Measuring inflation

Nikhil Patel and Agustín Villar 1

Abstract

This note focusses on key issues-both conceptual and practical- with regard to the
measurement of inflation such as the tradeoff between different measures and the
incorporation of prices not fully determined by market forces of supply and demand.
It also draws on a recent survey of emerging market economy central banks
conducted by the BIS to highlight specific issues faced by these economies and how
the resulting inflation indices differ across countries.
Keywords: Price index, Inflation measurement
JEL classifications: E31, P24, P44

1
Bank for International Settlements. Diego Urbina provided excellent research assistance. The paper
draws on a recent BIS survey of emerging market economy central banks that was conducted
specifically for this meeting.

BIS Papers No 89 9
Introduction
Broadly considered, there are two main rationales for measuring inflation. First,
inflation reduces welfare. Second, inflation provides an indication of the degree of
slack or short-term developments in the economy. 2 The weight given to these
respective elements will determine the definition of inflation used.
How to measure inflation is a highly technical issue and involves many choices
that could have significant consequences for the level and dynamics of the series one
would like to construct. At the same time, it is important to measure inflation in a
transparent and credible way, one that is understood by different constituencies of
society. That said, if the measurement of inflation relies on cost of living estimates
across different groups of the population, such a measurement could vary across
groups; targeting the cost of living index relevant to one specific group might not be
appropriate for another group.
This note reviews different aspects of inflation measurement. Section 1 discusses
what is to be measured. Section 2 reports on alternative prices indices and their
possible biases. Section 3 deals with the challenges faced by emerging market
economy (EME) central banks in appropriately defining and measuring inflation,
focusing in particular on the treatment of administered prices and owner-occupied
housing across a surveyed sample of EME central banks.

1. What are we trying to measure?

Inflation acts as a tax on money: high and volatile inflation rates affect negatively the
demand for money. To the extent that money provides services valued by consumers
and producers, inflation imposes a welfare cost. Moreover, inflation can result in
relative price changes in the presence of nominal price and wage rigidities (Vedrin
(2015)), which can create significant welfare costs. How large these welfare costs are
depends, inter alia, on the competitive structure of the economy, government
interference, market imperfections and financial factors. As a consequence, central
banks care about the level and variability of inflation.
The theoretical basis of inflation measurement is the concept of a “composite
good” whose price is measured by a cost-of-goods index. A special case of a cost-of-
goods index is the cost-of-living index (Wynne (2008)), which derives the choice of
goods to be included in the index and their weights from the theory of consumer
behaviour. The relative importance of different goods and services changes in
response to changes in prices and preferences. A fixed bundle of goods is therefore
likely to become less representative as time elapses. Popular cost-of-living indices are
the CPI and retail price index (RPI), both of which measure the rate of change in the
prices of goods and services consumed by households. 3

2
Of course, the degree of slack in an economy and other short-term developments are also affected
by other factors not reflected in current inflation rates.
3
The difference between these two indices is more historical than conceptual. The RPI originated as a
measure of the cost-of-living needed to maintain healthy living conditions in a working class
household in the United Kingdom. However, over time its scope expanded to include all major
categories of household expenditure. The CPI, which was originated in the United States, initially
valued the expenditure of a clerical urban wage-earner. Later on, coverage was extended to all urban

10 BIS Papers No 89
 While changes in a cost-of-living index come very close to what most people
intuitively understand by the term “inflation” (Lebow and Rudd (2008)), it is not clear
that this is what central banks want to target since monetary policy affects only a
subset of prices in the economy. For this reason, central banks usually do not react to
the first round effects of changes in prices resulting from changes in indirect taxes or
administered or regulated prices, even if they are included in the index targeted. One
possibility would be to exclude such prices from the index altogether. We will return
to this issue later.
A related problem is that the cost of living may fluctuate because of transitory
shocks that cannot be addressed by monetary policy. Many central banks therefore
use different measures of “core” inflation that eliminate some of the most volatile
price components or those most affected by transitory shocks (for instance, highly
seasonal food prices). We will discuss core inflation in the next section, when we look
at the actual indices used by EME central banks. 4
The downside of dropping specific price categories is that the resulting narrow
index may become less representative of the actual cost of living. This compounds
the problem that different parts of the population have different consumption
patterns and thus face price changes which can differ significantly even from the
broadest price index. This problem tends to be especially acute in EMEs, where a
larger fraction of income is spent on items with volatile prices (such as food).
Producer price indices (PPI) or the GDP deflator provide an alternative to the CPI
or similar cost-of-living indices. For economies vulnerable to terms-of-trade shocks,
targeting the PPI or the GDP deflator might help preserve price stability, as
fluctuations in the nominal exchange rate can counter movements in the terms of
trade. Central banks would then react to any changes in producer prices that are not
absorbed by exchange rate movements (Agénor and Pereira da Silva (2013)).

