. 1994. “On the First Step in the Calculation of a Con- Canada. Paper presented at the UNECE/ILO Meeting of the
sumer Price Index.” Paper presented at First Meeting of the Group of Experts on Consumer Price Indices, Geneva, May
International Working Group on Price Indices, Ottawa, October 7–9. https://www.unece.org.
31–November 4. http://www.ottawagroup.org. Beidelman, C. 1973. Valuation of Used Capital Assets. Sarasota,
Balk, B. M. 1995. “Axiomatic Price Index Theory: A Survey.” FL: American Accounting Association.
International Statistical Review 63: 69–93. Beidelman, C. 1976. “Economic Depreciation in a Capital Goods
Balk, B. M. 1996a. “A Comparison of Ten Methods for Multilat- Industry.” National Tax Journal 29: 379–90.
eral International Price and Volume Comparisons.” Journal of Beisteiner, A. 2008. “Optimal Allocation of Prices in Practice—
Official Statistics 12: 199–222. Using the Neumann Formula as Tool.” Paper presented at the
Balk, B. M. 1996b. “Consistency in Aggregation and Stuvel Indi- UNECE/ILO Meeting of the Group of Experts on Consumer
ces.” The Review of Income and Wealth 42: 353–63. Price Indices, Geneva, May 8–9. https://www.unece.org.
Balk, B. M. 1998a. Industrial Price, Quantity and Productivity Bentley, A., and F. Krsinich. 2017. “Towards a Big Data CPI for
Indices. Boston, MA: Kluwer Academic Publishers. New Zealand.” Paper presented at the Fifteen Meeting of the
Balk, B. M. 1998b. “On the Use of Unit Value Indices as Consumer International Working Group on Price Indices, Eltville am
Price Subindices.” Paper presented at the Fourth Meeting of the Rhein, May 10–12. http://www.ottawagroup.org.
International Working Group on Price Indices, Washington, DC, Bernardini, A., C. De Vitiis, A. Guandalini, F. Inglese, and M. D.
April 22–24. http://www.ottawagroup.org. Terribili. 2016. “Measuring Inflation Through Different Sam-
Balk, B. M. 2000a. “Divisia Price and Quantity Indexes 75 Years pling Designs Implemented on Scanner Data.” Paper presented
After.” Draft Paper, Department of Statistical Methods, Statis- at Meeting of the Group of Experts on Consumer Price Indices,
tics Netherlands, Voorburg. Geneva, Switzerland, May 2–4.
Balk, B. M. 2000b. “On Curing the CPI’s Substitution and New Berndt, E. R. 1991. The Practice of Econometrics: Classic and
Goods Bias.” Research Paper 0005, Department of Statistical Contemporary. Reading, MA: Addison-Wesley.
Methods, Statistics Netherlands, Voorburg. Berndt, E. R., D. Ling, and M. K. Kyle. 2003. “The Long Shadow
Balk, B. M. 2001. Aggregation Methods in International Com- of Patent Expiration: Generic Entry and Rx to OTC Switches.”
parisons: What Have We Learned? Report Series Research in In Scanner Data and Price Indexes, NBER Studies in Income
Management ERS-2001-41-MKT, Erasmus Research Institute and Wealth, edited by M. Shapiro and R. C. Feenstra. Chicago,
of Management. Rotterdam: Erasmus University. IL: University of Chicago Press, pp. 229–73.
Balk, B. M. 2005. “Price Indexes for Elementary Aggregates: The Berndt, E. R., L. T. Bui, D. H. Lucking-Reiley, and G. L. Urban.
Sampling Approach.” Journal of Official Statistics 21: 675–99. 1997. “The Roles of Marketing, Product Quality and Price Com-
Balk, B. M. 2008. Price and Quantity Index Numbers. Cambridge: petition in the Growth and Composition of the U.S. Anti-Ulcer
University Press. Drug Industry.” In The Economics of New Goods, NBER Studies
Balk, B. M. 2008. Price and Quantity Index Numbers: Models for in Income and Wealth, edited by T. Bresnahan and R. J. Gordon.
Measuring Aggregate Change and Difference. New York: Cam- Chicago and London: University of Chicago Press, pp. 232–77.
bridge University Press. Berndt, E. R., Z. Griliches, and N. J. Rappaport. 1995. “Economet-
Balk, B. M. 2010. “Lowe and Cobb-Douglas Consumer Price ric Estimates of Price Indexes for Personal Computers in the
Indexes and Their Substitution Bias.” Jahrbücher für Nation- 1990s.” Journal of Econometrics 68: 243–68.
alökonomie und Statistik (The Journal of Economics and Statis- Berry, S., J. Levinsohn, and A. Pakes. 1995. “Automobile Prices in
tics) 230 (6): 726–40. Market Equilibrium.” Econometrica 63 (4): 841–90; also pub-
Balk, B. M., and H. M. P. Kersten. 1986. “On the Precision of lished as NBER Working Paper No. W4264, July 1996. http://
Consumer Price Indices Caused by the Sampling Variability of www.nber.org.
462
Bibliography
Beuerlein, I. 2001. “The German Consumer Price Index for Tele- Bresnahan and R. J. Gordon. Chicago, IL: University of Chi-
communication Services: A User Profile Approach for Mobile cago Press, pp. 237–47.
Technology and Internet Access.” Paper presented at the Sixth Breton, R., G. Clews, L. Metcalfe, N. Milliken, C. Payne, J.
Meeting of the International Working Group on Price Indices, Winton, and A. Woods. 2016. “Research Indices Using Web
Canberra, April 2–6. http://www.ottawagroup.org/. Scraped Data.” Paper presented at the UNECE/ILO Meeting of
Bhardwaj, H., T. Flower, P. Lee, and M. Mayhew. 2017. Research the Group of Experts on Consumer Price Indices, Geneva, May
Indices Using Web Scraped Price Data. August 2017 Update, 2–4. https://www.unece.org.
UK Office for National Statistics, London. Brown, C., and A. Stockburger. 2006. “Item Replacement and Qual-
Bird, D., R. Breton, C. Payne, and A. Restieaux. 2014. Initial ity Change in Apparel Price Indexes.” Monthly Labor Review
Report on Experiences with Scanner Data in ONS. Report, (December): 35–45. https://www.bls.gov/opub/mlr/2006/12/
Office for National Statistics, Newport, UK. art3full.pdf.
Bode, B., and van J. Dalén. 2001. “Quality-Corrected Price Indexes Burnett-Isaacs, K. 2015. “Using Hedonics to Create Land and
of New Passenger Cars in the Netherlands, 1990–1999.” Paper Structure Indexes for the Ottawa Condominium Market.” Paper
presented at the Sixth Meeting of the International Work- presented at the Fourteen Meeting of the International Working
ing Group on Price Indices, Canberra, April 2–6. http://www. Group on Price Indices, Tokyo, May 20–22. http://www.ottawa
ottawagroup.org. group.org.
Boettcher, I. 2015. “Automatic Data Collection on the Internet Cage, R., J. S. Greenlees, and P. Jackman. 2003. “Introducing the
(Web Scraping).” Statistic Austria. Paper presented at the Four- Chained Consumer Price Index.” In International Working Group
teen Meeting of the International Working Group on Price Indi- on Price Indices (Ottawa Group): Proceedings of the Seventh
ces, Tokyo, May 20–22. http://www.ottawagroup.org. Meeting, edited by Thierry Lacroix. Paris: INSEE, pp. 213–46.
Böhm-Bawerk, E. V. 1891. The Positive Theory of Capital. Trans- Canning, J. B. 1929. The Economics of Accountancy. New York:
lated from the original German edition of 1888 by W. Smart. The Ronald Press Co.
New York: G. E. Stechert. Carli, G.-R. 1804. “Del valore e della proporzione dei metallic
Bortkiewicz, L. V. 1923. “Zweck und Struktur einer Preisin- monetati.” In Scrittori classici italiani di economia politica,
dexzahl.” Nordisk Statistisk Tidsskrift 2: 369–408. Vol. 13. Milano: G. G. Destefanis, pp. 297–366; originally pub-
Boskin, M. J. (Chair), E. R. Dullberger, R. J. Gordon, Z. Griliches, lished in 1764.
and D. W. Jorgenson. 1996. Final Report of the Commission Carruthers, A. G., D. J. Sellwood, and P. W. Ward. 1980. “Recent
to Study the Consumer Price Index, U.S. Senate, Committee on Developments in the Retail Prices Index.” The Statistician 29:
Finance. Washington, DC: U.S. Government Printing Office. 1–32.
Boskin, M. J. (Chair), E. R. Dullberger, R. J. Gordon, Z. Griliches, Cassel, E., and R. Mendelsohn. 1985. “On the Choice of Func-
and D. W. Jorgenson. 1998. “Consumer Prices in the Consumer tional Forms for Hedonic Price Equations: Comment.” Journal
Price Index and the Cost of Living.” Journal of Economic Per- of Urban Economics 18 (September): 135–42.
spectives 12 (1): 3–26. Castles, I. 1997. The OECD–EUROSTAT PPP Program: Review of
Böttcher, I., and S. Sergeev. 2014. “Austrian Scanner Data Practice and Procedures. Paris: OECD.
Project—Report Multipurpose Consumer Price Statistics: The Cavallo, A. 2016. “Are Online and Offline Prices Similar? Evi-
Use of Scanner Data.” Paper presented at the UNECE/ILO dence from Large Multi-Channel Retailers.” Massachusetts
Meeting of the Group of Experts on Consumer Price Indices, Institute of Technology. Paper presented at the UNECE/ILO
Geneva, May 26–8. https://www.unece.org. Meeting of the Group of Experts on Consumer Price Indices,
Bowley, A. L. 1899. “Wages, Nominal and Real.” In Dictionary Geneva, May 2–4. https://www.unece.org.
of Political Economy, Volume 3, edited by R. H. I. Palgrave. Cavallo, A., and R. Rigobon. 2016. “The Billion Dollar Project.
London: Macmillan, pp. 640–51. Using Online Data for Measurement and Research.” MIT &
Bowley, A. L. 1901. Elements of Statistics. Westminster: Orchard NBER. Paper presented at the UNECE/ILO Meeting of the
House. Group of Experts on Consumer Price Indices, Geneva, May
Bowley, A. L. 1919. “The Measurement of Changes in the Cost of 2–4. https://www.unece.org.
Living.” Journal of the Royal Statistical Society 82: 343–61. Cavallo, A., and R. Rigobon. 2016. “The Billion Prices Project:
Bradley, R. 1995. “The Use of Scanner Data as a Means to Reduce Using Online Prices for Measurement and Research.” Journal
the Mean Squared Error in the CPI.” Working paper, U.S. of Economic Perspectives 30 (2): 151–78.
Department of Labor, Bureau of Labor Statistics, Washington, Caves, D. W., L. R. Christensen, and W. E. Diewert. 1982a. “The
DC. Economic Theory of Index Numbers and the Measurement of
Bradley, R. 2007. “Analytical Bias Reduction for Small Samples Input, Output and Productivity.” Econometrica 50: 1393–414.
in the U.S. Consumer Price Index.” Journal of Business & Eco- Caves, D. W., L. R. Christensen, and W. E. Diewert. 1982b. “Mul-
nomic Statistics 25: 337–46. tilateral Comparisons of Output, Input and Productivity Using
Bradley, R., B. Cook, S. E. Leaver, and B. R. Moulton. 1997. “An Superlative Index Numbers.” Economic Journal 92: 73–86.
Overview of Research on Potential Uses of Scanner Data in the Cecchetti, S. G. 1997. “Measuring Inflation for Central Bankers.”
U.S. CPI.” Paper presented at the Third Meeting of the Interna- Federal Reserve Bank of St. Louis Review 79: 143–55.
tional Working Group on Price Indices, Voorburg, April 16–18. Chessa, A. G. 2017. “Comparisons of QU-GK Indices for Dif-
http://www.ottawagroup.org. ferent Lengths of the Time Window and Updating Methods.”
Braithwait, S. D. 1980. “The Substitution Bias of the Laspey- Paper prepared for the Second Meeting on Multilateral Methods
res Price Index: An Analysis Using Estimated Cost-of-Living Organised by Eurostat, Luxembourg, March 14–15, Statistics
Indexes.” American Economic Review 70 (1): 64–77. Netherlands.
Brand, M., C. Vermeulen, D. Fischbach, and Y. Carpy. 2017. Chessa, A. G. 2020. “A Comparison of Multilateral Index Exten-
“Swiss Residential Property Price Index: The Use of Geolocal- sion Methods on Transaction Data of Seasonal Items.” Paper
ised Information for Quality Adjustment in Location.” Paper presented at the meeting of the Ottawa Group, Rio de Janeiro,
presented at the Fifteenth Meeting of the International Work- May 8–10, 2019.
ing Group on Price Indices, Eltville, May 10–12. http://www. Chessa, A. G., and R. Griffioen. 2016. “A New Methodology for
ottawagroup.org. Processing Scanner Data in the Dutch CPI.” Eurona 1: 49–69.
Bresnahan, T. F. 1997. “Comment.” In The Economics of New Chessa, A. G., J. Verburg, and L. Willenborg. 2017. “A Compari-
Goods, NBER Studies in Income and Wealth, edited by T. F. son of Price Index Methods for Scanner Data.” Paper presented
463
Bibliography
at the Fifteen Meeting of the International Working Group Dalén, J. 1998b. “Studies on the Comparability of Consumer Price
on Price Indices, Eltville am Rhein, May 10–12. http://www. Indices.” International Statistical Review 66 (1): 83–113.
ottawagroup.org. Dalén, J. 1999a. “On Reliability, Uncertainty and Bias in Con-
Christensen, L. R., and D. W. Jorgenson. 1969. “The Measurement sumer Price Indexes.” In Proceedings of the Measurement of
of U.S. Real Capital Input, 1929–1967.” Review of Income and Inflation Conference, edited by M. Silver and D. Fenwick. Car-
Wealth 15 (4): 293–320. diff: Cardiff University, pp. 184–90.
Christensen, L. R., D. W. Jorgenson, and L. J. Lau. 1971. “Con- Dalén, J. 2017. “Unit Values and Aggregation in Scanner Data—
jugate Duality and the Transcendental Logarithmic Production Towards a Best Practice.” Paper presented at the Fifteen Meet-
Function.” Econometrica 39: 255–56. ing of the International Working Group on Price Indices, Eltville
Church, A. H. 1901. “The Proper Distribution of Establishment am Rhein, May 10–12. http://www.ottawagroup.org.
Charges, Part III.” The Engineering Magazine 21: 904–12. Dalén, J., and E. Ohlsson. 1995. “Variance Estimation in the Swed-
Clements, K. W., and H. Y. Izan. 1981. “A Note on Estimating Divi- ish Consumer Price Index.” Journal of Business and Economic
sia Index Numbers.” International Economic Review 22: 745–47. Statistics 13 (3): 347–56.
Clements, K. W., and H. Y. Izan. 1987. “The Measurement of Infla- Dalton, K. V., J. S. Greenlees, and K. J. Stewart. 1998. “Incorporat-
tion: A Stochastic Approach.” Journal of Business and Eco- ing a Geometric Mean Formula into the CPI.” Monthly Labor
nomic Statistics 5: 339–50. Review 121 (10): 3–7.
Cobb, C., and P. H. Douglas. 1928. “A Theory of Production.” Davies, G. R. 1924. “The Problem of a Standard Index Number For-
American Economic Review 18: 39–165. mula.” Journal of the American Statistical Association 19: 180–88.
Cochran, W. G. 1977. Sampling Techniques, 3rd edition. New Davies, G.R. 1932. “Index Numbers in Mathematical Economics.”
York: Wiley. Journal of the American Statistical Association 27: 58–64.
Cockburn, I. M., and A. H. Anis. 1998. “Hedonic Analysis and De Gregorio, C. 2012. “Sample Size for the Estimate of Consumer
Arthritic Drugs.” Working Paper 6574, National Bureau of Eco- Price Subindices with Alternative Statistical Designs.” Rivista
nomic Research, Cambridge, MA. di statistica ufficiale N° 1/2012.
Coggeshall, F. 1887. “The Arithmetic, Geometric and Harmonic De Haan, J. 2001. “Generalised Fisher Price Indexes and the Use of
Means.” Quarterly Journal of Economics 1: 83–6. Scanner Data in the CPI.” Unpublished paper, Statistics Neth-
Combris, P., S. Lecocqs, and M. Visser. 1997. “Estimation of a erlands, Voorburg.
Hedonic Price Equation for Bordeaux Wine: Does Quality Mat- De Haan, J. 2003. “Time Dummy Approaches to Hedonic Price
ter?” Economic Journal 107 (441): 390–402. Measurement.” Paper presented at the Seventh Meeting of
Commission of the European Communities (Eurostat), International the International Working Group on Price Indices, Paris, May
Monetary Fund, Organisation for Economic Co-operation and 27–29. http://www.insee.fr/.
Development, United Nations, and World Bank. 1993. System of De Haan, J. 2004. “Estimating Quality-Adjusted Unit Value
National Accounts 1993. Brussels/Luxembourg, New York, Paris, Indexes: Evidence from Scanner Data.” Paper presented at the
Washington, DC: United Nations. SSHRC International Conference on Index Number Theory and
Congressional Budget Office (CBO). 1994. “Is the Growth of the the Measurement of Prices and Productivity, June 30–July 3,
CPI a Biased Measure of Changes in the Cost of Living?” CBO Vancouver, Canada.
Paper, October, Washington, DC. De Haan, J. 2015. “A Framework for Large Scale Use of Scanner
Crawford, A. 1998. “Measurement Biases in the Canadian CPI: An Data in the Dutch CPI.” Paper presented at the Fourteen Meet-
Update.” Bank of Canada Review (Spring): 39–56. ing of the International Working Group on Price Indices, Tokyo,
Cropper, M. L., L. L. Deck, and K. E. McConnell. 1988. “On May 20–22. http://www.ottawagroup.org.
the Choice of Functional Form for Hedonic Price Functions.” De Haan, J. 2018. “Scanner Data in the CPI: The Imputation CCDI
Review of Economics and Statistics 70 (4): 668–75. Index Revisited.” Paper presented at the 18th EMG Workshop,
Crump, N. 1924. “The Interrelation and Distribution of Prices and November 29–30, UNSW, Sydney, Australia.
Their Incidence upon Price Stabilization.” Journal of the Royal De Haan, J., and F. Krsinich. 2014. “Scanner Data and the Treat-
Statistical Society 87: 167–206. ment of Quality Change in Nonrevisable Price Indexes.” Jour-
Cunningham, A. W. F. 1996. “Measurement Bias in Price Indices: nal of Business & Economic Statistics 32: 341–58.
An Application to the UK’s RPI, Bank of England.” Working De Haan, J., and F. Krsinich. 2018. “Time Dummy Hedonic and
Paper 47, Bank of England, London. Quality-Adjusted Unit Value Indexes: Do They Really Differ?”
Curry, B., P. Morgan, and M. Silver. 2001. “Hedonic Regressions: Mis- Review of Income and Wealth 64: 757–76.
specification and Neural Networks.” Applied Economics 33: 659–71. De Haan, J., and H. A. van der Grient. 2011. “Eliminating Chain
Czinkota, M. R., and I. Ronkainen. 1997. “International Business Drift in Price Indexes Based on Scanner Data.” Journal of
and Trade in the Next Decade: Report from a Delphi Study.” Econometrics 161: 36–46.
Journal of International Business Studies 28 (4): 827–44. De Haan, J., and H. van der Grient. 2009. “Eliminating Chain Drift
Dalén, J. 1992. “Computing Elementary Aggregates in the Swedish in Price Indexes Based on Scanner Data.” Statistics Nether-
Consumer Price Index.” Journal of Official Statistics 8: 129–47. lands. https://www.unece.org/fileadmin/DAM/stats/documents/
Dalén, J. 1994. “Sensitivity Analyses for Harmonizing European ece/ces/ge.22/2010/zip.8.e.pdf.
Consumer Price Indices.” Paper presented at the First Meeting De Haan, J., E. Opperdoes, and C. Schut. 1999. “Item Selection
of the International Working Group on Price Indices, Ottawa, in the Consumer Price Index: Cut-Off versus Probability Sam-
October 31–November 4. http://www.ottawagroup.org. pling.” Survey Methodology 25 (1): 31–41.
Dalén, J. 1995. “Quantifying Errors in the Swedish Consumer De Haan, J., R. Hendriks, and M. Scholz. 2016. “A Comparison
Price Index.” Journal of Official Statistics 13 (3): 347–56. of Weighted Time-Product Dummy and Time Dummy Hedonic
Dalén, J. 1997. “Experiments with Swedish Scanner Data.” Paper Indexes.” Graz Economics Papers 2016-13, Department of Eco-
presented at the Third Meeting of the International Working nomics, University of Graz, Austria.
Group on Price Indices, Voorburg, April 16–18. http://www. De Waal, T., J. Pannekoek, and S. Scholtus. 2011. Handbook of Sta-
ottawagroup.org. tistical Data Editing and Imputation. Hoboken, NJ: John Wiley
Dalén, J. 1998a. “On the Statistical Objective of a Laspeyres Price & Sons. https://leseprobe.buch.de/images-adb/5c/b5/5cb50a55-
Index.” Paper presented at the Fourth Meeting of the Interna- cc80-4e75-821b-591e49dc38bb.pdf.
tional Working Group on Price Indices, Washington, DC, April Deaton, A. 1998. “Getting Prices Right: What Should Be Done?”
22–24. http://www.ottawagroup.org. Journal of Economic Perspectives 12 (1): 37–46.
464
Bibliography
Denny, M. 1974. “The Relationship between Functional Forms for the Diewert, W. E. 1996c. “Sources of Bias in Consumer Price
Production System.” Canadian Journal of Economics 7: 21–31. Indexes.” Discussion Paper No. 96/4, School of Economics,
Diewert, W. E. (1983a) 1990. “The Theory of the Cost of Living University of New South Wales, Sydney.
Index and the Measurement of Welfare Change.” In Price Level Diewert, W. E. 1997. “Commentary on Mathew D. Shapiro and
Measurement, edited by W. E. Diewert and C. Montmarquette. David W. Wilcox: Alternative Strategies for Aggregating Price
Ottawa: Statistics Canada, pp. 163–233; reprinted in Price Level in the CPI.” The Federal Reserve Bank of St. Louis Review 79
Measurement, edited by W. E. Diewert. Amsterdam: North-Hol- (3): 127–37.
land, pp. 79–147. Diewert, W. E. 1998a. “Index Number Issues in the Consumer Price
Diewert, W. E. 1974a. “Applications of Duality Theory.” In Fron- Index.” The Journal of Economic Perspectives 12 (1): 47–58.
tiers of Quantitative Economics, Vol. II, edited by M. D. Intri- Diewert, W. E. 1998b. “High Inflation, Seasonal Commodities and
ligator and D. A. Kendrick. Amsterdam: North-Holland, pp. Annual Index Numbers.” Macroeconomic Dynamics 2: 456–71.
106–71. Diewert, W. E. 1999a. “Index Number Approaches to Seasonal
Diewert, W. E. 1974b. “Intertemporal Consumer Theory and the Adjustment.” Macroeconomic Dynamics 3: 48–68.
Demand for Durables.” Econometrica 42: 497–516. Diewert, W. E. 1999b. “Axiomatic and Economic Approaches
Diewert, W. E. 1976. “Exact and Superlative Index Numbers.” to Multilateral Comparisons.” In International and Interarea
Journal of Econometrics 4: 114–45. Comparisons of Income, Output and Prices, edited by A. Hes-
Diewert, W. E. 1978. “Superlative Index Numbers and Consistency ton and R. E. Lipsey. Chicago, IL: University of Chicago Press,
in Aggregation.” Econometrica 46: 883–900. pp. 13–87.
Diewert, W. E. 1980. “Aggregation Problems in the Measurement Diewert, W. E. 2000. “Notes on Producing an Annual Superlative
of Capital.” In The Measurement of Capital, NBER Studies in Index Using Monthly Price Data.” Discussion Paper No. 00-08,
Income and Wealth, edited by D. Usher. Chicago, IL: University Department of Economics, University of British Columbia,
of Chicago Press, pp. 433–528. Vancouver. http://www.econ.ubc.ca.
Diewert, W. E. 1983a. “The Theory of the Output Price Index Diewert, W. E. 2001. “The Consumer Price Index and Index Num-
and the Measurement of Real Output Change.” In Price Level ber Purpose.” Journal of Economic and Social Measurement 27:
Measurement, edited by W. E. Diewert and C. Montmarquette. 167–248.
Ottawa: Statistics Canada, pp. 1049–113. Diewert, W. E. 2002a. “The Quadratic Approximation Lemma and
Diewert, W. E. 1983b. “The Treatment of Seasonality in a Cost of Decompositions of Superlative Indexes.” Journal of Economic
Living Index.” In Price Level Measurement, edited by W. E. and Social Measurement 28: 63–88.
Diewert and C. Montmarquette. Ottawa: Statistics Canada, pp. Diewert, W. E. 2002b. “Similarity and Dissimilarity Indexes: An
1019–45. Axiomatic Approach.” Discussion Paper No. 02-10, Depart-
Diewert, W. E. 1986. “Microeconomic Approaches to the Theory ment of Economics, University of British Columbia, Vancouver.
of International Comparisons, Technical.” Working Paper No. http://www.econ.ubc.ca.
53. National Bureau of Economic Research, Cambridge, MA. Diewert, W. E. 2002c. “Harmonized Indexes of Consumer Prices:
Diewert, W. E. 1992a. “Fisher Ideal Output, Input and Productivity Their Conceptual Foundations.” Swiss Journal of Economics
Indexes Revisited.” Journal of Productivity Analysis 3: 211–48. and Statistics 138 (4): 547–637.
Diewert, W. E. 1992b. “Exact and Superlative Welfare Change Diewert, W. E. 2002d. “Notes on Hedonic Producer Price Indexes.”
Indicators.” Economic Inquiry 30: 565–82. Unpublished Paper, Department of Economics, University of
Diewert, W. E. 1993a. “The Early History of Price Index Research.” British Columbia, Vancouver.
In Essays in Index Number Theory, Vol. 1, Contributions to Eco- Diewert, W. E. 2002e. “Hedonic Regressions: A Review of Some
nomic Analysis 217, edited by W. E. Diewert and A. O. Naka- Unresolved Issues.” Unpublished Paper, Department of Eco-
mura. Amsterdam: North-Holland, pp. 33–65. nomics, University of British Columbia, Vancouver.
Diewert, W. E. 1993b. “Duality Approaches to Microeconomic Diewert, W. E. 2003a. “Hedonic Regressions: A Consumer Theory
Theory.” In Essays in Index Number Theory, Vol. 1, Contribu- Approach.” In Scanner Data and Price Indexes, NBER Studies
tions to Economic Analysis 217, edited by W. E. Diewert and A. in Income and Wealth, edited by R. C. Feenstra and M. D. Sha-
O. Nakamura. Amsterdam: North-Holland, pp. 105–75. piro. Chicago, IL: The University of Chicago Press, pp. 317–48.
Diewert, W. E. 1993c. “Symmetric Means and Choice under Diewert, W. E. 2003b. “Measuring Capital.” NBER Working Paper
Uncertainty.” In Essays in Index Number Theory, Vol. 1, Con- W9526, National Bureau of Economic Research, Cambridge,
tributions to Economic Analysis 217, edited by W. E. Diewert MA.
and A. O. Nakamura. Amsterdam: North-Holland, pp. 355–433. Diewert, W. E. 2004a. “On the Stochastic Approach to Linking the
Diewert, W. E. 1993d. “Overview of Volume 1.” In Essays in Index Regions in the ICP.” Discussion Paper No. 04-16, Department
Number Theory, Vol. 1, Contributions to Economic Analysis of Economics, The University of British Columbia, Vancouver,
217, edited by W. E. Diewert and A. O. Nakamura. Amsterdam: Canada.
North-Holland, pp. 1–31. Diewert, W. E. 2004b. “The Treatment of Owner Occupied Hous-
Diewert, W. E. 1995a. “Axiomatic and Economic Approaches ing and Other Durables in a Consumer Price Index.” Paper
to Elementary Price Indexes.” Discussion Paper No. 95–01, presented at the Eighth Meeting of the International Working
Department of Economics, University of British Columbia, Group on Price Indices, Helsinki, August 23–25. http://www.
Vancouver. http://www.econ.ubc.ca. ottawagroup.org.
Diewert, W. E. 1995b. “On the Stochastic Approach to Index Num- Diewert, W. E. 2011. “Alternative Approaches to Measuring House
bers.” Discussion Paper No. 95–31, Department of Economics, Price Inflation.” Paper presented at the Twelfth Meeting of the
University of British Columbia, Vancouver. http://www.econ. International Working Group on Price Indices, Wellington, May
ubc.ca. 4–6. http://www.ottawagroup.org.
Diewert, W. E. 1996a. “Price and Volume Measures in the National Diewert, W. E., and D. A. Lawrence. 2000. “Progress in Measuring
Accounts.” In The New System of National Economic Accounts, the Price and Quantity of Capital.” In Econometrics Volume 2:
edited by J. Kendrick. Norwell, MA: Kluwer Academic Pub- Econometrics and the Cost of Capital: Essays in Honor of Dale
lishers, pp. 237–85. W. Jorgenson, edited by L. J. Lau. Cambridge, MA: The MIT
Diewert, W. E. 1996b. “Seasonal Commodities, High Inflation and Press, pp. 273–326.
Index Number Theory.” Discussion Paper 96–06, Department Diewert, W. E., and K. J. Fox. 2017. “Substitution Bias in Mul-
of Economics, University of British Columbia, Vancouver. tilateral Methods for CPI Construction Using Scanner Data.”
465
Bibliography
Discussion Paper 17-02, Vancouver School of Economics, Uni- Edgeworth, F. Y. 1923. “The Doctrine of Index Numbers Accord-
versity of British Columbia, Vancouver, BC. ing to Mr. Correa Walsh.” The Economic Journal 11: 343–51.
Diewert, W. E., and R. Feenstra. 2017. “Estimating the Benefits Edgeworth, F. Y. 1925. Papers Relating to Political Economy, Vol.
and Costs of New and Disappearing Products.” Discussion 1. New York: Burt Franklin.
Paper Series 17-10, Vancouver School of Economics, Univer- Edwards, R. 1997. “Measuring Inflation in Australia.” In Bias in
sity of British Columbia, Vancouver, BC. the CPI: Experiences from Five OECD Countries, Prices Divi-
Diewert, W. E., D. Fixler, and K. Zieschang. 2011. “The Measure- sion Analytical Series, No. 10, edited by L. M. Ducharme.
ment of Banking Services in the System of National Accounts.” Ottawa: Statistics Canada, pp. 5–12.
Paper presented at the Twelfth Meeting of the International EFQM. 2019. “European Foundation for Quality Management
Working Group on Price Indices, Wellington, May 4–6. http:// Excellence Model.” European Foundation for Quality Manage-
www.ottawagroup.org. ment, Brussels. http://www.efqm.org.
Diewert, W. E., K. J. Fox, and J. de Haan. 2016. “A Newly Identi- Ehemann, C., A. J. Katz, and B. R. Moulton. 2002. “The Chain
fied Source of Potential CPI Bias: Weekly versus Monthly Unit Additivity Issue and the U.S. National Accounts.” Journal of
Value Price Indexes.” Economics Letters 141, 169–72. Economic and Social Measurement 28: 37–49.
Diewert, W. E., K. J. Fox, and P. Schreyer. 2017. “The Digital Eichhorn, W. 1978. Functional Equations in Economics. Reading,
Economy, New Products and Consumer Welfare.” Discussion MA: Addison-Wesley Publishing Company.
Paper 17-09, Vancouver School of Economics, University of Eichhorn, W., and J. Voeller. 1976. Theory of the Price Index, Lec-
British Columbia, Vancouver, BC. ture Notes in Economics and Mathematical Systems, Vol. 140.
Dikhanov, Y. 1997. “The Sensitivity of PPP-Based Income Esti- Berlin: Springer-Verlag.
mates to Choice of Aggregation Procedures.” Unpublished Eldridge, L. P. 1999. “How Price Indexes Affect BLS Productivity
Paper, International Economics Department, World Bank, Measures.” Monthly Labor Review 122 (2): 35–46.
Washington, DC. Eltetö, O., and P. Koves. 1964. “On an Index Number Computation
Dippo, C. S., and C. A. Jacobs. 1983. “Area Sampling Redesign Problem in International Comparison.” Statisztikai Szemle 42:
for the Consumer Price Index.” In Proceedings of the Survey 507–18 (in Czech).
Research Methods Section. Alexandria, VA: American Statisti- Epple, D. 1987. “Hedonic Prices and Implicit Markets: Estimat-
cal Association, pp. 118–23. ing Demand and Supply Functions for Differentiated Products.”
Divisia, F. 1926. L’indice monétaire et la théorie de la monnaie. Journal of Political Economy 95: 59–80.
Paris: Société anonyme du Recueil Sirey. European Commission, International Monetary Fund, Organisation
Dorfman, A. H., J. Lent, S. G. Leaver, and E. Wegman. 2006. “On for Economic Co-operation and Development, United Nations,
Sample Survey Designs for Consumer Price Indexes.” Survey and World Bank. 2009. System of National Accounts 2008. New
Methodology 32: 197–216. York: United Nations.
Drechsler, L. 1973. “Weighting of Index Numbers in Multilateral Eurostat. 1993. Classification of Products by Activity in the Euro-
International Comparisons.” Review of Income and Wealth 19: pean Economic Community (CPA) (Luxembourg).
17–34. Eurostat. 2013. Compendium of HICP Reference Documents,
Drobisch, M. W. 1871a. “Ueber die Berechnung der Veränder- Methodologies & Working Papers, Economy and Finance. Lux-
ungen der Waarenpreise und des Geldwerths.” Jahrbücher für embourg: Publications Office of the European Union. https://
Nationalökonomie und Statistik 16: 143–56. ec.europa.eu/eurostat/documents/3859598/5926625/KS-RA-
Drobisch, M. W. 1871b. “Ueber einige Einwürfe gegen die in 13-017-EN.PDF/.
diesen Jahrbüchern veröffentlichte neue Methode, die Verän- Eurostat. 2016a. Handbook on Price and Volume Measures in
derungen der Waarenpreise und des Geldwerths zu berechten.” National Accounts, Manuals and Guidelines, Economy and
Jahrbücher für Nationalökonomie und Statistik 16: 416–27. Finance. Luxembourg: Publications Office of the European Union.
Ducharme, L. M. 1997. “The Canadian Consumer Price Index and https://ec.europa.eu/eurostat/documents/3859598/7152852/KS-
the Bias Issue: Present and Future Outlooks.” In Bias in the GQ-14-005-EN-N.pdf.
CPI: Experiences from Five OECD Countries, Prices Division Eurostat. 2016b. The Treatment of Cross-Border Internet Pur-
Analytical Series, No. 10, edited by L. M. Ducharme. Ottawa: chases in the HICP. Directorate C: National Accounts, Prices
Statistics Canada, pp. 13–24. and Key Indicators, Unit C-4: Price statistics, Purchasing
Duggan, J. E., and R. Gillingham. 1999. “The Effect of Errors in Power Parities, Housing Statistics. Luxembourg: Publica-
the CPI on Social Security Finances.” Journal of Business and tions Office of the European Union. http://ec.europa.eu/
Economic Statistics 17 (2): 161–69. eurostat/documents/272892/7048317/HICP-recommendation-
Dulberger, E. R. 1989. “The Application of an Hedonic Model to on-internet+purchases-December-2016/.
a Quality-Adjusted Price Index for Computer Processors.” In Eurostat. 2017a. Practical Guide for Processing Supermarket
Technology and Capital Formation, edited by D. Jorgenson and Scanner Data. Brussels, Belgium: Eurostat.
R. Landau. Cambridge, MA: MIT Press. Eurostat. 2017b. Technical Manual on Owner-Occupied Housing and
Dulberger, E. R. 1993. “Sources of Price Decline in Computer House Price Indices. Brussels, Belgium: Eurostat. https://ec.europa.
Processors: Selected Electronic Components.” In Price Mea- eu/eurostat/documents/7590317/0/Technical-Manual-OOH-
surement and Their Uses, NBER Studies in Income and Wealth, HPI-2017/.
edited by M. Foss, M. E. Manser, and A. H. Young. Chicago, IL: Eurostat. 2018. Harmonised Index of Consumer Prices (HICP)
University of Chicago Press, pp. 103–24. Methodological Manual, 2018 Edition. Luxembourg: European
Dutot, C. 1738. Réflections politiques sur les finances et le com- Commission. https://ec.europa.eu/eurostat/en/web/products-
merce, Vol. 1. La Haye: Les frères Vaillant et N. Prevost. manuals-and-guidelines/-/KS-GQ-17-015.
Dwyer, L., P. Forsyth, and D. S. Prasada Rao. 2001. “PPPs and the Feenstra, R. C. 1994. “New Product Varieties and the Measurement
Price Competitiveness of International Tourism Destinations.” of International Prices.” American Economic Review 34: 157–77.
Paper presented at the Joint World Bank–OECD Seminar on Feenstra, R. C. 1995. “Exact Hedonic Price Indices.” Review of
Purchasing Power Parities: Recent Advances in Methods and Economics and Statistics 77: 634–54.
Applications, Washington, DC, January 30–February 2. Feenstra, R. C., and W. E. Diewert. 2001. “Imputation and Price
Edgeworth, F. Y. 1888. “Some New Methods of Measuring Varia- Indexes: Theory and Evidence from the International Price
tion in General Prices.” Journal of the Royal Statistical Society Program.” Working Paper No. 335, Bureau of Labor Statistics,
51: 346–68. Washington, DC. http://www.bls.gov.
466
Bibliography
Feenstra, R. C., and M. D. Shapiro. 2003. “High Frequency Substi- Fisher, I. 1921. “The Best Form of Index Number.” Journal of the
tution and the Measurement of Price Indexes.” In Scanner Data American Statistical Association 17: 533–37.
and Price Indexes, NBER Studies in Income and Wealth, edited Fisher, I. 1922. The Making of Index Numbers. Boston, MA:
by R. C. Feenstra and M. D. Shapiro. Chicago, IL: The Univer- Houghton-Mifflin.
sity of Chicago Press, pp. 123–46. Fisher, W. C. 1913. “The Tabular Standard in Massachusetts His-
Feenstra, R. C., and C. R. Shiells. 1997. “Bias in U.S. Import tory.” Quarterly Journal of Economics 27: 417–51.
Prices and Demand.” In The Economics of New Goods, NBER Fixler, D., J. Greenlees, and W. Lane. 2001. “Telecommunications
Studies in Income and Wealth, edited by T. F. Bresnahan and Indexes in the U.S. Consumer Price Index.” Paper presented at
R. J. Gordon. Chicago, IL: University of Chicago, pp. 249–76. the Sixth Meeting of the International Working Group on Price
Feldmann, B., and L. Sandberg. 2012. “Internet Purchases—A Indices, Canberra, April 2–6. http://www.ottawagroup.org.
Challenge for the HICP.” Paper presented at the UNECE/ILO Fixler, D., and K. D. Zieschang. 1992. “Incorporating Ancillary
Meeting of the Group of Experts on Consumer Price Indices, Measures of Processes and Quality Change into a Superlative
Geneva, May 30–June 1. https://www.unece.org. Productivity Index.” Journal of Productivity Analysis 2: 245–67.
Fenwick, D. 1997. “The Boskin Report from a United Kingdom Fixler, D., and K. D. Zieshang. 2001. “Price Indices for Financial
Perspective.” In Bias in the CPI: Experiences from Five OECD Services.” Paper presented at the Sixth Meeting of the Inter-
Countries, Prices Division Analytical Series, No. 10, edited by national Working Group on Price Indices, Canberra, April 2–6.
L. M. Ducharme. Ottawa: Statistics Canada, pp. 45–52. http://www.ottawagroup.org.
Fenwick, D. 2006. “Real World Quality Measures.” Paper pre- Flux, A. W. 1921. “The Measurement of Price Change.” Journal of
sented at the Ninth Meeting of the International Working Group the Royal Statistical Society 84: 167–99.
on Price Indices, London, May 14–16. http://www.ottawagroup. Forsyth, F. G., and R. F. Fowler. 1981. “The Theory and Practice
org. of Chain Price Index Numbers.” Journal of the Royal Statistical
Fenwick, D. 2008. “Collection and Processing of Price Data: The Society A 144 (2): 224–47.
Role of the Tukey Algorithm in Validation Procedures for Prices Friedman, David M. 2016. “A New Estimation System for the US
Data in a Consumer Prices Index.” UK Office for National CPI—Capabilities and Impacts.” Presentation at the Meeting of
Statistics. Paper presented at the UNECE/ILO Meeting of the the Group of Experts on CPI, Geneva. http://www.unece.org/
Group of Experts on Consumer Price Indices, Geneva, May index.php.
8–9. https://www.unece.org. Frisch, R. 1930. “Necessary and Sufficient Conditions Regard-
Fenwick, D. 2014. “Exploiting New Technologies and New Data ing the Form of an Index Number Which Shall Meet Certain of
Sources—The Opportunities and Challenges Associated with Fisher’s Tests.” Journal of the American Statistical Association
Scanner Data.” Paper presented at the UNECE/ILO Meeting of 25: 397–406.
the Group of Experts on Consumer Price Indices, Geneva, May Frisch, R. 1936. “Annual Survey of General Economic Theory:
26–28. https://www.unece.org. The Problem of Index Numbers.” Econometrica 4: 1–38.
Fenwick, D., A. Ball, M. Silver, and P. H. Morgan. 2003. “Price Frost, S. 2001. “The Construction of Price Indices for Deposit and
Collection and Quality Assurance of Item Sampling in the Loan Facilities.” Paper presented at the Sixth Meeting of the
Retail Price Index: How Can Scanner Data Help?” In Scanner International Working Group on Price Indices, Canberra, April
Data and Price Indexes, NBER Studies in Income and Wealth, 2–6. www.ottawagroup.org/.
edited by M. Shapiro and R. Feenstra. Chicago, IL: University Frost, S. 2003. “Experimental Price Indexes for Financial Ser-
of Chicago Press, pp. 67–87. vices.” Australian Bureau of Statistics, UNECE Expert Work-
Fenwick, D., A. Brueton, and A. Ball. 2003. “Internet Retail Chan- ing Group on Consumer Price Indices, Paper presented at the
nels in Price Indices: The Challenges Involved in Including UNECE/ILO Meeting of the Group of Experts on Consumer
Non-traditional Retailers in the UK Retail Prices Index.” Paper Price Indices, Geneva, December 4–5. https://www.unece.org.
presented at meeting of the Ottawa Group, Rio de Janeiro, May Funke, H., G. Hacker, and J. Voeller. 1979. “Fisher’s Circular Test
8–10, 2019. Reconsidered.” Schweizerische Zeitschrift für Volkswirtshaft
Fenwick, D., and B. Whitestone. 2001. The Use of Handheld Com- und Statistik 115: 677–87.
puters for the Collection of CPI Price Data. London: Office for Funke, H., and J. Voeller. 1978. “A Note on the Characterization of
National Statistics (UK). Fisher’s Ideal Index.” In Theory and Applications of Economic
Ferger, W. F. 1931. “The Nature and Use of the Harmonic Mean.” Indices, edited by W. Eichhorn, R. Henn, O. Opitz, and R. W.
Journal of the American Statistical Association 26: 36–40. Shephard. Würzburg: Physica-Verlag, pp. 177–81.
Ferger, W. F. 1936. “Distinctive Concepts of Price and Purchas- Gabor, E., and P. Vermeulen. 2015. “New Evidence on Elemen-
ing Power Index Numbers.” Journal of the American Statistical tary Index Bias.” Paper presented at the Fourteen Meeting of
Association 31: 258–72. the International Working Group on Price Indices, Tokyo, May
Ferrari, G., G. Gozzi, and M. Riani. 1996. “Comparing GEKS and 20–22. http://www.ottawagroup.org.
EPD Approaches for Calculating PPPs at the Basic Heading Garcke, E., and J. M. Fells. 1893. Factory Accounts: Their Princi-
level.” In Eurostat: Improving the Quality of Price Indices: CPI ples and Practice, Fourth Edition (First Edition 1887). London:
and PPP. Luxembourg: Eurostat. Crosby, Lockwood and Son.
Ferrari, G., and M. Riani. 1998. “On Purchasing Power Parities Geary, R. C. 1958. “A Note on the Comparison of Exchange Rates
Calculation at the Basic Heading Level.” Statistica LVIII: and Purchasing Power Between Countries.” Journal of the
91–108. Royal Statistical Society (Series A) 121: 97–9.
Filer, R., and J. Hanousek. 2002. “Survey-Based Estimates of Biases Genereux, P. A. 1983. “Working Group on Price Indices.” INSEE,
in Consumer Price Indices during Transition: Evidence from Paris, May 27–29. http://www.insee.fr.
Romania.” Journal of Comparative Economics 30 (3): 476–87. Gini, C. 1931. “On the Circular Test of Index Numbers.” Interna-
Fisher, F. M., and K. Shell. 1972. “The Pure Theory of the National tional Review of Statistics 9: 3–25.
Output Deflator.” In The Economic Theory of Price Indexes. Greenlees, J. 1997. “Expenditure Weight Updates and Measured
New York: Academic Press, pp. 49–113. Inflation.” Paper presented at the Third Meeting of the Interna-
Fisher, I. 1897. “The Role of Capital in Economic Theory.” Eco- tional Working Group on Price Indices, Voorburg, April 16–18.
nomic Journal 7: 511–37. Greenlees, J. S. 2011. “Improving the Preliminary Values of the
Fisher, I. 1911. The Purchasing Power of Money. London: Chained CPI-U.” Journal of Economic and Social Measurement
Macmillan. 31 (1–2): 1–18.
467
Bibliography
Greenlees, J., and R. McClelland. 2012. “New Evidence on Outlet Income and Wealth, edited by R. C. Feenstra and M. D. Shapiro.
Substitution Effects in Consumer Price Index Data.” Review of Chicago, IL: University of Chicago Press, pp. 17–38.
Economics and Statistics 93 (2): 632–46. Haworth, M. F., D. Fenwick, and R. Beaven. 1997. “Recent Devel-
Greenlees, J., and E. Williams. 2010. “Reconsideration of Weight- opments in the UK Retail Prices Index: Quality Management.”
ing and Updating Procedures in the U.S. CPI.”Jahrbücher für Paper presented at the Third Meeting of the International Work-
Nationalökonomie und Statistik (The Journal of Economics and ing Group on Price Indices, Voorburg, April 16–18. http://www.
Statistics) 230 (6): 741–58. ottawagroup.org.
Griffioen, R., and O. Ten Bosch. 2016. “On the Use of Internet Hicks, J. R. 1940. “The Valuation of the Social Income.” Econom-
Data for the Dutch CPI.” Statistics Netherlands. Paper presented ica 7: 105–24.
at the UNECE/ILO Meeting of the Group of Experts on Con- Hicks, J. R. 1941–42. “Consumers’ Surplus and Index Numbers.”
sumer Price Indices, Geneva, May 2–4. https://www.unece.org. The Review of Economic Studies 9: 126–37.
Griliches, Z. 1988. Technology, Education and Productivity: Early Hicks, J. R. 1946. Value and Capital, 2nd edition. Oxford: Clar-
Papers with Notes to Subsequent Literature. New York: Basil endon Press.
Blackwell. Hidiroglou, M. A., and J.-M. Berthelot. 1986. “Statistical Editing
Griliches, Z. 1990. “Hedonic Price Indices and the Measurement of and Imputation for Periodic Business Surveys.” Survey Method-
Capital and Productivity: Some Historical Reflections.” In Fifty ology 12 (1): 73–83.
Years of Economic Measurement: The Jubilee of the Conference Hill, R. J. 1995. “Purchasing Power Methods of Making Interna-
on Research in Income and Wealth, NBER Studies in Income tional Comparisons.” Ph.D. dissertation, University of British
and Wealth, edited by E. R. Berndt and J. E. Triplett. Chicago, Columbia, Vancouver.
IL: University of Chicago Press, pp. 185–206. Hill, R. J. 1999a. “Comparing Price Levels across Countries Using
Guðnason. 2001. “Telecommunications Services in the CPI—A Minimum Spanning Trees.” The Review of Economics and Sta-
Quantity Approach.” Paper presented at the Sixth Meeting of tistics 81: 135–42.
the International Working Group on Price Indices, Canberra, Hill, R. J. 1999b. “Comparing Price Levels across Countries using
April 2–6. http://www.ottawagroup.org. Spanning Trees.” In: International and Inter-Area Comparisons
Guðnason. 2003. “How Do We Measure Inflation? Some Measure- of Income, Output and Prices, edited by A. Heston and R. E.
ment Problems.” Paper presented at the Seventh Meeting of Lipsey, Studies in Income and Wealth, Vol. 61. Chicago, IL:
the International Working Group on Price Indices, May 27–29, University of Chicago Press.
Paris. http://www.insee.fr. Hill, R. J. 1999c. “International Comparisons using Spanning
Hansen, Carsten Bolden. 2007. “Recalculations of the Danish CPI Trees.” In International and Interarea Comparisons of Income,
1996–2006.” Proceedings of the Tenth Meeting of the Interna- Output and Prices, NBER Studies in Income and Wealth, edited
tional Working (Ottawa) Group on Price Indexes, Ottawa, Can- by A. Heston and R. E. Lipsey. Chicago, IL: University of Chi-
ada, October 9–12. www.ottawagroup.org. cago Press, pp. 109–20.
Hardy, G. H., J. E. Littlewood, and G. Pólya. 1934. Inequalities. Hill, R. J. 1999d. “Chained PPPs and Minimum Spanning Trees.”
Cambridge: Cambridge University Press. In International and Interarea Comparisons of Income, Out-
Harper, M. J., E. R. Berndt, and D. O. Wood. 1989. “Rates of Return put and Prices, NBER Studies in Income and Wealth, edited by
and Capital Aggregation Using Alternative Rental Prices.” In A. Heston and R. E. Lipsey. Chicago, IL: Chicago University
Technology and Capital Formation, edited by D. W. Jorgenson Press, pp. 327–64.
and R. Landau. Cambridge, MA: The MIT Press, pp. 331–72. Hill, R. J. 1999e. “Comparing Price Levels Across Countries Using
Haschka, P. 2003. “Simple Methods of Explicit QA for Services Minimum Spanning Trees.” The Review of Economics and Sta-
in Complex Pricing Schemes.” Paper presented at the Seventh tistics 81: 135–42.
Meeting of the International Working Group on Price Indices, Hill, R. J. 2001. “Measuring Inflation and Growth Using Spanning
May 27–29, Paris. http://www.insee.fr. Trees.” International Economic Review 42: 167–85.
Hausman, J. A. 1997. “Valuation of New Goods under Perfect and Hill, R. J. 2002. “Superlative Index Numbers: Not All of Them Are
Imperfect Conditions.” In The Economics of New Goods, NBER Super.” Discussion Paper No. 2002/04, School of Economics,
Studies in Income and Wealth, edited by T. F. Bresnahan and University of New South Wales, Sydney.
R. J. Gordon. Chicago, IL: University of Chicago Press, pp. Hill, T. P. 1988. “Recent Developments in Index Number Theory
209–37. and Practice.” OECD Economic Studies 10: 123–48.
Hausman, J. A. 1999. “Cellular Telephone, New Products, and Hill, T. P. 1996. Inflation Accounting: A Manual on National
the CPI.” Journal of Business and Economic Statistics 17 (2): Accounting under Conditions of High Inflation. Paris: OECD.
188–94. Hill, T. P. 1998. “The Measurement of Inflation and Changes in the
Hausman, J. A. 2002. “Sources of Bias and Solutions to Bias in Cost of Living.” Statistical Journal of the United Nations ECE
the CPI.” Working Paper 9298, National Bureau of Economic 15: 37–51.
Research, Cambridge, MA. Hill, T. P. 1999. “COL Indexes and Inflation Indexes.” Paper tabled
Hausman, J. A. 2005. “Consumer Benefits From Increased Compe- at the Fifth Meeting of the International Working Group on
tition in Shopping Outlets: Measuring the Effect of Wal-Mart.” Price Indices, Reykjavik, August 25–27. http://www.ottawa-
Working Paper 11809, National Bureau of Economic Research, group.org.
Cambridge, MA. Hill, T. P. 2006. “Household Production, Consumption and CPIs.”
Hausman, J. A., and E. Leibtag. 2004. “CPI Bias from Supercent- Paper presented at the Ninth Meeting of the International Work-
ers: Does the BLS Know that Wal-Mart Exists?” Working Paper ing Group on Price Indices, London, May 14–16. http://www.
10712, National Bureau of Economic Research, Cambridge, MA. ottawagroup.org.
Hawkes, W. J. 1997. “Reconciliation of Consumer Price Index Hillinger, C. 2002. “A General Theory of Price and Quantity
Trends in Average Prices for Quasi-homogeneous Goods Using Aggregation and Welfare Measurement.” CISifo Working Paper
Scanning Data.” Paper presented at the Third Meeting of the No. 818, University of Munich, Munich.
International Working Group on Price Indices, Voorburg, April Hoffmann, J. 1998. “Problems of Inflation Measurement in Ger-
16–18. http://www.ottawagroup.org. many.” Discussion Paper 1/98, Economic Research Group of
Hawkes, W. J., and F. W. Piotrowski. 2003. “Using Scanner Data the Deutsche Bundesbank, Deutsche Bundesbank, Frankfurt.
to Improve the Quality of Measurement in the Consumer Price Hoffmann, J. 1999. “The Treatment of Quality Changes in the
Index.” In Scanner Data and Price Indexes, NBER Studies in German Consumer Price Index.” Paper presented at the Fifth
468
Bibliography
Meeting of the International Working Group on Price Indices, Bank. 2004. Consumer Price Index Manual: Theory and Prac-
Reykjavik, August 25–27. http://www.ottawagroup.org. tice. Geneva: International Labour Office.
Hoffmann, J., and C. Kurz. 2002. “Rent Indices for Housing in International Monetary Fund (IMF). 2013. “General Data Dissemi-
West Germany: 1985 to 1998.” Discussion Paper 01/02, Eco- nation System (GDDS).” Washington, DC. http://dsbb.imf.org/
nomic Research Centre of the Deutsche Bundesbank, Deutsche Applications/web/gdds/gddshome/.
Bundesbank, Frankfurt. International Monetary Fund (IMF). 2013. “Special Data Dissemi-
Holdway, M. 1999. “An Alternative Methodology: Valuing Quality nation Standard (SDDS).” Washington, DC. http://dsbb.imf.org/
Changes for Microprocessors in the PPI.” Unpublished Paper, Applications/web/sddshome.
Bureau of Labor Statistics, Washington, DC. International Monetary Fund (IMF). 1993. Balance of Payments
Horvitz, D. G., and D. J. Thompson. 1952. “A Generalization of Manual, 5th edition. Washington, DC.
Sampling without Replacement from a Finite Universe.” Jour- International Monetary Fund (IMF). 2001. Government Finance
nal of the American Statistical Association 47: 663–85. Statistics Manual. Washington, DC.
Hotelling, H. 1925. “A General Mathematical Theory of Deprecia- International Monetary Fund (IMF). 2009. Balance of Payments
tion.” Journal of the American Statistical Association 20: 340–53. and International Investment Position Manual, 6th edition.
Houthakker, H. S. 1952. “Compensated Changes in Quantities and Washington, DC.
Qualities Consumed.” Review of Economic Studies 19: 155–64. International Monetary Fund (IMF). 2014. Government Finance
Hoven, L. 1999. “Some Observations on Quality Adjustment in Statistics Manual. Washington, DC.
the Netherlands.” Unpublished Paper, Department of Consumer International Standards Organization. 1994. ISO 9000. Geneva:
Prices, Statistics Netherland, Voorburg. International Standards Organization. http://iso.ch.
Howard, A., K. Dunford, J. Jones, M. van Kints, K. Naylor, and International Standards Organization. 2000. ISO 900. Geneva:
R. Tarnow-Mordi. 2015. “Using Transactions Data to Enhance International Standards Organization. http://iso.ch.
the Australian CPI.” Paper presented at the Fourteen Meeting of Ioannidis, C., and M. Silver. 1999. “Estimating Hedonic Indices:
the International Working Group on Price Indices, Tokyo, May An Application to UK Television Sets.” Journal of Economics.
20–22. http://www.ottawagroup.org. Zeitschrift für Nationalökonomie 69 (1): 71–94.
Hulten, C. R. 1973. “Divisia Index Numbers.” Econometrica 41: Ivancic, L. 2007. “Scanner Data and the Construction of Price
1017–26. Indices.” PhD thesis, University of New South Wales, Sydney,
Hulten, C. R. 1990. “The Measurement of Capital.” In Fifty Years Australia.
of Economic Measurement: The Jubilee of the Conference on Ivancic, L., W. E. Diewert, and K. J. Fox. 2009. “Scanner Data,
Research in Income and Wealth, NBER Studies in Income and Time Aggregation and the Construction of Price Indexes.” Dis-
Wealth, edited by E. R. Berndt and J. E. Triplett. Chicago, IL: cussion Paper No. 09-09, Department of Economics, University
University of Chicago Press, pp. 119–58. of British Columbia, Vancouver, Canada.
Hulten, C. R. 1996. “Capital and Wealth in the Revised SNA.” In Ivancic, L., W. E. Diewert, and K. J. Fox. 2011. “Scanner Data,
The New System of National Accounts, edited by J. W. Kendrick. Time Aggregation and the Construction of Price Indexes.” Jour-
New York: Kluwer Academic Publishers, pp. 149–81. nal of Econometrics 161: 24–35.
Hulten, C. R. 1996. “Issues in the Measurement of Economic Depre- Ivancic, L., and K. J. Fox. 2013. “Understanding Price Variation
ciation: Introductory Remarks.” Economic Inquiry 34: 10–23. across Stores and Supermarket Chains: Some Implications for CPI
Hulten, C. R., and F. C. Wykoff. 1981a. “The Estimation of Eco- Aggregation Methods.” Review of Income and Wealth 59: 629–47.
nomic Depreciation Using Vintage Asset Prices.” Journal of Jaluzot, L., and P. Sillard. 2016. “Échantillonnage des aggloméra-
Econometrics 15: 367–96. tions de l’IPC pour la base 2015.” Document de travail F160,
Hulten, C. R., and F. C. Wykoff. 1981b. “The Measurement of INSEE.
Economic Depreciation.” In Depreciation, Inflation and the Jensen, J. L. W. V. 1906. “Sur les fonctions convexes et les iné-
Taxation of Income from Capital, edited by C. R. Hulten. Wash- galités entre les valeurs moyennes.” Acta Mathematica 8: 94–6.
ington, DC: The Urban Institute Press, pp. 81–125. Jevons, W. S. (1863) 1884. “A Serious Fall in the Price of Gold
Imai, S., E. Diewert, and C. Shimizu. 2015. “Consumer Price Index Ascertained and Its Social Effects Set Forth.” Reprinted in
Biases—Elementary Index Biases vs. Sampling Biases.” Paper pre- Investigations in Currency and Finance. London: Macmillan
sented at the 14th Meeting of the International Working Group on and Co., pp. 13–118.
Price Indices (the Ottawa Group), Tokyo, Japan, May 20–22. Jevons, W. S. 1865. “The Variation of Prices and the Value of the
International Labour Organization (ILO). 1987. Report of the Currency since 1782.” Journal of the Statistical Society of Lon-
Fourteenth International Conference of Labour Statisticians. don 28: 294–320; reprinted in Investigations in Currency and
Geneva: ILO. Finance (London: Macmillan and Co., 1884), pp. 119–50.
International Labour Organization (ILO). 1990. ISCO-88: Inter- Jevons, W. S. 1884. “A Serious Fall in the Value of Gold Ascertained
national Standard Classification of Occupations. Geneva: ILO. and Its Social Effects Set Forth. 1863.” In Investigations in Cur-
International Labour Organization (ILO). 1998. “Guidelines rency and Finance. London: Macmillan and Co., pp. 13–118.
Concerning Dissemination Practices for Labour Statistics.” In Johannessen, R. 2004. “Owner Occupied Housing in the Norwe-
Report of the Sixteenth International Conference of Labour gian CPI.” Paper presented at the Eighth Meeting of the Interna-
Statisticians. Geneva: ILO. http://www.ilo.org/public/english/ tional Working Group on Price Indices, Helsinki, August 23–25.
bureau/stat/standards/guidelines/index.htm. http://www.ottawagroup.org.
International Labour Organization (ILO). 2003. Report III to the Johansen, I., and R. Nygaard. 2012. “Various Data Collection Meth-
Seventeenth International Conference of Labour Statisticians. ods in the Norwegian CPI.” Statistics Norway. Paper presented at
Geneva: ILO. the UNECE/ILO Meeting of the Group of Experts on Consumer
International Labour Organization, International Monetary Fund, Price Indices, Geneva, May 30–June 1. https://www.unece.org.
Organisation for Economic Co-operation and Development, Euro- Johnson, D. S., S. B. Reed, and K. J. Stewart. 2006. “Price Mea-
stat, UN Economic Commission for Europe, and the World Bank. surement in the United States: A Decade after the Boskin
2004. Producer Price Index Manual. Brussels/Luxembourg; Report.” Monthly Labor Review 129: 10–9.
Geneva; Washington, DC: International Monetary Fund. Johnson, P. 2015. “UK Consumer Price Statistics: A Review.” UK
International Labour Organization, International Monetary Fund, Statistics Authority, London. https://www.statisticsauthority.
Organisation for Economic Co-operation and Development, gov.uk/archive/reports---correspondence/current-reviews/uk-
UN Economic Commission for Europe, Eurostat, and the World consumer-price-statistics---a-review.pdf.
469
Bibliography
Jorgenson, D. W. 1989. “Capital as a Factor of Production.” In Krsinich, F. 2016. “The FEWS Index: Fixed Effects with a Win-
Technology and Capital Formation, edited by D. W. Jorgenson dow Splice.” Journal of Official Statistics 32: 375–404.
and R. Landau. Cambridge, MA: The MIT Press, pp. 1–35. Krueger, A. B., and A. Siskind. 1998. “Using Survey Data to Assess
Jorgenson, D. W. 1996. “Empirical Studies of Depreciation.” Eco- Bias in the Consumer Price Index.” Monthly Labor Review 121
nomic Inquiry 34: 24–42. (4): 24–33.
Jorgenson, D. W., and Z. Griliches. 1967. “The Explanation of Pro- Lamboray, C. 2017. “The Geary Khamis Index and the Lehr Index:
ductivity Change.” Review of Economic Studies 34: 249–83. How Much Do They Differ?” Paper presented at the Fifteen
Karanka, J., R. O’Neill, N. Weaden, R. Sanderson, C. Jenkins, Meeting of the International Working Group on Price Indices,
and D. Bird. 2013. “Developing a Definitive Measure of House Eltville am Rhein, May 10–12. http://www.ottawagroup.org.
Prices in the UK.” Paper presented at the Thirteenth Meeting of Lancaster, K. J. 1966. “A New Approach to Consumer Theory.”
the International Working Group on Price Indices, Copenhagen, Journal of Political Economy 74 (2): 132–56.
May 1–3. http://www.ottawagroup.org. Lancaster, K. J. 1971. Consumer Demand: A New Approach. New
Katz, A. J. 1983. “Valuing the Services of Consumer Durables.” York: Columbia University Press.
The Review of Income and Wealth 29: 405–27. Lane, W. 2001. “Addressing the New Goods Problem in the Con-
Kennedy, P. 1998. A Guide to Econometrics. Oxford: Blackwell sumer Price Index.” Paper presented at the Sixth Meeting of the
Publishers. International Working Group on Price Indices, Canberra, April
Keynes, J. M. 1930. A Treatise on Money in Two Volumes: 1: The 2–6. http://www.ottawa.org.
Pure Theory of Money. London: Macmillan. Laspeyres, E. 1871. “Die Berechnung einer mittleren Waarenpreis-
Khamis, S. H. 1970. “Properties and Conditions for the Existence steigerung.” Jahrbücher für Nationalökonomie und Statistik 16:
of a New Type if Index Numbers.” Sankhya, Series B 32: 81–98. 296–314.
Khamis, S. H. 1972. “A New System of Index Numbers for Lau, L. J. 1979. “On Exact Index Numbers.” Review of Economics
National and International Purposes.” Journal of the Royal Sta- and Statistics 61: 73–82.
tistical Society, Series A 135: 96–121. Le Gallo, F., and F. Magnien. 2003. “Measuring Price Change in
Khamis, S. H. 1984. “On Aggregation Methods for International Mobile-Telephony Services: An Arduous Task.” INSEE, Paper
Comparisons.” Review of Income and Wealth 30 (2): 185–205. presented at the Seventh Meeting of the International Working
Knibbs, Sir G. H. 1924. “The Nature of an Unequivocal Price Group on Price Indices, Paris, May 27–29. http://www.ottawa-
Index and Quantity Index.” Journal of the American Statistical group.org.
Association 19: 42–60; 196–205. Leaver, S. G., and D. Swanson. 1992. “Estimating Variances for
Kokoski, M. F., B. R. Moulton, and K. D. Zieschang. 1999. the U.S. Consumer Price Index for 1987–1991.” In Proceed-
“Interarea Price Comparisons for Heterogeneous Goods and ings of the Survey Research Methods Section. Alexandria, VA:
Several Levels of Commodity Aggregation.” In International American Statistical Association, pp. 740–45.
and Interarea Comparisons of Income, Output and Prices, Lebow, D. E., J. M. Roberts, and D. J. Stockton. 1994. “Monetary
NBER Studies in Income and Wealth, edited by A. Heston and Policy and the ‘Price Level’.” Unpublished Paper, Board of
R. E. Lipsey. Chicago, IL: University of Chicago Press, pp. Governors of the Federal Reserve System, Washington, DC.
123–66. Lebow, D. E., and J. B. Rudd. 2003. “Measurement Error in the
Kokoski, M. F., K. Waehrer, and P. Rozaklis. 2001. “Using Hedonic Consumer Price Index: Where Do We Stand?” Journal of Eco-
Methods for Quality Adjustment in the CPI: The Consumer nomic Literature 41: 159–201.
Audio Products Component.” Working Paper 344, Bureau of Lehr, J. 1885. Beiträge zur Statistik der Preise. Frankfurt: J. D.
Labor Statistics, Washington, DC. Sauerlander.
Konüs, A. A. 1924. “The Problem of the True Index of the Cost Lent, J., and A. H. Dorfman. 2009. “Using a Weighted Average
of Living.” The Economic Bulletin of the Institute of Economic of Jevons and Laspeyres Indexes to Approximate a Superlative
Conjuncture (in Russian), 9–10: 64–71; also published in Index.” Journal of Official Statistics 25(1): 129–49.
English in 1939 in Econometrica 7: 10–29. Leontief, W. 1936. “Composite Commodities and the Problem of
Konüs, A. A., and S. S. Byushgens. 1926. “K probleme pokupatel- Index Numbers.” Econometrica 4: 39–59.
noi cili deneg.” Voprosi Konyunkturi 2: 151–72. Lequiller, F. 1997. “Does the French Consumer Price Index
Koskimäki, T., and M. Ylä-Jarkko. 2003. “Segmented Markets and Overstate Inflation?” In Bias in the CPI: Experiences from
CPI Elementary Classifications.” Paper presented at the Seventh Five OECD Countries, Prices Division Analytical Series, No.
Meeting of the International Working Group on Price Indices, 10, edited by L. M. Ducharme. Ottawa: Statistics Canada, pp.
Paris, May 27–29. http:// www.insee.fr/. 25–43.
Koskimäki, T., and Y. Vartia. 2001. “Beyond Matched Pairs and Levy, F., H. Beamish, R. J. Murnane, and D. Aurtor. 1999. “Com-
Griliches Type Hedonic Methods for Controlling Quality puterization and Skills: Example from a Car Dealership,
Changes in CPI Subindices.” Paper presented at the Sixth Meet- Brookings Program on Output and Productivity Measurement
ing of the International Working Group on Price Indices, Can- in the Services Sector.” Workshop on Measuring the Output
berra, April 2–6. http://www.ottawagroup.org. of Business Services, May 14, Brookings Institution, Wash-
Kotler, P. 1991. Marketing Management, 7th edition. Englewood ington, DC.
Cliffs, NJ: Prentice Hall. Ley, E. 2003. “Comment.” In Scanner Data and Price Indexes,
Kravis, I. B., A. W. Heston, and R. Summers. 1982. World Product NBER Studies in Income and Wealth, edited by R. C. Feenstra
and Income: International Comparisons of Real Gross Domes- and M. D. Shapiro. Chicago, IL: University of Chicago Press,
tic Product. Baltimore, MD: Johns Hopkins University Press. pp. 379–82.
Krsinich, F. 2015. “Implementation of Consumer Electronics Liegey Jr., P. R. 1992. “Adjusting Apparel Indices in the CPI for
Scanner Data in the New Zealand CPI.” Paper presented at the Quality Differences.” In Price Measurements and Their Uses,
Fourteen Meeting of the International Working Group on Price NBER Studies in Income and Wealth, edited by M. F. Foss, M.
Indices, Tokyo, May 20–22. http://www.ottawagroup.org. Manser, and A. Young. Chicago, IL: University of Chicago Press.
Krsinich, F. 2015. “Price Indexes from Online Data Using the Liegey Jr., P. R. 1994. “Apparel Price Indexes: Effects of Hedonic
Fixed-Effects Window-Splice (FEWS) index.” Statistics New Adjustments.” Monthly Labor Review 117: 38–45.
Zealand. Paper presented at the Fourteen Meeting of the Inter- Liegey Jr., P. R. 2000. “Hedonic Quality Adjustment Methods
national Working Group on Price Indices, Tokyo, May 20–22. for Microwave Ovens in the U.S. CPI.” Methodology Paper,
http://www.ottawagroup.org. Bureau of Labor Statistics, Washington, DC.
470
Bibliography
Linacre, S. 2004. “Experimental Price Indexes for Financial Ser- Moulton, B. R. 2001. “The Expanding Role of Hedonic Methods
vices, 1998 to 2003.” Information Paper 6413.0, Australian in the Official Statistics of the United States.” In Proceedings
Bureau of Statistics, Canberra, Australia. of a Symposium on Hedonic Methods. Wiesbaden: Deutches
Linder, F. 1996. Reducing Bias in the Estimation of Consumer Bundesbank and German Federal Statistical Office.
Price Indices by Using Integrated Data. Research Report, Sta- Moulton, B. R., and E. P. Seskin. 1999. “A Preview of the 1999
tistics Netherlands, Voorburg. Comprehensive Revision of the National Income and Product
Lloyd, P. J. 1975. “Substitution Effects and Biases in Nontrue Price Accounts.” Survey of Current Business 79: 6–17.
Indices.” American Economic Review 65: 301–13. Moulton, B. R., and K. E. Moses. 1997. “Addressing the Quality
Lowe, J. 1823. The Present State of England in Regard to Agricul- Change Issue in the Consumer Price Index.” Brooking Papers
ture, Trade and Finance, 2nd edition. London: Longman, Hurst, on Economic Activity 1: 305–66.
Rees, Orme and Brown. Moulton, B. R., T. LaFleur, and K. E. Moses. 1999. “Research on
Lowe, R. 1996. “The Type and Extent of Quality Changes in the Improved Quality Adjustment in the CPI: The Case of Televi-
Canadian CPI.” In Proceedings of the Second Meeting of the sions.” In Proceedings of the Fourth Meeting of the Interna-
International Working Group on Price Indices, edited by J. tional Working Group on Price Indices, edited by W. Lane.
Dalén. Stockholm: Statistics Sweden, pp. 231–49. http://www. Washington, DC: Bureau of Labor Statistics, pp. 77–99. http://
ottawagroup.org. www.ottawagroup.org.
Lowe, R. 1999. “The Use of the Regression Approach to Quality Mudgett, B. D. 1955. “The Measurement of Seasonal Movements
Change for Durables in Canada.” Paper presented at the Fifth in Price and Quantity Indexes.” Journal of the American Statis-
Meeting of the International Working Group on Price Indices, tical Association 50: 93–8.
Reykjavik, August 25–27. http://www.ottawagroup.org. Muellbauer, J. 1974. “Household Production Theory, Quality, and
Maddala, G. S. 1988. Introduction to Econometrics. New York: the ‘Hedonic Technique.’” The American Economic Review 64
Macmillan. (6): 977–94.
Malmquist, S. 1953. “Index Numbers and Indifference Surfaces.” Müller, R. 2010. “Scanner Data in the Swiss CPI: An Alternative to
Trabajos de Estadistica 4: 209–42. Price Collection in the Field.” Paper presented at the UNECE/
Malpezzi, S., L. Ozanne, and T. Thibodeau. 1987. “Microeco- ILO Meeting of the Group of Experts on Consumer Price Indi-
nomic Estimates of Housing Depreciation.” Land Economics ces, Geneva, May 10–12. https://www.unece.org.
63: 372–85. Murray, J., and N. Sarantis. 1999. “Price–Quality Relationships
Manser, M. E., and R. J. McDonald. 1988. “An Analysis of Sub- and Hedonic Price Indexes for Cars in the United Kingdom.”
stitution Bias in Measuring Inflation, 1959–85.” Econometrica International Journal of the Economics of Business 6 (1): 1–23.
56 (4): 909–30. Muth, R. F. 1966. “Household Production and Consumer Demand
Marshall, A. 1887. “Remedies for Fluctuations of General Prices.” Functions.” Econometrica 34: 699–708.
Contemporary Review 51: 355–75. Nevo, A. 2001. “New Products, Quality Changes, and Welfare
Marshall, A. 1898. Principles of Economics, 4th edition. London: Measures Computed from Estimated Demand Systems.” NBER
The Macmillan Co. Working Paper W8425, National Bureau of Economic Research,
Matheson, E. 1910. The Depreciation of Factories and Their Valu- Cambridge, MA.
ation, 4th edition. London: E. & F. N. Spon. Norberg, A. 1999. “Quality Adjustment: The Case of Clothing.” In
McClelland, R., and M. Reinsdorf. 1999. “Small Sample Bias Proceedings of the Measurement of Inflation Conference, edited
in Geometric Mean and Seasoned CPI Component Indexes.” by M. Silver and D. Fenwick. Cardiff: Cardiff University. http://
Economic Working Paper, Bureau of Labor Statistics, Washing- www.cardiff.ac.uk.
ton, DC. Norberg, A., K. Strandberg, and O. Ståhl. 2018. En modell för
McCracken, P. M., J. Tobin, and others. 1999. Measuring Prices in resursallokering inom KPI. Stockholm, Sweden: Statistics
a Dynamic Economy: Re-Examining the CPI. New York: The Sweden.
Conference Board. Nordhaus, W. D. 1998. “Quality Change in Price Indexes.” Journal
Melser, D. 2018. “Scanner Data Price Indexes: Addressing Some of Economic Perspectives 12 (1): 59–68.
Unresolved Issues.” Journal of Business & Economic Statistics Nygaard, R. 2015. “The Use of Online Prices in the Norwegian
36: 516–22. Consumer Price Index.” Statistics Norway. Paper presented at
Mendelsohn, R. 1984. “Estimating the Structural Equations of the Fourteen Meeting of the International Working Group on
Implicit Market and Household Production Functions.” Review Price Indices, Tokyo, May 20–22. http://www.ottawagroup.org.
of Economics and Statistics 66 (4): 673–77. Obst, Carl. 2000. “A Review of Bias in the CPI.” Statistical Jour-
Mendershausen, H. 1937. “Annual Survey of Statistical Tech- nal of the United Nations ECE 17: 37–58.
nique: Methods of Computing and Eliminating Changing Sea- Office for National Statistics (UK). 1998. The Retail Prices Index:
sonal Fluctuations.” Econometrica 5: 234–62. A Technical Manual. London: Office for National Statistics.
Merkel, F. K. 2000. “Addressing New Item Bias in the Producer http://www.statistics.gov.uk.
Price Indexes: a PPI Quality Improvement Initiative.” Unpub- Office for National Statistics (UK). 2014. “Consumer Price Indi-
lished Paper, Bureau of Labor Statistics, Washington, DC. ces Technical Manual.” London: Office for National Statistics.
Metcalfe, E., T. Flower, T. Lewis, M. Mayhew, and E. Rowland. http://www.statistics.gov.uk.
2016. Research Indices Using Web Scraped Data: Clustering Oi, W. Y. 1997. “The Welfare Implications of Invention.” In The
Large Datasets into Price Indices (CLIP). London: Office of Economics of New Goods, NBER Studies in Income and Wealth.
National Statistics (ONS). Chicago, IL: University of Chicago Press, pp. 109–41.
J. Mehrhoff. 2007. “A Linear Approximation to the Jevons Index.” Okamoto, M. 2001. “Mid-Year Basket Index as a Practical Approx-
in: y.d. Lippe, P.M. imation to a Superlative Index.” Paper presented at the Sixth
Mitchell, W. C. 1927. Business Cycles. New York: National Bureau Meeting of the International Working Group on Price Indices,
of Economic Research. Canberra, April 2–6. http://www.ottawagroup.org.
Moulton, B. R. 1996a. Constant Elasticity Cost-of-Living Index in Opperdoes, E. 2001. “Some Empirical Experiments with CES
Share Relative Form. Washington, DC: Bureau of Labor Statistics. Functions.” Unpublished Paper, Statistics Netherlands,
Moulton, B. R. 1996b. “Bias in the Consumer Price Index: What Voorburg.
Is the Evidence?” Journal of Economic Perspectives 10 (4): Organisation for Economic Co-operation and Development
159–77. (OECD). 1997. “Synthesis Paper on Shortcomings of the
471
Bibliography
Consumer Price Index Measure of Inflation for Economic Pol- Seminar on Purchasing Power Parities: Recent Advances in
icy Purposes.” Paper prepared for Working Party No. 1 on Mac- Methods and Applications, Washington, DC, January 30–Feb-
roeconomic and Structural Policy Analysis, ECO/CPE/WP1 ruary 2.
(97)12, September, Paris. Price Statistics Review Committee. 1961. The Price Statistics of
Organisation for Economic Co-operation and Development the Federal Government. New York: National Bureau of Eco-
(OECD). 1998. “FISM, A Note by the OECD Secretariat.” nomic Research.
Prepared for the Joint OECD/ESCAP Meeting on National Rao, D. S. Prasada. 1990. “A System of Log-Change Index Num-
Accounts—1993 System of National Accounts: Five Years On, bers for Multilateral Comparisons.” In Comparisons of Prices
Bangkok, May 4–8. and Real Products in Latin America, edited by J. Salazar Car-
Organisation for Economic Co-operation and Development rillo and D. S. Prasada Rao. Amsterdam: North Holland.
(OECD). 1999. Purchasing Power Parities and Real Expendi- Rao, D. S. Prasada. 1995. “On the Equivalence of the Generalized
tures. Paris: OECD. Country Product-Dummy (CPD) Method and the Rao-System
Organisation for Economic Co-operation and Development (OECD). for Multilateral Comparisons.” Working Paper No. 5, Centre for
2001a. Measuring Productivity: Measurement of Aggregate and International Comparisons, University of Pennsylvania, Phila-
Industry-Level Productivity Growth. Paris: OECD. delphia, PA.
Organisation for Economic Co-operation and Development (OECD). Rao, D. S. Prasada. 1997. “Aggregation Methods for International
2001b. Measuring Capital: Measurement of Capital Stocks, Con- Comparison of Purchasing Power Parities and Real Income:
sumption of Fixed Capital and Capital Services. Paris: OECD. Analytical Issues and Some Recent Developments.” In Pro-
Osgood, W. F. 1925. Advanced Calculus. New York: Macmillan. ceedings of the International Statistical Institute, 51st Session.
Paasche, H. 1874. “Über die Preisentwicklung der letzten Jahre The Hague: International Statistical Institute, pp. 197–200.
nach den Hamburger Borsennotirungen.” Jahrbücher für Rao, D. S. Prasada. 2001a. “Integration of CPI and ICP: Method-
Nationalökonomie und Statistik 12: 168–78. ological Issues, Feasibility and Recommendations.” Paper pre-
Pakes, A. 2001. “A Reconsideration of Hedonic Price Indices with sented at the Joint World Bank–OECD Seminar on Purchasing
an Application to PC’s.” Working Paper 8715, National Bureau Power Parities: Recent Advances in Methods and Applications,
of Economic Research, Cambridge, MA, revised November Washington, DC, January 30–February 2.
2001. Rao, D. S. Prasada. 2001b. “Weighted EKS and Generalized Coun-
Palgrave, R. H. I. 1886. “Currency and Standard of Value in try Product Dummy Methods for Aggregation at Basic Heading
England, France and India and the Rates of Exchange between Level and Above Basic Heading Level.” Paper presented at the
These Countries.” Memorandum submitted to the Royal Com- Joint World Bank–OECD Seminar on Purchasing Power Pari-
mission on Depression of Trade and Industry, Third Report, ties: Recent Advances in Methods and Applications, Washing-
Appendix B, pp. 312–90. ton, DC, January 30–February 2.
Parker, P. 1992. “Price Elasticity Dynamics over the Adoption Life Rao, D. S. Prasada. 2005. “On the Equivalence of Weighted Coun-
Cycle.” Journal of Marketing Research XXIX: 358–67. try-Product Dummy (CPD) Method and the Rao-System for
Pierson, N. G. 1895. “Index Numbers and Appreciation of Gold.” Multilateral Price Comparisons.” Review of Income and Wealth
Economic Journal 5: 329–35. 51: 571–80.
Pierson, N. G. 1896. “Further Considerations on Index-Numbers.” Rao, D. S. Prasada, and K. S. Banerjee. 1984. “A Multilateral Index
Economic Journal 6: 127–31. Number System Based on the Factorial Approach.” Statistische
Pigou, A. C. 1920. The Economics of Welfare. London: Macmillan. Hefte 27: 297–313.
Pike, C., B. Nimmo, and L. Mariposa. 2009. “New Zealand 2006 Rao, D. S. Prasada, C. J. O’Donnell, and E. Ball. 2000. “Tran-
and 2008 Consumers Price Index Review: Price Updating.” sitive Multilateral Comparisons of Agricultural Output and
Room Document, Eleventh Meeting of the International Work- Productivity Using Minimum Spanning Trees and Generalized
ing (Ottawa) Group on Price Indexes, Neuchatel, Switzerland, EKS Methods.” Paper presented at the Workshop on Agricul-
May. http://www.ottawagroup.org/. tural Productivity: Data, Methods, and Measures, March 9–10,
Pollak, R. A. 1975. “Subindexes of the Cost of Living.” Interna- Washington, DC
tional Economic Review 16: 135–60. Rao, D. S. Prasada, and M. Timmer. 2000. “Multilateralisation
Pollak, R. A. 1980. “Group Cost-of-Living Indexes.” American of Manufacturing Sector Comparisons: Issues, Methods and
Economic Review 70: 273–78. Empirical Results.” Research Memorandum No. GD 47, Gron-
Pollak, R. A. 1981. “The Social Cost-of-Living Index.” Journal of ingen Growth and Development Centre, Groningen.
Public Economics 15: 311–36. Rasmussen, D. W., and T. W. Zuehlke. 1990. “On the Choice of
Pollak, R. A. 1989. “The Treatment of the Environment in the Cost- Functional Form for Hedonic Price Functions.” The Review of
of-Living Index.” In The Theory of the Cost-of-Living Index, edited Economics and Statistics 72: 668–75.
by R. A. Pollak. Oxford: Oxford University Press, pp. 181–85. Reese, M. 2000. “Hedonic Quality Adjustment Methods for Col-
Pollak, R. A. (1983) 1989, 1990. “The Theory of the Cost-of- lege Textbooks for the U.S. CPI.” Methodology Paper, Bureau
Living Index.” In Price Level Measurement, edited by W. E. of Labor Statistics, Cambridge, MA. http://www.bls.gov.
Diewert and C. Montmarquette. Ottawa: Statistics Canada, pp. Reinsdorf, M. B. 1993. “The Effect of Outlet Price Differentials
87–161; reprinted in The Theory of the Cost-of-Living Index, on the U.S. Consumer Price Index.” In Price Measurement and
edited by R. A. Pollak. Oxford: Oxford University Press, pp. Their Uses, NBER Studies in Income and Wealth, edited by M.
3–52; also reprinted in Price Level Measurement, edited by W. F. Foss, M. E. Manser, and A. H. Young. Chicago, IL: Univer-
E. Diewert. Amsterdam: North-Holland, pp. 5–77. sity of Chicago Press, pp. 227–54.
Pollak, R. A. 1998. “The Consumer Price Index: A Research Reinsdorf, M. B. 1994. “Price Dispersion, Seller Substitution and
Agenda and Three Proposals.” Journal of Economic Perspec- the U.S. CPI.” Working Paper 252, Bureau of Labor Statistics,
tives 12 (1): 69–78. Washington, DC.
Popkin, J. 1997. “Improving the CPI: The Record and Suggested Reinsdorf, M. B. 1996. “Constructing Basic Component Indexes
Next Steps.” Business Economics 32 (3): 42–7. for the U.S. CPI from Scanner Data: a Test Using Data on Cof-
Prais, S. J. 1959. “Whose Cost of Living?” The Review of Eco- fee.” Working Paper 277, Bureau of Labor Statistics, Washing-
nomic Studies 26: 126–34. ton, DC.
Prennushi, G. 2001. “PPPs and Global Poverty: Strengths and Reinsdorf, M. B. 1998. “Divisia Indices and the Representative
Weaknesses.” Paper presented at the Joint World Bank–OECD Consumer Problem.” Paper presented at the Fourth Meeting of
472
Bibliography
the International Working Group on Price Indices, Washington, Sellwood, D. 2001. “Improving Quality Adjustment in Practice.”
DC, April 22–24. http://www.ottawagroup.org. Paper presented at the Sixth Meeting of the International Work-
Reinsdorf, M. B. 2003. personal communication, September 9. ing Group on Price Indices, Canberra, April 2–6. http://www.
Reinsdorf, M. B., W. E. Diewert, and C. Ehemann. 2002. “Addi- ottawagroup.org.
tive Decompositions for the Fisher, Törnqvist and Geometric Selvanathan, E. A., and D. S. Prasada Rao. 1994. Index Numbers:
Mean Indexes.” Journal of Economic and Social Measurement A Stochastic Approach. Ann Arbor, MI: University of Michigan
28: 51–61. Press.
Reinsdorf, M. B., P. Liegey, and K. Stewart. 1996. “New Ways of Shapiro, M. D., and D. W. Wilcox. 1997a. “Alternative Strategies
Handling Quality Change in the U.S. Consumer Price Index.” for Aggregating Prices in the CPI.” Federal Reserve Bank of St.
Working Paper 276, Bureau of Labor Statistics, Washington, DC. Louis Review 79 (3): 113–25.
Reinsdorf, M. B., and B. R. Moulton. 1997. “The Construction of Shapiro, M. D., and D. W. Wilcox. 1997b. “Mismeasurement in
Basic Components of Cost-of-Living Indexes.” In The Econom- the Consumer Price Index: An Evaluation.” Working Paper No.
ics of New Goods, NBER Studies in Income and Wealth, edited W5590, National Bureau of Economic Research, Cambridge,
by T. F. Bresnahan and R. J. Gordon. Chicago, IL: University of MA. http:// www.nber.org.
Chicago Press. Shephard, R. W. 1953. Cost and Production Functions. Princeton,
Reinsdorf, M. B., Quirós, G., and STA Group. 2018. Measuring the NJ: Princeton University Press.
Digital Economy. IMF Staff Report, International Monetary Fund, Shephard, R.W. 1970. Theory of Cost and Production Functions.
Washington, DC. https://www.imf.org/en/Publications/Policy- Princeton, NJ: Princeton University Press.
Papers/Issues/2018/04/03/022818-measuring-the-digital-economy. Shepler, N. 2000. “Developing a Hedonic Regression Model for
Reinsdorf, M. B., and P. Schreyer. 2017. “Measuring Consumer Refrigerators in the U.S. CPI.” Methodology Paper, Bureau
Inflation in a Digital Economy.” Paper presented at the Fifth of Labor Statistics, Washington, DC. http://www.bls.gov/cpi/
IMF Statistical Forum, Washington, DC, November 16–17. cpirfr.htm.
https://www.imf.org/en/News/Seminars/Conferences/2017/05/ Shinozaki, K. 2017. “Web-Scraped Prices Comparison Method
03/5th-statistical-forum. Adopted for the PPI in Japan.” Bank of Japan. Paper presented
Ribe, M. 2003. “Financial Services in Swedish Price Indices.” at the Fifteen Meeting of the International Working Group
Paper presented at the Seventh Meeting of the International on Price Indices, Eltville am Rhein, May 10–12. http://www.
Working Group on Price Indices, Paris, May 27–29. http:// ottawagroup.org.
www.ottawagroup.org. Shinozaki, K. 2018. “Compilation of Experimental Price Indexes
Richardson, D. H. 2003. “Scanner Indexes for the Consumer Price Using Big Data and Machine Learning: A Comparative Analy-
Index.” In Scanner Data and Price Indexes, NBER Studies in sis and Validity Verification.” Paper presented at the UNECE/
Income and Wealth, edited by R. C. Feenstra and M. D. Shapiro. ILO Meeting of the Group of Experts on Consumer Price Indi-
Chicago, IL: The University of Chicago Press, pp. 39–65. ces, Geneva, May 7–9. https://www.unece.org.
Rosen, S. 1974. “Hedonic Prices and Implicit Markets: Product Shiratsuka, S. 1999. “Measurement Errors in the Japanese Con-
Differentiation and Pure Competition.” Journal of Political sumer Price Index.” Monetary and Economic Studies 17 (3):
Economy 82: 34–49. 69–102.
Rothwell, D. P. 1958. “Use of Varying Seasonal Weights in Price Sidgwick, H. 1883. The Principles of Political Economy. London:
Index Construction.” Journal of the American Statistical Asso- Macmillan.
ciation 53: 66–77. Silver, M. 1995. “Elementary Aggregates, Micro-Indices and
Ryten, J. 1998. The Evaluation of the International Comparison Scanner Data: Some Issues in the Compilation of Consumer
Project (ICP). International Monetary Fund, Washington, DC. Price Indices.” Review of Income and Wealth 41: 427–38.
Samuelson, P. A. 1953. “Prices of Factors and Goods in General Silver, M. 1999. “An Evaluation of the Use of Hedonic Regressions
Equilibrium.” Review of Economic Studies 21: 1–20. for Basic Components of Consumer Price Indices.” Review of
Samuelson, P. A., and S. Swamy. 1974. “Invariant Economic Index Income and Wealth 45 (1): 41–56.
Numbers and Canonical Duality: Survey and Synthesis.” Amer- Silver, M. 2002. “The Use of Weights in Hedonic Regressions: The
ican Economic Review 64: 566–93. Measurement of Quality Adjusted Price Changes.” Unpublished
Särndal, C.-E., B. Swensson, and J. Wretman. 1992. Model Assisted Paper, Cardiff Business School, Cardiff University, Cardiff.
Survey Sampling. New York: Springer-Verlag. Silver, M. 2011. “An Index Number Formula Problem: The Aggre-
Schlömilch, O. 1858. “Über Mittelgrössen verschiedener Ordnun- gation of Broadly Comparable Items.” Journal of Official Sta-
gen.” Zeitschrift für Mathematik und Physik 3: 308–10. tistics 27 (4): 553–67.
Schultz, B. J. (Szulc). 1996. “Treatment of Changes in Product Silver, M. 2018. “How to Measure Property Price Indices Better.”
Quality in Consumer Price Indices.” In Proceedings of the Eurostat Review of National Accounts and Macroeconomic
Second Meeting of the International Working Group on Price Indicators (EURONA), 1/2018: 35–66.
Indices, edited by J. Dalén. Stockholm: Statistics Sweden, pp. Silver, M., and S. Heravi. 2001a. “Scanner Data and the Mea-
209–29. http://www.ottawagroup.org. surement of Inflation.” The Economic Journal 111 (June):
Schultz, B. J. (Szulc). 1999. “Effects of Using Various Macro- F384–F405.
Index Formulae in Longitudinal Price and Comparisons: Silver, M., and S. Heravi. 2001b. “Hedonic Price Indices and the
Empirical Studies.” In Proceedings of the Fourth Meeting of Matched Models Approach.” Unpublished Paper, Cardiff Busi-
the International Working Group on Price Indices, edited by W. ness School, Cardiff University, Cardiff.
Lane. Washington, DC: Bureau of Labor Statistics, pp. 236–49. Silver, M., and S. Heravi. 2002. “Why the CPI Matched Models
http://www.ottawagroup.org. Method May Fail Us.” Working Paper 144, European Central
Schultze, C. L., and C. Mackie. (eds.). 2002. At What Price? Con- Bank, Frankfurt.
ceptualizing and Measuring Cost-of-Living and Price Indices. Silver, M., and S. Heravi. 2003. “The Measurement of Quality
Washington, DC: National Academy Press. Adjusted Price Changes.” In Scanner Data and Price Indexes,
Schut, C., and others. 2002. The Use of Scanner Data from Super- NBER Studies in Income and Wealth, edited by R. C. Feenstra
markets in the Consumer Price Index [in Dutch]. Statistics and M. D. Shapiro. Chicago, IL: University of Chicago Press,
Netherlands, Voorburg, The Netherlands. pp. 277–316.
Scrope, G. P. 1833. Principles of Political Economy. London: Silver, M., and S. Heravi. 2005. “A Failure in the Measurement
Longman, Rees, Orme, Brown, Green and Longman. of Inflation: Results from a Hedonic and Matched Experiment
473
Bibliography
Using Scanner Data.” Journal of Business & Economic Statis- by J. Eatwell, M. Milgate, and P. Newman. London: Macmillan,
tics 23: 269–81. pp. 630–34.
Silver, M., and S. Heravi. 2007. “Why Elementary Price Index Triplett, J. E. 1990. “Hedonic Methods in Statistical Agency Envi-
Number Formulas Differ: Price Dispersion and Product Hetero- ronments: An Intellectual Biopsy.” In Fifty Years of Economic
geneity.” Journal of Econometrics 140(2): 874–83. Measurement: The Jubilee of the Conference on Research in
Solomons, D. 1961. “Economic and Accounting Concepts of Income and Wealth, NBER Studies in Income and Wealth, edited
Income.” The Accounting Review 36: 374–83. by E. R. Berndt and J. E. Triplett. Chicago, IL: University of
Solow, R. M. 1957. “Technical Change and the Aggregate Produc- Chicago Press, pp. 207–38.
tion Function.” Review of Economics and Statistics 39: 312–20. Triplett, J. E. 1997. “Current Status of the Debate on the Consumer
Statistics Sweden. 2001. Swedish Consumer Price Index: A Hand- Price Index in the U.S.” In Bias in the CPI: Experiences from
book of Methods. Stockholm: Statistics Sweden. Five OECD Countries, Prices Division Analytical Series, No. 10,
Stiglitz, J. E., Sen, A., and Fitoussi, J. P., 2009. Report by the Com- edited by L. M. Ducharme. Ottawa: Statistics Canada, pp. 53–60.
mission on the Measurement of Economic Performance and Triplett, J. E. 1999. “The Solow Productivity Paradox: What Do
Social Progress. Commission on the Measurement of Economic Computers Do to Productivity?” Canadian Journal of Econom-
Performance and Social Progress. ics 32(2): 309–34.
Stoevska, V., and E. Namukasa. 2012. “Methodologies of Com- Triplett, J. E. 2001. “Should the Cost-of-Living Index Provide the
piling Consumer Price Indices.” 2012 ILO Survey of Country Conceptual Framework for a Consumer Price Index?” The Eco-
Practices. https://www.ilo.org/ilostat-files/SSM/SSM1_NEW/ nomic Journal 111: F311–F334.
CPI%20Summary%20report%202013.pdf. Triplett, J. E. 2002. Handbook on Quality Adjustment of Price
Stone, R. 1956. Quantity and Price Indexes in the National Indexes for Information and Communication Technology Prod-
Accounts. Paris: OECD. ucts. Draft, OECD Directorate for Science, Technology and
Summers, R. 1973. “International Price Comparisons Based Upon Industry. Paris: OECD.
Incomplete Data.” Review of Income and Wealth 19 (1): 1–16. Triplett, J. E. 2003. “Using Scanner Data in Consumer Price
Sundgren, B. 1993. “Statistical Metainformation Systems Prag- Indexes: Some Neglected Conceptual Considerations.” In
matics, Semantics, and Syntatctics.” Statistical Journal of Scanner Data and Price Indexes, NBER Studies in Income and
the United Nations Economic Commission for Europe 10 (2): Wealth, edited by R. C. Feenstra and M. D. Shapiro. Chicago,
121–42. IL: University of Chicago Press, pp. 151–62.
Szulc, B. J. (Schultz). 1964. “Index Numbers of Multilateral Regional Triplett, J. E. 2006. Handbook on Hedonic Indexes and Quality
Comparisons” (in Polish). Przeglad Statysticzny 3: 239–54. Adjustments in Price Indexes: Special Application to Informa-
Szulc, B. J. (Schultz). 1983. “Linking Price Index Numbers.” tion Technology Products. Paris: OECD Publishing.
In Price Level Measurement, edited by W. E. Diewert and C. Trivedi, P. K. 1981. “Some Discrete Approximations to Divisia
Montmarquette. Ottawa: Statistics Canada, pp. 537–66. Integral Indices.” International Economic Review 22: 71–7.
Szulc, B. J. (Schultz). 1987. “Price Indices below the Basic Aggre- Turvey, R. 1979. “The Treatment of Seasonal Items in Consumer
gation Level.” Bulletin of Labour Statistics 2: 9–16. Price Indices.” In Bulletin of Labour Statistics, Fourth Quarter.
Tauchen, H., and A. D. Witte. 2001. “Estimating Hedonic Mod- Geneva: ILO, pp. 13–33.
els; Implications of the Theory.” Technical Working Paper No. Turvey, R. 1995. “Three Kinds of Monthly CPI.” Paper presented
271. National Bureau of Economic Research, Cambridge, MA. at the Second Meeting of the International Working Group
http://www.nber.org. on Price Indices, Stockholm, November 15–17. http://www.
Teekens, R., and J. Koerts. 1972. “Some Statistical Implications of ottawagroup.org.
the Log Transformations of Multiplicative Models.” Economet- Turvey, R. 1996. “Elementary Aggregate (Micro) Indexes.” Paper
rica 40 (5): 793–819. presented at the Eurostat Seminar on Improving the Quality of
Tellis, G. J. 1988. “The Price Elasticity of Selective Demand: A Price Indices: CPI and PPP, Florence, December 18–20, 1995.
Meta-Analysis of Econometric Models of Sales.” Journal of Turvey, R. 1998. “New Outlets and New Products.” In Proceed-
Marketing Research 25: 167–77. ings of the Third Meeting of the International Working Group
Theil, H. 1954. Linear Aggregation of Economic Relations. on Price Indices, edited by B. Balk. Voorburg: Statistics Neth-
Amsterdam: North-Holland. erlands, pp. 97–110.
Theil, H. 1967. Economics and Information Theory. Amsterdam: Turvey, R. 1999. “Incorporating New Models into a CPI: PCs as
North-Holland. an Example.” In Proceedings of the Measurement of Inflation
Törnqvist, L. 1936. “The Bank of Finland’s Consumption Price Conference, edited by M. Silver and D. Fenwick. Luxembourg,
Index.” Bank of Finland Monthly Bulletin 10: 1–8. London, Cardiff: Eurostat, Office for National Statistics, Cardiff
Törnqvist, L., and E. Törnqvist. (1937) 1981. “Vilket är förhål- University. http://www.cardiff.ac.uk.
landet mellan finska markens och svenska kronans köpkraft?” Turvey, R. 2000. “True Cost of Living Indexes.” In Proceedings of
Ekonomiska Samfundets Tidskrift 39: 1–39; reprinted in Col- the Fifth Meeting of the International Working Group on Price
lected Scientific Papers of Leo Törnqvist (Helsinki: The Indices, edited by R. Gudnason and D. Gylfadóttir. Reykjavik:
Research Institute of the Finnish Economy), pp. 121–160. Statistics Iceland. http://www.ottawagroup.org.
Trajtenberg, M. 1989. Economic Analysis of Product Innovation: Turvey, R., and others. 1989. Consumer Price Indices: An ILO
The Case of CT Scanners. Cambridge, MA: Harvard University Manual. Geneva: ILO.
Press. U.S. Bureau of Labor Statistics. 1983. “Changing the Home
Triplett, J. E. 1981. “Reconciling the CPI and the PCE Deflator.” Ownership Component of the Consumer Price Index to Rental
Monthly Labor Review (September):3–15. Equivalence.” In CPI Detailed Report. Washington, DC: U.S.
Triplett, J. E. 1983. “Concepts of Quality in Input and Output Bureau of Labor Statistics.
Price Measures: A Resolution of the User-Value Resource-Cost U.S. Bureau of Labor Statistics. 1997. BLS Handbook of Methods,
Debate.” In The U.S. National Income and Product Accounts: Bulletin 2490. Washington, DC: U.S. Bureau of Labor Statistics.
Selected Topics, NBER Studies in Income and Wealth, edited U.S. Bureau of Labor Statistics. 1998. “Measurement Issues in
by M. F. Foss. Chicago, IL: University of Chicago Press, pp. the Consumer Price Index.” Statistical Journal of the United
269–311. Nations ECE 15: 1–36.
Triplett, J. E. 1987. “Hedonic Functions and Hedonic Indices.” In U.S. Bureau of Labor Statistics. 2001. “Direct Pricing of Health
The New Palgrave: A Dictionary of Economics, Vol. 2, edited Insurance in the Consumer Price Index.” Ottawa Group, 2001.
474
Bibliography
Paper presented at the Sixth Meeting of the International Work- Vartia, Y. O. 1976. Relative Changes and Index Numbers. Helsinki:
ing Group on Price Indices, Canberra, April 2–6. http://www. The Research Institute of the Finnish Economy.
ottawagroup.org. Vartia, Y. O. 1978. “Fisher’s Five-Tined Fork and Other Quantum
U.S. Bureau of Labor Statistics. 2015. Handbook of Methods, Con- Theories of Index Numbers.” In Theory and Applications of
sumer Price Indexes. Washington, DC: U.S. Bureau of Labor Economic Indices, edited by W. Eichhorn, R. Henn, O. Opitz,
Statistics. https://www.bls.gov/opub/hom/pdf/homch17.pdf. and R. W. Shephard. Würzburg: Physica-Verlag, pp. 271–95.
U.S. General Accounting Office. 2000. “Consumer Price Index: Verbrugge, R. 2008. “The Puzzling Divergence of Rents and User
Update of Boskin Commission’s Estimate of Bias.” Report GAO/ Costs, 1980–2004.” Review of Income and Wealth, Series 54,
GGD-00–50, U.S. General Accounting Office, Washington, DC. Number 4, Bureau of Labor Statistics, Washington, DC.
U.S. Senate, Committee on Finance. 1996. Final Report of the Viglino, L. 2003. “Insurance and Quality Adjustment: Excess and
Advisory Commission to Study the Consumer Price Index, Print Option-cost Method.” Paper presented at the Seventh Meeting
104–72, 104 Cong., 2nd Session. Washington, DC: Government of the International Working Group on Price Indices, Paris, May
Printing Office. 27–29. http://www.ottawagroup.org.
United Nations Economic Commission for Europe, International Ville, J. 1946. “The Existence-Conditions of a Total Utility Func-
Labour Office, International Monetary Fund, Organisation for tion” (in French). Translated in 1951 in The Review of Economic
Economic Development and Co-operation, Statistical Office Studies 19: 123–8.
of the European Communities, The World Bank, and the (UK) Vogt, A. 1977. “Zum Indexproblem: Geometrische Darstellung
Office for National Statistics. 2009. Practical Guide to Produc- sowie eine neue Formel.” Schweizerische Zeitschrift für Volk-
ing Consumer Price Indices. Geneva and New York: United swirtschaft und Statistik 113: 73–88.
Nations. http://www.unece.org/index.php. Vogt, A. 1978. “Divisia Indices on Different Paths.” In Theory
United Nations. 1990. “International Standard Industrial Classifi- and Applications of Economic Indices, edited by W. Eichhorn,
cation of All Economic Activities.” Statistical Papers Series M, R. Henn, O. Opitz, and R. W. Shephard. Würzburg: Physica-
No. 4, Rev. 3, United Nations, New York. Verlag, pp. 297–305.
United Nations. 1992. Handbook of the International Comparison Vogt, A. 1980. “Der Zeit und der Faktorumkehrtest als ‘Finders of
Program, Series F, No. 62. New York: United Nations. Tests.’” Statistische Hefte 21: 66–71.
United Nations. 1994. Fundamental Principles of Official Sta- Vogt, A., and J. Barta. 1997. The Making of Tests for Index Num-
tistics, Adopted by the UN Statistical Commission. UN Eco- bers. Heidelberg: Physica-Verlag.
nomic and Social Council, 1994, Report of the Special Session Von Auer, L. 2001. “An Axiomatic Checkup for Price Indices.”
of the Statistical Commission, New York, April 11–15, 1994, Working Paper No. 1/2001, Faculty of Economics and Manage-
E/1994/29. New York: United Nations. ment, Otto von Guericke University, Magdeburg.
United Nations. 1998a. “Principles and Recommendations for Von Auer, L. 2002. “Spurious Inflation: The Legacy of Laspeyres
Population and Housing Censuses.” Revision 1, Statistical and Others.” The Quarterly Review of Economics and Finance
Papers Series M, No. 67/Rev. 1, Sales No. E.98.XVII.8, United 42: 529–42.
Nations, New York. Von Auer, L. 2014. “The Generalized Unit Value Index Family.”
United Nations. 1998b. “Central Product Classification. CPC.” Review of Income and Wealth 60: 843–61.
Version 1.0, Statistical Papers Series M, No. 77, Ver. 1.0, United Von der Lippe, P. 2001. Chain Indices: A Study in Price Index The-
Nations, New York. ory, Publication Series Spectrum of Federal Statistics, Vol. 16.
United Nations. 1999. “Classifications of Expenditure Accord- Wiesbaden: Statistisches Bundesamt.
ing to Purpose.” Statistical Papers Series M, No. 84, United Walras, L. 1954. Elements of Pure Economics. Translated from
Nations, New York. French by W. Jaffe. London: George Allen and Unwin; first
United Nations. 2002. International Standard Industrial Classifi- published in 1874.
cation of All Economic Activities, ISIC, Revision 3.1. New York: Walschots, J. 2011. “The Treatment of Internet Purchases in Con-
United Nations Statistical Division. sumer Price Statistics, Harmonising Their Treatment in the
Van der Grient, H. A., and J. de Haan. 2010. “The Use of Super- HICP (European Commission, Eurostat).” Paper presented at
market Scanner Data in the Dutch CPI.” Paper presented at the the Twelfth Meeting of the International Working Group on
UNECE/ILO Workshop on Scanner Data, May 10, Geneva, Price Indices, Wellington, May 4–6. http://www.ottawagroup.
Switzerland. org.
Van der Grient, H., and J. de Haan. 2011. “Scanner Data Price Walsh, C. M. 1901. The Measurement of General Exchange Value.
Indexes: The ‘Dutch’ Method versus Rolling Year GEKS.” New York: Macmillan and Co.
Paper presented at the Twelfth Meeting of the International Walsh, C. M. 1921a. The Problem of Estimation. London: P. S.
Working Group on Price Indices, Wellington, May 4–6. http:// King & Son.
www.ottawagroup.org. Walsh, C. M. 1921b. “Discussion.” Journal of the American Statis-
Van Haren, G., F. Kleima, and J. de Haan. 2006. “The CPI and the tical Association 17: 537–44.
New Dutch Health Care System.” Paper presented at the Ninth Walsh, C. M. 1932. “Index Numbers.” In Encyclopedia of the
Meeting of the International Working Group on Price Indices, Social Sciences, Vol. 7, edited by E. R. A. Seligman. New York:
London, May 14–16. http://www.ottawagroup.org. The Macmillan Co., pp. 652–8.
Van Ijzeren, J. 1987. “Bias in International Index Numbers: a Math- Ward, M. 2001. “True Comparisons in Real and Money Terms.”
ematical Elucidation.” Dissertation for the Hungarian Academy Paper presented at the Joint World Bank–OECD Seminar on
of Sciences, Koninklijke Bibliotheek, The Hague. Purchasing Power Parities: Recent Advances in Methods and
Van Loon, K., and D. Roels. 2018. “Integrating Big Data in the Applications, Washington, DC, January 30–February 2.
Belgium CPI.” Belgium STATBEL. Paper presented at the Westergaard, H. 1890. Die Grundzüge der Theorie der Statistik.
UNECE/ILO Meeting of the Group of Experts on Consumer Jena: Fischer.
Price Indices, Geneva, May 7–9. https://www.unece.org. White, A. G. 1999. “Measurement Biases in Consumer Price
Van Mulligen, P. H. 2003. “Quality Aspects in Price Indices Indexes.” International Statistical Review 67 (3): 301–25.
and International Comparisons: Applications of the Hedonic White, A. G. 2000. “Outlet Types and the Canadian Consumer
Method.” Ph.D. thesis, University of Groningen, Groningen. Price Index.” Canadian Journal of Economics 33: 488–505.
http://www.cbs.nl/en/publications/articles/general/theses/the- Wold, H. 1944. “A Synthesis of Pure Demand Analysis, Part 3.”
ses.htm. Skandinavisk Aktuarietidskrift 27: 69–120.
475
Bibliography
Wold, H. 1953. Demand Analysis. New York: John Wiley. Young, A. 1812. An Inquiry into the Progressive Value of Money
Wooldridge, J. M. 1996. “Estimating Systems of Equations with in England as Marked by the Price of Agricultural Products.
Different Instruments for Different Equations.” Journal of Piccadilly: Hatchard.
Econometrics 74: 387–405. Yule, G. U. 1921. “Discussion of Mr. Flux’s Paper.” Journal of the
Woolford, K. 1999. “Measuring Inflation: A Framework Based on Royal Statistical Society 84: 199–202.
Domestic Final Purchases.” In Proceedings of the Measurement Zarnowitz, V. 1961. “Index Numbers and the Seasonality of Quan-
of Inflation Conference, edited by M. Silver and D. Fenwick. tities and Prices.” In The Price Statistics of the Federal Gov-
Cardiff: Cardiff University, pp. 534–43. ernment, G. J. Stigler (Chair). New York: National Bureau of
Woolford, K. 2001. “Financial Services in the Consumer Price Economic Research, pp. 233–304.
Index.” Paper presented at the Sixth Meeting of the Interna- Zhang, L., I. Johansen, and R. Nygaard. 2019. “Tests for Price
tional Working Group on Price Indices, Canberra, April 2–6. Indices in a Dynamic Item Universe.” Journal of Official Statis-
http://www.ottawagroup.org. tics, 35 (3): 683–97.
Wynne, M. A. 1997. “Commentary.” Federal Reserve Bank of St. Zieschang, K. D. 1988. “The Characteristics Approach to the Prob-
Louis Review 79 (3): 161–7. lem of New and Disappearing Goods in Price Indexes.” Work-
Wynne, M. A. 1999. “Core Inflation: A Review of Some Concep- ing Paper No. 183, Bureau of Labor Statistics, Washington, DC.
tual Issues.” Research Department Working Paper 99-03, Fed- Zieschang, K. D., P. A. Armnecht, and D. Smith. 2001. “Integrated
eral Reserve Bank of Dallas, Dallas, TX. InterArea and International Price Comparisons with Consumer
Wynne, M. A., and F. D. Sigalla. 1994. “The Consumer Price Price Index Compilation.” Paper presented at the Joint World
Index.” Federal Reserve Bank of Dallas Economic Review, Bank–OECD Seminar on Purchasing Power Parities: Recent
Second Quarter. Dallas: Federal Reserve Bank of Dallas, Advances in Methods and Applications, Washington, DC, Jan-
pp. 1–22. uary 30–February 2.
476
INDEX
abbreviations xvii–xviii block pricing 264

