G U I D E L I N E S F O R

Producing Statistics on
Asset Ownership from
a Gender Perspective
 ST/ESA/STAT/SER.F/119

Department of Economic and Social Affairs

Statistical Division

Studies in Methods Series F No. 119

Guidelines
for Producing Statistics
on Asset Ownership
from a Gender Perspective

United Nations
New York, 2019
Department of Economic and Social Affairs

The Department of Economic and Social Affairs of the United Nations is a vital inter-
face between global policies in the economic, social and environmental spheres and
national action. The Department works in three main interlinked areas: (i) it compiles,
generates and analyses a wide range of economic, social and environmental data and
information on which United Nations Member States draw to review common prob-
lems and to take stock of policy options; (ii) it facilitates the negotiations of Member
States in many intergovernmental bodies on joint courses of action to address ongoing
or emerging global challenges; and (iii) it advises interested Governments on the ways
and means of translating policy frameworks developed in United Nations conferences
and summits into programmes at the country level and, through technical assistance,
helps build national capacities.

Note

The designations employed and the presentation of the material in the present pub-
lication do not imply the expression of any opinion whatsoever on the part of the
United  Nations concerning the legal status of any country or of its authorities, or
the delimitations of its frontiers. The term “country” as used in this publication also
refers, as appropriate, to territories or areas. The designations of country groups are
intended solely for statistical or analytical convenience and do not necessarily express
a ­judgment about the stage reached by a particular country, territory or area in the
development process. Mention of the names of firms and commercial products does
not imply endorsement by the United Nations. The symbols of United Nations docu-
ments are composed of capital letters and numbers. Mention of such a symbol refers to
a United Nations document.

ST/ESA/STAT/SER.F/119
United Nations publication
Sales No. E.18.XVII.11
ISBN: 978-92-1-161637-8
eISBN: 978-92-1-363250-5
Copyright © United Nations 2019
All rights reserved
 iii

Preface

This publication of the present Guidelines for Producing Statistics on Asset Ownership
from a Gender Perspective (hereinafter referred to as the Guidelines) provides national
statistical agencies and policymakers with guidance on collecting, processing, ana-
lysing and disseminating individual-level data on asset ownership to inform three
broad sets of policy issues: fostering the empowerment of women; reducing poverty
and vulnerability; and understanding livelihoods. The Guidelines introduce the con-
cepts, definitions and data requirements for measuring asset ownership from a gender
perspective in household surveys and provide guidance on planning, organizing and
implementing a household survey, appending a module, or adding a minimum set of
questions on asset ownership to a nationally representative household survey. Coun-
tries may choose a particular modality for the implementation of the recommenda-
tions, depending on their own needs and capabilities, including the needs of data users
and the availability of data from other sources, such as administrative records.
The Guidelines present a framework for measuring asset ownership from a gen-
der perspective, in which ownership is conceptualized as a bundle of ownership rights,
including reported and documented ownership and the rights to sell and bequeath an
asset. The extent to which these ownership rights are vested in one individual varies
across and within countries, depending on their legal frameworks and social norms.
The Guidelines also outline key recommendations, including the list of assets for data
collection, highlighting priority and additional assets; the issue of valuing assets;
the rationale for self-reported or self-declared data collection rather than proxy data
(because collecting proxy data from the head or another member of the household,
as is standard in many countries, is likely to underestimate both women’s and men’s
ownership of assets); data collection strategies; recommended approaches for sample
design, focusing on within-household selection; and suggested data analysis and indi-
cators, relevant for gender analysis of asset ownership and control.
 v

Acknowledgements

The Guidelines for Producing Statistics on Asset Ownership from a Gender Perspec-
tive were prepared under the Evidence and Data for Gender Equality (EDGE) project,
which is aimed at accelerating existing efforts to improve the capacity of countries to
produce relevant and high-quality gender statistics. Building on the work of the Inter-
Agency and Expert Group on Gender Statistics, the six-year project (2013–2018), a
joint initiative of the Statistics Division, Department of Economic and Social Affairs,
and the United Nations Entity for Gender Equality and the Empowerment of Women
(UN-Women), was carried out in collaboration with the Asian Development Bank
(ADB), the Food and Agriculture Organization of the United Nations (FAO), the
International Labour Organization (ILO), the Organization for Economic Coopera-
tion and Development (OECD) and the World Bank. The project was funded by the
Governments of Australia, Canada, Germany, Ireland, the Republic of Korea and the
United States of America.

The Guidelines are the result of extensive consultations with national and inter-
national statistical and gender experts and are based on the results of testing and
piloting different methods in seven countries (Georgia, Maldives, Mexico, Mongolia,
Philippines, Uganda and South Africa). In particular, the Guidelines benefited greatly
from the many comments and suggestions made by national statistical offices and
other experts, through a number of technical meetings and discussions, including
during the forty-eighth session of the Statistical Commission, when the Guidelines
were presented in their draft version. Furthermore, lessons learned from the national
statistical offices of the seven countries that piloted the EDGE methodology were criti-
cal in shaping the final recommendations included in the Guidelines.

To ensure that the Guidelines were methodologically robust and sustainable, the
EDGE project worked in partnership with the national statistical agencies of the seven
pilot countries, the ADB and the World Bank to pilot the methodologies. The studies
provided an opportunity to test and refine key aspects of the methodologies, including
conceptual and measurement issues related to questionnaire design, respondent selec-
tion interview protocols and indicator constructs. In 2014, EDGE collaborated with
the World Bank Living Standards Measurement Study team in Uganda to conduct a
methodological survey experiment assessing the relative effects of interviewing differ-
ent household members about individual-level asset ownership and control, the find-
ings of which informed the EDGE pilot studies that were implemented over the next
two years. In 2015, with funding from the National Institute of Statistics and Geogra-
phy, Mexico appended a module on the ownership of a core set of assets to its national
household survey. Also in 2015, Georgia, Mongolia and the Philippines implemented
stand-alone surveys on the full range of financial and physical assets, with funding
and technical support from ADB. In 2016, with funding and technical support from
EDGE, Maldives appended a module on the core set of assets to its household, income
and expenditure survey, while South Africa piloted a stand-alone survey on the full
set of assets.
vi

The Guidelines were prepared by a team in the Social and Gender Statistics Sec-
tion of the Statistics Division, comprising Ionica Berevoescu, Haoyi Chen, Francesca
Grum, Lauren Pandolfelli and Gulab Singh. The report was produced under the direc-
tion and overall substantive guidance of Francesca Grum, Chief of Section.
Special gratitude is owed to the pilot countries and partner agencies for their
invaluable contributions in developing the methods presented in the Guidelines: Ten-
giz Tsekvava, project leader, Giorgi Kalakashvili, Tamar Gulua, Teimuraz Paksashvili,
Salome Tvalodze and Paata Giorgashvili of the National Statistics Office of Georgia;
Aishath Shahuda, project leader, Aishath Laila, Ashiyath Shazna, Fathimath Nihan,
Fathimath Riyaza, Hana Mansoor, Mariyam Mirfath, Aishath Aniya, Hudha Haleem,
Nazima Shareef, Shamila Rasheed and Mohamed Fathih of the National Bureau of
Statistics of Maldives; Félix Vélez, project leader, María Eugenia Patricia Gómez Luna,
Marco Antonio Gutiérrez Romero, Juan José Ríos, Ramón Bravo, Hugo Hernández
Ramos, Alexandra Stephanie Boyer, Óscar Joaquín Ramírez Álvarez and José Anto-
nio Mejía of the National Institute of Statistics and Geography (INEGI) of Mexico;
Oyunchimeg Dandar and Amarbal Avirmed, project leaders, Saranchimeg Byamba,
Khuslen Zorigt, Tamir Baldandorj, Ankhzaya Byamba, Myagmarkhand Erdene-
Ochir and Sengum Shinetugs of the National Statistics Office of Mongolia; Wilma
Guillen, project leader, Bernadette Balamban, Plenee Grace Castillo, Elpidio Mara-
mot, Anna Jean Casañas, Andrea Bibares, Edna Rapanot and Florante Varona of the
Philippine Statistics Authority; Isabel Schmidt, project leader, Constance Mabela and
Babalwa Mpho Nyangintsimbi of Statistics South Africa; James Muwonge, project
leader, Norah Madaya, Stephen Baryahirwa, Pamela Nabukhonzo, Vincent Sennono
and Diana Byanjeru of the Uganda Bureau of Statistics.
Special acknowledgment goes to Chiara Brunelli of FAO, who has contributed to
the development of the recommended methods since the inception of the EDGE pro-
ject and who drafted parts of the Guidelines, and to Kaushal Joshi and his team at ADB
for their overall contribution to the EDGE project over the years and for his leadership
in the pilot testing of the recommended methods in three countries (Georgia, Mon-
golia and the Philippines) funded by the Bank. The Guidelines benefited extensively
from the collaboration between EDGE and the World Bank, in particular through
the results of the Methodological Experiment on Measuring Asset Ownership from
a Gender Perspective (MEXA), carried out under the supervision of Talip Kilic of the
World Bank Living Standards Measurement Study programme, in collaboration with
the Uganda Bureau of Statistics.
Sincere appreciation goes to the following experts who reviewed the Guidelines
and provided technical advice: Krista Jacobs (United States Agency for International
Development), Papa Seck (UN-Women), Caren Grown (World Bank), Cheryl Doss
(Yale University and University of Oxford), Arturo Martinez (ADB), Jim Lepkowski
(University of Michigan), Hitomi Komatsu (consultant to the Statistics Division),
Urmilla Bob (University of Cape Town), Elisa Benes and Kieran Walsh (ILO), Pilar
Campos and Maria O’Keefe (consultants to the National Institute of Statistics and
Geography of Mexico), and Hermanus Smith and Andrew Smith (Statistics Division).
 vii

Abbreviations and acronyms

ACASI Audio-Computer-Assisted Self-Interview

ADB Asian Development Bank
CAPI Computer-Assisted Personal Interview
CASI Computer-Assisted Self-Interview
CATI Computer-Assisted Telephone Interview
deff design effect due to intracluster correlation
EDGE Evidence and Data for Gender Equality
ENH National Household Survey (Mexico)
FAO Food and Agriculture Organization of the United Nations
HIES Household, Income and Expenditure Survey
ILO International Labour Organization
MDG Millennium Development Goals
MEXA Methodological Experiment on Measuring Asset Ownership
from a Gender Perspective
NSO national statistical office
NUTS Nomenclature of Territorial Units for Statistics
OECD Organization for Economic Cooperation and Development
OECD Guidelines OECD Guidelines for Micro Statistics
on Household Wealth, 2013
roh intracluster rate of homogeneity
SDG Sustainable Development Goals
SEEA System of Environmental-Economic Accounting
SEEA Central System of Environmental-Economic
Framework Accounting 2012—Central Framework
SNA System of National Accounts
UN-Habitat United Nations Human Settlements Programme
UN-Women United Nations Entity for Gender Equality
and the Empowerment of Women
WCA 2020 World Programme for the Census of Agriculture 2020
2008 SNA System of National Accounts 2008
 ix

Contents

page

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Abbreviations and acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Purpose of the Guidelines and key recommendations . . . . . . . . . . . . . . . . . . . . . . . 1
Relevance of the Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Key objectives of data collection, policy questions and measures . . . . . . . . . . . . . 6
Development of the Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Relationship with international standards and other global guidelines . . . . . . . . 11
Users of the Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Organization of the Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

I. Conceptual framework for measuring asset ownership

from a gender perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1. Defining asset ownership . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.1. Bundle of ownership rights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2. Forms of ownership . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.3. Acquisition of assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2. Respondent rules for measuring asset ownership from a gender perspective 20
2.1. Differences between proxy and self-reported estimates
of women’s and men’s asset ownership . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2. Implications of respondent rules on interviewing protocols . . . . . . 23
3. Definition and coverage of assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1. What is an asset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2. Terms and definitions relating to specific types of assets . . . . . . . . . 26
4. Establishing the value of assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.1. Why valuing assets is important . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2. Principles in establishing value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.3. Which assets to value? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4. Who should provide values? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.5. Valuation of specific assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.5.1. Dwellings and other structures . . . . . . . . . . . . . . . . . . . . . . . 41
4.5.2. Land . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.5.3. Agricultural equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.5.4. Livestock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.5.5. Financial assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
x

Page

4.5.6. Liabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.5.7. Consumer durables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.5.8. Non-agricultural enterprise assets . . . . . . . . . . . . . . . . . . . . 47
4.5.9. Valuables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5. Units of observation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.1. Individuals as the unit of observation . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.2. Assets as the unit of observation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

II. Role of household surveys and other sources of data in collecting

individual-level data on asset ownership and control . . . . . . . . . . . . . . . . . . . . . . . 55
1. Role of household surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
1.1. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
1.2. Conceptual framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
1.3. Units of observation and measures of ownership . . . . . . . . . . . . . . . 56
1.4. Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2. Population and housing censuses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3. Agricultural censuses and surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.1. Scope of agricultural surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.2. Limitations of agricultural surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.3. Scope of agricultural censuses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.4. Limitations of agricultural censuses . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4. Administrative sources of data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.1. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2. Conceptual framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3. Units of observation and measures of ownership . . . . . . . . . . . . . . . 60
4.4. Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

III. Guidance for implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

1. Planning data collection on asset ownership at the individual level . . . . . . 63
1.1. Specifying data collection objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 63
1.2. Building the project team . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
1.3. Budget and timeline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2. Data collection strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.1. Appending a minimum set of questions to an existing survey
questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.2. Appending a survey module to an existing household survey . . . . . 68
2.3. Conducting a stand-alone survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.4. Choosing between the three data collection strategies . . . . . . . . . . . 70
3. Modes of data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.1. Basic modes of data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.2. Implementing face-to-face surveys using paper questionnaires
versus CAPI approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.2.1. Advantages of using a CAPI questionnaire . . . . . . . . . . . . . 74
3.2.2. Costs and risks associated with using CAPI . . . . . . . . . . . . 74
 Contents xi

Page

4. Sample design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.1. Principles in sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.1.1. Target population . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.1.2. Sampling frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.1.3. Sample size determination . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.1.4. Structure of the sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.2. Selecting individuals from households . . . . . . . . . . . . . . . . . . . . . . . . 84
4.2.1. Operational challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.2.2. Cost considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.2.3. Making decisions on individual respondent selection . . . . 94
5. Questionnaire design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.1. Background research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2. Questionnaire content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.2.1. Key information required . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.2.2. Notes on components of the questionnaire . . . . . . . . . . . . . 105
5.2.3. Specific considerations in the questionnaire design
for selected assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.2.4. Specific considerations in the questionnaire design
for different survey instruments . . . . . . . . . . . . . . . . . . . . . . 114
5.3. Testing the questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.3.1. Expert reviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.3.2. Focus groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.3.3. Cognitive interviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.3.4. Field pretests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.3.5. Randomized or split-ballot experiments . . . . . . . . . . . . . . . 118
5.4. Designing and testing the CAPI questionnaire . . . . . . . . . . . . . . . . . 119
5.4.1. Designing the CAPI questionnaire . . . . . . . . . . . . . . . . . . . . 119
5.4.2. Testing the CAPI questionnaire . . . . . . . . . . . . . . . . . . . . . . . 120
5.5. Survey manuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
5.5.1. Instruction manual for fieldworkers . . . . . . . . . . . . . . . . . . . 121
5.5.2. CAPI manuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
5.6. Translating survey instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
6. Field operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.1. Field organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.1.1. Recruitment and organization of field staff . . . . . . . . . . . . . 124
6.1.2. Publicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.1.3. Role of geospatial information in supporting data
collection operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.2. Training of field staff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
6.2.1. Training on the paper questionnaire . . . . . . . . . . . . . . . . . . 128
6.2.2. Training on CAPI-specific issues . . . . . . . . . . . . . . . . . . . . . 131
6.3. Fieldwork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.3.1. Workload distribution, and information
and management flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
xii

Page

6.3.2. Interview protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

6.3.3. Quality assurance during field operations . . . . . . . . . . . . . . 136

IV. Data processing, analysis and dissemination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

1. Data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
1.1. Data entry and organization of data sets . . . . . . . . . . . . . . . . . . . . . . . 139
1.2. Data editing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
1.3. Imputation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
1.4. Weighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
1.4.1. Adjusting for unequal selection probability . . . . . . . . . . . . . 143
1.4.2. Adjusting for unit non-response . . . . . . . . . . . . . . . . . . . . . . 144
1.4.3. Post-stratification weighting . . . . . . . . . . . . . . . . . . . . . . . . . 147
1.4.4. Developing weights for asset . . . . . . . . . . . . . . . . . . . . . . . . . . 148
2. Recommended indicators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
2.1. Level of monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
2.2. Level of measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
2.3. Indicator construct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
3. Data analysis and dissemination of results . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
3.1. Data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3.1.1. Organization of the data file based on units of observation
and analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3.1.2. Types of variables used in an analysis of asset ownership
from a gender perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
3.1.3. Key objective: measuring and understanding
the gender asset gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
3.1.4. Key objective: measuring the gender wealth gap . . . . . . . . 165
3.1.5. Key objective: intrahousehold analysis of asset ownership 171
3.2. Data dissemination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3.2.1. Data tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
3.2.2. Analytical publications, reports, articles and briefs . . . . . . 179
3.2.3. Gender indicator databases . . . . . . . . . . . . . . . . . . . . . . . . . . 179
3.2.4. Production of metadata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
3.2.5. Sharing of microdata for researchers and academics . . . . . 179

Annex 1.  Minimum set of questions for priority assets . . . . . . . . . . . . . . . . . . . . . . . . . 181

Annex 2.  Model questionnaire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
 Contents xiii

Page

Boxes
Box 1. Measuring women’s ownership of assets in the 2030 Agenda . . . . . . . . . . . . 4
Box 2. EDGE project . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Box 3. Overview of the methodological experiment on measuring . . . . . . . . . . .
asset ownership from a gender perspective in Uganda . . . . . . . . . . . . . . . . . . 22
Box 4. Challenges of constructing household rosters of assets from multiple
respondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Box 5. Key steps to consider when conducting a survey using a handheld device . 76
Box 6. Kish method for random selection of household members . . . . . . . . . . . . . . 86
Box 7. Challenges in conducting simultaneous interviews . . . . . . . . . . . . . . . . . . . . . 89
Box 8. Using propensity score method to adjust for unit non-response . . . . . . . . . . 145
Box 9. Reconciling reporting discrepancies when interviewing multiple
persons in the same household . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

Figures
Figure 1. Conceptual framework for measuring asset ownership and control
from a gender perspective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Figure 2. FAO (WCA 2020) classification of land use . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Figure 3. Approaches for selecting individuals from households . . . . . . . . . . . . . . . . . . 85
Figure 4. Decision tree for intrahousehold respondent selection . . . . . . . . . . . . . . . . . . 99
Figure 5. Prevalence of reported ownership of the principal dwelling among
the adult population, by gender (Uganda, 2014) . . . . . . . . . . . . . . . . . . . . . . . . 160
Figure 6. Share of women among owners of principal dwelling (Uganda, 2014) . . . . . 161
Figure 7. Percentage of reported owners who have documented ownership,
the right to sell and/or bequeath assets, by gender (Uganda, 2014,
and KwaZulu-Natal, South Africa, 2016) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Figure 8. Ownership of principal dwelling, by gender and documentation
(Georgia, 2015) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
Figure 9. Main methods of acquisition of agricultural land (percentage),
by gender (Georgia, 2015) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
Figure 10. Distribution of values of principal dwellings (Uganda, 2014) . . . . . . . . . . . . 167
Figure 11. Women’s share among owners, and women’s share of total wealth stored
in selected assets (Uganda, 2014) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Figure 12. Women’s share of wealth by couple’s wealth (Uganda, 2014) . . . . . . . . . . . . . 173

Tables
Table 1. Overview of EDGE pilots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Table 2. Units of observation and key measures that can be calculated . . . . . . . . . . . 50
Table 3. Overlap between couples on exclusive dwelling ownership status,
by gender of respondent, Mongolia and the Philippines (percentage) . . . . . 91
Table 4. Design effects for (a) up to three adult members from each household;
and (b) one adult member from each household . . . . . . . . . . . . . . . . . . . . . . . 93
Table 5. Weighting effect (1 + L) due to unequal selection probability within
households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Table 6. Required sample sizes and cost calculation for prevalence rate estimate . . . 96
xiv

Page

Table 7. Required sample sizes for intrahousehold gender analysis . . . . . . . . . . . . . . . 97

Table 8. Questionnaire content by data collection objectives . . . . . . . . . . . . . . . . . . . . 104
Table 9. Relevance of asset-related information by type of asset . . . . . . . . . . . . . . . . . 104
Table 10. Calculating non-response adjustment weight . . . . . . . . . . . . . . . . . . . . . . . . . . 145
Table 11. Illustrative example of post-stratification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Table 12. Indicators of the bundle of ownership rights: rationale and asset coverage 151
Table 13. Indicators of additional aspects of asset ownership: rationale
and asset coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
Table 14. Organization of the data file using individuals as the unit of observation . . 157
Table 15. Organization of the data file using assets as the unit of observation . . . . . . 157
Table 16. Organization of the data file using individuals as the unit of analysis . . . . . 158
Table 17. Distribution of adult population (age 18 and older) by gender and
reported ownership of the principal dwelling (Uganda, 2014) . . . . . . . . . . . . 160
Table 18. Distribution of reported owners and non-owners of the principal
dwelling by gender (Uganda, 2014) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
Table 19. Prevalence of reported ownership of principal dwelling, by gender
and age (Uganda, 2014) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
Table 20. Estimates of adjusted odds ratios in a logistic regression model
predicting ownership of principal dwelling (Mongolia, 2015) . . . . . . . . . . . . 165
Table 21. Mean and median values of principal dwelling owned by women
and men (Mongolia, 2015, and Uganda, 2014) . . . . . . . . . . . . . . . . . . . . . . . . . 167
Table 22. Prevalence of reported ownership, mean value of individual-level
wealth among owners and mean value of wealth (in millions of shillings
for Uganda and in millions of tugriks for Mongolia) among all persons,
by gender and asset type (Uganda, 2014, and Mongolia, 2015) . . . . . . . . . . . 168
Table 23. Individual wealth by household type for major assets (principal
dwelling, agricultural land, financial assets and non-farm enterprise
assets) (Uganda, 2014) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Table 24. Intrahousehold ownership of selected assets based on self-reporting
both spouses in couple households (Mexico, 2014, Uganda, 2014,
and Mongolia, 2015) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Table 25. Intracouple wealth difference and women’s share of wealth among
couples (Uganda, 2014) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
Table 26. Distribution of households based on self-reported dwelling ownership
of both spouses or partners in couple households, Georgia, 2015 . . . . . . . . . 174
 1

Introduction

Purpose of the Guidelines and key recommendations

1. The present publication provides national statistical agencies and poli-
cymakers with guidance on collecting, processing, analysing and disseminating
­individual-level data on asset ownership for the production of gender statistics related
to three objectives: measuring the gender asset gap, or the differential prevalence of
women’s and men’s asset ownership; measuring the gender wealth gap, or the differ-
ential total wealth held by women and men; and, in households where more than one
member is interviewed, understanding how asset ownership and wealth are distrib-
uted by gender within households.
2. Although agricultural surveys and administrative data sources are
briefly considered, the focus of the Guidelines is on household surveys for two rea-
sons.1 First, household surveys are the most flexible instrument for data collection. 1 See chapter II of these Guide-
They can accommodate almost any population-based social or economic subject in lines for a discussion about the
great detail and provide statistics that serve the needs of a wide range of users. Sec- advantages and limitations of
ond, within existing national programmes of data collection, household surveys are household surveys in collect-
ing data on asset ownership
the most developed and frequent source of data. Thus, collecting individual-level data
from a gender perspective in
on asset ownership through household surveys may prove to be more immediate and comparison with other data
less r­esource-intensive than through other sources. Consequently, the methodology sources.
presented in these Guidelines for measuring asset ownership and control from a gender
perspective has been tested in the context of household surveys in select pilot countries.
3. The Guidelines introduce the concepts, definitions and data requirements
for measuring asset ownership and control from a gender perspective and provide
guidance on planning, organizing and implementing a household survey, or append-
ing a module or a set of questions on asset ownership to a nationally representative
household survey. Countries may choose a particular modality for implementation of
the recommendations, depending upon their own needs and capabilities, including
the needs of data users and the availability of data from other statistical and adminis-
trative sources. Guidance on data analysis and dissemination is also provided.
4. These Guidelines provide detailed information on the key decisions to
make for producing statistics on asset ownership from a gender perspective through
household surveys, including the following:
•• Asset ownership should be conceptualized as a bundle of ownership rights,
including documented ownership, reported ownership and the rights to
sell and bequeath an asset. To capture gender differences in the ownership
and control of assets, many countries will need to measure ownership as a
combination of some, or all, of these rights.
•• The priority set of assets on which countries should collect information
are the following: principal dwellings; agricultural land; other real estate,
including non-agricultural land; and financial assets. Countries may also
wish to collect data on non-agricultural enterprises, livestock, agricultural
equipment, valuables, and liabilities and consumer durables, based on
their policy needs and the prevalence of each asset within the country.
2

•• To estimate wealth, each asset should be valued, item by item, at its

current market price.
•• Individual-level data on asset ownership should be reported by self
rather than proxy, owing to large discrepancies between proxy and
self-responses and the assignment of ownership by proxy to persons
who do not consider themselves owners.
•• Within-household selection of respondents should be determined by
survey objectives, data-collection strategies and sample size.
5. Each of these key points is explained in detail throughout the Guidelines.

Relevance of the Guidelines

6. The international guidelines presented in this publication contribute to
the development of gender statistics. Gender statistics are instrumental in building
an evidence base of the driving forces and consequences of gender inequality and in
informing the necessary policy approaches for fostering gender equality and other
development outcomes. Conversely, the lack of adequate gender data is a major impedi­
ment to informed and effective policies.
7. The importance of gender statistics has long been emphasized in global
partnerships. In 1995, in the Beijing Declaration and Platform for Action, 12 critical
areas of concern to women and girls were identified, and Governments were urged to
regularly collect statistics related to each of these areas, to serve as a basis for monitor-
2 Report of the Secretary- ing progress and evaluating the impact of policies. While considerable progress has
General on social statistics been made in producing gender statistics, basic gender data in some areas of critical
submitted to the Statistical interest for policymakers are still non-existent, insufficient or lacking comparability
Commission at its forty-eighth
across countries. Recognizing this data gap, in 2011, the Busan Partnership for Effective
session (E/CN.3/2017/11), avail-
able at https://unstats.un.org/
Development Cooperation called for renewed and accelerated efforts to collect harmo-
unsd/statcom/48th-session/ nized data, disaggregated by gender, for informing policy decisions and guiding invest-
documents/2017-11-Social- ments. Asset ownership was identified as one critical area with a large gender data gap.
Stats-E.pdf. 8. More recently, the 2030 Agenda for Sustainable Development, adopted in
3 Cheryl Doss and others, “The 2015 by the General Assembly of the United Nations, raised the profile and impor-
gender asset and wealth
tance of gender statistics and the need for better data disaggregation. Sustainable
gaps”, Development, vol. 57,
Nos. 3–4 (December 2014), pp. Development Goal 5 of the 2030 Agenda is dedicated to achieving gender equality and
400–409. Available at http:// empowering all women and girls, and 80 of the global indicators have been identi-
doi.org/10.1057/dev.2015.10. fied by the Inter-Agency and Expert Group on Gender Statistics as relevant for gender
4 Agnes Quisumbing and analysis.2 An overview of the relevant target under the Sustainable Development Goals
John Maluccio, “Resources at is set out in box 1.
marriage and intrahousehold
allocation: evidence from Bang- 9. Assets serve multiple functions. In their productive capacity, they gener-
ladesh, Ethiopia, Indonesia, and ate income and facilitate access to capital and credit. They also strengthen a house-
South Africa”, Oxford Bulletin of hold’s capacity to cope with and respond to shocks by enhancing its ability to diversify
Economics and Statistics, vol. 65, income and ease liquidity constraints. Moreover, assets comprise a store of wealth that
No. 3 (July 2003), pp. 283–327.
5 can be liquidated or passed on to future generations. Finally, assets may provide status
Kelly Hallman, “Mother-father
resource control, marriage and security to individuals or households.
payments, and girl-boy health 10. Despite substantial empirical evidence that household members do not
in rural Bangladesh”, Food
fully pool their resources, most official data on assets are collected at the household
Consumption and Nutrition
Division Discussion Paper 93
level, typically by asking a proxy respondent whether anyone in the household owns
(Washington, D.C., International land, housing or other key assets. Yet this approach provides only a partial—and
Food Policy Research Institute, potentially misleading—picture of how asset ownership influences individual and
2000). household welfare. Indeed, prior research3 has found that most assets are owned by
 Introduction 3

individuals, either solely or jointly, thus making individual-level data more reveal- 6 Keera Allendorf, “Do women’s
ing than household-level data for informing evidence-based policies and programmes. land rights promote empow-
Added to this, individual-level data enable gender analysis and also analysis along erment and child health in
Nepal?” World Development,
numerous other dimensions, such as age or marital status, that are important for vol. 35, No. 11 (November
understanding a range of policy issues. For example, while widows and single mothers 2007), pp. 1975–1988.
are recognized as particularly vulnerable groups, relatively little evidence is available 7 Detailed information on the
to understand their asset portfolios. Gender Asset Gap project,
11. Collecting asset data at the individual level, by asking respondents about including survey instruments
their ownership status, provides insights into three broad sets of policy issues: fos- and publications, is available at:
https://sites.google.com/view/
tering the empowerment of women; reducing poverty and vulnerability; and under- genderassetgap/home.
standing livelihoods. 8 Carmen Diana Deere, Gina
Alvarado and Jennifer Twyman,
Empowerment of women “Gender inequality in asset
ownership in Latin America:
12. The importance of women’s ownership and control of assets has long been female owners vs. household
recognized as a key element of the empowerment of women. A call to strengthen wom- heads”, Development and
en’s access to assets, in particular land and financial assets, was made in both the Con- Change, vol. 43, No. 2 (2012),
vention on the Elimination of All Forms of Violence against Women, in 1979, and the pp. 505–530.
9
Beijing Declaration and Platform for Action, in 1995. Ensuring women’s ownership Pradeep Panda and oth-
ers, Property Ownership and
and control of land and other resources is also a key target of the 2030 Agenda. Still,
Inheritance Right of Women for
relatively limited data exist on women’s ownership of assets, in particular data derived Social Protection—the South
from nationally representative surveys. Asia Experience (Washington,
13. The available evidence does find that women’s ownership of assets is posi- D.C., International Center for
tively associated with a number of important development outcomes for the house- Research on Women, 2006);
Stuart Gillespie and Suneetha
hold, including food security, child nutrition and education. For example, mothers’
Kadiyala, “HIV/AIDS and food
ownership of assets is related to the increased educational attainment of daughters in and nutrition security: interac-
Ethiopia, and of sons in Indonesia.4 In Bangladesh, a higher share of women’s assets tions and response”, American
is associated with better health outcomes for girls.5 And in Nepal, mothers who own Journal of Agricultural Econom-
land are less likely to have malnourished children.6 ics, vol. 87, No. 5 (2005), pp.
1282–1288.
14. Women’s ownership of assets is also associated with improvements in their 10 Manasi Bhattacharyya, Arjun
own well-being. Analysis of data collected in Ecuador and Ghana under the Gender
Bedi and Amrita Chhachhi,
Asset Gap project7 found that indicators of women’s asset ownership are correlated “Marital violence and women’s
with more egalitarian decision-making.8 Securing women’s property and inheritance employment and property
rights to land can promote women’s economic security and thus reduce their vulner- status: evidence from North
ability to unsafe sex and other AIDS‐related risk factors.9 While the evidence on the Indian villages”, World Develop-
relationship between asset ownership and spousal violence is mixed, several studies ment, vol. 39, No. 9 (2011),
also indicate that asset ownership can protect against spousal violence.10 pp. 1676–1689; Pradeep Panda
and Bina Agarwal, “Marital
15. Thus, by measuring asset ownership at the individual level, national statis- violence, human development
tical agencies better equip policymakers to understand the empowerment of women and women’s property status in
and their well-being; their economic vulnerability, in particular in the event of house- India”, World Development,
hold dissolution through death, divorce, separation or abandonment; and their bar- vol. 33, No. 5 (2005),
gaining power within the household. pp. 823–850; Shelly Grabe,
“Promoting gender equality:
the role of ideology, power,
and control in the link between
land ownership and violence in
Nicaragua”, Analyses of Social
Issues and Public Policy, vol. 10,
No. 1 (2010), pp. 146–170.
4

Box 1
Measuring women’s ownership of assets in the 2030 Agenda
In 2015, the General Assembly of the United Nations adopted a set of goals to end
poverty, protect the planet and ensure prosperity for all. Building upon the achievements
of the Millennium Development Goals, the 2030 Agenda for Sustainable Development
comprises 17 goals and 169 targets. From a gender perspective, the Sustainable Develop-
ment Goals represent a significant step forward from the Millennium Development Goals,
covering, for the first time, all core areas of the agenda for women's empowerment.
Recognizing that gender equality is critical to achieving the vision set out in the 2030
Agenda, and indeed an objective in its own right, Goal 5 is dedicated to achieving gender
equality and empowering all women and girls. Under Goal 5, target 5.a directs countries
to “undertake reforms to give women equal rights to economic resources, as well as access
to ownership and control over land and other forms of property, financial services, inherit-
ance and natural resources, in accordance with national laws”. FAO is the custodian agency
of target 5.a, which is monitored by two indicators: indicator 5.a.1, a de facto indicator on
women’s land tenure rights over agricultural land; and indicator 5.a.2, a de jure indicator
on women’s land rights in the legal framework.
The two indicators under target 5.a focus on land because this is a key economic
resource inextricably linked to the access, use and control of other economic and pro-
ductive resources. Ownership or, at least, control of land is critical for poverty reduction,
a food security, inclusiveness and overall sustainable development objectives, in many
There is a “growing body
of case study evidence countries. In terms of gender equality, an increase in women’s rights to land is closely
from Latin America connected to the empowerment of women. Indeed, owning or bearing rights to landa
demonstrating that if one reduces women’s reliance on partners and relatives who are men, increases their bar-
compares peasant women gaining power within the household,b improves their chances of obtaining extension
landowners with those services and credit and encourages them to undertake and expand their investments
who are landless, women and join producer organizations.
landowners have a much
greater choice of marriage
Indicator 5.a.1 is divided into two sub-indicators, formulated as follows:
partners and strike a (a) Proportion of total agricultural population with ownership or secure rights over
stronger marriage bargain. agricultural land, by gender;
Within marriage women
(b) Share of women among owners or rights bearers of agricultural land, by type of
landowners play a greater
role in both household
tenure.
and farm decision-making, While sub-indicator (a) measures the prevalence of people in the agricultural popula-
including productive tion with ownership or tenure rights over agricultural land (disaggregated by gender),
decisions and those sub-indicator (b) focuses on gender parity, measuring the extent to which women are
governing the disposition disadvantaged in ownership and tenure rights over agricultural land.
of what is produced and
how income so generated Indicator 5.a.1 focuses on agricultural land, which, in compliance with the classifica-
is used” (Carmen Diana tion provided by the World Programme for the Census of Agriculture 2020, includes “land
Deere and Magdalena Leon, under temporary crops”, “land under temporary meadows and pastures”, “land tempo-
“The gender asset gap: land rarily fallow”, “land under permanent crops” and “land under permanent meadows and
in Latin America”, World pastures”. All the forms of land that are not considered agricultural are excluded from the
Development, vol. 31, No. 6 indicator. According to the operational guidelines of the World Programme for the Census
(2003), pp. 925–947). of Agriculture 2020, the greenhouses and land in family gardens are included in the land
b For instance, evidence for under temporary crops or land under permanent crops.
this was found in South
Asia. See Bina Agarwal, A
Indicator 5.a.1 uses “agricultural population” as the reference population (denomina-
Field of One’s Own: Gender tor), instead of the total population, because tenure rights over agricultural land are rel-
and Land Rights in South Asia evant in particular for individuals whose livelihood relies on agriculture. Although there is
(Cambridge, Cambridge no official definition of “agricultural population”, an operational definition of this term has
University Press, 1994). been proposed by FAO for the scope of indicator 5.a.1. FAO suggests that the term “agri-
 Introduction 5

cultural population” should be interpreted as equivalent to “agricultural households”—in

other words, households that operated land for agricultural purposes or raised or tended
livestock in the past 12 months, regardless of the final destination of the production.
Accordingly, individuals are part of the reference population if they are adult and form
part of an agricultural household. The adoption of a household perspective is particularly
important from the gender viewpoint, because, in many agricultural households, women
often consider themselves as not being involved in agriculture, while they provide sub-
stantive support to the household’s agricultural activities.
Based on the recommendations from the seven EDGE field tests, three proxies have
been identified to measure ownership or land tenure rights:
•• Having own name on a legally recognized document;
•• Having the right to sell;
•• Having the right to bequeath.
Individuals are considered owners or holders of land tenure rights over agricultural
land if they present at least one of the three proxies. Since individuals may have the right
to sell or bequeath land even in the absence of legal documents, the indicator combines
legal documentation with alienation rights, in order to render it comparable across coun-
tries. The EDGE pilots show that these three proxies offer the most robust measure of
land tenure rights, ensuring comparability across countries with a diverse prevalence of
documentation.
Considering the recommendations above, the two 5.a.1 sub-indicators can be
expressed through the following mathematical formulas:

Sub-indicator (a)
Number of adult individuals in agricultural households
with legally recognized document on agricultural land
OR the right to sell it OR the right to bequeath it
*100, by gender
Total adult individuals in agricultural households

Sub-indicator (b)
Number of adult women in agricultural households
with legally recognized document on agricultural land
OR the right to sell it OR the right to bequeath it
*100
Number of people in agricultural households
with legally recognized document on agricultural land
OR the right to sell it OR the right to bequeath it

The appropriate data sources for monitoring indicator 5.a.1 are agricultural surveys or
multi-topic household surveys. If multi-topic household surveys are used, it is necessary
to identify agricultural households, which are the reference population of indicator 5.a.1.
In addition, pre-screening and oversampling may be needed, especially in countries or
regions with a low proportion of households engaged in agricultural production. Admin-
istrative data are not recommended for monitoring indicator 5.a.1, mainly because they
c
do not allow focusing on the reference population—namely, adults living in agricultural Additional information on
households.c the calculation of indicator
5.a.1 may be found at
While indicator 5.a.1 focuses on gender parity in ownership and tenure rights over agri- www.fao.org/sustainable-
cultural land, other Sustainable Development Goal indicators recognize the importance of development-goals/
strengthening secure tenure rights for all. Indicator 1.4.2, for instance, measures the “pro- indicators/5.a.1/en/ and
portion of total adult population with secure tenure rights to land, with legally recognized www.fao.org/elearning/#/
documentation and who perceive their rights to land as secure, by gender and by type of elc/en/course/SDG5A1.
6

tenure”. Indicators 5.a.1 and 1.4.2 show similarities and differences. While both relate to
individual rights and promote gender-disaggregated data, indicator 1.4.2 mentions “any
land” and refers to the total adult population, indicator 5.a.1 focuses on agricultural land
and refers to the adult agricultural population. FAO, the United Nations Human Settle-
ments Programme (UN-Habitat) and the World Bank are collaborating to align concepts,
definitions and data-collection tools, to assist countries in the collection and generation of
these indicators. In particular, a joint land tenure module has been designed to generate
the data for calculating both indicators 5.a.1 and 1.4.2.

Reducing poverty and vulnerability

16. Traditional poverty studies measure poverty as flows of income, con-
sumption or expenditure deprivation, but this approach is typically conducted at the
household level and often fails to capture the wide range of vulnerabilities experi-
enced by individuals. Because stocks of assets are accumulated by individuals over
time, an asset-based approach to the study of poverty can provide better insights into
how people manage their vulnerability to poverty than those produced by traditional
poverty studies. Research in this vein identifies households with few to no assets that
are trapped in poverty, households vulnerable to losing their assets and becoming
trapped in poverty and households that are temporarily poor but that will be able to
11 Michael Carter and Christo- acquire additional assets and move out of poverty.11 While the results of these studies
pher Barrett, “The economics vary across countries in respect of the presence of asset poverty traps, it is important
of poverty traps and persis- to note that all of them examine assets at the household level. Collecting data on asset
tent poverty: an asset-based ownership and control at the individual level would provide a more rigorous basis for
approach”, Journal of Develop-
an analysis of how poverty affects different household members.
ment Studies, vol. 42, No. 2
(2006), pp. 178–199.
Understanding livelihoods
17. Women’s lack of access to important productive resources for agricul-
ture—in particular land, agricultural equipment and livestock—hinders their agri-
cultural productivity. Women’s lack of productive assets also inhibits their ability to
become entrepreneurs, generate income and earn livelihoods. Individual-level data on
asset ownership and control can facilitate a better understanding of the conditions
under which women’s and men’s ownership of assets and the interlinkages of those
assets contribute to diverse livelihood activities. These data can provide the basis for
integrated policy packages designed to boost agricultural productivity and entrepre-
neurship.

Key objectives of data collection, policy questions and measures

18. A sound and gender-informed evidence base is essential for the develop-
ment of policies and programmes aimed at promoting gender equality and overall
development. When collecting individual-level data on asset ownership, a range of
statistics can be generated to answer key policy questions relating to three objectives:
measuring the gender asset gap, or the differential prevalence of women’s and men’s
asset ownership; measuring the gender wealth gap, or the difference in total wealth
held by women and men; and, in households where more than one member is inter-
viewed, understanding how asset ownership and wealth are distributed among cou-
ples or by gender within households. The following are examples of policy questions
that may be asked under each objective:
 Introduction 7

(a) Gender asset gap:

•• What is the prevalence of asset ownership among women and men, by
type of asset?
•• What is the share of women among asset owners, by type of asset?
•• Are women and men more likely to own certain assets exclusively or
jointly? Which assets?
•• Does joint ownership confer equal rights over the asset on co-owners?
•• Do men and women acquire assets differently? If so, in which ways?
•• Which individual characteristics are associated with asset ownership
and do these characteristics differ for women and men?
(b) Gender wealth gap:
•• Do women and men possess similar levels of wealth?
•• Is women’s wealth concentrated in the same types of assets as men’s
wealth?
•• How does the composition of wealth vary by gender among wealth
quintiles?
•• Are women overrepresented in the poorest wealth quintiles?
(c) Intrahousehold analysis:
•• How are assets owned by household members or by members of cou-
ples? What are the ownership dynamics?
•• How is asset wealth distributed among household members, by gender?
•• What share of couples’ total wealth is owned by women?
•• Which individual and household characteristics are associated with
agreement among spouses about their ownership status of key assets?
•• Are women’s asset ownership and share of household wealth associ-
ated with greater decision-making in the household (or other prox-
ies of empowerment or well-being depending on the additional topics
included in the survey)?
19. To answer the above questions, three main types of measures can be gener-
ated: first, prevalence gaps, which compare the proportion of the total population, by
gender, who are owners of a particular type of asset, such as dwellings or land; second,
the share of owners, which indicates which proportion of the people who own a par-
ticular type of asset are women or men. Both of these measures are useful for com-
parisons between men’s and women’s asset ownership over time, within and across
countries, but are limited in that they do not indicate whether the quality and quantity
of assets owned vary among men and women owners; and third, gender wealth gaps,
which require data on the value of assets, and account for the quantity and quality
of the assets owned by men and women. Chapter IV, section 2, of these Guidelines
presents a full list of indicators for monitoring men’s and women’s ownership of assets
at the global and national levels, whereas chapter IV, section 4, illustrates how data
analysis can be used to answer policy questions.

Development of the Guidelines

20. The present Guidelines are the culmination of a multi-year, ­multi-stakeholder
initiative led by the Evidence and Data for Gender Equality (EDGE) project to develop
methodological guidance on measuring asset ownership from a gender perspective
(see box 2 for an overview of the EDGE project). The importance of measuring asset
8

ownership at the individual level to facilitate analysis of women’s and men’s well-being
is increasingly recognized by the international community as essential for devising
evidence-based policies and programmes that promote gender equality and other key
development outcomes. The World Bank Living Standards Measurement Study—Inte-
grated Surveys on Agriculture, the Demographic and Health Surveys programme,
the agricultural censuses supported by FAO, the Gender Asset Gap project and the
Women’s Empowerment in Agriculture Index all collect some individual-level data
on the ownership and control of assets. Building upon the conceptual and operational
foundations of that work, and in collaboration with a wide range of national, regional
and global partners, the EDGE project developed the present Guidelines for national
statistical agencies, with a view to the regular production of individual-level data on
asset ownership and control.

Box 2
EDGE project
The Evidence and Data for Gender Equality (EDGE) initiative seeks to improve the integra-
tion of gender issues into the regular production of official statistics, with a view to inform-
ing better evidence-based policies. Building on the work of the Inter-Agency and Expert
Group on Gender Statistics, this multi-year initiative is jointly executed by the Statistics
Division of the Department of Economic and Social Affairs and UN-Women, in collabora-
tion with national statistical offices, ADB, FAO, ILO, OECD and the World Bank. The project
is guided by a steering committee composed of members of the donor community, the
Inter-Agency and Expert Group on Gender Statistics, regional commissions and regional
development banks. EDGE received funding from the Governments of Australia, Canada,
Germany, Ireland, the Republic of Korea and the United States of America.
From 2013 to 2018, EDGE aimed specifically to accelerate existing efforts to generate
internationally comparable gender indicators on health, education, employment, entre-
preneurship and asset ownership through two main activities: by contributing to the
development of the Minimum Set of Gender Indicators (see genderstats.un.org), dissemi-
nating gender-relevant data and metadata on education, employment, health, public life
and decision-making, and human rights; and by developing methodological guidelines
on measuring asset ownership and entrepreneurship from a gender perspective.
To develop methodological guidelines on measuring asset ownership and entrepre-
neurship from a gender perspective, the EDGE project consolidated technical inputs over a
multi-year process from a wide range of stakeholders, including national statistical offices,
regional and international agencies, and researchers with expertise in gender analysis,
asset ownership and entrepreneurship. The project then tested the proposed methodol-
ogy in seven pilot countries—Georgia, Maldives, Mexico, Mongolia, Philippines, Uganda
and South Africa—and refined the methodology based on the lessons learned from the
pilots. The Guidelines were presented to the United Nations Statistical Commission in 2017.
By developing and testing methodologies to collect data on assets and entrepreneur-
ship, the EDGE project has provided national statistical offices with the necessary tools to
include the collection of asset data in their regular statistical programmes and has contrib-
uted to the advancement of research on measuring entrepreneurship data from a gender
perspective.
Consistent with a clear imperative for evidence-based policymaking, the ultimate aim
of the EDGE initiative is to build a cost-effective and sustainable model for integrating
gender issues into regular statistical production while strengthening countries’ capacities
to produce gender data in all policy areas.
 Introduction 9

21. To ensure that the Guidelines are robust, feasible and sustainable, the EDGE
project worked in partnership with the national statistical agencies of seven coun-
tries—Georgia, Maldives, Mexico, Mongolia, Philippines, Uganda and South Africa—
to pilot the methodology. An overview of the EDGE pilots is presented in table 1. The
selection of pilot countries was based on three criteria: first, given the limited scope of
the project, countries had to possess adequate statistical capacity to contribute to the
development of a new methodology; second, countries had to have plans in place to
conduct a survey that could accommodate a module on asset ownership and control
or be willing to implement a stand-alone survey during the project time frame; and
third, countries had to express interest in producing better gender statistics, including
on asset ownership and control. The selected seven countries offered a variety of con-
texts that could influence asset ownership at the individual level, including in terms of
economies, gender norms, legal frameworks and rights to property.
22. Key partners provided financial and additional technical support: the
Asian Development Bank supported the pilots in Georgia, Mongolia and the Philip-
pines, and the World Bank provided technical support for the pilot in Uganda. Fund-
ing and additional technical support for the pilot in Mexico were provided by the
National Institute of Statistics and Geography. Funding and technical assistance for
the pilots in Maldives, Uganda and South Africa were provided by the EDGE project.
23. The seven pilot studies provided an opportunity to test and refine key
aspects of the methodology on measuring asset ownership from a gender perspec-
tive, including conceptual and measurement issues related to questionnaire design,
respondent selection interview protocols and indicator constructs. In Uganda, in
2014, the EDGE project worked in partnership with the World Bank Living Stand-
ards Measurement Study team to conduct a methodological survey experiment12 to 12 For additional information,
assess the relative effects of interviewing different household members about individ- see box 3.
ual-level asset ownership and control. The findings of the study informed the EDGE
pilots implemented over the next two years. In 2015, Mexico appended a module on a
core set of assets to a national household survey, and Georgia, Mongolia and the Phil-
ippines implemented stand-alone surveys on the full range of financial and physical
assets. In 2016, Maldives also appended a module on a core set of assets to a national
household survey and South Africa piloted a stand-alone survey.
24. Throughout this process, the EDGE project held a series of technical meet-
ings, a midterm review meeting and side events during the forty-fifth, forty-sixth,
forty-seventh and forty-eighth sessions of the United Nations Statistical Commission
to solicit input on the methodology from its stakeholders, including national statistical
agencies, the Asian Development Bank (ADB), the Food and Agriculture Organization
of the United Nations (FAO), the International Labour Organization (ILO), the United
States Agency for International Development, the World Bank, and subject-matter and
sampling experts from the Indian Institute of Management Bangalore, the University
of Oxford, United Kingdom, the University of KwaZulu-Natal, South Africa, and the
University of Michigan, United States.
25. The final methodology presented in the present Guidelines, including the
recommendation of key indicators for global and national monitoring, is informed
by the technical input of the EDGE project stakeholders as well as quantitative and
qualitative analyses of the pilot data and lessons learned from implementing the pilot
studies.
Table 1

10
Overview of EDGE pilots

Data collec- Data collection

Country tion dates strategy Asset coverage Sample size Within-household respondent selection
Uganda June– Stand-alone survey Principal dwelling, agricultural land, 2,720 households Five interview settings were tested:
(MEXA)a August livestock, agricultural equipment, non-farm 1. Self-identified most knowledgeable household member—interviewed
2014 enterprises, other real estate, consumer dura- alone, asked about assets owned, exclusively or jointly, by any household
bles, financial assets and liabilities, valuables member;2. Randomly selected member of the principal couple—interviewed
alone, asked about assets owned, exclusively or jointly, by any household
member;3. Principal couple—interviewed together, asked about assets
owned, exclusively or jointly, by any household member;4. Adult (18+)
household members—interviewed alone and simultaneously, asked about
assets owned, exclusively or jointly, by any household member;5. Adult
(18+) household members—interviewed alone and simultaneously, asked
about assets owned, exclusively or jointly, by individual respondent.
Mexico June– Modules appended Principal dwelling, agricultural land, An ENH Principal couple; self-reported and proxy data collection. In households with-
October to national livestock, agricultural equipment, non-farm subsample of out couples, the household member most knowledgeable about the assets
2015 household survey enterprises, other real estate, financial assets 8,204 households belonging to the household and a household member of the opposite gender
(ENH) and liabilities were interviewed.
Georgia September– Stand-alone survey Principal dwelling, agricultural land, 2,783 Principal couple plus a third randomly selected household member;
October livestock, large agricultural equipment, non- households, ­self-reported and proxy data collection. In households without couples, the
2015 farm enterprises, other real estate, consumer nationally household member most knowledgeable about the assets belonging to the
durables, financial assets and liabilities, representative household and two randomly selected respondents were interviewed.
valuables
Philippines September– Stand-alone survey Principal dwelling, agricultural land, 1,536 households, Principal couple plus a third randomly selected household member;
October livestock, agricultural equipment, non-farm representative for ­self-reported and proxy data collection. In households without couples, the
2015 enterprises, other real estate, consumer dura- Cavite province household member most knowledgeable about the assets belonging to the
bles, financial assets and liabilities, valuables household and two randomly selected respondents were interviewed.
Mongolia September– Stand-alone survey Principal dwelling, agricultural land, 2,983 Principal couple plus a third randomly selected household member;
November livestock, agricultural equipment, non-farm households, ­self-reported and proxy data collection. In households without couples, the
2015 enterprises, other real estate, consumer dura- nationally household member most knowledgeable about the assets belonging to the
bles, financial assets and liabilities, valuables representative household and two randomly selected respondents were interviewed.
Maldives May Module appended to Principal dwelling, agricultural land, An HIES One randomly selected adult household member; self-reported data
2016 household, income ­aquaculture, enterprises, other real estate, subsample of collection.
and expenditure financial assets and liabilities 272 households
Survey (HIES) on 3 islands
South August– Stand-alone survey Principal dwelling, agricultural land, livestock, 1,568 households In half of the sample, one randomly-selected adult household member;
Africa September agricultural equipment, non-farm enterprises, in KwaZulu-Natal ­self-reported data collection.
2016 other real estate, consumer durables, financial province In the other half of the sample, one randomly-selected adult household
assets and liabilities, household decision- member and his/her spouse/partner; self-reported data collection.
making module

Note: Principal couple consisted of the person in the household most knowledgeable about assets owned by household members and that person’s spouse or partner.
a For additional information on MEXA, see box 3.
 Introduction 11

Relationship with international standards

and other global guidelines
26. The present Guidelines constitute the first United Nations guidelines for
measuring asset ownership and control from a gender perspective. As such, care-
ful attention was given to ensuring consistency with existing internationally agreed
standards, including concepts and definitions, classifications and recommendations
for data collection. The methodological publications most relevant from this perspec-
tive include the System of National Accounts 2008 (2008 SNA); the Principles and Rec-
ommendations for Population and Housing Censuses, Revision 3; the Organization
for Economic Cooperation and Development OECD Guidelines for Micro Statistics
on Household Wealth (OECD Guidelines); and the operational guidelines of the FAO
World Programme for the Census of Agriculture 2020 (WCA 2020).
27. Differences between the present Guidelines and the publications refer-
enced above do exist, however. For example, differences relating to the coverage of
assets and definitions of ownership reflect the focus of the present Guidelines on the
­individual-level measurement of asset ownership and a strong gender perspective sup-
ported by prior empirical research on gender and property rights. In addition, the
present Guidelines differ from the OECD Guidelines in that they aim to provide guid-
ance to a wider set of countries, both developing and developed, and to emphasize
the operational aspects of collecting the required data through household surveys. All
these differences are explained in the relevant sections of the Guidelines.

Users of the Guidelines

28. The present Guidelines are targeted primarily at national statistical offices
aiming to produce statistics on asset ownership from a gender perspective using
household surveys. It considers conceptual and definitional aspects of measuring asset
ownership at the individual level, the practicalities of planning and implementing data
collection in the field, and hands-on approaches to data analysis and dissemination.
The Guidelines are prescriptive in terms of the conceptual framework that should guide
the measurement of asset ownership, but offer a menu of options for data collection
and analysis that could fit a range of objectives and resources. They provide a com-
mon platform for the range of specialists typically involved in a data-collection pro-
ject, including specialists in gender statistics, household surveys, sampling, national
accounts and agricultural land, data managers, specialists in field operations, and data
analysts. Lastly, the guidelines are designed to be applicable in countries at different
stages of statistical development and with different levels of experience in conducting
household surveys. Aspects specific to the production of asset ownership statistics are
emphasized across all sections, while indicating where these specific aspects fit within
the typical stages of the statistical production in a country.
29. The Guidelines are also relevant to users of data. Data on individual-level
asset ownership and control are important for a variety of users, including govern-
ments, civil society, researchers and the general public. The Guidelines are designed
to improve users’ understanding of how to interpret the available data, including by
taking into account conceptual and measurement issues. They can also improve the
dialogue between users and producers of data, including by creating a more precise
and efficient communication and showcasing targeted statistical products that are
designed to respond to specific users’ needs.
12

Organization of the Guidelines

30. The present Guidelines for Producing Statistics on Asset Ownership from a
Gender Perspective comprise four parts, as follows:
•• Chapter I presents a conceptual framework for measuring asset owner-
ship and control from a gender perspective. It conceptualizes ownership
as a bundle of ownership rights and discusses the importance of collecting
data on ­individual-level asset ownership by self-report rather than proxy.
Chapter I also presents a definition of assets consistent with the System of
National Accounts, discusses terms and definitions related to specific types
of physical and financial assets and presents recommendations for valuing
assets. Chapter I concludes with a discussion of the different units of obser-
vation that can be used to collect data in a survey on individual-level asset
ownership and the different measures that can be generated.
•• Chapter II discusses the role of household surveys and other data sources,
including agricultural censuses and surveys and administrative sources,
in collecting individual-level information on the ownership and control
of assets.
•• Chapter III provides guidance on planning, organizing and implementing
the collection of individual-level data on asset ownership. It reviews data
collection strategies and modes of data collection, provides sampling guid-
ance on within-household respondent selection and presents a set of ques-
tions on asset ownership that countries are encouraged to adapt for their
data collections. Chapter III concludes with a discussion of field opera-
tions, including the organizing of field work, the training of field staff and
the management of field operations.
•• Chapter IV discusses data processing, tabulation, analysis and the dissemi-
nation of results, and provides guidance on how to structure a household
survey data set on individual-level asset ownership and how to weight
the data to adjust for the unequal probability of selection and unit non-
response. Chapter IV also presents a set of indicators for monitoring wom-
en’s and men’s ownership and control of physical and financial assets at the
global and national levels. Lastly, chapter IV illustrates how data analysis
can be employed to answer policy-relevant questions on asset ownership
and discusses the dissemination of findings.
 13

Chapter I
Conceptual framework for measuring asset
ownership from a gender perspective

31. This chapter presents a conceptual framework for measuring asset own-
ership from a gender perspective. To ensure consistency with existing international
standards, the concepts and definitions presented are anchored in the System of
National Accounts (SNA), the internationally agreed conceptual and accounting
framework for recording economic activities for the purpose of analysing and evaluat-
ing the performance of an economy. To ensure that the framework orients data collec-
tion on asset ownership from a gender perspective, the concepts also build upon prior
empirical research on gender and property rights.

Figure 1
Conceptual framework for measuring asset ownership and control from a gender perspective

Bundle of
Reported Documented Right to Right to Women's
ownership rights,
ownership ownership sell bequeath empowerment
self-reported

Legal framework
(statutory law,
customary law,
marital regimes)
Women's Men's
Mode of
assets assets Sustainable
acquisition
Social norms livelihoods
Household assets

Type Individual wealth

•• Principal dwelling
•• Agricultural land (stock of respondent’s assets
•• Agricultural equipment less respondent’s liabilities)
•• Livestock
•• Other real estate Poverty
Household wealth
•• Non-farm enterprise assets alleviation
•• Valuables (stock of all household mem-
•• Financial assets bers’ assets less all household
•• Consumer durables members’ liabilities)

Country context Data collection and analysis Evidence-based policy

14

32. As illustrated in figure 1, the conceptual framework for measuring asset own-
ership from a gender perspective is concerned with assets held by households, including
adult women and men household members. As defined in the 2008 SNA, an asset is “a
store of value representing a benefit or series of benefits accruing to the economic owner
13 European Commission, Interna- by holding or using the entity over a period of time”.13 In the conceptual framework on
tional Monetary Fund, Organi- measuring asset ownership from a gender perspective, household assets may be owned
zation for Economic Coopera- exclusively by one household member or jointly by two or more household members
tion and Development, United or household members and non-household members. The form of ownership, whether
Nations and World Bank,
exclusive or joint, is represented by the overlapping circles in figure 1 labelled “women’s
System of National Accounts
2008 (New York, 2009). assets” and “men’s assets”. The type of ownership may consist of one or more compo-
nents of the bundle of ownership rights—reported ownership, documented ownership,
the right to sell and the right to bequeath—depicted by the overlapping ovals at the top of
figure 1. Both the type and the form of women’s and men’s ownership of assets are influ-
enced by the country context, including the legal framework and social norms governing
property rights, and also the modes by which the assets were acquired.
33. The collection of individual-level data on a range of financial and non-
financial assets is recommended, including those listed in figure 1 in the square
labelled “type”: principal dwellings, agricultural land, agricultural equipment, live-
stock, other real estate (including non-agricultural land), non-farm enterprise assets,
financial assets, valuables and consumer durables. This is because individual-level data
on women’s and men’s ownership of these assets can provide important insights for the
design of evidence-based policies and programmes, including those on the empower-
ment of women, sustainable livelihoods and poverty alleviation. In addition, countries
are encouraged to collect information on the value of assets to reflect additional attrib-
utes of the assets—such as size, quality or location—that are not revealed by a simple
count of women’s and men’s asset holdings, including for the purposes of understand-
ing differentials in the individual wealth held by women and men.
34. Each of these key concepts is discussed in detail in the following sections.

1. Defining asset ownership

35. Deriving an internationally comparable and locally relevant definition of
asset ownership is complicated by the myriad legal frameworks and social norms gov-
erning individuals’ rights to property in different contexts, and also by the challenge
of deriving a definition of ownership that is applicable across a range of physical and
financial assets. For example, tenure rights, or the rules stipulating how property is
allocated within a country, may not accord legal ownership of assets to individuals,
such as in Ethiopia and the United Republic of Tanzania, where the State maintains
nominal ownership of land, complicating both the notion of ownership and the com-
parability of ownership rights across countries. In addition, while documented owner-
ship may be applicable for assets with high economic value, such as land and housing,
it is not applicable for some smaller assets with less economic value, such as small
agricultural equipment and consumer durables. Thus, the present Guidelines concep-
tualize ownership as a “bundle” of ownership rights, comprising some or all of the
following components: documented ownership, reported ownership and the rights to
alienate the asset through sale or bequest.
36. The present section discusses the components of asset ownership in detail,
along with the additional information that countries are encouraged to collect with a view
to gaining an understanding of asset ownership from a gender perspective, including
whether assets are owned exclusively or jointly and the modes by which they are acquired.
 Conceptual framework for measuring asset ownership from a gender perspective 15

1.1. Bundle of ownership rights

37. As illustrated in figure 1, the present Guidelines conceptualize ownership
as a bundle of ownership rights mediated by the legal framework and social norms
governing an individual’s rights to property in a given context. As such, ownership
may comprise some or all of the following components:
•• Documented ownership: documented ownership refers to the existence of
any document recognized by the Government that an individual can use
to claim ownership rights in law14 over an asset by virtue of the individual’s 14 For agricultural land, docu-
name being listed as an owner on the document. For key assets, such as mented ownership may refer
land and housing, the type of documentation conferring ownership, and to ownership or use rights,
the rights accorded under that ownership, will vary according to the tenure given that in some countries
freehold tenure does not exist.
rights in a given country. Formal documentation may, however, include 15 Based on FAO, Multilingual
one or more of the following:15
Thesaurus on Land Tenure
•• Title deed, or a written or printed instrument that effects a legal disposi- (Rome, 2003). See also
tion; www.fao.org/fileadmin/
•• Certificate of occupancy or land certificate, or a certified copy of an user_upload/sustainable_
development_goals/docs/
entry in a land title system that provides proof of the ownership and Metadata_5a1__14022018.pdf.
encumbrances on the land;
•• Legally recognized purchase agreement, or contract between a seller
and buyer to dispose of the asset in question;
•• Legally recognized will or certificate of hereditary acquisition, or a cer-
tificate that provides proof of the land having been received through
inheritance;
•• Certificate of customary tenure, or an official State document recogniz-
ing a particular person as a rightful owner or holder of the land on the
basis of customary law that can be used as proof of legal right over the
land. These certificates include, among others, certificates of customary
ownership and customary use;
•• Certificate of perpetual or long-term lease from the State, or a contrac-
tual agreement between the State and the individual for the tenancy of
land. A lease or tenancy agreement is the contractual document used to
create a leasehold interest or tenancy; 16 Cheryl Doss and others, Gender
•• Certificate issued for adverse possession or prescription, or a certificate Inequalities in Ownership and
indicating that the adverse possessor (a trespasser or squatter) acquires Control of Land in Africa: Myth
the land after a prescribed statutory period. versus Reality (Washington,
D.C., Poverty, Health, and Nutri-
In many contexts, documented ownership may provide owners of land and tion Division, International
housing with better tenure security. For example, households with docu- Food Policy Research Institute,
mented ownership of land may be better able to withstand large-scale land 2013). Available at http://
acquisitions by the private sector than households with no documented land ebrary.ifpri.org/cdm/ref/
ownership. Furthermore, women are more likely to retain ownership of land collection/p15738coll2/
id/127957.
that is documented in their name in the event of household dissolution due to 17
divorce or the death of a spouse.16 Ruth Meinzen-Dick and others,
“The gender asset gap and its
In other contexts, documented ownership may not confer greater tenure implications for agricultural
security, in particular when the institutional frameworks meant to enforce and rural development”, in
property rights are weak and landowners have little understanding of their Agnes Quisumbing and others
rights.17 (eds.), Gender in Agriculture:
Closing the Knowledge Gap,
•• Reported ownership: reported ownership refers to the persons who con- (Rome, FAO; Dordrecht, Neth-
sider themselves to be owners of the asset in question, irrespective of erlands, Springer Science and
whether they possess legal, or documented, ownership of the asset. For Business Media B.V., 2014).
16

example, a respondent may consider herself to own the principal dwell-

ing even though only her husband is listed as an owner on the deed to the
dwelling or she may consider herself an owner of agricultural land that
18
in fact is owned by the State but to which she has long-term use rights.
Cheryl Doss, Ruth Meinzen-
Dick and Allan Bomuhangi,
Because reported ownership measures people’s self-perceptions about their
“Who owns the land? Perspec- ownership status, it need not—and cannot—be objectively verified. It is a
tives from rural Ugandans and key concept for understanding the empowerment effects of asset owner-
implications for large-scale ship from a gender perspective, since we expect the benefits and behaviour
land acquisitions”, Feminist related to asset ownership to be influenced by people’s perceptions of what
Economics, vol. 20, No. 1 (2013), they believe themselves to own.18 Reported ownership can also illuminate
pp. 76–100. Available at http://
important gaps between legislation granting women property rights and
doi.org/10.1080/13545701.2013
.855320. their implementation on the ground. Lastly, in contexts in which the prev-
19 Under some legal systems,
alence of documented ownership for applicable assets remains low, such as
people do not have the right in most of sub-Saharan Africa, reported ownership, along with the aliena-
to choose who will receive tion rights described below, may be the best available proxies of a person’s
the bequest—spouses or ownership status.
children may be guaranteed
•• Right to sell: the right to sell an asset refers to the ability of an individual
some portion of a person’s
inheritance. See Carmen to transfer the asset in question permanently, in return for cash or in-kind
Diana Deere and Cheryl Doss, benefits. This right may be held jointly with one or more individuals. The
“The gender asset gap: what right to sell an asset is the right most commonly associated with owner-
do we know and why does it ship, but the concept is not applicable in areas where laws or social norms
matter?” Feminist Economics, preclude the sale of assets, such as land. In such contexts, information on
vol. 12, Nos. 1–2 (January–April the right to rent out an asset may be collected. This right refers to the abil-
2006), pp. 1–50. Available
ity of individuals to bestow the use rights of the asset in question to other
at http://doi.org/10.1080/
13545700500508056. persons for a specific period of time, in return for cash or in-kind benefits.
20 The legal owner is defined in •• Right to bequeath: the right to bequeath an asset refers to the ability of
European Commission and oth- individuals to give the asset in question, by oral or written will, to other
ers, 2008 SNA, para. 10.5, as the persons after their death. This right may be held jointly with one or more
institutional unit entitled in law individuals. The right to bequeath is also an alienation right, one that may
and sustainable under the law
be more universal than the right to sell, since in many contexts owners can
to claim the benefits associated
with the asset. bequeath assets to their children or other persons even if they are prohib-
21 Ruth Meinzen-Dick and Rajen-
ited from selling them.19
dra Pradhan, “Legal pluralism 38. The conceptualization of ownership as a bundle of ownership rights is
and property rights”, CAPRi aligned with the concept of legal ownership employed in the 2008 SNA,20 but posits
Working Paper (Washington, that, for two key reasons, legal ownership alone is not sufficient for understanding
D.C., International Food Policy
the complexity of individual rights to assets from a gender perspective. First, in many
Research Institute, 2002).
22
countries, the prevalence of ownership documents, which confer upon the owner the
Krista Jacobs and Aslihan Kes,
“The ambiguity of joint asset
ability to claim the asset under the law, remains low. In the absence of any documen-
ownership: cautionary tales tation, the legal owner of a given asset is not easily identified and may be determined
from Uganda and South Africa”, only if an external claim to the asset is made. Second, even when ownership docu-
Feminist Economics, vol. 21, ments exist, claims of legal ownership are complicated by legal pluralism—namely,
No. 3 (2015), pp. 23–55. the coexistence of multiple (often contradictory) types of laws governing individu-
23 Allan Bomuhangi, Cheryl Doss als’ rights to property, including both statutory and customary laws.21 For example,
and Ruth Meinzen-Dick, “Who constitutional or national laws, such as in South Africa and Uganda, may guarantee
owns the land? Perspectives
women equal rights to land ownership, while customary laws or practices, such as
from rural Ugandans and impli-
cations for land acquisitions”;
religious law or long-standing traditions, may prohibit women’s ownership of land and
International Food Policy grant them access only through husbands, fathers, brothers or other relatives who are
Research Institute Discussion men.22 When conflict arises between different types of laws, local law often prevails
Paper 01136 (2011). over statutory law, according fewer property rights to women.23 In this way, a woman
may be a legal owner, nominally, of a given asset but possess few or none of the rights
or benefits associated with legal ownership.
 Conceptual framework for measuring asset ownership from a gender perspective 17

39. For the accounting focus of its macro statistics framework, the SNA also
recognizes economic owners, defined as the institutional units entitled to claim the
benefits associated with the use of the asset in question in the course of an economic
activity by virtue of accepting the associated risks. From a gender perspective, how-
ever, the ability of individuals to claim the benefits associated with the use of an asset
cannot be assumed by virtue of their accepting the risks if prevailing gender norms
allow husbands or relatives who are men to assume command of women’s assets at
their discretion. For example, a woman may assume the risks associated with growing
crops or rearing livestock, while a relative who is a man retains the economic proceeds
from the sale of the produce or animal products. Countries which want to further
tease out the extent to which asset owners retain the right to claim the benefits of an
asset may consider measuring a series of rights to the asset.
40. Central to the conceptualization of ownership as a bundle of rights are two
key notions. First, whether the full set of ownership rights is held in a given country, in
particular with regard to land, will depend on the tenure systems recognized within
that country. Generally, in countries where land markets are well developed and own-
ership is conveyed through individual title, such as in much of Latin America, North
America and Europe, ownership comprises the full bundle of rights. In contrast, in
sub-Saharan Africa and parts of Asia, where much of the land is not registered, indi-
viduals may not possess formal documentation conferring ownership but consider
themselves owners of the land and may even be able to alienate it. In other contexts,
individuals may possess documented ownership of customary land but not be able to
sell it, owing to legal restrictions prohibiting its sale, whereas in countries in which
land is vested in the State, individuals cannot legally own land but can be accorded
documented use rights and may be able to alienate the land through sale or bequest.
41. Second, even when the full set of ownership rights exists in a given con-
text, the rights may not all be vested in one individual. For example, a woman may
consider herself to be an owner of the dwelling in which she resides, and her husband
may agree, but her name may not be listed as an owner on the deed for the dwelling.
Alternatively, her name may appear as an owner on the deed, but she may lack de facto
authority to sell the dwelling owing to local norms mediating her rights to the asset.
42. While variations in the overlap of ownership rights will be observed across
countries, analysis of the data from six of the EDGE pilot studies finds that women
owners, on average, possess fewer of the ownership rights systematically. This holds
true across all types of applicable assets, irrespective of the type of ownership (docu-
mented or reported) or form of ownership (exclusive or joint). For example, in Uganda
76 per cent of men who consider themselves owners of a principal dwelling also report
the right to sell the dwelling, whereas only 46 per cent of women who report them-
selves as owning the dwelling also report the right to sell it. Similarly, 90 per cent of
men reporting ownership of agricultural land report the right to bequeath it, while
only 62 per cent of women reporting ownership of agricultural land also report this
right. In KwaZulu-Natal, South Africa, 75 per cent of men who reported owning agri-
cultural land also report having the right to bequeath the land, versus 67 per cent of
women reported owners.
43. In Georgia, Mexico, Mongolia and the Philippines, the overlap between
reported ownership and the rights to sell and bequeath assets is greater for both men
and women, but the differences between men and women in the degree of overlap are
still statistically significant. For example, in Georgia, 90 per cent of men reported dwell-
ing owners possess the right to sell the dwelling, compared with 80 per cent of women
reported dwelling owners, while in Mongolia, 97 per cent of men reported dwelling
18 Guidelines for Producing Statistics on Asset Ownership from a Gender Perspective

owners have the right to sell, compared with 90 per cent of women reported owners.
In Cavite, Philippines, the corresponding figures are 93 and 88 per cent for men and
women reported dwelling owners, respectively. Finally, in Mexico, 97 per cent of hus-
bands or partners who are men who reported owning agricultural land also reported
the right to sell the land, compared with 89 per cent of wives or partners who are women.
44. Documented ownership confers a higher share of alienation rights than
reported ownership on both men and women in the pilot studies. While in almost
all cases the share of documented women owners with the rights to sell or bequeath a
given asset is still lower than the share of documented men owners with these rights,
the overlap is 90 per cent or greater for both men and women in all pilot countries
except Uganda and KwaZulu-Natal, South Africa, and the magnitude of the differ-
ences between men and women is smaller for documented ownership than reported
ownership in all countries except Uganda. For example, in Georgia, 97 per cent of men
documented dwelling owners possess the right to sell the dwelling, versus 93 per cent of
women documented dwelling owners. Whereas, in Mexico, 98 per cent of men docu-
mented agricultural owners reported the right to bequeath the land, versus 92 per cent
of women documented owners. In KwaZulu-Natal, South Africa, 96 and 88 per cent
of men and women documented landowners, respectively, reported the right to sell
the land. In Uganda, however, only about 60 per cent of women documented dwelling
owners have the right to sell or bequeath the dwelling, compared with 95 per cent of
men owners.
45. Two key implications for countries measuring asset ownership from a gen-
der perspective emerge from the analyses of the EDGE pilot data. First, the extent
to which the bundle of ownership rights is vested in one individual (graphically, the
extent to which the dotted ovals representing the bundle of ownership rights will over-
lap in figure 1 above) can vary considerably across and within countries. Second, to
capture gender differences in asset ownership, many countries will have to measure a
combination of ownership rights. This is particularly true in countries with a low prev-
alence of documented ownership, such as those in sub-Saharan Africa, where multiple
land tenure systems complicate the ownership of land and housing.
46. Accordingly, these Guidelines recommend that, for comparability at
the international level, individuals should be considered asset owners if they have
documented ownership of the asset or the ability to alienate the asset through sale
or bequest. At the national level, and as detailed in chapter III of these Guidelines,
national statistical agencies will need to develop a thorough understanding of the
country context prior to collecting individual-level data on asset ownership, including
an understanding of the statutory and customary laws governing property rights and
the social norms mediating those rights, as represented by the square labelled “country
context” in figure 1. Equipped with such knowledge, national statistical agencies can
thus define ownership at the national level as the strongest bundle of rights available
in that country and may indeed wish to measure the full bundle of ownership rights
depending on the country’s policy objectives.