2. What indices do EME central banks use?

Responses to our survey questionnaire reveal that EME central banks use a relatively
narrow set of measures of inflation in their monetary policy assessment, with the
changes in the CPI playing a central role (Table 1). Most central banks target inflation
measures based on the CPI – 16 out of the 22 responses received. A small number of
central banks replied that they did not target inflation directly as measured by a price
index. This group of central banks base their monetary policy strategy on the
targeting of the external value of the currency although the CPI still plays a central
role as a measure of inflation. 5
None of the EME central banks that participated in the survey targets a measure
of inflation based on an index other than the CPI (for instance, producer prices or the
GDP deflator). This overwhelming preference for targeting the CPI suggests that
central banks do not see the shortcomings of the CPI, such as the high share of

consumers. In both cases, the prices used in the index are obtained by regular nationwide sampling
while the weights pertaining to each price category are derived from household budget surveys.
4
From a statistical point of view, more volatile price categories tend to reduce the precision of the
price index estimates. In some cases, price categories with a comparably small relative weigh may
make a big contribution to variation in the overall index.
5
Central banks targeting the exchange rate appear to have considerable leeway in keeping inflation
low and stable. However, they still need to wrestle with considerable challenges in securing price
stability. The note submitted for this meeting by the Hong Kong Monetary Authority describes some
of those challenges.

BIS Papers No 89 11
 regulated or administered prices in EMEs, as insurmountable obstacles in the
achievement of their monetary policy objectives. In particular, by choosing the CPI
over other measures of inflation, EME central banks appear to place greater weight
on the welfare cost rationale for measuring inflation than on measures of slack or
short-term developments.
The widespread use of the CPI as the preferred price index in inflation targeting
and other monetary policy strategies in EMEs seems to reflect a number of perceived
advantages. First, the CPI is relatively easy to understand and is the best available
measure of the cost of living faced by consumers. Second, it is familiar to large
segments of the population, regularly reported in the news media, used as a reference
for the provision of government benefits or contracts and is widely followed as an
indicator of macroeconomic stability. Finally, it is available at a relatively high
frequency and is not subject to many revisions, enhancing its transparency and use
in monetary policy (Moreno (2010)).

What measure of inflation and for what?

Summary of questionnaire responses Table 1

Other uses2
Target 1
Assessment of
Set policy Forecasting
policy stance
CPI 16 (22) 13 (15) 16 (18) 15 (18)
CPI (core) ... 10 (12) 13 (14) 13 (15)
Import prices 3
... 1 7 2
GDP deflator ... 1 2 2
Producer price index ... 2 8 7
1
Total number of countries that provided information in parenthesis. 2 Number of countries using the CPI (headline) for inflation
targeting and for other uses. 3 Chile uses external prices instead of import prices for forecasting purposes.

Source: BIS survey.

Central banks that target inflation as measured by the CPI also make use of the
inflation index for other aspects of their monetary strategy. Of the 16 central banks
that target consumer price inflation, 13 replied that they also make use of CPI inflation
to set their policy (Table 1). This reveals consistency between the measured target
and the use of this measure in setting the policy rate. The three central banks that do
not set policy on the basis of the CPI tend to use a narrower measure. A few central
banks do not target consumer price inflation in their monetary policy but instead
forecast it (three central banks) and use the forecasts to assess the policy stance (three
others) or set policy (two others).
Many central banks also rely on measures of “core inflation” to forecast inflation,
assess the policy stance and set policy. There is a large overlap in the central banks
that make use of a “core inflation” measure as part of their monetary policy strategy
with those that target a measure of consumer price inflation. But “core” can mean
different things to different countries. It may exclude highly volatile prices (eg those
of foodstuffs and fuel), non-monetary expenditures (eg imputed housing costs for
owner-occupied property or rents) or the effects of changes in taxes, subsidised
prices or rents, and administered prices.