access by the public 331 bonus offers 79–80
acquisition Boskin Commission 304, 305
definition 435 bulletin 330–331
and uses 35–36 bundled goods and services 49
acquisition approach 3 Division 12 of 50
housing 251–253
measuring owner-occupied housing 253–254 calculating consumer price indices 175
property insurance services 286 alternatives to fixed-weight indices 194–196
acronyms xvii–xviii axiomatic approach to elementary price indices 180–181
additivity 446 calculation of elementary price indices 177–180
administrative data calculation of elementary price indices using weights 189–190
research 456 calculation of geographic and national indices 194–198
sources 58–59 calculation of higher-level indices 190–191
advertised price, unavoidable costs not part of 81 chained vs direct indices for elementary aggregates 182
advisory committees 333 consistency in aggregation 182–183
aggregation, consistency in 182–183 construction of elementary aggregates 175–177
agriculture producer price index (APPI) 51 economic approach to elementary price indices 181–182
allocations, sampling 74–75 higher-level price indices as weighted averages of elementary
arithmetic average 21 price indices 192–193
audit report template 323 missing price observations 183–189
audit schedule, pro forma for 321 recommendations 198
audits of production processes 312–314 reference periods 191
audit report 313 typical calculation methods for higher-level indices 193–194
longer-term annual reviews 314 see also consumer price index (CPI)
monitoring performance 314 capital services 45
objectives of 313 Carli index 21, 22, 126, 179, 181
overview 312 definition 178
procedures and techniques 313 elementary aggregates 177
responsibilities of audit members 312–313 formula 2, 198
review systems 313–314 indices using relatives with missing prices 184
risk assessment 313 permanently disappeared varieties 186
role of audit team 312–313 replacing varieties with no overlapping prices 187
short-term monthly reviews 314 carryforward imputation, missing prices 129
automated data checking 109–111 cash back 80
catalog prices 78
back-checking, local price collection 91–92 central price collection 12, 78
backward imputation 134 central shops weights 8
backward linking factor 210, 211 chain drift 227, 228
bank-specific credit card offers 81 chain index, calculating 199–200
bargaining, price 81–83 chaining 162–163
base invariance 445 bilateral price indices and 227–228
base period sample account 293 change(s)
basic index number, formulas and terminology 453–454 analysis of contributions to 328
basket index 37, 191 large price 85–86
basket price index 453 link-to-show-no-price-change 135, 136, 148
benchmarking 309 quality change 163–164, 256, 303–305
bias 26 seasonal adjustment and smoothing of index 328
components of 302–305 tariff, for Internet prices 268
definition 301 time-series presentation of level and 327–328
elementary aggregate bias 303 see also quality change
new outlet 305 characteristicity 231, 445, 448
procedures to minimize 306 characteristics approach, rapid turnover of models 151
quality change and new goods bias 303–305 choice of item identifier 232
recommendations 306 Classification of Individual Consumption According to Purpose
summary of bias estimates 305–306 see COICOP
types of 301–306 classification systems, consumer price index 6, 7, 48–50
upper-level substitution 302 class mean imputation
bilateral price indices 227–228 quality adjustment 134–136
477
Index
rapid turnover of models 150 calculating 21–24