1.2. Forms of ownership

47. An asset may be owned exclusively by one person or jointly by two or more
persons. While joint ownership between spouses or couples is the most common form
of joint ownership, other patterns of joint ownership are also possible, such as joint
ownership between siblings or between parents and their adult children.
48. Measuring the form of ownership, whether exclusive or joint, is important
because the rights and benefits associated with ownership may differ if a person owns
 Conceptual framework for measuring asset ownership from a gender perspective 19

an asset exclusively or jointly. Further, while joint ownership typically confers some
rights on the owners, the joint owners may not have equal rights to, or benefit equally
from, the asset in question. To fully understand whether women may benefit more
from exclusive or joint ownership of assets, data are needed on both forms of owner-
ship and the rights held under exclusive and joint ownership.
49. Countries are also advised to develop an understanding of the laws regard-
ing property ownership within marriage, because they influence exclusive and joint
ownership among couples. Broadly speaking, marital regimes may be classified into
three types. In a common property regime, all property owned by either member of a
couple is joint property. In a partial community property regime, assets brought to the
marriage or inherited during the marriage remain exclusive individual property, while
all property acquired during the marriage is joint property. Finally, in a separation of
property regime, marriage does not confer any rights to the property of the spouse.
50. In many countries, there is a default regime, but a couple may choose a
different marital regime at the time of marriage. In addition, there may be different
marital systems with different marital property regimes within the same country. For
example, a couple may choose to marry under civil law, customary law or religious
law, and each may have different property arrangements. Collecting information on
which regime applies to a specific couple can assist the interpretation of data on asset
ownership among couples, for example, in assessing which types of marital systems
are associated with women’s ownership of key assets, such as land and housing.24 24
Information on marital regimes
was not collected in the EDGE
pilots. Countries are encour-
1.3. Acquisition of assets aged, however, to at least
undertake a qualitative study
51. In order to develop policies and programmes that promote women’s and on types of marital systems
men’s accumulation of assets, data are needed to understand how women and men and their association with
acquire assets and whether their modes of acquisition differ. The means of acquiring an women’s ownership of key
assets.
asset may also determine the ownership rights that are associated with it. For example,
in some contexts individuals who inherit land or acquire it from the State in perpetuity
may not be able to sell the land or transfer it to non-family members, while individuals
who purchase land may be able to exercise the full range of ownership rights.
52. While countries will need to customize the modes of acquisition accord-
ing to their specific contexts, as discussed in more detail in chapter III, there are a few
modes of acquisition that all countries should consider including in their data collec-
tion. These include allocation through marriage, with a view to ascertaining whether
women’s ownership of key assets is conditional upon their husband’s ownership, and
allocation through inheritance, purchase and government programmes, with a view
to assessing whether these channels can be used to strengthen women’s ownership of
assets. For example, in some countries, daughters and sons may have equal rights to
inherit land but, in practice, parents may bequeath more land to their sons, which sug-
gests that additional research may be needed to understand parental preferences and
whether programmatic opportunities exist to influence social norms around inherit-
ance.
53. Understanding the extent to which women acquire key assets through pur-
chase can also provide important insights into women’s access to land and housing 25 Carmen Diana Deere and Mag-
markets. For example, research in Latin America indicates that the most prevalent dalena León, “The gender asset
means of acquisition of land for women is inheritance. This may suggest that, in Latin gap: land in Latin America”,
America, markets have more of a gender bias than inheritance regimes, since women World Development, vol. 31,
are less likely to acquire land through purchase than through inheritance.25 No. 6 (2003), pp. 925–947.
20

Key points
•• Asset ownership should be conceptualized as a bundle of ownership rights, including
reported ownership, documented ownership and the rights to sell and bequeath an
asset. To capture gender differences in the ownership and control of assets, countries
will need to measure ownership as a combination of some, or all, of these rights.
•• Measuring the form of ownership, whether exclusive or joint, is important. Countries
are advised to develop an understanding of the laws regarding property ownership
within marriage, since they influence exclusive and joint ownership among couples.
•• Collecting data on modes of acquisition helps in understanding how men and women
acquire assets and whether their modes of acquisition differ and, subsequently, in
developing policies and programmes that promote women’s and men’s accumulation
of assets. The most common modes of acquisition include allocation through mar-
riage, allocation from the Government, inheritance and purchase.

2. Respondent rules for measuring asset ownership

from a gender perspective
54. It is recommended that information on individual-level asset ownership
be self-reported rather than collected by proxy, owing to large discrepancies between
proxy and self-response information, including the assignment of ownership by proxy
to persons who do not consider themselves owners. Each rationale, including the impli-
cations for data collection, is explained in detail below.

2.1. Differences between proxy and self-reported estimates

of women’s and men’s asset ownership
55. Central to collecting data at the individual level on the ownership and con-
trol of assets is the question of whether the information can be collected by proxy or
should be self-reported. While some large-scale household survey programmes, such
26 Proxy responses are accepted as the demographic and health surveys and labour force surveys,26 collect self-reported
for household members data from multiple household members, many national statistical agencies that col-
unavailable for interview in lect individual-level data from household surveys minimize costs by obtaining proxy
labour force surveys, but ILO information from the head of the household or the person most knowledgeable about
guidelines caution that proxy
the survey topic.
respondents may provide inac-
curate information, which can 56. Collecting information on self-reported asset ownership has an important
bias labour force statistics. RaIf implication for policy and programme design in such areas as women’s empowerment,
Hussmanns, Farhad Mehran livelihood strategies and poverty reduction. This is because the success of interven-
and Vijay Verma, Surveys of
tions is likely to be driven by people’s self-perceptions of what they own rather than
Economically Active Population,
Employment, Unemployment what other people think they own. Collecting individual-level information on asset
and Underemployment: An ownership by proxy may be problematic for several reasons. First, there may be an
ILO Manual on Concepts and incomplete pooling of information within households. For example, the head of house-
Methods (Geneva, International hold may be aware of the full stock of assets but unable to accurately identify who the
Labour Office, 1990). owners are. Second, prevailing gender norms about asset ownership may bias proxy
responses about the ownership status of household members. For instance, because
assets are a store of wealth and thus enhance the status of individuals, the head (who is
often a man) may inflate his ownership of assets relative to his wife’s or other women’s
in the household. Third, individual household members may have different under-
standing about who owns a particular asset, especially in countries where ownership
rights are not clearly delineated or for types of assets for which documentation of own-
 Conceptual framework for measuring asset ownership from a gender perspective 21

ership is not common, such as consumer durables. Lastly, because reported ownership
measures people’s perceptions of whether they consider themselves to be asset owners,
irrespective of documented ownership or alienation rights, it is assumed that other
household members are not fully privy to individuals’ thoughts about their reported
ownership status.27 As such, individual ownership of assets should be self-reported 27 Robert Groves, Survey Errors
rather than by proxy, unless evidence suggests that there is no difference in ownership and Survey Costs (Hoboken,
level and patterns collected through self-reporting or by proxy. New Jersey, John Wiley and
Sons, 1989).
57. Only a few studies have systematically assessed the effects of using proxy
data in lieu of self-reported data. For example, in an analysis of a randomized survey
experiment in the United Republic of Tanzania in which both self-reported and proxy
data were collected for a labour force module, response by proxy rather than self-report
had no effect on women’s labour force participation rates, but resulted in a decrease of
labour force participation by men by about 12 percentage points. The effects on labour
force participation by men are attenuated (although still large) when proxy respond-
ents are spouses, suggesting that spouses may have more accurate information on the
employment status of their partners than other household members.28 Still, proxy 28 Elena Bardasi and others, “Do
responses by spouses are likely to suffer from imperfect information sharing or response labor statistics depend on how
bias as demonstrated in an analysis of the effects of proxy versus self-reported data on and to whom the questions are
household income, in which, in 66 per cent of sampled households in Malawi, husbands asked? Results from a survey
experiment in Tanzania”.
underestimated the earnings of their wives by an average of 47 per cent.29 Similarly, in a 29 Monica Fisher, Jeffrey Reimer,
study assessing the effects of information asymmetries on farm production in Ghana, it
and Edward Carr, “Who should
was found that spouses poorly estimated each other’s income and expenditure.30 be interviewed in surveys of
58. As no similar studies had been done in the context of individual-level asset household income?” World
ownership, the EDGE project worked in partnership with the World Bank Living Development, vol. 38, No. 7
Standards Measurement Study team and the Uganda Bureau of Statistics to implement (2009), pp. 966–973.
30
a randomized survey experiment in Uganda that tested the relative effects of interview- Joyce Chen and LaPorchia
Collins, “Let’s talk about the
ing different household members and collecting proxy versus self-reported data on the
money: spousal communica-
ownership of assets (see box 3 for an overview of the experiment, formally known as tions, expenditures and farm
the Methodological Survey Experiment on Measuring Asset Ownership from a Gender production” American Journal
Perspective, or MEXA). An analysis of the extent of differences between self-reports and of Agricultural Economics,
proxy reports in MEXA found that the collection of proxy information from the house- vol. 96, No. 5 (October 2014),
hold head yielded estimates of men’s and women’s asset ownership that differed from pp. 1272–1290.
those obtained by asking respondents to self-report their ownership status. For exam-
ple, response by self-report rather than proxy increased women’s reported ownership of
the principal dwelling by 14 percentage points and men’s reported ownership by 11 per-
centage points. Response by self-report rather than proxy also increased both women’s
and men’s reported ownership of agricultural land in Uganda, although the increase
was greater for men (10 percentage points) than for women (5 percentage points).
59. Similar patterns were observed in the EDGE pilots in Georgia, Mongolia
and the Philippines, where self-reporting also increased the probability of ownership
of the principal dwelling for both women and men. For example, in Mongolia, self-
reporting increased men’s reported ownership of the principal dwelling by 10 percent-
age points and women’s by 5 percentage points. In Cavite, Philippines, self-reporting
increased women’s reported ownership of the principal dwelling by 7 percentage
points and women’s documented ownership by 6 percentage points. In Georgia, the
increase in reported ownership in self-reporting compared to proxy-reporting was 2
percentage points for men and 5 percentage points for women. The prevalence of agri-
cultural land ownership was estimated only in Georgia; there was a low prevalence of
agricultural land ownership in the Mongolia and Philippines samples. Self-reporting
increased both women’s and men’s reported ownership of agricultural land in Georgia
by 7 and 3 percentage points, respectively.
22

Box 3
Overview of the methodological experiment on measuring
asset ownership from a gender perspective in Uganda
In 2013, the EDGE project formally established a partnership with the Living Standards
Measurement Study programme for the design, implementation and analysis of a meth-
odological household survey experiment to test different respondent selection protocols
for collecting data on asset ownership and control at the individual level. The Uganda
Bureau of Statistics, an early partner of the EDGE project, was selected to implement the
experiment in Uganda given its strong statistical capacity and longstanding partnership
with the Living Standards Measurement Study. Formally known as the Methodological
Experiment on Measuring Asset Ownership from a Gender Perspective, or MEXA, the sur-
vey was implemented on the World Bank Survey Solutions Computer-Assisted Personal
Interviewing (CAPI) platform from May to August 2014, with in-country training, survey
management, field supervision, data processing and quality control support from the
Living Standards Measurement Study. The findings from MEXA and the operational chal-
lenges of implementing the experiment, both of which are discussed in these Guidelines,
informed the six EDGE pilot studies implemented over the following two years, and also
the continuing work by the Living Standards Measurement Study team. The totality of this
work forms the basis for the best practices recommended in these Guidelines.

Questionnaire design
The MEXA questionnaire consisted of two parts: first, a household questionnaire compris-
ing a household roster (of people, not assets) and a short module on dwelling character-
istics administered to the self-identified most knowledgeable household member; and,
second, an individual questionnaire comprising modules on the ownership and control of
the principal dwelling, agricultural land, large and small livestock, large and small agricul-
tural equipment, non-farm enterprises and enterprise assets, other real estate, consumer
durables, financial assets, liabilities, and valuables, administered to one or more respond-
ents through the survey treatment arm protocols (described in the section on experiment
design below).
For agricultural land, other real estate, non-farm enterprises, and financial assets and
liabilities, an inventory of assets belonging to the household was collected from each
respondent in the individual questionnaire by asking the respondent to itemize the given
assets at the start of each respective module (for example, each agricultural parcel owned
by any household member). The individual questionnaire asked questions on four main
topics: ownership and control of assets; acquisition of assets; valuation of assets; and hid-
den assets. Data were collected on a bundle of ownership rights, including reported and
documented ownership and the rights to sell the asset, bequeath the asset, use the asset
as collateral, make improvements to the asset and claim the economic benefits from the
sale of the asset.

Experiment design
In order to assess the relative effects of respondent selection protocols on key outcome
estimates of women’s and men’s asset ownership and control, MEXA tested the following
five survey treatment arms in which different household members were interviewed:
1. Self-identified most knowledgeable household member, interviewed alone, asked
about assets owned, exclusively or jointly, by any household member;
2. Randomly selected member of the principal couple, interviewed alone, asked about
assets owned, exclusively or jointly, by any household member;
3. Principal couple, interviewed together, asked about assets owned, exclusively or
jointly, by any household member;
 Conceptual framework for measuring asset ownership from a gender perspective 23

4. Adult (18+) household members, interviewed alone and simultaneously, asked

about assets owned, exclusively or jointly, by any household member;
5. Adult (18+) household members, interviewed alone and simultaneously, asked
about assets owned, exclusively or jointly, by the individual respondent.

Sample design
A key consideration in determining the sample size for MEXA was the requirement that
households allocated to treatment arms 2 and 3 had to include a couple (either married
or cohabitating) among the adult household members, by virtue of the requirement that
a randomly selected member of the principal couple be interviewed in treatment arm 2
and that both members of the principal couple be interviewed together in treatment
arm 3. Although a full household listing was conducted prior to sample selection, infor-
mation on whether a couple resided in the household was not collected, for reasons of
cost and timing constraints. Instead, the sample design oversampled across all treatment
arms to account for the rate of households with a couple in Uganda (being approximately
66 per cent). Factoring in a non-response rate of roughly 10 per cent at the enumeration
area level in the survey programme of the Uganda Bureau of Statistics, 544 households
were initially allocated to each treatment arm.
In total, the experiment attempted to cover 140 enumeration areas, with an urban-to-
rural split of 84 to 56, across Uganda, selected with a probability proportional to the size
of the enumeration area. The actual enumeration area coverage was 137. In each com-
pleted enumeration area, 20 households were selected, using systematic sampling with a
random start, and four households were randomly allocated to each of the five treatment
arms for a total sample size of 2,720 households.
In treatment arms 4 and 5, in which multiple adult household members were inter-
viewed, the number of respondents was capped at four for each household for logistical
reasons, which resulted in a negligible number of adults being missed. If a household had
more than four adult members who were eligible for an interview in treatment arms 4 and
5, the teams made sure to target the household head and, where applicable, the spouse,
with the rest of the respondents selected at random.

2.2. Implications of respondent rules on interviewing protocols

60. The collection of self-reported data has implications for how respond-
ents are selected for interview within households. Respondent selection is discussed
in detail in chapter III, section 4, on sample design, but two conceptual issues aris-
ing from the requirement to collect self-reported data are highlighted here. First, it
is recommended that respondents be interviewed alone, owing to the sensitivity of
questions about asset ownership and wealth and the potential bias introduced by the
presence of others during the interview. The collection of information about the value
of a person’s assets may indeed be quite sensitive, as noted in the OECD Guidelines,31 31 Organization for Economic
which contain recommended practices on measuring household wealth. In the pres- Cooperation and Develop-
ence of other household or non-household members, a respondent may be less inclined ment, OECD Guidelines for Micro
to reveal such information. For example, in the EDGE pilot in Georgia, some women Statistics on Household Wealth
(Paris, 2013).
respondents were afraid that their husbands, if nearby, would hear the answers that
they wanted to give to the interviewers. Non-respondents may also try to intervene
in the interview and provide answers for the respondent, as was the case in the EDGE
pilot in Cavite, Philippines, when wives were interviewed while their husbands were
at home. This phenomenon may be more pronounced in areas with strong customary
views of women’s roles. For instance, in three regions of Georgia with relatively high
proportions of ethnic minorities, field staff observed that husbands or other household
24

members who are men insisted on sitting in during the interviews of women house-
hold members and often “corrected” them when they reported owning assets, because
it is not the custom in their society for women to own assets.
61. Second, if national statistical agencies opt to interview more than one
respondent per household, it is recommended that the interviews be conducted con-
secutively (one immediately after another) to mitigate the contamination of data
that may result when household members discuss the content of the questionnaire
and coordinate their answers, accordingly. For example, if given the opportunity to
exchange notes between interviews, respondent 1 may inform respondent 2 that the
interview will be shorter and less burdensome if she or he reports that she or he owns
no, or few, assets. Or, upon finishing the interview, a respondent may instruct his or
her spouse to provide the same answers to the enumerator to avoid the appearance
of inconsistencies within the household. While it is difficult to quantify the effects of
contamination, national statistical agencies should be aware of such contamination
as a potential source of measurement error and take care to organize the field work to
enable consecutive interviewing, to the extent possible.

Key points
•• Data at the individual level on ownership and control of assets should be collected on
the basis of self-reported data only.
•• The collection of self-reported data has implications for the way in which respondents
are selected for interview within households. Respondents should be interviewed
alone, in view of the sensitivity of questions about asset ownership and wealth. When
more than one respondent per household are being interviewed, the interviews
should be conducted consecutively (one immediately after another) to mitigate the
contamination of data that may result when household members discuss the content
of the questionnaire and coordinate their answers accordingly.

3. Definition and coverage of assets

62. The terms and definitions related to assets presented in this section are
based on, and consistent with, the 2008 SNA, the internationally agreed conceptual
and macroeconomic accounting framework for recording economic activities for the
32 European Commission and purpose of analysing and evaluating the performance of an economy.32 Other global
others, 2008 SNA, 2009. methodological publications were also used, where relevant, including the System
33 United Nations, European of Environmental-Economic Accounting 2012—Central Framework (SEEA Central
Commission, FAO, OECD, ­Framework),33 the OECD Guidelines and the WCA 2020.34 Divergences from those
International Monetary Fund publications that reflect the focus of the present Guidelines on the individual-level
and World Bank, System of
measurement of asset ownership from a gender perspective are explained in the rel-
Environmental-Economic
Accounting 20212—Central evant sections.
Framework (New York, 2014,
Sales No. E.12.XVII.12). 3.1. What is an asset
34 Food and Agriculture Organiza-
63. Consistent with the 2008 SNA (para. 3.30), the present Guidelines define
tion of the United Nations,
World Programme for the Census an asset as “a store of value representing a benefit or series of benefits accruing to
of Agriculture 2020, vols. 1 and 2 the economic owner by holding or using the entity over a period of time”. Economic
(Rome, 2017 and 2018). benefits include primary income and possible holding gains or losses due to changes
in the prices of assets. Also consistent with the 2008 SNA, all assets covered refer to
 Conceptual framework for measuring asset ownership from a gender perspective 25

economic assets, including, for example, buildings, land, equipment, currency, securi-
ties, shares and other equity, loans and accounts receivable.
64. Owing, however, to their focus on measuring asset ownership at the indi-
vidual level, these Guidelines cover only assets held by households, including women
and men household members and the unincorporated household enterprises that they
run. Assets held by other institutional units that are important from the standpoint
of the SNA, including non-financial corporations, financial corporations, government
units and non-profit institutions serving households, are not covered.
65. The coverage of assets in the 2008 SNA is limited to those assets that can be
used in an economic activity repeatedly (for generally one year or more) and that are
subject to ownership rights. As such, resources such as human or social capital, which
are sometimes described in common parlance as “assets,” and also natural resources
that are not owned, such as the air or the oceans, are excluded from the SNA asset
boundary. Also excluded are consumer durables, because the services that they pro-
vide are produced for own use by the household’s members and thus fall outside the
35
production boundary. Countries may consider addi-
tional types of assets, based
66. Consistent with the 2008 SNA, the present Guidelines do not cover human on the prevalence of their
and social capital, although their importance for women’s empowerment, poverty alle- ownership among women and
viation and sustainable livelihoods is recognized. Similarly, natural resources that are men and their relevance for
not individually owned are not covered. Consumer durables are included, however, in policymaking. For example, the
the scope of assets for the purpose of the present Guidelines, in view of their impor- OECD Guidelines recommend
tance to individual and household well-being. This inconsistency with the 2008 SNA the inclusion of intellectual
property products such as
is only partial. Indeed, the 2008 SNA (para. 3.47) recognizes the analytical interest of computer software, databases
information on consumer durables and suggests that it appear as a memorandum item that allow resource-effective
in a country’s balance sheet. The coverage of consumer durables is also consistent with access to and use of the data,
the OECD Guidelines. and entertainment, literary
and artistic originals. Because,
67. Finally, in a manner consistent with the 2008 SNA (para. 2.35), the pre-
however, the prevalence of
sent Guidelines distinguish between financial and non-financial assets. Non-financial ownership of intellectual prop-
assets may be produced during a process that falls within the production boundary of erty rights in the household
the SNA (and may be further classified into fixed assets, inventories, and valuables), sector is likely to be minimal,
while other non-financial assets are non-produced (and further classified into natu- the present guidelines do not
ral resources; contracts, leases and licenses; and purchased goodwill and marketing cover them.
assets). Examples of non-financial assets held by households include dwellings as a 36 James Davies and others,
produced asset and land as a non-produced asset. Most non-financial assets generally “The world distribution of
household wealth”, in Personal
serve two purposes (ibid.). They are primarily objects usable in an economic activity
Wealth from a Global Perspec-
and, at the same time, serve as stores of value. tive (Oxford, Oxford University
68. Financial assets are necessarily and primarily stores of value, although Press, 2008).
they may also fulfil other functions. Some examples of financial assets held by house- 37 Cheryl Doss and others, “Les-
holds include bank deposits, shares, equity in unincorporated enterprises and pension sons from the field: implement-
fund entitlements. For almost all financial assets, there is a corresponding liability. A ing individual asset surveys
in Ecuador, Ghana, India and
liability is always financial and is established when one unit (the debtor) is obliged, Uganda”, Journal of Economic
under specific circumstances, to provide a payment or series of payments to another Inequality, vol. 11, No. 2 (June
unit (the creditor) (2008 SNA, para. 3.5). Loans are one of the most common examples 2013), pp. 249–265.
of liabilities at the household or individual level. 38 Household Finance and
69. Consistent with the OECD Guidelines, the present Guidelines further clas- Consumption Network,
“The Household Finance and
sifies non-financial assets as follows: principal dwellings, agricultural land, other real
Consumption Survey: results
estate—including non-agricultural land, non-agricultural enterprise assets, large and from the second wave”, ECB
small agricultural equipment, livestock, valuables and consumer durables.35 These Statistics Paper Series, No. 18
assets were selected on the basis of their relevance for the household sector and for (Frankfurt am Main, European
measuring asset ownership from a gender perspective, as explained later in the section. Central Bank, 2016).
26

70. Because the patterns of asset ownership vary across countries with differing
levels of wealth,36 each country will need to determine the assets on which information
should be collected. This decision should be based upon the needs of data users, the
consistency with the national SNA framework, the availability of individual-level, gen-
der-disaggregated data from other statistical and administrative sources and, lastly,
the resources available for collecting the data. These Guidelines suggest, however, that
countries collect information, at a minimum, on the following “priority” set of assets,
categorized as such because of their universal applicability (principal dwellings and
financial assets) or relevance for global development monitoring under the Sustain-
able Development Goal indicator framework (agricultural and non-agricultural land).
In addition, in prior studies, these assets have been found to constitute a substantial
portion of individual wealth in Ecuador, Ghana and Karnataka, India,37 and the bulk
of household wealth in European countries.38 The priority set of assets is as follows:
•• Principal dwelling;
•• Agricultural land;
•• Other real estate, including non-agricultural land;
•• Financial assets.
71. Based upon policy needs and the prevalence of each asset within the coun-
try, the latter of which can be determined by existing household-level or holding-level
data from household or agricultural surveys, countries may also wish to collect data on
additional assets. For example, countries whose economies are centred upon the pro-
duction, consumption, trade and sale of agricultural products may also wish to collect
information on the ownership of livestock and agricultural equipment, while industri-
alized economies may opt to collect information on non-agricultural enterprises and
valuables. The additional assets recommended for data collection are as follows:
•• Non-agricultural enterprise assets;
•• Livestock;
•• Large and small agricultural equipment;
•• Valuables;
•• Consumer durables.
It should be noted that, if countries plan to collect information on the value of each
“priority” or “additional” asset for the purposes of calculating individual-level wealth
measures, as discussed in detail in section 4, then data should also be collected on
liabilities.
72. The following section specifies terms and definitions related to each type
of asset.

3.2. Terms and definitions relating to specific types of assets

73. This section presents terms and definitions related to the specific types of
assets covered by these Guidelines. For each type of asset, two aspects are emphasized:
first, the importance of covering the asset and, second, consistency and differences
with existing international standards. Related measurement issues are discussed in
chapter III, section 5, on questionnaire design.

Dwellings
74. Dwellings are one of the most important assets owned by individuals and
households. They serve as a store of wealth and can provide a place to live for owners.
In particular for women, having secure tenure to a dwelling reduces their vulnerability
when the household is dissolved through divorce or death and it provides economic
 Conceptual framework for measuring asset ownership from a gender perspective 27

security. From a policy perspective, information on the ownership of dwellings is key

to an understanding of the forces driving homeownership and to the development of
national and local housing programmes that can reach both women and men.
75. Dwellings can also be occupied by household members on a regular or
occasional basis or used by the household for other purposes, including running and
operating an unincorporated enterprise. They can also be rented out, in whole or in
part, to earn money. These Guidelines distinguish between the principal dwelling,
defined as the main dwelling or housing unit occupied by the household and owned
by one or more of its members, regardless of whether the residence has a mortgage
or loan secured against it, and other dwellings. Other dwellings that are not used as
principal residences are captured, in these Guidelines, within the category of “other
real estate”, together with non-agricultural land. Agricultural land is captured as a
separate category of assets.
76. The definition of dwellings (regardless of their use) in these Guidelines is
consistent with the definition in the 2008 SNA: “dwellings are buildings, or desig-
nated parts of buildings, that are used entirely or primarily as residences, including
any associated structures, such as garages, and all permanent fixtures customarily
installed in residences” (para. 10.68). Some typical examples of dwellings are houses,
semi-detached houses and flats in a block of flats. “Houseboats, barges, mobile homes
and caravans used as principal residences of households are also included” (ibid.).
Furthermore, the definition of the principal dwelling and the distinction between the
principal dwelling and other dwellings adopted by these Guidelines is consistent with
that made in the OECD Guidelines and current practices of data collection on housing
units in household surveys and censuses.
77. One challenge faced in collecting data on dwellings is the need to deter-
mine whether the land on which a dwelling sits should be treated as a distinct category
of asset. Evidence from the EDGE pilot studies shows that, in some contexts, the plot
of land on which the dwelling is located may be owned together with the dwelling,
while in other contexts it may be owned separately. When the land and dwelling are
owned separately, provision needs to be made for a separate measurement. In addi-
tion, some areas of the plot of land on which the dwelling is located may be used for
agricultural production, such as a kitchen garden. Information on the use of the land
of the dwelling for agricultural purposes should also be recorded separately. Chapter
III, section 5, on questionnaire design, shows in detail how to deal with these measure-
ment challenges.

Agricultural land
78. The ownership and control of agricultural land are important for a range
of policy issues, including, for example, agricultural production, food security and
the development of rural communities. In recognition of the importance of this type
of economic resource, particularly for women, indicator 5.a.1 of Sustainable Develop-
ment Goal 5 on gender equality and empowerment for all women and girls directly
refers to the ownership and control over agricultural land (see box 1).
79. Accordingly, these Guidelines recommend that agricultural land be treated
as a distinct category, separate from land that may be used for non-agricultural pur-
poses, which is classified as “other real estate” in the Guidelines. This approach differs
from that of the 2008 SNA and the OECD Guidelines, which do not identify agricul-
tural land as a separate category. In the 2008 SNA (para. 10.175), the focus is on the
overall category of land, defined as a natural resource and a non-produced asset con-
sisting of “the ground, including the soil covering and any associated surface waters”.
28

Agricultural land, however, is a subset of the “land” category classified by use, accord-
ing to the SEEA Central Framework. In that framework, land use reflects the activities
undertaken and the institutional arrangements for a given area of land for the purpose
of economic production or the maintenance and restoration of environmental func-
39 SEEA Central Framework, 2014. tions.39

80. The present Guidelines are consistent with the land use classification in the
SEEA Central Framework and the WCA 2020 in covering the following classes of land
use under the category of “agricultural land”: (a) arable land under temporary crops
(with a less than one-year growing cycle); (b) arable land under temporary meadows
and pastures (cultivated with herbaceous forage crops for mowing or pasture); (c)
arable land that is temporarily fallow (because of crop rotation systems or temporary
unavailability for planting); (d) land under permanent crops; and (e) land under per-
40 FAO, World Programme for manent meadows and pastures.40
the Census of Agriculture 2020,
Volume 1, Programme, Concepts 81. Going further than the SEEA Central Framework, the WCA 2020 distin-
and Definitions (Rome, 2017). guishes the category “land under farm buildings and farmyards”, which refers to areas
under farm buildings such as hangars, barns, cellars, silos and buildings for animal
production such as stables, cow sheds, sheep pens, and poultry yards. Farmyards and
areas under the holder’s house and the yard around it are also included in this cat-
egory. “Agricultural land”, together with “land under farm buildings and farmyards”,
forms the WCA 2020 category “land used for agriculture”, which is equivalent to the
“agriculture” category in the SEEA Central Framework. This is presented schemati-
cally in figure 2.

82. The present Guidelines suggest that data be collected on the ownership of
all categories of agricultural land as described above (see figure 2). Additional infor-
mation is available in box 1 on the definition of agricultural land for measuring own-
ership of assets in the context of the 2030 Agenda.

83. Evidence from the Gender Asset Gap Project and the EDGE pilot surveys
shows that individuals may own one or more parcels of agricultural land. These par-
cels may vary in terms of use (as shown in figure 2), and other characteristics such as
tenure type, size, value, or existing improvements such as irrigation systems. Thus,
countries wishing to collect data on such aspects will have to record the information
parcel by parcel, as indicated in chapter III, section 5, on questionnaire design.

Figure 2
FAO (WCA 2020) classification of land use

Basic land use classes Aggregate land use classes

LU1. Land under temporary crops
LU1-3
LU2. Land under temporary meadows and pastures LU1-4
Arable land LU1-5 LU1-6
Source: FAO, WCA 2020, vol. 1, LU3. Land temporarily fallow Cropland Agricultural Land
2017, p. 68. land used for
LU4. Land under permanent crops
agriculture
Note: Greenhouses and land in LU5. Land under permanent meadows and pastures
family gardens are included in LU1
LU6. Land under farm buildings and farmyards
and LU4 and classified in one of
these categories depending on LU7. Forest and other wooded land
the crop types. LU8. Area used for aquaculture (including inland and coastal waters if part of the holding)
LU9. Other area not elsewhere classified
 Conceptual framework for measuring asset ownership from a gender perspective 29

Other real estate

84. Real estate other than the principal dwelling and agricultural land (already
covered above) includes other residential buildings and spaces, buildings for commer-
cial use, and non-agricultural land. These assets may serve several purposes, including
providing services to one or more household members (such as a secondary vacation
house), serving as a source of income by being rented out, or being used as assets in an
unincorporated enterprise for the purpose of producing and selling goods and services.
85. These Guidelines recommend that information on all categories of other
real estate listed above be collected by countries. Two additional definitional aspects
should be noted. First, consistent with the 2008 SNA (para. 10.71), incomplete dwell-
ings that may be used in the future as a primary residence for the owner should be
listed as other real estate and not in the category of principal dwellings. While they are
not yet used as a primary residence, they are still assets to the extent that the ultimate
user is deemed to have taken ownership, either because the construction is on an own-
account basis or as evidenced by the existence of a contract of sale or purchase.
86. Second, in a departure from the SNA, these Guidelines recommend that
information on the ownership and value of non-agricultural land improvements be
collected together with the ownership and value of the land on which such improve-
ments have been made. In the 2008 SNA (para. 10.79), improvements to land are
treated as a fixed asset separately from the natural asset (agricultural and non-agricul-
tural land included) in its unchanged state. Such improvements may be the result of
land clearance, land contouring, creation of wells and watering holes and other meas-
ures and their value is to be compiled separately in the accumulation accounts and the
balance of sheets of the SNA. This detailed approach might, however, unnecessarily
complicate data collection in household surveys. Instead, countries may consider col-
lecting information on existing improvements to land in additional questions describ-
ing the quality of the land owned.

Livestock

87. Livestock refers to all animals, birds and insects kept or reared in captiv-
ity primarily for agricultural purposes.41 The term includes the following categories: 41 Ibid.
cattle and buffaloes, sheep and goats, horses and other equines, camels and camelids,
poultry, bees and silk worms and others. Domestic animals that may be used as pets,
such as cats and dogs, are excluded, unless they are being raised for sale, food or other
agricultural purposes.42 42 Ibid.

88. The present Guidelines recommend that countries collect information

on asset ownership for categories of livestock that are most relevant in their context.
Those categories may be further refined to include, for example, categories defined by
the purpose of raising the livestock. For instance, the 2008 SNA distinguishes live-
stock that should be considered fixed assets from livestock considered as inventories
(para. 10.92). Included among fixed assets are breeding stocks, dairy cattle, draft ani-
mals, sheep or other animals used for wool production and animals used for transpor-
tation, racing or entertainment. Animals raised for slaughter, including poultry, are
considered “inventory”, a separate category of assets.
89. Countries should not, however, exclude categories of livestock that have a
higher monetary value and would contribute substantially to the wealth of individuals
and households, such as cattle, or categories of livestock that may be more often in the
ownership of women, such as poultry or sheep and goats.
30

Large and small agricultural equipment

90. Agricultural equipment constitutes a crucial asset for many households
and individuals, and is often central to the livelihoods of people living in rural areas.
These Guidelines recommend that countries collect data on the ownership of large
agricultural equipment and consider covering small agricultural equipment as well, if
relevant in their context. It should be noted that, although small agricultural equip-
ment is of limited value, its coverage can be useful for understanding agricultural pro-
ductivity, in particular in poorer developing countries. In addition, the gender gap in
asset ownership may be different where small agricultural equipment is concerned by
comparison to large agricultural equipment.
91. The above recommendation on the coverage of large and small agricul-
tural equipment is consistent with the 2008 SNA and the operational guidelines of
43 Ibid. the WCA 2020.43 In the 2008 SNA, agricultural equipment is a subcategory of fixed
assets relating to machinery and equipment. They are production assets that are used
repeatedly in agricultural production processes for one year or more. The 2008 SNA
recommends that tools that are small, inexpensive and used to perform relatively sim-
ple operations may be excluded from the asset boundary and be treated as materi-
als or supplies for intermediate consumption. Examples of such tools include saws,
spades, knives, axes, hammers, screwdrivers and spanners or wrenches. Nevertheless,
the 2008 SNA (para. 10.35) acknowledges that some flexibility is needed, depending
on the relative importance of such tools in a given country. In countries in which they
account for a significant part of the value of the total stock of an industry’s durable
producers’ goods, they may be treated as fixed assets.
92. A broad concept of machinery and equipment is also used for agricul-
tural censuses, covering all machinery, equipment and implements used as inputs to
44 Ibid. agricultural production.44 This includes everything from simple hand tools, such as
a hoe, to complex machinery, such as a combine harvester. According to WCA 2020,
countries should decide on the type of agricultural machinery and equipment that is
most relevant in their context. Developed countries may focus on machinery such as
tractors, and crop maintenance and harvesting machines. However, less developed
countries may be interested in some animal-powered or even hand-powered equip-
45 Ibid. ment and machinery.45
93. The following categories of agricultural machinery and equipment, distin-
guished in the WCA 2020 operational guidelines, are within the scope of the present
Guidelines and may be adapted to the country context, as is the case in the EDGE pilot
surveys: manually operated equipment such as seeding and fertilizing drills, trans-
planters, threshers, winnowers, sprayers, dusters; animal-powered equipment such
as wooden ploughs, steel ploughs, cultivators, disk harrows, animal carts; machine-
46 Ibid. A complete list of classes, powered equipment, including machines for general farm use, tractors, bulldozers
subclasses and type of machin- and other vehicles, crop machinery and equipment for land preparation, planting,
ery and equipment is included
crop maintenance, crop harvesting, post-harvest equipment; livestock machinery and
in annex A.
47
equipment; and aquacultural machinery and equipment.46
Cheryl Doss and others, “Les-
sons from the field: implement-
ing individual asset surveys
Non-agricultural enterprise assets
in Ecuador, Ghana, India and 94. Enterprises, defined as entities engaged in the production or distribution
Uganda”, 2013; Household
of goods and services mainly for the purpose of sale, are one of the major components
Finance and Consumption
Network, “The Household
of individual and household wealth.47 While enterprises may be considered “assets” in
Finance and Consumption the sense that holding them would bring a series of economic benefits to the owner,
Survey: results from the second these Guidelines, consistent with the 2008 SNA, consider enterprises as economic
wave”, 2013. institutional units that may hold financial and non-financial assets. Ownership of such
 Conceptual framework for measuring asset ownership from a gender perspective 31

productive assets that can be used to start or grow a business play an important role,
particularly for women, in creating self-employment, earning income, and reducing
poverty and inequality.

95. These Guidelines recommend collecting information on the ownership

and control of assets used only in non-agricultural and unincorporated enterprises,
for the following reasons. First, capturing agricultural enterprises is operationally
more challenging, including with regard to separating agricultural activities for own
consumption from activities mainly for the purpose of sale, and thus warrants a sepa-
rate set of recommendations. Moreover, these Guidelines recommend collecting data
on the ownership of key assets that are involved in agricultural production, including
agricultural land, agricultural machinery and equipment, and livestock, in addition
and separately from assets of non-agricultural enterprises.

96. The following definitions are used to distinguish between agricultural

and non-agricultural enterprises. Agricultural enterprises are enterprises engaged
in the production and sale of non-processed agricultural goods (such as milk, wool,
fruits, vegetables) produced on own farm. Non-agricultural enterprises are enterprises
engaged in the production or sale of goods and services other than own-produced,
non-processed agricultural products. It should be noted that the sale of by-products
of agricultural goods (such as cheese, beer, jam, sweaters and other products) is a non-
agriculture enterprise in the manufacturing sector. The sale or trade of agricultural
products purchased from non-household members is also a non-agricultural enter-
prise, in the trade sector. Other examples of non-agricultural enterprises are: making
mats, crafts and bead jewellery, bricks, or charcoal; working as a builder or carpenter;
selling firewood; metalworking; running a street corner stall; providing services such
as haircuts or massages; making local drinks, carpets or baskets; trading in any form
(in food, clothes or various articles), offering services for payment in cash or in-kind,
including for professional activity (e.g., a private lawyer or a doctor).

97. Second, the present Guidelines recommend focusing only on assets in

unincorporated enterprises, consistent with the 2008 SNA and the OECD Guidelines.
As mentioned earlier, the 2008 SNA distinguishes between the household sector and
other institutional sectors, including corporations, the Government and non-profit
institutions serving households. Assets owned by one or more members in a house-
hold are uniquely accounted for in the household sector. Assets owned by other enti-
ties are accounted for separately, in the entity’s corresponding institutional sector. For
instance, assets of incorporated enterprises are recorded in the sectors of non-financial
and financial corporations. Sources such as establishment surveys and business reg-
isters may be used to obtain information on incorporated enterprises, including the
assets that they hold.

98. An incorporated enterprise is defined as a legal entity, “created for the pur-
pose of producing goods and services for the market, that may be a source of profit or
other financial gains to its owner(s); it is collectively owned by shareholders who have
the authority to appoint directors responsible for its general management” (2008 SNA,
para. 4.39).48 An incorporated entity is recognized independently of the other insti- 48 An incorporated enterprise
tutional units that may own shares of its equity. The shareholders are entitled to may also be owned by one
dividends (shares of the enterprise’s income) and, in the event that the enterprise is shareholder, who would hold
wound up or liquidated, they are entitled to a share in the net worth of the corporation all the shares of the enterprise.
remaining after all assets have been sold and all liabilities paid. If, however, a corpora-
tion is declared bankrupt, the shareholders are not liable to repay the excess liabilities
with their own money (ibid., para. 4.40).
32

99. Unincorporated enterprises, on the other hand, often belong to the house-
hold sector. Households are primarily consumer units, but they can also engage in
production, including for the purpose of producing goods or services for sale or bar-
ter on the market. They can range from single persons working as street vendors or
shoe cleaners with virtually no capital or premises of their own to larger manufactur-
ing, construction or service enterprises with employees. When the production units
of households are not legal entities, they are described as household unincorporated
enterprises and they remain part of the same institutional unit as the household to
which they belong (ibid., para. 4.21). The liability of the household members for the
debts of the enterprises is unlimited, and all the assets of the household may be at risk
if the enterprise goes bankrupt. Household unincorporated market enterprises may
also include unincorporated partnerships, where the partners may belong to different
households (ibid., para. 4.156).
100. Some unincorporated enterprises may hold accounts similar to incorpo-
rated enterprises. An unincorporated enterprise can be treated as a corporation only
if it is possible to separate all financial and non-financial assets into those belonging to
the household in its capacity as a consumer from those belonging to the household in its
capacity as a producer (ibid., para. 4.157). The 2008 SNA advises that such unincorpo-
rated enterprises that maintain separate accounts be treated as quasi-corporations, and
the data be presented in the sectors of non-financial and financial corporations. In prac-
49 OECD, OECD Guidelines, 2013. tice, however, it is rare that unincorporated enterprises maintain separate accounts.49
101. These Guidelines recommend that all unincorporated enterprises, regard-
less of whether they maintain separate accounts or not, are treated similarly. This is
consistent with the OECD Guidelines, which argue that unincorporated enterprises
and quasi-corporations share key similarities, including the fact that the risks and
benefits associated with the ownership of assets and the running of the business stay
with the person and not with a legal entity. Therefore, assets and liabilities of any
unincorporated enterprise owned and operated by one or more household members
should be captured by individual-level measures of asset ownership and wealth. As
mentioned earlier, assets of incorporated enterprises are excluded, since these are not
owned by individuals within the household; equity shares in incorporated enterprises
should, however, be included among financial assets that a person may hold.
102. Nevertheless, it is important to note that collecting information on assets
of unincorporated enterprises can be operationally challenging. It is difficult to dis-
tinguish between the assets belonging to an unincorporated enterprise and those that
are used to provide goods and services for own use by the household members. A
dwelling, for example, may be used as a primary residence for the household mem-
bers but also as the place where products meant for market are prepared or crafted.
A vehicle owned by a household may be used not only for the transport of household
members but also to distribute to clients goods produced by the household enterprise.
Chapter III, section 5, on questionnaire design, indicates how to deal with these meas-
urement issues and to ensure that only assets not listed under previous categories of
assets should be included under the category of assets of unincorporated (and non-
agricultural) enterprises.

50 Household Finance and Financial assets and liabilities

Consumption Network, “The
Household Finance and Con-
103. Financial assets are a key component of the wealth of households and indi-
sumption Survey: results from viduals, in particular in industrialized countries.50 Examples of financial assets include
the second wave” (December cash and deposits, shares and debentures, bonds, and also loans made by households
2016). or individuals to others in cash and in kind. Several types of financial assets may be
 Conceptual framework for measuring asset ownership from a gender perspective 33

held by individuals or households, as distinguished and defined by the 2008 SNA and

the OECD Guidelines:
•• Currency and deposits: these consist of notes and coins of fixed nominal
values issued or authorized by the central bank or government and claims
represented by evidence of deposit. Typical forms of deposits relevant for
the household sector include saving deposits, fixed-term deposits, and non-
negotiable certificates of deposits (2008 SNA, paras. 11.52, 11.54 and 11.58).
•• Debt securities: these are negotiable instruments serving as evidence of a
debt. They include bills, bonds, negotiable certificates of deposit, commer-
cial paper, debentures, asset-backed securities, and similar instruments
normally traded in the financial markets (ibid., para. 11.64).
•• Equity and investment fund shares: equity comprises instruments and
records acknowledging claims on the residual value of a corporation after
the claims of all creditors have been met (ibid., para. 11.83). Investment
funds are collective investment undertakings through which investors pool
funds for investment in financial and non-financial assets (ibid., para. 11.94).
•• Insurance, pension and standardized guarantee schemes: these refer to
financial claims of policy holders, account holders or members who con-
tributed with funds to a financial institution in exchange for financial ben-
efits in the same or later periods. Among these types of financial assets,
life insurance, annuity entitlements and pension entitlements are the most
common at the household and individual level. Life insurance and annu-
ity entitlements are defined as “claims of policy holders on enterprises
offering life insurance or providing annuities, except those annuities pur-
chased from lump sums rolled over from pension schemes. These claims
include life insurance entitlements where the insurer guarantees to pay the
policy holder an agreed minimum sum or an annuity at a given date or
earlier if the policy holder dies beforehand” (ibid., para. 17.6; and OECD
Guidelines). Term insurance is a policy that provides a benefit in the case
of death within a given period but in no other circumstances is regarded
as a non-life insurance and is not covered by the measurement of wealth
and asset ownership (2008 SNA, para. 17.6). Pension entitlements refer to
claims of members and account holders on pension schemes such as retire-
ment plans or superannuation schemes and include “entitlements in both
employment-related social insurance pension schemes and private pension
schemes. These claims also include annuities purchased with lump sums
rolled over from pension funds regardless of the institution with which the
annuity is held” (OECD Guidelines). Excluded are entitlements in govern-
ment social security pension schemes.
•• Financial derivatives and employee stock options: these financial assets
are less frequently held by individuals and households. Financial deriva-
tives refer to financial instruments through which specific financial risks
(such as interest rate risk, currency, equity and commodity price risk,
and credit risk) can be traded in their own right in financial markets
(2008 SNA, paras. 11.111–11.112). Employee stock options are agreements
made on a given date under which an employee may purchase a given
number of shares of the employer’s stock at a stated price either at a stated
time or within a period of time immediately following (ibid., para. 11.125).
•• Other financial assets held by individuals or households may refer to loans
made to persons in other households.
34

104. The present Guidelines recommend, however, that national statistical

offices collect information on the ownership of financial assets using the list of types of
assets noted above along with subcategories of those assets, based on their prevalence
in the population and relevance from a policy perspective. In particular, subcategories
of “currency and deposits” may be defined relative to the institutional set-up and refer
to bank savings, savings and credit associations, post-office accounts, informal sav-
ing accounts, saving accounts operated through non-governmental organizations and
other arrangements. Use of such subcategories captures gender differences that can be
relevant to programmes designed to boost women’s access to financial services.
105. Countries are also encouraged to collect data on liabilities. This informa-
tion is needed to estimate the net worth of a person or household, by subtracting the
value of outstanding liabilities from the value of the asset held. By definition, a liabil-
ity (or debt) is established when one unit (the debtor) is obliged, under specific cir-
cumstances, to provide a payment or series of payments to another unit (the creditor)
(2008 SNA, para. 3.5). Most financial liabilities at the level of the household sector are
loans. Loans are defined as obligations that are created when a creditor lends funds
directly to a debtor and the creditor’s claims are evidenced by documents that are not
negotiable (ibid., para 11.72; and OECD Guidelines). Loans may be categorized into
short-term loans (with an original maturity of one year or less) and long-term loans.
106. In addition, countries should consider collecting information on the main
purpose for which the loan was taken out. For example, the OECD Guidelines sug-
gest the following types of loans: principal residence loans and other owner-occupied
loans; other real estate loans; financial asset loans; valuable loans; intellectual property
loans (loans to develop intellectual property products such as a computer software);
vehicle loans; other consumer durable loans; education loans; other loans and liabili-
51 OECD Guidelines, 2013. ties.51 Collecting this information would also enable analysis of whether women and
men borrow money for different reasons. Countries may also consider further splitting
the category “other loans and liabilities” into further subcategories, such as loans for
the purpose of paying medical bills, food and clothing, and so on, which may also be
relevant from a gender perspective. Information on the identity of the lenders, includ-
ing formal or informal institutions or persons from whom the money was borrowed,
is also important in providing evidence for gender-relevant policies and programmes
on access to financial services.

Consumer durables
107. Consumer durables are goods that may be used for the purposes of con-
sumption repeatedly or continuously over a period of a year or more (2008  SNA,
para.  9.42). Examples of consumer durables are cars and other vehicles, furniture,
kitchen equipment, laundry appliances, computers and entertainment equipment. It
should be noted that the same type of durable good may be considered an asset in one
circumstance and a consumer durable in another. For example, a car used as means
of transportation solely for the household members is a consumer durable, while a
car used for transportation of passengers for pay or profit is an asset in an enterprise
providing transport services. Similarly, a computer may be a consumer durable when
used in a household for educating children or paying personal bills or as an item of
personal entertainment but an asset in the equipment category when used to keep
business records for a household-operated enterprise.
108. As noted before, consumer durables are not regarded as assets in the
2008 SNA (but as a form of expenditure) because the services that they provide are not
within the production boundary. However, these Guidelines recognize the analytical
interest of information on the stock of consumer durables, including for the purpose
 Conceptual framework for measuring asset ownership from a gender perspective 35

of measuring household and individual-level wealth. This approach is also consistent

with the OECD Guidelines, which treat consumer durables as non-financial assets. The
OECD Guidelines highlight two main reasons for this treatment. On one hand, the
inclusion of consumer durables in the measurement of household wealth can signifi-
cantly affect the magnitude and distribution of wealth across households. On the other
hand, treating consumer durables as assets ensures greater symmetry with liabilities
data, since households often take out loans to purchase more expensive durables, such
as motor vehicles. In addition, an asset owned by a household may have multiple uses,
including for productive activities and other activities. Conceptually, these multiple
uses can complicate the categorization of a durable good as an asset or a consumer
durable. In practice, however, ownership of such goods that can be used in productive
activities can have a positive impact on livelihoods, particularly for women.
109. The present Guidelines recommend that countries determine which catego-
ries of consumer durables to include on the basis of their prevalence in the popula-
tion and also of the countries’ own policy needs. In general, however, countries should
include durables of high value such as motor vehicles (cars, motorcycles and boats),
together with those durables that are of lower value but that may be of particular impor-
tance to women, such as cell phones, kitchen equipment, or laundry appliances. While
the high-value durables are important from the perspective of the value stored in the
assets and the estimation of individual and household wealth, other durables may be
used in productive and non-productive activities that may be more often performed by
women.

Valuables
110. Valuables include precious metals and stones, fine jewellery, paintings,
antiques or other art objects, and other valuables. Valuables are acquired and held as
stores of value. They are expected to appreciate or at least not to decline in real value
or to deteriorate over time. Thus, although valuables are a type of non-financial assets
they have more in common with financial assets (2008 SNA, paras. 10.13 and A4.57).
111. These Guidelines recommend that the range of valuables covered be wider
than that prescribed by the SNA, including for the purpose of capturing types of valu-
ables that are more relevant for women. The intent of the 2008 SNA is to capture only
those items that can be regarded as alternative forms of investment. Valuables function
as a store of value but they can also be used as collateral in pawn markets or sold quickly
for cash. This can play an important role in consumption-smoothing and building indi-
vidual and household wealth.52 Valuables such as collections of stamps, coins, china, 52 Rania Antonopoulos and Maria
books and other objects that have a recognized market value, and fine jewellery, fash- Floro, “Asset ownership along
ioned out of precious stones and metals of significant and realizable value (e.g., gold) gender lines: evidence from
may be more often held by individual household members, with jewellery in particular Thailand”, Economics Working
Paper No. 418 (Annandale-
being an important asset for women in some countries.
on-Hudson, Levy Economics
Institute, Bard College, 2005).

Key points
•• The priority set of assets on which countries should collect information are the follow-
ing: principal dwellings; agricultural land; other real estate, including non-agricultural
land; and financial assets. Countries may also wish to collect data on non-agricultural
enterprises, livestock, agricultural equipment and valuables, and also on liabilities
and consumer durables based on their policy needs and the importance of each asset
within the country.
36

4. Establishing the value of assets

4.1. Why valuing assets is important
112. Establishing the value of assets allows for the calculation of wealth, or net
worth, which is defined in the 2008 SNA as the value of all non-financial and financial
assets owned by an institutional unit or sector less the value of all its outstanding lia-
bilities (para. 3.109). Similarly, at the micro level of individuals and households, wealth
represents the net value of economic resources held at a point in time by an individual
or a household, measured as the value of all assets owned less the value of all liabili-
53 OECD, OECD Framework for ties.53 The share in monetary terms contributed by each type of asset to the total wealth
Statistics on the Distribution of of an individual or a household is referred to as the composition of wealth.54 The level
Household Income, Consump- and the composition of wealth may vary over time as a result of changes in the assets
tion and Wealth (Paris, 2013). owned and changes in the market prices of assets.55 The valuation of assets reflects a
54 OECD, OECD Guidelines, 2013. range of asset attributes, such as size, quality or location, and allows for the calcula-
55 European Commission tion of a series of measures of wealth level, distribution and composition at individual,
and others, 2008 SNA;
household and macroeconomic levels. All assets can be valued in monetary terms. The
OECD, OECD Guidelines, 2013.
value of an asset represents the total of the benefits (current or future) embodied by the
asset, typically assessed as if the asset was acquired in a market transaction.
113. Wealth may be calculated at the level of a person, household, institutional
sector or an economy. Because the valuation of assets reflects the attributes of the asset,
the calculation of wealth at the individual level provides important information on
gender and asset ownership beyond a simple count of women’s and men’s asset hold-
ings. For example, women and men in a given community may own an equal number
of agricultural parcels, suggesting gender equality in land ownership, but the men’s
parcels may in fact be more valuable because of differences in the characteristics of
women’s and men’s parcels (such as size or soil quality). Expressing asset values in
monetary terms provides a method for summarizing differences between women’s
and men’s ownership of assets, by type of asset, or in an aggregated measure for all
assets. The aggregate measure, referred to as the gender wealth gap, reflects differences
between women and men not only in terms of whether they own assets but also in
terms of the number and quality of the assets that they own. The gender wealth gap
can thus be used to assess differences in women’s and men’s relative holdings, and also
to understand how women’s and men’s wealth relates to other outcomes of interest,
such as investment in agriculture, livelihood strategies, and resilience to shocks.
114. At the household level, the valuation of assets can serve two purposes.
First, valuation can provide the basis for estimating household wealth. This can be
done, as recommended by OECD, in an integrated framework that ensures consist-
ency in measuring household economic well-being along the dimensions of wealth,
56 OECD, OECD Framework, 2013. income and consumption.56 Second, information on the value of some assets, such
as the value of owner-occupied dwellings, may be incorporated into living standard
measures at the household level. For example, consumption analyses incorporate the
rent paid for housing, including an estimated rental equivalent for housing owned by
the occupants. While, however, rentals may be observed directly for renters, for owner-
occupied dwellings a rental value may be imputed when the value of the dwelling is
57 Margaret Grosh and Paul obtained.57 Lastly, information on the values of assets held in a household may be used
Glewwe, Designing Household to validate some of the statistics generated by the SNA, in particular for the balance
Survey Questionnaires for sheets in the household sector, and thus contribute to constructed measures of wealth
Developing Countries: Lessons at the level of the household sector and the national economy.
from Fifteen Years of the Living
Standards Measurement Study, 115. Regardless of the level at which wealth is estimated, it is important that
vol. 1 (Washington, D.C., World assets are not double-counted. In the SNA, for example, rules of accounting are fol-
Bank, 2000). lowed systematically to avoid counting the same asset as being owned in more than
 Conceptual framework for measuring asset ownership from a gender perspective 37

one institutional unit or sector. Data collection on asset ownership at the individual
level, for the purpose of wealth measures, should also ensure that assets are not dou-
ble-counted. If an asset is owned exclusively by an owner, that asset should be listed
only once as belonging to that and only that owner, and its total value should become
a share in the net worth of that owner. If an asset is owned jointly by more than one
owner, the asset should be listed as belonging to all joint owners and, for the purpose
of estimating wealth, its value should be divided into shares that can be apportioned
to the net worth of each owner.

4.2. Principles in establishing value

116. These Guidelines recommend several principles of establishing value for
the purpose of measuring wealth at the individual level. They are consistent with the
2008 SNA and the OECD Guidelines, ensuring the comparability of statistics and indi-
cators of wealth based on individual-level measurement with those based on house-
58
hold and macroeconomic levels of measurement. Alternative methods of
obtaining the values of assets,
117. The first principle is that assets and liabilities should be valued at market although departing to some
prices. Market prices are values at which assets are exchanged (or could be exchanged) extent from the concept of
in actual transactions, in other words, the amounts of money that willing buyers pay current market prices, include
to acquire something from willing sellers (2008 SNA, paras. 3.118–3.119). The second what is referred to as the
principle is that assets and liabilities should be recorded at current values, correspond- “quick sale price” (the price
that would be obtained if the
ing to the value that would be obtained for the asset in its current condition at the ref-
owner sells right away) or a
erence point date, or its closest equivalent, not the value that would have been obtained “reservation price” (the price
when the asset was acquired. that would cause an owner not
118. The third principle relates to the consistency of valuation across assets. intending to sell to be willing
Household surveys collecting information on the value of assets should aim to obtain to sell). Some of these methods
may undervalue the assets—
the information in a consistent manner across all assets, using the same principles of
the “quick sale” approach—or
valuation and time reference. A common approach, used in the World Bank Living overvalue them—the “reserva-
Standards Measurement Study—Integrated Surveys on Agriculture project and the tion price” approach (OECD,
Gender Asset Gap Project, and further tested in the EDGE pilot surveys, is to inquire OECD Guidelines, 2013).
about the amount of money that would be received if the asset were to be sold today.58 59 Cheryl Doss and others, “Do
This method is referred to as the potential sales value or realization value59 and is rec- men and women estimate
ommended by the present Guidelines as the preferred method for collecting data on property values differently?”
the prices of assets in household surveys. When a point estimate cannot be obtained, World Development, vol. 107,
issue C (2018), pp. 75–78.
an interval estimate may be required as a follow-up question. 60 heryl Doss and others, “Meas-
119. The potential sales value method, while not without its challenges, as dis- uring personal wealth in devel-
cussed below, has been shown to provide a more robust measure of asset prices by oping countries: interviewing
comparison to methods based on the construction value or rental value of the asset, men and women about asset
in countries covered by the Gender Asset Gap Project.60 It also allows for consistent values”, Gender Asset Gap
measures of wealth at the individual level, comparable with household and macroeco- Project, No. WPS 15 (Bangalore,
Indian Institute of Manage-
nomic levels of wealth. The application of the potential sales value method, including
ment, November 2013). Avail-
the exact formulation of questions, is illustrated in the section on questionnaire design able at www.researchgate.net/
in chapter III of these Guidelines. publication/270278605_
120. When current market prices are not available, alternative methods of valu- Measuring_Personal_Wealth_
ation may be considered. For instance, the SNA recommends that the observable cur- in_Developing_Countries_
Interviewing_Men_and_
rent market price for the asset in question be used to value non-financial assets. In the
Women_about_Asset_Values.
absence of an observed value, however, averages estimated from observed market val- 61 While the alternative methods
ues for similar assets may be used if the market is one on which the items in question recommended here have not
are regularly, actively and freely traded. Information from markets may also be used to been tested in the EDGE pilots,
price similar assets that are not traded (2008 SNA, para. 13.22). When assets cannot be they are consistent with OECD,
valued at the current acquisition price, as in the case of used assets, such as consumer OECD Guidelines, 2013.
38

durables, for which second-hand markets may not exist, their value may be given by
the current acquisition price of an equivalent new asset less the accumulated deprecia-
tion. This valuation is sometimes referred to as the “written-down replacement cost”
(ibid., para 13.23). Similarly, in the case of financial assets and liabilities, the 2008 SNA
recommends that financial assets and liabilities be valued at current prices if they are
regularly traded on organized financial markets. Financial claims that are not traded
on organized financial markets should, however, be valued by the amount that a debtor
must pay to the creditor to extinguish the claim (ibid., para. 13.54).
121. Countries may consider similar alternative methods for establishing the
value of assets,61 if deemed to be practical, for the purpose of estimating asset prices or
supplementing the information obtained using the potential sales value method. The
use of these methods, however, will depend on the availability of reliable statistical
information from other sources of data, preferably disaggregated at regional or other
subnational levels, and may be suitable for some assets only. Overall, three categories
of alternative methods may be considered:
•• Countries with reliable data on price indices reflecting changes in asset
prices over time may consider collecting information on acquisition prices
and the year when the asset was acquired. This method would typically
apply to real estate items, including the principal dwelling, agricultural
land and other real estate.
•• Countries with existing information on accumulated depreciation for
assets such as specific types of equipment, vehicles or other consumer
durables may consider using information on current acquisition prices of
an equivalent new asset less the accumulated depreciation.
•• Countries may also consider imputing asset values based on statistical
information obtained from other sources of data, including administra-
tive sources (such as property records for tax purposes, land registration
and cadastre systems), land-use surveys and statistical data collection
from local expert informants, such as community and price surveys.
While community and price surveys do not typically cover asset prices,
such items could be considered for inclusion and enumerators could be
trained in obtaining information on unit prices (for example, for different
categories of livestock, agricultural land or agricultural equipment) based
on community information interviews or the visiting of markets.
122. One argument for considering alternative sources for obtaining valuation
data is the potential challenge of obtaining this type of information from household
surveys due, for example, to respondents’ lack of information about asset prices and
the sensitivity of the data being collected. Results from the EDGE pilots show that
only a fraction of respondents report being aware of the existence of sales markets
and informed about recent market transactions. For example, in Uganda, women
owners report information on markets and recent transactions in the location of only
28 per cent of the dwellings and 40 per cent of the agricultural parcels they own. For
men owners, the corresponding proportions stand at 63 and 68 per cent, respectively.
A similar pattern is observed in Mongolia, while in Georgia and Cavite, Philippines,
the proportion of dwellings and agricultural parcels for which the owners have infor-
mation on markets and recent transactions is even lower.
123. In addition, respondents may be unwilling to disclose information per-
ceived as sensitive to enumerators, including the values of the assets that they own.
As a result, a high proportion of non-responses on questions of valuation may arise.
For example, in KwaZulu-Natal, South Africa, 53 per cent of women dwelling owners
 Conceptual framework for measuring asset ownership from a gender perspective 39

reported that they did not know the value of their dwelling and an additional 7 per cent
of women dwelling owners refused to disclose the value of their dwelling. The corre-
sponding proportions for men owners were 41 and 2 per cent, respectively. Similarly
for Cavite, Philippines, a higher percentage of women than men (55  per  cent versus
41 per cent) did not answer the question on value of dwelling.
124. The value of financial assets may be considered particularly sensitive by
respondents. In KwaZulu-Natal, South Africa, 13 per cent of women owners of finan-
cial assets reported that they did not know the value of their financial assets and an
additional 23 per cent of women financial asset owners refused to disclose the value.
The corresponding proportions for men were 18 and 16 per cent respectively.
125. Countries aiming to measure individual wealth will thus need to decide
which methods to use to obtain the values of assets based on an assessment of avail-
able sources and to plan accordingly before the household survey on asset ownership is
implemented in the field. If countries determine to collect asset valuation data through
household surveys, they should draw on any prior experience that the national statisti-
cal office has gained in the collection of valuation data through household surveys and
thoroughly pretest the questionnaire and train field staff to mitigate the challenges
described above. If additional sources on price information are used to impute asset
values in the household survey data set, a set of variables available in the external data
sources related to the characteristics of assets, together with other variables that are
correlated with the value of assets, must also be collected in the household survey. In
addition, if community and price surveys are to be used to obtain the prices of some
assets, the household survey and the community and price survey should be planned
in coordination.62 62 For a description of com-
munity and price surveys
126. Regardless of the method of valuation chosen by countries, it is important
and their integration with
that data be provided with information about the method of valuation used. This infor- multi-topic household surveys
mation is important in assessing the comparability of information across countries. see Elizabeth Frankenberg,
127. Finally, the last principle of valuation refers to the fact that assets should “Community and price data”,
in Designing Household Survey
be valued item by item, even when they belong to the same category. This is because
Questionnaires for Developing
each asset item may have distinct characteristics that determine the market value of Countries: Lessons from Fifteen
that particular asset. In addition, each asset item may be owned by a different number Years of the Living Standards
and set of owners, and this has implications for the calculation of individual wealth Measurement Study, vol. 1, Mar-
because the total value of the asset is to be apportioned between its owners. As dis- garet Grosh and Paul Glewwe,
cussed below, however, there are some instances where item-by-item valuation is not eds. (Washington, D.C., World
feasible, such as in the case of livestock and jewellery valuation, and in these cases it Bank, 2000).
may be more practical to obtain a bulk valuation.

4.3. Which assets to value?

128. The present Guidelines recommend that, in principle, all assets should be
valued. Collecting information on the value of all assets held by a person is key to
obtaining unbiased statistical measures of the distribution of wealth across gender and
other groups of the population.
129. In practice, if countries’ aim is to obtain a full set of measures of the dis-
tribution of wealth, including by gender and wealth quintiles or deciles, it is recom-
mended that the valuation of all assets is attempted. Wealth tends to be concentrated
in the hands of a fraction of the population, and a significant proportion of women
and men may hold no major assets, as evidenced by the EDGE pilots. In addition, the
Gender Asset Gap Project found that the poorest households typically hold the major-
ity of their wealth in consumer durables rather than in land and other real estate. Valu-
40

ing all assets would enable countries to answer more properly questions related to, for
example, gender inequality at the lower end of the wealth distribution.
130. Posing question about the value of all owned assets, large or small,
can, however, complicate data collection and jeopardize the quality of information
obtained. For instance, the EDGE pilot in Uganda included a detailed module on large
livestock that collected information on each animal owned, the number of owners and
the value of the animal, all essential information for calculating individual wealth held
in livestock. Despite focused training, however, the module proved difficult for the
63 In light of the experience enumerators to understand and cumbersome to implement in the field.63
gained in Uganda, the remain- 131. One practical strategy would therefore be to obtain valuation only for
ing EDGE pilots did not collect
major assets that form the bulk of wealth at the individual and household levels. For
valuation data on livestock or
other assets that it had proved example, a key result of the Gender Asset Gap Project was that the principal dwelling,
difficult to itemize. agricultural land and other real estate, including non-agricultural land, constituted
a substantive proportion of non-financial household wealth in Ecuador, Ghana and
64
Cheryl Doss and others, “Do Karnataka, India.64 Countries may consider using this approach and value only prior-
men and women estimate ity assets—the principal dwelling, agricultural land, other real estate and financial
property values differently?”, assets—if the objective is to obtain overall measures of the gender wealth gap or the
2018. average net worth of women and men.

4.4. Who should provide values?

132. The present Guidelines recommend that the self-reported owner of an asset
should also report the value of the asset. When, however, more than one household
member is interviewed, countries may wish to consider collecting information on the
value of non-financial assets from only one household member, ideally from an adult
knowledgeable about the assets belonging to the household and thus capable of con-
structing a household roster of assets (see section 5.2. below, for a discussion on the
rostering of assets). Financial assets should always be valued by their owners.
133. These recommendations are based on both operational feasibility and
analysis of the EDGE pilot data. Across all pilots in which valuation data were col-
lected from multiple household members, the household member most knowledgeable
about the assets belonging to the household had the same likelihood of providing the
potential sales value of the asset in question as the owner of the asset. In addition, in
three of the pilots—Georgia, Mongolia and the Philippines—the differences between
the mean values of the principal dwelling reported by the most knowledgeable house-
hold member and those reported by the respondent owner were not statistically sig-
nificant, suggesting that in these three pilots, valuation estimates were less sensitive to
variations in the individuals reporting the values.
134. In other contexts, however, widely differing asset values may be reported
when more than one household member is interviewed. For example, in Uganda, the
65 Ibid. average gap between the dwelling sales value reported by any respondent and the
66 While the EDGE pilots were dwelling sales value reported by the presumed most knowledgeable person stood at
not designed to test which 34 per cent of the value provided by the most knowledgeable person. Similar differ-
respondent’s evaluation would ences in reported asset values were observed for countries under the Gender Asset
be closest to a “true” asset mar- Gap Project.65 In such cases, reporting discrepancies will need to be reconciled to cal-
ket value, countries may wish culate measures of wealth and without the use of external or secondary sources of
to consider conducting their
valuation data to validate the information obtained from multiple respondents, and
own experiment to answer this
question, including by taking
the necessary methodology for this process is not available.66 Countries may wish to
into account reliable external consider conducting experiments to test which respondent’s valuation would be clos-
sources on valuation that can est to what may be deemed a “true” asset market value. Nevertheless, in the absence
be used for validation. of such evidence, collecting the information from one household respondent (either
 Conceptual framework for measuring asset ownership from a gender perspective 41

one randomly selected member self-reporting his or her assets and their value, or if
more than one member is interviewed, the most knowledgeable person about all assets
owned by any household member), is a pragmatic approach that eliminates the need
to reconcile discrepancies.

4.5. Valuation of specific assets

135. This section discusses issues specific to the valuation of each type of asset
recommended for data collection by these Guidelines. It addresses both the preferred
method of valuation—the current market price operationalized as the potential sales
value—tested by the EDGE pilots and alternative methods not tested by the EDGE
pilots, which may vary from one asset to another.
136. The recommended approaches are consistent with the OECD Guidelines
and the other international statistical guidelines noted above. Countries are encour-
aged to consider which valuation methods would provide the most reliable data in
their specific context and be most cost-effective. It is also recommended that the ques-
tions on valuation are thoroughly tested before the survey is implemented in the field.
137. As the number of countries collecting data on asset ownership and wealth,
including from a gender perspective, are increasing and new practices of valuation
are emerging, the treatment recommended by these Guidelines is expected to evolve
accordingly in the future.

4.5.1. Dwellings and other structures

138. The principal dwelling is considered one of the most valuable assets of a
household and every effort should be made to obtain its value for the purpose of meas-
uring wealth. The recommendation of these Guidelines is that dwellings (principal
dwellings and others) be valued together with the land on which the buildings stand,
unless the ownership of the dwelling differs from that of the land. This recommenda-
tion is consistent with the OECD Guidelines and reflects how real estate market trans-
actions take place (the dwelling and the land are traded as one asset) and, therefore,
how the value of dwellings is most often obtained in practice.67 67 See European Commission and
others, 2008 SNA, 2009. While
139. Houseboats, barges, mobile homes and caravans used as primary resi- the dwelling and the land on
dences of the household, which conceptually are included under the category of which it is located should be
principal dwellings, should be valued in manner similar to that used for vehicles (a valued separately, in practice,
subcategory of consumer durables).68 the information available on
property values for the house-
140. As with other assets, challenges related to the valuation of dwelling include hold sector often covers both
difficulties in obtaining, from household survey respondents, an objective estimate the dwelling and the land and
of the asset value, or any estimate at all, when the markets are thin or non-existing. statistical models are used to
Compared to certain other assets (such as agricultural land, agricultural equipment separate the two components
or livestock), dwellings are more likely to have specific features that render the use of for the purpose of providing
information useful for the SNA,
other sources of data on dwelling prices (such as those based on local expert opinion)
including for the estimation of
to impute valuation less precise. consumption of fixed capital
and net income.
Preferred valuation method 68 OECD, OECD Guidelines, 2013.

141. The preferred approach to the valuation of the principal dwelling and other
dwellings and structures is to use the potential sales value reported by the respondent.
The respondent may be prompted to take into account the price of similar dwellings
that have been sold recently in the location of the dwelling.
42

Alternative valuation methods

142. The following valuation methods may be used as alternatives or in addition
69 Based on OECD, to the preferred method of valuation:69
OECD Guidelines, 2013. •• When a price index for changes in dwelling prices over time is available
from alternative sources of data, the survey may collect information on the
purchase price and the year when the dwelling was acquired. Information
on the type of dwelling and its location may also be needed if the price
index is calculated with the disaggregation of those characteristics. This
method is difficult to use, however, when a large proportion of dwellings
are inherited from previous generations.
•• Valuation of the dwelling for the purpose of property taxes may be used to
establish value in household surveys, if such information is considered to be
relatively well known within the population of the country and the evalua-
tions made by the Government for tax purposes are frequently updated and
based on sound methods. Alternatively, existing information on property
taxes from administrative sources may be adjusted to give a better reflection
of current market prices and combined with the information obtained in the
household survey. For combining methods, the administrative source of data
and the household survey may be directly linked or statistically matched,
based on a common set of variables available in both sources of data.
•• Existing sources such as property records for tax purpose and transaction
sales databases may also be used in making an indirect estimate of the value
of dwellings reported as owned in the household survey. In this approach,
the existing databases are used to specify regression models that predict the
value of dwellings. The derived regression coefficients for a set of variables
are then used to impute the values of dwellings in the household surveys.
It is important that the set of variables specified in the regression model be
70 The list of characteristics collected in the household survey as well. These variables may refer to:70
provided are those used in •• Floor area of the dwelling;
Eurostat, International Labour
Organization, International
•• Location of the property;
Monetary Fund, Organization •• Age of the dwelling (construction year);
for Economic Cooperation and •• Type of structure of the dwelling (detached, semi-detached, unit in a
Development, United Nations multi-family residence building, apartment in a block of apartments, etc.);
Economic Commission for
Europe, World Bank, Handbook •• Materials used in construction (wood, brick, concrete, traditional mate-
on Residential Property Prices rials, etc.);
Indices (RPPIs) (Brussels, 2013). •• Other price determinant characteristics, such as the number of bed-
rooms and bathrooms, garage, swimming pool, air conditioning and
distance to amenities.
•• For selected types of dwellings, such as detached houses, the estimated
current construction cost of the asset (assuming that the size remains the
same and the construction materials are similar) may be requested. This
method may be used when housing markets are thin or non-existent, but
it is important that the value of land on which the dwelling is located is
also valued, and that it is valued separately. A variation of this method is to
obtain construction prices for typical housing units in a community and
price survey implemented at the same time with the household survey.
•• When markets are thin, other measures of housing values such as rental
rates may be collected in the household survey to facilitate the imputation of
data for unit non-response rates based on dwelling values reported by other
respondents in the survey, along with information on asset characteristics.
 Conceptual framework for measuring asset ownership from a gender perspective 43

4.5.2. Land
143. Obtaining the value of land poses challenges similar to those arising in
the valuation of dwellings, in particular in areas where markets are thin. In addition,
some land such as communal land or State land held in long-term lease by individuals
cannot be sold. Information on the value of land needs to be collected parcel by parcel.

Preferred valuation method

144. As with dwellings, the preferred approach in the valuation of land is the
potential sales value reported by the respondent. The respondent may be prompted to
take into account the sale prices of similar parcels of land sold in the area where the
parcel is located.

Alternative valuation methods

145. The following valuation methods may be used as alternatives if field test-
ing reveals that individuals are poorly aware of the cost of land or in addition to the
preferred method:
•• Countries may use prices per unit of agricultural land disaggregated by char-
acteristics of the land (such as type of land use and irrigation), at the subna-
tional level, if available for other statistical purposes (including, for example,
for the purpose of the SNA). In this case, the household survey would collect
information on the area and location of the land and characteristics of the
land but not on the value of the land. The prices per unit of land (per hectare
or equivalent) may be based on prices in actual transactions of land; net-
works of local experts (including local representatives of the ministries of
agriculture, local agents from real estate agencies, experts from the regional
statistical offices); or administrative sources (cadastres, land registries, tax
records).71 It is recommended, however, that the administrative sources be 71 Eurostat 2009 methodol-
carefully reviewed before use, including in terms of coverage, underreport- ogy on land prices and rents;
ing of prices, and availability of information disaggregated at the subna- Eurostat, “Agricultural land
tional level and by type of land.72 In the European context, for example, it prices and rents data for the
European Union”, Research
is recommended73 that, at a minimum, data on prices of agricultural land
Paper (Brussels, 2016).
be provided at the regional level (for Europe, following the nomenclature 72 Ibid.
of territorial units for statistics (NUTS) classification system, at the NUTS 73 Ibid.
2 level). The disaggregation of agricultural land by use should cover at least
the categories “arable land” and “permanent grassland”. Arable land should
further include the subcategories “irrigable arable land” (defined as “arable
land area which could, if necessary, be irrigated in the reference year using
the equipment and the quantity of water normally available on the holding”)
and “non-irrigable arable land” (defined as “arable land which cannot be irri-
gated due to the lack of water for irrigation on the holding”). Consistent with
the SNA, the prices considered should be the prices received or paid by the
holder in free trade without deduction of taxes and levies (except deductible
value added tax) and without the inclusion of subsidies.74 74 Ibid.
•• As with the valuation of dwellings, when a price index for changes in the
land prices over time is available from alternative sources of data, the sur-
vey may inquire about the purchasing price and the year in which the land
was acquired. Because land prices typically vary by the type of use of the
land, it is important that information on the use of land is also collected in
the household survey. This method may be used for both agricultural and
­non-agricultural land (excluding land on which dwellings or other build-
44

ings stand). As mentioned before, however, the method is difficult to use

when land is frequently acquired through inheritance.
•• When markets are thin, other measures of land values such as rental rates
may be collected in the household survey to facilitate the imputation of
data for unit non-response rates based on land values reported by other
respondents and land characteristics.

4.5.3. Agricultural equipment

146. The valuation of agricultural equipment may be difficult in the absence
of markets for used agricultural equipment. These Guidelines recommend that, at a
minimum, the value of large agricultural equipment be obtained, item by item, based
on the current condition of the equipment. Countries may also consider valuing small
agricultural equipment, in particular if small equipment is an important contribution
to the livelihood and wealth of a significant proportion of individuals and households,
as may be the case for households reliant on smallholder agriculture.