12 BIS Papers No 89
Biases in the measurement of inflation

While the CPI has a central role to play in the monetary policy framework of many
EMEs, there are reasons to believe that it may be biased. The most common
arguments suggest that it typically overestimates inflation. This bias reflects
substitution effects, household heterogeneity, and quality improvement and product
replacement effects. Unless the weights are revised frequently, the index will become
less representative as goods with larger price increases tend to be demanded less
and those with lower prices increases or price decreases tend to be demanded more.
The obvious solution lies in a more frequent updating of the index. However, this
comes at the cost of more frequent expenditure surveys and possibly also perceptions
of index manipulation.
Heterogeneity of household income is a significant potential source of bias in
EMEs. The weights of the CPI are often based on the average share of expenditure
from different groups in society. Because the distribution of expenditure tends to be
skewed, the index is likely to overweight the expenditure patterns of the more affluent
segments of society. This problem might be more serious in EMEs where the
distribution of income is more unequal. The impact on the inflation rate is less clear:
it would depend on the relative prices changes taking place in the economy. If the
prices of goods prominent in the expenditure of households at the higher end of the
income distribution scale are growing faster than those of groups at the lower end,
the measurement of inflation will be biased upwards.
The implications of overestimation have often been flagged in other areas of
public policy such as fiscal policy and the assessment of economic competitiveness.
For example, an upward bias may drive up pensions and social spending more rapidly
than justified by changes in the cost of living. If it is not anticipated, the systematic
overestimation of the inflation rate due to a bias can also have significant
consequences for perceived real interest rates and thus for saving and investment as
well as monetary policy.
The replies to the survey show that half of the 16 central banks have estimates
for the potential bias in CPI inflation, although the type of bias and even sometimes
the sign vary across countries. The central banks of Malaysia, Poland, Russia and
Thailand estimate an upward bias. In Chile, by contrast, the quality bias in the clothing
category imparts a negative bias to inflation as estimated by the CPI. In Malaysia, the
heterogeneity of income is a motive for considerable control of surveys carried out
at five-year intervals. The substitution bias is also monitored with surveys carried out
every six months in order to adjust expenditure shares.

3. Some practical issues in measuring inflation

Prices of goods and services that are not freely determined in the market present a
challenge to inflation measurement. Conceptually, it is not always clear whether and
to what extent these prices should be incorporated in an inflation measure used for
conducting monetary policy. Moreover, given the difficulty of estimating and inferring
prices in the absence of explicit markets, it is not always feasible to implement an
optimal index even if one attempts to base it on firm conceptual grounds. 6 Central
banks therefore face the complex task of striking a balance between practice and

6
For instance, while in theory it may be optimal to fully incorporate the cost of owner-occupied
housing (an issue discussed in detail below), in practice this may not be feasible in the absence of
good proxies for the cost of owner-occupied housing, given the lack of actual market prices.

BIS Papers No 89 13
theory when deciding how such prices should be treated in their preferred inflation
measure. This section covers two prominent prices that fall into this category, namely
administered and regulated prices, and the prices of owner-occupied housing
services. It also summarises how central banks in our survey deal with them.