seasonal products 237–238 classifications 48–50
clothing classification structure 6, 7
seasonal products 237–238, 241 concepts of 34–38
see also second-hand clothing conceptual basis of weights 53–55
COICOP (Classification of Individual Consumption According to data 294
Purpose) 6, 7, 29, 441 Data Quality Assessment Framework (DQAF) 307–308,
aggregation structure for CPI 176 317
disaggregation of 49 data sources 9–10, 57–59
financial services 279–280 definition 53
international standard 48–49 deriving weighting pattern 7–8
introducing new classification systems 218–219 deriving weights in practice 59–61
key changes from 1999 to 2018 49–50 developing 1
partial weight updates at COICOP class level 214 economic commentary and interpretation of 328–329
subaggregates 329 errors and biases 26
transport services 273 formula notations 27
weighting structure 55 frequency of weight updates 61–62
COICOP 1999 (Classification of Individual Consumption geographic coverage 4–5
According to Purpose 1999) imputed transactions and imputed prices 6
alcoholic beverages, tobacco and narcotics 348–349 inclusion of newly significant products 64
clothing and footwear 349 indexation 2
communication 355 inflation measurement 2
education 358 international standards of xii
food and nonalcoholic beverages 347–348 introducing new weights 24
furnishings, household equipment and household maintenance items requiring special treatment 62–64
351–353 linking previous CPI to new reference period 24–25
health 353–354 model quality report document 325
housing, water, electricity, gas and other fuels 350–351 national accounts deflation 2
insurance 360 organization and management 25
miscellaneous goods and services 359–360 overview of uses and needs 1–2
recreation and culture 355–358 owner-occupied housing 64
restaurants and hotels 358–359 planning and organizing price collection 105
social protection 360 popularity for economic analysis 33–34
transport 354–355 press release 330
COICOP 2018 (Classification of Individual Consumption price collection form 107
According to Purpose 2018) 361 price statistics in national accounts 51
alcoholic beverages, tobacco and narcotics 370–372 product exclusions 6
breakdown of consumption expenditure 361–425 publication and dissemination of 25–26
clothing and footwear 372–375 recommendations 50, 64
education 414–416 reference population for 5–6
food and nonalcoholic beverages 361–370 resolution of Seventeenth International Conference of Labour
furnishings, household equipment and routine maintenance Statisticians 427–434
380–387 scope of 38–48
health 387–395 seasonal adjustment and smoothing of 328
housing, water, electricity, gas and other fuels 375–380 special cases 26
implementation of 50 taxonomy of errors in 297
information and communication 401–405 time-series presentation 327–328
insurance and financial services 419–421 type of index number formula 2–3
personal care social protection and miscellaneous goods uses 29–34
421–425 weighting structure 8–9
recreation, sport and culture 405–414 weighting structure of 55–57
restaurants and accommodation services 417–419 weight reference period 61
transport 395–401 see also calculating consumer price indices; updating CPI
collection schedule price 86 weights
commensurability test 180 consumer profiles approach
comparable replacements 131–132 example 266
compensation index 262 mobile telephones 267
computer-assisted data collection (CADC) price collection 100–101 price measurement 265
confidentiality 332 quality adjustment 268
congestion pricing 263 consumer substitution, definition 435
construction producer price index (CPPI) 51 consumption, deflation 31–32
consumer goods, definition 435 consumption expenditures, definition 435
consumer price(s) contractual payments, indexation of 31
permanently unavailable 121 core inflation 329
temporarily unavailable 121 cost of living index (COLI) 2, 37–38
consumer price index (CPI) xi–xii counter seasonal imputation 237
acquisition, use and payment approach 3–4 country-product-dummy (CPD) method 447
automated data checking 109–111 weighted method 447
478
Index
country-product-representativity-dummy method 447 education 277–279