Preferred valuation method

147. The preferred valuation method for large agricultural equipment is to ask
the respondent to estimate the potential sales value of each piece of equipment owned
given its current condition. If countries wish to collect the values of small agricultural
equipment that cannot be easily itemized, a suggested approach—although not one
tested by the EDGE project—is to ask the respondent about the number of items in
each group of items and their average age. In this case, information on replacement
value may be obtained from other sources such as community and price surveys. A
general depreciation factor would need to be applied to obtain current practices. For
the purpose of calculating individual wealth, respondent owners should be asked to
estimate their own share of wealth from each group of small agricultural equipment.

Alternative valuation methods

148. The following alternative valuation methods may be considered:
•• When information on price indices and depreciation rates by type of equip-
ment are already available from other sources of data, countries may con-
sider collecting information on the historical cost of the equipment and its
age. This method may be less applicable for small agricultural equipment.
•• Countries may also consider using prices of second-hand agricultural
equipment collected from expert informants in community and price sur-
veys, or from existing sources. For example, countries that have robust
cost-of-production programmes in place that are aimed at measuring the
cost of agricultural production may have information on the value of agri-
cultural equipment. In this case, only data on the type and age of agri-
cultural equipment owned by respondents need to be collected from the
household survey on asset ownership.

4.5.4. Livestock
149. The potential market value for livestock should, in principle, be relatively
easy to obtain. In most places where people raise livestock, there is an active livestock
market. The challenge arising in the valuation of livestock is that, if a farmer owns five
head of cattle, for example, each animal may have a different sales price, depending on
its gender, age and condition. The farmer may also own some of the animals exclusively
and others jointly with one or more people and information must be collected on the
ownership arrangement for each animal in order to apportion its value to its owner.
 Conceptual framework for measuring asset ownership from a gender perspective 45

150. Depending on their policy needs, countries may still wish to itemize large
livestock and collect its value. The valuation of small livestock and poultry would,
however, be extremely burdensome and likely to yield poor-quality data. Thus, if coun-
tries wish to collect the values of livestock that cannot be easily itemized, a suggested
approach—although not one tested by the EDGE project—would be to collect for each
category of livestock only the average sales price per unit (for example, the average
price per goat, per chicken and so forth), either in the household survey or, if they have
been conducted, from community and price surveys. For the purpose of calculating
individual wealth, respondent owners should be asked to estimate their own share of
wealth from each group of livestock.

4.5.5. Financial assets

151. A general challenge for collecting information on financial assets is that
people may be reluctant to report these assets and provide their values. There are wide
differences between countries as to whether their respondents are willing to provide
account balances and whether the enumerators think that it is appropriate to make
such a request. Accordingly, the thorough training of field staff can make a difference
to the quality of the data obtained.
152. In addition, pensions that provide a stream of income over time (such as
annuities) are difficult for respondents to value. Respondents may, however, be able
to provide the amounts in pension savings accounts. Whenever possible, respondents
may be asked to consult their own financial records (such as annual statements) to
improve the accuracy of their responses.
153. These Guidelines recommend that the valuation of financial assets be
obtained directly from the respondent owners, item by item. As a matter of principle,
financial assets and liabilities should be valued at current prices if they are regularly
traded on organized financial markets. Financial claims that are not traded on organ-
ized financial markets should be valued by the amount that a debtor must pay to the
creditor to extinguish the claim.
154. The following specific valuation rules, derived from the OECD Guidelines,
should be observed:75 75 For detailed information on
•• Currency and overnight deposits that can be converted into cash and are valuing financial assets see
OECD, OECD Guidelines, 2013.
transferable on demand (such as transaction accounts, saving accounts,
fixed-term deposits and non-negotiable certificates of deposit, and special
saving accounts) should be valued at their nominal value.
•• Agreed maturity deposits should be valued at the present value of their
expected redeemable value.
•• Bonds and other debt securities should be valued at the market price.
Respondents may, however, report the face value, in which case, countries
may consider adjusting the data obtained by taking into account the mar-
ket interest rates and the interest rate of the securities.
•• The value of shares in corporations should be based on the quotation prices
of the shares for listed companies.
•• Mutual funds and other investments funds should be reported at market
values.
•• Agreed maturity deposits should be valued at the present value of their
redeemable value.
•• Life insurance funds are primarily seen as saving and investment vehicles,
where the entitlement of a payout at the policy maturity date is regarded as
an asset. The value of this payout should be reported.
46

•• For life insurance policies before maturity, the payout depends on the
nature of the policy and may be approximated, for example, by the current
equity of the policy holder in the fund or the surrender value. Countries
would need to determine what types of life insurance policies are most
relevant in their context and the corresponding valuation.
•• Annuities that function as an investment fund should be valued as the
equity remaining in the fund. For annuities with guaranteed payments
for the remainder of the beneficiary’s life, an asset value would need to be
derived by the statistical office based on the schedule of payments obtained
from the respondent and actuarial life expectancy data typically generated
by the statistical office based on demographic data sources.
•• The valuation of pension funds is complex and depends on the type of
pension benefits. A first step for national statistical offices is to determine
the types of pensions relevant in the country. Pension benefits may refer
to three categories: social assistance schemes, social-insurance pension
76 OECD, OECD Framework, 2013. schemes and private pension schemes.76 Social assistance schemes are
non-contributory and should not be considered assets or valued as assets.
Social-insurance pension schemes and private pension schemes are con-
tributory schemes and, from a conceptual point of view, are considered to
be assets. They may be further distinguished as follows:
•• Social-insurance pension schemes are typically defined-benefit schemes
in which the employer or employee or both contribute to a pension fund
throughout the employment time and the benefits reflect the wage level
at retirement and the length of participation in the scheme. The benefits
may be paid as a lump sum or as regular pension payments. If the benefits
are paid as lump sum, the asset value should be equal to the lump sum. If
the benefits are paid as regular pension payments, the asset value would
need to be derived by the statistical office based on the schedule of pay-
ments obtained from the respondent and actuarial life expectancy data
(for the owner and potential survivors that may have benefits) typically
77 OECD, OECD Guidelines, 2013. generated by the statistical office based on demographic data sources. 77
•• Private pension schemes are typically defined-contribution schemes that
function as saving and investment schemes. The benefits received at retire-
ment are lump sums that reflect the contribution made. At the simplest,
the pension benefits, and therefore the asset value, can be approximated
by the current equity accumulated in the fund, and the respondent owners
should be asked to indicate this amount. Countries may consider addi-
tional adjustments of this information according to the specific proce-
78 Ibid. dures determining the final retirement benefits in the scheme.78
•• For loans made to other people, the value of the loan should include the
amount of the original loan and any interest accrued to date (but not in
the future).
4.5.6. Liabilities
155. Liabilities, which are primarily loans taken out by a person individually
or jointly with someone else, should be valued item by item. The value of loans should
be collected from household survey respondents who have taken out the loan or are
responsible for paying back the loan. The value of liabilities is the outstanding balance
of the debt, including any outstanding interest that is currently due. For example, in
the case of a regular mortgage payment, the value of liability is the amount of principal
still outstanding.
 Conceptual framework for measuring asset ownership from a gender perspective 47

4.5.7. Consumer durables

156. Consumer durables are an important contribution to the wealth of indi-
viduals and households, in particular in the poorest households that may not own
major assets. They are even more difficult to value, however, than major assets such
as real estate. Only a small number of consumer durables are traded in second-hand
markets and very few people may be aware of those markets and the transactions tak-
ing place. As a result, most people are unable to estimate a current market value for
most consumer durables.
157. The present Guidelines recommend that, at a minimum, countries should
collect value information on big items likely to cost a significant amount of money,
such as vehicles, including cars, motorcycles, boats, caravans and aircraft. These assets
should be valued item by item in their current condition. The remaining types of con-
sumer durables may be valued group by group. This recommendation is consistent
with the recommendations in the OECD Guidelines on collecting valuation data for
consumer durables.

Preferred valuation method

158. The preferred valuation method for vehicles or other major consumer
durables that can be itemized and for which markets are likely to exist is to ask the
respondent to estimate the potential sales value of such items. If countries wish to
collect the values of smaller consumer durables that cannot be easily itemized, a sug-
gested approach—although not one tested by the EDGE project—is to ask the respond-
ent about the replacement value of each group of items and their average age. A general
depreciation factor would then need to be applied based either on information from
the respondents about how long they expect to keep those durables, or on a standard
factor established by the data analyst for general application.79 In addition, for the pur- 79 Ibid.
pose of calculating individual wealth, respondent owners should be asked to estimate
their own share of wealth from each group of durable goods.

Alternative valuation methods

159. The following three methods may be considered for the valuation of vehi-
cles or other major consumer durables that can be itemized:
•• When information on price indices and depreciation rates for vehicles by
type of vehicle is already available from other sources of data, countries
may consider collecting information on the historical cost of the vehicle
and its age.
•• When databases of second-hand vehicle prices exist, household surveys
may collect data on the type and age of vehicle.
•• When the insuring of vehicles or other consumer durables is a frequent
practice in the country and the valuation practices used by the insurance
companies reflect the value of the asset, the insured value of the asset may
be requested in the household survey.

4.5.8. Non-agricultural enterprise assets

160. Non-agricultural enterprise assets should be valued by category of asset,
including, first, all machinery, equipment and furniture used in the production pro-
cess; second, inventories of inputs and supplies, including raw materials; and, third,
inventories of finished merchandise (goods for sale). The recommended valuation
method is to ask the respondent to estimate the potential sales value of each category
of enterprise asset, by group of assets. For the purpose of estimating individual wealth,
48

respondent owners should be asked what percentage of the enterprise they own and
the apportioned value of the assets should be assigned to them accordingly.

4.5.9. Valuables
161. Valuables are acquired and held as stores of value. They are expected to
appreciate or at least not to decline in real value or deteriorate over time. Although
markets exist for valuables, average individual respondents in household surveys
(unlike sale and other industry experts) may find it difficult to estimate the prices of
such valuables.
162. If countries wish to value valuables, a suggested approach is to ask the
respondent for the potential sales value of each category of valuables. In addition, for
the purpose of calculating individual wealth, respondent owners should be asked to
estimate their own share of the sales value from each group of valuables.

Key points
•• If countries’ aim is to obtain a full set of measures of the distribution of wealth (for details,
refer to chap. IV, sect. 3.1.4), including by gender and wealth quintiles or deciles, it is rec-
ommended that the valuation of all assets be attempted. If the aim is to obtain overall
measures of the gender wealth gap, countries may consider valuing only major assets.
•• Assets should be valued item by item at current market prices.
•• The potential sales value method of valuation allows for consistent measures of
wealth across assets but, if sales markets do not exist, alternative methods may be
used, including the following:
•• Countries with reliable price index data reflecting changes in asset prices over time
may consider collecting information on acquisition prices and the year when the
asset was acquired from the survey on asset ownership. This method would typi-
cally apply to real estate items, including the principal dwelling, agricultural land
and other real estate.
•• Countries with existing information on accumulated depreciation for assets such
as specific types of equipment, vehicles or other consumer durables may consider
using information on current market prices of an equivalent new asset less the
accumulated depreciation.
•• Countries may also consider imputing asset values based on statistical informa-
tion obtained from other sources of data, including administrative sources (such
as property records for tax purposes, land registration and cadastre systems), land-
use surveys and statistical data collection from local expert informants such as
community and price surveys. While community and price surveys do not typically
cover asset prices, such items could be considered for inclusion and enumerators
could be trained in obtaining information on unit prices (for example, for different
categories of livestock, agricultural land or agricultural equipment) based on com-
munity information interviews or the visiting of markets.
•• The method of valuation should be consistent across assets and countries should pro-
vide information about the valuation method used.
•• When one randomly selected household member is interviewed about his or her
asset ownership, this same respondent should provide the value of the assets that he
or she owns. When more than one household member is interviewed the valuation
of non-financial assets should be obtained at the household level from a knowledge-
able person capable of constructing a household roster of assets. The value of finan-
cial assets should, however, always be reported by their owners.
 Conceptual framework for measuring asset ownership from a gender perspective 49

5. Units of observation
163. This section discusses the different units of observation that can be used to
collect data in a survey on individual-level asset ownership and control, namely, the
individual and the asset, and the different measures that can be generated from each
unit. Each option is explained in detail below.
164. Household surveys in general have households and individual household
members as their basic units of enumeration, observation and analysis. Households
may consist of one or more persons, and they are defined on the basis of the house-
keeping concept. According to the Principles and Recommendations for Population and
Housing Censuses, Revision 3,80 a one‐person household is defined as “a person who 80 United Nations, Principles and
makes provision for his or her own food or other essentials for living without combin- Recommendations for Popula-
ing with any other person to form part of a multiperson household”. A multiperson tion and Housing Censuses,
Revision 3 (New York, 2014).
household is defined as “a group of two or more persons living together who make
common provision for food or other essentials for living. The persons in the group
may pool their resources and have a common budget; they may be related or unrelated
persons or a combination of persons both related and unrelated.” Definitions of house-
holds may vary and countries are encouraged to use their own definitions, already
established and in use by the statistical offices, for the purpose of collecting data on
asset ownership.81 81 A discussion on the pros
165. Although in practice most households are composed of a single family, the and cons of using different
definitions of households and
concept of “household” differs from that of “family”. A family is defined as those per- population when designing a
sons “who are related, to a specified degree, through blood, adoption or marriage”.82 sample is presented in chap-
A household may contain a combination of one or more families together with one ter III, section 4.
or more non‐related persons, or may consist entirely of non‐related persons. A fam- 82 United Nations, Principles and
ily, however, will typically not comprise more than one household. There are excep- Recommendations for Popula-
tion and Housing Censuses,
tions, including, for example, the case of polygamous families in some countries, or
rev. 3, 2014.
the shared child custody and support arrangements in others.
166. These Guidelines recommend that households—not families—be used as
one of the key units of enumeration. This is consistent with common practices in con-
ducting surveys and censuses in most countries and existing international standards,
including the Principles and Recommendations for Population and Housing Censuses,
Revision 3, the 2008 SNA and the OECD Guidelines.
167. Individuals, as units of enumeration and observation in data collection, may
be identified, in principle, within households (where the majority of the population live)
or within institutions. Typically, household surveys are designed to represent only the
population living in households (in other words, the non-institutional population). For
the purpose of these Guidelines, an individual is defined as a person residing within a
household. As with other surveys, a roster of household members is constructed, namely,
a listing of all individuals identified as belonging to a household, and for each of them a
series of basic characteristics such as age and gender are collected. Other characteristics,
such as those pertaining to education and employment, are collected only for a subset of
household members, typically defined by an age threshold. Information on asset own-
ership is collected only for adult persons, defined as individuals aged 18 or above. The
threshold of 18 years follows international standards defining a child, as set out in the
Convention on the Rights of the Child, and is often the minimum age at which individu-
als can enter into legally binding contracts to own property. While these Guidelines offer
guidance on collecting asset ownership data for persons aged 18 or above, countries may
consider extending the data collection to younger people, as warranted.
50

168. When planning a survey on measuring asset ownership at the individual

level, countries must decide whether to collect the data using individuals or assets as
the unit of observation, based on their main policy needs and the related statistics that
they wish to generate. Table 2 below presents the key measures that can be calculated
and who should be interviewed when the unit of observation is individuals or assets.

Table 2
Units of observation and key measures that can be calculated
a Details on key measures and
Unit of observation Key measures that can be calculateda
indicators for the gender
analysis of asset ownership are Individual: minimum set Prevalence gaps: compare the proportion of individuals who are owners of a particular
presented in chapter IV, of questions type of asset, by gender
section 2. Share of owners: indicates how many of the people who own a particular type of asset
are women and men
b Household wealth data Forms of ownership: provide information on whether respondents are exclusive or
obtained from a survey can joint owners
be used to cross-check and Asset: appending All of the above, plus:
cross-validate the statistics modules to an existing Share of agricultural land area owned by women: requires data on the size of each
on the balance sheet of the survey or conducting agricultural parcel owned by women and men plus the number of owners for assets
households sector in the stand-alone surveys owned jointly
national accounts, because when only one respond- Proportion of men and women who acquire assets through a specific mode: provides infor-
the latter are usually residual ent is selected per mation on potential channels for strengthening women’s ownership of assets
household Differences in the characteristics of assets owned by women and men
estimates. Many countries are Gender wealth gap: requires data on the value of each asset owned by women and
able to obtain frequent data men plus the number of owners for assets owned jointly
to compile the balance sheets
Asset: if more than Level, composition and distribution of household wealth: can also validate the statistics
of the total economy and the
one respondent per on the balance sheet of the households sector, which is derived residually in the
other institutional sectors such household is selected national accounts of the country b
as corporations and general for interview on asset Analysis of intrahousehold gender inequality in asset ownership
government in their national ownership, asset
accounts. The data used to roster is produced at the
compile the balance sheet of household level
the households sector tend,
however, to be available only
intermittently. As a result, the
statistics on the balance sheet 5.1. Individuals as the unit of observation
of the households sector in
countries’ national accounts 169. A simpler approach is to use the individual as the unit of observation, as
are derived residually by this allows for the measurement of asset ownership through a minimum set of ques-
subtracting the balance sheets tions that ask whether respondents, women and men, own a given type of asset. Coun-
of the other institutional tries may consider this approach when they want to obtain information on gender
sectors from the balance sheet
prevalence gaps in asset ownership. For example, it can be used to monitor part (a) of
of the total economy.
Sustainable Development Goal indicator 5.a.1 on the proportion of the total agricul-
tural population with ownership or secure rights over agricultural land, by gender.
This approach has been used in demographic and health surveys for the purpose of
measuring land ownership, and for some assets in the EDGE pilot studies, includ-
ing livestock, small agricultural equipment, consumer durables and valuables. As pre-
sented in chapter III of these Guidelines, the recommended minimum set of questions
for measuring the prevalence of women’s and men’s ownership of key assets also uses
the individual as the unit of observation.

5.2. Assets as the unit of observation

170. Using the asset as the unit of observation is a more complex approach that
requires inventories, or rosters, of assets to be created for each type of asset (such as
agricultural land and other real estate), but it allows for the computation of additional
measures of asset ownership, such as the gender wealth gap, that can provide impor-
 Conceptual framework for measuring asset ownership from a gender perspective 51

tant policy insights. In addition, many household surveys, such as the Living Stand-
ards Measurement Study surveys, already collect asset rosters, to which a module on
individual-level asset ownership and control could be appended. In these cases, only
slight modifications of the host survey instrument would be required to align it with
the recommendations in the present publication.
171. Two types of asset rosters can be obtained, each providing different infor-
mation, as presented in table 2. A respondent roster of assets lists each asset owned
(whether exclusively or jointly) by the respondent randomly selected for interview. The
respondent provides this information to the enumerator in the individual question-
naire.83 A household roster of assets lists each asset owned (whether exclusively or 83 See chapter III on question-
jointly) by all household members. One person, ideally the person most knowledgeable naire design for illustrations,
about household assets, provides this information to the enumerator in the household including discussion of the
questionnaire (see box 4 on an alternative approach implemented in the EDGE pilot household and individual
questionnaires.
studies).
172. Countries will need to determine which assets to include in the roster based
on their policy needs, but it is suggested that they include the priority assets (principal
dwelling, agricultural land, other real estate and financial assets), along with liabilities
for the purposes of calculating individual-level net wealth.84 Once the roster of assets is 84 Because there is only one
established, two sets of information are collected: first, key characteristics of the asset, principal dwelling, it does not
such as its value, size, location or use; and, second, the identity of the owner of owners need to be itemized in a roster
of the asset. If the respondent roster is used, all of this information is obtained in the of assets.
individual questionnaire for assets owned by the selected respondent. If the household
roster of assets is used, information on the value and other characteristics of the asset
is obtained when the assets are listed in the household questionnaire, but informa-
tion on the identity of the owners is reserved for the respondents to the individual
questionnaire. This design approach is based on both operational feasibility and the
rationale that reporting the characteristics of assets is less sensitive to variations in
respondents (owner or non-owner) than reporting ownership status.
173. A respondent roster of assets enables countries to generate statistics that
take into account differentials in the size and quality of assets owned by women and
men. For example, as discussed in chapter I, section 4, gender wealth gaps can be
derived by collecting information on the value of each asset owned by men and women
respondents and on the number of owners for assets that are jointly owned. With this
information, both the total wealth gap, and wealth gaps by composition or type of asset,
such as agricultural land, can be calculated. Establishing a roster of assets also ena-
bles countries to develop asset-level indicators on the joint or exclusive ownership of
assets that provides insights into asset ownership patterns other than individual-level
prevalence indicators. For example, policymakers might be interested in knowing the
proportion of agricultural land that is jointly owned by spouses. If multiple respondents
are interviewed, multiple rosters of assets need to be merged ex post—and for cases with
discrepancies, reconciled—to create one household asset roster that does not double-
count assets. This has proved to be resource-intensive in the EDGE pilot surveys (box 4).
174. A household roster of assets, established by a knowledgeable person in the
household questionnaire, is therefore recommended, as it eliminates the complication
of having to merge multiple individual rosters of assets and resolving any discrepan-
cies among them. In addition, with a household roster of assets, information may be
produced for two additional areas of analysis important for policymaking: household
wealth and intrahousehold gender inequality in asset ownership and control. As dis-
cussed in chapter I, section 4, micro data on the level, composition and distribution of
household wealth is of increasing interest to policymakers, as it can inform the design
52

85 OECD, OECD Guidelines, 2013. and evaluation of a wide range of economic and social policies.85 Such data can also be
used to validate statistics generated for the balance sheet for the household sector in a
country’s system of national accounts. Countries wishing to also estimate household
wealth through a survey on individual-level asset ownership can obtain information
on the value of each asset from respondents who complete the household roster of
assets in the household questionnaire and then proceed to interview individual house-
hold members about their ownership status for the assets listed in the household roster
along with any other assets that were not captured in the roster.

Box 4
Challenges of constructing household rosters of assets from multiple respondents
A respondent roster of assets was tested in five of the EDGE pilot studies. In Georgia, Mex-
ico, Mongolia, the Philippines and Uganda, rosters of agricultural parcels, large agricultural
equipment, non-agricultural enterprises, other real estate and financial assets and liabili-
ties were collected in the individual questionnaire from each household member who was
interviewed by asking him or her to list each asset (e.g., each agricultural parcel) owned by
each member of the household (whether exclusively or jointly)). This decision was based on
the hypothesis that a household roster of assets created by one respondent in the house-
hold questionnaire might be incomplete because of information asymmetries within the
household, including the presence of hidden assets—namely, assets that household mem-
bers owned but reportedly kept hidden from other household members.
To assess the prevalence of hidden assets, enumerators in Georgia, Mongolia, the Phil-
ippines and Uganda asked respondents whether anyone 18 years of age or older did not
know about the respondent’s ownership of the assets that they reported owning, includ-
ing agricultural parcels, agricultural equipment, non-farm enterprises, other real estate,
financial assets and liabilities. Across the pilot studies, the proportion of hidden assets was
negligible for all assets considered, with the exception of financial assets and liabilities. For
example, in Uganda, about 25 per cent of men who had borrowed money reported that
at least one other member of the household did not know about their liability, while the
comparable figure for women was 18 per cent. In Georgia, while the percentage of hidden
liabilities was negligible, the percentage of man and woman owners of financial assets
reporting hidden assets was about 12 and 13 per cent, respectively, with slightly higher
proportions for both genders in urban areas than in rural areas. While the low prevalence
of hidden assets in the pilot studies could be driven by respondents’ reluctance to reveal
them to the enumerators, qualitative findings from the Gender Asset Gap Project support
the results of the EDGE pilot studies, as they revealed that, while individuals were likely to
know about the physical assets owned by other household members, they were less likely
a Cheryl Doss and others, to know about the financial assets of other household members.a
“Measuring personal wealth Moreover, when multiple respondents provide independent asset rosters, the informa-
in developing countries: tion must be merged ex post—and for cases with discrepancies, reconciled—to create
interviewing men and
one household asset roster that does not double-count assets. Counting each asset only
women about asset values”,
2013.
once is essential for the estimation of household wealth and the construction of indicators
at the asset level. In all the EDGE pilot studies, this exercise proved to be resource-inten-
sive with little additional information gained, suggesting that having multiple household
members create independent household rosters is not a better design approach than hav-
ing one person provide a list of all assets belonging to the household’s members. The one
exception would be for financial assets. If countries wish to obtain a complete household
roster of financial assets, it should be generated by asking all adult household members
about the financial assets that they own.
 Conceptual framework for measuring asset ownership from a gender perspective 53

175. A household roster of assets also provides the scope for intrahousehold
gender analysis of asset ownership so long as couples, or multiple household members,
are interviewed and self-report their ownership status for the assets listed in the house-
hold roster of assets. As discussed in chapter I, section 2, understanding how assets
are distributed among couples or all household members can provide policymakers
with important insights, including how household members may respond differently
to policy and program interventions based on their asset endowments. Countries
wishing to collect data for intrahousehold analysis can obtain the household roster of
assets, as described above, in the household questionnaire from one respondent and
then proceed to interview more than one household member about their ownership
status for the assets listed in the household roster and any other assets that were not
captured in the roster. For guidance on how to select multiple household respondents,
see chapter III on sample design.

Key points
•• Data on asset ownership and control can be collected at the individual or at the asset
level, contributing to gender analysis from different angles.
•• Two options are presented for countries that want to use the asset as the unit of
observation:
•• Respondent roster of assets: self-reported by the randomly selected adult house-
hold member that lists all of the assets that she or he owns;
•• Household roster of assets: reported by one adult household member in the house-
hold questionnaire that lists all the assets owned by all the household’s members.
•• With regard to the type of gender analysis and indicators that can be produced:
•• Countries interested in information on gender prevalence gaps in asset ownership
can ask a minimum set of questions that use the individual as the unit of analysis.
•• Countries interested in developing asset-level indicators that account for differen-
tials in the size and quality of assets owned by women and men need to establish
a roster of assets, at least for all priority assets. The roster can be established at the
household or individual level.
•• Countries interested in estimating household wealth need to establish a house-
hold roster of assets that includes information on the value of each asset.
•• Countries interested in analysing intrahousehold gender inequality in asset own-
ership can establish a household roster of assets and interview multiple people in
the household, as described in chapter III of these Guidelines.
 55

Chapter II
Role of household surveys and other sources
of data in collecting individual-level data on
asset ownership and control

176. Individual-level data on asset ownership and control can be collected prin-
cipally through household surveys, agricultural censuses, and surveys and adminis-
trative sources. In some countries, population and housing censuses may also play a
role; these are conducted only once every 10 years, however, and the magnitude of the
operations to conduct them limits the possibility of adding new topics and covering
these in any detail.
177. National statistical offices should consider all relevant sources in a com-
plementary manner and decide, within the context of the overall statistical plan, each
source’s role in collecting individual-level data on asset ownership and generating sta-
tistics relevant from a gender perspective. For this purpose, the sections below briefly
describe the type and detail of information each data source can provide in respect of,
first, the range of assets and types and forms of ownership that can be measured; sec-
ond, the conceptual framework used to assess ownership and control; third, the units
of observation and analysis; and, ultimately, the types of statistics and indicators that
can be generated.

1. Role of household surveys

178. Household surveys are a major source of social, demographic and eco-
nomic statistics in both developed and developing countries. A survey is defined as
“an investigation about the characteristics of a given population by means of collecting
data from a sample of that population and estimating their characteristics through the
systematic use of statistical methodology”,86 household surveys are specialized surveys 86 United Nations, Statistical
on a sample of households. Compared to other sources of data on asset ownership, Commission and Economic
household surveys are advantageous because they can cover a wide range of topics and Commission for Europe, Termi-
conceptual frameworks and generate a complete set of measures of asset ownership nology on Statistical Metadata
(Geneva, 2000), available at
from a gender perspective.
www.unece.org/fileadmin/
DAM/stats/publications/53met
1.1. Scope adaterminology.pdf.

179. Household surveys are the only source of data that can explore the full
range of physical and financial assets, including those recommended for data collec-
tion in the present Guidelines: dwellings, agricultural land, other real estate, livestock,
agricultural equipment, bank accounts or other financial assets, valuables and con-
sumer durables. The information obtained on women’s and men’s ownership of these
assets can be linked to information obtained on other topics covered by the same sur-
vey, such as education, health, employment, income or living arrangements. Integrat-
ing these relevant dimensions into data collection in the same household survey will
56

provide the most complete understanding of asset ownership patterns across different
groups of the population and the link between asset ownership and key development
outcomes for the household.

1.2. Conceptual framework

180. Household surveys can easily adopt the conceptual framework presented
in chapter I of these Guidelines for measuring asset ownership and control from a
gender perspective. In particular, household surveys can accommodate the set of ques-
tions for measuring the bundle of ownership rights, including reported ownership,
documented ownership and the rights to sell and bequeath an asset.
181. In addition, household surveys can relatively easily implement the respond-
ent selection protocols presented in the Guidelines. As discussed in chapter I, national
statistical agencies are recommended to collect self-reported information on the own-
ership and control of assets from household surveys by interviewing one or more
randomly selected adult household members or all household members. Among the
various types of surveys within a country’s household survey programme, some may
already select individuals on a random basis for interviewing or interview all house-
hold members, while others may be able to initiate the respondent selection protocols
87 The implications for the after data for the main survey have been collected.87 Other potential sources of data
organization of fieldwork are on the ownership and control of assets do not have the same flexibility in adjusting
discussed in chapter III. respondent selection protocols. For instance, agricultural censuses and surveys focus
on agricultural holdings and therefore have an interest in collecting information from
the person most knowledgeable about the agricultural holding, not from a randomly
selected household member who may or may not know about the agricultural hold-
ing. Censuses can cover topics of interest only briefly and have to rely heavily on proxy
responses in order to avoid inflating the length and cost of the census.

1.3. Units of observation and measures of ownership

182. Household surveys have households or individuals as their basic units of
enumeration, observation and analysis, and the data requirements for producing prev-
alence indicators can be as simple as asking the sampled persons whether they own
any of the different types of assets of interest. Population-based prevalence indicators
on asset ownership and control provide a basic picture of how many women and men
own dwellings, land, livestock or other physical and financial assets. Gender differ-
ences in the prevalence of ownership by type of asset, for the entire population of a
country or disaggregated by multiple relevant population groups, can only be assessed,
in most countries, on the basis of household surveys. This is essential information for
policymaking that most countries currently do not collect, but that national statistical
offices can produce in order to assess the extent of gender discrepancies in asset own-
ership. This information may also signal the need for additional data to understand
the factors causing these discrepancies, including data that may be provided by other
sources, such as agricultural surveys or administrative data.
183. Depending on the overall objectives of the data collection and the overall
tabulation and analysis plan envisioned by the national statistical office, data require-
ments may be more complex. Each respondent may be required to list assets owned
individually or jointly with somebody else. Household surveys are flexible enough to
allow the collection of information on an inventory of assets and their characteristics
item by item, as discussed in chapter I of these Guidelines. In this case, the different
types of assets listed become additional units of observation and analysis. Using assets
 Role of household surveys and other sources of data in collecting individual-level data on asset own 57

as units of analysis in addition to the individual allows for a much broader range of
indicators and analyses to be developed from the data, including measures of wealth
distribution by gender (as described in chap. I), in addition to the measures of owner-
ship prevalence.

1.4. Limitations
184. It should be noted, however, that data collection through household sur-
veys has implications in terms of cost, data quality, sampling errors and the ability to
provide data for small areas or population groups. These challenges and limitations are
typical of all household surveys.
185. Some non-observation errors may have a specific impact on the estimates
of asset ownership. For example, the richest and poorest households may be excluded
from some household surveys, either by design or because they are more likely to
refuse to respond to the surveys.88 This omission can have an impact on the estimated 88 United Nations, Handbook
wealth distribution across the population, and, to a lesser degree, the estimated prev- of Household Surveys, revised
alence of ownership. This aspect should be taken into account in survey operation edition, Studies in Methods,
activities, including the sample design, the training of enumerators, data processing Series F, No. 31, New York, 1984;
European Commission and oth-
and weighting.89
ers, 2008 SNA, 2009.
89 Sample design and field
operations are discussed in
2. Population and housing censuses chapter III. Data processing is
discussed in chapter IV.
186. “A population census is the total process of planning, collecting, compil- 90 United Nations, Principles and
ing, evaluating, disseminating and analysing demographic, economic and social data Recommendations for Popula-
at the smallest geographic level pertaining, at a specified time, to all persons in a coun- tion and Housing Censuses,
try or in a well-delimited part of a country.”90 They are conducted every 10 years, on rev. 3, 2014.
the principle of complete enumeration, and based on large-scale operations.
187. Population censuses and household surveys cover, in principle, the same
population and employ the same units of enumeration, that is, households and indi-
viduals. For their part, however, censuses are less well-equipped to collect complex or
detailed information on specific topics that would require intensive training, more
specialized interviewers and a higher burden for the field staff. Furthermore, the
census interview relies heavily on proxy respondents. The requirement to collect self-
reported data from one or more randomly selected adult household members or from
all household members (as is needed for measuring asset ownership at the individual
level) would considerably increase the burden, length and cost of the census.
188. Nevertheless, many countries have designed population censuses to com-
bine, first, a full field enumeration, based on a short-form questionnaire, and, second,
a large sample attached to the census, where a long-form questionnaire can be used
to cover a range of issues in greater depth. Collecting information during the census
on additional topics from a sample of households is a cost-effective way of broadening
the scope of the census to meet the expanding demands for statistics. Countries may
explore this sample-based modality of data collection to obtain individual-level data
on the ownership and control of a core set of assets. For example, adding questions on
whether women, men or both own selected assets listed in the housing questionnaire
would enable the calculation of some basic asset-based measures of wealth distribution
by gender; while adding questions on the ownership of selected assets for adults listed
in the household roster would enable the calculation of population-based prevalence
measures of ownership. Such attempts, however, would first need to assess carefully
the possibility of using only self-reported information on ownership.
58

3. Agricultural censuses and surveys

189. As defined by FAO, a “census of agriculture is a statistical operation for
collecting, processing and disseminating data on the structure of agriculture, cover-
ing the whole or a significant part of the country. Typical structural data collected in
a census of agriculture are size of holding, land tenure, land use, crop area, irrigation,
livestock numbers, labour and other agricultural inputs.” Data are normally collected
on the basis of a complete enumeration of all agricultural holdings (i.e., the agricul-
91 FAO, WCA 2020, vol. 1, 2017. tural production unit), every 10 years.91
190. Agricultural surveys collect structural and production data from a sample
of agricultural holdings. They are based on the same units of analysis as those used
in agricultural censuses. With much smaller workloads and the opportunity to train
fewer personnel more intensively, however, agricultural surveys can examine topics in
much greater detail. They are conducted more frequently than agricultural censuses
and can therefore provide more timely data.
191. The key concepts in agricultural censuses and surveys are the agricultural
holding and the agricultural holder. The agricultural holding is defined as “an economic
unit of agricultural production under single management comprising all livestock kept
and all land used wholly or partly for agricultural production purposes, without regard
to title, legal form or size” and the agricultural holder as “the civil person, group of civil
persons or juridical persons who makes the major decisions regarding resource use and
92 Ibid. exercises management control over the agricultural holding operation.”92
192. The statistical unit for the agricultural census and surveys is the agricul-
tural holding. Overall, data are collected at the agricultural holding level, but second-
ary units of observation and analysis are used. For example, some of the information
on agricultural land may be collected at the parcel level, including, for example, area
of land, land use, land tenure and terms of renting.

3.1. Scope of agricultural surveys

193. Because of their focus on agricultural holdings, agricultural surveys can
be extremely efficient in obtaining detailed information on a series of aspects related
to agricultural production and to the structure of the agricultural sector. Agricultural
surveys are more likely than agricultural censuses to collect data on ownership of
agricultural land and equipment. They may be an appropriate vehicle for obtaining
individual-level data on ownership of assets used in agricultural production, including
agricultural land, livestock and agricultural equipment, and also on their character-
istics. In addition, the inclusion of questions on the owners of these assets is relatively
straightforward, since the surveys normally already collect holding-level information
on ownership of agricultural land, livestock and agricultural equipment.
194. In addition, agricultural surveys can be an optimal vehicle for monitoring
Sustainable Development Goals indicator 5.a.1, since the reference population for this
indicator consists of the adult individuals living in agricultural households and there
is normally a one-to-one relationship between the holdings in the household sector
and the agricultural households.

3.2. Limitations of agricultural surveys

195. Ownership indicators constructed on the basis of data collected in agri-
cultural surveys may have limitations related to the coverage of the data collection
method. As mentioned above, agricultural surveys use the “holding” as their unit of
analysis and cannot therefore be used to generate estimates on the asset ownership of
 Role of household surveys and other sources of data in collecting individual-level data on asset own 59

women and men that are representative for the entire population: in particular, indi-
viduals residing in urban areas may not be captured.93 93 Ibid.

196. In addition, agricultural surveys may be less flexible than household sur-
veys when it comes to collecting individual data or applying the respondent selec-
tion protocols presented in these Guidelines. The information that individuals would
provide about the ownership status of other persons in the households may be biased.
Furthermore, the ownership of women holding small-size areas of land may not be
captured either because, in some countries, agricultural surveys have a minimum size
limit for the holdings covered by the data collection or because they are restricted to
holdings conducting commercial agricultural activities.94 94 Ibid.

3.3. Scope of agricultural censuses

197. The emphasis in agricultural censuses is on the management of the agri-
cultural holdings, and not the ownership of agricultural assets. The identification of
the agricultural holder provides the basis for comparing the characteristics of holdings
operated by women and men, which is important for understanding issues of decision-
making and agricultural productivity. Agricultural censuses rarely identify the own-
ers of agricultural assets, although land and livestock ownership within the holding is
proposed as a new topic for data collection, together with intrahousehold distribution
of decision-making, in the WCA 2020 operational guidelines.

3.4. Limitations of agricultural censuses

198. Like the population and housing censuses, agricultural censuses are com-
plex and large-scale operations. For that reason, they are not flexible when it comes to
collecting individual data or applying the respondent selection protocols presented in
these Guidelines. Indeed, WCA 2020 recommends that the respondent be someone
sufficiently knowledgeable to give accurate answers to questions on the agricultural
holding and its components—in other words, ideally the holder or the manager.95 95 Ibid.

199. For the agricultural census, the size limitation for holdings covered is justi-
fied on the grounds that there are usually a large number of very small units making
little contribution to total agricultural production and it is not cost-effective to include
them in the agricultural census. Nevertheless, an alternative to setting minimum size
limits is to cover all units regardless of size, but to ask only some very limited questions
for small units.96 The questions put to small units should, however, include questions 96 Ibid.
on asset ownership and control from a gender perspective.

4. Administrative sources of data

200. Administrative sources are defined as “data holdings containing infor-
mation that is not primarily collected for statistical purposes”.97 Typically, they are 97 United Nations, Economic
developed and maintained by government administrative authorities for the purpose Commission for Europe, Using
of implementing government services and regulations.98 Administrative sources have Administrative and Secondary
several key advantages in comparison to other sources of data: first, their running cost Sources for Official Statistics:
A Handbook of Principles and
is low once they have been set up; second, when complete, they can provide accurate
Practices (Geneva, 2011).
and detailed information at the level of small geographic areas; third, they can gener- 98 Ibid.
ate statistics at frequent and regular intervals; and, fourth, they can eliminate survey
errors and non-response.
60

4.1. Scope
201. Statistical information on asset ownership may be derived from some
administrative sources such as land registration and cadastre systems (or some sort
of land information system), dwelling property records, property taxation records
and vehicle registration records. These sources may provide information on registered
assets (such as a description of the land parcels or dwellings and their value) and some
characteristics of their owners (such as their name and national identification). These
sources of information are typically developed by formal institutions and can provide
statistics mainly on documented ownership. For example, in a country where access
to land is governed by a mix of formal and customary institutions, information on
formal legal rights to land is probably recorded in some form of land registration and
99 FAO, Land tenure and rural cadastre system.99 While that information may be retrieved for the purpose of generat-
development, FAO Land Tenure ing statistics, there are no corresponding systematic and consolidated records reflect-
Studies 3 (Rome, 2002). ing customary tenure.
202. Several countries have developed farm registers (listings of farms or agri-
cultural holdings) and some have attempted to create statistical farm registers, includ-
ing for selecting samples for agricultural surveys and for generating statistics, among
other purposes. Farm registers may be developed and updated based on agricultural
censuses and surveys or on administrative records (such as tax records, cadastral
records, directories from farmers’ associations). Statistical farm registers are in use in
100 Stephen Clarke, “Improv- many European countries.100 At the same time, however, the development of statistical
ing the quality of EU farm farm registers with regularly updated and maintained records of holdings and hold-
registers”, paper presented at ers that would generate statistics in agriculture remains a difficult enterprise. Statisti-
the Seminar on Registers in cal farm registers usually contain information about the name of the holder and the
Statistics, Helsinki, 21–23 May
address of the holding, gender of holder, total area of holding, main land uses and
2007.
types of animals kept. Owners of specific agricultural assets are usually not recorded.

4.2. Conceptual framework

203. Information available in administrative sources is not regarded as a pri-
mary source for statistical purposes. Therefore, the recording of information about
assets covered and their owners does not follow a predefined conceptual framework.
It should be noted, however, that administrative sources such as property and taxation
records typically cover assets that are registered and documented, thus corresponding
to one of the types of ownership presented in these Guidelines.

4.3. Units of observation and measures of ownership

204. The unit of record in administrative sources is typically the asset. In the-
ory, having the asset as a unit of record enables the calculation of both population-
based indicators and asset-based indicators of gender differences in documented asset
ownership. In practice, the calculation of these two types of indicators depends on the
accurate and complete recording of all owners of an asset and their gender, and also
on the ability to determine whether one person owns multiple plots, as the plots may
be registered under different names and thus viewed as owned by two separate people.

4.4. Limitations
205. Administrative sources can only be useful if they are kept current, and if
the gender of the owner or holder is recorded. This is not the case for many adminis-
trative sources. Besides differences in conceptual frameworks, one of the key limita-
 Role of household surveys and other sources of data in collecting individual-level data on asset own 61

tions in using administrative sources for statistical indicators and analysis of asset
ownership is the incomplete coverage of assets and incomplete information on all
owners of an asset, including their gender or other demographic characteristics. For
instance, land registry records may not systematically incorporate information that
can be used to establish whether the landowners are women or men. A review101 of 101 FAO and World Bank, Gender
land registry databases in five countries in the Western Balkans102 showed that the Disaggregated Data—Western
gender of the owner is not typically recorded as a stand-alone variable and often can- Balkans. Statistical Reports
not be deduced from other information that may be specified in the records, such as 2005–2013 (2014).
102 Albania, Bosnia and Herze-
the identity details or the first name of the owners.
govina, Montenegro, North
Macedonia and Serbia.
 63

Chapter III
Guidance for implementation

206. Careful planning and execution are critical to the success of any survey or
survey module on measuring asset ownership and control at the individual level. Gen-
eral principles and rules for all statistical sample surveys are applicable to surveys on
asset ownership but, in addition, specific considerations should be taken into account
in order to ensure the quality and reliability of results on individual-level asset owner-
ship. Topics addressed in this chapter include the planning process, data collection
strategies, sample design, questionnaire design and field operations.

1. Planning data collection on asset

ownership at the individual level
207. While some countries collect data on assets or durable goods owned by, or
used in, the household, very few collect individual-level data on asset ownership and
control. Countries adopting this new gender-relevant approach will need to decide
which mechanisms of data collection are at their disposal to gather such data. As dis-
cussed in chapter II, household surveys are the preferred data source for regularly
estimating the prevalence of asset ownership and other key measures within the popu-
lations of women and men, but other potential sources include agricultural surveys
and censuses and administrative records. All of these sources should be assessed to
determine their capacity for delivering the data needed, the costs involved and the
technical expertise required.
208. The desired frequency of producing statistics on asset ownership and of
ensuring comparability with future statistics through the use of the same methodol-
ogy of data collection over time should also be assessed at this early stage and should
inform the decision on which sources of data to use. Asset-ownership prevalence is
expected to be fairly stable over short periods of time, unless there are economic crises
that deeply affect the wealth of a large portion of the population or new government
programmes that target assets. In the absence of such events, monitoring the preva-
lence of ownership once every five to seven years should be sufficient to construct reli-
able trends.
209. Nevertheless, similar to other data collections, the quality of data collected
will depend in part on the quality of the planning process, starting with specifying
clear objectives for data collection, bringing together the right people and developing
realistic budgets and timelines. These aspects are covered in the following sections.

1.1. Specifying data collection objectives

210. A clear statement of the data collection objectives should be developed in
consultation with stakeholders, including data users. The objectives should indicate
topics and policy issues that need to be addressed; the statistical information on which
they are based; the geographical and population coverage of the results; and the man-
64

ner in which the results will be used. Data collection objectives give a rough idea of
the expected scale of the survey and are a crucial input in deciding the sample size and
structure, the amount and complexity of information to be collected, and the required
resources of time, human skill and funding.
211. Data collection objectives may initially be formulated as key questions to
which the collection exercise is seeking answers. They may vary from very simple ones
that provide a basic picture of asset ownership to more complex ones, as in the follow-
ing examples:
•• What is the prevalence of asset ownership among women and men, by
type of asset? A simple description of asset ownership can be developed by
using data collected through a minimum set of questions on whether the
respondent owns specific assets or not.
•• Are women more likely to own assets exclusively or jointly? Are men? Are
women owners as likely as men owners to possess the full bundle of owner-
ship rights, including the right to sell and bequeath assets? Do women and
men acquire assets in different ways? To address these policy issues, a few
more questions can be added to a short module appended to a household
survey.
•• What is the monetary value of assets owned by women and men? Is wom-
en’s wealth concentrated in the same types of assets as men’s wealth? On
average, do men possess more net wealth than women? Addressing these
types of questions requires that a respondent roster of assets is created and
information obtained on the form of ownership and the value of each asset.
•• Are assets equally distributed among adult women and men living in the
same household? Is the value of assets owned by women who are married
or in partnerships similar to the value of assets owned by their husbands or
partners? Data requirements to answer these questions are more complex
and require that a household roster of assets be created and information
be obtained about each asset listed. In addition, couples and possibly other
adult members of the household, depending on the focus of analysis, will
have to be interviewed about their own asset ownership.
•• Do women who own assets, whether exclusively or jointly, have more deci-
sion-making power than women who do not own assets? Are they more
likely to be entrepreneurs or to have their own income? Do they invest
more often in the education of their children? Are they less likely to be vic-
tims of domestic violence? Answering these policy questions requires the
inclusion in the survey of additional questions that need to be analysed in
relation to the questions on asset ownership.
•• Have government programmes on housing subsidies or land allocation
had different impacts on women’s and men’s ownership of these assets?
Addressing questions of programmatic impact may require including in
the questionnaire items specifically referring to the programme of inter-
est or fielding separate survey waves before and after the programme is
implemented.
212. Several objectives may be accommodated in the same data collection if they
are consistent with one another and their number and complexity do not compromise
the quality of the overall data collection. For example, Statistics South Africa included a
module on decision-making in the EDGE pilot survey in order to analyse the relation-
ship between women’s and men’s asset ownership and household decision-making.
 Guidance for implementation 65

213. Nevertheless, covering too many objectives may prove challenging for
sample and questionnaire designs and may cause budget overruns. As such, care must
be taken not to overload the data collection with too many competing goals, and a
clear statement of the objectives will help to keep the project focused throughout all
stages of development, implementation, data analysis and dissemination.
214. Similar to other data collections, once the objectives have been defined,
they should be ranked by their importance and feasibility, including through the use
of a tabulation and data analysis plan. A tabulation and data analysis plan explains in
detail what data are needed to attain the objectives (namely, to answer the questions)
set out for the data collection and what indicators can be derived from the data col-
lected. It also ensures that no unnecessary questions are included and no essential
analyses are omitted, therefore maintaining compatibility between the data require-
ments and the final survey design. Survey designers must refer to this plan constantly
when working out the details of the survey questionnaire.
215. The final set of objectives to be covered in the data collection should be cho-
sen on the basis of capacity for data collection, amount and quality of data expected
from other sources and the funding available. If countries choose to append a module
rather than implementing a stand-alone survey, as discussed in chapter III, section 2,
it is important that the module itself be designed on the basis of a clear set of objectives
that can be accommodated by the host survey.

1.2. Building the project team

216. The planning of a data collection exercise is usually carried out by a rela-
tively small group of subject-matter specialists and technical and administrative staff
members of the central statistical office, in close collaboration with key stakeholders.
A small team of key stakeholders should be involved from the early planning stages of
the data collection, including when formulating the scope and objectives. Consulta-
tions should also be undertaken at other specific stages. By including this group, the
communication between the data users and data producers will be greatly increased.
Engaging stakeholders throughout all phases of the project also ensures that there will
be knowledgeable experts who are prepared to use the basic statistics obtained and the
results of deeper analysis for policymaking.
217. Key stakeholders may include researchers, women’s and gender advocates,
policy analysts and policymakers, and donors. Data and analysis on asset ownership
is generally relevant for the development and monitoring of government policies and
programmes related to poverty, livelihoods and entrepreneurship, agriculture, wom-
en’s empowerment and gender equality, and housing and distribution or titling of land.
Analysts involved in these aspects of policymaking can provide technical expertise
and contribute to specifying the objectives of the survey. When individual-level data
on asset ownership is collected through a module attached to a survey, it is important
that the group of stakeholders for the overall host survey includes people representing
those institutions.
218. Staff requirements for the various aspects of the survey are a crucial plan-
ning consideration in any survey, and their assignment to the project must be decided
at an early point. A team of experts, subject-matter specialists and data analysts must
be formed at the very beginning of survey planning to ensure that no aspect of the
survey is neglected and that it is given priority in the national statistical office’s survey
programme. This team should include senior staff of the national offices, including
specialists in gender statistics, household surveys, sampling, national accounts and
agriculture, along with data managers and other specialists in field operations. The
66

services of additional specialists may be enlisted from outside the statistical agency,
including the stakeholders referenced above.
219. This team of experts has a crucial role to play, especially when it is the first
time that the national statistical office is collecting data on asset ownership from a
gender perspective. New concepts and definitions will need to be applied, reflected in
the questionnaire design in a way that makes sense to the country context, and com-
municated in an effective way to the enumerators and supervisors during the training
of field staff. The sample design needs to account for potential variations of the tenure
system across the country and may involve new respondent selection protocols. The
individual-level perspective used in data collection makes it possible to analyse asset
ownership in new ways that could lead to better-articulated policies; at the same time,
however, the data structure and data analysis have a certain degree of complexity and
analytical reports that inform policymaking will need to be carefully drafted.
220. Lastly, the group of data collectors has an indispensable role to play in
ensuring the quality of data. This group includes interviewers, supervisors, data entry
staff and computer technicians. They may be part of the staff of the central agency or of
regional offices. Additional interviewers and supervisors may be hired from the field.
Some representatives of this group may be involved in some aspects of the question-
naire design, including the optimal formulation of some questions or instructions to
ensure that they are clearly understood by both enumerators and respondents.

1.3. Budget and timeline

221. One of the first tasks in planning a survey is to draw up a draft budget that
approximates the cost of the survey based on some assumptions of the sample size and
the average time needed to interview one or more household members. This exercise is
normally accomplished by looking at budgets of similar surveys already implemented
in the country or in similar countries.
222. Typically, there are two types of survey costs: fixed and variable. Examples
of fixed costs are costs associated with the developing and testing of the question-
naire and other survey instruments. Variable costs refer to expenses that are highly
dependent on the sample size and structure, including those related to the employ-
ment of the field staff and their transportation and accommodation in the field, or, for
example, the number of electronic devices needed when computer-assisted interview-
ing is used. The variable costs, mainly driven by the sample size and structure, will
usually dominate the survey budget. As extensively discussed in chapter III, section 4,
the sample size and structure will depend on the level of precision required for the
key estimates of asset ownership and wealth, the number and level of population sub-
groups for which estimates need to be produced, and the prevalence of asset ownership
in the targeted population or population subgroups. Another important factor is the
use of sampling techniques such as stratification and clustering, which are cost-effec-
tive ways of reducing costs associated with field-staff travel without compromising the
possibility of obtaining estimates representative at the level of the population groups
desired and with the required precision.
223. More reliable and detailed cost estimates can be developed once the overall
scale of the survey and a detailed timetable of activities are in place. The timetable
of activities should be comprehensive and include details on the time frame of each
activity, keeping in mind existing constraints, such as other surveys being developed
at the same time or the preferred time of the year when the fieldwork should take place.
There should also be a clear specification of which activities are to be done in-house—
 Guidance for implementation 67

performed by the regular staff of the statistics office—and which are to be outsourced
to other individuals or institutions.
224. Financial resources can be a major constraint, limiting how many house-
holds can be surveyed, how many interviewers can be employed and how much time
they can spend within any given enumeration area. Some elements of the survey may
need to be adjusted depending on the available budget, including sample size and
structure and questionnaire length and complexity. The quality of the data, however,
needs to be preserved. Survey errors need to remain at an acceptable minimum for the
specified survey objective; the data collection instruments need to be properly devel-
oped or customized; and the staff involved in data collection need to be adequately
skilled and properly trained.

2. Data collection strategies

225. Countries choosing to measure asset ownership and control at the individ-
ual level through household surveys have three options for collecting the data. These
vary both in their complexity and in the range of measures that can be produced from
the data items. First, and simplest, they can append a minimum set of questions using
the individual as the unit of observation to an existing household survey. Second, they
can append a more detailed module using the asset as the unit of observation to an
existing household survey. Third, and most complex, they can implement a stand-
alone, or dedicated, survey comprising a household roster of assets and individual
questionnaires on asset ownership.
226. Each option, including the advantages and limitations of the data collec-
tion strategy, is discussed in detail below.

2.1. Appending a minimum set of questions

to an existing survey questionnaire
227. Countries may choose to append a minimum set of questions on asset
ownership to an existing nationally representative household survey, in order to meas-
ure the prevalence of asset ownership by gender. For each asset, five questions in total,
using the individual as the unit of observation, can be appended to an existing ques-
tionnaire to measure the full bundle of ownership rights, as illustrated in chapter III,
section 5, on questionnaire design. It is advised that countries adopting this approach
limit data collection to priority assets including principal dwellings, agricultural land,
other real estate and financial assets. Countries may also choose to collect individual-
level data for non-priority assets covered in the Guidelines, depending on their policy
needs, by including one question each to measure reported ownership of agricultural
equipment, livestock, consumer durables and valuables. The collection of information
about these additional assets will, however, require the inclusion of additional ques-
tions, since they will have to be disaggregated by type within their asset type to yield
meaningful information. For example, asking the question “Do you own any con-
sumer durables?” will not yield information on the types of durables that the respond-
ent owns and is thus less useful for analysis. Suggested questions to collect information
on ownership of different assets are provided in chapter III, section 5.
228. There are several advantages to appending a minimum set of questions to
an existing survey questionnaire. First, the data collection strategy enables countries
to measure the prevalence of ownership for priority assets in the population, by gen-
der, with minimal increases in data collection or response burden. For example, coun-
tries can adopt this approach for measuring Sustainable Development Goal indicator
68

5.a.1 on the proportion of the agricultural population with ownership or secure rights
to agricultural land, by gender, so long as the sample design treats the agricultural
population as a subgroup of interest. Second, appending a minimum set of questions
to an existing household survey may be more sustainable than appending a detailed
module to a host survey or conducting a stand-alone survey, as the latter approaches
are often more susceptible to budgetary cuts and competing priorities within a survey
programme. One disadvantage of this approach, however, is that it only allows for the
calculation of prevalence estimates of women’s and men’s ownership of priority assets.
If countries wish to collect information on additional assets or to derive additional
indicators, such as on the gender wealth gap or modes of asset acquisition, they should
consider appending a survey module to an existing household survey.

2.2. Appending a survey module to an existing household survey

229. Alternatively, countries may opt to append a survey module on asset own-
ership and control, using the asset as the unit of observation, to an existing nationally
representative household survey. Depending on the level of precision needed at lower
administrative units, the module may be appended to the entirety of the sample for the
main survey or to a subsample of the main survey, so long as national representation
is retained. Appending a module to an existing survey is often less resource-inten-
sive and costly than conducting a stand-alone, dedicated survey on asset ownership,
because the bulk of the costs are borne by the main survey; as discussed in chapter III,
section 4, on sample design, the marginal costs of the survey module are a function
of the total number of interviews on asset ownership. Appending a module on asset
ownership to a household survey, in particular a multi-topic survey, will also provide
a rich source of data for the analysis of relationships between asset ownership and key
variables of interest to policymakers.
230. One disadvantage of this approach, however, is that it affords less flexibil-
ity than a stand-alone survey, because the data collection is subject to the parameters
of the main survey. For example, the sample size is determined by the objectives of
the main survey, not by the module on asset ownership, which has implications for
within-household respondent selection, as detailed in chapter III, section 4. In addi-
tion, the content of the module on asset ownership, including the number of both the
assets that can be covered and the questions asked, will be determined in part by the
length of the main survey questionnaire and the need to minimize response burden.
As a result, this approach is more suitable to collecting data on the priority set of assets
rather than the full range of assets covered in the publication.
231. Countries planning to append a module on asset ownership and control to
an existing household survey should consider which surveys in their existing survey
programme would be a good fit for hosting the module. There are several factors to
consider. First, the topic or topics of the main survey should complement the mod-
ule’s focus on asset ownership and control, so as to ensure the continuity of the inter-
view. Integrated or multipurpose household surveys, which collect data on multiple
topics relevant to policy analysis in one survey, such as the Living Standards Meas-
urement Study survey, are a natural fit. Also appropriate are household income and
expenditure surveys, which collect data on the flow of monetary and non-monetary
resources of households and individuals and may already include a module on asset
ownership at the household level. Because these surveys tend to require repeat visits
by the enumerators to the households to collect the data, respondents may be more
comfortable answering sensitive questions, such as on asset valuation, than in surveys
requiring enumerators to visit the household only once. For example, in the qualita-
 Guidance for implementation 69

tive assessment of the EDGE pilot study in Mexico, it was observed that the module on
asset ownership should be appended to a household survey that allows enumerators
to develop a rapport with respondents through repeat visits to the enumeration area.
232. Second, the host survey must be able to accommodate the respondent selec-
tion protocols necessary for collecting individual-level data on asset ownership and con-
trol. Furthermore, as previously discussed, information on asset ownership should be
self-reported, not reported by proxy (see chap. I, sect. 2, on respondent selection rules).
233. As detailed in chapter III, section 4, on sample design, for the purposes of
deriving nationally representative estimates of asset ownership and wealth, one ran-
domly selected adult respondent should be interviewed. In this case, unless the host sur-
vey randomly selects adult respondents for interview, field protocols must allow for the
random selection of household members to complete the module on asset ownership.
234. If countries aim to undertake an intrahousehold analysis of asset owner-
ship via an appended module, more than one respondent per household must be inter-
viewed, and the selection of respondents will depend on the sample size of the host
survey. When more than one respondent per household is interviewed, it is advised
that a roster of assets be collected at the household level, as discussed in chapter I.
Thus, countries adopting this approach must ensure that, in addition to accommodat-
ing the respondent selection protocols, the main survey is able to integrate a household
roster of assets into its household questionnaire.
235. Third, countries should assess how often they will need to collect data on
individual-level asset ownership and select a host survey that can accommodate this fre-
quency. As discussed in chapter III, section 1, data on asset ownership can be collected
every five to seven years unless a country has a policy need, such as assessing public
interventions, for more frequent monitoring. Thus, the main survey to which the mod-
ule on asset ownership will be appended should be administered with similar frequency.

2.3. Conducting a stand-alone survey

236. A stand-alone survey on individual-level asset ownership and control
includes, at a minimum, two parts: the first consists of a household questionnaire com-
prising, first, a household roster that lists and collects basic sociodemographic informa-
tion about all household members; and, second, a household roster of assets that lists
all the assets owned by the adult members of the household and collects information
on the characteristics of the assets, including their value. The second part consists of an
individual questionnaire on asset ownership. It is advised that the household question-
naire be completed by a person knowledgeable about the household’s asset holdings.
The roster of assets is then fed forward to the individual-level interviews, which are
administered to all adult household members as described in chapter III, section 4, on
sample design. In the individual questionnaires, respondents self-report their owner-
ship status for the assets listed in the household roster and also for any additional assets
owned by them that were not captured in the household roster of assets.
237. A dedicated survey on asset ownership and control enables countries to
collect a comprehensive set of data from all adult household members for the pur-
pose of informing policies and programmes aimed at promoting gender equality
both within and across households. In comparison to a minimum set of questions
or a survey module, a stand-alone survey can collect information on a larger range
of assets and their characteristics, including, for example, the tenure type and size
of agricultural parcels. It can also measure all components of the conceptual frame-
work presented in chapter I, including types and forms of asset ownership, modes of
70

asset acquisition and asset valuation. Notably, because a household roster of assets and
their values are collected and all adult household members are interviewed about their
ownership status (see figure 4), a stand-alone survey also enables analysis of how total
household wealth is distributed among all household members.
238. A dedicated survey on asset ownership also has the flexibility of including
modules on additional topics, such as education, health or decision-making, so that
data users can analyse the relationship between asset ownership and key development
outcomes of interest to policymakers. Moreover, because the data collection is not sub-
ject to the sample design and fieldwork organization of a host survey, more control can
be exercised over these components of the survey process to ensure data quality. Lastly,
because the survey’s focus is on asset ownership, the data collected in a dedicated sur-
vey are less likely to suffer from respondent fatigue than data collected from a module
appended to a household survey on another topic.
239. A disadvantage of stand-alone surveys is that they typically require more
resources to implement than appending a minimum set of questions or a module to an
existing household survey. In countries that conduct many official surveys in one year,
it may also be difficult to find the time and resources to include a dedicated survey on
asset ownership in the survey pipeline, in particular within the constraints of a limited
survey budget.

2.4. Choosing between the three data collection strategies

240. In choosing between the three data collection strategies, countries should
first determine which assets to collect data on by assessing their policy needs, includ-
ing the types of measures needed by data users, in consultation with the survey team’s
stakeholders comprising policymakers, researchers, gender specialists and other data
users. In tandem, countries should assess whether they have existing data, either
through statistical or administrative sources, on the prevalence of each asset type (e.g.,
principal dwellings or agricultural land) within the country to determine for which
assets individual-level data should be collected. For example, an industrialized coun-
try in which less than 5 per cent of the population owns agricultural equipment may
choose not to collect data on agricultural equipment, while a country with a largely
agrarian economy and a higher percentage of agricultural equipment may opt to do so.
241. If no such data exist for each of the types of assets of interest, a country may
choose to add a few questions on asset ownership at the household level to an existing
and nationally representative survey that will be implemented well in advance of the
data collection on individual-level asset ownership. The information on asset preva-
lence obtained at the household level through this exercise can be used to determine
which assets should be included in future data collection on individual-level asset own-
ership, along with the required sample size. As an alternative but less robust option,
countries may choose to conduct a series of focus group discussions with individuals
to identify which assets women and men own. While, owing to the non-probabilistic
approach in which focus groups are selected, this approach will not yield prevalence
estimates of asset ownership, it can still identify the assets that women and men con-
sider important to their livelihoods and well-being that should be included in national
data collections. If national statistical agencies choose this approach, skilled facilitators
should be employed to lead the focus groups and a sufficient number of groups should
be conducted in various regions of the country to ensure variation in land tenure sys-
tems and gender norms and other characteristics that may influence asset ownership.
242. Similarly, each national statistical agency will need to determine the
relevance of the bundle of ownership rights to its country context. As discussed in
 Guidance for implementation 71

chapter I, this assessment should entail an analysis of the legal framework, including
statutory and customary laws, on property rights and also of the social norms medi-
ating those rights. If countries opt to conduct focus group discussions, the bundle of
ownership rights and other themes important to the data collection, such as modes of
asset acquisition, can be explored. The qualitative information obtained would serve
both to inform whether data should be collected on the full bundle of ownership rights
and subsequent questionnaire design and also to interpret the quantitative findings of
the survey.
243. With a clear understanding of which assets and which measures are
needed, countries should consider the following guidance:
•• If countries are only interested in deriving prevalence measures, by gender,
on the ownership of priority assets, including principal dwellings, agricul-
tural land, other real estate and financial assets, they are advised to append
the minimum set of questions, using individuals as the unit of analysis, to
an existing household survey.
•• If countries want to collect data on priority assets to calculate additional
measures, such as the gender wealth gap, which requires that assets be val-
ued asset by asset, they are advised to append a module, using assets as the
unit of observation, to an existing household survey.
•• If countries wish to collect information on the full range of physical and
financial assets, and on their characteristics, they are advised to implement
a stand-alone, dedicated survey on asset ownership.

Key points
•• Data on individual-level asset ownership and control may be collected through,
first, appending a minimum set of questions on ownership and rights to an existing
household survey; second, appending a module on asset ownership, rights and asset
characteristics, to an existing household survey; and, third, conducting a stand-alone
survey. Each of these data collection strategies has its advantages and limitations, as
summarized in the following table:

Integrating Stand-alone survey

a minimum set Appending on ownership, rights,
of questions on a module on asset characteristics,
ownership and ownership, rights, within-household
rights asset characteristics decision-making
Data and indicators to Gender asset gap – Gender asset gap – Gender asset gap
be produced (based – Gender wealth gap – Gender wealth gap
on objectives of data – Intrahousehold analysis – Intrahousehold analysis
collection) (if more than one respond-
ent from each household)
Asset roster needed? No Yes Yes
Assets covered Priority assets Priority assets All relevant assets
Flexibility in sample Less flexible Less flexible Flexible
design
Cost Less costly Costly More costly
72

244. Ultimately, the determination of which data collection strategy to adopt

should be considered in light of the objectives of data collection, the resources available
for the data collection, including the capacity to implement the required respondent
selection protocols, and the overall work programme of the national statistical agency.

3. Modes of data collection

245. This section briefly introduces modes of data collection typically used
in household surveys, including face-to face interviews, telephone interviews,
­self-enumeration methods and computer-assisted interviewing. It then focuses on
face-to-face interviews, including the advantages and limitations of using paper ques-
tionnaires versus computer-assisted personal interviewing (CAPI) questionnaires.
The focus on face-to-face interviewing is attributable to three factors: first, it is the
most common method; second, it meets key sampling and field operations require-
ments when implementing a survey on asset ownership from a gender perspective;
and, third, it achieves a high degree of cooperation, resulting in higher response rates
and data that are more complete and accurate. The discussion on the advantages and
limitations of a paper questionnaire versus a CAPI approach summarizes a few general
issues related to data quality and timeliness, and emphasizes aspects that are most rel-
evant for data collection on asset ownership, including how these methods deal with
the complexity of using multiple rosters (of individuals and assets) in the household.

3.1. Basic modes of data collection

246. The modes of data collection used by national statistical offices vary across
countries and across surveys within the same country. A household survey may
employ one mode of data collection or a combination of two or more methods. The
mode of data collection has implications in terms of logistical requirements for the
survey operations, procedures related to sampling, number and qualifications of the
enumerators, training needs, and, consequently, the cost of the survey. Accordingly,
a decision on the mode of data collection should be made by countries early in the
planning stage of the household survey, based on the objectives and scope of the sur-
vey, previous experience in data collection, available resources, characteristics of the
population, such as literacy rates and coverage of phone and internet services, and
availability of sampling frames.
247. There are three basic modes of data collection used in household surveys
that countries may consider: face-to-face interviews, telephone interviews and self-
administered questionnaires. Face-to-face interviews are the most common method,
in particular in developing countries and in population groups with significant illiter-
acy rates. In this method, information is obtained from one or more household mem-
bers and entered in the questionnaire by an enumerator (field interviewer) designated
to visit that household and conduct an interview for the purpose of data collection. A
high degree of cooperation from respondents is usually achieved, which translates into
103 Robert Groves and others, response rates that are typically higher than those of other methods.103 Data obtained
Survey Methodology, 2nd ed. in face-to-face interviews may also be more complete and accurate because of the
(Hoboken, New Jersey, John potential for interaction between the enumerator and the respondent, and the oppor-
Wiley and Sons, 2009). tunity to clarify some of the questions and probe for more adequate answers. At the
same time, face-to-face interviews require highly trained enumerators and are likely
to be more costly than other data collection modes, primarily because of the need to
travel to respondents’ residences.
 Guidance for implementation 73

248. Telephone interviews are increasingly used, but require that telephone
services have broad coverage. Surveys based on telephone interviewing are cheaper
than face-to-face surveys and may be completed faster than surveys involving a
­self-administered questionnaire. Their main limitation is their incomplete coverage,
resulting in a high proportion of non-responses. In addition, telephone interviewing
may result in higher coverage error when the survey requires a listing of all household
members with a subsequent random selection of a person in the household.104 104 Ibid.
249. Self-administered questionnaires are more often used in developed coun-
tries. In the self-enumeration method, questionnaires are distributed to households
selected in the survey sample and collected by mail, email or posting on an Internet site.
The major responsibility for entering the information in the questionnaire is given to a
person in the household. The sample population must be literate and, in the case of web-
based surveys, able to access the Internet through computers or handheld devices. Self-
enumeration questionnaires need to be limited in length in order to avoid confusion
and reduce non-response. In addition, when using self-enumeration, there are no estab-
lished methods for meeting key sampling and operational requirements for collecting
the data on asset ownership presented in these Guidelines; methods range from ran-
domly selecting one person in the household to interviewing multiple respondents.105 105 These requirements are
250. More recently, computer-assisted versions of these three methods have explained in detail in the
sections of these guidelines
been developed and countries are increasingly using an electronic questionnaire. Com- on sample design and field
puter-assisted interviewing may take the form of a computer-assisted personal inter- operations.
view (CAPI), a computer-assisted telephone interview (CATI), a computer-assisted
self-interview (CASI) or an audio-computer-assisted self-interview (ACASI).106 106 Groves and others, Survey
251. There are several advantages to using technology in data collection. Data Methodology.
will be available much faster for analysis because the data are transferred to a central
database immediately or soon after data collection in a household. Electronic forms
reduce the amount of material (such as questionnaires) to be printed, distributed and
returned, and reduce data entry costs and errors. In addition, the need to securely
store completed paper questionnaires is eliminated, thus contributing to the privacy
of respondents and the confidentiality of data. Instead, national offices need to ensure
that online transmission is encrypted and secured for confidentiality purposes. Most
important, electronic forms can improve data quality by implementing validation rules
on individual questions, cross‐validation between questions and automatic sequenc-
ing of questions (leading the operator to the next appropriate question). More options
in pull‐down lists may be implemented, thus capturing more detailed data. Finally,
electronic questionnaires can also give enumerators access to material provided by a
“help” function that can be used during interviews.

3.2. Implementing face-to-face surveys using

paper questionnaires versus CAPI approach
252. As mentioned above, face-to-face interviews, the most common method of
data collection in surveys, have the advantage of achieving higher response rates and
obtaining data that are more complete and accurate. This is particularly the case when
the questionnaires and the interviewer protocols are complex, as in the EDGE pilot
studies. All seven EDGE pilot studies were based on face-to-face interviews, reflect-
ing the typical mode of data collection used by the national statistical offices in those
countries. Five countries used paper questionnaires, while two countries, South Africa
and Uganda, collected data using the CAPI software developed by the World Bank
known as Survey Solutions. Each of the two methods proved to have its own advan-
tages and challenges in collecting data on asset ownership.
74

253. Paper-based data collection has been used for decades, and many countries
have accumulated extensive experience with this mode of data collection in terms of
designing and testing questionnaires, building networks of skilled enumerators and
trainers and implementing quality assurance procedures for field operations. Never-
theless, use of a paper questionnaire in collecting data on asset ownership involves
certain specific challenges when complex survey instruments are used. One of these
relates to the creation and use of multiple rosters, one roster for household members
and additional rosters for the different inventories of assets that are collected. For
example, one of the requirements in constructing asset rosters is the need to list all
asset items in a roster before starting to record specific information for each asset item.
This technique prevents the underreporting of asset items due to respondent fatigue.
While this aspect can be emphasized during training, enumerators may not necessar-
ily follow the rule in the field and proceed to complete all the questions related to an
asset before listing a second asset.