3.1 Administered and regulated prices

Governments typically regulate the prices of certain goods and services. These
obviously include goods and services provided by the government itself, such as
public transport, health care and education, but also other goods that are considered
essential. 7 It is in the treatment of administered prices that the tension between the
welfare cost and slack measurement rationales for measuring inflation come to the
fore. On the one hand, if such prices are mainly driven by social and political concerns,
they cease to be good indicators of slack in the economy or of short-term
developments more generally. On the other hand, these prices represent a significant
share of the consumption basket in most EMEs and hence warrant inclusion as a
gauge of the cost of living.
According to questionnaire responses, the share of administered and regulated
prices varies substantially across countries (Appendix Table 3.1). It is as high as 34.6%
in Thailand and less than 5% in India. As expected, food, energy and government
services are the most prominent categories.
The survey also highlights how central banks differ in their treatment of
administered and regulated prices, and in balancing the trade-offs mentioned above.
At one extreme, these prices are incorporated fully in the inflation measure (eg in
Peru). At the other extreme, they are dropped from the index altogether
(eg in Hungary and Poland). 8 In between, many central banks choose to downplay
administered and regulated prices through different means, often retaining scope to
exercise discretion and judgment in the presence of large movements. 9
The Bank of Thailand provides a slightly different and interesting alternative. It
uses an inflation measure that attempts to explicitly exclude only government
measures (such as an increase in excise taxes) from regulated prices. This has the
advantage of retaining goods with administered prices to maximise the coverage of
the inflation measure while at the same time preventing government actions (which
are presumably not directly linked to the state of the economy) to affect it. Taking
this approach further, another option might be to estimate shadow prices for
regulated goods and services by using, for instance, estimates of the cost of
production. That said, this would make the measure less transparent and timely and
no longer indicative of the cost of living.
Regulated prices often give rise to black markets and underground economies.
More generally, the informal sector of an economy, which in some cases is fairly
sizeable (39–42% of GDP for the Philippines according to the estimates of Schneider

7
This, for instance, is the case for fuels such as kerosene and liquefied petroleum gas (LPG) in India.
8
The core inflation measure computed by the National Bank of Poland excludes administered and
regulated prices. As discussed above, different countries work with different definitions of core
inflation.
9
For example, while regulated and administered prices are incorporated in the inflation measure used
by Korea, steps are taken to exclude transitory changes due to institutional shifts unrelated to
economic forces.

14 BIS Papers No 89
et al (2010)), presents a challenge for inflation measurement since these prices are
excluded from the computation of economic indices.

3.2 Measuring the price of housing services

Housing services, whether explicitly accounted for by rental payments or implicitly

incorporated in owner-occupied housing estimates, account for a large proportion of
total household expenditure. But houses tend to be big-ticket items that are
purchased infrequently, so the cost of the implicit services provided must be
estimated. 10 This could be done by including house prices directly in the measure of
inflation, for instance through the net acquisition approach that measures the average
change in the price paid by a household to acquire a home, or by estimating implicit
rents. Yet another option is to follow the user cost approach, which covers various
costs of home ownership such as interest payments on mortgages, taxes, insurance,
repairs and maintenance costs (see McCarthy and Peach (2010) and Cecchetti (2007)
for discussions of how those issues are being addressed in advanced economies). 11
Our survey responses show that EME central banks address these issues in very
different ways. As a result, the share of housing in the inflation measures varies widely,
ranging from 1.2% in Hungary to 31.7% in Hong Kong SAR (Appendix
Table 3.2).
While rental prices are relatively easy to obtain and are incorporated by all the
surveyed countries, with the exception of Peru, Saudi Arabia and the United Arab
Emirates, the extent to which owner-occupied housing is covered differs widely. Some
central banks choose to exclude owner-occupied housing completely (Chile, Hungary,
Korea, Malaysia, the Philippines, Poland, Russia and Turkey). 12 This could be either
because they prefer to work with an inflation measure that is representative only of
market prices or because it is not feasible to obtain satisfactory measures. Among the
central banks that do incorporate owner-occupied housing in their inflation measure,
the rental equivalence approach is nearly the unanimous choice (Appendix
Table 3.2). 13 This is consistent with Gillingham (1983), who argues that this is
preferable on both theoretical and empirical grounds.
Even in the case of rents, central banks in our survey highlight potential sources
of bias. Typically, the longitudinal nature of most surveys implies that newly
constructed houses are often underrepresented, leading to a downward bias in
measured inflation. Central banks address this issue in several ways. For instance,
some countries report using rotating samples and replace a certain fraction of the
sample at a specific frequency. 14 Recognising the problem posed by lags between
construction of new housing units and their full incorporation in the rental

10
From the point of view of national income accounting, homeowners living in their own houses are
assumed to pay themselves a market rent, which appears as consumption expenditure in GDP. The
rationale is that if some or all homeowners become renters or vice versa, GDP should not be affected.
11
See Poole et al (2005) for a detailed description and comparison of these methods.
12
This is also the approach taken in computing the Harmonised Index of Consumer Prices (HIPC), the
primary inflation measure used by the ECB to set monetary policy for the euro area.
13
The only exceptions are Israel, which in addition to the rental equivalence approach gives a small
weight to the cost of insurance and legal services; Chile, which reports using repairs and maintenance
services; and the Czech Republic, which reported using the user cost approach in the past, but
switched to the rental equivalence approach in 2007.
14
One thirty-sixth of the sample is replaced in Korea every month, while one third is replaced by Turkey
every year.