country target index 195 discounting forms 279
coverage, definition 435 pricing 278
coverage of medical insurance, health services 276–277 scholarships 279
credibility checking, price collection 97 electronic reporting 100
credit card 80 elementary aggregate bias 303
fees 283–284 elementary aggregate index 166, 168
Internet purchases 245 elementary aggregates
research of bank data and 456 aggregation structure 176
currency exchange 280–281 calculating price indices for 175–190
current cost accounting 32–33 chained vs direct indices for 182
customer profiles, telecommunications 271–272 construction of 175–177
cutoff sampling 67 formula for estimating 208
definition 435 weights within 177
elementary index error 439
data elementary index formulas 2–3
access to 332 elementary price indices 21–22
timeliness of release vs accuracy of 332 aggregation of 197, 198
data editing 13–14 axiomatic approach to 180–181
price collection 93–95 calculation of 177–180
data output checking 97–98 calculation of chain index 199
Data Quality Assessment Framework (DQAF) 307–308, calculation of using weights 189–190
317 economic approach to 181–182
data validation formulas applicable to scanner data 190
in field 89–90 formulas for 189–190
field supervisors 90 sampling properties of 179–180
price collection 93–95 unit value indices 190
visual 96 enhanced General Data Dissemination System (e–GDDS) 307,
debit card fees 283–284 315, 331, 332
deflation 31–32 errors 26
delivery charges, Internet purchases 245 estimation 299
democratic weights 54–55 estimation of variance 298–299
CPI 5–6 measurement 300
deposit and loan facilities 282–283 measuring 298–299
deposit product, calculation of price index for 293–295 nonobservation errors 298
digitalization 458 nonsampling 297–298
direct extension method 231 procedures to minimize 299–301
disaster recovery, price collection 99–100 processing 298
discount(s) 48 recommendations 306
discounted prices 79 response 298
discounting, education 279 sampling 297
dissemination 25–26, 233 types of 297–298, 439
international standards for, of CPI 331 estimation error 297, 299
recommendations 333–334 European Foundation for Quality Management (EFQM)
timeliness of release vs data accuracy 332 Excellence Model 309
timing of, of CPI 331–332 European Union (EU) 236, 328
Dissemination Standards Bulletin Board (DSBB) 315 Harmonised Index of Consumer Prices (HICP) 33, 343–354
documentation Eurostat-Organisation for Economic Co-operation and
control 311–312 Development 446
control template 113, 319 evolutionary new products 164, 165–166, 169–170
overview 311 forced replacement 166
domestic concept 37, 243, 244 sample augmentation 166
domestic coverage, consumer price index (CPI) 4 sample rotation 165
dual pricing 80–81 sample rotation in high-turnover groups 165–166
Dutot index 21, 22, 67, 126, 178, 179, 198 updating and chaining the index 165
definition 177 evolutionary product 19, 20
elementary aggregates 177 expenditure
formula 2, 136 domestic and national concept of 36–37
indices using averages with imputed prices 185 gambling 43
indices using averages with missing prices 184 gifts and subscriptions 42
indices using relatives with imputed prices 185 goods and services 34–35
permanently disappeared varieties 186 insurance 43
replacing varieties with no overlapping prices 188 licenses 42
use of 180, 181 social security contributions and taxes 42
dynamic universe 227 tips and gratuities 42
transactions in financial assets 43–44
economic analysis, popularity of CPIs for 33–34 transfers 42
economic policy purpose, CPI for 30 unimportant or difficult-to-measure products 60–61
479
Index
expenditure weights GEKS (Gini-Eltetö-Köves-Szulc) method 228–229, 231,