3.2.1. Advantages of using a CAPI questionnaire

254. Use of the CAPI approach can address some of the challenges that are com-
mon in household surveys or specific to the measurement of asset ownership. As with
other computer-assisted modes of data collection, CAPI considerably reduces the time
lag between data collection and data analysis because data entry and certain data vali-
107 The term “handheld electronic dation procedures can be embedded in the process of recording information obtained
device” typically refers to a
from respondents. In addition, the CAPI questionnaire can be designed to facilitate
small device that provides
computing and information better data quality through the way in which the rosters of assets are constructed and
storage along with retrieval displayed. By design, the enumerator has to complete a roster of assets before recording
capabilities. The typical further information for each asset. When collecting information regarding the joint
handheld electronic device owners of a particular asset, the name or identification of the asset in question remains
has a touch-screen interface displayed, so that the enumerators are reminded of the subject of observation. The ros-
for input and output, along ter of household members, including their names, is also displayed, reducing the poten-
with a miniature or a vir-
tial for errors in recording the joint owners of that particular asset.
tual keyboard. Most handheld
electronic devices have an 255. Better data quality is also made possible by embedding in the CAPI ques-
operating system and can run tionnaire design an algorithm for within-household respondent selection based on
various types of application randomization procedures. When using a paper questionnaire, some enumerators
software. Most are equipped
may have difficulties in correctly using the method prescribed for randomly selecting
with capabilities for connec-
tion to cellular networks and a respondent (such as the Kish selection method or a selection method based on birth
for establishing connectivity to date), as observed in some of the EDGE pilot studies, with a negative impact on the
the Internet and other devices quality of the estimates obtained in the survey.
such as a personal computer
256. In addition, the CAPI approach can capture a range of operational infor-
and other mobile devices
through mechanisms includ- mation that can be used to monitor operations and analyse responses. For example,
ing Wi-Fi and the Bluetooth, although the duration of an interview may be recorded manually in a paper question-
Infrared Data Association naire, use of CAPI enables a detailed analysis of the time taken by the exercise, includ-
(IRDA) and near field commu- ing the duration of the entire questionnaire and the time spent on each module and
nication (NFC) technologies. question. Lastly, use of CAPI is also conducive to the more efficient management of
The synchronization func- interviewers, including the updating of enumerators’ assignments and checking of the
tion of these devices allows
completed questionnaires by the supervisors.
the exchange of data with a
personal computer or other
devices. Handheld devices are 3.2.2. Costs and risks associated with using CAPI
available in a variety of forms,
including personal digital
257. At the same time, however, there are costs and risks associated with using
assistants, tablet computers, CAPI instead of a paper questionnaire. When considering the CAPI method, the cost of
smart phones and ultra-mobile providing all interviewers with the electronic device107 used to administer the question-
personal computers. naire must be incorporated into the project budget. All interviewers must have their
 Guidance for implementation 75

own tablet computer for data collection, which can represent a substantial initial invest-
ment. Less expensive tablets and notebooks are becoming widely available, however,
enabling computer costs to be offset by savings derived from eliminating the printing,
editing and transport of the questionnaires and the transfer of data from paper forms to
an electronic database. These devices can also be reused in future surveys. Accordingly,
use of CAPI is typically more expensive for the first one or two surveys, but subsequent
surveys should be far less expensive. Additional costs when using CAPI may include the
human resources and time needed for programming and additional CAPI training for
interviewers, field supervisors and headquarters staff; cost of access to server hardware,
software and server maintenance; and technical support.
258. Furthermore, more preparation time is needed before starting data col-
lection in the field. The additional time to be allocated to field preparation activities
should not be underestimated. It also should be emphasized that, when not enough
time is allocated for the development and testing of the CAPI questionnaire, data qual-
ity may be severely compromised. Based on the EDGE pilot studies in South Africa
and Uganda, additional time needs to be allocated to field preparation when using
CAPI, to allow for CAPI design, testing, and training in CAPI-specific issues.
259. Nevertheless, when using CAPI, statistical offices are strongly encouraged
to develop a paper questionnaire first. The paper questionnaire will serve several pur-
poses. First, a draft questionnaire will need to be developed in parallel with the tabu-
lation and data analysis plan and the objectives of the survey, to ensure consistency
among all three elements. The paper questionnaire can be shared with the entire team
and stakeholders to ensure exchange of ideas and communication.
260. Second, the paper questionnaire will provide a full picture of the organiza-
tion of the questions in modules and sections, the flow from one section to another
and the sequence of questions within each section. It is important that the paper ques-
tionnaire contains all the questions and skip patterns needed. Having the entire logi-
cal design on paper will significantly facilitate implementation of the right sequence of
questions and logical validations in CAPI. It is important that the paper questionnaire
is implemented in CAPI only after it has been tested, finalized and approved. This
will prevent going back and forth in numbering the questions and redoing validation
checks in the CAPI questionnaire.
261. Third, the paper questionnaire is an invaluable tool to be used during
the training of the enumerators. Trainers and trainees can easily refer to the paper
questionnaire for a variety of purposes, including understanding the scope of data
collection and how key concepts are operationalized, illustrating the sequencing of
questions and emphasizing difficult questions that need to be probed further, without
getting distracted by the use of technology. Fourth, the paper questionnaire can be
made available by itself or can accompany the statistical publications and products
developed after data collection.
262. The CAPI questionnaire will be used specifically for the following pur-
poses: test the flow of the questionnaire and validation rules in the field and the com-
munication between the different components of the system involved in data transfer;
conduct the training of enumerators and supervisors on CAPI-specific issues; conduct
the field practice for enumerators and supervisors; and collect the data once the field-
work commences.
263. Countries with little experience in conducting surveys with the use of a 108 United Nations, Guidelines on
handheld device may refer to box 5, which outlines a list of key issues to be considered Use of Electronic Data Collec-
when planning such surveys, and to the forthcoming guidelines on the use of elec- tion Technologies in Censuses
tronic data collection technologies in censuses,108 for additional information. (forthcoming).
76

Box 5
Key steps to consider when conducting a survey using a handheld device
Successful data collection exercises using handheld devices require careful planning and
a number of elements need to be considered, including:
(a) Timetable: this should be adjusted to fit the needs of collecting data using a hand-
held device, particularly when the technology is being used for the first time. More
time is needed to develop and test the application; set up data transfer and process-
ing systems; procure, programme and test the handheld devices; plan edit checks
in the program; design and test the electronic questionnaire; test and debug the
software; and train field staff to ensure proficient use of the handheld devices;
(b) Budget: using an electronic questionnaire may save the cost of printing paper
questionnaires and data capture. The additional cost of using handheld devices
should also be considered, however, including system design, software devel-
opment, hardware acquisition, communications, system maintenance, technical
support, human resource planning and the additional training required;
(c) Questionnaire development process: developing an electronic questionnaire is an
iterative process. Both the technical and content elements must be developed,
tested, revised and then tested again, repeating the cycle until the questionnaire
works as intended. It is critical that the subject matter specialists work closely with
the programmers throughout the process to ensure that there is clear communi-
cation and understanding regarding the purpose of the questionnaire content,
layout and design, data validation, and other specifications;
(d) Infrastructure considerations such as availability of electricity and Internet access
should be assessed before the electronic data collection process is launched. In
the early planning stage, areas that lack electricity or Internet coverage or both
should be identified. If a cellular or Wi-Fi network needs to be used to transmit
data, the speed of the data transmission should be tested. Contingency plans
should be developed for charging and backing up devices for sampling areas
where electricity or Internet access is not available;
(e) System design, software development and hardware acquisition: it is necessary
to take into consideration the minimum requirements for installing and operat-
ing the electronic questionnaire, in addition to any restrictions on the operating
system on which it works;
(f) Data transfer from the field: if the data collection strategy requires transmitting
data and monitoring fieldwork directly from the handheld devices in the field, then
reliable cellular coverage is essential. If cellular network coverage is poor, mecha-
nisms need to be developed for data transfer from the field to a central server by
establishing multiple data collection stations with reliable Internet connection;
(g) Data security: after completing each interview, data should be saved and secured
until they can be transmitted to the central database. Data collection through
handheld devices requires investments in data security and staff training to pre-
vent unauthorized access and the loss of sensitive personal data. Security con-
cerns include failure in hardware and software, human error and accidents. Data
transfer protocols from the field should be designed with specific security fea-
tures, including encryption;
(h) Technical skills and capacity development: careful consideration should be given
Source: United Nations, to the type of expertise needed to build, integrate and implement a handheld col-
Guidelines on Use of Electronic lection system. This requires evaluating the technical skills of, and the distribution
Data Collection Technologies in of responsibilities among, the staff of the national statistical office and developing
Censuses (forthcoming). training programmes or hiring external contractors.
 Guidance for implementation 77

Key points
•• The modes of data collection typically used in household surveys include face-to-face
interviews, telephone interviews, self-enumeration methods, and computer-assisted
interviewing.
•• A decision on the mode of data collection should be made by countries early in the
planning stage of the household survey, based on the objectives and scope of the
survey, previous experience in data collection, available resources, characteristics of
the population, such as literacy rates and coverage of phone and Internet services,
and availability of sampling frames.
•• When the questionnaire and the interviewers’ protocols are complex, as in the case
of surveys on asset ownership from a gender perspective, face-to-face interviews are
preferred.
•• Face-to-face interviews on asset ownership can be carried out through the traditional
paper questionnaire or through an electronic questionnaire conducted on a hand-
held electronic device (CAPI method).
•• Countries interested in using electronic questionnaires should take into consideration
a number of important elements during the planning stage of the household survey
(see box 5).

4. Sample design
264. Sample design is a process that specifies how to select a sample of elements
from a sampling frame and how to compute estimates using sample data. The goal is
to provide estimates of certain properties in the population from which the sample
was drawn and make statements about the uncertainty of those estimates, because a
sample rather than a complete enumeration of all elements was selected.
265. Official statistical systems generally prefer that the elements in the sam-
ple be randomly selected with a non-zero probability, properly representing the target
population, along with key subgroups of the target population. A survey collecting
information on asset ownership, as with a survey on any topic, needs to satisfy survey
objectives, take into account the mode of data collection and the fieldwork constraints,
be efficient in terms of cost and the precision of the survey estimates and be practically
feasible in a country.
266. In most countries, no comprehensive population or household register
is available. Hence a stratified multistage area sample design is used. The sample is
selected in stages so that locations where interviews are conducted require less travel
and the households are chosen efficiently. To ensure the representation of population
subgroups in the sample, first-stage sampling units, such as enumeration areas defined
in a population census, are divided into mutually exclusive strata, based on informa-
tion that is available for every element in the first stage frame. Clusters are then selected
independently across the strata. Within selected clusters, households are selected from
a list of households in the selected cluster created at the time of household selection or
obtained from official sources, to keep costs at a manageable level. For surveys requir-
ing the sampling of individuals, a last step of the sampling process involves the selec-
tion of one or more individuals from selected households, who are then interviewed.
267. In countries that maintain a comprehensive and up-to-date population
register, the selection of individuals may be carried out either through systematic
sampling from a purposively ordered list of people registered in the system or system-
atic sampling from a purposively ordered list of addresses followed by the selection
78

of individuals from within the household. Even if a population or address register is

available, individuals or addresses may be clustered into first stage sampling units, as
in stratified multistage area sampling, and multiple individuals or addresses can be
selected from chosen sampling units. If, on the other hand, a selection of individuals
is carried out directly from a register, the ordering of people in the register should
be carefully sorted by key characteristics available in the register that are correlated
with the measurement of asset ownership. Systematic selection applied to such a sorted
list is often called implicit stratification and is helpful in achieving smaller standard
errors for virtually the same cost as systematic sampling from a randomly ordered or
unsorted register. Information on relevant correlates is available below in the section
on stratification.
268. Countries measuring asset ownership at the individual level may choose to
implement a stand-alone or dedicated survey, append a survey module to an existing
household survey, or integrate a minimum set of questions into the questionnaire of
an existing household survey. If a stand-alone survey is implemented, sample selection
involves selecting sample households and selecting individuals from households. If a
survey module is appended to an existing household survey, sample households would
have already been selected and the sampling process for the purpose of collecting data
on individual-level asset ownership would only involve the selection of individuals
from the sampled households.
269. Designing the sample for a survey requires comprehensive knowledge of
the principles and techniques of sampling. Survey managers without such compre-
hensive knowledge must seek assistance from a specialized statistician early in the
planning stage.
270. While a thorough review of the principles and techniques is beyond the
scope of these Guidelines, this section will cover basic principles in sampling house-
holds and individuals, along with issues that need to be taken into consideration for
household surveys collecting individual-level asset ownership data. This section will
also cover in detail the selection of individuals within households.

4.1. Principles in sampling

271. This section outlines basic principles in sampling households and indi-
viduals, including, first, the target population; second, the sampling frame; third, the
sample size; and, fourth, the structure of the sample. Whenever relevant, issues relat-
ing to measuring asset ownership from a gender perspective will be highlighted.

4.1.1. Target population

272. The target population refers to the population or universe that is the focus
of the study. Depending on the objective of the survey, the target population could be
person-based or entities other than persons, such as establishments for establishment
surveys and agricultural holdings for agricultural surveys. Population-based surveys
are typically used for surveys of persons, and often limit their focus to a specific target
population such as people residing in the country, including those living in house-
holds or institutions, excluding population subgroups such as the homeless or those
109 The present Guidelines recom-
in the military.
mend collecting data on asset
ownership for individuals aged
273. More specific objectives of the survey could further limit the scope of the
18 and over. Countries may, target population. For example, as these Guidelines focus on collecting information on
however, use a different age asset ownership at the individual level, the target population will be limited to usual
cut-off as appropriate. residents living in households who are above a certain age.109 If the objective were to
 Guidance for implementation 79

understand decision-making and power dynamics between husbands and wives or

between unmarried partners, the target population would be persons who are either
married or residing with a partner.
274. Other restrictions may arise depending on the survey focus. For example,
the measurement of women’s land ownership, as presented in the 2030 Agenda (under
Sustainable Development Goal indicator 5.a.1 on the proportion of the total agricul-
tural population with ownership or secure rights over agricultural land, by gender),
could restrict the population to the agricultural population. This can be done by limit-
ing data collection to individuals only in a specific population, or the subpopulation
may be treated as a subgroup that is of interest in a survey with a broader target popu-
lation.110 110 See box 1 for an overview of

the 2030 Agenda and indicator

4.1.2. Sampling frame 5.a.1.

275. Sampling frames are lists or source materials used to select the sample.
Ideally, the sampling frame is a perfect match to the target population. In a multistage
sample design, the sampling frame is different for each stage. Surveys on asset owner-
ship could, as noted above, require an area sampling frame composed of lists and maps
of geographical units for the first stage of sampling, a list and a map of households for
the second stage, and a list of individuals in selected households for the final stage of
sampling.
276. An area sampling frame consists of geographical units arranged hierar-
chically. An area frame may include provinces, districts, tracts, wards and villages
(rural areas) or blocks (urban areas). For census purposes, these administrative subdi-
visions are further divided into enumeration areas. The enumeration area is typically
the smallest geographical unit that is defined and delineated in a country, making it
a natural and convenient choice for the primary sampling unit in household sample
surveys. The use of census frame lists of households is not an uncommon practice in
the field. If household census information is outdated, the frame can still be used, but a
new or updated listing of households within the selected enumeration areas is advised.
277. For surveys designed to collect data on asset ownership at the individual
level, the last stage sample frame includes a list of individuals in selected households.
It is recommended that the individuals listed be those aged 18 years or older. The list is
constructed by asking one informant, such as the head of household or a knowledge-
able household member, to identify individuals who are residents of the household
and provide their birthdates so those who are aged 18 and older can be identified.
Residency rules must be specified so that the enumerators can determine, based on the
information provided by the informant, whom to list as residents.
278. Two different residency rules are generally used. One is a de facto resi-
dence, based on the place where the person resides at the time of the data collection,
usually the night before data are collected. The other is a de jure residence, based on
the place where the person usually resides. De facto residence is more straightforward
for the informant to report—anybody who spent the night prior to data collection in
the household would be listed as a household member. Frames based on de facto resi-
dents generally fit better with surveys that take a relatively short period of time. If the
enumeration is extended over a period of weeks or months, the risk of either overcoun-
ting or undercounting household members increases. For example, one person who
sleeps at multiple locations might be included multiple times under a de facto rule, or
this person may not be included at any location.
279. Usual residents of a household are defined as persons who have lived in
the household for at least a specified period of time, or who intend to stay there for
80

some time. The minimum duration of stay, either actual or intended, that is required
to qualify as a usual resident of a household varies from country to country. It is rec-
ommended that a threshold of 12 months be applied when considering place of usual
111 United Nations, Principles and residence.111 Even with a very clear cut-off threshold, it is not always easy to identify
Recommendations for Popula- the usual residents of a household. The informant might understand the question dif-
tion and Housing Censuses, ferently from the survey’s intention, in particular for certain groups of people whose
rev. 3, 2014, para. 2.50. residence is difficult to define. For example, people who maintain multiple residences
and students who attend boarding school and stay away from the family while main-
taining a close tie with the family are types of individuals who are either missed or
112 For a more complete list of overrepresented under a de jure rule.112 Whichever residency rule is used, clear and
these population groups, see specific instruction should be provided to enumerators and respondents about the
United Nations, Principles and manner in which different types of resident groups should be treated.
Recommendations for Popula-
tion and Housing Censuses, 280. Another important consideration in deciding whom to include as a resi-
rev. 3, 2014, para. 4.43. dent of a household is to maintain comparability with population censuses and other
household surveys conducted in the country. It is usually plausible to keep the same
residence definition throughout all data collections, unless there is a specific reason for
using a criterion for a given survey that differs from the criterion used in the rest of the
surveys conducted within the same country.

281. Those designing surveys on individual-level asset ownership should also

consider errors in frames and remedies to address them during data collection or esti-
mation. For example, non-coverage error arises when a sampling frame fails to cover
all of the target population. Non-coverage can occur at the level of the primary sam-
pling unit—the household—and at the individual level. For developing and transition
countries, non-coverage is a more serious problem at the household and individual
113 Ibrahim Yansaneh, “Overview levels.113 Listing of households within selected primary sampling units immediately
of sample design issues for before data collection is helpful in reducing household-level frame non-coverage. For
household surveys in develop- individual-level non-coverage, clear guidance to both the enumerator and respond-
ing and transition countries”, ents on whom to include in the household roster is crucial.
in United Nation, Household
Sample Surveys in Developing 282. An additional frame issue concerns blank elements in the frame, when
and Transition Countries, Stud-
ies in Methods, Series F, No. 96
some listings in the frame contain no elements of the target population. In the case of
(New York, 2008). surveys on asset ownership from a gender perspective, households headed by children
114 Around 1 per cent of the or those younger than the target age range would fall into this category.114 This would
households in KwaZulu-Natal, not be discovered until after the household has been visited and a roster of eligible
South Africa were headed by individuals created. These households should be removed from the sample once indi-
children under the age of 18. viduals are listed, unless the country has a policy interest in collecting data on asset
See Statistics South Africa, ownership for younger age cohorts.
Census 2011 Municipal Report—
KwaZulu-Natal (Pretoria, 2012). 283. The clustering of target population elements within the frame is another
frame problem to consider. Clustering arises when a single listing in the sampling
frame actually consists of multiple elements in the target population. For example,
in the EDGE pilot survey in South Africa, in some instances multiple households or
families appeared in one dwelling unit. In these cases, all households and families in
the selected dwelling unit were included in the sample.

284. One final frame problem that arises is duplicate listings. This problem is
less likely to occur in household surveys requiring personal visits than in telephone
sampling frames in which one person has multiple phone numbers. Yet, it is still possi-
ble in household surveys that a person might be included in more than one household
during the data collection. This further highlights the importance of following strict
guidelines on listing household members under de jure rules, as mentioned earlier.
 Guidance for implementation 81

285. Countries could also use a population register as a sampling frame. In this
case, however, there might be over- or undercoverage of the target population, even in
countries with very well-maintained population registers.
286. There is another issue associated with the use of population registers as
frames in respect of the information needed for surveys on asset ownership from a
gender perspective. Individual-level registries may not include data about the usual
private household status for the individual, also referred to as “housekeeping house-
holds” data.115 Population registers do not group individuals by private household, 115 United Nations, Economic

more typically instead providing information only about “dwelling households”. Commission for Europe, Confer-
Dwelling households include all persons living in the same housing unit as members of ence of European Statisticians
the same household. Dwelling households could therefore include one or more house- Recommendations for the 2020
Censuses of Population and
keeping households. For asset ownership dynamics within the household, the con-
Housing (Geneva, 2015).
cept of “dwelling households” is problematic, as respondents selected from the same
dwelling household but different housekeeping households do not provide meaning-
ful information on intrahousehold ownership dynamics. Lastly, another key piece of
information that is usually missing from population registers is the status of partners
living together, as is the case when only legal marital status is recorded in the system.

4.1.3. Sample size determination

287. Factors that must be considered when determining the appropriate sample
size for a survey on asset ownership from a gender perspective are similar to those that
must be considered for any statistical sample survey. They include the following:
•• Level of precision required for the key estimates to be obtained from the
survey (of which there are usually several);
•• Number of planned subgroups of the population for which estimates will
be produced—for example, estimates may be needed separately for urban
and rural areas, geographical regions and population subgroups, such as
age groups and minority groups;
•• Population variance, which requires prior knowledge about the approxi-
mate prevalence in the population of the key characteristics to be meas-
ured, usually obtained from past data or data from another country;
•• The extent of anticipated levels of non-response among households or indi-
viduals.
288. The level of precision is a major consideration when determining the size
of the survey sample. As a general rule, the more precise or reliable the survey esti-
mates need to be, the larger the sample. It must be noted that, in estimating precision,
sampling error needs to be estimated in a manner that takes into account the sample
design that is used. Clusters in selection increase sampling variance, while stratifi-
cation may reduce sampling variance. In a typical multi-stage cluster survey, using
clusters would require a larger sample size than in surveys that use simple random
sampling in order to achieve the same precision in the cluster sample as obtained
in a simple random selection. The size of the cluster sample depends on how closely
associated or alike cluster elements are, relative to elements in other clusters. Data on
intracluster association or correlation, and its impact on sample size, is discussed later
when the use of clusters is examined.
289. The need for estimates for subgroups of the population increases the sam-
ple size required as well. The subgroups are generally analytical subgroups for which
equally reliable data are wanted. For surveys on measuring ownership from a gender
perspective, the two essential subgroups are women and men. Another subgroup that
82

is relevant for the ownership of agricultural land, agricultural equipment and livestock
is the agricultural population (see, for example, box 1 on Sustainable Development
116 There is no international Goal indicator 5.a.1).116 Separate estimates by region would also be relevant for coun-
agreement on the definition tries that have different marital regimes and land tenure systems in different regions
of agricultural population. The of the country. While it is desirable to have data for many subgroups, the number of
most recent proposal by FAO subgroups has to be carefully considered, as an impracticably large sample size may be
defining agricultural popula-
required to produce reliable estimates for a large number of subpopulations.117
tion for measuring Sustainable
Development Goal indicator 290. The prevalence of the key variables of interest also plays an important role
5.a.1 can be seen in box 1. in calculating the sample size for the survey. When a proportion is to be estimated,
117 United Nations, Designing
such as the proportion of women owning assets, and if prior knowledge on ownership
Household Survey Samples: prevalence among women and men is available, it is possible to calculate the sample
Practical Guidelines, Studies in
size required for the survey to reach the required precision. Measuring a rare or very
Methods, Series F, No. 98
(New York, 2008). low prevalence event requires a much larger sample than an event of medium preva-
lence. For example, the ownership of dwellings is usually quite common, and meas-
uring it would not require as large a sample as measuring the prevalence of owning
agricultural land, which can vary greatly from one country to another. In Uganda
118 International Labour Organiza- some 70 per cent118 of the population are employed in the agricultural sector, and the
tion, Employment by sec- reported ownership of agricultural land is around 60 per cent for men and 30 per cent
tor—ILO modelled estimates, for women. In Mongolia, on the other hand, while around 19 per cent of men living
May 2018. Available at in rural areas report exclusive or joint ownership of agricultural land, only 5 per cent
www.ilo.org/ilostat/faces/
of rural women report owning this asset.119 A larger sample would thus be needed to
wcnav_defaultSelection?_
afrLoop=704653335497001&_ produce a reliable estimate of the prevalence of owning agricultural land in Mongolia
afrWindowMode=0&_ than in Uganda.
afrWindowId=ww2zuc2wi_ 291. If women’s ownership of assets is significantly lower than men’s, oversam-
1#!%40%40%3F_afr
pling women within households is one strategy that may be used to increase sample
WindowId%3Dww2zuc2wi_
1%26_afrLoop%3D70465333 size. Ultimately, determining sample sizes is a process that must balance the overall
5497001%26_afrWindow budget and the objectives of the survey. If the prevalence of owning a particular asset is
Mode% 3D0%26_adf.ctrl- extremely low, it might be in the interests of the survey planners, in consultation with
state%3Dww2zuc2wi_74. policymakers and other stakeholders, to reconsider inclusion of this asset.
119 Findings from the EDGE pilot
292. On the other hand, the expected prevalence of individual-level asset own-
study on measuring asset own-
ership from a gender perspec- ership may not be known in advance, especially if it is the first time that a survey on
tive in Mongolia. this topic is being conducted in a country. In such cases, the national statistical office
should first assess whether any prior surveys have collected data on asset ownership
at the household level. From such data, an initial estimate of the prevalence can be
obtained for sample size calculation. If this is not possible, a rough estimate may be
calculated on the basis of household surveys conducted in other countries with similar
cultural contexts and property rights.
293. A final factor to consider when calculating sample size is anticipated non-
response. For surveys on individual-level asset ownership, non-response needs to be
taken into account at both the household and the individual levels. Refusals, non-
contacts and break-offs by a household are considered non-response at the household
level. If a household questionnaire is completed but not all selected respondents are
interviewed, then there is non-response at the individual level. Chapter IV discusses
how to handle non-response at both the household and the individual level. Non-
response is likely to vary by country and should be calculated on the basis of national
survey experience.
294. One important consideration for surveys on individual-level asset owner-
ship is the need to ensure that households in the entire spectrum of wealth are rep-
resented in the sample. Wealthy households tend to have lower response rates than
other households and, without proper representation of wealthy households, the over-
 Guidance for implementation 83

all estimates of asset ownership might be skewed downward. A higher non-response

rate might also occur in households that have very little wealth and hence low owner-
ship of assets. Sample selection should take this into consideration by oversampling
those households at both ends of the spectrum of wealth, which tend to have a high
­non-response rate in household surveys.

4.1.4. Structure of the sample

295. The structure of the sample for a survey of asset ownership at the individual
level will be broadly similar to that of other national household surveys within a coun-
try for sample selection up to the household level. In most countries, a stratified multi-
stage area sample design could be used. In a stratified multi-stage area sample design,
sampling efficiencies are achieved by techniques such as stratification and sampling
in stages. Each of these techniques figures prominently in national household surveys
that employ the traditional census frame or other frames based on household or indi-
vidual listings. The following discussion will briefly cover the basic principal of each
technique. For more details, readers should refer to standard textbooks on sampling or
a sampling specialist in their country.

4.1.4.1. Stratification
296. Stratification improves efficiency by reducing sampling variances. It also
occurs when separate estimates for each stratum are required, and it can be applied
to any stage of sampling. It divides the units to be sampled into mutually exclusive
and collectively exhaustive subgroups or strata, based on auxiliary information that
is known about the full population. Sample elements are selected from each stratum
independently.
297. One of the purposes of stratification is to reduce sampling variances and
gain efficiency. The gains in efficiency are guaranteed when strata sample sizes are pro-
portional to the strata population size, the strata formed are as different as possible
from each other and the units within the same stratum are as homogeneous as possible
in respect of the characteristics of interest in the survey. For surveys on asset owner-
ship from a gender perspective, regions that have different marital regimes and land
tenures should be placed in different strata. Dividing populations into urban and rural
residence is also preferred, since the ownership of agricultural land, agricultural equip-
ment and livestock would be very different for people living in urban and rural areas.
298. Another benefit of stratification is to guarantee the representation of
important domains and special subpopulations. The level of asset ownership at the
individual level is closely linked to the level of household wealth. It is therefore impor-
tant to reach respondents representing an entire spectrum of wealth levels, even
though in practice it is often difficult to reach the two extremes of the wealth distribu-
tion, the extremely poor and the extremely wealthy.

4.1.4.2. Cluster sampling

299. Cluster sampling is a random sampling technique whereby the study pop-
ulation is divided into clusters and a sample of those clusters is chosen. These clusters
are often naturally occurring units or groups, such as neighbourhoods, villages, enu-
meration areas or city blocks. The final sample of elements is then drawn from the
selected clusters.
300. Sometimes a household can also be considered a cluster when more than
one person from each household is selected. In practice, however, households are typi-
cally selected within enumeration areas. Then the cluster sample becomes multi-stage,
84

with a stage for the selection of enumeration areas, a stage for the selection of house-
holds within selected enumeration areas and a stage for selection of persons from
selected households.
301. In household surveys, the sample design will invariably and necessarily use
120 United Nations, Designing some form of cluster sampling if survey costs are to be contained.120 Cluster sampling
Household Survey Samples: is particularly cost-effective in face-to-face interview situations with widely dispersed
Practical Guidelines, 2008. populations, where the clustering of interviews in specific geographic areas can signifi-
cantly reduce travel costs and, hence, the overall costs of the survey. The disadvantage
of cluster sampling is that it decreases the reliability of the estimates, because people
living in the same cluster tend to be relatively alike in the characteristics under study.
Correlation among units within the same cluster inflates the variance—and therefore
lowers the precision—of the survey estimates.
302. The effects of clustering are measured by the design effect, which expresses
how much larger the sampling variance for the cluster sample is, compared to that for
a simple random sample of the same size. The design effect is generated by two fac-
tors, the intracluster homogeneity measurement (roh) and the size of the cluster. The
higher the intracluster homogeneity and the larger the size of each cluster, the higher
the design effect and the lower the precision of estimates. The intracluster homogene-
ity varies by the variables of interest. For example, studies across samples in different
countries show that intracluster homogeneity is higher for socioeconomic characteris-
121 Groves and others, Survey tics than for variables on attitudes and behaviour.121
methodology, 2009. 303. Before designing a survey, it is advised to use the intracluster homogeneity
122 Leslie Kish, Survey Sampling rate—or roh, the abbreviation coined by Leslie Kish122—as the value for calculating the
(New York, John Wiley and optimal cluster size. The roh can be based on information obtained from prior national
Sons, 1965). household surveys that investigated similar topics. If countries are conducting surveys
on the topic of asset ownership for the first time, it is possible to use the roh value from
another country for the initial calculation. Particular caution must, however, be exer-
cised in borrowing a value for roh, as it is more transferable for some variables of inter-
est, such as demographic variables, than others, such as variables on socioeconomic
conditions. For example, estimates of roh for a given variable in the demographic and
health surveys are fairly transferable across countries if sample designs are compa-
rable. But when the variable measures socioeconomic conditions such as household
consumption and ownership of household durables, the roh tends to vary from one
123 Martin Vaessen and others, country to another.123
“The demographic and health 304. A standard survey sample design uses criteria that lead to the design effect
surveys”, Household Sample
being kept as low as possible, given the cost constraints. An optimal survey sample
Surveys in Developing and Tran-
sition Countries; Hans Petterson design will use clusters with a large number of units from which selections can be
and Pedro Luis do Nasci- taken, and will sample a small number of units that are as diverse as possible, from
mento Silva, “Analysis of design each cluster. For example, if at the last stage of selection of households, there is a choice
effects for surveys in devel- between selecting a geographically dispersed sample or groups of households that are
oping countries”, Household closer together, the geographically dispersed sample is preferred, even though it may
Sample Surveys in Developing cost more to collect data from widespread households.
and Transition Countries.

4.2. Selecting individuals from households

305. In the conceptual framework on measuring asset ownership from a gender
perspective (see chap. I), it is recommended that information on individual-level asset
ownership be self-reported rather than reported by proxy. The selection of respondents
within households therefore needs to be dealt with carefully so that the individual
respondents are selected with a known probability and representative estimates can
be derived.
 Guidance for implementation 85

306. Depending on the objectives of the data collection exercise, the flexibility
to follow specific field protocols and the absorption of related costs within the house-
hold, selection of respondents for asset ownership studies can be achieved by, first,
selecting one person at random from all adult household members or, second, select-
ing more than one adult household member for interview. The second approach covers
a variety of options, ranging from interviewing two adult members to interviewing all
adult members.
307. Four different options are proposed under the second approach, where
more than one adult member is interviewed within the same household. In the first
option, all household members are interviewed, regardless of the household size. To
keep the number of interviews to a maximum of three within each household, in
the second option, one couple is selected at random from couples within the sam-
pled households and then a randomly selected individual is added from among the
adult household members who do not have a spouse or partner. In option three, one
respondent is selected at random from the household and, if that person has a spouse
or a partner, he or she will also be interviewed. In the fourth option, a fixed number
of respondents are selected randomly from each household. This option can be used
when EDGE questions are attached to a small household survey. Interviewing more
than one person in the household will increase the sample size and improve the pre-
cision of the estimates obtained. Other considerations should also be kept in mind,
however, when deciding whom to interview at the household level (see figure 3).

Figure 3
Approaches for selecting individuals from households

Approach 1: interview one adult Approach 2: interview more than one

member randomly from each household adult member from each household

Interview one
couple randomly
Interview one Interview
selected from
person randomly more than one
couples in the
Interview all selected from (fixed number)
household; and
adult household each household individual
one randomly
members plus that randomly
selected
person’s partner, selected from
individual from
if available the household
the non-couple
members

308. Making decisions on how respondents should be selected within house-

holds depend on a number of factors. The most important of these is the objective
of the data collection exercise. For the purposes of deriving nationally representative
indicators of women’s and men’s asset ownership, including Sustainable Development
Goal indicators 5.a.1 (a) and (b) (see box 1), and also of the gender wealth gap, either
one or multiple adult household members from each household could be selected for
interview. If the objective is to understand how assets are allocated and owned within
households, in other words, to conduct an intrahousehold analysis of asset ownership,
two or more adult household members must be interviewed. If the focus is asset allo-
86

cation among couples or partners within the household, then a sufficient number of
couples should be interviewed when designing individual-level sample selection.
309. Other factors to consider when deciding whom to interview include opera-
tional challenges and the cost of data collection. The following subsections discuss the
advantages and trade-offs of different within-household selection protocols in rela-
tion to these two aspects. The discussion will also reflect how operational challenges
and cost vary with different data collection strategies, namely, whether a stand-alone
survey on asset ownership is conducted or asset ownership modules or questions are
appended to an existing survey.

4.2.1. Operational challenges

310. Operational challenges for the different within-household individual selec-
tion protocols revolve around three aspects. The first relates to the random selection of
individuals. The second is associated with challenges in field operations when multiple
persons are to be interviewed separately and independently to avoid contamination.
The last aspect pertains to challenges in dealing with reporting discrepancies at the
data analysis stage when multiple household members are interviewed.

4.2.1.1. Random selection procedure within the household

311. The challenge of randomly selecting individuals to interview exists mainly
for surveys where face-to-face or telephone data collection is carried out using paper-
124 and-pencil methods. Before selecting respondents randomly within the household, a
Countries are advised to use
the random selection method household roster or list must be completed. Then a random selection process will be
that has been adopted in previ- carried out by the enumerators, ideally using the Kish selection method (as set out in
ous surveys. box 6).124

Box 6
Kish method for random selection of household members
For interviewers who are not trained statisticians, random selection under field conditions
a Leslie Kish, “A procedure is not at all easy to implement. In the Kish method of “objective respondent selection”,a
for objective respondent the selection of a single person from each household is made before the interviewer
selection within the arrives at the household. Two problems must be overcome, though, to ensure that the
household,” Journal of
process is ultimately objective and random.
the American Statistical
Association, vol. 44, No. 247 First, households may be of different sizes, that is, may have different numbers of eligi-
(1949), pp. 380–387. ble persons within them. For instance, a household may include only one adult, or it may
have two, three, four or more. At the doorstep or during a telephone interview, where
interviewers have a paper form to guide them, they are told on that form which person to
select. If there is only one eligible person, the interviewer is told to attempt an interview
with that person. But if there are two, the interviewer is told to list the persons, and then
interview either the first listed person, or the second. Whether the selected person is the
first or second is determined by a central office and printed on the form. Similarly, if there
are three eligible persons, the central office determines in advance whether the first, sec-
ond or third listed person is to be interviewed.
To implement the Kish procedure, the form must have a table that tells the interviewer
to first list the number of eligible persons, then count the total number. For each possible
number (one, two, three, four and so on), the form tells the interviewer which person on
the list to interview.
 Guidance for implementation 87

The Kish procedure introduces an additional requirement in the process. To avoid

introducing subjective procedures into the selection, it does not allow interviewers to
determine the order of the list of eligible persons. Instead, it requires them to list persons
in a particular order. For example, women may have to be listed first, from youngest to
oldest. Since the order in which persons are listed is objective (i.e., not determined by the
interviewer), the person selected to be interviewed would also have been determined
objectively.
There is one other feature of the Kish procedure that was designed to help organiza-
tions handle the problem of selecting respondents in a central office, then printing the
selection on a form that is assigned for each household. A very large number of possible
selection tables could be used. For example, suppose that, in a particular country, no more
than six eligible persons are expected to be in a household. A very large number of selec-
tion tables would be needed, any one of which could be assigned to a given household.
One selection table might tell the interviewer to select the only person listed, if one person
is in the household at the time of the interview; the first person, if two people are in the
household; the first person, if three, four, five or six people are in the household. A second
possible table might indicate to select the only person listed, if one person is in the house-
hold; the first person listed, if two, three, four or five people are in the household; but
the second person listed, if six people are in the household. A third possible table might
indicate to select the only person listed, if only person is in the household; the first person
listed, if two, three, four or five people are in the household; but the third person listed, if
six people are in the household. Altogether there would be 720 (6 x 5 x 4 x 3 x 2) possible
tables. The central office would assign one of the 720 tables to the first household in the
sample, another table, selected at random, to the second household, and so on.
In practical terms, however, keeping track of 720 tables and making selections for each
household in advance would be difficult in a paper-and-pencil operation. It means that
there would have to be 720 forms informing the interviewer which person to select for
each household, and these would have to be printed household by household. That is,
each of the 720 forms would be assigned to one 720th of the households. This process
would dramatically drive up the cost of the process of printing forms.
Kish proposed one further modification of the process to reduce printing costs in
paper-and-pencil operations. He devised a procedure whereby, for up to six possible per-
sons in a household, only eight forms would be needed. He found that, if a subset of the
720 possible forms were assigned to households in the right proportions, the right selec-
tion process was achieved. That is, with just eight forms, he managed to find a subset that
gave an equal chance to each of the two people in a household made up of two eligible
persons, and an equal chance to each of the three people in a household made up of three
eligible persons, and so on.
In order to do this, he determined that the eight forms would need to be distributed
not to one eighth of the households each, as in the 720-form case, but using a varying
fraction. Two of the eight forms were assigned to one third of the households, two forms
to one fourth, two forms to one sixth, and two forms to one twelfth. When properly bal-
anced, the end result was that each person in any household with one, two, three, four
and six eligible persons had the same chance of being selected. There was a slight dif-
ference in selection chances for people in five-person households, but the difference in
chances for the five people was not considered to be large enough to be of concern.

312. Training the enumerators in the Kish intrahousehold selection method is

no easy task. Sufficient training and exercises should be provided to enumerators to
ensure that the procedure is fully adopted by the enumerators in the field.
88

313. The challenge of using the Kish method, however, could be alleviated by
using computer-assisted interviewing techniques, such as CAPI (for further discus-
sion of CAPI, see chap. III, sect. 5). In computer-assisted data collection, the eligible
persons in the household can be listed in any order, and the computer can be pro-
grammed to select the person to be interviewed, using a random process. The proce-
dure is then objective—list order does not matter and the interviewer will not know
who is to be interviewed until after the list is entered and the computer program makes
the selection.
314. The EDGE pilot countries adopted different methods for the within-­
household random selection of respondents, following the usual practice in the
respective countries. The Kish method was used in Maldives and South Africa for the
selection of one respondent from each household, and in Uganda for the selection of
non-principal couple respondents. Both South Africa and Uganda implemented the
method through programs embedded in the CAPI platform. In Maldives, enumera-
tors were asked to list all household members by gender and age, and a respondent was
selected randomly following a randomly assigned Kish table for intrahousehold selec-
tion. In Georgia, Mongolia and the Philippines, the nearest birthday method was used
125 For a comparison of different for the selection of the non-principal couple respondent.125
methods of within-household 315. Either the Kish procedure or the computer-assisted random selection pro-
respondent selection, see
Cecilie Gaziano, “Comparative
cedure would yield approximately the same number of women and men in the sample,
analysis of within-household before non-response occurs. This gender balance is extremely important for the meth-
respondent selection odology proposed in these Guidelines. In some countries, however, asset ownership
techniques”, Public Opinion among women may occur much less frequently than among men. Having the same
Quarterly, vol. 69, No. 1 (spring number of women and men in the sample can be problematic for the estimation of key
2005), pp. 124–157. measures of asset ownership in countries with large differences between women’s and
men’s asset ownership; both the asset ownership prevalence and the estimates on asset
value for women will tend to have lower precision than that for men.
316. If there is prior information on the prevalence of ownership indicating that
women own fewer assets, then national statistical agencies may choose to oversample
women to improve the precision of estimates of both women’s ownership of assets and
126 Oversampling can be the value of those assets.126
accomplished by assigning a 317. Using oversampling by gender as a means of selecting persons within
higher probability of selec-
tion for women than for men
households will also be easier with computer-assisted systems. Sampling rates can be
when selecting one person set by gender in the selection program, to give women a higher chance of selection
per household. Such selection than men. It is also possible in the Kish selection method to select women at a higher
is more straightforward with rate than men, but this is more difficult than with computer-assisted selection.
the use of a computer-assisted
random selection procedure. It 4.2.1.2. Avoiding contamination when interviewing more
cannot be accomplished when than one adult member within the household
all members of the household
are selected, because the 318. The operational challenge of interviewing more than one adult member in
gender balance of the sample the household is related to the organization of the fieldwork. As discussed in chapter
is determined by the gender I, when multiple interviews are to be conducted within the same household, field pro-
distribution within the popula-
tion.
tocols should ensure, first, that respondents are interviewed alone—in other words, no
other adult member is present during the interview, which eliminates the impact of the
presence of another adult member on the respondent’s answers to the questionnaire;
and, second, that respondents within the same household not be given the chance to
discuss the content of the interview before their interviews are completed.
319. These two requirements are not easy to implement in the field if there is
only one enumerator present in the household, unless there are only two people to
be interviewed. In this case, the enumerator may conduct the two interviews inde-
 Guidance for implementation 89

pendently and consecutively. If, however, there are more than two respondents in the
household, two enumerators would be needed, to ensure that the two above require-
ments are satisfied. For example, if there are two enumerators and four respondents,
two respondents can be interviewed first, and the remaining two respondents inter-
viewed after. This would preclude any discussion among respondents before their
interviews are completed.
320. Additional complications are caused by the likelihood that not all respond-
ents will be available to be interviewed consecutively. In the field, a certain number of
days are usually allocated for each enumeration area. Decisions then need to be made
as to whether interviews for some households are to be completed in multiple visits,
acknowledging that there might be some contamination in the responses as discus-
sions among respondents before all interviews cannot be avoided.
321. In principle, when more than one household member must be interviewed,
the members should be interviewed simultaneously to avoid potential contamination
of responses that may occur when household members exchange information with
one another about the survey questionnaire. At the same time, while a contamination
effect has long been surmised for opinion-based or attitudinal surveys, it is difficult to
quantify this source of measurement error. In addition, in practice, as demonstrated
by the EDGE pilots, simultaneous interviewing is difficult to achieve (see box 7 for
further details on this issue).

Box 7
Challenges in conducting simultaneous interviews
In the Uganda pilot, up to four adult members per householda were interviewed in a The principal couple was
treatment arms 4 and 5,b and the field protocol required that the respondents be inter- selected with a probability
viewed alone and that the interviews be carried out simultaneously. Non-simultaneous of 1, and two additional
interviews were allowed only when the enumerator had confirmed that not all eligible persons were selected
randomly from the
household members would be available for simultaneous interviews within the allocated
remaining adult household
time for enumeration in the enumeration area. A similar approach was adopted for the members.
pilots in Georgia, Mongolia and the Philippines, where three adult members per house- b See box 3 for more
hold were interviewed.
information about
In all four pilots, a team of enumerators was dispatched to the field, including an expe- treatment arms 4 and 5 in
rienced supervisor who oversaw the assignment of enumerators to different households the Uganda pilot survey.
in the enumeration area. Multiple visits were made to households, so that all eligible
members could be interviewed simultaneously. Despite these efforts, it proved difficult
to achieve simultaneity in all the EDGE pilots. In most cases, it was not possible to achieve
simultaneity for all households in the sample. In addition, as shown in the following table,
the more household members there were to interview simultaneously, the less likely sim-
ultaneity was achieved. For example, in the Georgia pilot, among all two-adult house-
holds that were interviewed, simultaneity was achieved in 71 per cent of the households.
The percentage was lower for households where three adult members were interviewed
(57 per cent). Similarly, for the EDGE pilots in Mongolia, the Philippines and Uganda, the
percentage of two-adult households that achieved simultaneous interviews ranged
between 38 and 57 per cent. For three-adult households, the percentage of households
with simultaneous interviews was lower, in the range of 20 to 30  per  cent. In Uganda,
where four adult members of a household were to be interviewed, enumerators were able
to conduct simultaneous interviews in only 8 per cent of selected households.
90

Percentage of households in which all eligible respondents were interviewed and interviewed simultaneously,
by size of household
Uganda
Georgia Mongolia Philippines Arm 4 Arm 5
Number of two-adult households interviewed 926 1 285 622 237 248
Proportion of all eligible adults interviewed 84% 74% 89% 58% 54%
Proportion of all eligible adults interviewed simultaneously 71% 42% 57% 47% 38%
Number of three-adult households interviewed 1 399 1 341 790 54 58
Proportion of all eligible adults interviewed 75% 39% 76% 37% 40%
Proportion of all eligible adults interviewed simultaneously 57% 27% 32% 22% 26%
Number of four and more-adult households interviewed 60 60
N/A (a maximum of three adult members were interviewed
Proportion of all eligible adults interviewed 23% 25%
in these countries)
Proportion of all eligible adults interviewed simultaneously 8% 8%

An additional challenge encountered by the EDGE pilot studies when testing simulta-
neous interviewing was a lack of space in the household when multiple interviews needed
to be conducted. This was a problem in particular in urban areas, where it was often dif-
ficult to find separate rooms where interviews would not interfere with each other.

322. Given the difficulties in achieving simultaneous interviews, along with

the higher costs associated with assigning additional enumerators per household
to conduct simultaneous interviews, these Guidelines do not include simultaneous
interviewing among the recommended field protocols. Countries should, however,
be aware of the potential bias introduced as a result of forfeiting simultaneous mul-
tiple interviews in the same household. On the other hand, from the experience of
EDGE pilot surveys, independent interviews seem to be feasible. In Uganda and South
Africa, the proportion of individual interviews conducted alone was very high across
127 Talip Kilic and Heather Moylan, all modules, at 90 per cent or above, for both women and men.127 In the EDGE pilots
Methodological Experiment in Georgia, Mongolia and the Philippines, the corresponding proportions were 92, 96
on Measuring Asset Owner- and 95 per cent, respectively. The number of callbacks required to achieve such a high
ship from a Gender Perspec- level of independent interviews varied among those countries, ranging from an aver-
tive (MEXA): Technical Report
age of less than 1 in Georgia and the Philippines, to an average of 2 in South Africa.
(Washington, D.C., World Bank,
2016). 323. Despite these high success rates, qualitative assessments of the fieldwork
in the pilot countries revealed cases in which household members, usually men, inter-
rupted interviews, demanded to listen in, answered some of the questions themselves,
or amended the answers provided by women respondents to what they considered to
be correct answers. Thus, enumerators should be carefully trained in how to tactfully
handle such scenarios, so as to ensure that respondents can indeed be interviewed
alone.
4.2.1.3. Reporting discrepancies within households
324. Interviewing multiple persons within a household also poses challenges at
the analysis stage. Reporting discrepancies exist when more than one person is inter-
viewed within the same household. For example, in the Philippines pilot survey, the
spouses or partners of 26 per cent of men respondents who claimed exclusive own-
ership of the dwelling also claimed ownership. Similarly, for 37  per  cent of women
respondents in Mongolia who claimed exclusive ownership of their dwelling, their
partners also claimed ownership (table 3).
 Guidance for implementation 91

Table 3
Overlap between couples on exclusive dwelling ownership status, by gender
of respondent, Mongolia and the Philippines (percentage)
Respondent’s self-reported status
Mongolia Philippines
Respondent spouse’s
self-reported status Exclusive owner Exclusive owner
Men Women Men Women
Owner (exclusive or joint) 16% 37% 26% 23%

Exclusive owner 2% 13% 0% 1%

Joint owner 14% 23% 26% 22%

Not owner 84% 64% 74% 77% Source: EDGE pilots, Mongolia
and the Philippines, self-reported
Number of observations 874 122 120 40
data.

325. While discrepancies among respondents are an interesting area of analysis

in their own right, the discrepancies must be reconciled when deriving individual-
level indicators on forms of ownership (exclusive or joint) and the gender wealth gap
and asset-based indicators.128 Given that there is usually no external source to verify 128 More details on generating
the true owner of a given asset and the associated rights, and that certain dimen- weights for asset-level analysis
sions of the bundle of ownership rights are based on self-perception, reconciliation can be found in chapter IV,
of reporting discrepancies has proved difficult. Examples of different reconciliation section 1.4.
methods can be found in chapter IV.

4.2.1.4. Implications of operational challenges on the within-household individual selection

326. Operational challenges discussed above have implications as to who should
be interviewed within households. The difficulty of implementing random-selection
procedures suggests that countries might wish to avoid the random-selection proce-
dure when the interview is carried out using the paper-and-pencil approach. This, in
turn, means that all adult household members would be selected for interview to avoid
random selection. Within-household selection is much more flexible when the inter-
view is conducted through the CAPI approach, as the random selection can be inte-
grated in the CAPI program.
327. If the sample design requires that women have a higher selection probabil-
ity than men because women have lower prevalence of owning a key asset than men,
then interviewing all adult household members should be avoided as the selection
probability of women and men is predefined by the gender structure of the popula-
tion. If a certain number of adult members are randomly selected from the household,
it is possible to integrate the unequal selection probability by gender into the CAPI
program. The use of paper-and-pencil approach is discouraged, however, as it will be
too complicated to implement the unequal individual selection probability within the
household using this mode of data collection.
328. To avoid contamination as far as possible, the number of respondents
within households should be minimized, under the required precision level for key
estimates. This principle would also help to offset the challenge of reconciling discrep-
ancies in the post-data collection phase. Of course, the ultimate decision on how many
people to interview within a household should always be guided by the objective of
the data collection exercise, overall available budget and the precision required of the
estimates.
92

4.2.2. Cost considerations

329. A comparison of the cost of different selection protocols can be carried out
129 Kish, Survey Sampling, 1965. using the following cost function:129
Total cost = ACa + abCb + nc
where a is the number of clusters, Ca is the cost per cluster, b is the average number
of households within each cluster and Cb is the cost associated with household-level
activities, such as contacting households for interviews and completing the household
questionnaire. The total number of individual respondents in a given household is
denoted by n, and c is the average cost per individual interview. In the proposed sample
designs, n varies from one to the number of all members of the household (see figure 3).
330. As shown in the above formula, for a stand-alone household survey, the
overall cost of a sample design is defined by the cost structure at cluster, household
and individual levels, and also by the number of clusters, households and individual
respondents. If data on asset ownership are collected by adding questions to an exist-
ing household survey, then the resulting additional cost would only be a function of
the total number of interviews on asset ownership (nc). Section 4.2.2.1 discusses the
cost of different intrahousehold selection methods (as set out in figure 3), assuming
that the same level of precision of estimates is to be achieved.

4.2.2.1. Impact of intracluster correlation

331. In a typical multistage cluster sample design, the smaller the cluster size,
the higher the precision achieved if the same number of respondents are interviewed.
As discussed earlier under cluster sampling (section 4.1.4), the effects of clustering are
measured by the design effect, which expresses how much larger the sampling vari-
ance for the cluster sample is compared to that of a simple random sample of the same
size. The design effect is generated by two factors, the intracluster homogeneity meas-
urement (roh) and the number of individuals interviewed within the cluster. With the
selection of the same number of households within a cluster, the more individuals
are selected from each household and the higher the number of respondents from the
same cluster. This, in turn, results in a higher design effect and a lower precision of
estimates.
332. To illustrate how the design effect increases with the increase of respond-
ents within a household, table 4 compares the design effect for two within-household
130 This was the design adopted designs: the first has up to three adult members from each household130 and the second
in the Georgia pilot, although has one adult member randomly selected from each household. It is assumed that 20
larger households with at least households were sampled from each cluster and, as an example, information on the
three adult members were distribution of households by the number of adults in the household was obtained
oversampled, so that, in each
from the Georgia integrated household survey in 2013.
cluster, there were 10 house-
holds with one or two adult 333. The two within-household selection designs generate different levels of
members and 10 households
design effect. When the roh value is 0.05, the design effect is 3.3 for design (a), when up
with three or more adult
members. to three respondents are interviewed from each household, and 2 for design (b), when
one person is randomly selected from each household. This means that, compared
with a simple random sample design with the same overall number of respondents,
the sampling error is 3.3 times as high for design (a) and 2 times as high for design
(b). When the value of roh increases to 0.1, the sampling error is 5.7 times as high for
design (a) and 2.9 times as high for design (b), compared to the sampling error for a
simple random sample design.
 Guidance for implementation 93

Table 4
Design effects for (a) up to three adult members from each household;
and (b) one adult member from each household
Households by number
Number of adults of adult members
in household: 1 2 3 4 5+ Total
Distribution of households in
each cluster by size (Georgia inte- 3 6 5 4 2 20
grated household survey, 2013)

Design effect Effective sample size

Number of respondents
Designs in each cluster roh= 0.05 roh= 0.1 roh= 0.05 roh= 0.1
(a) Up to three adult members
3 12 15 12 6 48 3.3 5.7 14 8
from each household
(b) One adult member
3 6 5 4 2 20 2.0 2.9 10 7
from each household

Design effect (deff) = 1 + (number of respondents in each cluster – 1) × roh

Effective sample size = number of respondents in each cluster/design effect

334. One can also interpret the design effect through what is termed the “effec-
tive sample size”, which is calculated as the total number of respondents divided by
the design effect. The effective sample size of a particular sample design is interpreted
as the sample size under a simple random sample design that would yield the same
sampling variance as that achieved by the actual design. For instance, in the above
example under design (a), with a roh value of 0.05, the sample size required in the case
of a simple random sample—the effective sample size—is 14 individuals. This is con-
trasted to the more complex sample design and its effects, where 48 respondents would
be required in the cluster to achieve the same precision. Similarly, under design (b),
the effective sample size is 10, while 20 respondents would be needed in the cluster to
achieve the same precision, owing to the effect of the adopted sample design.
335. The calculation illustrates that design (a), in which more adult members are
interviewed in each household, has an advantage over design (b) in reducing the num-
ber of clusters required to achieve the same precision. The magnitude of the advantage
varies by the size of the intracluster correlation roh. Yet this advantage does not nec-
essarily translate into a reduction in cost, as more individual interviews are required
under design (a) than under design (b). For example, with the overall required effective
sample size 140 and a roh value of 0.05, a total of 10 clusters are needed for design (a)
and 14 clusters for design (b). The total number of interviews required is 480 (48 × 10)
for design (a) and 280 (20 × 14) for design (b). Design (a) requires more interviews but
fewer clusters, while design (b) requires fewer interviews but more clusters. The overall
cost ultimately depends on the cost ratio of reaching a cluster and conducting an indi-
vidual interview. A simulation of the cost calculation, taking into consideration the
design effect, may be found in table 6.

4.2.2.2. Impact of unequal intrahousehold selection probability

336. As discussed above, interviewing one adult member from each household
creates a lower design effect than when more than one person is selected for interview-
ing. Operationally, interviewing one person selected at random also helps in avoiding
data contamination (see section 4.2.1). It is important to note, however, that selecting
one person at random from each household creates an unequal selection probability
at the household level, which is usually offset through the use of weights. This in turn
94

will increase the variance of the estimates produced. Calculation of such effect, usually
referred to as the weighting effect (1 + L), is illustrated in table 5, using data on the dis-
tribution of households by the number of adult members from Georgia and Mongolia.

Table 5
Weighting effect (1 + L) due to unequal selection probability within households

Size of households (h)

Weighting
1 adult 2 adults 3 adults 4 adults 5+ adults effect (1 + L)
Source: Georgia integrated Distribution of households
household survey 2013. Methods 9 15 22 26 28 —
by size (Wh)
for calculating the weighting ef- Selection probability,
fect available in Kish, Survey Sam- 1 0.5 0.33 0.25 0.2 1.26
1 person per household (kh)
pling, and Leslie Kish, Questions/
Selection probability,
Answers (1978–1994) from the Sur- 1 1 1 0.75 0.6 1.05
3 persons per household (kh)
vey Statistician (Paris, International
Association of Survey Statisticians, Wh
Weighting effect (1 + L) = ∑h × ∑h Wh kh
1995), number 17.1. kh

337. As shown in table 5, selecting one person from each household generated a
weighting effect in the magnitude of 1.3 for both Georgia and Mongolia. If up to three
adult members are selected, as in the case of the pilot surveys in those two countries,
the weighting effect due to unequal selection probability is only 1.05, meaning that the
sampling error resulting from the current sample design is just slightly higher than
that obtained using simple random sampling, given the same number of respondents
in both designs. It should be noted that the magnitude of the weighting effect due
to unequal selection probability depends only on the distribution of households by
the number of adults in the household. In a society that has smaller households, the
131 A large proportion weighting effect will be correspondingly smaller.131
(73 per cent) of households in 338. The increase in sampling error due to unequal selection probability within
the United States of America
households when one person is selected at random from the household indicates that
had two adults in 1957, and the
calculated design effect was the total number of respondents should be increased to achieve the necessary precision
around 1.1 (see Kish, Survey (see table 6).
Sampling, 1965).
4.2.3. Making decisions on individual respondent selection
339. While considering the advantages and limitations of different selection
methods within households (see figure 3), national statistical agencies need to keep in
mind the ultimate objectives of the survey. The choice of method should also depend
on how the information on asset ownership from a gender perspective will be col-
lected, whether through a stand-alone survey, an attached module, or a minimum set
of questions added to an existing household survey.

4.2.3.1. Objective 1: Ownership prevalence and gender wealth gap

340. As discussed in section 4.2.2, the overall cost of a data collection varies in
terms of the different within-household respondent selections; whether data are col-
lected through a stand-alone survey or through appending to an existing household
survey; and a number of key factors, including the prevalence of asset ownership in
the population; the required precision (coefficient of variation); the intracluster vari-
ation (roh value); the distribution of households by number of adult members in the
household; and the non-response rate in the survey.
341. Table 6 provides a simulation of the calculation of the sample size and
related costs for two major within-household respondent selection protocols proposed
 Guidance for implementation 95

in the Guidelines—selecting all adult members (assuming 2.5 adult members per
household on average)132 and selecting one adult member at random. For both selec- 132 The averages are 2.4 for Mon-
tion protocols, and to illustrate the simulation, the following assumptions are made: golia, and 2.8 for Georgia.
the value of roh is assumed to be 0.1;133 two different levels of prevalence of asset own- 133 Estimates of roh range
ership are used, 0.1 and 0.3; the coefficient of variation is set at 0.15; the individual between 0.02 and 0.3, with the
non-response rate is assumed to be 20 per cent; and the number of households in a lowest value in Cavite province,
cluster is fixed at 20. Philippines, and the highest in
Georgia.

342. As shown in the table below, the higher the roh value, the more respond-
ents are required to achieve the same precision. The more respondents required, the
smaller the prevalence of asset ownership. The calculation takes into consideration the
increase in sample error introduced by the intrahousehold correlation—or “deff 1”, the
design effect expressed as a function of roh, and also by the unequal within-household
selection probability—or “deff 2”, the weighting effect.

343. In general, if all adult members in each household are interviewed, some
3,400 women and men need to be approached and some 1,400 households are required
if the ownership prevalence is at 10 per cent, the intracluster correlation is 0.1 and the
coefficient of variation is 0.15. Under the same assumptions, some 2,500 households
and individuals are required if one person is randomly selected from each household.
Accordingly, interviewing one person from each household requires interviewing
more households but fewer total individuals, compared to interviewing all adult mem-
bers in each sampled household.

344. The corresponding cost implications vary, depending on whether data are
collected through a stand-along survey or through a module attached (or questions
added) to an existing survey. For a stand-alone survey, the cost is much higher if only
one randomly selected person is interviewed from each household. For that reason,
these Guidelines do not recommend this respondent selection approach for stand-
alone surveys.

345. If, however, the overall objective of collecting data on asset ownership is to
produce statistics on asset ownership prevalence and the related gender wealth gap, the
set of questions proposed in these Guidelines are attached to an existing survey that
interviews a sufficient number of households,134 and the marginal cost of the data collec- 134 Totalling 2,500 households in
tion exercise will depend primarily on the number of interviews. In this case, interview- the simulation (table 6), if the
ing one person randomly from each household costs less than interviewing all adult prevalence of ownership is
members and this strategy should therefore be adopted. If, on the other hand, the host 10 per cent.
survey does not have a sufficiently large number of households, an alternative strategy
is to interview more than one person in the selected households for the main survey, to
ensure a number of respondents sufficient to derive representative estimates. A similar
calculation can be performed (as in table 6), based on assumptions applied to national
circumstances such as the average household size and prevalence of asset ownership.

346. Caution should be exercised when considering the cost implication of vari-
ous within-household selection protocols. The calculation of cost in table 6 is based on
one specific cost structure at the cluster, household and individual levels. A very dif-
ferent overall cost might be obtained should the cost structure change. In addition, the
additional cost due to callbacks and repeated visits to households was not featured in 135 The average number of call-
the calculation.135 If a significant number of callbacks is required to capture individual backs ranges between 0.5 to 2
respondents, then the cost of those repeated visits would further drive up the total cost. in EDGE pilot countries.
96

Table 6
136 Cost of the data collection is Required sample sizes and cost calculation for prevalence rate estimate136
calculated on the basis of the
following assumption: $2,500 Selecting all adult members Selecting one person at
to reach an enumeration in the household random in the household
area (Ca), $80 to reach a house- Percentage owning assets (p) Percentage owning assets (p)
hold (Cb) and $20 for each 10 30 10 30
interview (C). The calculation is
for illustrative purposes only; 1. Number of women in the cluster (b) 25 25 10 10
the value for the overall cost 2. Design effect due to intracluster
3.4 3.4 1.9 1.9
will change if different assump- correlation (deff = 1 + (b - 1) × roh)
tions are made. 3. Weighting effect due to weighting
for unequal selection probability
1.0 1.0 1.3 1.3
within household (1 + L, calculation
in table 5)
4. Number of women required taking
into consideration design effect and 1 360 353 988 256
weighting effect
Note: The following assump-
5. Number of women required taking
tions made for the simulation
into consideration the non-response 1 700 441 1 235 320
in this table: 2.5 adult members rate (at 20 per cent)
per ­household; roh = 0.1; assets
6. Total number of households required 1 360 353 2 470 640
prevalence = 10% and 30%;
coefficient of variation (cv) = 0.15; 7. Number of clusters 68 18 124 32
­non-response rate = 20% and 8. Total cost (stand-alone) $333 200 $86 385 $545 870 $141 522
cluster size = 20 households. 9. Cost after reaching the household $54 400 $14 104 $39 520 $10 246

Row 1: b = 20 households in the cluster and 2.5 adult members (1.25 women and 1.25 men on average) per house-
hold. The total number of adult women in the cluster would therefore be 20 x 1.25 = 25, if all adults are selected. If
only one adult member is selected at random, then on average there will be 10 women in the cluster.
Row 2: design effect (deff) = 1 + (b – 1) x roh
Row 3: weighting effect (1 + L), calculated as in table 5
s2 p × (1 – p)
Row 4 = × deff x (1 + L) = × deff x (1 + L), the value for deff and (1 + L) is in rows 2 and 3, respectively.
p2 × cv2 p2 × 0.152
row 4 row 4
Row 5 = =
(1 – non-response rate) (1 – 0.2)
row 5
Row 6 = , where 1.25 is the average number of adult women in each household. Note that 1.25 is calculated
1.25
on the basis of the assumption that there are on average 2.5 adult members in the household and half of them are
women.
row 6 row 6
Row 7 = = , based on the assumption that there are 20 households per cluster
No. of households per cluster 20
Row 8 = No. of clusters × Ca × No. of households × No. of individual interviews × C = row 7 × Ca + row 6 × Cb + row 4 ×
2 × C. Note that this is the total cost of interviewing all women and men. The costing is based on the a­ ssumption
of $2,500 reaching an enumeration area (Ca), $80 reaching a household (Cb) and $20 for each interview (C). The
­calculation is for illustrative purpose only; the value for the overall cost will change if different assumptions are made.
Row 9 = No. of individual interviews × C = row 4 × 2 × C. It is assumed that the cost of interview is $20.

4.2.3.2. O
 bjective 2: Intrahousehold analysis dynamics
of ownership and decision-making
347. If the objective of the survey also includes analysing the intrahousehold
dynamics of ownership and decision-making, multiple persons from each household
must be selected. If the dynamics between spouses or partners are of interest, then
a sufficient number of couples are needed for the analysis. The following table illus-
trates how the required number of couples can be calculated for a set of outcome vari-
ables related to couples. Examples of these types of variables include the proportion
 Guidance for implementation 97

of couples that both own assets or the share of women’s wealth among the couple’s
total wealth (more examples can be found in chap. IV, sect. 3, on data analysis). It is
assumed that the required coefficient of variation for a given estimate is 0.15 and that
20 households are selected from each cluster. The calculation is carried out for a num-
ber of scenarios, with the intracluster correlation, roh, fixed at a level of 0.1, and the
values for the outcome variables set at 0.1 and 0.3, respectively.

Table 7
Required sample sizes for intrahousehold gender analysis

Example of intrahousehold
gender analysis: percentage of
couples that both own assets (y)
10 30
1. Number of couples within each cluster (b), assuming
50% of households have a couple) 10 10
2. Design effect due to intracluster correlation (deff = 1 + (b - 1) x roh) 1.9 1.9
3. W
 eighting effect due to weighting for unequal selection
probability within household (1 + L) 1 1
4. N
 umber of couples required under the required coefficient
of variation (cv=0.15), under the simple random sample design 400 104 Note: The following assumptions
5. Total number of couples taking into consideration the were made for the simulation
design effect and weighting effect 760 197 in this table: 50% of households
6. T otal number of couples taking into consideration have a couple; roh = 0.1; coeffi-
non-response (at 36% per couple) 1 188 308 cient of variation (cv) = 0.15; non-
response rate = 20% and cluster
7. Total number of households required 2 375 616
size = 20 households.
Row 1: Number of couples in each cluster b = 20 x 50% = 10
Row 2: Design effect deff = 1 + (b – 1) x roh = 1 + (10 – 1) x 0.1 = 1.9
Row 3: Weighting effect (1 + L) = 1 as the selection probability of a couple from each household is 1.

s2 y × (1 – y)
Row 4 = = , this is the number of couples required under the simple random sample design.
y2 × cv2 y2 × 0.152
Row 5 = row 4 × deff × (1 + L) , where the value for deff and (1 + L) is in rows 2 and 3, respectively.

row 5 row 5
Row 6 = =
(1 – non-response rate) (1 – 0.36)
Row 7 = row 6 x 2, assuming that only half of the households have a couple.

348. The calculation differs from that in table 6 in two respects. First, not every 137 Eurostat database, Private

household has a couple. In the calculation, it is assumed that 50 per cent of households households by type, tenure
include a couple among their members. This assumption in the above simulation is status and NUTS 2 region,
based on a calculation made for 31 European countries, in which the percentage of based on 2011 population
households with a couple ranges between 43 and 67.137 The second way in which the censuses (extracted in August
calculation differs from earlier calculations is in the area of non-response probability. 2017). The percentage is much
lower in South Africa according
Not only is it necessary to capture a sufficient number of households with couples, but to its 2011 census data, which
both members of the couple need to be available and to answer the questions to enable indicated that only 38 per cent
couple-based analysis. Given the assumption in table 6 that the individual response of households had at least one
rate is 80 per cent, the response rate used in the calculation for both members of the couple. Minnesota Population
couple is 64 per cent (or 0.8 x 0.8). Center, Integrated Public Use
Microdata Series, International,
349. Again, as shown in the earlier calculation for the prevalence of asset own- data set version 6.5 (Minne-
ership, the required number of couples is higher when only 10 per cent of couples both apolis, University of Minnesota,
own assets, compared to when the proportion is 30 per cent. A comparison of tables 7 2017). Available at https://doi.
and 8 shows that, for the same estimated value of the outcome variable, the intracouple org/10.18128/D020.V6.5.
98

analysis requires more households than the design where all household members are
interviewed for a given prevalence.
350. The example here assumes that one couple is selected from sampled house-
holds, if available. It is unlikely, however, that countries would only be interested in
the analysis among couples. To derive both the nationally representative ownership
prevalence and to conduct an intracouple analysis, it is advised that the respondents
include a couple and an additional adult randomly drawn from selected households. If
there is a sufficient number of households in the survey, this within-household selec-
tion protocol keeps the maximum number of interviews to three for each household,
making it easier for the supervisors of field operations to manage (see section 4.2.1 for
further details on the operational challenges arising when multiple interviews are to
be conducted within households).
351. Another option for attaining both objectives—deriving the prevalence rate
and conducting an intrahousehold analysis—is randomly selecting one adult mem-
ber from the sample household and that member’s spouse or partner, if she or he is
married or in a union. This approach was tested in the South Africa EDGE pilot. It
ensures that there are no more than two interviews per household and, even with only
one enumerator, the interviews can be conducted with minimal potential contamina-
tion (see section 4.2.1 above). One disadvantage of this approach is that it requires
an increase in the number of sample households, because a certain percentage of the
population living in households with couples do not have a spouse. For example, they
may be the adult child of a couple or a parent-in-law living in the household with a
couple. According to the data from the 2011 South Africa census, among people 18
138 Minnesota Population Center, years or older living in a household with couples, 25 per cent do not have a partner.138
Integrated Public Use Micro- When one of those individuals is selected for interview, no partner will be available. To
data Series, International, data capture the same number of couples as when interviewing one couple and a third indi-
set version 6.5, 2017. vidual from each household, this approach requires a greater number of households.

4.2.3.2. Putting the puzzle together

352. Making decisions on how within-household respondents should be
selected requires striking a balance between data-collection costs and field-operation
feasibility, while keeping sampling and non-sampling variance to an acceptable level.
To that end, the following diagram aims to guide countries on respondent selection
within households for data collection on asset ownership from a gender perspective.
353. The illustration is based on a coefficient of variation of 0.15. It is assumed
that 20 households are selected from each cluster. The interest variable, both at the
individual level (e.g., the asset ownership prevalence) and at the couple level (e.g., the
share of the couple’s total wealth held by the partner who is a woman or the propor-
tion of couples who both own assets) is set at 10 per cent. The intracluster variation is
assumed to be 0.1.
354. In the illustration, the decision on whom to interview is made on the basis
of the calculation in tables 7 and 8, as well as other practical considerations, as out-
lined in section 4.2.1 above. For example, if the ownership prevalence rate is of interest
and the host survey has a large enough number of households (2,500 households based
on the assumptions made for the simulations described above), then interviewing one
random adult member from each household is a sound choice. The design not only
helps in achieving the required coefficient of variation of 0.15, but also eliminates the
possibility of contamination and other field operation problems. Of course, if the host
survey is small (fewer than 2,500 households), it is recommended that several people
be interviewed, in order to achieve the required precision.
 Guidance for implementation 99

Figure 4
Decision tree for intrahousehold respondent selection

Are you interested in the prevalence of asset ownership or in

both ownership prevalence and intrahousehold dynamics?

Prevalence of asset ownership only Both ownership prevalence and intrahousehold dynamics

Append to an existing Append to an existing

Stand-alone survey
household survey household survey

How many households Interview all adult How many households

does the host survey members from each does the host survey
cover? household cover?

More than
More than Fewer More than 2,500 Fewer
2,500 than 2,500 3,500 households, than 2,500
households households households but fewer than households
3,500

Interview one
Interview more person Interview one
Interview one
than one randomly couple and
adult member
randomly selected from one ­additional Reassess
randomly
selected adult each house- person survey
selected
member hold, plus randomly objectives
from each
from each the spouse or ­selected from
household
household partner, the household
if available
100

355. From the calculation in table 7, the required number of households is also
2,500 households for a variable of 10 per cent and roh of 0.1. Accordingly, if a host
survey is larger than 2,500 households, interviewing a couple and a third randomly
selected person from these households would provide reliable estimates for both the
asset ownership prevalence and the intrahousehold analysis. Alternatively, when
139 The requirement of 3,500 the host survey covers more than 3,500 households,139 an adult member randomly
households is calculated selected from each household and that member’s partner should produce estimates
based on an assumption that meet both objectives: asset ownership prevalence and intrahousehold analysis.
that 75 per cent of the adult This selection method has an operational advantage—a maximum of two interviews
members living in coupled
are required within each household. This helps in reducing contamination bias and
households have a partner.
This percentage holds for also makes it easier for the survey team to assign enumerators.
KwaZulu-Natal, South Africa, 356. It should be noted that the guiding sample sizes in the diagram above are
but may change for other for illustrative purposes only. Countries under different circumstances are advised to
countries. It is suggested that calculate the optimal sample sizes on the basis of the different scenarios applicable to
countries should adjust the
calculation when planning
their national context.140
the data collection exercise in 357. When the individual-level asset ownership data are collected through
accordance with their national appending a minimum set of questions or a module to an existing survey, another
circumstances. consideration that is not featured into the diagram above is the respondent selec-
140 See https://unstats.un.org/
tion and interview protocol of the host survey. Existing surveys vary greatly in these
edge for a worksheet on cal- dimensions, and sometimes vary by modules within the same survey. While consider-
culating required sample sizes
ing the suggested respondent selection protocols illustrated in the diagram, countries
with user input on different
parameters. should also be flexible in making adjustments to adapt to the circumstances specific to
the host survey. For example, the diagram shows that, if a country is interested in both
the national estimates of the asset ownership prevalence and intrahousehold dynam-
ics and the host survey is large enough, then a randomly selected adult member and
that member’s partner should be interviewed. If, however, the host survey is already
collecting self-reported data from all adult members of the household, then the EDGE
module or questions could be put to all adult members within the household. In this
case, fewer households would need to be interviewed for the EDGE module and ques-
tions if there is need to control for the additional data collection cost.

Key points
•• If a stand-alone survey is conducted to collect data on asset ownership, the sample
selection process up to the household level is similar to any other household surveys.
At the same time, however, the following elements should be taken into consideration:
•• It is recommended that the individuals interviewed be limited to those aged 18
and older;
•• The two essential population subgroups in the sample for gender analysis of asset
ownership are women and men. If there is prior information on the prevalence of
ownership indicating that women own fewer assets, then national statistical agen-
cies may choose to oversample women to improve the precision of estimates of
both women’s ownership of assets and the value of those assets.
•• Regions that have different marital regimes and land tenure systems should be placed
in different strata. Dividing populations into urban and rural is also advised, since the
ownership of agricultural land, agricultural equipment and livestock would be very
different for people living in urban and rural areas.
•• The number of individuals to be interviewed in the selected households is deter-
mined by the following factors:
 Guidance for implementation 101

•• Survey objectives: whether the focus of the data collection is only to generate
asset ownership prevalence rates, by gender, or also to study intrahousehold
gender dynamics in asset ownership;
•• Data collection instrument: whether asset ownership will be collected through
a stand-alone survey or through a module or set of questions appended to an
existing survey. In the latter scenario, the sample size of the host survey and the
related interview protocol will influence the sample selection for the study of
asset ownership;
•• Field operation feasibility: whether it is possible to collect asset ownership data
from more than one respondent in the sampled household, while keeping con-
tamination bias to a minimum.
•• Once the above factors are assessed, the basic principles in sampling within
household are:
•• If the objective includes studying intrahousehold dynamics of asset ownership,
more than one respondent is required within a given household. If the only
objective is ownership prevalence, a stand-alone survey is not recommended
and one or more respondents from each household may be selected, depend-
ing on the sample and field protocol of the host survey and other considera-
tions (see figure 4);
•• If there is a sufficient number of households in the sample, the number of
household members to be interviewed should be kept at a minimum to ensure
the necessary precision. This would reduce the contamination bias introduced
by interviewing multiple persons in the same household;
•• If conducting a stand-along survey is an option for a country, it is recommended
to also collect information on intrahousehold dynamics of asset ownership.

5. Questionnaire design
358. In order to produce reliable measurements of women’s and men’s ownership
and control of assets, the conceptual framework presented in chapter I of these Guide-
lines must be explicitly operationalized in the questionnaire used to collect the data.

5.1. Background research

359. Although the present Guidelines provide a model for collecting data on
asset ownership and control at the individual level, including a detailed questionnaire
as presented in the next section and in the annex to these Guidelines, countries are
encouraged to conduct their own background research before adapting the proposed
generic model to their country context. In general, background research to a house-
hold survey has the role of informing the survey design and providing the context in
which to interpret the results of the survey. It can include a desk review of relevant
national quantitative and qualitative research studies, together with new qualitative
research commissioned by the statistics office or key stakeholders.
360. The desk review should cover such topics as:
•• The legal framework and customary norms that govern property rights,
including those related to marital and inheritance regimes, across different
areas of the country;
•• The link between asset ownership and other development issues, includ-
ing poverty, livelihoods, entrepreneurship, agriculture, women’s empower-
ment and gender equality;
102

•• Government programmes and policies related to key core assets, including

on housing and distribution or titling of land;
•• Existing quantitative studies providing information on the prevalence of
asset ownership, including at individual and household levels, and also
wealth distribution across different population groups;
•• Studies related to land tenure systems across the country.
361. In addition, new qualitative research may be conducted. This research may
be limited in scope and focused on adapting or improving the questionnaire design
or more comprehensive, providing a stand-alone qualitative study complementing the
statistical results obtained in the survey. Statistics offices are more likely to focus their
efforts on the first approach, owing to constraints of cost and time. Conducting com-
prehensive qualitative research also requires a set of skills more often found among staff
in research and academic institutions than in the statistics office. Accordingly, these
experts should be sought out to provide technical guidance for any qualitative research
undertaken by national statistical offices. The statistics offices should at the very least
conduct interviews with key informants and hold focus-group discussions for the pur-
pose of improving questionnaire design. These methods can explore how participants
think about asset ownership and the terms that they use in talking about them. Indi-
vidual interviews may vary in format, ranging from informal discussions, used for the
purpose of gaining a broad understanding of the issues related to asset ownership, to
structured interviews with a predetermined set of questions covering key topics related
to asset ownership, such as types and forms of ownership and acquisition of assets. It
is important that the key respondents are chosen to represent a range of viewpoints
and concerns and income levels. Similar information may be obtained in focus group
discussions. For example, in the Gender Asset Gap Project, four themes were covered
during the focus group discussions: the accumulation of assets over the individual life
cycle; the importance of assets; the market for assets; and household decision-making
over asset acquisition and use. In terms of respondents covered, it is important that
several groups are formed, each of them involving a moderator and from 6 to 10 par-
ticipants with relatively homogeneous background and experience.

5.2. Questionnaire content

362. This section provides recommendations on the content and formulation of
the questionnaire. It is important to note that countries need to carefully review the
proposed questions and decide on the final formulations based on the following steps:
agree on the principle concepts to be measured; identify key information required;
and examine how this can be translated into specific series of questions. The content of
the questionnaire should be developed in accordance with the objectives and required
final outputs of the survey. Other important considerations also include the length of
the completed interview, the mode of interviewing, the need for skip and filter ques-
tions, the importance of establishing rapport with respondents, and the wording and
ordering of questions.