BIS Papers No 89 15
computation programme, the Czech Republic makes annual adjustments based on
finalised construction works and liquidated housing.
Survey attrition, small sample coverage and changes in neighbourhood quality
are some of the other sources of bias that affect the measurement of housing services
for both tenant and owner-occupied units. Israel, for instance, identifies the small
sample of rental housing units as a major potential source of bias and uses a hedonic
estimation routine to address it.
Sometimes the biases are not easy to minimise, and central banks may be left
with little choice but to use an inflation index that excludes housing inflation to reflect
inflationary pressure, as is done by Thailand.

16 BIS Papers No 89
 Appendix

Administered and regulated prices

Summary of survey responses; as a share of preferred inflation measure Table 3.1

Overall weight on preferred Sectors

inflation measure
Chile 9.9 Gasoline 3.5
Electricity 2.7
Drinking water 1.9
Transportation by subway 1.3
China ... Agricultural goods 3.0–5.0
Petrol 3.0–5.0
… ...
… ...
Colombia 15.2 Urban public transportation 4.8
Fuel 2.9
Electricity 2.9
Water/sewage 2.6
Czech Republic 16.9 Housing, water, electricity, gas and other fuels 11.6
Health 1.7
Transport 1.2
Restaurants and hotels 0.9
Hungary 17.8 Regulated services and medicines 10.5
Regulated energy 7.3
… ...
… ...
India ... Fuel group – administered kerosene prices and
subsidised liquefied petroleum gas prices 1.8
Food group – prices of food items distributed
through the Public Distribution System (PDS) 0.6
… ...
… ...
Indonesia 18.0 Transport, communication, and financial services 9.7
Housing, water, electricity, gas, and fuel 5.1
Prepared food, beverages, cigarette, and tobacco 3.1
Health 0.0
Israel ... Education 4.2
Energy 3.1
Food 2.4
Municipal taxes 2.3
Korea ... Public services 8.6
Electricity and water and gas 4.9
… ...
… ...
Malaysia ... Utilities and energy (excluding fuels for personal
transport equipment) 4.5

BIS Papers No 89 17
 Alcoholic beverages and tobacco 2.2
Food (controlled price) 1.1
Transport services 0.9
Mexico 14.8 Gasoline 3.7
Electricity 3.6
Public transportation 2.0
Liquefied petroleum gas 1.6
Peru ... Electricity 2.9
Telephone 2.9
Water 1.6
… ...
Poland ... Energy 9.4
Services 5.2
Electricity 4.4
Gas 2.5
Russia ... Public utilities 5.7
Housing services other than apartment rentals 2.6
Local railway and municipal transportation 1.6
Vital medicines 0.8
Singapore ... Electricity 3.2
Bus fares 1.4
Train fares 1.2
Household services and supplies: government levy
for foreign domestic worker 0.8
South Africa 18.5 Petrol 5.7
Electricity 4.1
Education 3.0
Communication 76.0
Thailand 34.6 Core 19.9
Energy 11.4
Raw foods 3.3
… ...
Turkey ... Energy 6.9
Alcoholic beverages and tobacco 4.8
… ...
… ...
United Arab Emirates ... Gas, electricity and water 5.2
...
...
...
Source: BIS survey responses.