conceptual basis of 53–55 445–446, 448
Internet purchases 242–243 general consumption index 262
Expert Group on Consumer Price Indices 172 Generic Statistical Business Process Model (GSBPM) 309
explicit quality adjustment 230, 437 geographic approach
export price index (XPI) 51 consumer price index (CPI) 4–5, 196
extras 79–80 seasonal products 240
geographic coverage 39, 53
feature/option costs, differences in 138–139 expanding CPI 219–220
fee schedule 293 HICP 343
fees of agents and brokers 46 Internet purchases 243–244
field supervisors, data validation 90 geometric averages 21–22
final consumption 260, 336 geometric Laspeyres index 23
financial intermediation services indirectly measured (FISIM) geometric Paasche index 23, 229
279–280, 282 geometric Young index 23, 193–195
financial services 279–280 gifts and subscriptions 42
COICOP (1999) 360 globalization 459
COICOP (2018) 419–421 Global Trade Item Numbers (GTINs) 298, 304–305
credit and debit card fees when abroad 283–284 glossary 335–342
currency exchange 280–281 goods and services, undesirable, informal or illegal 46–47
deposit and loan facilities 282–283 gross fixed capital formation 260
introduction 279–280 Group of Twenty (G20) 33
investment funds 282 Guidelines concerning dissemination practices for labor statistics
property insurance services 285–287 331
real estate agency services 284–285
stockbroking services 281–282 Handbook on Hedonic Price Indexes and Quality Adjustments in
Fisher price index 3, 23–24, 177, 190–191, 194–195, 200–202, Price Indexes (OECD) 18
227–229, 302–303, 305, 335, 338, 342, 431, 445, 457 handheld computers, price collection 100–101
fixed basket, principle of 12–13 harmonic mean of price relatives 189
fixed-basket index 191, 192 Harmonised Indices of Consumer Prices (HICP) 14, 33, 36, 39,
fixed basket principle 83 46, 236, 315, 327
fixed-weight approach, seasonal products 235, 236–237 coverage 343–344
fixed-weight indices 201 definition of 200
annual chaining 194 European Union (EU) 343–354
arithmetic vs geometric aggregation for 194–195 household final monetary consumption expenditure 343
index formulas 194 index formulas 345
long-term and short-term links calculating CIP 195–196 methodological manual 345
retrospective superlative indices 195 prices 344–345
follow-up training, price collection 98–99 quality adjustment 345
foreign currency release, timeliness and revisions 345
denominated prices 81 sampling 344
price 81 weights 344
purchases and sales of 43 health services 274–275
formula(s) COICOP (1999) 353–354
aggregation, for elementary price indices 124–125 COICOP (2018) 387–395
basic index number 453–454 coverage of medical insurance 276–277
Carli index 2, 198 doctors and dentists 275–277
Dutot index 2, 136 education 277–279
elementary index 2–3 health services (doctors and dentists) 275–277
estimating higher-level indices 208 introduction 274–275
elementary price indices 189–190 pricing 276
for estimating elementary aggregates 208 sampling 275–276
formula for estimating 208 social protection 279
HICP index formulas 345 weights 277
Jevons index 2, 22, 127, 177, 198 hedonic approach
Laspeyres index 200, 345 equivalences of 152
notations 27 implementation of 142–144
type of index number formula 2–3 imputation 151–152
Young index 192 interpretation of estimated hedonic coefficients 142
forward imputation 134 limitations of the 144–146
forward linking factor 210, 211 patching 139–142, 148
free gifts 79–80 quality adjustment 139–146
free products 48 rapid turnover of models 150–152
fruits and vegetables, seasonal products 240 regression imputation of new variety's price 143, 144
Fundamental Principles of Official Statistics (UN) 331 regression imputation of old variety's price 144
hedonic geometric mean characteristics 151
gambling 43 hedonic geometric mean imputation 151
Geary-Khamis (GK) method 119, 229, 448 hedonic indices 162–163
480
Index
hedonic price indices of social security benefits 31