5.2.1. Key information required

363. The information required to fulfil all major objectives of collecting data
on individual-level asset ownership, as described in chapter I, is summarized below:
•• List or roster of household members, established at the household level;
•• Individual sociodemographic characteristics, including age and gender,
collected at the household level;
 Guidance for implementation 103

•• Individual sociodemographic characteristics, including education, mari-

tal status, economic status and other characteristics to identify population
groups that are of policy interest (such as the agricultural population), col-
lected at household or individual level, depending on the survey design;
•• Within-household decision-making process, collected at the individual level;
•• Asset and asset characteristics: roster of assets and their characteristics,
including their value. This information can be collected at the individual
or household level, depending on the respondent selection protocol, within
the household. Additional details are available below under the respective
subsections on roster of assets and asset characteristics;
•• Individual ownership of assets—reported and documented ownership, own-
ership rights, modes of acquisition of assets, collected at the individual level.
364. The roster of household members and information on their age and gender
is necessary when respondents are selected at random within households. Character-
istics of individuals allow data users to calculate a set of gender indicators on asset
ownership and further to investigate asset ownership by key covariates for a more
nuanced understanding of who owns and controls assets. Data collected on power
and decision-making of women and men within the household enable an analysis of
the association between asset ownership and decision-making at the individual level.
365. Collecting information on who owns the asset and whether there is an
ownership document for the asset allows countries to begin to monitor both the gen-
der patterns of asset ownership and policies to improve women’s property rights. The
additional questions about asset characteristics, value, alienation rights and acquisition
make further gender analysis possible. Posing questions about rights over the assets will
make it possible to understand the extent to which rights are shared under joint owner-
ship. In addition, information on rights may indicate the extent to which the full range
of rights are correlated with ownership and the extent to which women may have some
ownership rights, but not others. The valuation and asset characteristics data allow for
the computation and analysis of gender wealth gaps. Lastly, because men and women
often acquire assets through different means, understanding the modes of acquisition
may provide insights for developing policies to ensure women’s ability to acquire them.
366. Table 8 below illustrates how, as the objectives of data collection vary from
producing estimates of asset ownership prevalence to assessing the gender wealth gap
and intrahousehold asset ownership dynamics, the content of the questionnaire varies
accordingly. It should be noted that “essential”, as used in the table, means that such
information should be collected. “Additional” information is not necessary but would
enrich the understanding of asset ownership from a gender perspective. Intrahouse-
hold decision-making information is essential only when one of the objectives of the
data collection is the intrahousehold analysis of asset ownership.
367. In cases where the only objective of data collection is to derive ownership
prevalence by gender, basic information on reported and documented ownership and
ownership rights would suffice, without the need to itemize each asset. For instance,
using the example of agricultural land, the basic question on individual-level owner-
ship would be whether or not the respondent owns (reported and documented) and
has the right to sell and bequeath any agricultural land. It would not be necessary to
ask these questions against each single parcel owned in the household. Accordingly,
information on “asset roster” and “asset characteristics” is considered not essential
in table 8. Individual characteristics other than age and gender may be essential for
the prevalence of asset ownership by gender, if asset ownership is to be calculated for
specific population groups.
104

Table 8
Questionnaire content by data collection objectives

Data collection objectives

Gender Gender asset Gender asset gap,
asset gap and gender gender wealth gap and
Information to be collected gap wealth gap intrahousehold analysis
A. Information on household and individuals
Roster of household members Essential Essential Essential
Characteristics of individuals: gender and age Essential Essential Essential
Characteristics of individuals: Additional Additional Essential
education, marital status, economic activity
Decision-making within the household Additional Additional Essential
B. Information on assets
Roster of assets Additional Essential Essential
Characteristics of assets including value Additional Essential Essential
Asset ownership (reported and documented) Essential Essential Essential
Ownership rights Essential Essential Essential
Asset acquisition Additional Essential Essential

368. Table 8 defines broadly what information is to be collected for each data
collection objective, and should be used with caution. This is because the information
to be collected on assets varies by type of asset. For example, there is only one prin-
cipal dwelling for each household and a roster of dwellings is not necessary. Certain
assets, such as livestock, are difficult to itemize; for that reason, a roster of every single
animal is not recommended, although broad categories of livestock might be listed, as
explained in the section covering the establishment of a roster of assets. Furthermore,
documented ownership applies only to the assets that have documents (dwellings,
agricultural land and other real estate). A summary of the applicability of each type of
information by asset can be seen in table 9, and a more detailed discussion is covered
in the following sections.
Table 9
Relevance of asset-related information by type of asset

Roster Characteristics Reported Documented Right to sell and/or Asset

of assets of assets ownership ownership bequeath asset acquisition Valuation
Priority assets
Principal dwelling × × × × × ×
Agricultural land × × × × × × ×
Other real estate × × × × × × ×
Financial assets × × × ×
Additional assets
Large agricultural equipment × × × × ×
Non-agricultural enterprise assets × × × ×
Liabilities × × ×
Livestock ×
Small agricultural equipment ×
Consumer durables × ×
Valuables × ×
 Guidance for implementation 105

369. The questionnaire design also varies by the method used to collect asset
ownership data—whether through appending questions to an existing survey or con-
ducting a stand-alone household survey. Certain components such as the roster of
household members and sociodemographic characteristics of individuals are basic
questions covered by all household surveys. When appending a module on asset own-
ership to an existing survey, that basic information will have already been collected
and will not need to be asked again. Additional details are provided in the subsections
below.

5.2.2. Notes on components of the questionnaire

370. This section provides detailed comments and notes that explain, for each
type of information identified above, why questions are phrased in a particular way
and how each can be altered to fit specific needs and contexts. It is important to note
that questions related to assets and asset ownership tend to vary by asset. The discus-
sion below highlights commonalities in question formulation across assets while also
covering deviations and special circumstances for particular assets. Aspects that are
very specific to certain assets are further discussed in section 5.2.3 and a model ques-
tionnaire can be found in the annex to these Guidelines.

5.2.2.1. Roster of household members

371. The roster of household members should include the name of every house-
hold member and a unique identification code should be assigned to each person. This
information is critical as the respondent chosen to complete the individual question-
naire on asset ownership will be randomly selected from the roster, as discussed in
section 4 above. Each country should define household members according to the
standards employed by the national statistical agency to ensure comparability with
censuses and other household surveys administered in-country.141 For within-house- 141 See the section on sampling

hold individual random selection, information on the age and gender of the house- in chapter III of these guide-
hold members should be collected if the Kish within-household selection method is lines for further discussion of
applied. The household roster is collected at the household level, with the information the definition of “household
members”.
provided by the most knowledgeable person.
372. If the individual-level asset ownership information is collected through
appending questions to an existing household survey, the roster of household mem-
bers should be already available from the host survey and can be used for the selection
of individual respondents.

5.2.2.2. Characteristics of individuals

373. As is standard in most household surveys, basic sociodemographic infor-
mation should also be collected for each respondent, including age, gender, education
level, employment status, marital status, relationship to household head and ethnicity.
The information obtained from these variables will allow data users to calculate a set
of gender indicators on asset ownership and further investigate asset ownership by key
covariates for a more nuanced understanding of who owns and controls assets.
374. Countries using this survey instrument to assess individual-level asset
ownership for population groups should also incorporate questions that help to iden-
tify these groups. Examples of this need can be seen in the data required to monitor
Sustainable Development Goal indicators 5.a.1 (a) on the proportion of the total agri-
cultural population with ownership or secure rights over agricultural land, by gender,
and 5.a.1 (b) on the share of women among owners or rights-bearers of agricultural
land, by type of tenure. In this case, the survey design will also need to ensure that the
106

household questionnaire permits the identification of the agricultural population, as

142 The definition of agricultural defined by the metadata for the indicators.142
population for Sustainable
375. Sociodemographic characteristics of individual respondents are collected
Development Goal indicator
5.a.1, proposed by FAO, can be by most household surveys. If the EDGE questions and modules are appended to the
found in box 1. existing household surveys, this will mean that those questions are in all likelihood
already covered by the host survey and do not need to be asked again. If, however, the
host survey does not cover certain characteristics crucial to the collection of individ-
ual-level asset ownership data, additional questions could either be incorporated in
the host survey, or in the individual questionnaire collecting asset ownership data at
the individual level.

5.2.2.3. Roster of assets

376. The roster of assets records each asset (physical and financial, including lia-
bilities) owned by individual respondents. It is used when it is possible for one person
to own multiple assets of the same type. For example, a roster of agricultural parcels
would be helpful for respondents to report on their ownership of each parcel, and also
on their rights over the parcels that they own. For financial assets, it would be useful
for respondents to list all their individual bank accounts, pensions or other types of
financial assets, where applicable. Such a roster is important when the data collection
objective moves beyond simply assessing gender gaps in the prevalence of ownership
of assets. In the case of agricultural land, by establishing a roster of assets, an assess-
ment can be made of gender differences in the quantity, quality and value of agricul-
tural land when key characteristics and also the value are collected for each parcel.
377. On the roster, the assets are usually required to be listed in a specific order.
For example, the agricultural parcel should be listed from the largest to smallest, large
agricultural equipment should be listed by year of acquisition, from the most recent
to the oldest, and financial assets should be listed from the most valuable to the least
valuable. More instructions on how to order the assets may be found in the annex to
these Guidelines.
378. The roster of assets is necessary for agricultural land, large agricultural
equipment, other real estate, financial assets and liabilities, but not for the other assets
recommended in the publication. There is only one principal dwelling for each house-
hold, so there is no need for a roster. For livestock, small agricultural equipment, con-
sumer durables and valuables, a detailed roster of assets is not recommended because
itemization is extremely tedious and items tend to have different configurations of
rights and ownership. Countries, however, may include broad categories of those
assets for respondents to provide ownership information. For livestock, for example,
there could be such broad categories as cattle, goats, sheep, pigs, horses and others.
More instructions on what to list under livestock, small agricultural equipment, con-
sumer durables and valuables may be found in the model questionnaire, in the annex
to these Guidelines.
379. A roster should also be created to collect data on non-agricultural enter-
prise assets but the angle is slightly different. Instead of building a roster of assets, the
roster is about non-agricultural enterprises. Respondents are asked to list the non-
agricultural enterprises that they own, indicating the kind of activities each enterprise
is engaged in; whether the enterprise is of limited or non-limited liability and the type
of records or accounts that the enterprises maintain. Establishing a roster in this case
serves the purpose of not only itemizing the enterprises but also screening out incor-
porated enterprises, as their assets are usually owned by enterprises rather than by
individuals.
 Guidance for implementation 107

380. The roster of assets can be completed at household or at individual level,

depending on the survey instrument used. If there are multiple respondents within the
household providing information on their asset ownership status, the roster of assets
should be built at the household level, by asking the most knowledgeable person to
provide information on assets owned by each household member. This would avoid
the need for, and eventual complication of, matching assets reported by different indi-
vidual respondents. Once collected at the household level, the roster of assets should
be fed into the individual questionnaire and questions on ownership and rights at
the individual level should be asked against each asset listed on the roster. In surveys
when only one respondent is selected from each household, the roster of assets can be
collected at household (for assets owned by the household) or individual (for assets
owned by the individual) level. However, the roster of financial assets should always
be collected at the individual level, owing to the sensitive nature of such assets and the
difficulties in getting accurate information on financial items.

5.2.2.4. Asset characteristics

381. Characteristics of assets are collected for principal dwellings and agricul-
tural land. For dwellings, information is collected on the type of dwelling, such as
whether it is detached or semi-detached, the materials used for the roof, the walls and
floor, the type of toilet available in the household, and other features. For agricultural
land, data on parcel size, availability of an irrigation system and the primary use of the
parcel during the last cropping season should be collected. Data on these character-
istics are important for a number of reasons. First, they may serve as proxies for the
difference in quality of assets owned by women and men. Second, as discussed in the
section in chapter I on establishing values, good valuation data are crucial to estab-
lishing the gender wealth gap, yet are often difficult to obtain. In the absence of good
valuation data, characteristics of assets could be used to impute value or to assess the
quality of the valuation data.
382. Characteristics of principal dwellings are well covered in many household
surveys and are often used as proxies for household economic status. If individual-
level asset ownership data are collected through appending a module or a set of ques-
tions to an existing survey, then it is not necessary to collect such information again.
Agricultural parcel characteristics should always be added next to the parcel roster,
whether collected at household or individual level.

5.2.2.5. Reported ownership

383. A number of questions are used to collect information on whether an indi-
vidual is a reported owner of an asset; if an owner, whether the asset is owned exclu-
sively or jointly; and, for joint ownership, how many other people are joint owners, and
if the spouse or partner is one of them.
384. The question on reported ownership is usually phrased as “Do you [cur-
rently] own [asset]?”143 Sometimes more specific information is provided in the ques- 143 Regarding agricultural land,

tion explaining what is covered under the asset. For instance, agricultural parcels cover for a broader measurement,
those owned by the respondent, including those that are cultivated by the household, the question may also be
and those that are currently fallow, rented out or loaned out for nothing in return, on a formulated as follows: “Do you
currently hold, have, use or
temporary basis. Examples of large agricultural equipment (such as tractors, ploughs,
occupy agricultural land?” Such
irrigation systems or trailers) and financial assets (e.g., bank account, microfinance formulation accommodates
account, informal savings account and so forth) are also provided. situations in countries where
385. Data collected through the questions on reported ownership make pos- strict ownership of agricultural
land does not exist.
sible the calculation of various indicators related to the gender asset gap. They provide
108

information on whether the respondent, and not the respondent’s household, owns
the asset. Reported ownership captures the respondent’s self-perception of his or her
ownership status, irrespective of whether his or her name is listed as an owner on an
ownership document or having the relevant rights over the asset, such as the right to
sell and bequeath.
386. Reported ownership questions relating to assets that are small and for
which the preparation of a roster is a potentially tedious process usually integrate sev-
eral major categories of the specific asset into the question’s formulation. For exam-
ple, for consumer durables, the question may be phrased as “Do you own any [broad
or major consumer durable category], exclusively or jointly?” The respondents would
answer, for each category of consumer durables, whether they are owners.
387. The question on reported ownership usually serves as a filtering question
for all the other questions. In other words, questions on documented ownership and
related ownership rights (see discussion below) will not be asked if respondents do not
report themselves as owners. For the collection of data on non-agricultural enterprise
assets, the filtering questions are slightly different. Instead of asking about ownership
of assets, three questions are asked to assess whether the respondent owns one or more
non-agricultural businesses, before the questions on ownership and rights over non-
agriculture enterprise assets are asked.
388. The question on reported ownership also measures the form of ownership,
in other words, whether the respondent owns the asset exclusively or jointly with one
or more persons, by allowing the respondent to select relevant answering categories,
such as: “Yes, alone”; “Yes, jointly with someone else”; “No, someone else is the owner”;
“Refuse to answer”. Because the benefits of ownership may differ if a person owns the
asset alone or jointly, countries are encouraged to distinguish between individual and
joint ownership. If the respondent indicates joint ownership, then a follow-up question
will be asked: “How many other people jointly own [this asset] with you, including
household members and non-household members?” Information collected through
this additional question is needed for the calculation of the gender wealth gap, as dis-
cussed in chapter IV of these Guidelines.
389. Identifying whether or not respondents jointly own assets with their
spouses or partners allows for the construction of an indicator on joint (reported)
ownership between spouses, the most common form of joint ownership for dwellings.
The question is usually phrased as follows: “Is one of these joint owners your spouse
or partner?” Other patterns of joint ownership are also possible, such as between sib-
lings or a parent and an adult child, and countries that are interested in identifying
these patterns are encouraged to ask: “Who are the joint owners, including household
members and non-household members?” The personal identification codes assigned
to household members in the household roster should be recorded for each household
member who is a joint reported owner and each non-household member who jointly
owns the asset should be assigned a standard non-household member identification
code (for example, 100).

5.2.2.6. Documented ownership

390. Questions about documented ownership are asked in respect of three
assets: principal dwellings, agricultural land and other real estate. It does not apply
to other assets covered in the Guidelines, as documents for those assets usually do not
exist. Assessing whether the respondent is a documented owner can be done through
a number of questions structured in a form similar to those relating to reported own-
ership.
 Guidance for implementation 109

391. Before asking about documented ownership, the respondent is first asked
the following question: “Is there an ownership document for [the asset]?”144 The 144 Regarding agricultural land,

response categories enable the collection of additional information on the different for a broader measurement,
types of documents that exist. There may be a range of document types that provide the question may also be
formal evidence of ownership, and national statistical offices will need to customize formulated as follows: “Is there
a formal document for any of
the response categories according to their country context. Titles and deeds are com-
the agricultural land you hold/
mon forms of ownership documents. Registration certificates document rights over have/use/occupy, issued by
property. In addition, where titling or registration is not complete, documents, includ- the land registry/cadastral
ing wills or sales receipts, provide some form of documented claim. agency?” This formulation
accommodates situations in
392. If an ownership document exists for the asset, it should be recorded, inde- countries where strict owner-
pendent of whether or not it includes the name of someone in the household. If there ship of agricultural land does
is more than one type of document, the one that is held by someone in the household not exist. Then the follow-up
should be recorded. For example, if there is a deed, but the household members do not question can be: “Is your name
have it in their possession, but instead have an invoice or sales receipt, these should be listed as an owner or holder on
recorded. any of the legally recognized
documents?”
393. The question that follows: “Are you listed as an owner on the ownership
document for this [asset]?” is then used to measure “documented ownership” of the
asset. Documented ownership refers to the existence of any document that an indi-
vidual can use, in law, to claim ownership rights over the asset, by virtue of the indi-
vidual’s name being listed as an owner in the document. Because individual names
can be listed also as witnesses in an ownership document, it is important to ask if the
respondent is specifically listed “as an owner” in the document. While countries may
ask the respondent to show the document to the enumerator so that he or she can con-
firm that the respondent’s name is listed on it, these Guidelines recommend that the
measure of documented ownership should not be conditional on the document being
checked or kept within the home.145 145 In the EDGE pilot study in

Uganda, where respondents

394. The question, along with its response categories, also measures the form of were asked to produce the
documented ownership of the asset; in other words, whether the respondent owns the ownership documentation,
asset exclusively or jointly with one or more persons. Because the benefits of owner- they were able to do so in only
ship may differ if a person owns the asset exclusively or jointly, countries are encour- 25 per cent of interviews that
aged to collect information on the form of documented ownership. reported documentation for
at least one asset. The low
395. The third question in this group, “How many other people are listed as prevalence may be due to the
owners in the ownership document, including household members and non-house- respondent’s refusal or their
hold members?” obtains information on the number of joint documented owners. inability to locate the docu-
As mentioned above for questions on reported ownership of assets, a question is also ment (Kilic and Moylan, MEXA,
asked as to whether the spouse or partner of the respondent is one of the joint owners 2016).
of the asset.

5.2.2.7. Ownership rights

396. Ownership rights refer to the right to sell and bequeath, which are impor-
tant from a gender perspective as part of the bundle of rights (more information on
this can be found in chap. I). Ownership rights are relevant for principal dwelling,
agricultural land, large agricultural equipment and other real estate.
397. There are two questions under each ownership right. The first one asks:
“Do you have the right to [sell/bequeath] this [asset]?” When respondents have the
right to sell an asset, it means that they have the right to transfer the asset to another
person or entity permanently for cash or in-kind benefits. When respondents have the
right to bequeath an asset, it means that they have the right to give the asset by oral or
written will to another person or persons upon their death.
110

398. To assess gender differences in the right to sell or bequeath the asset, it is
useful to assess whether the right to sell or bequeath can be executed alone or jointly
with others and, if the person does not have the right, whether someone else has the
right or whether the asset cannot be sold or bequeathed because of cultural or legal
rules or standards.
399. If respondents indicate that they have the right to sell or bequeath jointly
with other persons, a follow-up question is put: “Is your spouse or partner one of the
persons who jointly has the right to [sell/bequeath] the asset?”
400. Collecting information on whether the spouse or partner jointly has the
right to sell or bequeath the asset makes it possible to analyse whether joint owners
have the same rights to the asset. If countries choose to collect information on all
joint reported and documented owners, then they can ask, in place of that question:
“Which other household members also have the right to [sell/bequeath] this [asset]?”
The personal identification codes assigned to household members in the household
roster should be recorded for each household member who has the right to sell or
bequeath the asset.

5.2.2.8. Asset acquisition

401. The question on asset acquisition helps in assessing how women and men
acquire assets differently. The question is usually asked in the form: “How did you
acquire this [asset]?” and the options for respondents include “purchased”, “inherited”,
“received as a gift”, “built it” (if relevant for the asset), “allocated by government pro-
gramme” or “acquired through marriage”. The question refers to when the respondent
first came into possession of the asset and presumably began deriving economic benefit
from it. Because men and women often acquire assets through different means, under-
standing the modes of acquisition may provide insights for developing policies to ensure
that women have the ability to acquire assets. For that reason, national statistical agen-
cies should include all relevant modes of acquisition, and may want to add additional
codes, for the case where assets are received as an inheritance or as a gift, to indicate
who gave the inheritance or gift (e.g., the respondent’s natal family or the spouse’s fam-
ily). This is particularly useful for gender analyses, since the information collected can
indicate whether the asset was received from the husband’s family or the wife’s family.
402. Asset acquisition questions are relevant for principal dwellings, agricul-
tural land, large agricultural equipment and other real estate.

5.2.2.9. Value of assets

403. As discussed in chapter I, the valuation of assets is important as it reflects
a range of asset attributes. From the gender perspective, data on asset values enable to
assess the difference in the quality and quantity of assets owned by women and men.
The question on valuation is usually phrased as follows: “If this [asset] were to be sold
today, how much could be received for it?” For financial assets and liabilities, the ques-
tion is the following: “What is the current value [or remaining amount to be repaid]?”
404. It is recommended that data on the valuation of assets be collected for most
assets, including principal dwellings, agricultural land, other real estate, financial
assets and liabilities, large agricultural equipment, non-agricultural enterprise assets
and consumer durables. The full amount that would be received in the sale should
be listed, regardless of whether or not all of it would be kept by the respondent. If the
respondent is not sure how to answer, enumerators should probe on this question by
encouraging the respondent to consider the price received for similar assets sold in the
community. It should be noted that questions about the price that would be received
 Guidance for implementation 111

today refer to the current value. If there are large areas of the country with no market
for the asset, other measures may be considered. For dwellings and other real estate,
such measures could include the cost of constructing a similar asset (investigators
should specify whether the cost of the stand should be included), or the amount that
they could receive if they rented it out. Investigators may also want to use information
on the characteristics of assets so that a value can be imputed.
405. With the exception of financial assets, values for all other relevant assets
may be collected either at the household or individual level. If there are multiple
respondents within the household, values of assets should be collected at the house-
hold level. This is to avoid complications and the need for reconciliation if different
values are reported within the household for the same asset, such as a jointly owned
principal dwelling. This recommendation follows the same logic as the approach rec-
ommended in determining the person for whom a roster of assets should be collected
when there are multiple respondents within the same household.
406. Collecting information on the value of financial assets presents a great chal-
lenge because of its sensitive nature. Enumerators may think it is inappropriate to ask
for account balances of financial assets and respondents may be reluctant to provide
values. For example, the qualitative assessment of the Mexico EDGE pilot revealed
that information about savings tended to be a private matter, related to personal goals,
and thus was not usually subject to questioning or public discussion.146 In Maldives, 146 United Nations, Statistics

where a fraudulent scheme involving bank accounts came to light during the period of Division, United Nations Entity
data collection for the EDGE pilot, the National Bureau of Statistics opted not to ask for Gender Equality and the
respondents to provide account balances. Empowerment of Women
and National Statistics and
407. Given the sensitivity of asking about the value of financial assets, an alter- Geographic Institute of Mexico,
native approach in collecting such sensitive data is to request a range of values as “Assessing Mexico’s pilot sur-
response categories and to use the average of the data obtained for calculation of the vey on measuring individual
gender wealth gap. The response category “Refuses to respond” should be included in level asset ownership and
entrepreneurship from a
both approaches.
gender perspective”, EDGE
408. Enumerators should be trained accordingly on how to solicit sensitive final report (New York, October
information. The training should include the need to emphasize to respondents the 2016).
security and confidentiality with which such information will be treated. It is also
recommended that the module on financial assets be placed near the end of the ques-
tionnaire and that financial asset values be reported at the individual level rather than
at the household level.
409. For small agricultural equipment and livestock, the model questionnaire
does not include any question on values. This is because it is so labour-intensive to
itemize these types of assets (further details on this can be found in the discussion on
the roster of assets above). It is suggested that, if a country is interested in collecting
value for those specific assets because of their important contribution to the wealth
of a significant proportion of individuals, a total value of each type of asset and the
proportion that is owned by the respondent might be collected.

5.2.2.10. Within-household decision-making

410. As discussed in chapter I of these Guidelines, studies have found that
women’s asset ownership is correlated with a stronger role for women in household
decision-making. To assess such correlation, questions on within-household decision-
making can be asked, which were piloted in the EDGE survey in South Africa. The
questions might ask about who takes decisions on how income is to be used, seek-
ing heath care, making major household purchases, visiting family and relatives, and
about domestic violence. These questions are standard and have been used in demo-
112

graphic and health surveys. The questions used in the EDGE pilot in South Africa are
available online at https://unstats.un.org/edge, and examples of the questions are also
available in the annex to these Guidelines.

5.2.3. Specific considerations in the questionnaire design for selected assets

411. In designing questionnaires for collecting asset ownership at the individ-
ual level, a few issues specific to certain types of assets need to be taken into considera-
tion. These are further discussed in this section.

5.2.3.1. Principal dwelling

412. The dwelling structure may be owned separately from the land on which
it sits, and the plot and the structure may have been acquired at different times and in
different ways. For example, a person may have inherited a plot of land and may own
it individually; then the person may have built a house—the dwelling—jointly with
his or her spouse on the land. Or, the plot may not be legally owned by a household
member, but a household member may own the dwelling. Countries should collect this
type of information through qualitative research prior to designing the questionnaire.
If the dwelling and the plot on which it stands are considered separate properties, then
the questions related to the principal dwelling that were discussed in the above section
should be also asked with reference to the plot of land on which the dwelling is located.

5.2.3.2. Agricultural land

413. One specific data item related to agricultural land concerns security of
tenure. WCA 2020 defines land tenure as “the arrangements or rights under which
147 FAO, WCA 2020, vol. 1, 2017, the holder operates the land”.147 WCA 2020 recognizes that there are many differ-
para. 8.2.36. ent systems of formal and informal land tenure around the world and the distinction
between legal and non-legal ownership (one of the keys to tenure security) is often
blurred. Consequently, only four broad categories of land tenure are proposed by
WCA 2020 as follows:
•• Legal ownership or legal owner-like possession: this refers to legal own-
ership obtained through either a formal land title system or customary
land tenure arrangements that are registered or certified in some way. Such
arrangements might include possession of an ownership title by the holder;
operation of the land by the holder under hereditary tenure arrangements;
perpetual or long-term lease (with nominal or no rent); or the land is held
148 Adriana Neciu, “Approaches under tribal or traditional form of tenure recognized by the State.148
to measuring asset ownership
and control in agricultural •• Non-legal ownership or non-legal owner-like possession: this covers situ-
censuses and surveys”, paper ations where the holder operates the land without interruption for a long
prepared for the EDGE project period of time without any legal form of ownership, title, long-term lease
(Rome, FAO, 2013). or rent payment; is operating land owned by the State without any legal
rights; or is operating land held under tribal or traditional form of tenure
that is not recognized by the State.
•• Rented from someone else: land may be rented for an agreed amount of
money or produce or both, for a share of the produce, or in exchange for
services; or land may be granted for free. The categories are the following:
rented for an agreed amount of money or produce or both; rented for a
share of produce; provided in exchange for services; and provided under
other rental arrangements.
 Guidance for implementation 113

•• Other types of land tenure include the following: land operated on a squat-
ter basis; land operated under transitory tenure forms, such as trusteeship;
land received by members of collective holdings for individual use; and 149 For the purposes of collecting
land under inheritance proceedings. data for Sustainable Develop-
414. Because the module proposed in these Guidelines collects information only ment Goal indicator 5.a.1, the
on agricultural land considered to be owned by the respondent, the tenure status of suggested list of legally recog-
the parcels reported by the respondent should fall under either legal ownership or legal nized documents includes: title
deed, certificate of occupancy
owner-like possession, or non-legal ownership or non-legal owner-like possession.149
or land certificate, legally
415. The categories of land tenure identified in WCA 2020 are purposely broad recognized purchase agree-
so that each country could use its own categories of land tenure, which would allow ment, legally recognized will or
for a more detailed analysis. For example, the categories used in the EDGE pilot study certificate of hereditary acqui-
in Uganda were: “Mailo”, “Customary”, “Leasehold” and “Freehold”,150 while in South sition, certificate of customary
tenure, certificate of perpetual
Africa, the categories were “Owns and farms the land”, “Owns and rents out the land”,
or long-term lease or rental
“Owns and sharecrops out the land”, “Tribal authority”, “State land” and “Other”. It agreements, and certificate
should be noted that land rented or sharecropped should not be included in the mod- issued for adverse possession
ule. These categories may be added to the list, however, if countries consider them or prescription. It should be
relevant. The tenure status categories used by the countries can then be collapsed into noted that, while the present
the categories designated in WCA 2020, as warranted by additional analysis. Guidelines focus on owner-
ship, the methodology under
416. Two questions on tenure security of agricultural land were piloted in South Sustainable Development Goal
Africa. The first: “What is the tenure status of this [parcel]?” The second: “What could indicator 5.a.1 also covers use
make you lose ownership of this parcel over the next five years?”151 rights. See FAO, “Measuring
417. Sources of perceived tenure insecurity may include contestation from individuals’ rights to land: an
integrated approach to data
within households, families or communities, or as a result of the actions of govern-
collection for SDG indicators
ments, companies or other private land claimants. Individuals holding land under 1.4.2 and 5.a.1” (forthcoming).
customary systems may perceive their rights as secure despite the absence of legal rec- 150 These categories are offi-
ognition or formal documentation, while those with formal documentation may still cial land tenure systems as
perceive some insecurity depending on the robustness of the institutions enforcing enshrined in the 1995 Ugandan
documented land rights. National statistical agencies should customize the response constitution (chap. 15, art. 237).
categories according to their country context. “Mailo” and “freehold” are
forms of documented owner-
5.2.3.3. Non-agricultural enterprise assets ship, but “mailo” is essentially
feudal in character, and recog-
418. Collecting information on non-agricultural enterprise assets differs from nizes occupancy by tenants,
data collection on all the other assets in many ways. Those differences have been cov- whereas “leasehold” is owner-
ered in various places whenever relevant (see section 5.2.2). A summary of those dif- ship for a particular period of
time, and “customary” usually
ferences is also presented below.
means ownership without land
419. First, the scope of non-agricultural enterprise assets is limited to different titles.
categories of enterprise assets for unincorporated enterprises including the following: 151 The question proposed to

the current stock of physical capital, including all machinery, equipment, and fur- measure perceptions of
niture used for the business that were not listed earlier in any of the other modules; tenure security under Sus-
the current stock of inputs or supplies, including raw materials; and the current stock tainable Development Goal
of finished merchandise (goods for sale). For each category of enterprise assets, the indicator 1.4.2 identifies the
likelihood of the respondent
respondents will also report on the total value if all the assets in that category were to involuntarily losing ownership/
be sold today. While enterprises may be considered “assets” in the sense that holding use rights to the parcel in the
the enterprises would bring a series of economic benefits to the owner, the SNA con- next five years “on a scale from
siders enterprises as economic institutional units, not assets. 1 to 5, with 1 being not at all
likely and 5 being extremely
420. Second, slightly different from all the other assets with reported ownership
likely”. See FAO, “Measuring
of asset as filter question, the data collection for non-agricultural enterprise asset starts individuals’ rights to land: an
with a few questions assessing whether the respondent owns a business and estab- integrated approach to data
lishing the sector in which the business operates, distinguishing between agricultural collection for SDG indicators
activities and non-agricultural activities. 1.4.2 and 5.a.1” (forthcoming).
114

421. Once a respondent is identified as owning a business, a roster of enter-

prises—rather than a roster of assets—should be established. The roster should list
each non-agricultural enterprise described by the respondent to create a respond-
ent roster of non-agricultural enterprises. The roster of enterprises also includes two
questions to determine whether the enterprise is incorporated (i.e., the production
unit is a separate legal entity from its owners) or unincorporated. Consistent with the
2008  SNA and the OECD Guidelines, the assets owned by incorporated enterprises
cannot be owned by the respondent and thus, in these Guidelines, are excluded from
the measurement of wealth at the individual or household level. If the enterprise is a
limited liability enterprise and keeps formal accounts, it is an incorporated enterprise
and the enumerator should skip to the next enterprise or the next module.
422. Last, only those assets that were not listed in the previous modules should
be included here to avoid the double-counting of assets. Any motor vehicles used for
the enterprise should be listed and valued in the module on consumer durables. Any
land and buildings used for enterprises should be listed and valued in the module on
other real estate.

5.2.4. S pecific considerations in questionnaire design for different

survey instruments
423. Data on asset ownership at the individual level may be collected through
appending questions to an existing household survey or a stand-alone household sur-
vey. Regardless of the type of survey instruments, countries will need to customize
them according to their local context. At a minimum, for many of the questions, each
country will have to determine the appropriate response categories or coding. For
example, under the question as to whether there is an ownership document for a given
asset, each country should list the various types of ownership documents that are used
locally. The possible modes of acquisition will also differ by country and by asset. To be
more thorough, countries should draw upon qualitative research and prior quantita-
tive research to customize the questionnaire, as discussed previously.

5.2.4.1. Appending a minimum set of questions on asset ownership and control

424. If the objective of the data collection exercise is to understand gender dif-
ferences in owning specific types of assets, countries may choose to append a mini-
mum set of questions to a nationally representative household survey.
425. These Guidelines recommend restricting the minimum set of questions
to priority assets, including principal dwellings, agricultural land, other real estate
and financial assets. The minimum set of questions ask about the reported and docu-
152 The applicability of the ques- mented ownership and the ownership rights over those assets, whenever applicable.152
tions to different types of asset A roster of assets is considered unnecessary, given that the number of questions is to be
is discussed in section 5.2.1 minimized. Accordingly, the question to be asked on the ownership of dwellings and
below. agricultural land is the following: “Do you own [this dwelling/any agricultural land]?”,
with the following response categories: “Yes, exclusively”; “Yes, jointly”; “No”. If the
answer is “Yes, exclusively,” or “Yes, jointly”, a question will be asked at to whether
there is an ownership document for the dwelling or for any of the agricultural land
owned by the respondent, followed by the question: “Are you listed as an owner in any
of the ownership documents?” Ownership rights questions are also asked in relation
to the asset owned by the respondent. It should be noted that, for principal dwellings,
those questions make it possible to calculate the prevalence of reported and docu-
mented ownership and the right to sell and bequeath, as well as the overlap of those
different forms of ownership of and rights over the dwelling. For agricultural land,
however, the only estimates that can be produced are the prevalence of each single
 Guidance for implementation 115

form of ownership and rights. There is no guarantee that the different forms of owner-
ship or rights refer to the same agricultural parcel. It is not possible to study the overlap
between different forms of ownership of and ownership rights over agricultural lands.
426. Incorporating the minimum set of questions for other real estate and
financial assets is a more complex undertaking. Asking such questions as: “Do you
own any other real estate?” or “Do you own any financial assets?” yields little informa-
tion that is of policy relevance. In this case, a question on the ownership and rights
could be formulated as follows, with the corresponding response categories:
“Do you own any of the following categories of other real estate?”
•• Dwelling “Yes, exclusively” “Yes, jointly” “No”
•• Non-agricultural land “Yes, exclusively” “Yes, jointly” “No”
•• Other categories
considered important “Yes, exclusively” “Yes, jointly” “No”
Similar questions can be formulated for the gathering of information on documented
ownership rights to sell and bequeath. It should be noted, however, that for financial
assets only the reported ownership is relevant.
427. Thus far, the Guidelines have recommended that agricultural land be
treated as a distinct category, separate from land that may be used for non-agricultural
purposes, which is classified as “other real estate”. This recommendation is prompted
by the importance that the ownership and control of agricultural land have for a range
of policy issues, including, for example, agricultural production, food security and
the development of rural communities. Land may also, however, be treated as a single
entity, covering both agricultural and non-agricultural land. The question would then
be formulated as “Do you own any land?”, followed by other questions on documented
ownership and the right to sell or bequeath.

5.2.4.2. Appending a module on asset ownership and control at the individual level
428. A more elaborative list of questions may be appended to an existing house-
hold survey if the objective of the data collection exercise goes beyond the calculation
of prevalence rates of asset ownership. Countries may want to collect data on the full
range of physical and financial assets included in the stand-alone survey, but limit the
number of questions asked about each asset. For instance, some countries may wish
to ask questions about the types and forms of ownership rights for each asset only,
which would enable countries to begin to monitor gender patterns of asset ownership
and to assess the extent to which the full range of rights is correlated with ownership
in the country. Other countries might wish to also ask questions about the value of
assets, making possible an analysis of gender wealth gaps, since the value of men’s and
women’s assets may differ. Lastly, some countries may wish to collect data on a few key
assets only, such as principal dwellings, agricultural land and other real estate, but to
ask the full set of questions for each asset.
429. Once a national statistical agency has a clear understanding of the survey’s
objectives, as discussed earlier in chapter III of these Guidelines, it can refer to table 8
above to design a module on asset ownership and control. A careful assessment of the
host survey questionnaire should be carried out to avoid any duplication of questions. For
example, a household roster and individual characteristics are usually already collected
in most household surveys. The roster and characteristics of assets are also included in
some household surveys and should not be asked again. Given its sensitive nature, infor-
mation about financial assets should always be sought at the individual level.
116

430. To minimize the complication resulting from the need to reconcile asset
ownership within the household when multiple respondents are selected and provide
information, efforts should be made to ensure that the roster of assets (with the excep-
tion of financial assets) and their characteristics is collected at the household level. If
there is only one respondent randomly selected from each household, questions on the
roster of assets and their characteristics can be collected from respondents at the indi-
vidual (assets owned by the individual) or household (assets owned by all household
members) level.

5.2.4.3. Stand-alone survey

431. Collecting information on asset ownership at the individual level through
a stand-alone survey should serve all three objectives: deriving estimates on the gen-
der asset gap and the gender wealth gap and conducting an intrahousehold analysis
of asset ownership. The survey should collect all the information outlined in table 8.
432. As is standard in household surveys, the questionnaire should start with
a statement of purpose that explains the survey to the households selected for inter-
view. The statement of purpose should be read before the household questionnaire
is administered and again to the adult household member randomly selected for the
individual interview, if this is a different person. Before proceeding to the question-
naires, respondents should be given time to ask questions about the survey.
433. Below is a template for a statement of purpose for a survey on asset owner-
ship and control. Each country should customize it accordingly, and translate it into
local languages, as warranted, ensuring that the term “asset ownership” is clearly
understood. It is recommended that countries do not describe the survey as a survey
on gender or gender equality, as this may be off-putting to some respondents.
The [name of NSO] is conducting a survey of households across [country] to bet-
ter understand asset ownership in the country. The findings from the survey will
provide important information to the Government for developing policies and pro-
grammes to improve the lives of men and women in [country]. Your household
was selected as one of those to which the survey questions will be put. You were
not selected for any specific reason. Rather, your household was selected randomly
from a list of all of the households in this village.
All the information that your household provides is strictly confidential. It will not
be shared with any other government agency, and it will only be used for statistical
purposes by the [NSO] or under its supervision. To ensure that the most accurate
information is collected, it is very important that we interview the specific house-
hold member selected for the interview and that we interview him or her alone,
without family or neighbours present. If, during the interview, any family members
or neighbours come within hearing distance of the interview, please ask them to
kindly come back later after the interview has been completed. Please spare some
time to answer the questions. We thank you in advance for your time.
434. The stand-alone survey also includes a household questionnaire and an
individual questionnaire. The household questionnaire comprises the household
roster, which captures key sociodemographic information about all members of the
household, and a roster of key assets, together with the characteristics of those assets.
As more than one respondent will be selected from each household, the inclusion of
the roster and characteristics of assets at the household level will eliminate the compli-
cation of reconciling assets reported by multiple respondents. The household question-
naire can be completed by any household member, but ideally should be administered
to a person knowledgeable about the characteristics of all household members and
 Guidance for implementation 117

the assets that they own. The last component of a stand-alone survey questionnaire is
the individual-level questionnaire that is administered to selected respondents within
the household, following the selection protocol described in chapter III, section 4, on
sample design. It is again recommended that, given its sensitive nature, information on
financial assets should be collected at the individual level only.

5.3. Testing the questionnaire

435. As previously noted, countries are encouraged to conduct their own back-
ground research for the purpose of customizing the model questionnaire presented
above to their country context. After the questionnaire has been customized, several
methods for testing the questionnaire should be considered, including expert reviews,
cognitive interviewing, field pretests and randomized experiments. The decision as
to which testing methods to employ will be based on the available survey budget and
whether the survey questions are being used for the first time. At a minimum, statisti-
cal offices should use expert reviews and field pretesting of the questionnaire. When,
however, a survey on asset ownership is being implemented for the first time with a
new questionnaire, focus groups and cognitive interviews should also be conducted.
436. Testing of the questionnaire should assess three aspects: first, whether it
covers the assets and aspects of asset ownership that are relevant in the country context
and uses terms that are clearly understood by the respondents (content standards); sec-
ond, the ability of respondents to formulate answers to individual questions (cognitive
standards); and, third, the ability of interviewers and respondents to easily complete the
entire questionnaire (usability standards).153 National statistical offices can use several 153 Groves and others, Survey

methods to evaluate draft survey questionnaires, as described below.154 Methodology, 2009.

154 Ibid.

5.3.1. Expert reviews

437. Expert reviews ensure that the questionnaire collects the information
needed to achieve the objectives of the survey and in the form needed, including the
proper structure of the questionnaire and flow of the questions, wording of the ques-
tions, response categories, instructions to interviewers and skip patterns. This method
is easiest to carry out at minimal expense and was implemented by all the EDGE pilots.
Reviewers should be questionnaire design experts and subject matter experts, some of
whom may be in the group of stakeholders involved in the planning and implementa-
tion of the survey. Subject matter experts and research analysts can have a key role in
identifying aspects of asset ownership that are relevant in the country context but that
have not yet been included in the questionnaire. They can also ensure that the data are
collected in the format and detail needed to achieve the objectives of the survey.
438. Potential problems in formulating questions and categories of answers can
also be revealed, including unclear purpose; reference to information that respondents
are unlikely to know or recall; complex syntax; vague, ambiguous or imprecise terms
or unfamiliar technical terms; misleading or incorrect presuppositions; and mismatch
between the questions and response categories.155 155 Ibid.

5.3.2. Focus groups

439. Focus groups may be involved even before a survey questionnaire is devel-
oped and when the customizing of a given model questionnaire is being considered.
Focus group discussions can explore what members of a target population think about
asset ownership and what terms they use in talking about them. Focus groups consist
of a small number of participants (6 to 10) and a moderator. Usually, the participants
118

are selected to form a homogeneous group, and more than one focus group should be
mounted to cover different subpopulations in the country. During discussions, the
moderator follows a set of pre-identified topics (but no scripted questions or prob-
ing questions) and the participants are encouraged by the moderator to freely express
their point of view on those topics.

5.3.3. Cognitive interviews

440. Cognitive interviews are conducted with individuals for the purpose of
understanding how respondents understand the questions in the draft questionnaire
and how they formulate their answers. The person conducting the cognitive interview
may be a research scientist, cognitive psychologist, expert in survey question methodol-
ogy, or an interviewer with special training or experience in question evaluation. Cog-
nitive interviewing may involve different techniques, including requiring respondents
to think aloud and verbalize their thoughts as they answer a question or after they have
answered a set of questions or a section of a questionnaire; paraphrase some questions
in their own words or even provide definitions for key terms in the question; answer
additional probing questions to reveal why specific answers were given; and rate how
confident they were in giving answers to specific questions. Cognitive interviews may
be recorded by video or audio; alternatively, interviewers may take notes. The informa-
tion obtained can be used to revise the questions and the response categories.

5.3.4. Field pretests

441. Field pretests consist of a small number of interviews, typically up to 100,
using field procedures similar to those of the full-scale survey. The purpose is to eval-
uate the entire questionnaire in different settings (for example in large cities, small
towns, rural areas or other areas of the country that have different tenure systems or
different marital regimes). Interviewing protocols may also be tested at the same time.
The pretest interviews may be conducted by the statistical office staff or field supervi-
sors. At the end of the field pretests, the interviewers should be debriefed on which
questions worked in the field and which did not. Interviewers can often offer sugges-
tions on how to improve questions and categories of answers. Data obtained during
the pretest may be entered and tabulated to identify items with high rates of missing
data that may need to be revised or removed.
442. In some field pretests the observations on how questions are asked and
answered may be more systematic, using the technique known as behaviour coding.
Interviews may be recorded, with the permission of the respondent. After the field pre-
test the behaviour of the interviewer and that of the respondent are coded consistently
across interviews using the same categories of assessment. Alternatively, the coding
is carried out during the interview by a third person present while the questionnaire
is being administered. The technique makes it possible to tabulate rates of specific
respondent behaviour, such as seeking clarifications, giving answers that inadequately
respond to the questions, or interrupting the reading of questions, which may indi-
cate that the questions were poorly phrased. At the same time, the technique provides
information on the interviewer’s behaviour (such as whether the question was read in
such a way that its meaning was altered), which may be controlled for when analysing
information on respondents’ behaviour.

5.3.5. Randomized or split-ballot experiments

443. Generally, randomized or split-ballot experiments can be conducted for
the purpose of comparing different versions of the questionnaire or different methods
 Guidance for implementation 119

of data collection and field procedures. These different questionnaires or procedures

are covered separately in random portions of the sample. One example can be seen in
the Methodological Experiment on Measuring Asset Ownership from a Gender Per-
spective (MEXA), which was implemented in Uganda as part of the EDGE project. As
described in box 3 of these Guidelines, MEXA tested the relative effects of five different
approaches to survey respondent selection on individual-level measurement of asset
ownership. One of the key findings that informed these Guidelines was that informa-
tion on individual-level ownership reported by proxy by the household head yielded
estimates of women’s and men’s asset ownership that differed from those obtained by
asking respondents to self-report their ownership status.
444. Nevertheless, randomized experiments can be costly and they need tight
supervision in the field to control other factors that may influence the results obtained
in the different samples. More important, although these experiments can demon-
strate that the different versions of the instruments or procedures produce different
results, they cannot resolve the question as to which version produces better data,
unless external validation data can be used to check survey responses or strong theo-
retical reasons form the basis for decisions that one version of the questions is better
than another.

5.4. Designing and testing the CAPI questionnaire

445. This section presents general issues related to the design and testing of
the CAPI questionnaire. As discussed in the section on modes of data collection, an
increasing number of countries are considering using computer-assisted interviewing
for the purpose of increasing the quality and timeliness of data. In countries choosing
the CAPI approach for the collection of data on asset ownership, the designers of the
questionnaire should pay particular attention to two aspects. First, they should ensure
that the complexity of the questionnaire, given by the multiple units of observation
and rosters, is reflected in the CAPI questionnaire, through the proper nesting of vari-
ous sections and subsections. This aspect should also be verified at the testing stage. A
second consideration relates to the use of automatic procedures for the random selec-
tion of the person to be interviewed about his or her own assets.

5.4.1. Designing the CAPI questionnaire

446. Designing a CAPI questionnaire involves more than simply replicating the
paper questionnaire on the screen of a handheld device.156 The CAPI questionnaire 156 For more information on

should be developed after the paper questionnaire is finalized and it should include CAPI design see World Bank,
all the information planned to be collected and covered in the paper questionnaire. Survey Solutions. Questionnaire
It should be noted, however, that what might appear as a single question in a paper Designer-User’s Guide (Wash-
ington, D.C., 2015). Available at
questionnaire may correspond to one or more questions in the CAPI questionnaire.
http://siteresources.worldbank.
For that reason, CAPI designers should have a good understanding of the types of org/INTCOMPTOOLS/Resources
variables in the database that will correspond to each question in the questionnaire. /8213623-1380598436379/
447. Most of the questions included in the CAPI questionnaire are for the pur- designer.pdf.
pose of information gathering during the interview and, once the interview starts,
those questions are the only ones displayed. Some questions, however, may have dif-
ferent purposes. Some are prefilled with information that is useful to enumerators
in completing their field assignments, such as the identification of enumeration area,
household address, household identification number and enumerator identification
number. Other questions are to be completed by the supervisors, once the enumerator
completes the interview and sends it for checking and approval, while others have the
sole purpose of being used in validation and enabling conditions.
120

448. How information is displayed on the device screen is important. More

than one question should be displayed at a time, to help orient the interviewers in
the overall flow of the questionnaire. The interviewers should be able to navigate eas-
ily through the questionnaire and immediately understand, for example, which text
they should read out loud and which contains instructions for them. This task can be
accomplished by using different graphic characters and colour variation in displaying
information that has different purposes. Interviewers should also be able to immedi-
ately spot messages generated by the software application when some questions have
been left unanswered or the answers given are not within the range expected.
449. One of the most important tasks in designing a CAPI questionnaire is the
proper implementation of skip patterns and validation checks. In a survey on asset
ownership, not every respondent will be asked every question in the questionnaire.
Depending on the answers to some questions, some questions will be skipped. For
example, if a person does not own a particular asset, questions related to the mode
of acquisition or value of that asset will be skipped. In CAPI questionnaire design,
skip patterns are implemented through the use of enabling conditions for a question,
which determine whether that question is displayed or not contingent on answers in
one or more previous questions. It should be stressed that, once errors are made in the
structure of the CAPI questionnaire, enumerators have no power to correct them in
the field. For that reason, care must be taken to ensure that skips are correctly placed,
so that relevant questions are not excluded in the interview. This requires a thorough
review of the logical conditions implemented in the CAPI questionnaire design and
testing of the questionnaire.
450. Implementation of validation conditions, which determine whether an
answer recorded is acceptable, in other words, whether it is within the expected range
of values, is particularly important in the CAPI questionnaire design. Control rules
that would be performed during the data entry stage in a survey using a paper question-
naire can be implemented in the design of the electronic questionnaire. There are two
types of errors that can be managed: range errors (e.g., a person’s age being recorded
as 157) and inconsistency errors (e.g., a person’s age not being equal to the difference
between the current date and that person’s date of birth). When errors occur, error
messages can be displayed to alert the interviewer to the need to probe the respondent
or to correct the answer that has been wrongly entered.
451. Another important aspect is the flow of the questionnaire, in other words,
the order in which the questions appear on the screen and are administered to the
respondent. This aspect is particularly relevant in the context of complex question-
naires dealing with multiple interrelated sections and rosters (including a household
members’ roster and potentially several separate rosters of assets owned). This is the
case of the questionnaire on asset ownership. When designing the CAPI question-
naire, the nesting of sections, subsections and questions needs to be carefully man-
aged, to ensure that the questions appear on the screen as planned. Similarly, when
more than one person is interviewed in each household, adequate nesting is also
required so that it is only possible to proceed to the next individual interview once
the previous individual interview has been completed. Lastly, some elements of design
may involve using more complicated macros or developing new functions. This may
be the case, for example, when using automatic procedures for the random selection of
the person to be interviewed about own assets.

5.4.2. Testing the CAPI questionnaire

452. An initial test of the CAPI questionnaire should be conducted by survey
team members who are familiar with the questionnaire. This will help to resolve the
 Guidance for implementation 121

most obvious errors prior to the field pretest. A key aspect to be verified at this stage is
whether the questionnaire flows in the intended order, in other words, whether proper
nesting was used to integrate the multiple rosters and sections of the questionnaire on
asset ownership.
453. The field pretest will be the first opportunity to test the entire CAPI-based
data collection system in the environment in which it is to be used. All components
of the system should be checked, including synchronization with the headquarters,
access to work assignments, the completion of several interviews with real households,
the transfer of completed questionnaires, making provision for supervisors’ checks
and approval, the receipt of the data at headquarters and the conduct of additional
checks, the implementation of additional validation rules and the testing of the data-
base structure.
454. The questionnaire incorporating the revisions suggested by the field pre-
test should be used to train the interviewers and supervisors, including in the field
practice. The field practice will also test the communication and data transfer proce-
dures and the entire network infrastructure. Complete interviews implementing the
interviewing protocol established for data collection in the field should be followed. A
final questionnaire should be prepared based on observations from the field practice.

5.5. Survey manuals

455. For face-to-face interviews, a detailed instruction manual should be pre-
pared for supervisors and interviewers, covering all aspects of the survey. When the
interview is to be conducted using CAPI, practical guidance should also be prepared
on how enumerators and supervisors can perform CAPI-related tasks.

5.5.1. Instruction manual for fieldworkers

456. As with any survey conducted by the national statistical agency, a detailed
instruction manual should be prepared for supervisors and interviewers prior to the
start of field training. The manual should be prepared in a language that a typical
interviewer can easily understand and serve as guidance during training, along with a
reference document during field operations.
457. All aspects of the survey should be covered in the manual. Some compo-
nents will be similar to those of other surveys, including, for example, administrative
responsibilities of interviewers, rules of proper behaviour and dress, and strategies for
minimizing non-response in the survey. Other aspects, however, should be specific to
data collection on asset ownership. The manual for interviewers should cover the fol-
lowing aspects:157 157 United Nations, Handbook

of Household Surveys, 1984.

(a) General background and purposes of the survey, the scope of the informa-
tion, and the general type and coverage of the sample;
(b) Administrative responsibilities of interviewers, in terms of managing
materials, proper planning and organization of their workload, procedures
for reporting progress, and problems and preparation of necessary admin-
istrative forms;
(c) Basic interviewing rules, in terms of proper behaviour and dress, the need
for proper identification, courtesy in interviewing persons of all walks of
life and appropriate ways of introducing the survey;
(d) Instructions for the selection of households and appropriate respondents
in each household;
122

(e) Importance of, and strategies for, minimizing non-response in the sur-
vey, including arranging for return visits and procedures for dealing with
refusals;
(f) Detailed instructions and specifications for each item in the questionnaire,
permissible types of probing, tactful ways of dealing with inconsisten-
cies, methods of recording information, types of notes and explanations
needed.
458. Additional information is required for field supervisors to aid them in
158 Ibid. their supervisory responsibilities.158 These include:
(a) Procedures for organizing and controlling the flow of materials to and
from the field;
(b) Means of monitoring fieldwork, importance of adherence to timetables,
procedures for the field review of completed questionnaires and applica-
tion of quality control procedures;
(c) Steps to take when serious errors are discovered;
(d) If supervisors are involved in the recruitment and training of interviewers,
additional provisions on these matters should be covered.
459. In addition to the manuals, other training materials may be developed,
including materials that interviewers can study at home before attending training ses-
sions, including instructional materials and test exercises; materials for group training
sessions, including test exercises, recordings of illustrative interviews, slides and other
visual aids that can show mapping materials, questionnaire forms and the like.

5.5.2. CAPI manuals

460. In addition to the manuals of instructions for fieldworkers, manuals should
be developed that provide practical guidance on how enumerators and supervisors can
159 See World Bank, Survey Solu- perform CAPI-related tasks.159 These manuals may cover practical issues on how enu-
tions Interviewer Manual (Wash- merators should sign in and out of the software application used (thus barring access
ington, D.C.), available at http:// by others to the sensitive data recorded on the device), and how to use the tablet for the
siteresources.worldbank.org/ purpose of opening and managing their work assignment. Information on managing
INTCOMPTOOLS/Resources/
work assignments may include details on how to check the status of each assignment
8213623-1380598436379/
Interviewer_manual.pdf; and (including whether or not the assignment has been completed by the enumerator and
World Bank, Survey Solutions approved or rejected by the supervisor); how to open, close or resume an assignment;
Supervisor Manual (Washing- and how to transfer completed questionnaires and receive new assignments through
ton, D.C.), available at http:// a synchronization procedure. A section of the manual may be dedicated to examples
siteresources.worldbank.org/ of key issues to which the enumerators should pay particular attention in the field,
INTCOMPTOOLS/Resources/ including, for example, types of questions that are more complicated and how to use
8213623-1380598436379/
Supervisor_manual.pdf, for an
rosters. A separate section may also be dedicated to quality checks, including checking
illustration of issues that may that all questions have been completed and answers are valid, and how to leave com-
be covered in a CAPI manual. ments that supervisors can check.
461. Practical guidance tailored to supervisors should also be included. Super-
visors have a key role in the survey workflow. They receive the survey assignments
from headquarters and allocate them to the interviewers in their team. Once the
questionnaires have been completed by the interviewers, the supervisors review those
questionnaires to confirm that all questions are answered and the answers are coher-
ent and plausible. This review may result in the questionnaire being approved (and
therefore transferred to headquarters) or rejected (and therefore returned to the inter-
viewer for corrections, completion, or explanatory notes). The manual may illustrate
how the software application can be used by supervisors at each step of this process.
 Guidance for implementation 123

The manual should also illustrate how the application should be used to conduct regu-
lar checks of the overall status of the fieldwork and actively manage the workload dis-
tribution of the team members through assignments and reassignments. Lastly, the
manual should provide support on how to troubleshoot problems that interviewers
may have, including, for example, with synchronization of their tablets, checking for
the updates and manually backing up the data collected.

5.6. Translating survey instruments

462. Countries are likely to have in place translation protocols for their national
survey programme and these protocols can be followed for collecting individual-level
data on the ownership of assets. In general, if the data are to be collected in more
than one language, best practice indicates that the survey questionnaire should be
translated into the main languages spoken in the area where the survey will be imple-
mented, then back-translated to ensure proper translation. This should occur prior
to the training of enumerators, so that enumerators are accustomed to the translated
materials. If countries choose instead to rely on oral translation by enumerators in the
field during interviews, they may wish to prepare a small packet of translated mate-
rials prior to the training of enumerators to aid them in the field. This may include
translations of the glossary definitions found in the manual of instruction, the state-
ment of purpose to be read at the start of each interview and key questions that appear
throughout each module. It is particularly important that terms related to ownership
and control of assets are thought out in each language to ensure consistency in data
collection across all interviewers.

Key ponts
•• A generic set of questions is proposed for countries to collect data on asset owner-
ship. Countries are encouraged to conduct their own background research for the
purpose of adapting the generic set to the country context.
•• After the questionnaire has been customized, several methods for testing the ques-
tionnaire should be considered, including expert reviews, cognitive interviewing,
field pretests and randomized experiments.
•• In countries choosing the CAPI data collection method for asset ownership, the design-
ers of the CAPI questionnaire should pay particular attention to two aspects. First, they
should ensure that the complexity of the questionnaire, consisting in multiple units of
observation and rosters, is reflected in the CAPI questionnaire, through the proper nest-
ing of various sections and subsections. This aspect should also be verified at the testing
stage. The second aspect is related to the use of automatic procedures for the random
selection of the person to be interviewed about his or her own assets.
•• A detailed instruction manual should be prepared for supervisors and interviewers
prior to the start of field training. The manual should be prepared in a language that
a typical interviewer can understand easily and serve as guidance during training,
along with a reference document for use during field operations. All aspects of the
survey should be covered in the manual.
124

6. Field operations
463. Guidance on field operations, in these Guidelines, covers the following
aspects: field organization, training of field staff and various aspects of the field work.
While such guidance is typical of household surveys, aspects that are particularly rel-
evant for the collection of data on asset ownership from a gender perspective are high-
lighted.

6.1. Field organization

464. The organization of fieldwork can greatly affect the quality and cost of the
survey and the staff involved in planning the survey should give considerable attention
to these activities. This section addresses issues related to four key activities: recruit-
ment of field staff, publicity, use of geospatial information to support enumeration and
printing of field materials. In addition to aspects that are typical of the field organiza-
tion in any household survey, the section includes details on three aspects specific to
implementing a survey on asset ownership: the organizing of the field staff into small
teams, with a high ratio of supervisors to enumerators; assessment of the necessity
to match the genders of the interviewers and the respondents; and the content of the
publicity material.

6.1.1. Recruitment and organization of field staff

465. Field staff typically comprise supervisors and interviewers. Field supervi-
sors may be full-time statistics officers or other employees assigned to posts related to
survey operations in central or regional offices, or they may be employed on a tempo-
rary or part-time basis. In the latter case, they may be selected from the ranks of those
interviewers who have experience and exhibit the ability and willingness to take on
more responsibilities.
466. Interviewers are less likely to be permanent staff of statistical offices and
most of the time they are selected and employed for fieldwork in specific surveys. Sta-
tistical offices usually develop a network of experienced interviewers who can be called
as needed.
467. Desirable traits for interviewers in household surveys would typically
160 United Nations, Handbook include:160
of Household Surveys, 1984. •• Sufficient education (the exact requirements may vary from country to
country);
•• Absence of highly opinionated views, in particular with regard to the sub-
jects covered by the survey;
•• Willingness to accept instructions and to adhere to rules;
•• Knowledge of local languages and dialects used in the areas where they are
going to conduct interviews;
•• When using CAPI, familiarity with computers and keyboard skills;
•• Availability for travel and work in the evening and on weekends, when the
respondents may be available for the interview.
468. Costs related to field staff are one of the largest components of the survey
budget; accordingly, decisions related to the number of interviewers and supervisors
and their selection should be made early in the planning stage of the survey. The num-
ber of interviewers who need to complete the fieldwork within the specified survey
period should be estimated on the basis of the size of the sample, its geographical
distribution and the number of interviewers needed per household. The number of
 Guidance for implementation 125

interviews per interviewer may vary across areas where the survey is implemented,
depending on differences in travel distances and time, access, and the likelihood of
finding respondents at home.
469. How the interviewers are selected also has implications for the survey
budget. The interviewers may be selected locally, which may minimize the travel costs.
In that case, it is important that all interviewers have access to the same type of train-
ing to ensure consistency in data collection. It should be noted, however, that the use
of locally based staff, in particular in small areas, towns and rural areas, increases the
probability that the interviewers and respondents would be acquainted, which may
affect the survey results.
470. The ratio of supervisors to interviewers depends to some extent on the geo-
graphical spread of the fieldwork and the complexity of the survey operations. It is
generally considered that, in situations where the sample is widely dispersed, difficult
communications and complex field protocols, the ratio should not exceed 6 or 8 to 1.161 161 Ibid.

Where close supervision is required, higher supervisor-interviewer ratios should be

considered. In EDGE pilot surveys, the ratio was one supervisor for every four enu-
merators or less.
471. Similarly, the organization of interviewers and supervisors in teams may
vary depending on the complexity of survey operations, in particular the interviewing
protocols used. There are two extremes that can be described, with various arrange-
ments in-between:
(a) Small mobile teams consisting of one supervisor and a few interviewers
who move from one area to the next as fieldwork proceeds. In this case, the
supervisors have a key role in assigning work for each of the team mem-
bers within each area, and planning and scheduling interviews. The use
of mobile teams is generally conducive to better supervision and control
of fieldwork. It can also permit a more efficient sample design since, with
mobility, a given number of interviewers can cover a more dispersed sam-
ple. The organization of field staff in small teams is particularly suitable
for data collection based on interviewing more than one person in each
sampled household and when gender matching between interviewers and
respondents is required. Use of small teams, however, may be associated
with higher travel costs, to cover transport facilities for each team and tem-
porary accommodation in areas covered by the sample;
(b) Use of fixed interviewers, often recruited locally, each working singly in
a fixed sample area for an extended period. With fixed interviewers, the
supervisor may be located elsewhere and visit each interviewer periodi-
cally. In this case, it is important that the interviewers are experienced and
can function independently, as close supervision is not available on a daily
basis. This arrangement is more suitable for simple interviewing protocols,
in which only one respondent is selected for each sampled household and
gender matching between interviewers and respondents is not required.
472. Another aspect for consideration when selecting and assigning interview-
ers is whether customs and traditions make it necessary to match interviewers and
respondents in terms of ethnicity, tribal affiliation, gender or other characteristics.
There may be situations where cooperation cannot be obtained unless the two par-
ties are matched based on those criteria. In general, statistical offices should follow
the protocols that they have developed for this matter. In particular, it is important to
assess whether gender matching between interviewers and respondents is required,
given that the data collection focuses on measuring asset ownership from a gender
126

perspective and both women and men are going to be interviewed. In some contexts,
for example, some men respondents may not be willing to talk to interviewers who
are women, or women respondents may not be allowed to talk to interviewers who
are men. There are, however, other considerations to be borne in mind, including the
safety and security of both interviewers and respondents. For example, in some con-
texts, respondents may feel safer and may be more willing to share information when
the interviewer is a woman. In other contexts, it may be less safe for a woman than a
man to be a field staff member and to travel alone.
473. Gender matching between interviewers and respondents requires a bal-
anced distribution by gender of the overall number of interviewers and within each
team dispatched in the field. Some of the EDGE pilots show that this matching is feasi-
162 Kilic and Moylan, MEXA, 2016. ble. In Uganda,162 for example, the field staff consisted of 16 men and 14 women organ-
ized in seven mobile teams, comprising one supervisor and two to four interviewers.
Gender matching was encouraged on the basis of qualitative fieldwork conducted prior
to the survey by experts on gender and land rights, which showed that respondents were
more comfortable disclosing information on asset ownership when the interviewers
were of the same gender as themselves. The gender matching had a high success rate.
Overall, about 75 per cent of men respondents were interviewed by interviewers who
were men and 82 per cent of women respondents were interviewed by interviewers who
were women.
474. In other countries, women constituted the majority of interviewers and
high rates of gender matching were obtained for women respondents only. In Georgia,
91 per cent of women respondents were interviewed by interviewers who were women,
while only 18 per cent of men respondents were interviewed by interviewers who were
men. In Mongolia, the corresponding proportions were 74 and 40 per cent, and in the
Philippines, 76 and 24 per cent, respectively.
475. Similarly, in Mexico, 68 per cent of women respondents were interviewed
by enumerators who were women and 31 per cent of men respondents by enumerators
who were men. An analysis of several dimensions of quality of interviewing, including
enumerators’ perceptions of fluidity of the interview and resistance to the interview
and the proportion of incomplete interviews, showed that, in the Mexican context, the
less satisfactory gender match-up had no significant impact.

6.1.2. Publicity
476. A survey requires the cooperation of the households selected to be inter-
viewed, and an effort should be made to inform those households in advance about
the survey. To this end, the national statistical office may undertake a selection, as
163 United Nations, Handbook appropriate, of the following tasks:163
of Household Surveys, 1984. (a) Preparation of materials that interviewers can share with respondents,
including pamphlets or articles, in local languages;
(b) Newspaper publicity about the survey; in this case, the interviewers should
be provided with copies of the newspaper article;
(c) Radio or television announcements that may be referenced by the inter-
viewers in the field;
(d) Information disseminated through local government bodies, professional
associations or similar groups;
(e) Securing the approval of local officials, such as village heads, chiefs of
nomadic tribes, including through personal contacts. If the statistical
agency has a regional office structure, these local contacts might more
readily be made by the regional officials;
 Guidance for implementation 127

(f) Where literacy levels are reasonably high, sending advance letters to
selected households, describing the survey briefly and announcing the
approximate time period for the data collection. Caution must be exer-
cised with this approach, however, as, in some contexts, depending on local
conditions and customs, such notices may create suspicion and hostility.
477. Where content is concerned, national offices should assess the sensitivity
of the topics that would be covered in the publicity materials. In some communities,
mentioning issues such as asset ownership or women’s empowerment may create nega-
tive reactions and lead to non-participation in the survey. If these issues are perceived
as sensitive, the publicity materials should avoid them. Instead, they may refer to non-
controversial issues, such as how the findings of the survey would provide important
information for developing policies and programmes to improve the lives of women
and men, and they must emphasize the confidentiality of the information provided.
Furthermore, it is important that the materials are translated into the local languages
of the households covered by the survey, to make sure that their messages will reach
the intended audience.

6.1.3. Role of geospatial information in supporting data collection operations

478. In surveys based on face-to-face interviews, maps and other geographi-
cal information (such as the geocodes of buildings and dwelling units) may need to
be developed for the purpose of sample selection and field administration. Compre-
hensive geographical information, including maps, are usually prepared and updated
in the course of population and housing censuses, typically by geographers, cartog-
raphers, geospatial information specialists and support staff. Census geography can
serve as the starting point for meeting mapping needs for survey purposes. When
census maps are incomplete, inadequate or substantially out of date, however, they
may need to be updated for the current exercise, or new ones may need to be produced.
For this purpose, additional geographical information from other sources (mapping
agencies, civilian and military agencies, or other agencies associated with the national
spatial data infrastructure, if one exists) should be obtained.
479. Once maps are assembled from the census and other sources, the next step
is to appraise and validate them from the standpoint of completeness, accuracy and
currency. This may be done by comparison of different maps and imageries (such as
those from satellites and aerial photography) for the same area, or through on-the-spot
field checks. Geographical information obtained from satellite imageries should be
field-validated before use in data collection operations. The knowledge and experience
of geographers, cartographers and the geospatial information specialists of regional
and local offices is important for appraisal and validation. Special training may also be
needed on use of the maps, or the updating of old listings from previous surveys, or the
collecting of new geographical information, if this is included in the survey operation.
The training of fieldworkers is especially important when data collection is carried out
with handheld electronic devices such as tablets. Geospatial information collected as
part of a survey operation could be useful for updating the national geographic infor-
mation database.

6.2. Training of field staff

480. Training plays a key role in obtaining good-quality data in household
surveys. Typically, during training, interviewers will learn about the purpose and
the structure of a survey, the key constructs and concepts used, and the role of each
question in measuring them. They will also learn (or will be reminded) about how to
128

approach communities and households, correctly select the persons to be interviewed,

and successfully schedule and complete the interviews with those persons.
481. The training of interviewers for the collection of data on asset ownership
from a gender perspective should follow similar principles, while emphasizing issues
specific to the topic. While many interviewers may have experience collecting house-
hold-level data on asset ownership, they may have no prior experience of collecting data
about asset ownership at the individual level and they may not be familiar with some
of the concepts employed such as “rights to” an asset. It is recommended that training
be designed on the premise that interviewers have little pre-existing knowledge of the
topic, and that it is always useful to cover general data collection techniques such as
approaching communities and households and successfully conducting an interview.
482. The following set of issues specific to surveys on asset ownership should
be emphasized during the training: first, what assets are measured and how they are
defined; second, how ownership rights to assets are defined and measured; third,
guidance on how to refrain from assuming answers based on gender-biased presump-
tions on ownership of assets; fourth, systematic guidance on refraining from assum-
ing answers to questions on ownership when moving from one type of ownership to
another, and refraining from attempting to reconcile responses in the field when more
than one person is interviewed in the same household; fifth, delivering the statement
of the purpose of the survey; and sixth, how to select eligible respondents.
483. This section includes two parts. The first addresses training on the paper
questionnaire, including the content of the training and training approaches. The sec-
ond addresses training on CAPI-specific issues. In countries using a CAPI question-
naire, the training should be organized in two parts, starting with training on the
paper questionnaire, followed by training on the CAPI questionnaire. Training on the
paper questionnaire should cover topics such as overview of objectives of the survey;
content of the questionnaire, including organization of the questionnaire, key con-
cepts and a detailed review of the questionnaire; and operational procedures regarding
approaching communities and households, identifying eligible household respond-
ents for interview and conducting successful interviews. For this part of the training,
as much as one week should be reserved in a stand-alone survey. Training on the CAPI
questionnaire should follow and cover CAPI-specific issues, while recapitulating the
key points made during training in the paper questionnaire. This part of the training
may also continue for as long as one week. Lastly, separate training sessions for super-
visors should be scheduled.

6.2.1. Training on the paper questionnaire

484. The section covers the content of training on the paper questionnaire as
well as different approaches that can be used to ensure that training materials are fully
grasped by supervisors and interviewers. If the survey is conducted using CAPI, addi-
tional training on the CAPI platform should also be provided (section 6.2.2).

6.2.1.1. Content of the training

485. The following content should be covered in the training on the paper ques-
tionnaire.
486. Overview of objectives of the survey: Understanding the objectives of the
survey will enable the interviewers to introduce the survey to the respondents and
answer their questions with confidence. If applicable, a session on the objectives of the
survey should discuss how asset ownership data have been routinely collected in the
past at the household level (highlighting that some interviewers participating in the
 Guidance for implementation 129

training may have that experience) and explain why it is important to collect these data
at the individual level—reaffirming the policy relevance of measuring asset ownership
from a gender perspective, as presented in the introduction of these Guidelines.
487. Introduction of key concepts: Two sets of issues lie at the core of the train-
ing on key concepts: first, which assets are covered in the survey and how are they
defined; and, second, how ownership of assets is defined and measured. Understand-
ing these issues is key to ensuring the accuracy of data collected. The main objective of
a session covering key concepts is to familiarize interviewers with these concepts, so
that they recognize them and understand why they are being asked during the detailed
review of the questionnaire, later in the training. A session on this topic should first
define “assets” and introduce the interviewers to the range of assets covered by the
survey. The session should then cover the concepts related to asset ownership. These,
as explained in chapter I of these Guidelines, relate to types of ownership—including
reported ownership, documented ownership, the right to sell and bequeath the asset;
forms of ownership—exclusive and joint ownership; modes of acquiring assets; and
valuation of assets.
488. Detailed review of the questionnaire: The detailed review of the question-
naire may be covered in one or more sessions, depending on the length of the ques-
tionnaire. During these sessions, the interviewers should learn about the structure and
organization of the questionnaire, the way in which the different sections relate to
the survey objectives, and the purpose of each question. With regard to ownership
questions, interviewers should receive systematic guidance on the importance of not
assuming answers relating to ownership, when moving from one type of ownership to
another, and of not attempting to reconcile responses in the field, when more than one
person is interviewed in the same household. Interviewers should also understand the
logic of the filter questions and skip patterns, be aware of which questions may require
more probing, how to engage in such probing without prompting or suggesting the
response to the respondent, and of what strategies to pursue when some questions are
perceived as sensitive, including by emphasizing the confidentiality of information.
During these sessions, the facilitators should emphasize possible stereotyping or pre-
conceived ideas on the part of interviewers that may affect the way in which questions
are asked and the answers are noted, including, for example, the assumption that the
man is the owner of all the assets in the household, or that the woman is, by default, a
joint owner.
489. Approaching communities and households and conducting successful
interviews: Training in this area is crucial to obtaining high response rates from the
households and respondents in each household. Topics to be covered should include
avoiding conflicts at the community and household levels, in the context of soliciting
sensitive information; introducing the survey to respondents, including awareness-
raising and consent forms; building a rapport with respondents; interviewing selected
respondents alone; scheduling callbacks, if selected respondents are not available
when first approached; and dealing with difficult scenarios, such as how to respond if
respondents refuse to participate. The training may stimulate discussion and partici-
patory formulation of solutions that could be employed in the field, as was the case in
the Uganda EDGE pilot.164 164 Kilic and Moylan, MEXA, 2016.

490. All interviewers should be able to deliver the statement of purpose of the
survey, which should be read or presented to all respondents in all households sam-
pled. All EDGE pilots emphasized the importance of doing this, and also the problems
faced by some interviewers in delivering it. The statement of purpose may refer to non-
controversial issues, for example, how the findings of the survey would provide impor-
130

tant information to the Government for the development of policies and programmes
to improve the lives of men and women; it should emphasize the confidentiality of the
survey; indicate how the particular household was selected for the survey; and stress
that the interview should be conducted alone, without family or neighbours present.
Issues known to create negative reactions in the community, including for example
women’s empowerment or asset ownership (in particular the ownership of land), and
which may negatively influence participation in the survey, should not be mentioned.
491. Identifying eligible household respondents for interview: This ses-
sion should provide a detailed explanation of how interviewers will identify eligi-
ble respondents in each household. Interviewers should understand the difference
between the household questionnaire and the individual questionnaire and who is eli-
gible to complete each one. As mentioned before, the household questionnaire will be
completed by a single person, ideally, a person knowledgeable about the topics covered
in the household questionnaire, including for example, sociodemographic character-
istics of the household members or, where applicable, the assets owned by members of
the household. The respondents for the individual questionnaire are one or more per-
sons randomly selected in each household or all adult household members (for details,
see chapter III, section 4.2, on selecting individuals from households).
492. If the survey is administered on the basis of a paper questionnaire, and one
or more persons need to be randomly selected, sufficient time should be dedicated to
learning how to correctly apply the selection method used in the survey.
493. Supervisor training: Field supervisors will need to understand all aspects
of the interviewing and data collection processes and must therefore receive extensive
training so that they are prepared to manage the workload and to monitor and support
interviewers during the data collection process. In some countries and surveys, the
training of supervisors precedes interviewer training, in others the supervisors will be
trained at the same time as interviewers. It is very important, however, to plan separate
training sessions for field supervisors.
494. The following topics should be covered: roles and responsibilities in the
field; process for distributing assignments; protocols for non-responsive households;
data quality assurance, including the importance of reviewing questionnaires to
ensure that all questions have been asked and answers have been recorded; steps to
take if mistakes are found in completed questionnaires; how to deal with problems
that may arise in the field; maintaining contact with the national statistical office
headquarters, and so forth.
495. If countries are collecting data based on a CAPI questionnaire, the sessions
targeted to the supervisors should be scheduled after the CAPI-specific training has
been conducted and some should be focused on how to use electronic tools to manage
the workload of their teams and perform quality checks.