18 BIS Papers No 89
Housing rental cost in the CPI Table 3.2

Imputed rent based on:

Market rent Net Rental User cost Other Share in
acquisition equivalent approach price index
approach approach (%)

Chile YES ... ... ... ... 4.20

China YES ... YES ... ... 20.00
Colombia YES ... YES ... ... 18.59
Czech Republic YES ... YES ... ... 13.82
Hong Kong SAR YES ... YES ... ... 31.70
Hungary YES ... ... ... ... 1.20
India YES ... YES ... ... 10.10
Indonesia YES ... ... ... ... 8.57
Israel YES ... YES ... ... 24.80
Korea YES ... ... ... ... 9.28
Malaysia YES ... ... ... ... 17.20
Mexico YES ... YES ... ... 17.20
Peru ... ... YES ... ... 2.40
Philippines YES ... ... ... YES 13.80
Poland YES ... ... ... ... 1.20
Russia YES ... ... ... ... 2.90
Saudi Arabia ... ... ... ... ... 20.50
Singapore YES ... YES ... ... 22.90
South Africa YES ... YES ... ... 0.00
Thailand YES ... YES ... ... 15.40
Turkey YES ... ... ... ... 5.10
United Arab Emirates ... ... ... ... ... 39.00
Source: BIS survey responses.

BIS Papers No 89 19
Core inflation measures used by central banks
Summary of survey responses Table 3.3

Chile CPI (core): CPI without foods and energy

China …
Colombia CPI (core):
Czech Republic CPI (core): excluding fuels
Hong Kong SAR …
Hungary CPI (core): adjusted for the effects of indirect taxes
India …
Indonesia CPI (core)
CPI (volatile foods)
CPI (index-administered prices)
Israel CPI (core): excluding energy, food, fruits and vegetables
Korea CPI (core)
Malaysia CPI (core)
Mexico CPI (core)
Peru CPI (core)
Poland CPI (core): CPI net of food and non-alcoholic beverages and energy
Inflation net of administered prices
Philippines CPI (core)
Russia CPI (core)
Saudi Arabia …
Singapore CPI (core): excludes accommodation and private road transport costs
South Africa CPI (core)
Administered prices
Thailand CPI (core)
Core inflation excluding government measures and rent
Turkey CPI (core): excludes unprocessed food and alcohol-tobacco
United Arab Emirates CPI excluding rent
Source: BIS survey responses.

20 BIS Papers No 89
References
Agénor, P-R and L Pereira da Silva (2013): Inflation targeting and financial stability: a
perspective from the developing world, Inter-American Development Bank and Centro
de Estudios Monetarios Latinoamericanos.
Cecchetti, S (2007): “Housing in inflation measurement”, voxeu.org, 13 June.
Deaton, A (1998): “Getting prices right: what should be done?”, Journal of Economic
Perspectives, vol 12, no 1, pp 37–46.
Diewert, E (2010): “On the stochastic approach to index numbers”, in Price and
productivity measurement: vol 6 – Index number theory, pp 235–62.
Frankel, J (2011): “A comparison of product price targeting and other monetary
anchor options for commodity exporters in Latin America”, Harvard Kennedy School
Faculty Research Working Paper Series, no 11-027, July.
Gillingham, R (1983): “Measuring the cost of shelter for homeowners: theoretical and
empirical considerations”, The Review of Economics and Statistics, vol 65, no 2, May,
pp 254–65.
Hanson, C and I Weir (1992): “Measures of inflation”, in The New Palgrave Dictionary
of Money and Finance, vol 2, pp 685–8.
Lebow, D and J Rudd (2008): “Inflation measurement”, in The New Palgrave Dictionary
of Economics, Second Edition.
McCarthy, J and R Peach (2010): “The measurement of rent inflation”, Federal Reserve
Bank of New York, Staff Reports, no 425, January.
Moreno, R (2010): “Some issues in measuring and tracking prices in emerging market
economies”, BIS Papers, no 49, January, pp 13–51.
Poole, R, F Ptacek and R Verbrugge (2005): Treatment of owner-occupied housing in
the CPI, Federal Economic Statistics Advisory Committee (FESAC), 9 December.
Schneider, F, A Buehn and C Montenegro (2010): “New estimates for the shadow
economies all over the world”, International Economic Journal, vol 24, no 4,
pp 443–61.
Vedrin, A (2015): “Inflation targeting and financial stability: providing policymakers with
relevant information”, BIS Working Papers, no 503, July.

Wynne, M (2008): “Core inflation: review of some conceptual issues”, Federal Reserve
Bank of St Louis, Review, Part 2, May/June, pp 205–28.

BIS Papers No 89 21