matched indices and 152–153 type of index for 30
rapid turnover of models 150–152 of wages and pensions 30–31
time dummy variable approach 150–151 index compilation, research 456–457
Hidiroglou and Berthelot method 95 index dispersion report 93
higher-level indices index number formula, consumer price index concepts
calculation of 190–191 2–3
calculation of chain index 199–200 index reference period 191, 200, 335
formula for estimating 208 deciding on 207–208
typical calculation methods for 193–194 definition 435
as weighted averages of elementary price indices 192–193 keeping the old 211, 212
hire purchase 43, 44 indices for current accounts 295
household actual final consumption 35 inflation
definition 34 accounting under 32–33
household budget surveys (HBSs) 37, 53 consumer price indices and 29–30
adjusting 60 core 329
as data source 57–58 international comparisons of 33
household expenditure measurement 2
discounts, rebates, loyalty programs and “free” products 48 perceptions 33–34
fees of agents and brokers 46 inputs 45
imputed expenditure on goods and services 47–48 inquiry management system, price collection 86–87
luxury goods and services 47 institutional households 5, 38
second-hand goods 47 insurance 43, 63
undesirable, informal or illegal goods and services 46–47 HICP 344
household final consumption expenditure (HFCE) 32, 34, 35, 39 integrity 34, 331
household final monetary consumption expenditure (HFMCE) 35, interest, indexation of 31
36, 47, 343, 344 interest data 294
household production 45–46 interest payments 44
households intermediate consumption 260, 336
scope of CPI 38–39 International Comparison Program (ICP) 32
see also own-account production aggregation at basic heading level 446–450
housing aggregation methods for spatial comparisons 444–445
accounting for missing observations 255–256 benefits from CPI and ICP integration 450–451
acquisitions approach 251–253 data requirements for comparisons 444
calculating rent index 255 differences between temporal and spatial comparisons 443–444
depreciation, major improvements and quality change 256 index number methods for spatial comparisons 446
introduction to 245 notation and conceptual framework 445–446
measuring owner-occupied, under acquisitions approach purchasing power parities (PPPs) 443–445, 447–449
253–254 subnational PPPs 450–451
owner-occupied services 245–254 International Conference of Labour Statisticians (ICLS) xii
payments approach 248–251 resolution (17th) concerning CPIs adopted by 427–434
prices and rent surveys 254–255 International Labour Organization (ILO) xi, 307, 331
rented accommodation 254–256 Resolution 6, 14
services 45, 260, 261 Resolution (2003) 235, 256
use approach 246–248 Survey (2012) 298
weights 254 International Monetary Fund (IMF) xi, 53, 307, 315, 331
see also own-account production Internet 73
hyperinflation, price collection 85–86 coherence and data integration 242
determining location of online retailer 243–244
ICP see International Comparison Program (ICP) estimating expenditure weights for 242–243
Iklé-Kidhanov-Balk (IDB) method 448 geographical coverage 243–244
impact checking 97–98 introduction to 241–242
implicit quality adjustment 229–230, 437 national vs domestic concept 243
import price index (MPI) 51 price collection 244
imputation 126 purchases 63, 241–245
class mean 19 treatment of additional costs for 244–245
missing products 16 Intersecretariat Working Group on Price Statistics (IWGPS) xi,
overall mean 19 xii, xiii
term 34 intersection universe 161
imputed expenditure on goods and services 47–48 investment funds 282
imputed expenditures, definition 435 ISO 9001, 309–310
imputed prices 6, 435 item replacement 161–162
imputed transactions 6 item substitution 161–162
income, deflation of 31–32
independence 34 Jevons-Gini-Eltetö-Köves-Szulc (Jevons-GEKS) method 446
indexation 30–31 Jevons index 67, 126, 178, 179, 181, 227
of consumer price index 2 elementary aggregates 177
of interest, rents and contractual payments 31 formula 2, 22, 127, 177, 198
481
Index
indices using averages with imputed prices 185 illustrative variety codes for price collection 125
indices using averages with missing prices 184 long-term vs short-term comparisons 124
indices using relatives with imputed prices 185 multiplicative vs additive adjustment 123
indices using relatives with missing prices 184 noncomparable replacements 132
permanently disappeared varieties 186 observations 183–189
replacing varieties with no overlapping prices 187 overall mean imputation 126–128, 155–157
permanently observations 131–132
Knibbs, G. H. 23 price reference vs current period adjustment 124
recommendations 153–154
labor cost index (LCI) 51 role of price collectors 125
landline telephones, matched models 265, 266 targeted mean imputation 128–129, 155–157
Laspeyres index 2, 23, 38, 67, 192 treatment of permanently disappeared varieties
formula 200 186–189
geometric 23, 192 treatment of temporarily 183–185
Laspeyres-type formula 345 treatment of temporarily, observations 126–129
new weights 205–206 treatment of temporarily and permanently 125–126
licenses 42 see also quality adjustment
life insurance 43, 279, 280, 344 Mitchell, W. C., 23
COICOP (1999) 360 mobile telephones
COICOP (2018) 419–420 consumer profiles 266, 267
linking, definition 435 price collection 100–101
link-to-show-no-price-change 135, 136, 148 Model Quality Report Document for CPI 325
list prices 78 models see rapid turnover of models
Lloyd-Moulton index 190 monetary purpose, CPI for 30
local currency price 81 monetary transactions 36, 54
local price collection 12, 77–78 monetary transactions index 262
back-checking 91–92 monitoring performance 314
procedures for 86–87 mortgage interest charges 250
quality checks 91 movement splice 231
loss of characteristicity 231 multilateral indices, revisions in 231–232
Lowe index 22–23, 37, 38, 191, 192, 198, 200–201 multilateral methods
new weights 207 assessing 232
loyalty programs 48 calculating indices 232–233
luxury goods and services 47 communicating with users and stakeholders 233
dissemination of 233
management 25 implementation of 232–233
market intelligence 59 operational choices 232
matched-model method (MMM) 121 producing empirical results 232–233
hedonic indices and matched indices 152–153 publication of 233
index number bias 159–160 multilateral price index methods 228–229
missing varieties 122 bilateral price indices and chaining 227–228
new products 123 defining the variety 226–227
potential errors in 122–123 Geary-Khamis (GK) method 229
price index bias and pricing strategies 159 GEKS Method 228–229
sampling issues 122–123 multipurpose goods and services 49
matched samples 264
example 265–266 national accounts 10
maintenance and 161–163 household final consumption expenditure (HFCE) 32
price measurement 264 use of price statistics in 51
quality adjustment 267 national accounts data, as data source 58
tariffs 264, 265–266 national accounts deflation 2
telecommunications 270–271 national concept 37, 243
matrix consistency 446 national coverage, consumer price index (CPI) 4,
maximum overlap price index 239–240 196–197
mean splice 231 national pricing strategies 73
measurement errors 300 national statistical offices (NSOs) xi, 1, 121, 200
metadata net price indices 40–41
explaining index quality 333 new goods and services, quality adjustment 268–269
maintaining the sample 172–173 new goods bias, quality change and 303–305
methodological statement 330–331 new goods error 439
presentation of 332–333 new outlet bias 305
missing prices 97, 121–122 new outlet error 439
aggregation formula for elementary price indices 124–125 new products 19–21
carryforward imputation 129 CPI in between basket revisions 167–168
comparable replacements 131–132 distinguishing between new elementary aggregates and
concepts for treatment of 123–125 170–171
general considerations 129 evolutionary 165–166, 169–170
guide to treatment of 146 example introducing a new elementary aggregate 168
482
Index
example introducing a new weights for higher-level patching