6.2.1.2. Training approaches

496. To ensure that interviewers and supervisors fully understand the objec-
tives of the survey, the questionnaire, and their roles and responsibilities in the field,
three main types of training approaches should be used. First, interactive plenary
presentations and discussions ensure that all key elements regarding the content of
the questionnaire have been emphasized by the facilitators and understood by the
interviewers. Second, role play and mock interviews are an effective technique for
developing interviewing skills and the ability to recognize and react effectively and
professionally to challenging situations. They provide opportunities to observe and
rehearse a variety of scenarios and to discuss strategies for meeting challenges. Finally,
 Guidance for implementation 131

field practice is key to enabling interviewers to experience typical interview situations

in the process of survey data collection.
497. In countries where multiple languages are used for data collection, the sta-
tistical offices may find it useful to translate the questionnaire and the manuals into
those languages. At a minimum, a glossary should be prepared with key terms used
in the questionnaire. During training, after the concepts have been understood in the
primary language, additional explanations and exercises in other major languages
should be considered.
498. Quizzes may be used to evaluate interviewers’ learning and, where appro-
priate, to select the top performers if more interviewers participate in the training than
are needed for the field operations, or for the purpose of identifying supervisors.

6.2.2. Training on CAPI-specific issues

499. Training in CAPI-specific issues should be conducted immediately after
the training on the paper questionnaire. The training on CAPI should focus on all
practical steps that interviewers will typically have to follow in the field, starting from
signing into the software application used, accessing their work assignment, using the
information given in the work assignment to identify the next household to be visited,
completing the questionnaire, transferring the completed questionnaires to supervi-
sors, and signing out of the application.
500. Training on how to complete the CAPI questionnaire should be allocated
sufficient time (three to five days for a stand-alone survey). During this time, inter-
viewers should become familiar with the display of the electronic questionnaire on the
device and the formulation and flow of questions. They should know the significance
of the different formatting standards used across the questionnaire, including for the
information that needs to be read, information that contains instructions for them and
potential probing questions or further explanations, and error messages. Key concepts
learned during the training on the paper questionnaire should be recapped during
the CAPI training and the similarities and differences in the formulation of ques-
tions in the paper questionnaire versus the CAPI questionnaire should be emphasized.
If applicable, the feature of automatically selecting one or more persons to be inter-
viewed should be explained.
501. Additional training for supervisors is crucial. It will emphasize their role
in managing the workload in the field and will enable them to use specific case man-
agement tools. The following practical steps should be covered: how to synchronize
with the headquarters server to receive the workload for which they and their team
are responsible and to transfer completed and supervisor-approved questionnaires;
how to review the questionnaires completed and submitted by the interviewers and to
decide whether they should be approved and sent to the headquarters or rejected and
sent back for proper completion to the interviewers; how to troubleshoot problems that
interviewers may have with their devices; and how to manage the assignments and
reassignments of the total workload, so that all interviews planned to be completed
have indeed been completed, reviewed, and approved by both the supervisors and the
headquarters, by the end of the fieldwork period.

6.3. Fieldwork
502. This section discusses aspects of the fieldwork that are typical of house-
hold surveys, including workload distribution, information and management flow,
and quality assurance during field observations. It also addresses the interviewing
132

protocol that should be used in surveys collecting data on asset ownership, including
specific issues such as identifying eligible respondents in the household, interviewing
respondents alone, and interviewing respondents consecutively when more than one
person per household is interviewed.

6.3.1. Workload distribution, and information and management flow

503. In order to establish efficient and effective survey control, a management
information system or survey management system should be set up, and measures
must be taken to ensure that the workload is distributed efficiently.

6.3.1.1. Setting up a management information system

504. An element that is key to the successful monitoring of field activities is
an efficient management information system that integrates all types of information
required for field operations and connects multiple members of the project team. The
system relies on the continuous exchange of information between the coordinating
office and field supervisors and between supervisors and interviewers. It enables the
coordinating office to evaluate if the field operations are proceeding according to
schedule, to make necessary adjustments to the distribution of the workload across the
teams, and to correct some of the problems detected in the field, almost in real time.
505. Close control over the flow of materials and information to and from the
field is essential for a rapid and orderly progress of data collection in the field. This
task is usually performed by some central administrative unit in the statistical agency.
This central unit is responsible for sending instructional and training materials, blank
forms and questionnaires and other necessary supplies to field personnel. This same
unit should be responsible for receiving completed questionnaires and other materi-
als from the field. Records must kept of what has been sent and to whom and what
has been received and from whom. Of particular importance is a control record iden-
tifying each and every household in the sample that should be interviewed and the
outcome of the interview. Where regional offices exist, the material and information
can flow back and forth between the central office and supervisors and interviewers
through these offices. It is important that each intermediate channel (regional officer
or supervisor) maintains careful control records of its own.
506. A convenient way of controlling survey materials is to prepare a folder for
each final sample cluster, identifying the corresponding geographical unit and indi-
cating the supervisor or interviewer to whom it is assigned. This folder can include the
relevant maps, the blank questionnaires and forms to be used and any special instruc-
tions. When the completed materials are received back by the administering unit, they
should be checked against the control records. The most important step is to account
for every ultimate sample household either as interviewed or not interviewed for some
specific reasons. Where there are discrepancies, the matter must be followed up imme-
diately with the field personnel.

6.3.1.2. Methods of communication and feedback loops

507. The management information system may use a variety of methods of
communication and feedback loops, including telephone helplines, instant messages
or text messages, online forums, or social media. For example, in the Mexico EDGE
pilot, the National Institute of Statistics and Geography employed a web-based plat-
form (forum) to facilitate communication between supervisors in the field and the
central and state office staff involved in the survey. The forum is a standard approach
in surveys implemented by the National Institute of Statistics and Geography. It is
 Guidance for implementation 133

used to access survey materials, such as manuals, coding catalogues, training presen-
tations or other support materials, informant collaboration certificates, work sched-
ules and progress reports. It is available continuously during the field operations and
accessed on a regular basis by both supervisors and office staff to ensure close to real-
time responses to supervisors’ queries and optimization of the fieldwork.

6.3.1.3. Survey management system in surveys using CAPI

508. Alternatively, in countries using a CAPI questionnaire, a survey manage-
ment system could be designed to control the flow of information through the use
of electronic tools available for each level of survey staff, including on the handheld
electronic devices of interviewers and supervisors and on the headquarters’ computers
to which the field managers have access. Data and information received from the field
through the synchronization procedures can be easily aggregated and used to generate
progress reports feeding into the survey management system. Thus, one of the biggest
advantages of such a system is the ability to continuously track the progress of data
collection and identify immediately the teams that encounter challenges in the field.

6.3.1.4. Workload distribution

509. The workload for each interviewer may be decided by the coordinating
team of the survey, regional offices involved in field operations, or the field supervi-
sors. Before starting the fieldwork, interviewers should have available all necessary
materials and information, including a list of households where they should conduct
interviews and rules for the development or updating of such lists. In cases where a
recent census has been conducted in a country, the list of households to be approached
by interviewers in an enumeration area may already have been established by the coor-
dinating unit. When the census is not recent, however, the listing of households in the
enumeration area may need to be developed or updated. The new listing will be the
basis on which the set of households to be interviewed will be extracted or updated,
respectively. The supervisors may also have a role in assigning interviews to the inter-
viewers, in particular when the field staff is organized into small teams and when more
than one person is interviewed in each household. As the fieldwork is progressing,
some interviews may also be reassigned by supervisors or the coordinating unit, in
order to complete the fieldwork in the time allocated for data collection.

6.3.2. Interview protocol

510. This section aims to provide practical guidance on interview protocols
that follow the principles and recommendations outlined throughout the Guidelines,
including respondent selection procotols, sampling strategies to selecting individual
respondents, and minimizing contamination bias when multiple respondents within
the households are selected for interview. The discussion also takes into considera-
tion situations under different data collection strategies (minimum set of questions,
appending modules or stand-alone surveys).

6.3.2.1. Identifying eligible respondents in the household

511. After successfully approaching a household, the interviewer must then
identify the members of the household who should be interviewed. The basic rules for
selecting the eligible respondents in each household should be formulated in advance
and emphasized during the training, so as to ensure that interviewers encounter no
problems when following them in the field. The following discussion highlights some
134

considerations for the data collection strategies and related respondent selection pro-
tocols described earlier in these Guidelines.

6.3.2.2. Minimum set of questions or appended module administered

to one randomly selected adult household member
512. When data on asset ownership are collected through a minimum set of
questions or a module attached to a main survey, and one adult respondent is ran-
domly selected for interview, the eligible respondent will only be identified after the
household questionnaire has been completed and a listing (or roster) of all the house-
hold members has been obtained. The household questionnaire is typically collected
during the main survey and the respondent who completes the household question-
naire is selected on the basis of the rules established in the main survey. The listing of
household members should include information on the age and gender of the house-
hold members, among other characteristics, as it is the basis used by the interviewer
(or by a software routine when using CAPI) to select the adult household member for
interview on asset ownership. A method of selection such as the Kish method should
be consistently applied across all households to ensure a probabilistic selection of the
respondents, as explained in section 4 above, on sample design.

6.3.2.3. Appended module administered to more than one adult household member
513. When an appended module is administered to more than one adult house-
hold member, respondents eligible for individual interviews will only be identified
after the household questionnaire has been completed, as described above. The listing
of household members should include information on the age, gender and marital
status of the household members, as this is the basis for selecting either one randomly
selected adult household member and that member’s spouse or partner or one couple
and an additional randomly selected adult household member for individual inter-
views, as presented in section 4 on sample design.
514. At the same time, however, a roster of assets will also need to be collected
at the household level, as discussed in chapter III, sections 2 and 4, that the interviewer
will then feed forward into the individual interviews. When the roster of assets is con-
structed at the household level and incorporated in the household questionnaire, it is
important to specifically choose as the respondent a household member who is knowl-
edgeable about assets owned by all household members. That person should be identi-
fied by the interviewer immediately after establishing contact with the household and
obtaining the members’ consent to conduct an interview. A simple direct question
may be used, such as “Who in the household is most familiar with (or knowledge-
able about) assets owned by all household members?” If that person is different from
the respondent selected to complete the household questionnaire for the main survey,
then efforts should be made to build the household roster of assets with the knowl-
edgeable member, upon completion of the household questionnaire.

6.3.2.4. Stand-alone survey administered to all adult household members

515. When a stand-alone survey on asset ownership is implemented, a household
questionnaire comprising a household roster of respondents and a household roster of
assets should be administered to one adult household member and the individual ques-
tionnaire on asset ownership should be administered to all adult household members.
516. As described above, when the roster of assets is constructed at the household
level and incorporated in the household questionnaire, it is important that a household
member who is knowledgeable about the assets owned by all household members is
 Guidance for implementation 135

chosen as the respondent. The entire household questionnaire should be administered

to that person.
517. The individual-level questionnaire, which includes questions related to
asset ownership that require self-reporting, as discussed in section 4, will be adminis-
tered to all household members aged 18 and older. The suggestion that all adult house-
hold members should be interviewed is aimed at reducing costs when a stand-alone
survey is conducted. Other options of respondent selection are possible, such as inter-
viewing a couple and a third randomly selected adult if there is a sufficient number of
households. A more extensive discussion of within-household respondent selection
may be found in section 4. The interviewer will be able to identify the persons to whom
the individual-level questionnaire is to be administered only after the household ques-
tionnaire has been completed and a roster of all the household members has been
obtained. This roster includes information on the age and gender of each household
member and it is the basis used by the enumerator (or by a software routine when
using CAPI) to select the adult household members for interview.

6.3.2.5. Interviewing respondents alone

518. Questions about asset ownership can be sensitive and interviewers should
make every effort to establish an interview setting that is conducive to disclosure of
information. It is very important that interviews on the individual questionnaire be
conducted with each respondent alone. When other people are around, the respond-
ent may be less inclined to report ownership of certain types of assets or other people
may try to influence the respondent’s responses. For example, in the EDGE pilot in
the Philippines, when wives were interviewed while their husbands were at home, the
husbands often tried to intervene in the interviews. Even neighbours may be curious
and want to be present at an interview, in particular if the interview is being imple-
mented on a handheld CAPI device with which community members are not familiar.
A good deal of tact on the part of the interviewer will be required to operate in those
kinds of situations, and tips for conducting interviews alone should be discussed dur-
ing the training of field staff. For example, the interviewer should state clearly that
the interview needs to be conducted alone and encourage the respondent to ask other
persons to leave the interview setting, if possible. In some cases, the interviewer may
choose momentarily to stop asking questions until all other persons have withdrawn.

6.3.2.6. Interviewing multiple household members consecutively

519. When the sample design requires more than one adult household member
to be interviewed, countries should aim to interview respondents consecutively. As
discussed earlier in these Guidelines, consecutive interviewing mitigates the potential
contamination of answers from one respondent to another in the same household; its
success will largely depend, however, on the number of respondents to be interviewed
per household and the number of interviewers deployed per household.
520. When two household members, such as one randomly selected respondent
and that respondent’s spouse or partner are interviewed, it should be fairly easy for
one interviewer to interview the respondents consecutively, assuming they are both
home at the same time. If three respondents are to be interviewed, such as the mem-
bers of a couple and an additional, randomly selected adult, two interviewers will have
to be dispatched to the household and the task of ensuring consecutive interviewing
may be further complicated by the need to match the genders of the interviewer and
respondent. If any more adult household members are interviewed, for example, when
a stand-alone survey is being conducted and all adult household members are to be
136

interviewed, consecutive interviewing of all adult respondents will not be possible. In

addition, in many instances, multiple respondents will not be home at the same time
and available for consecutive interviewing when the interviewer calls at the household.
In such cases, together with their supervisors, interviewers will have to decide whether
to schedule a return visit to the household when interviews can be conducted consecu-
tively or to proceed to interview the respondent who is at home and forgo connective
interviewing. While there are no set rules, field staff should be trained in these issues
and data specialists should be aware of the potential biases introduced to the data
when consecutive interviewing is not possible.

6.3.2.7. Gender matching

521. Gender matching between the interviewer and the respondent may also be
important in some countries. As explained earlier, an assessment should be made by
the core team in charge of planning and implementing the survey regarding whether
gender matching would have an impact on the level of household non-response and
the quality of responses obtained. If the assessment concludes that gender matching
is important, teams comprised of men and women interviewers should be deployed
to the same enumeration area, in order to interview women and men respondents. In
this case, supervisors will have a key role in efficiently assigning households to differ-
ent team members.

6.3.2.8. Callbacks
522. Every effort should be made by interviewers to successfully interview all
the households in the sample. If a significant proportion of households are omitted
by interviewers, the probability sample chosen may lose its representative character,
because households (and individuals) that are not interviewed may have different
characteristics from those interviewed. Even if procedures exist for the adjustment of
non-response (as presented in chap. IV, sect. 1.4, on weighting), the survey results are
165 United Nations, Handbook likely to be biased if more than a small percentage of cases is omitted.165
of Household Surveys, 1984.
523. One of the most frequent reasons for non-response at the household level is
the inability to find anyone at home to interview. The usual approach in this instance
is to plan a return visit on a different day or at a different time of day when it is likely
that somebody in the household will be present. Interviewers may be able to obtain
information from neighbours or landlords about a more suitable time for a repeat visit.
It is recommended that at least two repeat visits be attempted, unless excessive costs
166 Ibid. are involved.166 Statistical agencies should follow their typical protocol with regard to
the number of callbacks allocated for each household and the amount of time that a
team can allocate to an enumeration area. In the EDGE pilot surveys, for example,
the interviewing protocols required the interviewers to have a minimum of three call-
backs before considering a case to be a non-response. Information on the reasons for
non-response was also collected.
524. There may be cases where respondents refuse to be interviewed. Inter-
viewers should attempt to provide further explanations about the survey’s purpose,
and reiterate the guarantee of confidentiality. They should also offer to schedule or
reschedule interviews at a more appropriate time. Interviewers should also refer the
case to the field supervisor, who may then make an attempt to obtain cooperation.

6.3.3. Quality assurance during field operations

525. Quality management in any household survey should be comprehensive
and applied at all stages of survey implementation, including planning, sample design,
 Guidance for implementation 137

questionnaire design, field operations, data processing and analysis. Each stage of sur-
vey implementation plays an important role in obtaining quality data. Field operations,
in particular, have a great impact on the accuracy of data collected, which is defined
as “the degree to which the information gathered correctly describes the phenomena
it was designed to measure”.167 Training and quality control mechanisms during the 167 Statistics Canada, Statistics

fieldwork (such as the supervision and monitoring of interviewers’ activities) enable Canada Quality Guidelines,
the achievement of key quality standards, including high response rates from house- 4th ed. (Ottawa, 2003).
hold and individuals (household response rate), high response rates for all questions
asked (item response rate) and consistency in asking questions and recording answers.
Achieving these standards is very important. Household non-response will affect the
statistics obtained in the survey, because households that are not immediately avail-
able for an interview or that refuse to participate may differ from those that agree to
participate, in terms of characteristics that are relevant to the topic of the survey. In
other words, the sample will not be representative of the population from which it was
derived. Item non-response and errors in asking the questions or coding the answers
received will also affect the accuracy of the data, because they may provide a partial or
biased representation of what the survey tried to measure.

526. Quality assurance procedures may be used before the fieldwork starts and
during the process of field data collection, while assessments of the quality of data col-
lection may be conducted during the data collection exercise or after it has concluded.
Various strategies may be involved in reducing non-response rates before the fieldwork
starts. To ensure high item response rates and consistency in administering the ques-
tionnaire, training of the interviewers and supervisors plays a crucial role, as described
in section 6.2 of this chapter. In order to achieve high household response rates, sev-
eral strategies may be employed by national statistical offices in addition to training,
including the use of publicity (see sect. 6.1.2), sending advance letters to inform poten-
tial respondents about the upcoming survey, and offering incentives.

527. During data collection, field supervisors have a key role to play in quality
control, as they monitor all aspects of data collection in the field on a daily basis, check
the questionnaires submitted by the interviewers for completeness and errors in cod-
ing answers and may observe some of the interviews. Supervisors also play a key role in
identifying interviewers who underperform, and providing a solution by supporting or
replacing those interviewers. For example, item non-response relating to questions that
are not answered may occur as a result of the respondents’ lack of knowledge or under-
standing about the questions, which cannot be addressed while in the field. But it may
also indicate that interviewers have skipped questions or respondents have become
disengaged because the interviewer failed to create a rapport with them.

528. Observation of interviewers is particularly important in a survey on asset

ownership, and supervisors should schedule this activity at the beginning of the field-
work. If working only on the basis of completed questionnaires, it would be difficult to
judge whether interviewers have correctly asked the questions about different types of
ownership or whether, on the contrary, they have assumed information about owner-
ship or tried to reconcile answers with information on asset ownership obtained from
other household members.

529. In this regard, countries may use an additional mechanism for quality con-
trol, in particular when they are conducting a data collection exercise on asset owner-
ship at the individual level for the first time. For example, in some EDGE pilot studies,
teams of national statistical office headquarters staff who were participating in the
training process as trainers and facilitators were deployed during the first week of data
138

collection to observe the teams in the field, discuss challenges and provide additional
guidance where necessary.
530. Lastly, supervisors should make sure that field assignments progress as
planned and all respondents are reached and interviewed, which would ensure a high
household response rate. Interviewers are responsible for correctly identifying house-
holds and individuals who should be interviewed, and ensuring their participation in
the survey, including through introduction of the survey. When the respondents are
not immediately available for interview, interviewers should be flexible in schedul-
ing or rescheduling interviews. Their efforts to this end may not always be success-
ful, however. In cases where respondents are quite reluctant to participate, supervisors
may step in. In addition, a sample of each interviewer’s work should be spot-checked to
verify that the interview has been implemented only in the households in the sample.
531. At the end of the fieldwork, countries should make an overall assessment
of its quality, based on supervisors’ quality checks and observations during the field-
work. They may also consider conducting repeat interviews on a small subsample of
households and comparing the responses obtained with those obtained during the
field data collection for the same subsample. This will give an indication of response
reliability and help in assessing the work of particular interviewers. For example, in
the EDGE pilot in Georgia, 12 per cent of the household sample, or two households per
interviewer, were reinterviewed after the fieldwork by staff from the national ­statistical
office. The reinterview survey covered both urban and rural areas and used a subset of
items from the questionnaire administered during the original data collection.

Key points
•• It is important to assess whether gender matching between interviewers and respond-
ents is required, given that the data collection processes covered in these guidelines
focus on measuring asset ownership from a gender perspective and that both women
and men are going to be interviewed. If the assessment concludes that gender match-
ing is important, teams comprised of women and men interviewers will have to be
deployed to the enumeration areas to interview women and men respondents.
•• The following issues specific to surveys on asset ownership should be emphasized
and guidance provided during the training:
•• Which assets are measured and how they are defined;
•• How the ownership rights to assets are defined and measured;
•• Refraining from assuming answers based on gender-biased preconceptions on
ownership of assets;
•• Refraining from assuming answers to questions about ownership when moving
from one type of ownership to another, and refraining from attempting to recon-
cile responses in the field when more than one person is interviewed in the same
household;
•• Delivering the statement of the purpose of the survey;
•• How to select eligible respondents.
•• Special interviewing protocols should be used in surveys aimed at collecting data on
asset ownership, so as to cover such areas as identifying eligible respondents in the
household, interviewing respondents alone, and interviewing respondents consecu-
tively when more than one person per household is interviewed.
 139

Chapter IV
Data processing, analysis and dissemination

1. Data processing
532. Data processing refers to a range of activities aimed at converting informa-
tion collected in the field through a survey instrument into a database that can be used
for tabulation and analysis of data. Typically, it encompasses data entry, data editing,
data imputation and assessments of the data quality and the precision of the survey.
Data processing has an impact on the quality of the final survey results, and its efficiency
is key to obtaining those results within a reasonable time period after data collection.
533. The overall processing plan should be developed early in the stage of plan-
ning the survey, and the data managers involved should have a good understanding of
the survey’s objectives and the questionnaire design. They will be key decision makers
with regard to the data processing activities to be implemented, the timetable for each
activity, the required personnel, equipment and computer software packages, and how
best to organize the data in an electronic format.
534. Countries are encouraged to follow their typical protocol for data process-
ing in household surveys when conducting a survey on asset ownership and control
from a gender perspective. Two aspects of data processing need particular attention,
however. First, the data structure is complex, dealing with several statistical units of
observation and analysis, including households, individuals and assets (for those assets
that are itemized, such as land, large agricultural equipment, other real estate, finan-
cial assets and unincorporated enterprises). It is important that the structure of the
survey data set reflect the hierarchical relationships between the different statistical
units, minimize the storage requirements, and interface well with statistical software
at the analytical phase (see sect. 1.1, on data entry and organization of the data sets).168 168 Juan Muñoz, “A guide for data
Second, it is important that no gender bias be introduced in any of the data processing management of household
steps, including data adjustment activities to increase the consistency of data, and that surveys”, in United Nations,
the survey sample results be representative of the population targeted (see sects. 1.2, Household Sample Surveys
in Developing and Transition
1.3 and 1.4 on data editing, imputation and weighting, respectively).
Countries (New York, 2008).

1.1. Data entry and organization of data sets

535. Data entry refers to the transfer of the information collected on the paper
questionnaire into an electronic file that can be used for data tabulation, analysis and
sharing. A more general term for the process is data capture, which, in the case of a
paper questionnaire, may consist of manual data entry by clerical staff, mark-character
recognition and optimal-character recognition. In computer-assisted interviewing, the
processes of interviewing and electronic data capture occur simultaneously, which is
one of the advantages of using computer-assisted interviewing.
536. In surveys using paper-based data collection, data entry may be done cen-
trally or in the field. Data entry in the field, by enumerators, with supervisors still in
an enumeration area, can improve the quality and timeliness of the data. Errors and
140

inconsistencies identified during data checking and entry may be resolved by revis-
iting households in that area. By comparison, office data entry may not accurately
reflect the realities observed in the field. Data capture in the field may also shorten the
process of preparing data for tabulations and analysis. Field staff must, however, be
organized into teams that can ensure both data collection and data entry, and these
teams must be trained in data entry before the start of data collection. The approach
also requires that the data entry and editing programs be developed, tested and final-
ized before field operations commence.
537. Data entry programs and the structure of the resulting data sets should be
carefully considered. A household survey on asset ownership and control is a complex
survey that collects information about a major statistical unit—the household—along
with a variety of subordinate units within the household—individuals (persons) and
itemized assets, including land, large agricultural equipment, other real estate, finan-
cial assets and liabilities, and unincorporated enterprises. Data entry and editing pro-
grams should be able to adequately handle this complexity and provide additional
capabilities related to data tabulation, data analysis and conversion of data files into
the most commonly used statistical programs, such as CSPro, SAS, SPSS, R and Stata.
538. The structure of the survey data sets must reflect the hierarchical relation-
ships between the different statistical units in a survey on asset ownership and control.
Most of the questions on asset ownership refer to subordinate statistical units that
appear in variable numbers within each household, including individuals and differ-
ent types of assets. Individuals are both units of observation and of analysis nested
within the households units, and the assets held by individuals are both units of obser-
vation and of analysis nested within the individual units. The number of individuals in
a household varies across households and the number of assets owned by an individual
varies across individuals and households.
539. The data corresponding to the three units of observation and analysis
(households, individuals, assets) should not be stored in one simple rectangular file
(called a “flat file”), with one row for each household and columns for each of the fields
on the questionnaire. A flat file is adequate only if all the questions refer to one statisti-
cal unit (e.g., the household), but this is not the case in a survey that measures asset
ownership at the individual level. Storing information related to individuals (persons)
and assets at the household level would be wasteful and extremely cumbersome at the
analytical stage.
540. Instead, the data structure should maintain a one-to-one correspondence
between each statistical unit observed and the records in the computer files, using
a different record type for each kind of statistical unit. For example, to manage the
data listed in the household roster, a record type should be defined for the variables in
the roster and the data corresponding to each individual should be stored in a sepa-
rate record of that type. A similar logic should be applied to the roster of assets. For
example, to manage the data in the agricultural land module, a record type should be
defined to include the variables in the module, and the data corresponding to each
parcel of land should be stored as a separate record.
541. Across the data sets, each record should be uniquely identified by a code
in four parts:
(a) Part 1 denotes the record type and appears at the beginning of each record;
it indicates whether the information is from the cover page, the household
roster, or one of the asset modules (e.g., agricultural land module or finan-
cial asset module);
(b) Part 2 refers to the household number;
 Data processing, analysis and dissemination 141

(c) Part 3 refers to the household member’s identification number;

(d) Part 4 refers to the code of the particular asset item.
542. The survey data sets may need to be organized as separate flat files, one for
each record type, for dissemination purposes. For example, one flat file would com-
prise records of all the parcels of agricultural land owned by the respondents, while
another flat file would comprise records of all financial assets owned by the respond-
ents in the survey. The identification codes for the individual records described above
will enable the linking of the data across the flat files.
543. When using a CAPI questionnaire, the organization of data follows the
structure of the CAPI questionnaire design, reflecting the specified hierarchy and
nesting of the different units of observation. After data collection has concluded and
all the questionnaires have been delivered to headquarters, data may be exported into
formats that are compatible with the most commonly used statistical programs. For
example, when using Survey Solutions (as was used in the EDGE pilots in Uganda and
South Africa), the data may be exported in .sav (for SPSS), .dta (for Stata) or .tab files.
There should be one file for each distinct level of hierarchy and unit of observation in
the questionnaire.

1.2. Data editing

544. The files obtained at the end of data entry may be further checked and
improved by means of data editing, imputation and weighting, for the purpose of
obtaining a database that can be analysed or shared.
545. A field review of questionnaires by supervisors (see chap. III, sect. 6),
should identify and rectify many of the problems and errors in the data. At the same
time, some further checking is nearly always needed at the data-processing stage to
catch remaining errors. The general types of content errors likely to be found at this
stage are omissions, inconsistencies across different questions and variables, unrea-
sonable entries and impossible entries.169 169 United Nations, Handbook
of Household Surveys, 1984.
546. Generally, the best procedure is to resolve the problem on the basis of the
information in the questionnaire. Some errors may have occurred in the course of
data transfer and the correct information may actually appear in the questionnaire. In
other cases, the information that appears to be incorrect may be corrected in the light
of other information recorded in the questionnaire. For example, omissions for the
variable on gender may be resolved from the name of the person, or omissions in age,
from information on date of birth. Inconsistencies can sometimes be resolved by con-
sidering the whole range of information and deciding which of the conflicting entries
appears most reasonable. For example, based on data on education, marital status and
occupation, it would seem logical that a 13-year-old tertiary-educated married person
working as a wage employee is more likely to be 31 years of age.
547. Using information from the questionnaire to resolve errors is the procedure
typically applied to variables relating to the demographic characteristics of household
members. It should also be applied to questions on asset ownership and control when
errors have occurred during data transfer and the correct information may be found
in the questionnaire. This procedure should not be applied, however, when it comes
to ambiguities, inconsistencies or omissions relating to types and forms of ownership.
Different answers given to questions on different types of ownership for a given asset
should not be treated as inconsistencies. Similarly, omissions of answers to questions
on certain types of ownership (e.g., documented ownership) should not be remedied
142

by reference to information on other types of ownership (e.g., reported ownership and

the rights to sell or bequeath).
548. Data editing processes should also aim to obtain a clear distinction
between sample units (households and individuals) that are respondents, eligible non-
respondents, ineligible units or non-responding units of unknown eligibility. A clear
account of the status of each household and individual in the sample is required for the
computation of survey weights, as described in section 1.4 on weighting. For the pur-
pose of weighting, it is also important to ensure that the sampling information—such
as respondent unit (individual), primary sampling unit (household) and stratum—is
available for each respondent data record.

1.3. Imputation
549. Imputation refers to the process of placing estimated answers into data
fields that have missing information or information that is assessed to be incor-
rect or implausible. In general, decisions on whether to impute values or work with
“unknown” answer categories depend on a number of circumstances. A general rule
of thumb is to make imputations for certain basic demographic items that are essential
in analysis and where the error rates are comparatively low. Another rule of thumb is
to impute responses or values where considerable prior complementary information
corresponding to the record or observation is available, but otherwise to assign a value
of “unknown” when such information does not exist.
550. The decision as to whether to impute values has implications. On the one
hand, when the statistical agency does not use imputations, users are induced to make
their own imputations for the “unknown” categories and may do so on the basis of less
adequate information than is available to the statistical agency. On the other hand,
imputed values may be perceived as made-up data, regardless of how well the imputa-
tion might be carried out.
551. Examples of imputations typically used include replacing the missing
value with the mean or median for that variable, preferably at the level of a population
subgroup with characteristics similar to those of the respondent with missing values,
or a value estimated on the basis of a regression procedure; and borrowing a replace-
ment value from a case in the data set that is most similar to the case with missing data
on a set of relevant variables (a process known as “hot-deck” imputation). In any of
these methods, it is particularly important that the estimates of replacement values be
based on responses from respondents of the same gender as the respondent with miss-
ing data. In addition, when subgroups of a population or a regression-based method
are considered to estimate the replacement value, individual characteristics that may
be associated with one or the other gender (such as marital status or education) should
be taken into consideration among other variables that may be relevant.
552. Alternative methods of imputation, including the use of information col-
lected from other sources of data, may be carefully considered for items with a high
non-response rate. For example, a variable crucial for the calculation of the gender
wealth gap is the value of assets owned. Yet, as shown by the results of the EDGE pilots
discussed in chapter I, valuation questions have a high non-response rate.
553. It should be noted that removing cases with missing values for variables in
the analysis is equivalent to an implicit imputation. In this case, it is assumed that the
results obtained for the respondents apply to the non-respondents as well, which is not
necessary the case. The alternative is to make the imputation explicit and transparent,
and inform the users about the method for imputation. In this case, the analyses pre-
 Data processing, analysis and dissemination 143

sented in a publication will be consistently based on the same number of cases, and all
the data collected are used in the analysis.
554. Where imputations are made, it is important to provide information on the
extent of the imputation (i.e., the proportion of item non-response) and the method
used. It is also valuable to create an imputation flag variable. Thus, tabulations can be
made with and without the imputations, and external users are given the opportunity
to decide for themselves whether they want to use imputed values or not, based on the
objectives of their analyses.

1.4. Weighting170 170 Guidance on constructing

sample weights is beyond
555. Weighting is a process to adjust for unequal probability of selection and unit the scope of these Guidelines.
non-response. Post-stratification weights could also be used to align the sample popula- Readers interested in such
tion distribution with a target population distribution. The purpose of using weights in guidance should consult
estimating asset ownership is to produce estimates that correspond as closely as possi- Groves and others, Survey
ble to the real values in the target population. When weights are not used to compensate Methodology, 2nd ed., (Hobo-
for differential selection rates and for sample imperfections due to issues such as non- ken, New Jersey, Wiley and
Sons, 2009); and Richard Valli-
response, the resulting estimates of population parameters will, in general, be biased.
ant, Jill Dever and Frauke Kreu-
ter, Practical Tools for Designing
1.4.1. Adjusting for unequal selection probability and Weighting Survey Samples
556. The first step in constructing weights is to adjust for unequal probabil- (New York, Springer, 2013).
ity of selection of sampling units at different stages of sample selection. The selection
probability is determined by the sample design, and the base weight of a respondent
is calculated as the reciprocal or inverse of its probability of selection. For example,
a respondent selected with probability 1/100 represents 100 in the population from
which the sample was drawn. Thus, the base weight assigned to the respondent would
be 100, the inverse of its selection probability. The sum of the sample weights provides
an unbiased estimate of the total number of people in the target population.
557. For a multistage stratified sample design for surveys of asset ownership
from a gender perspective, the base weights must reflect the probabilities of selection
at each stage of selection, including the last stage in which individual respondents are
selected from households.
558. As discussed in chapter III, two main within-household respondent selec-
tion protocols are recommended for measuring asset ownership from a gender per-
spective: first, randomly selecting one adult household member for interview; second,
interviewing several household members. In the first approach, the weight assigned
is the inverse of the selection probability within the household. For example, if there
are four adult members in the household, then the probability of selecting one person
would be 1/4, and the within-household weight adjustment should be 4. In the second
approach, if all eligible persons in a household are selected, no weight adjustment is
required because everyone in the household has a conditional selection probability of 1.
559. With the alternative option of selecting one couple randomly from all cou-
ples in the household and an additional person randomly from the remaining non-
coupled adult household members, the selection probability of both members of the
couple is 1 over the total number of the couples in the household. If there is only one
couple in the household, then both members of the couples have a selection probability
of 1. If there are two couples in the household, then selecting one couple gives both
member of the selected couple a selection probability of 1/2. The selection probability
for the additional adult is 1 out of the total number of non-coupled adult household
members. The weight for adults for unequal selection probability within the household
is the inverse of their selection probability.
144

560. For the within-household selection design that selects one adult member
randomly and selects the spouse of this person for interview, if present, the selec-
tion probability varies depending on whether there is a spouse to be interviewed. For
example, if there are four adult members in the household; one person is randomly
selected, and this person does not have a partner. The selection probability for this
person is 1/4. If this person has a spouse to be interviewed, the selection probability for
both members of the couple is 1/2. Weights assigned to each respondent would be the
inverse of their selection probability.
561. If more than one adult member is selected randomly from the household,
the selection probability for each can be calculated accordingly. For example, if three
out of five eligible household members are selected randomly from a household, the
selection probability would be 3/5 and the weight would be 5/3.

1.4.2. Adjusting for unit non-response

562. A second step in the weighting procedure is to adjust for unit non-response.
If this is not done, estimates may be biased in some way by the undercoverage or over-
coverage of certain groups in the target population. The magnitude of bias due to
non-response is associated with two factors: the overall non-response rate; and the
size of the difference in asset ownership between the respondent group and the non-
responding group.
563. As in many household surveys, evidence from the EDGE pilot surveys
showed that people living in urban areas tend to have higher non-response rates than
those in rural areas, and men were more likely to be non-respondents than women (as
was the case in Georgia, Mongolia and the Philippines). The Uganda pilot also showed
that respondents were on average older. A similar pattern was observed in the three
Asian pilot countries. Overall, when compared to the original selected sample, there
was overrepresentation among respondents who were women, inhabitants of rural
areas and older people.
564. To compensate for such overrepresentation due to non-response, an assump-
tion is made in survey non-response weighting that generates the same kind of weight-
ing adjustment as that discussed for the unequal probability of selection. It is assumed
that respondents are, in some sense, a random sample of the overall selected sample.
Depending on the nature of the assumption, the inverse of the response rate can be used
171 Groves and others, Survey as a weight to restore the respondent distribution to the original sample distribution.171
Methodology, 2009. 565. The variables that are used to align the distribution of the respondent sample
and the original sample should be correlated with outcome variables, and they should be
available for both respondents and non-respondents. For this reason, in surveys on asset
ownership at the individual level, the response status of each household and selected
person within the household must be carefully recorded. Data should be collected
about both responding and non-responding households in order to aid household non-
response adjustment later. Data on key sociodemographic characteristics of household
members should also be collected in the household roster before selecting individual
respondents in order to facilitate non-response adjustment at the individual level.
566. For the measurement of individual-level asset ownership, variables that are
collected on the household questionnaire are good candidates. They should include
gender, among other characteristics, such as region, urban or rural residence, age, rela-
tionship to head of household, marital status, education and economic activity.
567. Table 10 shows how non-response adjustment weight can be calculated
using auxiliary variables such as gender, education and age. Data for the table were
drawn from the Mongolia pilot survey. From the table, it is evident that younger people,
in particular young men, are less likely to respond than older people, and people with
 Data processing, analysis and dissemination 145

lower education levels are more likely to respond than people with higher education
levels. If we can assume that, within these classes disaggregated by gender, education
and age, the respondents are a random sample of all sampled persons—in other words,
the non-respondents are also missing at random—then the non-response adjustment
weights can be calculated as the inverse of the response rate for people in each class.
568. A similar weighting adjustment can be made for household-level non-
response. Variables that can be used for the adjustment are often limited to geographi-
cal location and residential area, whether urban or rural, as other information is often
not available. For this adjustment, account must also be taken of households that are
of unknown eligibility.172 172 For more information on
eligibility, see American
Table 10 Association for Public Opinion
Calculating non-response adjustment weight Research, Standard Definitions:
Final Dispositions of Case Codes
Non-response and Outcome Rates for Surveys
Auxiliary variables: Response adjustment (Chicago, 2016). Available at
gender, education and age Sample Respondents rate (Ri) weight (1/Ri) www.aapor.org/AAPOR_Main/
Men, primary or lower, 18–34 276 191 0.69 1.45 media/publications/Standard-
Definitions20169theditionfinal.
Men, high school or higher, 18–34 1 084 677 0.62 1.60
pdf.
Men, primary or lower, 35–55 445 383 0.86 1.16
Men, high school or higher, 35–55 958 770 0.80 1.24
Men, primary or lower, 56+ 209 182 0.87 1.15
Men, high school or higher, 56+ 313 285 0.91 1.10
Women, primary or lower, 18–34 175 145 0.83 1.21
Women, high school or higher, 18–34 1 174 870 0.74 1.35
Women, primary or lower, 35–55 399 370 0.93 1.08
Women, high school or higher, 35–55 1 281 1 176 0.92 1.09
Women, primary or lower, 56+ 272 236 0.87 1.15 Source: Data from the Mongolia
EDGE pilot survey, self-reporting
Women, high school or higher, 56+ 336 307 0.91 1.09
only.

Box 8
Using propensity score method to adjust for unit non-response
Another method in adjusting for unit non-response is the propensity score method. The
method uses multivariate logistic models to estimate the probability of responses for
each respondent, using a number of auxiliary variables available for both respondents
and non-respondents. Ideally those auxiliary variables are related to both the propensity
to respond and outcomes being measured. A description is provided below of the basic
steps for calculating propensity score adjustment weights, illustrated with an example
from the Philippines.
(a) Preparing the auxiliary variables: The list of variables considered for the Philip-
pines pilot data includes gender, age, primary couple or not, education, urban or rural
residence, religion, ethnicity, employment status, household size and age structure, and
whether there is a major housing asset in the household. Data for these variables are col-
lected in the household questionnaire and are available for both respondents and non-
respondents. Missing values for these variables were assessed. Values on marital status
were missing for only two persons, and these two persons were dropped from the dataset
for propensity score weight adjustment.
(b) Checking the two-way association between the auxiliary variables and the
binary variable of respondent/non-respondent: In the example, age, education, marital
146

status and employment are significant, while urban versus rural residence is only signifi-
cant at a level of 10 per cent.
(c) Calculating propensity scores: This is carried out by fitting a logistic regression
model, where the outcome variable is binary, that is 1 for respondents and 0 for non-
respondents. The model may be fit with or without base weight. It is advisable, however,
not to use weights as the model relates to response propensity among sampled persons.
While fitting the final model, a stepwise approach is suggested until all main effects and
two-way interactions are significant. The propensity score for respondents and non-
respondents will be calculated using the fitted model and their values for the auxiliary vari-
ables. In other words:
exp(∑âi xij)
^j =
The propensity score for the j th person is φ i
1 + exp(∑âi xij)
i
where x represents the value of auxiliary variables and â is the fitted coefficient value from
the logistic regression model. The fitted logistic model showed that the main effects of
age, gender, marital status, education, employment are all significant.
(d) Using estimated propensity score to create adjustment classes: Propensity is
scored from low to high, and the individuals divided into five groups based on quantiles,
so there are about the same numbers of individuals (respondents and non-respondents).
Then a table is constructed as follows for each of the five groups.

Estimated propensity scores to create adjustment class

Unweighted Weighted Unweighted 1/median

Range of Number average average Unweighted Weighted median propensity
Adjustment estimated of sample estimated estimated response response estimated (inverse of
class propensities persons propensity propensity rate rate propensity (7))
(1) (2) (3) (4) (5) (6) (7) (8)
1 [0.54, 0.88] 747 0.83 0.84 0.82 0.83 0.84 1.19
2 [0.88, 0.93] 746 0.91 0.91 0.92 0.92 0.91 1.10
3 [0.93, 0.95] 747 0.94 0.94 0.93 0.94 0.94 1.06
4 [0.95, 0.97] 747 0.96 0.96 0.97 0.96 0.96 1.04
5 [0.97, 0.99] 745 0.98 0.98 0.98 0.98 0.98 1.02

Note: column (1) shows the range of estimated propensity scores for the individuals in
each class; column (2) indicates the number of individuals in each class; column (3) is the
simple (unweighted) average of estimated propensity scores for all individuals in the class;
column (4) is the average of the estimated propensity scores weighted by the weights
produced by following the steps taken to adjust for non-responses at the individual
level; column (5) is the unweighted raw (true) response rate for each class, which refers
to the proportion of respondents within the class; column (6) is the weighted response
rate, calculated as the sum of weights for all respondents divided by sum of weights for
the sampled individuals (respondents and non-respondents); column (7) is the median of
unweighted average of estimated propensity scores for all individuals in the class.
(e) Calculating weight based on estimated propensity scores: The weights are calcu-
lated as the inverse of the estimated propensity score. In this example, the weights are cal-
culated as the inverse of the median of estimated propensity score for each class. It is also
possible to calculate the weight as the inverse of the mean estimated propensity score
(column (3)). If, however, there is a large variation among estimated propensity scores
within the same class, use of the median is preferred.
 Data processing, analysis and dissemination 147

The estimated response propensity itself can be used to derive weight for individual
^ Another way of creating weights is to group units
respondent, which is the inverse of φj.
into classes according to their estimated propensity scores. The advantage of using pro-
pensity classes is that it avoids situations where respondents with very low values of
estimated propensity scores receive large weights that can inflate the variance of survey
estimates excessively.
For further discussion on the calculation of weights using propensity scores, see Rich-
ard Valliant, Jill Dever and Frauke Kreuter, Practical Tools for Designing and Weighting Sur-
vey Samples.

1.4.3. Post-stratification weighting

569. In addition to adjusting for unequal probability of selection and non-
response in the sample, data can be weighted up to the total target population. This
procedure is called “post-stratification”, and it uses weights to assure that the sample
distribution defined by some key variables follows the same distribution as that of
some external population. Data for the variables used for post-stratification weighting
should be available for both the sample population and the external total population.
Commonly used variables include region, urban versus rural residence, gender, age,
education and economic activity.
570. Continuing with the above example from the Uganda EDGE pilot survey,
to align the distribution of the sample population with the total population in terms
of education, a post-stratification exercise is illustrated in table 11. A comparison of
the population distribution and sample distribution by education showed that for
both women and men, the sample overrepresents the group with primary education,
while it underrepresents the group with secondary education.173 Post-stratification 173 Please note that this is only a
weights are then derived to compensate for the difference between the sample and hypothetical exercise, as the
the population. population data are drawn
from the 2002 Uganda popula-
tion census.
Table 11
Illustrative example of post-stratification

Population Sample
Population distribution by Sample distribution by Post-stratification
Gender Education size education (ai) size education (bi) weights (ai/bi)

Women None 549 872 0.15 236 0.20 0.76

Primary 696 242 0.19 580 0.49 0.39

Secondary 2 194 102 0.60 298 0.25 2.41

Higher 206 541 0.06 79 0.07 0.86

Total 3 646 757 1.00 1 193 1.00

Men None 696 242 0.13 96 0.09 1.39

Primary 860 821 0.16 510 0.48 0.32 Source: Data for the sample
from the Uganda EDGE pilot
Secondary 3 646 757 0.66 350 0.33 2.00
survey, arms 4 and 5 combined,
Higher 343 331 0.06 107 0.10 0.61 self-reporting only. Data for the
Total 5 547 151 1.00 1 063 1.00
population from the 2002 Uganda
population and housing census.
148

1.4.4. Developing weights for asset

571. For asset-based analysis, additional calculation of weights is needed for
assets identified by respondents. The calculation of the asset weights depends on how
respondents own a particular asset, whether exclusively or jointly, and, if jointly, how
many persons share the ownership. It also matters whether those who share ownership
with the respondent are household members or not.
572. Taking reported ownership as an example, it is assumed that one respond-
ent was selected randomly within a household of three adult members. The asset
reported by the respondent as exclusively owned will be assigned the same weight (i.e.,
3, the inverse of the intrahousehold selection probability 1/3) for the respondent (as
discussed in chap. III, sect. 4, on sample design).
573. For assets that are jointly owned by the respondent and others, however,
multiplicity weights would need to be calculated, depending on the number of joint
owners and whether the joint owners are household members or not. For example, if
the randomly selected respondent reports joint ownership of a parcel of agricultural
land with two additional persons—one household member and one from another
household—the weight for this parcel will be calculated as the inverse of the total
selection probability of all joint owners. In this example, the selection probability is
1/3 for both the respondent and the within-household joint owner. Since the selection
probability for the non-household joint owner cannot be derived, it will be assumed
that this person has the same selection probability as the respondent. Accordingly, for
an indicator such as the share (percentage) of reported agricultural parcels owned by
women out of the total reported agricultural parcels owned by women and men, the
weight assigned to this particular parcel will be the inverse of (1/3 × 2 + 1/3), which is 1.
574. It is more complicated when there is more than one respondent within the
household. The respondents might not necessarily provide consistent reports of asset
ownership. In this case, a decision has to be made as to how to reconcile the informa-
tion before calculating those weights.

Key points
•• Data processing refers to a range of activities aiming to convert the information col-
lected in the field through the survey instrument into a database that can be used for
tabulation and analysis of the data. Typically, it includes data entry, data editing, data
imputation and an assessment of data quality and the precision of the survey.
•• Countries are encouraged to follow their typical protocol for data processing in
household surveys when conducting a survey on asset ownership and control from a
gender perspective. Two aspects of data processing need particular attention:
•• The data structure is complex, dealing with several statistical units of observation
and analysis, including households, individuals and assets (such as agricultural
land, large agricultural equipment, other real estate, financial assets and unincor-
porated enterprises). It is important that the structure of the survey data set reflect
the hierarchical relationships between the different statistical units, minimize the
storage requirements and interface well with statistical software at the analytical
phase;
•• It is important that no gender bias be introduced in any of the data processing
steps, including data adjustment activities undertaken to enhance the consist-
ency of data and that the survey sample results are representative for the targeted
population.
 Data processing, analysis and dissemination 149

2. Recommended indicators
575. This section presents a set of indicators for monitoring women’s and men’s
ownership and control of physical and financial assets at the global and national levels.
The selection of the indicators is consistent with the definitions and recommendations
presented in these Guidelines and based on the following statistical criteria:174 174 See United Nations, Inter-
•• The indicators should be relevant to policymaking and sensitive to policy agency and Expert Group
on MDG Indicators, “Lessons
interventions at the appropriate level (global, regional, national and local).
learned from MDG monitoring
•• The indicators should be clear and easy to understand for policymakers, from a statistical perspective:
the general public and other stakeholders. report of the task team on les-
•• The indicators should be disaggregated by gender and can be further dis- sons learned from MDG moni-
toring of the IAEG-MDG”, 2013,
aggregated by geographical region, income or special population groups,
and Canadian International
where applicable and relevant. Development Agency, Guide
•• The indicators should be a direct and unambiguous measure of progress in to Gender-Sensitive Indicators
gender equality in asset ownership in society over time. (Quebec, Hull, 1997).
•• The indicators should be consistent with and complementary to each other.
576. The following additional criteria are used for the selection of indicators for
global monitoring:
•• The number of indicators for global monitoring should be small.
•• The indicators should be relatively comparable across countries.
•• The indicators should be broadly consistent with global lists of indicators,
such as the Sustainable Development Goals, and avoid imposing an unnec-
essary burden on Governments and other partners.
•• The indicators should be able to be measured in a cost-effective and practi-
cal manner by countries.
•• A regular data collection mechanism should have already been or should
be able to be developed with reasonable costs and by involving the official
statistical system.

2.1. Level of monitoring

577. The present Guidelines distinguish between global indicators and national
indicators. Global indicators are the standardized indicators for which all countries
are encouraged to collect data to measure the prevalence of women’s and men’s owner-
ship and control of the priority assets: principal dwellings, agricultural land and other
real estate including non-agricultural land, and financial assets. An indicator of the
prevalence of women’s and men’s ownership of mobile phones is also proposed in line
with Sustainable Development Goal indicator 5.b.1. The data for all the global indica-
tors listed below can be obtained by integrating the minimum set of questions into an
existing household survey, as discussed in chapter III, thereby enabling national sta-
tistical agencies to monitor progress towards women’s ownership and control of assets
in a cost-effective manner.
578. National indicators are complementary indicators for which countries may
wish to compile data based on their policy needs and the resources available for data
collection. These indicators will of necessity be customized by countries to be relevant
to policymaking and sensitive to policy interventions within countries. For example,
countries may wish to derive an indicator of joint documented ownership of agricul-
tural land to monitor national land titling programs. Countries in which livestock
constitutes an important component of the agrarian economy may choose to develop
prevalence indicators of women’s and men’s ownership of cattle or other large livestock,
150

while countries in which women store a bulk of their wealth in jewellery may opt to
develop prevalence indicators of women’s ownership of jewellery and other valuables.

2.2. Level of measurement

579. All indicators proposed in these Guidelines use the individual as the unit of
analysis. Two different types of statistics on the prevalence of asset ownership are pro-
posed: first, the proportion of women (or men) who own assets; and, second, the share
of women among owners of assets. Additional aspects of asset ownership can also be
measured through indicators of exclusive or joint ownership of assets and on modes
of acquisition of assets. Lastly, the gender wealth gap is also an important dimension
to be measured to account for gender differentials in the quantity and quality of the
assets owned by women and men.
580. Indicators of the proportion of women (or men) who own assets are calcu-
lated using as the numerator, the number of women (or men) who own the asset, either
individually or jointly, and as the denominator, the total number of women (or men)
in the population. For example, for the indicator of the proportion of women (or men)
with documented ownership of principal dwellings, the numerator is the total number
of women (or men) in the population who are documented owners of principal dwell-
ings and the denominator is the total number of women (or men) in the population.
581. Indicators of the share of women among asset owners are calculated using
as the numerator, the number of women who own the asset, either individually or
jointly, and as the denominator, the total number of women and men who own the
asset, either individually or jointly. For example, for the indicator of the share of
women among documented owners of agricultural land, the numerator is the number
of women who have documented ownership, either individually or jointly, of agri-
cultural land and the denominator is the total number of women and men who have
documented ownership of agricultural land.
582. Although often used interchangeably, these types of indicators provide dif-
175 Cheryl Doss and others, Gen- ferent information.175 Indicators of the proportion of women (or men) who own assets
der inequalities in ownership indicate how widespread ownership is in the population and are useful for gender
and control of land in Africa, comparisons across time and countries. The proportions for women and men must
2013. be presented together so that they can be compared to produce the gender asset gap,
a gender inequality measure of the differential prevalence of women’s and men’s asset
ownership. Indicators of the share of women owners indicate how many of the people
who own assets are women and are useful for showing the underrepresentation of
women among asset owners.
583. Forms of ownership can be measured using the distribution of individuals
by forms of ownership—exclusive owner, joint owner and non-owner. It is important
to note that an individual can be an exclusive and a joint owner at the same time. For
example, the person can be an exclusive owner of one parcel of land and a joint owner
of another parcel. If data are being presented by forms of ownership in a distribution,
a certain order needs to be set up in assigning a specific ownership form to persons
who hold more than one form of ownership for a specific asset. If data are presented as
“Proportion of individuals who are exclusive owners of agricultural land” and “Pro-
portion of individuals who are joint owners of agricultural land”, then individuals
holding both forms of ownership will be covered in the numerator for both indicators;
as a result, the two proportions, with the addition of the proportion of individuals who
do not own agricultural land, will not add up to 100 per cent.
584. How women and men differ in the manner in which their assets were
acquired may be measured through indicators of “Proportion of owners who acquired
 Data processing, analysis and dissemination 151

[asset] through [purchase/marriage/inheritance]”. Similarly, as in the case of indicators

of forms of ownership, individuals might acquire the same type of asset in different
ways as they might own more than one count of a specific asset. For example, a per-
son might own two agricultural parcels and, while one parcel was acquired through
inheritance, the other was acquired through purchase. Because of this situation, the
proportions for all acquisition modes are not expected to add up to 100 per cent.
585. Gender wealth gap is an indicator that measures the quantity and quality
of assets owned by women and men. More information on how to calculate the indica-
tor is available in section 3, on data analysis. To calculate the gender wealth gap, the
data collection should move beyond the minimum set of questions and the roster of
assets is essential, as covered in chapter III, section 5, on questionnaire design. The
value of assets should also be collected.
586. The present Guidelines do not cover indicators that use the asset as the unit of
analysis. One possible formulation could, however, be the share (percentage) of the total
number of documented (reported) agricultural land parcels owned by women out of the
total number of documented (reported) agricultural land parcels owned by women and
men. Countries interested in tracking such an indicator are encouraged to compile data
for the indicator, taking into consideration data processing challenges associated spe-
cifically with asset-based indicators. General guidance on how to weight assets properly
to avoid overcounting them can be found in section 1, on data processing.

2.3. Indicator construct

587. In substantive terms, the set of recommended indicators measures key
concepts of asset ownership and wealth presented in chapter I, including the bundle
of ownership rights, forms of ownership, acquisition of core assets, and their quality
and quantity. Table 12 presents indicators of the bundle of ownership rights to assets,
including reported ownership, documented ownership, and the rights to bequeath and
sell, along with the rationale for the construction of each indicator.

Table 12
Indicators of the bundle of ownership rights: rationale and asset coverage

Indicator Rationale Asset coverage

Proportion of individuals Provides the broadest indicator of asset ownership, as it measures people’s perceptions of All assets
with reported ownership whether they consider themselves owners, which has implications for the behaviour that they
of [asset], by gender adopt vis-à-vis the asset
Proportion of individuals Provides a measure of the ability to claim ownership rights, in law, over an asset owing to the Principal dwelling; agricultural
with documented owner- individual being listed as an owner on the ownership document, and can be used to monitor land; other real estate, includ-
ship of [asset], by gender national programmes and policies on housing and land titling reforms ing non-agricultural land
Proportion of individuals Not all persons who consider themselves owners of an asset have alienation rights over that Principal dwelling; agricultural
with the right to sell or be- asset. As evidenced by the EDGE pilot data, men reported owners are more likely to have the land; other real estate, includ-
queath [asset], by gender rights to sell or bequeath assets than women reported owners ing non-agricultural land
Proportion of total popula- Provides a measure of the ability to claim ownership rights, in law, over an asset that is compara- Principal dwelling; agricultural
tion with documented ble across countries (or areas within countries) with disparate rates of documentation. Because land; other real estate, includ-
ownership of [asset] or the individuals may still have the right to sell or bequeath an asset in the absence of documented ing non-agricultural land
right to sell or bequeath ownership, as evidenced by analysis of the EDGE pilot data, the indicator combines documented
[asset], by gender ownership with the right to sell or bequeath, in order to render it comparable across countries

588. Table 13 presents indicators measuring additional aspects of asset owner-

ship, including forms of ownership, acquisition of assets, and the quality and quantity
of assets owned, together with the rationale followed in covering those aspects. This
type of indicator is used for a subset of indicators for national monitoring.
152

Table 13
Indicators of additional aspects of asset ownership: rationale and asset coverage

Indicator Rationale Asset coverage

Distribution of individuals by forms Provides a measure for monitoring national policies and pro- Principal dwellings, agricultural land,
of ownership (do not own, own grammes to increase women’s ownership of land and housing non-agricultural land
exclusively, own jointly) and gender through joint titling
Proportion of owners who acquired Provides information on how women and men acquire assets, and Principal dwelling, agricultural land,
[asset] through [specific mode of whether their modes of acquisition differ, with a view to develop- non-agricultural land
acquisition], by gender ing policies and programmes that promote women’s and men’s
­accumulation of assets. Policy-relevant specific modes of acquisi-
tion may refer to inheritance, purchase, government programmes
Share (percentage) of documented Accounts for gender differentials in the size of the agricultural land Agricultural land
(reported) agricultural land areas owned by women and men
owned by women out of total docu-
mented (reported) agricultural land
areas owned by women and men
Gender wealth gap Accounts for gender differentials in the quantity and quality of the Principal dwelling, agricultural land, other real
core assets owned by women and men estate including non-agricultural land, non-
agricultural enterprise assets, financial assets

589. The following list presents the global and national indicators organized by
type of asset.

Principal dwellings
Global indicators:
•• Indicator G1: proportion of total population with documented owner-
ship of the principal dwelling or the right to sell or bequeath the principal
dwelling, by gender
•• Indicator G2: share of women among individuals with documented own-
ership of the principal dwelling or the right to sell or bequeath the princi-
pal dwelling
National indicators:
•• Indicator N1: proportion of total population with reported ownership of
principal dwelling, by gender
•• Indicator N2: share of women among reported owners of principal dwelling
•• Indicator N3: proportion of total population with reported ownership of
principal dwelling and the right to sell or bequeath the principal dwelling,
by gender
•• Indicator N4: share of women among individuals with reported ownership
of principal dwellings and the right to sell or bequeath the principal dwelling
•• Indicator N5: proportion of total population with documented ownership
of principal dwelling, by gender
•• Indicator N6: share of women among documented owners of principal
dwelling
•• Indicator N7: distribution of individuals by form of ownership (do not
own, own exclusively, own jointly) of principal dwelling and by gender
•• Indicator N8: proportion of individuals who acquired ownership of the
dwelling through [specific mode of acquisition] (e.g., purchase; inherit-
ance; government programme), by gender
•• Gender wealth gap: difference in the wealth that women and men hold in the
principal dwelling (see sect. 3, on data analysis and dissemination of results).
 Data processing, analysis and dissemination 153

Agricultural land
Global indicators:
•• Indicator G3: proportion of total population with documented ownership
of agricultural land or the right to sell or bequeath agricultural land, by
gender
•• Indicator G4: share of women among individuals with documented own-
ership of agricultural land or with the right to sell or bequeath agricultural
land
National indicators:176 176 The global indicators of

•• Indicator N1: proportion of total population with reported ownership of agricultural land proposed in
these Guidelines only inform,
agricultural land, by gender
as proxies, the measurement of
•• Indicator N2: share of women among reported owners of agricultural land Sustainable Development Goal
•• Indicator N3: proportion of total population with reported ownership of indicators 5.a.1 (a) and (b), that
agricultural land and the right to sell or bequeath agricultural land, by refer to “agricultural popula-
tion”. Additional information
gender
on the methodology for the
•• Indicator N4: share of women among individuals with reported ownership Sustainable Development
of agricultural land and the right to sell or bequeath agricultural land Goal indicators can be found
•• Indicator N5: proportion of total population with documented ownership in box 1.
of agricultural land, by gender
•• Indicator N6: share of women among documented owners of agricultural
land
•• Indicator N7: distribution of individuals by form of ownership (do not
own, own exclusively, own jointly) of agricultural land and by gender
•• Indicator N8: proportion of individuals who acquired ownership of agri-
cultural land through [specific mode of acquisition] (e.g., purchase; inher-
itance; government programme), by gender
•• Indicator N9: share (percentage) of documented agricultural land area
owned by women out of total documented agricultural land area owned
by women and men
•• Indicator N10: share (percentage) of reported agricultural land area
owned by women out of total reported agricultural land area owned by
women and men
•• Gender wealth gap: difference in the wealth that women and men hold in
agricultural land

Non-agricultural land (or other categories of other real estate, as applicable)

Global indicators:
•• Indicator G5: proportion of total population with documented ownership
of non-agricultural land or the right to sell or bequeath non-agricultural
land, by gender
•• Indicator G6: share of women among individuals with documented owner-
ship of non-agricultural land or with the right to sell or bequeath agricul-
tural land
National indicators:
•• Indicator N1: proportion of total adult population with reported owner-
ship of non-agricultural land, by gender
•• Indicator N2: share of women among reported owners of non-agricultural
land
154

•• Indicator N3: proportion of total adult population with reported owner-

ship of non-agricultural land and the right to sell or bequeath non-agricul-
tural land, by gender
•• Indicator N4: share of women among individuals with reported ownership of
non-agricultural land and the right to sell or bequeath non-agricultural land
•• Indicator N5: proportion of total adult population with documented own-
ership of non-agricultural land, by gender
•• Indicator N6: share of women among documented owners of non-agricul-
tural land
•• Indicator N7: distribution of individuals by forms of ownership (do not
own, own exclusively, own jointly) of non-agricultural land and by gender
•• Indicator N8: proportion of individuals who acquired ownership of
­non‑agricultural land through [specific mode of acquisition] (e.g., pur-
chase; inheritance; government programme), by gender
•• Gender wealth gap: difference in the wealth that women and men hold in
non-agricultural land
Financial assets
Global indicator:
•• Proportion of individuals who have [specific type of financial asset] in
their name, by gender
National indicator:
•• Gender wealth gap: difference in the wealth that women and men hold in
financial assets
Non-agricultural enterprise assets
National indicator:
•• Proportion of individuals who own non-agricultural enterprise assets
•• Gender wealth gap: difference in the wealth that women and men hold in
non-agricultural enterprise assets
Consumer durables
Global indicator:
•• Proportion of individuals who own a mobile telephone, by gender
National indicator:
•• Proportion of individuals who own [specific type of consumer durable],
by gender
Livestock
National indicator:
•• Proportion of individuals who own [specific type of livestock], by gender
Large agricultural equipment
National indicators:
•• Proportion of individuals who own [specific type of large agricultural
equipment], by gender
•• Proportion of individuals who acquired ownership of [large agricultural
equipment] through [specific mode of acquisition] (e.g., purchase; inherit-
ance; government programme), by gender
 Data processing, analysis and dissemination 155

Small agricultural equipment

National indicator:
•• Proportion of individuals who own any small agricultural equipment, by
gender

Valuables
National indicator:
•• Proportion of individuals who own [specific type of valuable], by gender

Overall indicator of wealth

National indicator:
•• Gender wealth gap: the total net worth of key assets owned by women
(value of their key assets less the value of their outstanding liabilities) as a
share of the total net worth of key assets owned by women and men. The
gender wealth gap could also be calculated for each key asset separately.

Key points
•• This section presents a set of indicators for monitoring women’s and men’s ownership
and control of physical and financial assets at the global and national levels.
•• Global indicators are generally phrased as the proportion of total population with
ownership of a specific asset, and the share of women among owners. For compara-
bility at the international level, ownership of physical priority assets, such as principal
dwelling, agricultural land, other real estate, including non-agricultural land, is best
defined as “documented ownership or the rights to alienate the asset”.
•• At the national level, countries will need to assess the legal frameworks and social
norms governing access to assets and consider:
•• Measuring the full set of ownership rights depending on the policy objective;
•• Addressing all issues that are of policy relevance, including whether assets are
owned exclusively or jointly with others, how women and men differ in asset
acquisition and in wealth stored in a particular type of asset or in a combination of
different types of assets.

3. Data analysis and dissemination of results

590. Data analysis and dissemination are steps that need to be taken in prepar-
ing and communicating the key findings of the survey to stakeholders and a range
of users. Data analysis refers to the process of transforming raw data into statistics
and indicators presented in the form of numbers, tables and graphs and interpreted
in analytical reports that discuss data patterns and trends and their significance for
programme development and policymaking. This is the stage of the survey that aims
to answer the questions formulated as the survey objectives. The dissemination of data
refers to the release of survey findings through various statistical and analytical prod-
ucts, and also to the sharing of data files and associated metadata.
591. The first part of this section, focusing on data analysis and presentation,
illustrates how data analysis can be used to answer policy-relevant questions on asset
156

ownership and how to present the findings in a form that tells a story about existing
gender differences in a particular country context. While data and statistics obtained
from the EDGE pilots and other sources are used, it should be noted that the EDGE
pilots tested different iterations of the final methodology presented in these Guide-
lines. As a result, the findings presented below should be interpreted with caution and
are not fully comparable across countries. Readers are encouraged to consult national
EDGE reports prepared by the national statistical offices for a more detailed analysis.
592. The second part of this section focuses on the dissemination of results and
covers aspects related to dissemination of products that would typically be prepared at
the end of a household survey.

3.1. Data analysis

593. Data analysis is the component of the survey process that aims to pro-
vide answers to the overarching questions on asset ownership that were identified by
stakeholders at the outset of the project. As discussed in the introduction to these
Guidelines, these questions can be broadly categorized under three objectives: first,
understanding the gender asset gap; second, understanding the gender wealth gap;
and, third, when more than one household member is interviewed, undertaking intra-
household analysis of asset ownership to better understand how assets are allocated
within households, in particular within couples, and how asset ownership affects
intrahousehold decision-making powers.
594. It is important that data analysts on the survey team formulate, in consul-
tation with stakeholders, an analytical plan that is relevant to policy concerns in the
country. This should be done early in the planning stages of the survey, because both
the sample design and the structure of the questionnaire strongly influence the type
and range of analysis that can be undertaken. For example, if the division of asset
ownership between spouses or partners is identified as a policy concern for land and
housing titling reforms, the sample must be designed to yield a sufficient number of
spouses or partners who will be interviewed. Likewise, if one of the survey objectives
is to measure the gender wealth gap, the questionnaire will need to be designed with
the asset, not the individual, as the unit of observation. This has implications for how
the data file is organized and prepared for analysis.

3.1.1. Organization of the data file based on units of observation and analysis
595. As discussed in chapter I, section 5, on units of observation, a household
survey collecting data on asset ownership from a gender perspective has a hierarchical
set of “units of observation”—including the household, the individual and the asset—
that defines the levels at which information is collected. The household and individual
are common units of observation in household surveys. Information typically col-
lected using the household as the unit of observation includes household identification
variables and housing characteristics. Typical information collected at the individual
level includes demographic, social and economic characteristics of the members of
each household.
596. The individual is the unit of observation for information on asset owner-
ship in surveys using a minimum set of questions, such as “Do you own any [asset]?”
When the individual is used as the unit of analysis, the data file is organized in such a
way that a unique record exists for each individual. This record contains basic demo-
graphic, social and economic characteristics, such as gender, age and marital status,
together with a set of variables that identifies, for example, whether specific individuals
 Data processing, analysis and dissemination 157

are the reported or documented owners of particular types of assets, and whether they
have the right to sell or bequeath the asset. This organization is illustrated in table 14.

Table 14
Organization of the data file using individuals as the unit of observation
Characteristics of individuals

Reported Documented
owner of owner of
Individual Household Marital principal principal
ID ID Gender Age status … dwelling dwelling …

Records of
individuals {
597. In asset ownership surveys that collect information on characteristics of
assets—such as size, value or quality of assets—the asset is the unit of observation,
in addition to the household and the individual. A unique record exists for each asset
and includes information about the characteristics of the asset. In the example given
in table 15, there is a record for each parcel of agricultural land, which includes infor-
mation such as its owners, area of the parcel, use of the parcel and value of the parcel.

Table 15
Organization of the data file using assets as the unit of observation
Characteristics of assets

Number of Area of Use of Value of

Asset ID Household ID Owner ID owners parcel parcel parcel …

Records of
assets { Parcel 1
Parcel 2
…
Parcel n

598. The units of observation to be used as units of analysis are determined by

the questions guiding the analysis. Questions such as “What is the prevalence of asset
ownership among women and men?”, “Are women owners as likely as men owners
to possess the full bundle of ownership rights?”, “Do women and men acquire assets
in different ways?”, “Is the level of women’s wealth the same as men’s?”, “Is women’s
wealth concentrated in the same types of assets as men’s wealth?” require that the
ultimate analysis be carried out at the individual level to enable a comparison between
women and men. In contrast, a question such as “What is the distribution of agricul-
tural land parcels by type of use?” requires that the analysis be carried out at the asset
level, which, in this case, is the level of agricultural parcels.
599. Because these Guidelines are concerned with measuring and analysing
asset ownership from a gender perspective, the indicators and analysis presented
below use individuals as the unit of analysis. Even if the data are collected using assets
as the unit of observation, as shown in table 15, the data are transformed into a struc-
ture similar to that presented in table 14, for the purpose of analysing the data at the
individual level. This is reflected in table 16. For example, if one person appears in the
data set as an owner of two parcels listed in the roster of agricultural land, the analyst
158

may compute a variable of “ownership status for agricultural land” at the individual
level. This computed new variable indicates whether or not a person is an owner of
agricultural land, and it is used in the statistical analysis to answer a question such as
“What is the prevalence of ownership of agricultural land among women and men?”
600. Furthermore, when the value of the two parcels of land and the number of
owners for each of them are known, the analyst may compute a variable indicating the
177 When assets are used as units
individual wealth stored in agricultural land that can be attributed to each individual
owner. This computed new variable is used in the statistical analysis to answer ques-
of analysis of data obtained
from household surveys,
tions such as “What share of the total value of agricultural land can be attributed to
weights for assets, calculated women land owners?” To answer such questions, weights computed at the individual
as described in chapter IV, level177 to adjust for unequal selection probability or unit non-response (see sect. 1.4)
section 1.4, should be applied. should be used.
Table 16
Organization of the data file using individuals as the unit of analysis
Characteristics of individuals

Reported Documented Value of agri-

owner of owner of cultural land
Individual Household Marital agricultural agricultural attributed to Individual
ID ID Gender Age status … land land … individual … weight

Records of
individuals {
3.1.2. T ypes of variables used in an analysis of asset
ownership from a gender perspective
601. At a minimum, the essential variables for a gender analysis of asset owner-
ship and wealth are whether an individual owns a type of asset (e.g., principal dwell-
ing), characteristics of the asset and the monetary value of the asset, and gender of
owner. In addition, the following variables may be considered for a more nuanced
understanding of asset ownership from a gender perspective:
(a) Age, marital status, and type of household are variables typically found in
household surveys. They are key to using a life cycle perspective in the anal-
ysis. A life cycle perspective is particularly important when undertaking a
gender analysis of asset ownership. For example, in most societies, wealth
tends to accumulate as people get older, and demographic events that may
result in the dissolution of the household, such as marriage, divorce or the
death of a spouse, can have a significant impact on the acquisition or loss
of an asset, in particular for women;
(b) Education and employment are basic individual characteristics that are
likely to vary with the ownership of certain types of assets. Employment
history, in addition to the family trajectories noted above, can also shed
light on the driving forces behind wealth accumulation. Employment in
entrepreneurial activities may also be linked with the holding of assets in
their productive capacity, including land, livestock and machinery, or cer-
tain types of consumer durables (such as vehicles and equipment);
(c) Variables reflecting the decision-making power of women are key to under-
standing the empowerment that asset ownership may foster by increasing
women’s bargaining power in the household. Such variables may be related
 Data processing, analysis and dissemination 159

to participation in intrahousehold decision-making on the allocation of

economic resources, personal decision-making (on an individual’s own
health for example), or decision-making related to the use of assets in their
productive capacity;
(d) Context variables, such as urban versus rural areas or geographical and
administrative areas of a country, may reflect different formal laws or cus-
tomary rules regulating access to assets.

3.1.3. Key objective: measuring and understanding the gender asset gap
602. All individual-level data collection on asset ownership should aim to
measure the gender asset gap, or the differential distribution of asset ownership among
women and men in a society. The gender asset gap comprises, first, the prevalence of
asset ownership among women and men, which measures how frequently women and
men own a given type of asset; and, second, the share of women among asset owners,
which measures whether women and men are equally represented among the owners
of the asset type. Prevalence indicators are useful for representing gender differences
in asset ownership across time and countries, while share indicators are useful for
highlighting gender disparities among owners.
603. This section demonstrates how analysis could be carried out to study the
overlap of different types of ownership, namely, reported ownership, documented
ownership and the right to sell and bequeath (see chap. I, sect. 1.2, for more informa-
tion on types of ownership). Such analysis helps in determining the type or combina-
tion of types of ownership that best convey gender differences in a specific country.
604. This section also illustrates how, once the overall gender asset gap has been
identified, it varies within different population subgroups. It outlines the key charac-
teristics of owners and non-owners of assets.

3.1.3.1. Measuring the gender asset gap

605. The gender asset gap can be measured in terms of the prevalence of asset
ownership and the share of women among asset owners. To measure the prevalence of
asset ownership, the base proportions are calculated as the relative frequencies of asset
owners for each gender, with women’s and men’s respective totals in the sample used
as the denominators. The respective formulae for calculating the prevalence of asset
ownership, by gender, are the following:
Number of women owners
× 100 and  Number of men owners × 100
Total number of women Total number of men

606. As shown in table 17, out of 869 women respondents in Uganda, 306
reported owning the principal dwelling, yielding a prevalence of ownership of the prin-
cipal dwelling among women of 35 per cent (306/869 × 100). Out of 653 men respond-
ents, 389 reported owning the dwelling, giving an ownership prevalence among men
of 60 per cent (389/653 × 100).
607. The prevalence of reported ownership can also be presented in a graph, as
in figure 5. The gender gap in the prevalence of asset ownership is usually calculated
as the differential prevalence, that is, the proportion of men who are owners minus the
proportion of women who are owners, expressed in percentage points. In this exam-
ple, the prevalence of owning the principal dwelling in Uganda is 60 per cent for men,
compared to 35 per cent for women. The gender difference is 25 percentage points; in
other words, in Uganda, men’s ownership of the principal dwelling is 25 percentage
points higher than that of women.
160

Table 17
Distribution of adult population (age 18 and older) by gender and reported ownership
of the principal dwelling (Uganda, 2014)
Distribution of population of women and men
Number of respondents
by ownership status (percentage)
Women Men Women Men
Owner 306 389 35 60
Non-owner 563 264 65 40
Source: EDGE pilot, Uganda, treat-
Total 869 653 100 100
ments 4 and 5, self-reported data.

Figure 5
Prevalence of reported ownership of the principal dwelling among the adult population,
by gender (Uganda, 2014)
70

60

50

40
Per cent

30

20

10
Source: EDGE pilot, Uganda, treat-
ments 4 and 5, self-reported data. 0
Women Men

608. To determine whether the difference between women’s and men’s asset
ownership is statistically significant, the independence of gender and ownership sta-
tus should be tested, or a t-test carried out for the difference in asset ownership preva-
lence. For the Uganda example on ownership of the principal dwelling, a statistical test
showed that the gender difference was statistically significant.
609. To measure the share of women among asset owners, the numerator is
the number of women owners, while the denominator is the total number of owners
(both men and women). The following formula is used to calculate the share of women
among all owners:
Women owners
× 100
Women and men owners

610. As shown in table 18, out of 695 reported owners of principal dwellings,
306 are women. Women thus represent 44 per cent of adults in Uganda who own the
principal dwelling (306/695 × 100). The share of women among owners could also be
presented graphically (see figure 6).