aggregates 169 hedonic approach 139–146, 148
incorporation of 163–172 term 142
major concerns 164 payment, definition 435
methods for incorporating 164–168 payments approach 3
planning the introduction of 20 housing 248–251, 254
quality adjustment 268–269 property insurance services 285
quality changes vs 163–164 peak-load pricing 263
recommendations 173 pensions, indexation of 30–31
responding to limitations in overlap method use 171–172 period-of-time price collection 84–85
revolutionary 166–167, 170 permanently unavailable prices 121
timing the introduction of 20–21 plutocratic CPI 5
noncomparable replacements 132 plutocratic weights 54–55
nondurable good 36 point-in-time price collection 84–85
nonmonetary transactions 36, 54 point-of-purchase surveys 59
nonobservation errors 298 population censuses 59
nonprobability sampling techniques 67–68 population coverage 38–39, 53–54
nonprofit institutions serving household (NPISH) 34, 35, 38, practice see calculating consumer price indices
251, 274 presentation
nonresponse 301 alternative and subaggregate indices 329–330
nonsampling errors 297–298, 299 core inflation 329
nontransaction prices 78 example of illustrative press release 330
of methodology 332–333
observation errors 298 press release, bulletin and methodological statement
online prices, collection 101–102 330–331
optimal allocation, sampling 74–75 recommendations 333–334
Organisation for Economic Co-operation and Development related and alternative measures 329–330
(OECD) xi, 18, 282, 304, 446 role of advisory committees 333
organization 25 time-series, of level and change 327–328
Ottawa Group 172 press release 330–331
outlets timeliness vs data accuracy 332
coverage 39–40 price(s)
definition 435 bargaining 81–83
outlet-type weights 57 collecting and editing 12–14
product substitution 15 coverage 40–41
replacement 162 data editing 13–14
rotation bias 40 defining 78–81, 435
scope of CPI 38–39 discrimination 79
second-hand goods 258 Harmonised Index of Consumer Prices (HICP) 344–345
substitution bias 42 principle of a fixed basket 12–13
substitution error 439 principles of price collection 12–13
outliers review, price collection 96 second-hand goods 259
outputs 45, 46 variety specifications 13
overall mean imputation 155–157 price collection 72–73, 77
missing prices 126–128, 155–157 calculating average price from data in elementary aggregate
seasonal products 237–238 103
overcoverage 298 calculation of average product price 115
overlap method central 78
assumptions and concerns on use of 134–136 checklist for second-hand clothing 289
forward vs backward imputation 134 collection by telephone 100
imputed overlap prices 133–134 collector training 98–99
quality adjustment 132–134 computer-assisted data collection (CADC) 100–101
use of overlap price 132 continuity and use of codes 88–89
using actual price of replacement in preceding period 133 data validation and editing 93–98
own-account production defining the price 78–81
background 260–261 disaster recovery 99–100
construction of “compensation” index 262 documentation control template 113
construction of “general consumption” index 262 electronic reporting 100
construction of “monetary” transactions index 262 example of form 107
household services 262–263 fixed basket principle 83
introduction of 259–260 frequency 84
scope and choice of index 261–263 Internet purchases 244
owner-occupied housing 64, 344 item specifications 83–84
research 457–458 letter to second-hand clothing retailer 291
services 261 local 77–78
measuring hyperinflation or large price changes 85–86
Paasche index 23, 30, 38, 190, 192, 195, 201, 230, 239, 303, 340 median and quartile values in 94–95
geometric 23, 229 missing prices 97
483
Index
online prices 101–102 pricing gross insurance premiums 286–287