Table 18
Distribution of reported owners and non-owners of the principal dwelling by gender
(Uganda, 2014)
Distribution of owners and non-owners,
Number of respondents
by gender (percentage)
Total Men Women Total Men Women
Owner 695 389 306 100 56 44
Non-owner 827 264 563 100 32 68
Source: EDGE pilot, Uganda, treat-
Total 1522 653 869 100 43 57
ments 4 and 5, self-reported data.
 Data processing, analysis and dissemination 161

Figure 6
Share of women among owners of principal dwelling (Uganda, 2014)

Women
Men

44%

56%

Source: EDGE pilot, Uganda, treat-

ments 4 and 5, self-reported data.

3.1.3.2. Analysing the overlap of ownership rights

611. As presented in the conceptual framework for measuring asset ownership
from a gender perspective, set out in chapter I, to capture meaningful gender differ-
ences in asset ownership, many countries will need to measure a combination of own-
ership rights, including reported ownership, documented ownership and the rights to
sell or bequeath the asset. This is because the full set of ownership rights may not all
be vested in one individual, with the overlap of rights most likely favouring men, as
shown below.
612. As a preliminary analysis, countries are thus advised to examine the prev-
alence of different types of ownership rights, followed by a study of overlaps of own-
ership rights to determine which measure or combination of measures best conveys
gender differences in asset ownership.
613. Figure 7 depicts differences in the overlap of ownership rights for men and
women in Uganda and KwaZulu-Natal, South Africa. In both places, men reported
owners are more likely than women reported owners to have other types of ownership
rights. For example, in Uganda, 89 per cent of men who reported owning the dwelling
also reported having the right to bequeath the dwelling, whereas only half of reported
women owners reported having this right. While the differences in the overlap of
ownership rights for men and women are less pronounced in KwaZulu-Natal, South
Africa, than in Uganda, they are still higher for men. For example, only 27 per cent of
women reported owners of the principal dwelling have documented ownership, com-
pared to 47 per cent of men reported owners. Although not presented, the patterns
of overlap are similar for agricultural land. Taken together, these findings suggest
that a prevalence indicator measuring reported asset ownership may not be sufficient
in and of itself for capturing policy-relevant gender differences in asset ownership.
Accordingly, although reported ownership is presented throughout this section for
ease of illustration, countries are advised to assess the overlap of different types of
ownership rights, as described above, before deciding which constructs of ownership
to disseminate.
162

Figure 7
Percentage of reported owners who have documented ownership, the right to sell and/or
bequeath assets, by gender (Uganda, 2014, and KwaZulu-Natal, South Africa, 2016)

Women 100
Men

80

60
Per cent

40
Source: EDGE pilot, Uganda, treat-
ments 4 and 5, self-reported data;
20
EDGE pilot survey, South Africa.
Note: For illustration purposes, the
data presented are not weighted. 0
Have Have the right Have the right Have Have the right Have the right
documented to sell to bequeath documented to sell to bequeath
ownership ownership

Uganda KwaZulu-Natal, South Africa

3.1.3.3. Gender asset gap by key characteristics

614. Additional analyses of the gender asset gap, including descriptive multiway
tabulations and more sophisticated multivariate inferential analysis, will be useful for
unpacking the gender asset gap. Analyses of these types can point to existing obstacles
to asset ownership and distinct patterns in the accumulation and loss of assets that
may be different for women and men and among different groups of women and men.
615. Table 19 illustrates how gender difference in ownership of the principal
dwelling varies by age in Uganda. The prevalence of dwelling ownership by age shows
that older respondents, both men and women, are more likely to own the principal
dwelling than their younger counterparts. For instance, the prevalence of ownership
for women increases from 19 per cent in the age category 18–34 to over 70 per cent
in the age category 60 and over. For men, the respective proportions increase from
40 per cent to over 80 per cent. A test shows that the increase with age in ownership of
the dwelling is statistically significant. Gender difference is also significant throughout
all age groups.

Table 19
Prevalence of reported ownership of principal dwelling, by gender and age
(Uganda, 2014)
Prevalence of ownership
Number of respondents Number of owners by age (percentage)
Age Women Men Women Men Women Men
18–34 473 324 88 131 19 40
35–59 296 255 145 196 49 77
60+ 100 74 73 62 73 84
Source: EDGE pilot, Uganda, treat-
All ages 869 653 306 389 35 60
ments 4 and 5, self-reported data.
 Data processing, analysis and dissemination 163

616. Similar to the above example, gender differences in asset ownership should
be analysed across different population subgroups defined by contextual variables
(such as urban versus rural areas, regions with different customs, or different ethnic
groups). The analysis should also apply a lifecycle perspective by using information on
age, marital status and type of household.178 178 Data analysts should keep in
mind that, when gender is the
focus of the analysis, while
3.1.3.4. Forms of ownership variations in the prevalence
of ownership across different
617. Because individuals’ rights and benefits associated with ownership may groups are important, the
differ if they own an asset exclusively or jointly, it is also important to assess whether highlighted differences (includ-
assets are owned exclusively or jointly, in addition to measuring the level of ownership. ing the differences observed
In Georgia, for example, women are less likely to be documented owners than men. In and their statistical sig-
addition, among documented owners, women are also more likely to be joint owners, nificance) should be between
while men are more likely to be exclusive owners (see figure 8). women's and men’s prevalence
of ownership.

Figure 8
Ownership of principal dwelling, by gender and documentation (Georgia, 2015)

100 Exclusive documented owner

Joint documented owner
90
Non-documented owner
80

70

60
Per cent

50

40

30

20

10 Source: EDGE Pilot, Georgia,

self-reported data.
0
Women Men

3.1.3.5. Asset acquisition

618. As discussed in chapter I, women and men may acquire assets in different
ways. Analysis of asset acquisition should highlight the most policy-relevant chan-
nels of acquisition in a country, which may include sales markets, government pro-
grammes relating to land and housing reform, inheritance (in particular from parents
and spouses) or acquisition through marriage, depending on the marital regimes in
the country. This information is important for the design of policies and interventions
aiming to strengthen mechanisms that provide equal opportunities in access to asset
ownership.
619. As illustrated in figure 9, in Georgia, the two most common ways for
women to acquire agricultural land are purchase and marriage, while for men, they
are purchase and allocation by family. This suggests that, if women do not have the
resources to purchase land, their ownership is largely conditional on marriage.
164

Figure 9
Main methods of acquisition of agricultural land (percentage), by gender (Georgia, 2015)

60

50

40
Per cent

30
Source: EDGE Pilot, Georgia,
­self-reported data. Percentages
20
do not add up to 100 owing to to
the fact that respondents might
be reporting different methods of 10
­acquisition for different agricul-
tural parcels they own. 0
Purchased Acquired Allocated Inherited Purchased Allocated Inherited Acquired
through from from through
marriage household household marriage
member member
Women Men

3.1.3.6. Creating profiles of owners and non-owners of major assets

620. A multivariate analysis, based on multiway tables and regression analysis,
can be used to understand the characteristics of individuals who are owners and non-
owners of certain assets, in terms of their gender, age, education, marital status, house-
hold income status, location or other relevant variables. This type of analysis may be
performed for major categories of assets (e.g., housing, land and financial assets) and
should respond to policy questions in specific sectors (such as those related to housing
and land titling programmes).
621. Table 20 presents the estimated odds ratio of owning principal dwellings
in Mongolia, by fitting a logistic regression model with the ownership of a dwelling
as the independent variable. An odds ratio larger than 1 means that, compared to the
reference category, people in this category have a higher likelihood of owning dwell-
ing. If the odds ratio is smaller than 1, people in this category have a lower likelihood
of ownership compared to the reference category. The 95-per-cent confidence interval
of the odds ratios provides information on whether the increase or decrease in the
probability of ownership, due to given characteristics (gender, age, education, etc.), is
statistically significant at the level of 95 per cent. For example, the confidence interval
for the comparison between the wage employee or employer and those who are not
employed is (0.88, 1.48). As the confidence interval covers 1, the analysis shows that
ownership prevalence does not differ between these two groups.
622. The model shows that, in addition to a significantly lower prevalence of
principal dwelling ownership for women compared to men, attributes such as age,
education, marital status, household wealth, employment status and province are all
important predictors of dwelling ownership. For example, older people are more likely
to own a dwelling than younger people. Married people are more likely to own a dwell-
ing than those who are not married. The model also includes an interaction term of
gender and other variables—the only significant interaction is employment status by
gender. What the model does show is that, for women, there is no difference in owner-
 Data processing, analysis and dissemination 165

ship prevalence, whether they are not employed or employed as own-account work-
ers, contributing family workers or casual labourers. For men, however, those who are
employed as own-account workers, contributing family workers or casual labourers
are more likely to own the principal dwelling than those who are not employed.

Table 20
Estimates of adjusted odds ratios in a logistic regression model
predicting ownership of principal dwelling (Mongolia, 2015)

Predictor Category Odds ratio 95% confidence intervals

Gender Woman 0.31 (0.24, 0.40)
Age Age 1.2 (1.16, 1.23)
Education College or higher 1.37 (1.14, 1.65)
Marital status Married 1.47 (1.15, 1.90)
Widow 2.48 (1.70, 3.61)
Household size Continuous variable 0.94 (0.89, 0.98)
Housing asset 2nd housing asset quartile 1.24 (1.02, 2.05)
quartile
3rd housing asset quartile 1.58 (1.21, 2.05)
4th housing asset quartile 1.47 (1.09, 1.99)
Employment Own-account worker, contributing family
status worker, casual labourer 2.3 (1.75, 3.02)
Waged employee, employer 1.14 (0.88, 1.48)
Province Bayan-Olgii 2.12 (1.46, 3.08)
Darkhan 1.22 (0.91, 1.63)
Dornod 0.88 (0.66, 1.16)
Khenti 1.73 (1.25, 2.41)
Source: EDGE pilot survey, Mon-
Khuvsgul 1.19 (0.91, 1.55) golia, self-reported data. Refer-
Selenge 1.44 (1.07, 1.94) ence predictors are gender (men),
education (less than college),
Umnugovi 1.47 (0.92, 2.35)
marital status (single), housing
Uvurkhangai 0.58 (0.44, 0.77) asset quartile (first housing asset
Uvs 0.77 (0.58, 1.02) quartile), employment status (not
Employment Woman: own-account worker, contributing employed), province (Ulaan-
and gender family worker or casual labourer 0.4 (0.28, 0.57) baatar). Housing asset quartile
Woman: waged employee, employer 1.25 (0.90, 1.74)
constructed using the dwelling
characteristics.

3.1.4. Key objective: measuring the gender wealth gap

623. Gender asset gap indicators presented in the previous section provide a
basic picture of whether or not assets are owned by women and men, but they do not
take into account the quantity, value or any other characteristics of the owned assets.
For example, similar proportions of women and men may own agricultural land in
a given area, but the number of parcels and their size and quality may vary greatly
between women and men owners.
624. By collecting individual-level data on the value of assets owned by women
and men, it is possible to assess the overall gender disparities in owning assets, in
terms of the ownership prevalence and also of the attributes of the assets owned by
women and men. Values of assets owned by women and men also provide insights
into women’s and men’s standards of living, bargaining power and vulnerability to
166

economic shocks, furthering our understanding of economic inequality from a gender

perspective.
625. Individual wealth or net worth (i.e., the total value of all assets minus the
total value of all liabilities) aggregates value across assets, enabling wealth compari-
179 Cheryl Doss and others, “Meas- sons for women and men who own different types of assets.179
uring personal wealth in devel-
oping countries: interviewing 3.1.4.1. Calculation of individual-level wealth
men and women about asset
values”, 2013. 626. Wealth analysis requires data collection using individual assets as the unit
180 Assets can also be used as units of observation,180 which is more complex than collecting data solely for the purpose
of analysis, which requires that of calculating the prevalence of ownership. The data file would typically be organized
weights for assets, calculated as shown in table 15. For each asset, information on the number of co-owners and its
as shown in section 1.4, are value is needed.
used. Examples of analysis
using assets as units of analysis 627. Based on these asset-level data, a variable on individual wealth can be cre-
include the distribution of agri- ated at the individual level. The wealth of an individual will be calculated as the total
cultural land parcels by type of value attributed to that individual less the value of liabilities of that individual. For
use, and forms of ownership. example, if an individual owns a principal dwelling with a spouse, and the value of the
dwelling is estimated to be 100,000 local currency units, the value attributed to him or
her is half, or 50,000 currency units. The same individual may exclusively own one agri-
cultural parcel of land valued at 10,000 currency units, which will be attributed entirely
to the individual. Finally, the individual and his or her spouse may owe 30,000 currency
units in bank loans for the purpose of purchasing the principal dwelling, which means
that a share of 15,000 units can be attributed as individual liability. In that case, the
level of wealth or net worth of the individual in local currency units is 50,000 + 10,000
– 15,000 = 45,000.
628. In order to calculate the composition of net wealth, or the wealth held by
type of asset, information on the value of liabilities specific to each type of asset is
181 See chapter III, section 5, on needed.181 In the case of the individual in the example above, the equity in the princi-
questionnaire content, for the pal dwelling is 35,000 (50,000 – 15,000) currency units, representing 78 per cent of the
implications of such calculation individual’s total wealth (35,000 × 100 / 45,000).
for the design of the question-
naire. 629. It is important to note that, when self-reported information is collected
from more than one respondent per household, it will be necessary to reconcile any
disagreements in ownership status before calculating the gender wealth gap (see box 9).

3.1.4.2. Average wealth of a specific asset held by owners and among overall population
630. The value of each asset and liability that can be attributed to each indi-
vidual is calculated by dividing the value of the asset by the number of owners, or the
amount of the liability by the number of persons responsible for repaying the loan. It
should be noted that the decision to allocate the value of jointly owned assets evenly
among owners is somewhat arbitrary. Countries that collect information on the own-
ership share of each joint owner may wish to use additional information to distribute
the overall value. In any event, it is important to ensure transparency in how value and
liability are distributed among joint owners.
631. Given the challenge in collecting good-quality data on values, a basic anal-
ysis of the valuation data should be conducted before any other analysis on wealth.
This entails analysing the percentage of missing data for the valuation variable, iden-
tifying possible data points that are outliers, calculating the mean with and without
trimming and comparing the value of the mean and the median to assess the degree
of inequality within the study population. A box plot of the valuation variable would
also be helpful.
 Data processing, analysis and dissemination 167

632. Table 21 presents, for owners of principal dwellings, median and mean
values of the dwelling with and without trimming. Trimming the outliers reduces
the mean values substantially. For example, the mean values with no trimming (for
women and men) are reduced by almost half when the top 5 per cent of values are
trimmed. This suggests outliers for both women and men at the higher end.

Table 21
Mean and median values of principal dwelling owned by women and men
(Mongolia, 2015, and Uganda, 2014)
Mongolia (in millions of tugriks) Uganda (in millions of shillings)
Measures Women Men Women Men
Mean No trimming 55.2 46.5 14 17.9
Top 1% values trimmed 51.9 43.5 9.4 12.7 Source: EDGE pilot survey, Mon-
Top 5% values trimmed 44.4 38.2 7.6 8.7 golia, self-reported data; Kilic and
Moylan, MEXA, 2016, treatments 4
Median 40 30 3.0 5.0
and 5, self-reported data.

633. The difference between the median and the mean is an indication of
wealth inequality in the population studied. For example, the average value of princi-
pal dwellings owned by women in Uganda is 7.6 million shillings after trimming the
top 5 per cent, while the median value is 3 million shillings. This suggests that a large
proportion of the dwellings owned by women are clustered at the lower end of the dis-
tribution of dwelling values. A similar situation also applies to men in Uganda. This
can be visualized in figure 10.

Figure 10
Distribution of values of principal dwellings (Uganda, 2014)

Uganda
60

50
Value of principal dwelling,
in millions of shillings

40

30

20

10 Source: EDGE pilot, Uganda,

treatments 4 and 5, top 5 per cent
0 values trimmed.
Men Women

634. The mean value of wealth vested in any specific asset takes into account the
attributes of such an asset. When coupled with the indicator of the prevalence of own-
ing such an asset, this indicator provides a more comprehensive picture of women’s
and men’s ownership of assets. To illustrate this point, table 22 shows that not only are
women less likely to own principal dwelling and agriculture land in Uganda, but the
average values of their assets are also lower than of those belonging to men. Financial
assets represent the only exception to this rule: there is no difference between women
168

and men in terms of their ownership of financial assets, nor in the value of those assets.
A similar analysis of data from Mongolia indicated that women are less likely to own
principal dwellings than men (30 per cent versus 53 per cent), although, on average,
the value of the principal dwellings owned by women is similar to that of those owned
by men.

Table 22
Prevalence of reported ownership, mean value of individual-level wealth among owners
and mean value of wealth (in millions of shillings for Uganda and in millions of tugriks for
Mongolia) among all persons, by gender and asset type (Uganda, 2014, and Mongolia, 2015)
Prevalence of Mean value of wealth Mean value of wealth
reported ownership among owners with among all persons with
(percentage) 95% confidence interval 95% confidence interval
Source: EDGE pilot, Uganda,
treatments 4 and 5. Asset values Assets Women Men Women Men Women Men
and number of owners are based Uganda
on self-reported data; EDGE pilot,
Principal dwelling 35 (32, 38) 60 (56, 63) 7.7 (5.9, 9.5) 13.1 (10.6, 15.6) 2.5 (1.9, 3.2) 7.6 (6.1, 9.2)
Mongolia, self-reported owner-
ship data. Top 1 per cent of values Agricultural land 18 (15, 20) 41 (37, 45) 4.4 (3.3, 5.5) 10.9 (8.4, 13.4) 0.7 (0.5, 0.9) 4.2 (3.1, 5.2)
trimmed Financial assets 31 (28, 34) 34 (31, 38) 0.2 (0.1, 0.3) 0.3 (0.2, 0.4) 0.065 (0.04, 0.09) 0.095 (0.07, 0.12)
Note: The calculations presented Mongolia
for agricultural land for Uganda Principal dwelling 30 (27, 32) 53 (51, 56) 33.5 (29.6, 37.5) 32.9 (29.0, 36.8) 8.6 (7.3, 9.9) 16.3 (14.2, 18.4)
do not include home gardens.

635. It is also possible to devise an overall measure of wealth that reflects both
the prevalence level of owning an asset and the mean value of wealth vested in the
asset among owners. The measure is calculated as the overall wealth, vested in one
specific or multiple assets, divided by total population (women or men). This measure
is also equivalent to the product of the prevalence of ownership and mean value of
wealth among owners, as shown in the last column of table 22. In table 22, the data
for Mongolia show that, among owners of principal dwellings, the average value of
the dwellings is the same for women and men. If, however, the prevalence of dwelling
ownership is taken into consideration, the overall wealth measure for women and men
would be different. That is, on average women store 8.6 million tugriks of their wealth
in principal dwellings, compared to 16.3 million tugriks for men.

3.1.4.3. Share of women’s wealth

636. Another way of presenting wealth data from a gender perspective is to
measure the share of individual wealth that is held by women. The indicator may be
calculated as:
Sum of wealth of women
× 100
Sum of wealth of women and men

637. This indicator can be calculated for a specific asset type, such as princi-
pal dwelling or agricultural land, a group of major assets, or all assets. The indicator
can be presented along with the share of women among owners for a more detailed
analysis. As illustrated by figure 11, in Uganda, women’s share of the value of princi-
pal dwellings, agricultural land and financial assets is lower than the women’s share
among all the owners of each of these assets. Women represent 44 per cent of dwelling
owners, but their share of total dwelling value is only 31 per cent. Likewise, 38 per cent
of agricultural land owners are women, but they possess only 17 per cent of the total
wealth stored in agricultural land.
 Data processing, analysis and dissemination 169

Figure 11
Women’s share among owners, and women’s share of total wealth stored in selected
assets (Uganda, 2014)

50 Women’s share among owners

44 Women’s share of wealth
45

40 38

35
31
30
Per cent

25

20 17

15

10 Source: EDGE pilot, Uganda, treat-

ments 4 and 5. Asset values and
5 number of owners are based on
self-reported data.
0
Principal dwelling Agricultural land

3.1.4.4. Gender wealth gap by household type

638. By analysing the distribution of wealth by gender and type of household,

it is possible to assess whether the overall gender wealth gap observed in the popula-
tion is driven by lower levels of individual wealth held by women in selected house-
hold types, such as those in one-person and single-parent households, or if the gender
distribution of wealth in favour of men persists among couple households as well. It
should be noted that this analysis requires a sufficiently large number of households in
each type of household included in the analysis.
639. Table 23 provides an example based on data from Uganda showing the
average individual wealth of respondents who belong, first, to one-person or single-
parent households; and, second, to nuclear households that consist of only one mar-
ried couple or partners in a consensual union, with or without children. The results
suggest that, in Uganda, gender wealth inequality varies between those living in
single-adult households and those in couple households, with or without children.
There is no difference when comparing women and men living in single-headed
households, while among nuclear couple households, women’s average wealth is
much smaller than that of men. They also demonstrate that women in single-parent
or single-person households own more wealth than women in couple households,
while men in single-parent or single-person households own much less than those
living in couple households.
640. The analysis does not cover respondents living in other types of house-
holds, such as a couple living with in-laws or adult siblings or a single mother living
with her parents. It is still possible, however, to assess how women and men differ in
owning assets for other household types. When sample size permits, it is advisable to
break those households further down into different constructs for similar analysis.
170

Table 23
Individual wealth by household type for major assets (principal dwelling, agricultural
land, financial assets and non-farm enterprise assets) (Uganda, 2014)
Average wealth Share of women’s wealth
Number of among respondents =
respondents (in millions of shillings) average women’s wealth/
(average women’s wealth
Type of household Women Men Women Men + average men’s wealth)
Single adult households
Source: EDGE pilot, Uganda, treat- 106 77 3.5 5.4 0.394
(with or without children
ments 4 and 5, self-reported data. Couple households with no other
301 253 1.4 10.3 0.118*
adult (with or without children)
* Significant at 99 per cent level.

641. Instead of calculating the ratio of the overall wealth of women to the over-
all wealth of both women and men, the gender wealth gap in the analysis is calculated
as the ratio of the average value of women’s wealth divided by the sum of the average
value of women’s wealth and the average value of men’s wealth (see formula below).
This is because the total number of women and men respondents differ under the two
types of households.
Average wealth of women
× 100
Average wealth of women + average wealth of men

642. It is important to note that use of the ratio of the overall wealth of women
to all wealth will introduce a bias if the total numbers of women and men differ in the
study population. If there are more women than men in the study population, compar-
ing the total wealth of women with that of men will give women an advantage. Usu-
ally, however, this would not be an issue for country-level estimates, as the numbers of
adult women and men would be about the same after adjusting for non-response with
the use of appropriate weights.
643. It might be problematic when the target population is a population sub-
group of the country and there is an inherent imbalance in the gender composition of
that subgroup. For example, globally, women constitute 54 per cent of the population
182 United Nations, The World’s aged 60 and over, and 62 per cent of the population aged 80 and over.182 If one objective
Women 2015: Trends and Statis- of the data collection is to assess asset ownership among the older population, account
tics, 2015. should be taken of the preponderance of women in the study population. Simply cal-
culating the share of women’s wealth among all wealth will tend to underestimate
the real gender gap due to the higher number of women. In this case, it is advisable to
use the average wealth among the subgroup of respondents (older women, older men)
instead of their total wealth (as in the formula above).
644. Similarly, in table 23, women constitute some 58 per cent of all respond-
ents in single-adult households, compared to 54 per cent of respondents in one-couple
households. This again justifies calculating the share of women’s wealth with the use
of averages rather than totals.
645. Averaging wealth over the number of respondents avoids underestimating
or overestimating the gender wealth gap that may be introduced by imbalances in the
number of respondents. When there is an equal number of women and men respond-
ents, the two formulae provide the same estimate.
646. Although not presented here, in a manner similar to the analyses of the gen-
der asset gap presented in section 4.1.3 above, the gender wealth gap can be calculated
and analysed in relation to certain additional contextual and household-level variables,
such as education or age, and to population subgroups, such as rural versus urban.
 Data processing, analysis and dissemination 171

3.1.5. Key objective: intrahousehold analysis of asset ownership

647. Interviewing more than one adult per household about their asset owner-
ship makes possible an intrahousehold analysis of how asset ownership and wealth are
distributed among household members, thus furthering the current understanding of
intrahousehold economic inequality. A component of inequality in general and gen-
der inequality in particular, intrahousehold economic inequality has largely remained
unexplored in official statistics. A main impediment has been the fact that much of
the economic data are collected at the household level. Now, however, the collection of
information on asset ownership and wealth at the individual level from more than one
household member is creating new opportunities for data analysis.

648. The analysis of gender inequality in asset ownership involves collecting

self-reported information from more than one household member or from partnered
(married or cohabiting) women and men living in the same household. The examples
of intrahousehold analysis presented below focus on spouses and partners living in
the same household. As a result, households without couples are excluded, but the
analyses can be extended to look at all adult women and men within households with
more than one adult member, depending on common household compositions within
a country and the related policy needs.

649. With self-reported information from more than one household member,
new variables can be constructed that measure the differences in asset ownership and
wealth between women and men. The variables may be categorical, indicating—for
example—whether one or both spouses own a particular type of asset, or continuous,
indicating the number of assets or asset types held by each spouse or how large the
wealth difference is between spouses. The unit of analysis becomes the couple, and
both descriptive and inferential analyses can be undertaken to enrich the analysis of
intrahousehold gender differences in ownership and wealth.183 183 At its basic level, intrahouse-
hold inequality is identified
along dimensions of age and
3.1.5.1. Within-couple inequality in owning assets gender. Although this section
refers only to gender differ-
650. In table 24, households in Uganda, rural Mexico (agricultural land), ences, as captured mainly by
the differences between wives
and Mongolia (principal dwelling) are classified according to the ownership status
and husbands, it is recognized
by both members of couples. The sample is based on self-reported data only and is that intergenerational differ-
restricted to couple-only households in which both spouses are respondents. In rural ences are also important by
Mexico, 73 per cent of the couples do not own any agricultural land. In 18 per cent of themselves or in combination
the rural couple households, both members of the couple own agricultural land. In with a gender analysis.
2 per cent of the rural couple households, only the wife owns agricultural land, while
in 7 per cent of rural couple households, only the husband owns agricultural land. In
Uganda, in 26 per cent of couple households, both members own agricultural land;
in 54 per cent, only the husband owns the land and, lastly, in 5 per cent of couple
households, only the wife owns the land. For Mongolia, in half of the households
(50 per cent), only the husband owns the dwelling, and in only 6 per cent of all house-
holds, only the wife owns the dwelling. In around 21 per cent of households, both
members of the couple own the principal dwelling.

651. Even when both spouses own a specific asset, a gender difference exists.
For example, in Uganda, in 26 per cent of all couples, both members own agricultural
land. However, women’s plots are much smaller in size at 1 acre, on average, compared
to 2.3 acres for those of men.
172

Table 24
Intrahousehold ownership of selected assets based on self-reporting by both spouses
in couple households (Mexico, 2014, Uganda, 2014, and Mongolia, 2015)

Percentage of couples

Ownership by wife and husband Mexico: rural, Uganda: Mongolia:

in couple households agricultural land agricultural land principal dwelling

Neither spouse owns 73 15 22

Only wife owns 2 5 6

Only husband owns 7 54 50

Source: EDGE pilot surveys, Both own 18 26 21

Mexico, Mongolia and Uganda, No. of couple households 1 953 323 1 671
self-reported data.

3.1.5.2. Intrahousehold wealth inequality

652. Intrahousehold wealth inequality can be measured in different ways. The

example in table 25 below assesses whether the woman own less than the man within
the same couple and, if so, by how much. For wealth calculated on the basis of the
principal dwelling, agricultural land, financial assets and non-business assets, in
82 per cent of the 286 couples in the Uganda pilot the wife’s wealth was lower than
the husband’s.
653. While the data showed no significant variations in whether women own
less than their spouses in terms of their urban versus rural residence or by education,
among couples that own assets, women with higher levels of education have a higher
share of a couple’s wealth than those with lower levels of education (see table 25).

Table 25
Intracouple wealth difference and women’s share of wealth among couples (Uganda, 2014)

Source: Kilic and Moylan, MEXA, Urban or rural residence Education

2016, treatments 4 and 5, self-
Primary Secondary
reported data. Top 1 per cent of
Total Urban Rural or lower or higher
values trimmed.
Percentage of couples where
* Excludes households that have women’s wealth is lower than
no wealth. Average women’s men’s wealth 82% 76% 84% 83% 81%
share of wealth within the
couple differs significantly Average women’s share of
between those with primary or wealth within the couple* 18% 21% 17% 17% 20%
less education and those with Number of couple households 286 63 223 193 93
secondary or higher education.

654. In addition to the intracouple gender difference in wealth shown in the

above table, the scatter plot in figure 12 below of women’s share of wealth within a
couple by the overall wealth of the couple provides a graphic representation of how
such gender difference varies by the total wealth of the couple. As the couple’s wealth
increases, the women’s share of wealth decreases.
 Data processing, analysis and dissemination 173

Figure 12
Women’s share of wealth by couple’s wealth (Uganda, 2014)
1.0

0.9

0.8

0.7
Women’s share of wealth

0.6

0.5

0.4

0.3

0.2
Source: EDGE pilot, Uganda, treat-
0.1
ments 4 and 5, self-reported data.
0 Top 1 per cent of values trimmed.
0 10 20 30 40 50 60 70 80 90 100

Couple’s wealth, in millions of shillings

3.1.5.3. Discrepancy in responses

655. As discussed in chapter IV, when more than one of the household members
self-report their asset ownership status, there may be reporting discrepancies among
respondents. For example, a man respondent may report that he exclusively owns the
principal dwelling, while his wife may also claim exclusive ownership of the same
dwelling. Alternatively, he may report that he jointly owns the dwelling with his wife,
while his wife reports that she does not consider herself an owner of the dwelling. Table
26 categorizes households according to the self-reported dwelling ownership status of
both spouses in couple households in Georgia. The last column indicates whether or
not the responses provided by the respondents conflict with their spouses’ responses,
thus creating a reporting discrepancy. Among households with couples, 17 per cent
had reported discrepancies in ownership status, and in 8 per cent men underestimated
women’s ownership.
656. While data analysts will need to reconcile discrepancies in responses to
construct key measures, including the gender asset and wealth gaps, disagreements
among respondents about their ownership status can be expected and constitute use-
ful information in their own right (see box 9 for details about the reconciliation of dis-
crepancies). For example, development interventions may benefit from knowing when
spouses agree or disagree about who owns major assets, such as the principal dwelling
or agricultural land.
657. To assess whether certain individual or household characteristics are asso-
ciated with disagreements between spouses about their ownership status, a statistical
model can be used to predict cases in which there is a discrepancy in ownership status
reported by members of the couple, or to predict cases in which the member who is a
man underestimates his wife’s or partner’s ownership status, as presented in table 26 for
Georgia. Potential predictive variables include whether the residence of the household is
urban or rural, the wealth level of the household, age, education and employment status
of the men and women partners and the age difference of the men and women partners.
174

Table 26
Distribution of households based on self-reported dwelling ownership of both spouses or partners in couple households,
Georgia, 2015
Men underestimating
Distribution Discrepancies in women’s ownership
Type of households based on respondents self-reporting of households (%) ownership status? status?
1. Man non-owner, woman non-owner 11.8 No No
2. Man non-owner, woman exclusive owner 0.6 No No
3. Man non-owner, woman joint owner with man 2.4 Yes No
4. Man non-owner, woman joint owner with someone else 0.3 No No
5. Man joint owner with woman, woman non-owner 6.4 Yes No
6. Man joint owner with woman, woman exclusive owner 0.3 Yes Yes
7. Man joint owner with woman, woman joint owner with man 66.1 No No
8. Man joint owner with woman, woman joint owner with someone else 0.2 Yes No
9. Man joint owner with someone else, woman non-owner 0.6 No No
10. Man joint owner with someone else, woman exclusive owner 0 Yes Yes
11. Man joint owner with someone else, woman joint owner with man 1.2 Yes Yes
12. Man joint owner with someone else, woman joint owner with someone else 0 Yes Yes
13. Man exclusive owner, woman non-owner 3.7 No No
14. Man exclusive owner, woman exclusive owner 0 Yes Yes
15. Man exclusive owner, woman joint owner with man 6.3 Yes Yes
16. Man exclusive owner, woman joint owner with someone else 0 Yes Yes
Total number of households with couples 1 545
Percentage of households with discrepancies 17
Percentage of households where men underestimate women’s ownership 8

Source: EDGE pilot survey, Georgia, self-reported data.

Box 9
Reconciling reporting discrepancies when interviewing multiple persons
in the same household
As described in chapter III, one or more of the adult household members may be selected
as respondents to self-report their ownership of assets. For example, interviewing more
than one adult member within the household is required if intrahousehold (or intracou-
ple) analysis of asset ownership and wealth is an objective of the data collection exercise.
Interviewing more than one household member may result in discrepancies in the
reporting of ownership of a given asset. For example, in respect of the same asset a hus-
band might report exclusive ownership while his wife may report joint ownership. This
reporting discrepancy does not affect the calculation of the overall prevalence of owner-
ship for women and men. As emphasized in chapter I, individuals are owners if they report
themselves as owners. In the example above, each of the adults interviewed is considered
an owner and counted when the prevalence of ownership by gender is calculated. This
discrepancy must, however, be resolved or reconciled, first, when deciding how to assign
exclusive and joint ownership to each of the household members; and, second, when ana-
lysing the distribution of household wealth among individual members or calculating the
gender wealth gap. What fraction of the asset value should be attributed to each of the
two adults interviewed? Should they have equal shares or not?
The type and magnitude of discrepancies, and also the choice of reconciliation
method, may have an impact on the results obtained in the analysis. Where the types of
 Data processing, analysis and dissemination 175

discrepancies are concerned, if the focus is on couples only, there are 16 possible ways
that members of a couple could report their ownership about a given asset, 10 of which
would result in discrepancies (see the shaded rows in the table later in this box, presenting
all possible combinations of answers by each member of a couple about the ownership of
the principal dwelling). If there are more than two respondents per household, the num-
ber of possible combinations of answers increases, and with them the number of different
types of discrepancy.
The magnitude of discrepancies can vary from one country to another. Evidence gath-
ered through four EDGE pilot surveys (in Georgia, Mongolia, the Philippines and Uganda)
that collected ownership data from both members of a sample of couples illustrates dif-
ferent magnitudes of discrepancy and disagreement among couples about who owns the
dwelling. The percentage of couples with discrepancies in reported ownership ranges
from 9 per cent in the Philippines to 31 percent in Uganda.
Some discrepancies are more likely to occur than others. For example, across all four
pilot countries, the most likely discrepancies occured either when the husband reported
joint ownership of the principal dwelling with his wife, while his wife did not consider
herself an owner (category 5 in the table below), or when the husband reported exclusive
ownership of the dwelling, while his wife reported jointly owning the dwelling with her
husband (category 15).
Two approaches to the reconciliation of reported discrepancies may be considered:
Approach 1—ignoring discrepancies and assigning ownership according to the
ownership status reported by each individual respondent: for example, in category
(5) (see table), the husband indicated that he was a joint owner of the principal dwelling
with his wife, but his wife reported that she was not an owner. Using the ignoring rule,
the couple will be assigned ownership as reported, in other words, the husband is a joint
owner and the wife is not an owner. In terms of apportioning the value of the dwelling, the
husband will be apportioned half of the value of the dwelling, while the wife will receive
no value. In the case of category (15), the husband reported being an exclusive owner and
his wife reported being a joint owner with the husband, the ignoring rule will consider the
husband an “exclusive owner” and the wife a “joint owner”. As a result, the full value of the
dwelling will be assigned to the husband and an additional half of the value to his wife.
It should be noted that this reconciliation method may lead to results inconsistent with
the calculated value of the household-level wealth. In the first example (category 5), the
total wealth of the couple vested in the principal dwelling is only 50 per cent of the overall
value, while in second example (category 15), the total wealth of the couple vested in the
principal dwelling exceeds the overall value of the dwelling by 50 per cent. As a result, the
household-level wealth should not be calculated on the basis of the aggregated value of
individual wealth, while using this reconciliation method.
Approach 2—overriding any discrepancies in the ownership reported by individ-
ual respondents on exclusive or joint ownership: with this method of reconciliation,
persons will be considered exclusive owners if they are the only persons reporting owning
an asset in the same household, regardless of whether they report exclusive or joint own-
ership. If other household members also report owning the same specific asset, these per-
sons, together with all the others reporting owning the asset, will be considered as joint
owners. A person who does not report owning the asset will be considered a non-owner,
consistent with the rule used to calculate the prevalence of ownership. For category (5) in
the table, adoption of the overriding rule entails that the wife will be a non-owner, while
the husband will be considered an exclusive owner and will be apportioned the full value
of the dwelling. For category (15), both members of the couple will be considered joint
owners and will be apportioned half of the value of the dwelling, if no other member of
the household reports owning the same asset.
176

It should be noted that, although this reconciliation method does not overestimate the
household wealth, it also comes with limitations. When there is a discrepancy of ownership
among multiple household members, the reconciliation rule overrides the self-reported
joint or exclusive ownership status of one of the respondents, leading to inconsistencies in
the original self-reported information on exclusive or joint ownership. The wealth assigned
to household members following this reconciliation method will be different from the
wealth that would have been assigned to individual respondents according to their own
reporting. The overall prevalence of owning an asset would not be affected.

Distribution of households based on self-reported ownership of principal dwelling by each member of sampled
­couples, EDGE pilot surveys

Ownership under the

Discrepancies
Households by type of responses in ownership Ignoring rule Overriding rule
from members of couples status? Georgia Mongolia Philippines Uganda Men Women Men Women
1. Man non-owner, woman non-owner No 11.8 25.5 44.3 20.5 No No No No
2. Man non-owner, woman exclusive owner No 0.6 4.8 3.2 1.0 No Exclusive No Exclusive
3. M an non-owner,
woman joint owner with man Yes 2.4 1.4 2.1 1.3 No Joint No Exclusive
4. M an non-owner,
woman joint owner with someone else No 0.3 0.3 0.4 0.0 No Joint No Exclusive
5. Man joint owner with woman,
woman non-owner Yes 6.4 5.6 3.1 5.6 Joint No Exclusive No
6. M an joint owner with woman,
woman exclusive owner Yes 0.3 1.6 0.9 1.0 Joint Exclusive Joint Joint
7. M
 an joint owner with woman,
woman joint owner with man No 66.1 9.9 32.9 4.0 Joint Joint Joint Joint
8. Man joint owner with woman,
woman joint owner with someone else Yes 0.2 0.1 0.1 0.0 Joint Joint Joint Joint
9. M
 an joint owner with someone else,
woman non-owner No 0.6 2.3 1.4 1.0 Joint No Exclusive No
10. M an joint owner with someone else,
woman exclusive owner Yes 0.0 0.0 0.0 0.3 Joint Exclusive Joint Joint
11. M an joint owner with someone else,
woman joint owner with man Yes 1.2 0.5 0.1 0.3 Joint Joint Joint Joint
12. Man joint owner with someone else,
woman joint owner with someone else Yes 0.0 0.0 0.0 0.0 Joint Joint Joint Joint
13. Man exclusive owner, woman non-owner No 3.7 41.0 8.5 42.7 Exclusive No Exclusive No
14. M an exclusive owner,
woman exclusive owner Yes 0.0 0.9 0.1 1.7 Exclusive Exclusive Joint Joint
15. Man exclusive owner,
woman joint owner with man Yes 6.3 6.1 3.0 19.9 Exclusive Joint Joint Joint
16. M an exclusive owner,
woman joint owner with someone else Yes 0.0 0.0 0.0 0.7 Exclusive Joint Joint Joint
Percentage of households with discrepancy 16.9 16.2 9.3 30.8
Total number of households 1 545 1 463 1 030 302

Neither of the two methods is perfect. The ignoring rule underestimates or overesti-
mates the overall household wealth; while the overriding rule overrules the self-reported
ownership status for some of the respondents, affecting self-reported prevalence indi-
cators of joint and exclusive ownership, whenever ownership discrepancies occur within
the household. Furthermore, the extent of gender differences in the prevalence of joint
or exclusive ownership, and also in wealth, may vary when one method is favoured over
 Data processing, analysis and dissemination 177

the other. Accordingly, before adopting either of the described reconciliation methods, an
assessment of the impact of each method on the analysis outcomes should be conducted.
The following analysis illustrates how such an assessment can be carried out by cal-
culating wealth generated by different reconciliation methods. In addition to the two
methods described above, a third method is also considered that randomly takes one of
the responses, either from the husband or the wife, for households that have ownership
discrepancies. It should be noted that the assessment is carried out for reported owner-
ship of the principal dwelling and is limited to Mongolia and Uganda, because of the high
percentage of missing valuation data for Georgia and the Philippines.
As shown in the assessment, the impact of different reconciliation rules on the overall
wealth of women and men is much smaller for Mongolia than for Uganda. This is to be
expected, because of the smaller percentage of couples that have discrepancies in Mon-
golia. The significantly lower wealth of men in Uganda when the overriding rule is used,
compared to the ignoring rule, is due to the very high proportion (20 per cent) of couples
that fell into category (15). In that category, the husband self-reported having exclusive
ownership of the dwelling, while his wife claimed joint ownership with her husband.
When the overriding rule is used, the value assigned to the husband is only half that of the
full value under the ignoring rule. On the other hand, women’s share of wealth is lower
(mainly for Uganda) when calculated using the ignoring rule than when using the over-
riding rule, as categories (5) and (15) in the example, the two categories with the largest
discrepancies, both reflect cases when the husband underestimates his wife’s ownership
of assets. Because the data are limited to those from Mongolia and Uganda, it is difficult
to ascertain why women’s share of wealth calculated using the random approach is lower
than the value calculated through the other two methods.
It should be noted that none of the reconciliation methods is perfect, as there is no
golden rule to be applied in resolving ownership discrepancies. Before adopting any
method for reconciling the discrepancies, an assessment needs to be made on the overall
magnitude of the discrepancies and their impact on the wealth assigned to individuals.
When final results are published, information should be provided on whether discrepan-
cies were reconciled and, if so, which method was used.

Wealth stored on principal dwelling calculated on the basis of different reconciliation rules, Mongolia and Uganda
Approach 1: Approach 2: Approach 3:
Ignore discrepancies and ap- Override discrepancies and Random approach
portion value according to self- apportion value based on and value appor-
reported ownership status assigned ownership status tioned accordingly
Men Women Men Women Men Women
Mongolia
Number of respondents 1 463 1 463 1 463 1 463 1 463 1463
Sum of value apportioned to owner (in millions of tugriks) 6.3 3.0 6.4 3.1 6.4 3.0
Average share of women’s wealth among couple’s wealth* 0.323 0.324 0.316
Uganda
Number of respondents 302 302 302 302 302 302
Sum of value apportioned to owner (in millions of shillings) 2 207 656 1 760 580 1 829 498
Average share of women’s wealth among couple’s wealth* 0.229 0.248 0.214

Note: The average value is used in the calculation when both members of the couple reported the value of the dwelling.
* The share is calculated as (women’s wealth)/(women’s wealth + men’s wealth).
178

658. A probit model (not shown) estimating the likelihood that men in Georgia
underestimate their partner’s ownership showed that older men were more likely than
younger men to underestimate their spouses’ ownership of dwelling. For women who
are employed as farmers, contributing family workers and casual labourers, the own-
ership of dwellings is more likely to be underestimated by their spouses than it is for
women employees, employers or own-account workers.

3.1.5.4.Asset ownership dynamics and women’s

decision-making power within the household
659. Individual-level information on asset ownership can be used to assess
the effects of women’s increased bargaining power on household outcomes, such as
children’s health and education, and women’s own well-being. Questions on decision-
making include those on how own income is used; how the spouse’s income is used;
the respondent’s own health; major household purchases; and visits to the family.
These five questions were included in the South Africa pilot. Other studies included an
184 Carmen Deere and Jennifer additional question on who makes the decision on whether or not to work.184
Twyman, “Asset ownership and
660. To assess the impact of asset ownership on decision-making power within
egalitarian decision making
in dual-headed households the household, a statistical model can be built using the decision-making variables as
in Ecuador”, Review of Radical outcome, while incorporating other auxiliary variables such as age, spouses’ education,
Political Economics, vol. 44, employment status, and other attributes. A study in Ecuador, for example, examined
No. 3 (September 2012), the relationship between a couple’s egalitarian (joint) decision-making on how to spend
pp. 313–320. income and the wife’s share of the couple’s wealth using a logistic regression model.185
185 Ibid.
661. The study found that the likelihood of egalitarian (joint) decision-making
on how to spend one’s own income increased as the wife’s share of wealth increased, until
the wife’s share was 45 per cent, at which point the likelihood of join ­decision‑making
began to decrease. It is possible that women were more likely to make their decisions
alone (without consulting their husbands) as the share of women’s wealth was larger
than 45 per cent, although further analysis is needed. The study also found that, if both
spouses owned real estate (either jointly or individually) then there was a greater likeli-
hood that the spouses decided jointly about how to spend their own income (1.5 times
the odds), compared to when neither owned real estate.

3.2. Data dissemination

662. The second part of this section is focused on the dissemination of results
and covers aspects related to dissemination strategies and products that would typi-
cally be prepared at the end of a household survey. It is recommended that the prod-
ucts described be developed.

3.2.1. Data tables

663. Similar to other surveys, a tabulation plan should be developed at the stage
of planning a survey on asset ownership, and implemented after the data has been
cleaned, edited and weighted, as necessary.
664. Data tables provide access to data to a large number of users who wish to
further analyse the results of surveys or to carry out research on specific topics, when
they do not have access to microdata or lack the technical expertise or resources to con-
duct their own microdata statistical analysis. The data disseminated may be detailed
and organized in large tables. It is important that both absolute numbers and calculated
proportions, shares and averages be provided. Absolute numbers may give specialists
more flexibility in conducting their own analysis. Some tabulated data may be included
 Data processing, analysis and dissemination 179

in publications, and more detailed data should be included in online databases, so as to

enable users to browse the data tables and choose statistics that are of interest to them.

3.2.2. Analytical publications, reports, articles and briefs

665. Analytical publications, including reports, articles and briefs, may be
intended for statisticians, researchers or policymakers. When preparing such publi-
cations, data analysts must take care to present their findings in a format that can be
clearly understood by the intended audience.
666. In those publications, the disseminated data are usually presented in small
summary tables and charts, and discussed in accompanying text. Large tables with
more detailed data may be provided in annexes, as in the household survey report.
Material can also be integrated into publications focusing on gender, which may have a
different target audience (e.g., gender specialists, gender advocates and policymakers).
Instead of presenting data that readers must analyse to draw their own conclusions
(as in the case of dissemination products focused on providing data tables), analytical
publications emphasize the main results of the data analysis, and give an interpretation
and implications for policymaking. They should be designed to be user-friendly and
drafted in accessible language, with simple and attractively presented tables and charts.

3.2.3. Gender indicator databases

667. Countries disseminating gender statistics through dedicated databases of
gender indicators or more comprehensive ones on social indicators or human develop-
ment indicators should also consider this avenue of dissemination. Data disseminated
in this format can be useful to specialists interested in ready-processed indicators that
facilitate comparisons over time and between various groups of the population, or
analysis across indicators. When these databases include disaggregation by key vari-
ables such as regions or age groups, more meaningful analysis can be conducted by the
users. Information on the calculation of indicators included in the database, underly-
ing definitions and concepts used, and indication of the sources of data should be
made available with the database.

3.2.4. Production of metadata

668. Metadata provide essential technical information to data users about the
records contained in a data file, including the data source and the method used to col-
lect the data. Detailed metadata ensure appropriate use and accurate interpretation of
the data.
669. Information contained in the metadata includes the data collection method,
format of the file, sample design, unit of count, relationships among records, reference
period, aggregation of records, restrictions on the use of the data, indicators of data
quality and the names and definitions of all the variables in the file, including derived
variables that are essential for replicating the key survey outputs.
670. Indicators of data quality and accuracy will include response rates, item
non-response, imputations, sampling error and variation coefficients that will deter-
mine the reliability of the estimates.
671. Most national statistical agencies have developed standards and guidelines
for producing metadata that should be followed for surveys on asset ownership at the
individual level.

3.2.5. Sharing of microdata for researchers and academics

672. Researchers and academics interested in more complex analyses than
those presented in the dissemination products prepared by national statistical offices
180

will want to have access to the microdata. The microdata may only be made available
to interested third parties in a survey data file by the national statistical agency if the
confidentiality and anonymity of survey respondents can be guaranteed. No informa-
tion that could allow individuals to be identified should ever be made available.
673. The creation of a data file for access by people outside the survey team
requires an additional effort in order to produce high-quality documentation and
clean data files. Plans and policies for archiving, accessing and using the data should
be discussed and agreed upon before the data collection process begins. If agreements
about data release are not made at the beginning of the process, it will become increas-
ingly difficult for this to happen at later stages of the survey.

Key points
•• Data analysis is the component of the survey process that aims to provide answers to
the overarching questions on asset ownership that were identified by stakeholders
at the outset of the project. Those questions can be broadly categorized under three
objectives: first, understanding the gender asset gap; second, understanding the gen-
der wealth gap; and third, when more than one household member is interviewed,
undertaking intrahousehold analysis of asset ownership to better understand how
assets are allocated within households, in particular within couples, and the impact
of owning assets on intrahousehold decision-making power.
•• A household survey collecting data on asset ownership from a gender perspective has
a hierarchical set of units of observation—including the household, the individual (per-
son) and the asset—which defines the levels at which the information is collected.
•• When characteristics of assets—such as size, value or quality—are collected, the
asset is the unit of observation, in addition to the household and the individual. While
organizing the data file, a unique record exists for each asset and contains informa-
tion about the characteristics of the asset.
•• In addition to essential variables, such as whether a person owns a type of asset (e.g.,
the principal dwelling), characteristics of the asset, monetary value of the asset and the
gender of the ownders for a gender analysis of asset ownership and wealth, there are
other variables that may be considered for a nuanced understanding of asset owner-
ship from a gender perspective. These include age, marital status, type of household,
education, employment, intrahousehold decision-making power and other contextual
variables, such as urban versus rural or geographic and administrative areas.
•• The dissemination of data refers to the release of survey findings through various
statistical and analytical products, as well as the sharing of data files and associated
metadata. Dissemination products should include:
•• Data tables in both absolute numbers and in calculated proportions, share and
averages, made available in both publications and online databases;
•• Analytical publications that are clearly understood by the intended audiences;
•• Ready-processed indicators in gender indicator databases, to facilitate compari-
sons over time and across various population subgroups;
•• Microdata for more complex analysis.
•• All data products should be accompanied by appropriate metadata, including all
or some of the following items: data collection method, format of the file, sample
design, unit of count, relationships among records, reference period, aggregation of
records, restrictions on the use of the data, indicators of data quality, and names and
definitions of all variables in the file, including derived variables that are essential for
replicating the key survey outputs.
 181

Annex 1
Minimum set of questions for priority assets

Principal dwelling
D1. Do you own this dwelling?
••Yes, exclusively
••Yes, jointly
••No —> end of questions on ownership of principal dwelling
D2. Is there an ownership document for the dwelling?
••Yes
••No —> D4
D3. Are you listed as an owner on the ownership document for the dwelling?
••Yes, exclusively
••Yes, jointly
••No
D4. Do you have the right to sell this dwelling?
••Yes, exclusively
••Yes, jointly
••No
D5. Do you have the right to bequeath this dwelling?
••Yes, exclusively
••Yes, jointly
••No
Agricultural land
AL1. Do you own any agricultural land?
••Yes, exclusively
••Yes, jointly
••No —> end of questions on ownership of agricultural land
AL2. Is there an ownership document for the agricultural land?
••Yes
••No —> AL4
AL3. Are you listed as an owner on the ownership document for any agricultural land?
••Yes, exclusively
••Yes, jointly
••No
AL4. Do you have the right to sell any agricultural land?
••Yes, exclusively
••Yes, jointly
••No
182

AL5. Do you have the right to bequeath any agricultural land?

••Yes, exclusively
••Yes, jointly
••No
Other real estate
OL1. Do you own any of the following categories of other real estate?
Dwelling ••Yes, exclusively ••Yes, jointly ••No
Non-agricultural land ••Yes, exclusively ••Yes, jointly ••No
Other categories ••Yes, exclusively ••Yes, jointly ••No (if no to all
considered important categories —>
end of module)
OL2. Is there an ownership document for the following categories of other real estate?
Dwelling ••Yes ••No
Non-agricultural land ••Yes ••No
Other categories ••Yes ••No (if no to all categories —>
considered important OL4)
OL3. Are you listed as an owner on the ownership document for any of the following
categories of other real estate?
Dwelling ••Yes, exclusively ••Yes, jointly ••No
Non-agricultural land ••Yes, exclusively ••Yes, jointly ••No
Other categories ••Yes, exclusively ••Yes, jointly ••No
considered important
OL4. Do you have the right to sell any of the following categories of other real estate?
Dwelling ••Yes, exclusively ••Yes, jointly ••No
Non-agricultural land ••Yes, exclusively ••Yes, jointly ••No
Other categories ••Yes, exclusively ••Yes, jointly ••No
considered important
OL5. Do you have the right to bequeath any of the following categories of other real
estate?
Dwelling ••Yes, exclusively ••Yes, jointly ••No
Non-agricultural land ••Yes, exclusively ••Yes, jointly ••No
Other categories ••Yes, exclusively ••Yes, jointly ••No
considered important

Financial assets
FA1. Do you own any of the following categories of financial assets?
Currency and deposits ••Yes, exclusively ••Yes, jointly ••No
Microcredit ••Yes, exclusively ••Yes, jointly ••No
Equity ••Yes, exclusively ••Yes, jointly ••No
Informal savings group ••Yes, exclusively ••Yes, jointly ••No
Stocks and bonds ••Yes, exclusively ••Yes, jointly ••No
Insurance and pension fund ••Yes, exclusively ••Yes, jointly ••No
Other categories considered ••Yes, exclusively ••Yes, jointly ••No
important
 183

Annex 2
Model questionnaire

STRICTLY CONFIDENTIAL

EDGE PILOT SURVEY ON MEASURING ASSET OWNERSHIP FROM A

GENDER PERSPECTIVE
HOUSEHOLD QUESTIONNAIRE

PSU NUMBER SEGMENT:

DWELLING UNIT NUMBER:
RANDOM ADULT (I)..................................1
TYPE OF SELECTION: RANDOM ADULT AND SPOUSE/PARTNER (C)...............2
PROVINCE:
DISTRICT MUNICIPALITY:
LOCAL MUNICIPALITY:
ASSIGNMENT NUMBER:
MAIN INTERVIEW..................................1
QUESTIONNAIRE TYPE: SECONDARY INTERVIEW.............................2
HOUSEHOLD NUMBER FOR THIS HOUSEHOLD
HOUSEHOLD IDENTIFICATION [1A]
1A-2. TOTAL NUMBER OF HOUSEHOLDS AT SELECTED DWELLING
1A-3. PHYSICAL IDENTIFICATION OF THE DWELLING UNIT:
1A-4. GPS COORDINATES OF DWELLING:
N=1 S=2 D M
LAT

LONG

1A-5. MAIN LANGUAGE SPOKEN AT HOME: See Codesheet

1A-6. LANGUAGE OF INTERVIEW: See Codesheet
1A-7. DATE OF INTERVIEW (DAY, MONTH, YEAR): d d m m y y
184

STAFF DETAILS
ASSIGNMENT DATE
SUPERVISOR
NO. CHECKED d d m m y y
ASSIGNMENT DATE
2. Quality Control
NO. CHECKED d d m m y y
3. NAME OF SURVEY OFFICER:
4. REMARKS BY SURVEY OFFICER

5. REMARKS BY SUPERVISOR

Read the following statement of purpose confidently, and then give time for the respondent to ask questions.

The [name of NSO] is conducting a survey of households across [country] to better understand asset ownership in the country. The findings
from the survey will provide important information to the Government for developing policies and programmes to improve the lives of men
and women in [country]. Your household was selected as one of those to which the survey questions will be put. You were not selected for
any specific reason. Rather, your household was selected randomly from a list of all of the households in this village.

All the information that your household provides is strictly confidential. It will not be shared with any other government agency, and it will only
be used for statistical purposes by the [NSO] or under its supervision. To ensure that the most accurate information is collected, it is very
important that we interview the specific household member selected for the interview and that we interview him or her alone, without family or
neighbours present. If, during the interview, any family members or neighbours come within hearing distance of the interview, please ask
them kindly to come back later after the interview has been completed. Please spare some time to answer the questions. We thank you in
advance for your time.
 HOUSEHOLD ROSTER

201. 202. 203. 204. 205. 206. 207. 208.

IN ORDER TO
MAKE A
COMPREHENSIVE NAME Is What population What religion, if any, does RELATIONSHIP TO Has the household member stayed What is [NAME]'s exact How old is [NAME] in
LIST OF P
INDIVIDUALS
[NAME] group does [NAME] [NAME] practise? HEAD OF HOUSEHOLD: in this household for at least four date of birth? completed years?
E
CONNECTED TO PLEASE WRITE DOWN a male or belong to? nights on average per week for the
R CHRISTIAN.............1 last four weeks? CONFIRM THE
THE HOUSEHOLD, THE NAMES OF ALL female?
S MUSLIM................2 NUMBER YOU HAVE
USE THE
FOLLOWING PROBE O HOUSEHOLD MEMBERS ANCESTRAL, TRIBAL, ANIMIST,
ENTERED. CHECK
OR OTHER TRADITIONAL AFRICAN
QUESTIONS: N WHO STAYED IN THIS RELIGION......3 THE RESPONDENT'S
A HOUSEHOLD FOR AT BLACK AFRICAN...1
HINDU.................4 BIRTHDAY AGAINST
Just to make BUDDHIST..............5 HEAD................1 THE DATE OF THE
sure that I L LEAST FOUR NIGHTS ON COLOURED........2
BAHAI.................6 SPOUSE / PARTNER....2 YES...1
INDIAN/ ASIAN...3 INTERVIEW AND
have a complete AVERAGE FOR THE LAST WHITE...........4
JEWISH................7 SON/DAUGHTER........3 NO....2
listing: OTHER(SPECIFY).96
ATHEIST...............8 GRANDCHILD..........4 RECALCULATE THE
I FOUR WEEKS. START AGNOSTIC..............9 STEPCHILD...........5 AGE.
D WITH HEAD OF OTHER(SPECIFY).......96 PARENT OF HEAD OR
(a) Are there NOTHING IN SPOUSE..............6
any other HOUSEHOLD. ORDER OF PARTICULAR...........97 SISTER/BROTHER OF
persons such as C DO NOT KNOW..........98 HEAD OR SPOUSE......7
REST SHOULD BE FROM REFUSED..............99 NEPHEW/NIECE........8
small children O
or infants that D
ELDEST TO YOUNGEST. M....1
OTHER RELATIVE......9
we have not NON-RELATIVE.......10 DD MM YYYY NO. OF YRS
E F....2 OTHER (SPECIFY)....96
listed?

(b) Are there

any other 01
persons who
stayed in this 02
household for
at least four 03
nights on
average for the 04
last four
weeks? 05

06

07

Model questionnaire
08

09

10

11

12

13

14

15

16

17

185
 HOUSEHOLD ROSTER (CONTINUED)

186
209. 210. 211a. 211b. 211c. 211d. 211e. 212. 213. 214. 215.

What is [NAME]'s present marital

Is [NAME]'s Who is [NAME]'s spouse? What date did Does [NAME] Who is [NAME]'s What date did Does [NAME] How many ASK FOR In the last week (Monday to
P status?
spouse/ partner a IF [NAME] marry or have another second [NAME] marry or have a spouses/part ALL HH Sunday), did [NAME] work for a
E member of this RESPONDENT DOES NOT form a consensual spouse/partner spouse/partner? form a consensual spouse/partner ners does MEMBERS wage, salary, commission or any
NEVER MARRIED......1>>214
R DIVORCED...........2>>214 household? KNOW THE EXACT DATE union with that lives in this union with second living outside [NAME] have AGED 5 payment in kind (including paid
S SEPARATED..........3>>214 AND MONTH OF spouse/partner? household now? spouse/partner? this household living outside YEARS AND domestic work), even if it was for
OLLOWING PROBE O WIDOWED............4>>214 MARRIAGE, SELECT 1 now? this OLDER. only one hour? Examples: a
CUSTOMARY/ RELIGIOUS YES..1
JANUARY OF THE YEAR household? regular job, contract, casual or
N MARRIAGE,MONOGAMOUS.....5 NO...2>>211c
OF MARRIAGE. CHECK YES..1 What was the piece work for pay.
A CUSTOMARY/ RELIGIOUS NO...2>>212
L MARRIAGE, POLYGAMOUS....6 WHETHER THIS DATE highest
CIVIL MARRIAGE, MAKES SENSE IN LIGHT grade/class
ave a complete MONOGAMOUS..............7 YES..1
OF THE PERSON'S BIRTH that [NAME]
I CIVIL MARRIAGE, NO...2>>214
YEAR. completed?
D POLYGAMOUS..............8
COHABITING,
SINGLE PARTNER..........9 [Code book YES..1
ersons such as C COHABITATING, omitted.] NO...2>>217
O MULTIPLE PARTNERS......10
r infants that D NUMBER OF
E SPOUSES /
PARTNERS

01

02

03
verage for the 04

05

06

07

08

09

10

11

12

13

14

15

16

17
 HOUSEHOLD ROSTER (CONTINUED)

Ask 215-223 for all household members aged 15 years and older.

216. 217. 218. 219. 220. 221. 222.

Was this work (or any of In the last week (Monday to Sunday), did [NAME] Does [NAME] have a business that he or Is [NAME]’s business in: In the last week (Monday to Was this work (or any In the last week (Monday to
P this work) in farming, run or do any kind of business, big or small, for she will definitely return to? Sunday), did [NAME] help without of this work) in Sunday), did [NAME] do any
E forestry, raising animals or himself or herself or with one or more partners, Examples: Commercial farming, selling being paid in any kind of business farming, forestry, work on their own or the
R fishing? even if it was for only one hour? things, making things for sale, READ run by the household, even if it raising animals or household’s plot, farm or
S Examples: Commercial farming, selling things, construction, repairing things, guarding was for only one hour? fishing? food garden?
Farming, forestry,
OLLOWING PROBE O construction, repairing things, guarding cars, cars, brewing beer, collecting wood or raising animals or Examples: Commercial farming, Examples: ploughing,
N brewing beer, collecting wood or water for sale, water for sale, hairdressing, creche fishing….........1 help to sell things, make things for harvesting, looking after
hairdressing, creche businesses, taxi or other businesses, taxi or other transport Sector other than sale or exchange. livestock
A
agriculture or
L transport business, running a legal or medical business,
practice, performing in public, operating a public running a legal or medical practice,
ave a complete
phone shop, etc. performing in public, operating a public
I
phone shop, etc. YES..1 YES..1
D YES..1
ENUMERATOR: IF [NAME] HAS MORE NO...2>>222 NO...2 NO...2
YES..1 THAN 1 BUSINESS, ENTER CODE 2 IF
C YES..1 YES..1
ersons such as NO...2
NO...2>>220
NO...2 AT LEAST 1 BUSINESS IS IN SECTOR
O
r infants that OTHER THAN AGRICULTURE OR
D
FISHING.
E

01

02

03
verage for the 04

05

06

07

Model questionnaire
08

09

10

11

12

13

14

15

16

17

187
 188
END OF HOUSEHOLD QUESTIONNAIRE

223. 224a. ENUMERATOR: ENTER RESPONSE CODE FOR COMPLETION STATUS OF HOUSEHOLD QUESTIONNAIRE:

In the last week (Monday to Sunday), did [NAME] catch any fish, prawns, shells, wild animals or other food for household
P consumption? COMPLETED..............................11>>228
E PARTIALLY COMPLETED....................12>>end of interview
NON-CONTACT............................21>>end of interview
R NOT INTERVIEWED, REFUSAL...............22
S OTHER NON-RESPONSE.....................23
UNOCCUPIED DWELLING....................31>>end of interview
FOLLOWING PROBE O
VACANT DWELLING........................32>>end of interview
N DEMOLISHED.............................33>>end of interview
A NEW DWELLING UNDER CONSTRUCTION........34>>end of interview
STATUS CHANGED.........................35>>end of interview
L LISTING ERROR..........................36>>end of interview
have a complete NOT INTERVIEWED, NON-HOUSEHOLD MEMBER..37>>end of interview
I
D YES..1 224b. REASON(S) FOR INTERVIEW NOT FULLY COMPLETED OR NOT ADMINISTERED SHOULD BE EXPLAINED.
NO...2

persons such as C
O 225. NAME OF PRIMARY RESPONDENT.
or infants that D
E

01

02

03
average for the 04

05

06

07

08

09

10

11

12

13

14

15

16

17
 Model questionnaire 189

STRICTLY CONFIDENTIAL

EDGE PILOT SURVEY ON MEASURING ASSET OWNERSHIP FROM A

GENDER PERSPECTIVE
INDIVIDUAL QUESTIONNAIRE
(a) THE NAME OF THE RANDOMLY SELECTED RESPONDENT IS LISTED BELOW AS THE FIRST OPTION.
(b) IF APPLICABLE, THE NAME OF THE SPOUSE OR PARTNER IS LISTED IN THE SECOND OPTION GENERATED
BY CAPI.
1. NAME OF SELECTED ADULT HOUSEHOLD MEMBER(S):
2. NAME(S) OF ENTREPRENEURS IN THE HOUSEHOLD:
3. LIST ALL THE RESPONDENTS TO BE INDIVIDUALLY INTERVIEWED ON THIS TABLET BY GIVING
THE NAME AND THE TYPE OF RESPONDENT (EXAMPLE, JOHN – RANDOM PERSON; JANE –
SPOUSE; KATE – ENTREPRENEUR):
4. SELECT WHICH HOUSEHOLD MEMBER IS BEING INTERVIEWED:
RANDOMLY SELECTED PERSON............1
5. SELECT THE TYPE OF RESPONDENT [NAME]: SPOUSE/PARTNER OF RANDOMLY SELECTED
PERSON..............................2
ENTREPRENEUR........................3
YES...1
6. IS THE RESPONDENT AVAILABLE FOR THE INTERVIEW NO....2
DURING YOUR TIME IN THE PSU?
7. DATE OF INTERVIEW:
THIS SURVEY IS BEING CONDUCTED BY [NAME OF NSO]
If the respondent is different from the one completing the household questionnaire, please read confidently the statement of purpose
given below, and then give time for the respondent to ask questions, before soliciting information.
The [name of NSO] is conducting a survey of households across [country] to better understand asset ownership in the country. The
findings from the survey will provide important information to the Government for developing policies and programmes to improve the
lives of men and women in [country]. Your household was selected as one of those to which the survey questions will be put. You were
not selected for any specific reason. Rather, your household was selected randomly from a list of all of the households in this village.
All the information that your household provides is strictly confidential. It will not be shared with any other government agency, and it will
only be used for statistical purposes by the [NSO] or under its supervision. To ensure that the most accurate information is collected, it is
very important that we interview the specific household member selected for the interview and that we interview him or her alone, without
family or neighbours present. If, during the interview, any family members or neighbours come within hearing distance of the interview,
please ask them kindly to come back later after the interview has been completed. Please spare some time to answer the questions. We
thank you in advance for your time.
 190

DWELLING CHARACTERISTICS Characteristics of principal dwellings are well covered in many household surveys and are often used as proxies for
household economic status. If individual-level asset ownership data are collected through appending a module or a set of
questions to an existing survey, then it is not necessary to collect such information again.
If there are multiple respondents within the household, it is suggested that values of dwellings are collected at the

DA1. DA2. DA3. DA4. DA5.

What type of dwelling is this? What type of material is mainly What type of material is mainly What type of material is mainly What type of toilet is mainly used in
used for construction of the roof? used for construction of the walls? used for construction of the floor? your household?
DETACHED HOUSE ...........1
SEMI-DETACHED HOUSE.......2 IRON SHEETS...........1 CONCRETE/STONES.......1 EARTH.................1
FLAT IN A BLOCK OF TILES.................2 CEMENT BLOCKS.........2 CEMENT SCREED.........2 FLUSH TOILET..............1
FLATS.....................3 ASBESTOS..............3 BRICKS................3 CONCRETE..............3 VIP LATRINE...............2
ROOM OR ROOMS OF A MAIN HOUSE CONCRETE..............4 WOOD..................4 TILE..................4 COVERED PIT LATRINE WITH A
..................4 TRADITIONAL MUD AND POLES.........5 BRICK.................5 SLAB......................3
TRADITIONAL TIN...................5 TIN/IRON SHEETS.......6 STONE.................6 COVERED PIT LATRINE WITHOUT A
DWELLING/HUT/STRUCTURE MADE THATCH................6 OTHER (SPECIFY)..........96 WOOD..................7 SLAB....................4
OF TRADITIONAL OTHER (SPECIFY)..........96 OTHER (SPECIFY)..........96 UNCOVERED PIT LATRINE WITH A
MATERIALS.................5 SLAB....................5
OTHER (SPECIFY)..........96 UNCOVERED PIT LATRINE WITHOUT A
SLAB............6
NO FACILITY - BUSH, POLYTHENE
BAGS, BUCKET,
ETC.......................7
OTHER (SPECIFY)..........96

The response should refer to the The response should refer to the The response should refer to the The response should refer to the Refers to use rather than
characteristics of the biggest part material that covers the largest part material that covers the largest part material that covers the largest part ownership.
of the dwelling unit of the roof. The quality of the of the wall. The quality of the of the floor. The quality of the
material does not matter. material does not matter. material does not matter.

DWELLING CHARACTERISTICS (CONTINUED)

DA6.

is mainly used in If this dwelling were to be sold today, how much could be received for it?

T LATRINE WITHOUT A

Should be the current value based on the location and condition of their particular dwelling. The full amount that would be received in the sale should be listed, regardless whether or not all of it
would be kept by the respondent.

If the respondent is not sure how to answer, enumerators should probe on this question by encouraging the respondent to consider the price received for similar houses sold in the community.

Note that asking about the price that would be received today refers to the current value, not a lower price that would be obtained from a distress sale or fire sale that had to take place
immediately.

If there are large areas of the country with no market for dwellings, other measures may be considered. Such measures could n i clude the cost of constructing a similar dwelling (investigators
should specify whether the cost of the lot should be included), or the amount that they could receive if they rented it out.If there is no market, investigators may want to use the information
collected in the household questionnaire on the characteristics of the dwelling to impute a value.
DWELLING OWNERSHIP
DB1. DB2. DB3. DB4. DB5.

Do you own this dwelling? How many other people jointly own Is one of these joint owners your Is there an ownership document for Are you listed as an owner on the
this dwelling with you, including spouse or partner? the dwelling? ownership document for the
household members and non- YES, A TITLE DEED.........1 dwelling?
household members? YES, A CERTIFICATE OF
CUSTOMARY OWNERSHIP.......2
YES,A CERTIFICATE
YES,
YES...........1 OF OCCUPANCY..............3 YES,ALONE..........1>>DB8
ALONE................1>>DB4
NO............2 YES, A WILL...............4 YES, JOINTLY WITH SOMEONE
YES, JOINTLY WITH SOMEONE
DON'T KNOW....98 YES, A SALES AGREEMENT....5 ELSE...............2
ELSE.................2
YES, OTHER (SPECIFY).....96 NO.................3>>DB8
NO, SOMEONE ELSE IS THE
NO...................6>>DB8 REFUSES TO
OWNER...............3 >>next
RESPOND............97>>DB8
module
DON'T KNOW.........98>>DB8
REFUSES TO
RESPOND.............97>>next
module
The question captures the If the respondent shares reported The question allows for the If an ownership document exists for Important to ask if the respondent
respondent's self-perception of ownership of the dwelling, construction of an indicator on joint the dwelling, it should be recorded is listed “as an owner” on the
his/her ownership status, information on the number of joint (reported) ownership between regardless whether or not it has the document because individual
irrespective of whether his/her reported owners is needed for spouses/partners, the most name of someone in the household names can be listed as witnesses
name is listed as an owner on an calculation of the gender wealth common form of joint ownership. on it. on an ownership document.
ownership document for the gap.
dwelling. Countries interested in identifying If there is more than one type of While countries may want to ask
joint ownership between siblings or document, the one that is held by the respondent to produce the
The question is also a screening a parent and an adult child are someone in the household should be document for the enumerator so
question for this module. If the encouraged to ask “Who are the recorded. For example, if there is a that the enumerator can confirm
response is "no" or "refuse to joint owners, including household deed, but the household member that the respondent’s name is
respond", move to the next asset. members and non-household does not have it, but has an invoice or listed on the document, these
members?” The personal IDs sales receipt, list the invoice, not the guidelines recommend that the
assigned to household members in deed. measure of documented

Model questionnaire
the household roster should be ownership not be conditional on
recorded for each household Titles and deeds are one form of the document being checked or
member who is a joint reported ownership document. Registration kept within the home.
owner and each non-household certificates document rights over
member who jointly owns the property. In addition, where titling or
dwelling should be assigned a registration is not complete,
non-household member ID code documents including wills or sales
(e.g. 100). receipts provide some form of
documented claim. If the dwelling is a
co-op, then the person may have
shares in the co-op rather than a
deed.

Countries should customize the

response categories according to

191
 DWELLING OWNERSHIP (CONTINUED)

192
DB6. DB7. DB8. DB9.

How many other people are listed Is one of these joint owners your Do you have the right to sell this dwelling? Is one of the persons who jointly has
as owners on the ownership spouse or partner? the right to sell the dwelling your
document, including household spouse or partner?
members and non-household
members?
YES...........1 YES, ALONE............1>>DB10 YES...........1
NO............2 YES, JOINTLY WITH SOMEONE NO............2
DON'T KNOW....98 ELSE.....................2 DON'T KNOW....98
NO, SOMEONE ELSE HAS THIS
RIGHT...................3>>DB10
NO, IT CANNOT BE SOLD...4>>DB10

he respondent If the respondent shares The question allows for the Having the right to sell the dwelling If the respondent’s spouse/partner Having the right to bequeath the
documented ownership of the construction of an indicator on joint means that the person has the right to was identified as a joint reported or dwelling means that the person has
dwelling, information on the documented ownership between permanently transfer the dwelling to documented owner of the dwelling, the right to give the dwelling by oral
d as witnesses number of joint reported owners is spouses/partners, the most another person or entity for cash or in collecting information on whether or written will to another person or
needed for calculation of the common form of joint ownership. kind benefits. the spouse/partner jointly has the persons upon their death.
gender wealth gap. right to sell the dwelling enables
Countries interested in identifying To assess gender differences in the analysis of whether joint owners To assess gender differences in the
joint ownership between siblings or right to sell the dwelling, it is useful to have the same rights to the right to bequeath the dwelling, it is
a parent and an adult child are distinguish between the two “no” dwelling. useful to distinguish between the two
encouraged to ask “Who else is answers, identifying if the respondent is “no” answers, identifying if the
listed as an owner on the not the one who can sell the dwelling If countries choose to collect respondent is not the one who can
ownership document, including (but someone else can sell it) or if the information on all joint reported and bequeath the dwelling (but someone
household members and non- dwelling cannot be sold (for example, documented owners in Qs DB3 else can bequeath it) or if the
household members?” The owing to cultural or legal norms). and DB7, respectively, then dwelling cannot be bequeathed (for
personal IDs assigned to countries can ask, in place of this example, because of cultural or legal
household members in the question, “Which other household norms).
household roster should be members also have the right to sell
recorded for each household this dwelling?” The personal IDs
member who is a joint documented assigned to household members in
owner and each non-household the household roster should be
member who jointly owns the recorded for each household
dwelling should be assigned a non- member who has the right to sell
household member ID code (e.g. the dwelling.
100).
Note that information on the
number of non-household
members who have this right is not
needed to calculate the gender
wealth gap for dwellings.
 DWELLING OWNERSHIP (CONTINUED)
DB10. DB11. DB12. DB13.

Do you have the right to bequeath Is one of the persons who jointly has How did you acquire this dwelling? Do you grow any food, either from crops or
this dwelling? the right to bequeath the dwelling animals, or raise any livestock on the plot of
your spouse or partner? land on which the dwelling is located, mainly
for the household's consumption?