organization options 77–78 use approach 285–286
period-of-time or point-in-time 84–85 using gross premiums as proxy for net insurance service
planning and organizing 105 287
practical aspects of managing 86–93 proportionality test 180
price bargaining 81–83 publication 25–26, 233
principles of 78–86 public transportation 273–274
quality management 89–93 purchasing power parities (PPPs) 32, 443–445, 447–449
questionnaire design 87–88 research agenda 458
recommendations 103 subnational 450–451
timing of 85 pure price change, definition 435
unavoidable costs not part of advertised price 81
web scraping 102–103, 117–119 qualitative assessment, nonsampling errors 299
work instructions 99 quality 331
price collectors 82, 300 adjusting price for differences 17–19
price index, calculation for deposit product 293–295 class mean imputation 19
price quote report 93 comparable replacements 17
price reductions 78–81 definition 315
price reference period 191, 200, 335 direct or explicit adjustments methods 17–18
annual 202–203 explaining index 333
annual average 210–211 implicit adjustment 18–19
definition 435 overall mean imputation 19
monthly 203, 209–210 overlap pricing 18–19
quarterly 203, 209–210 quality-adjusted unit value index 230
price statistics, use in national accounts 51 quality adjustment
price updating of weights 435 choice between methods 146–148
price variation 41–42 class mean imputation method 134–136
between outlets 41–42 definition 435
outlet substitution bias 42 differences in feature/option costs 138–139
price discrimination 41 differences in production costs 139
price vs capacity, scatter diagram 140 explicit methods of 136–146, 230, 437
pricing hedonic approach 139–146
education 278 HICP 345
health services 276 implicit methods of 132–136, 229–230, 437
pricing gross insurance premiums, property insurance services link-to-show-no-price-change 135, 136, 148
286–287 overlap methods 132–134
private households 5, 38 quantity adjustment 136–138
probability proportional to size (PPS) sampling 66 research 456
probability sampling second-hand goods 259
definition 435 tariffs 267–269
techniques 66–67 using replacement and price overlap 157
processing errors 298, 300 see also missing prices
producer price index (PPI) 51 quality change
product(s) definition of 130
adjusting for quality differences 17–19 error 439
calculation of average prices 115 housing 256
comparable replacements 17 nature of 130–131
direct or explicit quality-adjustment methods 17–18 new goods bias and 303–305
implicit quality adjustment 18 new products and 163–164
missing 14–17 quality checks
new 19–21 back-checking 91–92
overall mean imputation 19 data reports 92–93
overlap pricing 18–19 by head office 92
permanently missing 16–17 price collection 90–93
selection of replacement for pricing 16–17 role of auditors 91
substitution procedures 15 role of head office 90–91
temporarily (nonseasonal) missing 15–16 quality management 308
weights 8–9 benchmarking 309
see also new products; seasonal products frameworks for reporting and improving CPI 314–316
product coverage, HICP 343 price collection 89–93
product exclusions, CPI 6 prototype of system 310–311
production costs, differences in 139 systems 308–310
production processes see audits of production processes total (TQM) 308–309
pro forma for audit schedule 321 work programs 316
projected current period sample account 294 quantity adjustment 136–138
property insurance services 285–287 questionnaire design, price collection 87–88
acquisitions approach 286 quota sampling 67–68
payments approach 285 definition 435
484
Index
rapid turnover of models 148 price measurement 265

characteristics/repricing approach 151 telecommunications 272
difference between hedonic and matched indices sampling 65
152–153 central price collection 72–73
hedonic imputation approach 151–152 cutoff sampling 67
hedonic price indices 150–152 errors 297
illustration of 149 health services 275–276
matching and overlap method for markets with 149–150 of items 70
recommendations 154 of locations 69–70
time dummy variable approach 150–151 nonprobability techniques 67–68
use of class mean imputation 150 of outlets 70–71
ratio of harmonic mean prices 189 probability techniques 66–67
real estate agency services 284–285 quota sampling 67–68
rebates 48, 80 recommendations 75
reference households 38 representative item method 68
reference periods 191 sample maintenance 73–74
reference population stages in CPI 69–72
consumer price index (CPI) 5–6 stratification 68–69
definition 435 techniques 65–69
refresher training, price collection 98–99 in time 72
refunds 80 variance estimation and optimal allocation 74–75
regional coverage, consumer price index (CPI) 4–5 of varieties 71–72
regional weights 55–57 scanner data 59, 72, 165, 221
release, HICP 345 assessing and preparing for use 222–223
rental equivalence approach 247, 254 benchmark index 119
rented accommodation benefits and challenges of using 223
accounting for missing observations 255–256 bilateral price indices and chaining 227–228
calculating rent index 255 classifying 222–223
housing 254–256 for data validation and quality assurance 223–224
rent surveys 254–255 defining the variety 226–227
updating rent sample 256 explicit quality adjustment 230
weights 254 formulas applicable to 190
rents 73 Geary-Khamis (GK) method 229
indexation of 31 GEKS method 228–229
replacement universe 161 implementation of 223–226
reporting, CPI 316 implementation of multilateral methods 232–233
representative item method 68 for implementing new CPI compilation methods 226
repricing approach, rapid turnover of models 151 implicit quality adjustment 229–230
Research multilateral methods 228–229
agenda for consumer price index (CPI) 455–459 obtaining sets 221–222
price-updating expenditure weights 201–202 online prices 102
services 457–458 quality assurance of 223
response errors 298 for replacing field-collected prices 224
retail trade statistics 59 research 455
revolutionary new products 164, 166–167, 170 revisions in multilateral indices 231–232
revolutionary product 19, 20 sources of weights 189
risk assessment 313 time product dummy (TPD) method 229
rolling window methods 231 for updating index structures 225–226
for updating pricing samples 224–225
sale prices 79 web-scraped data 118, 152, 165, 173
sales taxes 80 scholarships, education 279
sample scope of the index, definition 435
approaches to drawing 11–12 seasonal products 62–63, 235–241
augmentation 166, 169, 170, 171 challenges 241
designing the 10–11 clothing 238, 241
maintaining and updating 14 fixed or seasonal weights 239
rotations 162–163 fixed-weight approach 236–237
replenishment 17 fruits and vegetables 240
second-hand goods 258 implementation problems 241
sample maintenance imputation of prices under fixed-weight approach 237–238
information requirements 172–173 intractable problems 241
item replacement or substitution 161–162 introduction of 235–236
matching and 161–163 lack of well-defined seasons 241
outlet replacement 162 monthly approach 239–240
rotation, chaining and hedonic indices 162–163 replacing domestic with imported products when out of season
see also new products 241
sample of bills approach research 456
example 266 seasonal-weight approach 238–239
485
Index
seasonal-weight approach, seasonal products 235, 236, tax schedule 294

238–239 telecommunications
second-hand clothing customer profiles 271–272
price collection checklist for 289 example of user profile for mobile phone services 272
price collection letter to retailer 291 examples of specifications 270
second-hand goods 6, 47, 63–64 illustrative index structure for 270
alternative approaches 259 introduction of 269–270
following prices of basket of items 258–259 matched samples 270–271
initial sample selection 258 sample of bills 272
introduction 256–258 unit values 271
items 258 telephone, price collection 100
outlets 258 template
prices of new goods 259 audit report 323
quality adjustment 259 documentation control 319
wholesalers’ prices 259 temporarily unavailable prices 121
selection error 297, 439 terminology, basic index number 453–454
semidurables 36 terms, glossary of 335–342
services, research 457–458 time dummy variable approach 150–151
services producer price index (SPPI) 51 timeliness, HICP 345
Seventeenth International Conference of Labour Statisticians, time product dummy method 229
resolution concerning CPIs adopted by 427–434 time reversal test 180
simple random sampling (SRS) 66 tips and gratuities 42
social protection 279 for services 80
COICOP (1999) 360 Törnqvist price index 3, 23–24, 177, 190–191, 194–196,
COICOP (2018) 421–425 201–202, 227–230, 335, 342, 431, 445, 457
social security benefits, indexation of 31 total quality management (TQM) 308
social security contributions and taxes 42 tourism expenditure surveys 59
special cases 26 trade associations 59
Special Data Dissemination Standard (SDDS) 307, 315, 331 trade-ins 80
special group discounts 264 training, price collection 98–99
special offer prices 79 transaction prices, reductions 78–81
stamps 80 transactions in financial assets 43–44
static universe 227 transfers 42
Statistical Office of the European Union (Eurostat) xi transitivity 444, 445
stockbroking services 281–282 test 180–181
stratification, sampling technique 68–69 transport services
stratum weights 8 COICOP (1999) transport 354–355
subsistence agriculture 45, 260, 260–261, 261 COICOP (2018) 395–401
superlative indices (Fisher, Tornqvist and Walsh) 177, 190–191, introduction 272–273
195, 227 public transportation 273–274
supervisor training, price collection 99 Tukey algorithm 95
supply and use, price statistics in national accounts 51 two-part tariffs 263
survey of purchasers 82
sustainability 458 undercoverage 298
symmetric indices 23–24 United Nations Economic Commission for Europe (UNECE) xi,
systematic sampling 66 xii, xiii, 172n5
System of National Accounts (SNA) 29, 243, 336 unit prices 89
bargaining 81 unit value approach
research agenda 458–459 example 266
price measurement 264–265
tablets, price collection 100–101 quality adjustment 267–268
targeted mean imputation telecommunications 271
missing prices 128–129, 155–157 unit value index 190
seasonal products 237–238 updating CPI weights
tariffs 72 aggregation across linked series 213
changes for Internet prices 268 annual updating and linking 215–218
classifications 269 annual weight updates 217–218
consumer profiles 265, 266, 268 approaches to 200
Internet services 268 calculating chain index 199–200
introduction 263–264 deciding on index reference period 207–208
main measurement issues 264 deciding on price reference period for 202–203
matched samples 264, 265–266 developing update factors 203–205
new goods and service providers 268–269 example of calculating new series at first level of aggregation
price measurement methods 264–265 212–213
quality adjustment 267–269 expanding CPI geographic coverage 219–220
sample of bills approach 265, 266 formula for estimation 208
unit values 264–265, 266 frequency of weight updates 214–215
taxes and subsidies 40 introducing new classification systems 218–220
486
Index
introducing new weights 205–207 price collection 102–103

linking new series to old index reference period 211, 212 process of 117–118
linking previous CPI to new price index reference period 208–213 research 456
price reference period as single month or quarter 209–210 typical data structure 118
price reference period as yearly average 210–211 weight(s)
recommendations 220 conceptual basis of 53–55
research on price-updating expenditure weights 201–202 definition 435
using Lowe or Young index 200–201 see also updating CPI weights
upper-level or section weights 9 weighted country-product-dummy (CPD) method 449
upper-level substitution bias 302 weighting
upper-level substitution error 439 democratic and plutocratic weights 54–55
use, definition 435 deriving the pattern 7–8
use approach 3 frequency of weight updates 61–62
property insurance services 285–286 geographic coverage 53
user consultation implicit weights within elementary aggregates 9
different uses of CPI 333 introducing new 24
role of advisory committees 333, 334 monetary and nonmonetary transactions 54
user cost approach, housing 246–248, 254 national vs domestic concept 53
utility gain 170 outlet-type weights 57
population coverage 53–54
variance reference period 61
estimation of 298–299 regional weights 55–57
sampling 74–75 structure 8–9
visual data validation 96 structure of CPI 55–57
see also updating CPI weights
wages, indexation of 30–31 weight reference period 191, 200, 335
Walsh price index 3, 23–24, 117, 190–191, 195, 200–201, 339, definition 435
342, 431, 445, 453, 457 welfare gain 170
washing machine data wholesalers’ prices
estimated equation of price against capacity 141 second-hand goods 259
hedonic regression estimates for 141 survey of trends in 82
scatter diagram of price vs capacity 140 window splice 231
web prices, research 455–456 work instructions, price collection 99
Web Robots 102 work programs 316
web scraping 115, 117 World Bank xi, 32, 33, 443
benefits of 117
classification of data 118 Young index 23, 37, 192, 195, 198, 200–203, 453
country case study 119 definition 342
limitations of 117 formula 192
options for aggregating prices 119 geometric 23, 193–195
options for defining individual products 118–119 modified 339
practical considerations 118 new weights 206–208
487
PUBLICATIONS
Consumer Price Index Manual


Cpi Manual Concepts and Methods

Uploaded by

Copyright:

Available Formats

Cpi Manual Concepts and Methods

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cpi Manual Concepts and Methods

Uploaded by

Copyright:

Available Formats

CONSUMER PRICE

International Monetary Fund | International Labour Organization

Eurostat Catalogue number: OA-04-20-634-EN-C (Paper)

Please send orders to:

2 Uses, Concepts, Scope, and Classifications of Consumer Price Indices . .......................................................... 29

3 Expenditure Weights and Their Sources ............................................................................................................. 53

5 Price Collection and Validation ............................................................................................................................ 77

7 Maintaining the Sample ........................................................................................................................................ 161

8 Calculating Consumer Price Indices in Practice ................................................................................................ 175

10 Scanner Data .......................................................................................................................................................... 221

11 Selected Special Cases ........................................................................................................................................... 235

12 Errors and Bias ...................................................................................................................................................... 297

13 Data Quality Management and Reporting .......................................................................................................... 307

14 Publication, Dissemination, and User Relations ................................................................................................. 327

A Glossary of Main Terms .................................................................................................................................... 335

11.7 Bus Fares: Old and New Tariffs ...................................................................................................................... 267

Rafael Diez de Medina

Louis Marc Ducharme

The Consumer Price Index

International Standards for Consumer Price Indices

The Background to the Present Update

The Organization of the Update

2008 SNA System of National Accounts 2008

TQM Total Quality Management

Permanently Missing Products 6

A Range of Possible Consumer Consumer Price Indices and General

Household Actual Final Consumption

Insurance rendered by financial auxiliaries such as brokers, banks,

08 has been renamed Information and Communication Key Recommendations

Domestic Intermediate Final Consumption Fixed Capital Formation

Agriculture APPI APPI APPI APPI APPI CPI PPI — — XPI/PPI

Table 3.1 Example of Plutocratic versus Democratic Weights

(a) (b) (c) (d) (e) (f) (g)

Scanner Data Deriving the Weights in Practice

Cutoff Sampling (a) (b) (c) (d) (e)

Rents website can be larger than the number of prices collected in

Table 5.1 Determining Purchaser Price When Bargaining Takes Place

Carrots Carrots Lettuce Onions

sample of sellers in the same market or at different outlets Item Specifications

Figure 5.1 Price Changes in Plot Chart during Sales Season

No Price Change the predefined movement limits described earlier, may be

Annex 5.3 Table A5.2 Parameters and Derived Limits (Example 1)

Automated Data Checking RQ1 0.94488 −0.03907

DM = (RQ3 − RQ1)/2 (A5.2) R

(a) (b) (c) (d) (a) (b) (c) (d) (e)

0.81380 −0.22880 Extreme Extreme 0.8138 Extreme

Annex 5.4 only.” It provides checks, background (including explana-

Template • More efficient production of documentation as it helps

Figure A5.3 Documentation Control Template

Local Shop (traditional collection) 500 480 489.898

Local Shop (traditional collection) 450 430 439.886 89.79

Annex 5.6 • Web scraping may be time-consuming for large retailers

Table A5.8 Web Scraping—Typical Data Structure Classification of Web-Scraped Data

T Temporarily Missing: The Variety Is Unavailable but

specifications is sold in six outlets, A to F, with different I J (I Dec 2019

missing in outlet F’s price collection for March 2020, and

impute the missing price for March 2020. This is obtained

Table 6.3 Temporarily Missing Price Observations and Imputed Prices

Dec.-19 Jan.-20 Feb.-20 Mar.-20 Apr.-20 May-20 Jun.-20 Jul.-20

A Supermarket 5.25 5.25 5.49 5.49 5.49 5.49 5.49 5.49

DM = (RQ3 − RQ1)/2 (A5.2) R

pmt + 4 pmt +3 α4 pnt+ 4

I Dc = …. = (8.4) The chained Carli should be avoided because it has a known,