YES, ALONE............1>>DB12 YES...........1 PURCHASED...................................1

YES, JOINTLY WITH SOMEONE YES...........1
NO............2 INHERITED ..................................2
ELSE.....................2 NO............2
DON'T KNOW....98 RECEIVED AS A GIFT..........................3
NO, SOMEONE ELSE HAS THIS DON'T KNOW....98
BUILT IT....................................4
RIGHT...................3 ALLOCATED BY GOVERNMENT PROGRAMME...........5
NO, IT CANNOT BE THROUGH MARRIAGE............................6
BEQUEATHED...4>>DB12 OTHER(SPECIFY).............................96

Having the right to bequeath the If the respondent’s spouse/partner The question refers to when the respondent first came The question captures whether the plot of
dwelling means that the person has was identified as a joint reported or into possession of the asset and presumably began land on which the dwelling is located is a
the right to give the dwelling by oraldocumented owner of the dwelling, deriving economic benefit from it. family garden. Because women commonly
or written will to another person or collecting information on whether rely on these parcels for subsistence
persons upon their death. the spouse/partner jointly has the Countries should include all relevant modes of production but such parcels are often not
right to bequeath the dwelling acquisition and may want to add additional codes for measured by agricultural surveys because of
To assess gender differences in the enables analysis of whether joint when dwellings are received as an inheritance or as a their small size, collecting information on
right to bequeath the dwelling, it is owners have the same rights to the gift to indicate who gave the inheritance/gift (e.g. the family gardens can provide a better
useful to distinguish between the two dwelling. respondent’s natal family or the spouse’s family). This is understanding of their prevalence and
“no” answers, identifying if the particularly useful for gender analyses, since the contribution to agricultural productivity and
respondent is not the one who can If countries choose to collect information collected can indicate whether the dwelling women’s food security.
t reported and bequeath the dwelling (but someone information on all joint reported and was received from the husband’s family or the wife’s
else can bequeath it) or if the documented owners in Qs DB3 family. According to the World Census of Agriculture
dwelling cannot be bequeathed (for and DB7, respectively, then 2020 guidelines, the land in family gardens

Model questionnaire
example, because of cultural or legal countries can ask, in place of this are considered agricultural land as well.
norms). question, “Which other household Therefore if the answer to this question is
members also have the right to "Yes", then the plot of land on which the
bequeath this dwelling?” The dwelling is located is considered as
personal IDs assigned to agricultural land, and should be included in
household members in the the module on agricultural land.
household roster should be
recorded for each household
member who has the right to
bequeath the dwelling.

Note that information on the

number of non-household
members who have this right is not
needed to calculate the gender
wealth gap for dwellings.

193
 AGRICULTURAL LAND CHARACTERISTICS

194
This module can be included at the household level or at the individual level. If more than one respondent is
interviewed for ownership of assets within the household, it is suggested that questions pertaining to the agricultural
land roster and characteristics be asked at the household level.

If this module is incorporated at the household level, the respondent should be asked to list all agricultural parcels
that are owned by the household.

If the module is asked, at individual level, of one randomly selected person within the household, the respondent
should be asked to list all agricultural parcels that the respondent owns

AL-A1. AL-A2. AL-A3. AL-A4. AL-A5. AL-A6.

PARCEL NAME AND Is this [agricultural parcel] What is the size of this [parcel]? Is this [parcel] irrigated? What was the primary use of this [parcel] If this [parcel] were to be sold
DESCRIPTION located inside or outside of the during the last cropping season? today, how much could be
country? YES..........1 received for it?
LIVESTOCK
NO...........2
PRODUCTION.............1
INSIDE...........1 POULTRY
PRODUCTION.............2 Respondents should
OUTSIDE..........2
GRAINS AND LEGUMES.....3 estimate the current value
COMMERCIAL CROPS.......4 based on the location and
FRUIT AND VEGETABLE quality of their particular
This question is only In the absence of GPS-based This question measures PRODUCTION............5 parcel. The full amount
The response should be listed whether the parcel uses FODDER, GRAZING PASTURE OR
necessary if countries wish area parcel measurements, this that would be received in
from largest to smallest any form of irrigation or GRASS FOR ANIMALS..6
to use the data collected on question measures the size of the sale should be listed,
P agricultural parcels owned. The relies entirely on rainfall FISH FARMING/
agricultural land for the parcel as self-reported by regardless whether or not
response should include the and can serve as a proxy
AQUACULTURE...........7
A updating their national the respondent and can serve FORESTRY PLANTATION...8 all of it would be kept by
name for each parcel (e.g. “road to assess differences in
R accounts as land located as a proxy to assess differences FALLOW................9 the respondent. If the
parcel” or “swamp parcel”) that the quality of land owned
C outside of the country is not in the quality of land owned by WOODLAND/ FOREST.....10 respondent is not sure
the respondent reports owning, by women and men. SWAMP................11
included in the 2008 SNA. women and men. The how to answer,
E and a brief description of each RENTED OUT...........12
parcel that can be referenced respondent should estimate the enumerators should
L DON'T KNOW...........98
during the interview. This size of the parcel in the units of OTHER (SPECIFY)......96 probe on this question by
information should be provided measure commonly used in the encouraging the
I national survey programme This question captures the primary use of respondent to consider
for each agricultural parcel that
D (e.g. hectares, acres). the agricultural parcel during the last the price received for
the respondent reports owning
cropping season and can provide for other parcels that have
before proceeding to the next
further disaggregation of gender statistics been sold in the area. If
question.
on land ownership. If the parcel had more markets are thin for land,
than one use, the activity constituting the investigators may want to
If more than one person is
largest use of the land should be use information on the
selected for interview it is
considered primary. characteristics of the plot
suggested that a roster is
collected at the household level. (such as the size of the
parcel and irrigation
status) so that a value
can be imputed.

AREA IN HECTARES

A01 __ __ __ __ . __ __

A02 __ __ __ __ . __ __

A03 __ __ __ __ . __ __

A04 __ __ __ __ . __ __

A05 __ __ __ __ . __ __

… __ __ __ __ . __ __
AGRICULTURAL LAND OWNERSHIP This is the screening question to determine whether the respondent should complete the module. It also measures
reported ownership of agricultural land.
YES..........1
Countries may want to consider a different formulation of the question such as "Do you currently hold, have, use or
AL-B1. Do you own any agricultural parcels? NO...........2 >> END OF occupy agricultural land?” in order to broaden the scope of the question and embrace rights over the land which are
MODULE not ownership sensu stricto.
AL-B2. AL-B3. AL-B4. AL-B5. AL-B6.
Does anyone jointly own this How many other persons Is one of these joint owners Is there an ownership document
List all parcels owned by the respondent, Referring [parcel] with you, including jointly own this [parcel] your spouse or partner? for this [parcel]?
to the parcel ID if parcel roster and characteristics household members and non- with you?
are collected at the household level household members?
YES, A TITLE DEED.....1
YES........1 YES, A CERTIFICATE OF
NO.........2 CUSTOMARY OWNERSHIP...2
YES...1 YES, A CERTIFICATE
NO....2>>AL-B6 OF OCCUPANCY..........3
YES, A WILL...........4
YES, A PURCHASE
P AGREEMENT..5
YES, OTHER (SPECIFY).96
A
NO.............7>>AL-B10
R
C
E
L
The question measures the If the respondent shares Identifying whether the If an ownership document
I form of reported ownership reported ownership of respondent jointly owns the exists for the parcel, it should
of the agricultural parcel; i.e. the agricultural parcel, agricultural parcel with be recorded independent of
D information on the her/his spouse or partner whether it has the name of
whether the respondent
owns the parcel exclusively number of joint reported allows for the construction of someone in the household on it.
or jointly with one or more owners is needed for an indicator on joint If there is more than one type of
persons. Because the calculation of the (reported) ownership of document, the one that is held
benefits of ownership may gender wealth gap, as agricultural land between by someone in the household
differ if an individual owns discussed in part four of spouses, the most common should be recorded. For
the parcel alone or jointly, these guidelines. form of joint ownership. Other example, if there is a deed, but
countries are encouraged to patterns of joint ownership the household member does
collect information on the are also possible, such as not have it, but has an invoice
form of reported ownership. between siblings or a parent or sales receipt, list the invoice,
and an adult child, and not the deed.
countries that are interested
in identifying these patterns
are encouraged to ask, in

Model questionnaire
place of AL-B4 and AL-B5,
“Who are the joint owners,
including household
members and non-household
members?” The personal IDs
assigned to household
members in the household
roster should be recorded for
each household member who
is a joint reported owner and
each non-household member
who jointly owns the
agricultural parcel should be
assigned a non-household
member ID code (e.g. 100).

195
 AGRICULTURAL LAND OWNERSHIP (CONTINUED) A

196
AL-B7. AL-B8. AL-B9. AL-B10. AL-B11. AL-B12. A
Are you listed as an owner on How many other people are listed as Is one of these joint owners Do you have the right to sell this [parcel]? Is one of the persons who jointly Do you have the right to bequeath this Is
the ownership document for owners on the ownership document, your spouse or partner? has the right to sell this [parcel] [parcel]? jo
this [parcel]? including household members and your spouse or partner? th
non-household members? pa
YES, ALONE..............1>>AL-B12 YES, ALONE............1>>AL-B14
YES, ALONE...1>>AL-B10 YES.........1 YES, JOINTLY WITH ONE YES, JOINTLY WITH ONE
YES, JOINTLY WITH ONE OR NO..........2 OR MORE PERSONS......2 OR MORE PERSONS....2
MORE PERSONS......2 YES.........1
NO, SOMEONE ELSE NO, SOMEONE ELSE
NO...........3>>AL-B10 NO..........2
HAS THIS RIGHT.......3>>AL-B12 HAS THIS RIGHT.....3>>AL-B14
NO, IT CANNOT NO, IT CANNOT
BE SOLD..............4>>AL-B12 BE SOLD............4>>AL-B14

If the respondent’s
spouse/partner was
identified as a joint reported
Documented ownership If the respondent shares Identifying whether both the Refers to the respondent's right to or documented owner of the When a respondent has the right to
refers to the existence of any documented ownership of the respondent’s name and the permanently transfer the parcel to parcel, collecting information bequeath the parcel, it means that
document an individual can parcel, information on the name of his or her another person or entity for cash or in on whether the she/he has the right to give the
use to claim ownership number of joint documented spouse/partner are listed as kind benefits. spouse/partner jointly has parcel by oral or written will to
the household on it. rights in law over an owners is needed for owners on the ownership the right to sell the parcel another person or persons upon the
ore than one type of agricultural parcel by virtue calculation of the gender document allows for the To assess gender differences in the right enables analysis of whether death of the respondent.
of the individual’s name wealth gap, as discussed in construction of an indicator on to sell the parcel, it is useful to joint owners have the same
being listed as an owner on part four of these guidelines. joint (documented) ownership distinguish between the two “no” rights to the parcel. If To assess gender differences in the
the document. Because between spouses/partners, answers, identifying if the respondent is countries choose to collect right to bequeath the parcel, it is
here is a deed, but individual names can be the most common form of not the one who can sell the parcel (but information on all joint useful to distinguish between the two
listed as witnesses on an joint ownership. Other someone else can sell it) or that the reported and documented “no” answers, identifying if the
ownership document, it is patterns of joint documented parcel cannot be sold (for example, owners in AL-B4 and AL-B8, respondent is not the one who can
eipt, list the invoice, important to ask if the ownership are also possible, because of cultural or legal norms). respectively, then countries bequeath the parcel (but someone
respondent is listed “as an such as between siblings or a can ask, in place of this else can bequeath it) or that the
owner” on the document. parent and an adult child, and question, “Which other parcel cannot be bequeathed (for
countries that are interested household members also example, because of cultural or legal
While countries may want to in identifying these patterns have the right to sell this norms).
ask the respondent to are encouraged to ask, in [parcel]?” The personal IDs
produce the document for place of AL-B8 and Al-B9, assigned to household
the enumerator so that the “Who else is listed as an members in the household
enumerator can confirm that owner on the ownership roster should be recorded
the respondent’s name is document, including for each household member
listed on the document, household members and non- who has the right to sell the
these guidelines recommend household members?” The parcel. Note that information
that the measure of personal IDs assigned to on the number of non-
documented ownership not household members in the household members who
be conditional on the household roster should be have this right is not needed
document being checked. recorded for each household to calculate the gender
member who is a joint wealth gap for agricultural
documented owner, and each land.
non-household member who
jointly owns the parcel should
be assigned a non-household
member ID code (e.g. 100).
AGRICULTURAL LAND OWNERSHIP (CONTINUED)

AL-B13. AL-B14. AL-B15. AL-B16. AL-B17. AL-B18.

Is one of the persons who Are you the decision-maker Is one of the joint decision-makers for this How did you acquire this [parcel]? What could make you lose ownership of this SELECT THE METHOD WHICH
jointly has the right to bequeath on this [parcel] regarding the [parcel] your spouse or partner? parcel over the next five years? WAS USED TO INTERVIEW THE
this [parcel] your spouse or timing of crop activities, crop RESPONDENT.
partner? choice, and input use? PURCHASED.............1
YES........1 INHERITED.............2
NO.........2 RECEIVED AS A GIFT....3
ALLOCATED BY GOVERNMENT EVICTION BY FAMILY OR
YES, ALONE......1>>AL- CLAN................1
B16 PROGRAM...............4
YES..........1 THROUGH MARRIAGE......5 SOMEONE WITH POLITICAL RESPONSE CODES:
YES, JOINTLY WITH ONE OR INFLUENCE WANTS THIS PIECE
NO...........2 MORE PERSONS.......2 OTHER (SPECIFY)......96
OF LAND.......2 ALONE........................1
ALONE........................1
NO, SOMEONE ELSE IS THE DEATH OF HOUSEHOLD
DECISION MAKER..3>>AL- WITH ADULT FEMALES
HEAD................3 PRESENT......................2
PRESENT......................2
B16 DEATH OF SPOUSE.....4
NO, THE PARCEL IS RENTED WITH ADULT MALES PRESENT.....3
LAND REDISTRIBUTION BY
OUT.............4>>AL- GOVERNMENT..........5
WITH ADULTS MIXED SEX
B16 EVICTION BY PRESENT......................4
SHERIFF/COURT.......6 WITH CHILDREN
NOTHING.............7 PRESENT......................5
If the respondent’s spouse/partner was Refers to when the respondent first came into I BELIEVE I WILL OWN THIS WITH ADULTS MIXED SEX
If the respondent’s identified as a joint reported or possession of the asset and presumably began LAND IN 5 AND CHILDREN PRESENT.........6
PRESENT.........6
spouse/partner was This question measures documented owner of the parcel, deriving economic benefits from it. Because women YEARS...............8
identified as a joint whether the owner of the collecting information on whether the and men often acquire assets by different means, OTHER (SPECIFY)....96
reported or documented agricultural parcel is also spouse/partner is a joint decision- understanding the modes of acquisition may provide
owner of the parcel, the decision-maker, or maker regarding the timing of crop insights for developing policies to ensure women’s
collecting information on among the decision- activities, crop choice, and input use ability to acquire them. As such, NSOs should include
whether the makers, for the enables analysis of whether joint all relevant modes of acquisition and may want to add
spouse/partner jointly has operations of the owners jointly make decisions additional codes for when parcels are received as an Response should reflect the
the right to bequeath the agricultural parcels. concerning agricultural production. inheritance or as a gift to indicate who gave the respondent’s perception of his or her
parcel enables analysis of Knowing whether the inheritance/gift (e.g. the respondent’s natal family or tenure security or the likelihood that a
whether joint owners have owners of the parcel If countries choose to collect the spouse’s family). claim will be made against his/her
the same rights to the (and their sex) make information on all joint reported and
ownership rights.
parcel. If countries choose agricultural production documented owners in AL-B4 and AL- This is particularly useful for gender analyses, since
to collect information on all decisions is useful for B8, respectively, then countries can (Reason(s) interview not administered with the
the information collected can indicate whether the This question was tested in the South
joint reported and policy targeting ask, in place of this question, “Which parcel was received from the husband’s family or the respondent(s) alone should be explained in the
African pilot survey, The question
documented owners in AL- increased household other household members are the joint wife’s family. survey officer remarks section).
proposed to measure perceptions of
B5 and AL-B9, agricultural productivity. decision-makers for this [parcel]
tenure security under Sustainable
respectively, then regarding the timing of crop activities,
Development Goal indicator 1.4.2
countries can ask, in place crop choice, and input use?” The
identifies the likelihood of the
of this question, “Which personal IDs assigned to household
respondent involuntarily losing
other household members members in the household roster
ownership/use rights to the parcel in
also have the right to should be recorded for each household
the next five years “on a scale from 1
bequeath this [parcel]?” member who is a joint decision-maker.

Model questionnaire
to 5, with 1 being not at all likely and 5
The personal IDs assigned Note that information on the number of
being extremely likely”.
to household members in non-household members who are joint
Source: Food and Agriculture
the household roster decision-makers is not needed to
Organization of the United Nations,
should be recorded for calculate the gender wealth gap for
“Measuring individuals’ rights to land:
each household member agricultural land.
an integrated approach to data
who has the right to
collection for SDG indicators 1.4.2 and
bequeath the parcel. Note
5.a.1” (forthcoming).
that information on the
number of non-household
members who have this
right is not needed to
calculate the gender
wealth gap for agricultural
land.

197
 LIVESTOCK

198
Livestock are an important source of income and means of wealth accumulation. Livestock are valued for breeding, for the various foods and goods that they
produce, and for their role in transport and work-related activities. Given, however, the variety of livestock often owned by households and individuals, this
category poses a number of challenges regarding identifying ownership. Among agro-pastoralist and pastoralist households, there may be many animals with
various configurations of rights and ownership. For this reason, the simplest approach, presented below, is to ask whether the respondent owns any of the
categories of livestock on which the country wants to collect data. This approach only provides information on the prevalence and form (exclusive or joint) of

LV-1 LV-2 LV-3

LIVE LIVESTOCK Do you own any [livestock category], exclusively or jointly? SELECT THE METHOD WHICH WAS USED TO INTERVIEW THE
STOCK NAME RESPONDENT.
CODE YES, EXCLUSIVELY ONLY......................1
YES, JOINTLY ONLY..........................2 RESPONSE CODES:
YES, BOTH EXCLUSIVELY AND JOINTLY..........3
NO..........................4>>NEXT CATEGORY
ALONE........................................1
WITH ADULT FEMALES PRESENT...................2
WITH ADULT MALES PRESENT.....................3
WITH ADULTS MIXED SEX PRESENT................4
This question measures reported ownership of livestock, and also the WITH CHILDREN PRESENT........................5
form of ownership, i.e. whether the livestock is owned exclusively or WITH ADULTS MIXED SEX AND CHILDREN PRESENT...6
jointly by the respondent. Categories may include but not be limited to
those that are listed, and distinctions may also be made between exotic
and indigenous breeds. Countries will need to determine the categories
of livestock to include based on prevalence rates from prior agricultural (Reason(s) interview not administered with the respondent(s) alone
or household surveys and also on policy needs. should be explained in the survey officer remarks section).

For some policy purposes, the approach presented above may not be
sufficient because it may be important to gather more detail about
ownership patterns and to establish the value of livestock. Establishing
the value of livestock tends to be easier than obtaining values for land.
In most places where people raise livestock, there is an active livestock
market. The challenge with valuing livestock is that, if a person owns
L01 Cattle five head of cattle, they may each have a quite different sales price,
depending on their sex, age, and condition. Thus, asking about how
L02 Goats much would be received if one of the animals is sold may not reflect
L03 Sheep the average price of the animals. In the EDGE pilot in Uganda, the
livestock module was separated into large livestock and small livestock
L04 Pigs and poultry. For large livestock, the respondents were asked to report
the total number of each type of livestock that they owned, how many
L05 Horses
animals were owned exclusively or jointly, the personal IDs of each
L06 Donkeys joint owner and the total amount that could be received if all of the
livestock for a given ownership arrangement were sold in the market.
L07 Chicken Because the module proved operationally difficult to implement, it is not
L08 Other poultry/ birds (ducks, geese, etc.) presented here, but countries can refer to the EDGE survey
instruments for Uganda for more detail.
L09 Game livestock (antelope, buffalo, etc.)
All EDGE survey instruments are available at:
L10 Other (specify)
http://unstats.un.org/edge
L11 Not applicable/ no livestock
 Model questionnaire 199

LARGE AGRICULTURAL EQUIPMENT

LAE-1. Do you own any large agricultural equipment, such as tractors, ploughs, irrigation systems or trailers?

YES.........1
NO..........2>>NEXT MODULE
REFUSES TO RESPOND....97>>NEXT MODULE

This is the screening question to determine whether the respondent should complete the module. It measures the reported owner ship of any agricultural equipment (not by type of
agricultural equipment). Categories of large agricultural equipment may include, but are not limited to, ploughs, ox -ploughs, tractors, trailers, threshers, irrigation systems and
spraying machines.

Countries will need to determine the categories of large agricultural equipment to include based on prevalence rates from pri or agricultural or household surveys, and also policy
needs. If the respondent does not own any large agricultural equipment, skip to the next module as the remainder of this modu le is only administered to the respondent if she/he
self-reports owning any large agricultural equipment.

The roster of large agricultural equipment can be established at the household level or at the individual level. If more than one respondent is interviewed for ownership of assets
within the household, it is suggested that the roster be asked at the household level. In this case, the household level resp ondent should be asked to list all pieces of large
agricultural equipment that are owned by the household. If the module is asked, at individual level, of one randomly selected person within the household, the respondent should
be asked to list all large agricultural equipment that the respondent owns

LAE-2. LAE-3. LAE-4.

E Please list each piece of large agricultural equipment that Does anyone jointly own this [agricultural equipment] with you, including How many other persons jointly own this
you own exclusively or jointly with someone else. household members and non-household members? [agricultural equipment] with you?
Q
U
C
I
O
P YES..........1
D NO...........2>>LAE-6
M
E
E
N
By recording each piece of agricultural equipment
T
owned by the respondent, a respondent roster of
The question measures the form of reported If the respondent shares reported
large agricultural equipment is created. This
ownership of the agricultural equipment; i.e. ownership of the agricultural
information also measures reported ownership of
whether the respondent owns the piece of equipment, information on the number
agricultural equipment, by type of equipment.
equipment exclusively or jointly with one or more of joint reported owners is needed for
E01 persons.
If more than one piece of the same type of calculation of the gender wealth gap,
agricultural equipment (e.g. two tractors) are as discussed in part four of these
Because the benefits of ownership may differ if an guidelines.
owned by the respondent, ask the respondent to
individual owns the equipment alone or jointly,
E02 provide a brief description of each piece and name
countries are encouraged to collect information on
them accordingly so they can be easily
the form of reported ownership.
distinguished during the interview.

E03 The list of each piece of agricultural equipment

owned by the respondent should be provided
before proceeding to the next question.
E04 If more than one person is selected for interview it
is suggested that a roster is collected at the
household level.
E05

…
 200

LARGE AGRICULTURAL EQUIPMENT (CONTINUED)

LAE-5. LAE-6. LAE-7. LAE-8.

Is one of these joint owners your spouse or partner? Do you have the right to sell this [agricultural Is one of the persons who jointly has the Do you have the right to bequeath this
equipment]? right to sell this [agricultural equipment] [agricultural equipment]?
your spouse or partner?
YES, ALONE...............1>>LAE-8 YES, ALONE.........1>>LAE-10
YES, JOINTLY WITH ONE OR MORE YES, JOINTLY WITH ONE OR
PERSONS..................2 MORE PERSONS.......2
YES...........1 NO, SOMEONE ELSE HAS THIS YES..........1
NO...........2 NO, SOMEONE ELSE HAS THIS
NO............2 RIGHT....................3>>LAE-8 RIGHT..............3>>LAE-10
NO, IT CANNOT BE NO, IT CANNOT BE
SOLD.....................4>>LAE-8 BEQUEATHED.........4>>LAE-10
If the respondent’s spouse/partner
was identified as a joint reported or If the respondent’s spous
Identifying whether the respondent jointly documented owner of the piece of This question obtains information on identified as a joint reporte
This question obtains information on whether the respondent believes that documented owner of the
owns the agricultural equipment with his/her agricultural equipment, collecting
whether the respondent believes that she or she or he has the right to bequeath the equipment, collecting info
spouse or partner allows for the information on whether the
he has the right to sell the piece of agricultural equipment. When a whether the spouse/partn
construction of an indicator on joint spouse/partner jointly has the right to
agricultural equipment. When a respondent respondent has the right to bequeath the right to bequeath it en
(reported) ownership of agricultural sell it enables analysis of whether
has the right to sell the equipment, it means the equipment, it means that she or he analysis of whether joint o
equipment between spouses /partners, the joint owners have the same rights to
LAE-5. that she or he has the right to permanently
LAE-6. LAE-7. LAE-8.
has the right to give the equipment by the same rights to the equ
most common form of joint ownership. the equipment. If countries choose to
transfer it to another person or entity for oral or written will to another person or
Other
Is one patterns
of these jointofowners
joint ownership
your spouseare or
alsopartner? Do you collect information on all joint
cashhave
or inthe right
kind to sell this [agricultural
benefits. Is one of the persons who jointly has the Do you have
persons uponthe theright to bequeath
death of the this If countries choose to coll
possible, such as between siblings or a reported owners in LAE-4, then
equipment]? right to sell this [agricultural equipment] [agricultural
respondent.equipment]? information on all joint rep
parent and an adult child, and countries that countries can ask, in place of this
To assess gender differences in the right to your spouse “Which
or partner? in LAE
are interested in identifying these patterns question, other household
sell agricultural equipment, it is useful to
YES, ALONE...............1>>LAE-8 To assess gender differences in the place of this question, “W
are encouraged to ask, in place of LAE 4-5, members also have the right to sell YES, ALONE.........1>>LAE-10
distinguish
YES, JOINTLYbetween
WITH the
ONEtwoOR “no” answers,
MORE right to JOINTLY
bequeathWITH agricultural household members also
“Who are the joint owners, including this [agricultural equipment]?” The YES, ONE OR
identifying if the respondent is not the one
PERSONS..................2 equipment,
MORE it is useful to distinguish
PERSONS.......2 to bequeath this [agricultu
household members and non- household
YES...........1 personal IDs assigned
YES..........1 to household
who SOMEONE
NO, can sell the equipment
ELSE HAS THIS (but someone NO, SOMEONE
between the twoELSE “no” HAS THIS
answers, equipment]?” The persona
members?” membersNO...........2
in the household roster
LARGE AGRICULTURAL EQUIPMENT (CONTINUED)
NO............2 RIGHT....................3>>LAE-8
else can sell it) or that the equipment
should be recorded for each
RIGHT..............3>>LAE-10
identifying if the respondent is not the assigned to household me
NO, IT be
cannot CANNOT BE example, because no
sold (for NO, IT CANNOT BE
The personal IDs assigned to household SOLD.....................4>>LAE-8 household member who has the one who can bequeath the equipment household roster should b
markets exist for used agricultural BEQUEATHED.........4>>LAE-10
members in the household roster should be If theto
right respondent’s spouse/partner
sell the equipment. Note that (but someone else can bequeath it) or for each household memb
equipment or because of cultural or legal was identified that the equipment cannot be If
thethe respondent’s
right to bequeath spous
the a
recorded for each household member who information onas
thea number
joint reported
of non-or
Identifying whetherowner
the respondent jointly norms). documented owner ofwho
the have
piecethis
of This question(for
bequeathed obtains information
example, becauseonof identified
equipment. asNote
a joint reporte
that info
is a joint reported and each non- This question obtains information on household members
owns the agricultural equipment with the
his/her agricultural equipment, collecting whether
cultural orthe respondent
legal norms). believes that documented owner
the number of non of the
household member who jointly owns whether the respondent believes that she or right is not needed to calculate the
spouse or partner allows for the information on gap
whether the she or he has the right to bequeath the equipment,
who have this collecting info
right is not
agricultural equipment should be assigned he has the right to sell the piece of gender wealth for agricultural
construction of anmember
indicatorIDoncode
joint(e.g. spouse/partner jointly has the right to agricultural equipment. When a whether
calculatethethespouse/partn
gender weal
a non-household agricultural equipment. When a respondent equipment.
(reported) ownership of agricultural sell it enables analysis of whether respondent has the right to bequeath the right to bequeath
agricultural equipment. it en
100). has the right to sell the equipment, it means
equipment between spouses /partners, the joint owners have the same rights to the equipment, it means that she or he analysis of whether joint o
that she or he has the right to permanently has the right to give the equipment by the same rights to the equ
most common form of joint ownership. the equipment. If countries choose to
transfer it to another person or entity for oral or written will to another person or
Other patterns of joint ownership are also collect information on all joint
cash or in kind benefits. persons upon the death of the If countries choose to coll
possible, such as between siblings or a reported owners in LAE-4, then
parent and an adult child, and countries that countries can ask, in place of this respondent. information on all joint rep
To assess gender differences in the right to in LAE
are interested in identifying these patterns question, “Which other household
sell agricultural equipment, it is useful to To assess gender differences in the place of this question, “W
are encouraged to ask, in place of LAE 4-5, members also have the right to sell
distinguish between the two “no” answers, right to bequeath agricultural household members also
“Who are the joint owners, including this [agricultural equipment]?” The
identifying if the respondent is not the one equipment, it is useful to distinguish to bequeath this [agricultu
household members and non- household personal IDs assigned to household
who can sell the equipment (but someone between the two “no” answers, equipment]?” The persona
members?” members in the household roster
else can sell it) or that the equipment identifying if the respondent is not the assigned to household me
should be recorded for each
LARGE AGRICULTURAL
The personal IDs assigned to household EQUIPMENT (CONTINUED)
cannot be sold (for example, because no
household member who has the one who can bequeath the equipment household roster should b
markets exist for used agricultural (but someone else can bequeath it) or for each household memb
members in the household roster should be right to sell the equipment. Note that
equipment or because of cultural or legal that the equipment cannot be the right to bequeath the a
recorded for each household member who information on the number of non-
LAE-9. norms). equipment. Note that info
is a joint reported owner and each non- LAE-10. LAE-11. members who have this
household bequeathed (for example, because of
household member who jointly owns the right is not needed to calculate the cultural or legal norms). the number of non
Is one of the persons
agricultural equipment whoshould
jointlybehas
assigned How did you acquire this [agricultural equipment]? If this [agricultural
gender wealth gap for equipment]
agriculturalwere who have this right is not
the right to bequeathmember
a non-household this [agricultural
ID code (e.g. to be sold today, how much could
equipment. calculate the gender weal
equipment]
100). your spouse or partner? be received for it? agricultural equipment.
PURCHASED...................................1
INHERITED...................................2
RECEIVED AS A GIFT..........................3
ALLOCATED BY GOVERNMENT PROGRAMME...........4
THROUGH MARRIAGE............................5
YES..........1 OTHER(SPECIFY).............................96
NO...........2
This question measures the value
If the respondent’s spouse/partner was This question measures the mode of acquiring the of the agricultural equipment.
identified as a joint reported or agricultural equipment. It refers to when the respondent Respondents should estimate the
documented owner of the agricultural first came into possession of the asset and presumably current value based on the
ht to bequeath the equipment, collecting information on began deriving economic benefits from it. location and quality of their
whether the spouse/partner jointly has particular equipment. The full
the right to bequeath it enables Because women and men often acquire assets through amount that would be received in
he analysis of whether joint owners have different means, understanding the modes of acquisition the sale should be listed,
LAE-9.
the same rights to the equipment. LAE-10.
may provide insights for developing policies to ensure LAE-11.
regardless whether or not all of it
or Is one of the persons who jointly has women’s ability to acquire them. As such, NSOs should would be kept by the respondent.
How did you acquire this [agricultural equipment]? If this [agricultural equipment] were
If countries choose to collect include all relevant modes of acquisition.
the right to bequeath this [agricultural to
If be
thesold today, how
respondent much
is not surecould
how
information on all joint reported owners
equipment] your spouse or partner?
in LAE-4, then countries can ask, in be
to received for it?
answer, enumerators should
PURCHASED...................................1 probe on this question by
place of this question, “Which other INHERITED...................................2
household members also have the right RECEIVED AS A GIFT..........................3 encouraging the respondent to
to bequeath this [agricultural ALLOCATED BY GOVERNMENT PROGRAMME...........4
consider the price received for
equipment]?” The personal IDs THROUGH MARRIAGE............................5 other pieces of agricultural
YES..........1
assigned to household members in the OTHER(SPECIFY).............................96 equipment that have been sold in
NO...........2 the area.
household roster should be recorded This question measures the value
for each household member who has of the agricultural equipment.
If the respondent’s spouse/partner was This question measures the mode of acquiring the
the right to bequeath the agricultural Respondents should estimate the
identified as a joint reported or agricultural equipment. It refers to when the respondent
equipment. Note that information on current value based on the
documented first came into possession of the asset and presumably
the number ofowner of the agricultural
non-household members location and quality of their
ht to bequeath the equipment, collecting information on began deriving economic benefits from it.
who have this right is not needed to particular equipment. The full
whether the spouse/partner jointly
calculate the gender wealth gap for has
the right to bequeath it enables Because women and men often acquire assets through amount that would be received in
agricultural equipment. the sale should be listed,
he analysis of whether joint owners have different means, understanding the modes of acquisition
the same rights to the equipment. may provide insights for developing policies to ensure regardless whether or not all of it
or women’s ability to acquire them. As such, NSOs should would be kept by the respondent.
If countries choose to collect include all relevant modes of acquisition.
information on all joint reported owners If the respondent is not sure how
in LAE-4, then countries can ask, in to answer, enumerators should
place of this question, “Which other probe on this question by
household members also have the right encouraging the respondent to
to bequeath this [agricultural consider the price received for
equipment]?” The personal IDs other pieces of agricultural
assigned to household members in the equipment that have been sold in
the area.
 Model questionnaire 201

SMALL AGRICULTURAL EQUIPMENT

SAE-1. SAE-2.

E E Do you own any [small agricultural SELECT THE METHOD WHICH WAS USED TO
Q Q equipment], exclusively or jointly? INTERVIEW THE RESPONDENT.
U U
I I YES, EXCLUSIVELY............1
P P YES, JOINTLY................2
YES, BOTH EXCLUSIVELY RESPONSE CODES:
M M AND JOINTLY...............3
E E NO..........................4 ALONE............................1
N N >>NEXT CATEGORY WITH ADULT FEMALES PRESENT.......2
T T WITH ADULT MALES PRESENT.........3
WITH ADULTS MIXED SEX PRESENT....4
This question measures reported WITH CHILDREN PRESENT............5
C N ownership of small agricultural WITH ADULTS MIXED SEX
O A equipment and also the form of AND CHILDREN PRESENT........6
D M ownership; i.e. whether the
E E equipment is owned exclusively or
jointly by the respondent. Categories
may include but not be limited to
axes, hoes, spades, watering cans, (Reason(s) interview not administered with the
S01 HOE rakes, wheelbarrows, pruners,
respondent(s) alone should be explained in the
S02 AXE weeders, etc. Countries will need to
determine the categories of small survey officer remarks section).
S03 MACHETE agricultural equipment to include
S04 SLASHER based on prevalence rates from
existing agricultural or household
S05 WATERING CAN/ PAIL surveys and also on policy needs.
S06 WHEELBARROW
S07 PRUNING KNIFE
S08 PRUNING SAW
S09 CHAIN/ HANDSAW
S10 SHELLER
S11 SPADE
S12 FORK HOE/ GARDEN FORK
S13 SPRAYER
S14 MILK CAN
S15 LANTERN/ TORCH
S16 CULTIVATOR
S17 RAKE
S18 WEEDER
S19 PLANTER
S20 OTHER (SPECIFY)
202

NON-AGRICULTURAL ENTERPRISE ASSETS

NAE-1. In the last week (Monday to Sunday) did you run or do any kind of business, big or YES...........1 >>NAE-3
small, for yourself or with one or more partners, even if it was for only one hour? NO............2

This is the first of three screening questions in this module to determine whether the respondent owns a business. Countries should
assess in advance through focus group discussions or testing of the questionnaire whether this question needs to be rephrased to capture
smaller economic units that respondents may not identify as businesses, but that nonetheless should be measured in this modul e. For
example, in Karnataka, India under the Gender Asset Gap Project, businesses were referred to as “economic activities” rather than
“businesses” when it became clear during field testing that respondents did not consider small informal business activities t o be
businesses. Alternatively, rather than asking if the respondent owned a business, the EDGE pilots in Georgia, Mongolia, the Philippines
and Uganda asked a detailed list of seven screening questions to capture the businesses owned by the respondent. See more inf ormation
on EDGE project website

NAE-2. Do you have a business to which you will definitely return?

YES...........1
NO............2>>NEXT MODULE

This question aims to capture business owners who may have been absent from the business owing to leave, illness or any other
commitments during the 7-day reference period in NAE-1, but who will definitely return to it. If the respondent answered “no” to NAE1 and
NAE2, skip to the next module as the remaining questions in this module are only asked if the respondent reports owning an
enterprise/business.

NAE-3. Is your business in?

FARMING..........................1
FORESTRY.........................2
RAISING ANIMALS OR FISHING.......3
A SECTOR OTHER THAN AGRICULTURE..4

The question establishes the sector in which the business operates, distinguishing between agricultural activities and non -agricultural
activities. The production and sale of non-processed agricultural goods (such as milk, wool, fruits, and vegetables) produced on own farm
is an agricultural enterprise while the sale or trade of agricultural products purchased from non -household members and the sale of by-
products of agricultural goods (such as cheese, beer, jam, or sweaters) are considered non -agricultural (manufacturing) enterprises. If the
enterprise owned by the respondent is engaged in the agricultural sector, skip to the next module as the remaining questions in this
module are only asked if the respondent owns a non-agricultural enterprise. If the respondent owns more than one enterprise, at least one
enterprise must be in the non-agricultural sector to proceed with the module. This question also measures reported ownership of non-
agricultural enterprises.

NAE-4. NAE-5. NAE-6.

Please describe the kind of activity in Is this business a…? What type of records or accounts does this
which each enterprise is engaged. [ENTERPRISE] maintain?
E
N LIMITED LIABILITY ENTERPRISE.........1 NO WRITTEN ACCOUNT KEPT..............1
T NON-LIMITED LIABILITY ENTERPRISE.....2 INFORMAL RECORDS FOR PERSONAL USE....2
SIMPLIFIED ACCOUNTING FORMAT
E REQUIRED FOR TAX PAYMENT..........3
R DETAILED FORMAL ACCOUNT
P (BALANCE SHEET
The enumerator should list AND INCOME STATEMENTS)............4
R each non-agricultural
I enterprise described by the
Questions NAE-5 and NAE-6 measure whether the enterprise is incorporated (i.e. the production
S respondent to create a
unit is a separate legal entity from its owners) or unincorporated. Consistent with the 2008 SNA
respondent roster of non-
E and the OECD Guidelines for Micro Statistics on Household Wealth, the assets owned by
agricultural enterprises. The
incorporated enterprises cannot be owned by the respondent and thus are excluded from the
list of each non-agricultural
measurement of wealth at the individual or household level in these guidelines. If the enterprise is
I enterprise owned by the
a limited liability enterprise and keeps formal accounts, it is an incorporated enterprise and the
D respondent should be
enumerator should skip to the next enterprise or the next module.
provided before proceeding
to the next question.

B01

B02

B03

B04

…
 Model questionnaire 203

NON-AGRICULTURAL ENTERPRISE ASSETS (CONTINUED)

NAE-7. NAE-8. NAE-9. NAE-10.
Do you currently own any of the following How many persons jointly own [category of Is one of these joint owners your spouse/partner? What is the total value that would be received
[category of enterprise assets] that are used by enterprise asset]? for this [category of enterprise asset] if it were
the [enterprise]? sold today?
YES, ALONE......1 >> NAE-10 YES.....................1
YES, JOINTLY....2 NO......................2
NO..............3 >> NEXT ENTERPRISE
(if more than 1 owned
by respondent)/NEXT MODULE

If the respondent jointly owns [category Identifying whether the respondent’s For each category of enterprise
The categories of enterprise assets are: (a)
of enterprise assets], information on the spouse/partner is a joint owner of the assets that the respondent reported
the current stock of physical capital,
number of joint owners is needed for category of enterprise assets allows for owing in NAE-7, she or he should
including all machinery, equipment, and
calculation of the gender wealth gap, as the construction of an indicator on joint estimate in local currency how much
furniture used for the business that were
discussed in part four of these ownership of enterprise assets between would be received in total if the all of
not listed earlier in any of the other
guidelines. spouses/partners, the most common the assets in that category (e.g. all
modules; (b) the current stock of inputs or
form of joint ownership. Other patterns finished merchandise) were sold
supplies, including raw materials; and (c)
Note that this question should be asked of joint documented ownership are also today. Include codes for “does not
the current stock of finished merchandise
for each category of enterprise assets possible, such as between siblings or a know” and “refuses to answer.”
(goods for sale). If the enterprise does not
own assets in any of these three that the respondent reports owning in parent and an adult child, and countries
NAE-7. NAE-8 to NAE-10 should be that are interested in identifying these As discussed above, because the
categories, skip to the next enterprise if the
asked for each category before patterns are encouraged to ask, in place SNA does not consider enterprises
respondent reported owning more than one
proceeding to the next category. of NAE-8 to NAE-9, “Who else is listed to be assets, the module does not
enterprise or to the next module.
as an owner on the ownership include a question on valuing
document, including household unincorporated enterprises in
Note that only those assets that were not
members and non-household addition to valuing the assets held
listed in the previous modules should be
members?” The personal IDs assigned by the enterprise.
included here to avoid the double-counting
of assets. Any motor vehicles used for the to household members in the household
enterprise should be listed and valued in roster should be recorded for each
the module on consumer durables. Any household member who is a joint
land and buildings used for enterprises documented owner, and each non-
should be listed and valued in the module household member who jointly owns the
on other real estate. category of enterprise assets should be
assigned a non-household member ID
code (e.g. 100).

Finished Finished Finished

Physical capital Inputs or supplies Physical capital Inputs or supplies Physical capital Inputs or supplies
merchandise merchandise merchandise
204

OTHER REAL ESTATE

ORE-1. Do you own any other real estate, including other residential dwellings or buildings, commercial buildings or non-agricultural plots
of land? YES...........1
NO............2>>NEXT MODULE

This is the screening question to determine whether the respondent should complete the module. It measures the reported ownership
of any other real estate (not by type of real estate). Categories of real estate may include residential dwellings or buildings,
commercial buildings or non-agricultural plots of land. Countries will need to determine the categories of other real estate to include
based on prevalence rates from prior household surveys and also policy needs and may want to further disaggregate the suggested
categories based on analytical needs. If the respondent does not own any other real estate, skip to the next module as the remainder
of this module is only administered to the respondent if she/he self-reports owning any other real estate.

ORE-2. ORE-3. ORE-4. ORE-5.

Please list each piece of other real estate that Is this [other real Does anyone jointly own this How many other persons jointly
you own, exclusively or jointly with someone estate] located inside [other real estate] with you, own this [other real estate] with
else. or outside the country? including household you?
members and non-
By recording each piece of other real estate household members?
INSIDE.......1
owned by the respondent, a respondent OUTSIDE......2
YES........1
roster of other real estate is created. This
NO.........2>>ORE-7
information also measures reported
ownership of other real estate, by type of
The question measures
real estate.
Note that this the form of reported
question is only ownership of the other If the respondent shares
If more than one piece of the same type of
necessary if real estate; i.e. whether reported ownership of the
other real estate is owned by the
countries wish to the respondent owns the real estate, information on
respondent (e.g. two non-agricultural plots
use the data real estate exclusively or the number of joint reported
of land), ask the respondent to provide a
collected on other jointly with one or more owners is needed for
brief description of each piece and name
real estate for persons. Because the calculation of the gender
them accordingly so they can be easily
updating their benefits of ownership wealth gap, as discussed in
distinguished during the interview. The list
national accounts may differ if individuals part four of these
of all pieces of other real estate owned by
as real estate own the real estate guidelines.
the respondent should be provided before
located outside of alone or jointly,
proceeding to the next question.
the country is not countries are
included in the encouraged to collect
If more than one person is selected for
2008 SNA. information on the form
CODE interview it is suggested that a roster is
of reported ownership.
collected at the household level. If this is the
case, characteristics (ORE-3) of the asset
R01 should also be collected at household level.

R02

R03
 Model questionnaire 205

OTHER REAL ESTATE

ORE-1. Do you own any other real estate, including other residential dwellings or buildings, commercial buildings or non-agricultural plots
of land? YES...........1
NO............2>>NEXT MODULE

This is the screening question to determine whether the respondent should complete the module. It measures the reported ownership
of any other real estate (not by type of real estate). Categories of real estate may include residential dwellings or buildings,
commercial buildings or non-agricultural plots of land. Countries will need to determine the categories of other real estate to include
based on prevalence rates from prior household surveys and also policy needs and may want to further disaggregate the suggested
categories based on analytical needs. If the respondent does not own any other real estate, skip to the next module as the remainder
of this module is only administered to the respondent if she/he self-reports owning any other real estate.

ORE-2. ORE-3. ORE-4. ORE-5.

Please list each piece of other real estate that Is this [other real Does anyone jointly own this How many other persons jointly
you own, exclusively or jointly with someone estate] located inside [other real estate] with you, own this [other real estate] with
else. or outside the country? including household you?
members and non-
By recording each piece of other real estate household members?
INSIDE.......1
owned by the respondent, a respondent OUTSIDE......2
YES........1
roster of other real estate is created. This
NO.........2>>ORE-7
information also measures reported
ownership of other real estate, by type of
The question measures
real estate.
Note that this the form of reported
question is only ownership of the other If the respondent shares
If more than one piece of the same type of
necessary if real estate; i.e. whether reported ownership of the
other real estate is owned by the
countries wish to the respondent owns the real estate, information on
respondent (e.g. two non-agricultural plots
use the data real estate exclusively or the number of joint reported
of land), ask the respondent to provide a
collected on other jointly with one or more owners is needed for
brief description of each piece and name
real estate for persons. Because the calculation of the gender
them accordingly so they can be easily
updating their benefits of ownership wealth gap, as discussed in
distinguished during the interview. The list
national accounts may differ if individuals part four of these
of all pieces of other real estate owned by
own the real estate guidelines.
OTHER REAL ESTATE
the respondent (CONTINUED)
should be provided before as real estate
alone or jointly,
proceeding to the next question. located outside of
the country is not countries are
included in the encouraged to collect
If more than one person is selected for
2008 SNA. information on the form
CODE interview it is suggested that a roster is
of reported ownership.
collected at the household level. If this is the
case, characteristics (ORE-3) of the asset
…

OTHER REAL ESTATE (CONTINUED)

ORE-6. ORE-7. ORE-8.

Is one of these joint owners your spouse or Is there an ownership document for this [other real Are you listed as an owner on the ownership document
partner? estate]? for this [other real estate]?

YES, A TITLE DEED............1 YES,ALONE...................1>>ORE-11

YES.....1 YES, A CERTIFICATE OF CUSTOMARY YES, JOINTLY WITH ONE OR MORE
NO......2 OWNERSHIP....................2 PERSONS.....................2
YES, A CERTIFICATE OF NO..........................3>>ORE-11
OCCUPANCY....................3
YES, A WILL..................4
YES, A PURCHASE AGREEMENT....5 This question measures documented ownership of
Identifying whether the respondent
YES, OTHER (SPECIFY).........6 other real estate. Documented ownership refers to
jointly owns the real estate with her/his
ORE-6. ORE-7.
NO...................7>>ORE-11 the existence of any document that an individual
ORE-8.
spouse or partner allows for the
can use to claim ownership rights in law over the
Is one of these joint
construction of anowners
indicatoryour
on spouse
joint or Is there an ownership document for this [other real Are you listed as an owner on the ownership document
real estate by virtue of the individual’s name being
(reported) ownership other real estate
partner? estate]? forlisted
this [other
This question identifies whether there is an as anrealowner estate]?
on the document. Because
between spouses/partners, the most ownership document for the other real estate and
YES, type
A TITLE DEED............1 individual names can be listed as witnesses on an
YES,ALONE...................1>>ORE-11
common form of joint ownership. Other what of document this is. There may be a
YES.....1 YES, A CERTIFICATE OF CUSTOMARY ownership
YES, JOINTLY document, WITHitONE
is important
OR MOREto ask if the
patterns of joint ownership are also
NO......2 range of types of documents that provide formal PERSONS.....................2
OWNERSHIP....................2 respondent is listed “as an owner” on the
possible, such as between siblings or a evidence of ownership,OF and national statistical NO..........................3>>ORE-11
YES, A CERTIFICATE document. As discussed above, while countries
parent and an adult child, and countries agencies will need to customize the response
OCCUPANCY....................3 may want to ask the respondent to produce the
that are interested in identifying these categories according to their country context.
YES, A WILL..................4 document
This questionfor the enumerator
measures so that the
documented ownership of
patterns
Identifyingarewhether
encouraged to ask, in place
the respondent Titles
YES, A and deeds are
PURCHASE one form of ownership
AGREEMENT....5
YES, OTHER (SPECIFY).........6
enumerator
other real estate.can confirm that theownership
Documented respondent’srefers to
of ORE-5
jointly ownsandtheORE-6, “Whowith
real estate areher/his
the document. Registration certificates document
NO...................7>>ORE-11 name
the existence of any document that an guidelines
is listed on the document, these individual
joint
spouseowners, including
or partner household
allows for the rights over property. In addition, where titling or recommend
can use to claimthat the measure
ownership of documented
rights in law over the
members
constructionandofnon-household
an indicator on joint registration is not complete, documents including ownership
real estate not be conditional
by virtue on the document
of the individual’s name being
members?” The personal
(reported) ownership otherIDs
realassigned
estate wills or
This sales receipts
question identifiesprovide
whether some
thereform
is an of
CODE being as
listed checked.
an owner on the document. Because
to household
between members in the
spouses/partners, thehousehold
most documented
ownership claim. for the other real estate and
document
roster should individual names can be listed as witnesses on an
common formbe of recorded for each
joint ownership. Other If an ownership
what type of document document thisexists for the
is. There may other
be areal The question also measures the form
household ownership document, it is important toofask if the
R01 patterns of member who is are
joint ownership a joint
also estate,ofit types
range shouldofbe recorded regardless
documents that providewhether
formal documentedisownership of the otheronreal
reported owner and each non- respondent listed “as an owner” theestate; i.e.
possible, such as between siblings or a or not it has
evidence the name ofand
of ownership, someone
nationalinstatistical
the whether the respondent owns the real estate
household document. As discussed above, while countries
parent and member
an adult who
child,jointly owns the
and countries householdwill
agencies onneed
it. If there is more than
to customize one type of
the response exclusively or jointly with one or more persons.
R02 other real estate should be assigned may want to ask the respondent to produce the
that are interested in identifying thesea document, according
categories the one that to is held
their by someone
country context. in the Because
non-household member ID documentthe forbenefits of ownership
the enumerator so thatmay
thediffer if
patterns are encouraged to code
ask, in (e.g.
place household
Titles and deedsshouldare be one
recorded.
form ofFor example, if
ownership individuals own
100). enumerator can the real estate
confirm that thealone or jointly,
respondent’s
R03 of ORE-5 and ORE-6, “Who are the there is a deed, but the household
document. Registration certificates document member does countries are encouraged to collect information on
name is listed on the document, these guidelines
joint owners, including household not have
rights over it, property.
but has an In invoice
addition, orwhere
sales receipt,
titling orlist the form of documented ownership.
recommend that the measure of documented
members and non-household the invoice, is
registration notnotthe deed. documents including
complete, ownership not be conditional on the document
members?” The personal IDs assigned wills or sales receipts provide some form of
CODE being checked.
to household members in the household documented claim.
roster should be recorded for each If an ownership document exists for the other real The question also measures the form of
R01 household member who is a joint estate, it should be recorded regardless whether documented ownership of the other real estate; i.e.
reported owner and each non- or not it has the name of someone in the whether the respondent owns the real estate
household member who jointly owns the household on it. If there is more than one type of exclusively or jointly with one or more persons.
R02 other real estate should be assigned a document, the one that is held by someone in the Because the benefits of ownership may differ if
non-household member ID code (e.g. household should be recorded. For example, if individuals own the real estate alone or jointly,
100). there is a deed, but the household member does
206

OTHER REAL ESTATE (CONTINUED)

OTHER REAL ESTATE (CONTINUED)

ORE-9. ORE-10.
How many other people are listed as owners Is one of these joint owners your spouse or partner?
on the ownership document, including
household members and non-household
members? YES.....1
NO......2

Identifying whether both the respondent’s name and the name of

the respondent's spouse/partner are listed as owners on the
ORE-9. ORE-10. document allows for the construction of an indicator on
ownership
If the
How manyrespondent sharesare
other people documented
listed as owners joint
Is one(documented) ownership
of these joint between
owners your spouses/partners,
spouse or partner? the most
ownership of the real estate, information common form of joint ownership. Other patterns of joint
ononthethe
ownership
number ofdocument, including
joint documented documented ownership are also possible, such as between siblings
household
owners ismembers
needed forand non-household
calculation of the or a parent and an adult child, and countries that are interested in
members?
gender wealth gap, as discussed in part identifyingYES.....1
these patterns are encouraged to ask, in place of ORE-9
four of these guidelines. and ORE-10,NO......2
“Who else is listed as an owner on the ownership
document, including household members and non-household
members?” The personal IDs assigned to household members in
the household roster should be recorded for each household
member who is a joint documented owner, and each non-
Identifying
household whether
member whoboth theowns
jointly respondent’s name
the other real andshould
estate the name of
be the respondent's
assigned spouse/partner
a non-household memberare
ID listed as owners
code (e.g. 100). on the
CODE ownership document allows for the construction of an indicator on
If the respondent shares documented joint (documented) ownership between spouses/partners, the most
ownership of the real estate, information common form of joint ownership. Other patterns of joint
R01 on the number of joint documented documented ownership are also possible, such as between siblings
owners is needed for calculation of the or a parent and an adult child, and countries that are interested in
gender wealth gap, as discussed in part identifying these patterns are encouraged to ask, in place of ORE-9
R02
four of these guidelines. and ORE-10, “Who else is listed as an owner on the ownership
document, including household members and non-household
R03 members?” The personal IDs assigned to household members in
the household roster should be recorded for each household
member who is a joint documented owner, and each non-
household member who jointly owns the other real estate should
be assigned a non-household member ID code (e.g. 100).
CODE

R01

R02

R03
 OTHER REAL ESTATE (CONTINUED)

Model questionnaire 207

OTHER REAL ESTATE (CONTINUED)

ORE-9. ORE-10.
How many other people are listed as owners Is one of these joint owners your spouse or partner?
on the ownership document, including
household members and non-household
members? YES.....1
NO......2

Identifying whether both the respondent’s name and the name of

ORE-9. ORE-10.
the respondent's spouse/partner are listed as owners on the
How many other people are listed as owners Is one of thesedocument
ownership joint owners
allowsyourforspouse or partner?of an indicator on
the construction
If the
on the respondent
ownership shares documented
document, including joint (documented) ownership between spouses/partners, the most
ownership of the real estate, information common form of joint ownership. Other patterns of joint
household members and non-household
on the number of joint documented documented ownership are also possible, such as between siblings
members?
owners is needed for calculation of the YES.....1
or a parent and an adult child, and countries that are interested in
gender wealth gap, as discussed in part NO......2
identifying these patterns are encouraged to ask, in place of ORE-9
four of these guidelines. and ORE-10, “Who else is listed as an owner on the ownership
document, including household members and non-household
members?” The personal IDs assigned to household members in
Identifying whether
the household rosterboth the respondent’s
should be recorded for name and
each the name of
household
the respondent's
member who is aspouse/partner
joint documented areowner,
listed as owners
and on the
each non-
ownership
household document
member who allows for owns
jointly the construction of an
the other real indicator
estate on
should
If the respondent shares documented joint (documented)
be assigned ownership between
a non-household member spouses/partners,
ID code (e.g. 100).the most
CODE ownership of the real estate, information common form of joint ownership. Other patterns of joint
on the number of joint documented documented ownership are also possible, such as between siblings
owners is needed for calculation of the or a parent and an adult child, and countries that are interested in
… gender wealth gap, as discussed in part identifying these patterns are encouraged to ask, in place of ORE-9
four of these guidelines. and ORE-10, “Who else is listed as an owner on the ownership
document, including household members and non-household
members?” The personal IDs assigned to household members in
the household roster should be recorded for each household
member who is a joint documented owner, and each non-
household member who jointly owns the other real estate should
be assigned a non-household member ID code (e.g. 100).
CODE

OTHER REAL ESTATE (CONTINUED)

…
ORE-11. ORE-12. ORE-13. ORE-14.
Do you have the right to sell this [other real Is one of the persons who jointly has the Do you have the right to bequeath this Is one of the persons who jointly has the right
estate]? right to sell this [other real estate] your [other real estate]? to bequeath this [other real estate] your spouse
spouse or partner? or partner?
YES, ALONE..............1>>ORE-13 YES, ALONE...........1>>ORE-15
YES, WITH ONE OR MORE YES.......1 YES, JOINTLY WITH ONE OR MORE YES.......1
PERSONS.................2 NO........2 PERSONS..............2 NO........2
NO, SOMEONE ELSE HAS THIS NO, SOMEONE ELSE HAS THIS
RIGHT...................3>>ORE-13 If the respondent’s spouse/partner RIGHT................3>>ORE-15
was identified as a joint reported or NO, IT CANNOT BE If the respondent’s spouse/partner was
documented owner of the real BEQUEATHED...........4>>ORE-15
identified as a joint reported or documented
estate, collecting information on owner of the other real estate, collecting
This question obtains information on whether the spouse/partner jointly information on whether the spouse/partner
whether the respondent believes that he has the right to sell the real estate This question obtains information on jointly has the right to bequeath the real
or she has the right to sell the real enables analysis of whether joint whether the respondent believes that estate enables analysis of whether joint
estate. When a respondent has the right owners have the same rights to the she or he has the right to bequeath owners have the same rights to the real
to sell the real estate, it means that he or real estate. If countries choose to the real estate. When a respondent estate. If countries choose to collect
she has the right to permanently transfer collect information on all joint has the right to bequeath the real information on all joint reported and
it to another person or entity for cash or reported and documented owners in estate, it means that she or he has documented owners in ORE-5 and ORE-9,
in kind benefits. To assess gender ORE-5 and ORE-9, respectively, the right to give the real estate by oral respectively, then countries can ask, in
differences in the right to sell the real then countries can ask, in place of or written will to another person or place of this question, “Which other
estate, it is useful to distinguish between this question, “Which other persons upon the death of the household members also have the right to
the two “no” answers, identifying if the household members also have the respondent. To assess gender bequeath this [other real estate]?” The
respondent is not the one who can sell right to sell this [other real estate]?” differences in the right to bequeath personal IDs assigned to household
the real estate (but someone else can The personal IDs assigned to the real estate, it is useful to members in the household roster should be
sell it) or that the real estate cannot be household members in the distinguish between the two “no” recorded for each household member who
sold (for example, because of cultural or household roster should be answers, identifying if the respondent has the right to bequeath the real estate.
legal norms). recorded for each household is not the one who can bequeath the Note that information on the number of
member who has the right to sell the real estate (but someone else can non-household members who have this
real estate. Note that information on bequeath it) or that the real estate right is not needed to calculate the gender
CODE
the number of persons who have cannot be bequeathed (for example, wealth gap for other real estate.
this right is not needed to calculate because of cultural or legal norms).
the gender wealth gap for other real
R01 estate.

R02

R03

R04

R05
208

OTHER REAL ESTATE (CONTINUED)

ORE-11. ORE-12. ORE-13. ORE-14.

Do you have the right to sell this [other real Is one of the persons who jointly has the Do you have the right to bequeath this Is one of the persons who jointly has the right
estate]? right to sell this [other real estate] your [other real estate]? to bequeath this [other real estate] your spouse
spouse or partner? or partner?
YES, ALONE..............1>>ORE-13 YES, ALONE...........1>>ORE-15
YES, WITH ONE OR MORE YES.......1 YES, JOINTLY WITH ONE OR MORE YES.......1
PERSONS.................2 NO........2 PERSONS..............2 NO........2
NO, SOMEONE ELSE HAS THIS NO, SOMEONE ELSE HAS THIS
RIGHT...................3>>ORE-13 If the respondent’s spouse/partner RIGHT................3>>ORE-15
was identified as a joint reported or NO, IT CANNOT BE If the respondent’s spouse/partner was
documented owner of the real BEQUEATHED...........4>>ORE-15
identified as a joint reported or documented
estate, collecting information on owner of the other real estate, collecting
This question obtains information on whether the spouse/partner jointly information on whether the spouse/partner
whether the respondent believes that he has the right to sell the real estate This question obtains information on jointly has the right to bequeath the real
or she has the right to sell the real enables analysis of whether joint whether the respondent believes that estate enables analysis of whether joint
estate. When a respondent has the right owners have the same rights to the she or he has the right to bequeath owners have the same rights to the real
to sell the real estate, it means that he or real estate. If countries choose to the real estate. When a respondent estate. If countries choose to collect
she has the right to permanently transfer collect information on all joint has the right to bequeath the real information on all joint reported and
it to another person or entity for cash or reported and documented owners in estate, it means that she or he has documented owners in ORE-5 and ORE-9,
in kind benefits. To assess gender ORE-5 and ORE-9, respectively, the right to give the real estate by oral respectively, then countries can ask, in
differences in the right to sell the real then countries can ask, in place of or written will to another person or place of this question, “Which other
estate, it is useful to distinguish between this question, “Which other persons upon the death of the household members also have the right to
the two “no” answers, identifying if the household members also have the respondent. To assess gender bequeath this [other real estate]?” The
respondent is not the one who can sell right to sell this [other real estate]?” differences in the right to bequeath personal IDs assigned to household
the real estate (but someone else can The personal IDs assigned to the real estate, it is useful to members in the household roster should be
sell it) or that the real estate cannot be household members in the distinguish between the two “no” recorded for each household member who
sold (for example, because of cultural or household roster should be answers, identifying if the respondent has the right to bequeath the real estate.
legal norms). recorded for each household is not the one who can bequeath the Note that information on the number of
member who has the right to sell the real estate (but someone else can non-household members who have this
real estate. Note that information on bequeath it) or that the real estate right is not needed to calculate the gender
CODE
the number of persons who have cannot be bequeathed (for example, wealth gap for other real estate.
this right is not needed to calculate because of cultural or legal norms).
the gender wealth gap for other real
…

OTHER REAL ESTATE (CONTINUED)

ORE-15. ORE-16.
How did you acquire this [other real estate]? If this [other real estate] were to be sold today, how much could be
received for it?

PURCHASED.........................................1
INHERITED.........................................2
RECEIVED AS A GIFT................................3
ALLOCATED BY GOVERNMENT PROGRAMME.................4
THROUGH MARRIAGE..................................5
OTHER(SPECIFY)...................................96
This question measures the value of the other real estate.
Respondents should estimate the current value based on the
location and quality of their particular real estate. The full amount
This question measures the mode of acquiring the other real that would be received in the sale should be listed, regardless
estate. It refers to when the respondent first came into possession whether or not all of it would be kept by the respondent. If the
of the asset and presumably began deriving economic benefits respondent is not sure how to answer, enumerators should probe
from it. Because women and men often acquire assets through on this question by encouraging the respondent to consider the
different means, understanding the modes of acquisition may price received for other pieces of real estate of the same type
provide insights for developing policies to ensure women’s ability that have been sold in the area.
to acquire them. As such, NSOs should include all relevant modes
of acquisition and may want to add additional codes for when the
real estate is received as an inheritance or as a gift to indicate who
gave the inheritance or gift. This is particularly useful for gender
analyses, since the information collected can indicate whether the
real estate was received from the husband’s family or the wife’s
family.

CODE

R01

R02

R03

R04

R05
 Model questionnaire 209

OTHER REAL ESTATE (CONTINUED)

ORE-15. ORE-16.
How did you acquire this [other real estate]? If this [other real estate] were to be sold today, how much could be
received for it?

PURCHASED.........................................1
INHERITED.........................................2
RECEIVED AS A GIFT................................3
ALLOCATED BY GOVERNMENT PROGRAMME.................4
THROUGH MARRIAGE..................................5
OTHER(SPECIFY)...................................96
This question measures the value of the other real estate.
Respondents should estimate the current value based on the
location and quality of their particular real estate. The full amount
This question measures the mode of acquiring the other real that would be received in the sale should be listed, regardless
estate. It refers to when the respondent first came into possession whether or not all of it would be kept by the respondent. If the
of the asset and presumably began deriving economic benefits respondent is not sure how to answer, enumerators should probe
from it. Because women and men often acquire assets through on this question by encouraging the respondent to consider the
different means, understanding the modes of acquisition may price received for other pieces of real estate of the same type
provide insights for developing policies to ensure women’s ability that have been sold in the area.
to acquire them. As such, NSOs should include all relevant modes
of acquisition and may want to add additional codes for when the
real estate is received as an inheritance or as a gift to indicate who
gave the inheritance or gift. This is particularly useful for gender
analyses, since the information collected can indicate whether the
real estate was received from the husband’s family or the wife’s
family.

CODE

…
 CONSUMER DURABLES

210
CD-1. CD-2. CD-3.
D D
U U Do you own any [consumer durable category], If this [category of consumer durable] were to SELECT THE METHOD WHICH WAS USED TO
R R exclusively or jointly? be sold today, how much could be received for INTERVIEW THE RESPONDENT.
A A it?
B B
L L YES, EXCLUSIVELY...................1
E E YES, JOINTLY.......................2
YES, BOTH EXCLUSIVELY AND JOINTLY..3 RESPONSE CODES:
NO....................4>>NEXT MODULE
C N
ALONE............................1
O A WITH ADULT FEMALES PRESENT.......2
This question measures reported ownership of These guidelines recommend collecting
D M consumer durables and also the form of WITH ADULT MALES PRESENT.........3
valuation data for all motor vehicles and WITH ADULTS MIXED SEX PRESENT....4
E E ownership; i.e. whether the durable is owned other high-value consumer durables. WITH CHILDREN PRESENT............5
exclusively or jointly by the respondent. Respondents should estimate the WITH ADULTS MIXED SEX
D01 Bed Categories may include but not be limited to current value based on the quality of AND CHILDREN PRESENT........6
cars, motorcycles, bicycles, computers and their particular durable. The full amount
D02 Charcoal stove laptops, cell phones, and radios. Countries will that would be received in the sale should
need to determine the categories of consumer be listed, regardless whether or not all of
D03 Electric stove
durables to include based on prevalence rates it would be kept by the respondent. If
D04 Electric iron from existing data sources and also on policy the respondent is not sure how to (Reason(s) interview not administered with the
needs. In general, however, countries should answer, enumerators should probe on respondent(s) alone should be explained in the
D05 Microwave include durables of high value, such as motor this question by encouraging the survey officer remarks section).
vehicles, together with those durables that are of respondent to consider the price
D06 Electric kettle lower value but that may be of particular received for other durables of the same
D07 Refrigerator importance to women, such as cell phones. Also type that have been sold in the area.
include a category of “other (specify)" to create
D08 Radio an exhaustive module on consumer durables.
Because some durables may be owned
D09 Television collectively by all household members, while
others are owned individually, countries can
D10 Cell phone
exclude those durables that are reported as
D11 Computer or laptop being collectively owned during focus group
discussions or pilot testing of the questionnaire.
D12 Bicycle

D13 Motorcycle

D14 Car

D15 Pickup

D16 Built-in kichen sink

D17 Washing machine

D18 Home security system

household member who jointly owns the right is not needed to calculate the cultural or legal norms). the number of non
D19 Home theatre system equipment should be assigned
agricultural gender wealth gap for agricultural who have this right is not needed to
a non-household member ID code (e.g. equipment. calculate the gender wealth gap for
D20 Other (specify)
100). agricultural equipment.
 A suggested approach to value valuables is to ask the respondent for the potential sales value of each category of valuables. In addition, for the purpose of
VALUABLES calculating individual wealth, respondent owners should be asked to estimate their own share of the sales value from each group of valuables.

V-1. V-2. V-3.

Do you own any [valuable category], exclusively or If this [category of valuables] were to be sold SELECT THE METHOD WHICH WAS USED TO
V V jointly? today, how much could be received for it? INTERVIEW THE RESPONDENT.
A A
L C L N
U O U A YES, EXCLUSIVELY...................1
A D A M YES, JOINTLY.......................2
YES, BOTH EXCLUSIVELY AND JOINTLY..3 RESPONSE CODES:
B E B E
NO....................4>>NEXT MODULE
L L
ALONE............................1
E E WITH ADULT FEMALES PRESENT.......2
WITH ADULT MALES PRESENT.........3
WITH ADULTS MIXED SEX PRESENT....4
WITH CHILDREN PRESENT............5
Jewellery WITH ADULTS MIXED SEX
V01
AND CHILDREN PRESENT........6
Semi-precious and precious metals This question measures reported ownership of These guidelines recommend collecting
V02
valuables and also the form of ownership; i.e. valuation data for all high-value
Semi-precious and precious stones whether the valuable is owned exclusively or consumer durables. Respondents
V03 should estimate the current value based
jointly by the respondent. Categories may
include but not be limited to jewellery, semi- on the quality of their particular
V04 Paintings (Reason(s) interview not administered with the
precious and precious metals or stones, valuables. The full amount that would be
received in the sale should be listed, respondent(s) alone should be explained in the
Other (specify) paintings and other artefacts. Countries will
V05 regardless whether or not all of it would survey officer remarks section).
need to determine the categories of valuables to
include based on prevalence rates from existing be kept by the respondent. If the
data sources and also on policy needs. Include respondent is not sure how to answer,
a category of “Other (specify)" to create an enumerators should probe on this
exhaustive module on valuables. Because some question by encouraging the respondent
valuables may be owned collectively by all to consider the price received for other
household members while others are owned valueables of the same type that have
individually, countries can exclude those been sold in the area.

Model questionnaire
durables that are reported as being collectively
owned during focus group discussions or pilot
testing of the questionnaire.

211
212

FINANCIAL ASSETS
FA-1. Do you currently own any of the following: a bank account, a microfinance account, an informal savings
programme, stocks and shares, a pension fund, life insurance or another type of account?

YES.......................1
NO........................2

This is the screening question to determine whether the respondent should complete the module. If the respondent does not own any financial assets, skip to the next module as the remainder
of this module is only administered to the respondent if she/he self -reports owning any financial assets. The question also measures the reported ownership of any financial assets (not by
type of financial asset). Categories of financial assets may include, but not be limited to, bank savings, savings and credit associations, post office accounts, informal savings accounts,
savings accounts through NGOs, stocks, bonds, pension funds, and insurance funds. Sums of money that respondents lend to fami ly or friends are also financial assets and should be
included in estimates of the gender wealth gap.
Countries will need to determine the categories of financial assets to include based on prevalence rates from prior financial or household surveys and also on policy needs. For example,
countries with limited access to financial services may want to include a category for cash savings.
Countries should also assess how best to present the categories of financial assets to the respondent. An approach adopted in the EDGE pilot in Mexico was to divide the module on financial
assets into formal and informal financial assets, beginning with informal assets. This approach was suggested because respond ents with restricted access to formal financial services were
often daunted when confronted with the array of service providers to whom they do not have access and, as a result, were relu ctant to discuss their apparently “insignificant” (i.e. informal)
savings with the enumerators.
It is recommended that this module on financial assets should be placed near the end of the questionnaire and that informatio n on financial assets should always be collected at the individual
level.

FA-2. FA-3.
Please list each financial asset that you own, exclusively or jointly. Is your name on the account as an owner?

YES, ALONE............................1>>FA-6
A YES, JOINTLY WITH ONE OR MORE PERSONS..2
S NO..............................3>>FA-6

S
E
By recording each financial asset owned by the respondent, a This question measures whether the respondent is a documented
T
respondent roster of financial assets is created. This owner of the financial asset by virtue of his or her name being listed
information also measures reported ownership of financial on ownership documents for the account. It also measures the form
C assets, by type of financial asset. If more than one of the same of documented ownership of financial assets; i.e. whether the
O type of financial asset is owned (e.g. two bank accounts), each respondent owns the financial asset exclusively or jointly with one
D one should be listed, starting with the most valuable one. The or more persons. Because the benefits of ownership may differ if an
E list of all financial assets owned by the respondent should be individual owns financial assets alone or jointly, countries are
provided before proceeding to the next question. encouraged to collect information on the form of documented
ownership.

F01

F02

F03

F04

F05

F06

F07

…
 Model questionnaire 213

FINANCIAL ASSETS (CONTINUED)

FA-4. FA-5. FA-6.

How many other persons’ names are on the account Is the name of your spouse or partner listed on the account as an What is the current value of the financial asset?
for this [financial asset], including household owner for this [financial asset]?
members and non-household members?

Identifying whether the respondent jointly owns the financial

This question measures the value of financial assets.
asset with her/his spouse or partner allows for the construction
YES..........1 Respondents should be encouraged to estimate the current
of an indicator on joint (reported) ownership of financial assets
NO...........2>>FA-7 value of the account in the currency in which it is held. As
between spouses/partners, the most common form of joint
discussed above, respondents may be reluctant to provide
ownership. Other patterns of joint ownership are also possible,
If the respondent jointly owns the financial asset, account balances, and enumerators should be trained
information on the number of joint reported such as between siblings or a parent and an adult child, and
accordingly on how to solicit sensitive information. The training
owners is needed for calculation of the gender countries that are interested in identifying these patterns are
should include the need to emphasize to respondents the
wealth gap, as discussed in part four of these encouraged to ask, in place of Qs 5-6, “Who are the joint
security and confidentiality of providing such information. An
guidelines. owners, including household members and non-household
alternative approach is to provide a range of values as response
members?” The personal IDs assigned to household members
categories and use the average for calculation of the gender
in the household roster should be recorded for each household
wealth gap. A response category for “refuses to respond” should
member who is a joint reported owner and each non-household
be included with either approach.
member who jointly owns the financial asset should be
assigned a non-household member ID code (e.g. 100).
FA-4. FA-5. FA-6.
How many other persons’ names are on the account Is the name of your spouse or partner listed on the account as an What is the current value of the financial asset?
for this [financial asset], including household owner for this [financial asset]?
members and non-household members?

Identifying whether the respondent jointly owns the financial

This question measures the value of financial assets.
asset with her/his spouse or partner allows for the construction
YES..........1 Respondents should be encouraged to estimate the current
of an indicator on joint (reported) ownership of financial assets
NO...........2>>FA-7 value of the account in the currency in which it is held. As
between spouses/partners, the most common form of joint
discussed above, respondents may be reluctant to provide
ownership. Other patterns of joint ownership are also possible,
If the respondent jointly owns the financial asset, account balances, and enumerators should be trained
information on the number of joint reported such as between siblings or a parent and an adult child, and
accordingly on how to solicit sensitive information. The training
owners is needed for calculation of the gender countries that are interested in identifying these patterns are
should include the need to emphasize to respondents the
wealth gap, as discussed in part four of these encouraged to ask, in place of Qs 5-6, “Who are the joint
security and confidentiality of providing such information. An
guidelines. owners, including household members and non-household
alternative approach is to provide a range of values as response
members?” The personal IDs assigned to household members
categories and use the average for calculation of the gender
in the household roster should be recorded for each household
wealth gap. A response category for “refuses to respond” should
member who is a joint reported owner and each non-household
be included with either approach.
member who jointly owns the financial asset should be
assigned a non-household member ID code (e.g. 100).

FINANCIAL ASSETS (CONTINUED)

FA-7. Does any person or any YES..........1

business owe you any money? NO...........2>>NEXT MODULE

This is the screening question to assess whether the remaining questions in this module should be asked. If no other persons
currently owe the respondent any money, skip to the next module.

FA-8. FA-9. FA-10.

Who was the money lent to? Did anyone jointly lend the money with you, How many other people jointly lent the money with
including household members and non- you?
FAMILY.......................1 household members?
FRIEND.......................2
CLIENT/ CUSTOMER.............3
YES..........1
EMPLOYEE.....................4
NO...........2
L OTHER(SPECIFY)..............96 YES............1
O NO.............2>>FA-12
A
N If the respondent jointly lent money, information
The information collected in this question allows If the respondent jointly lent money, on the number of joint lenders is needed for
for analysis of patterns of lending, by sex of the information on the number of joint lenders is calculation of the gender wealth gap, as
N lender. NSOs should customize the response needed for calculation of the gender wealth discussed in part four of these guidelines.
O categories according to the country context. gap, as discussed in part four of these
guidelines.

WRITTEN DESCRIPTION CODE

F11

F12

F13

…
FINANCIAL ASSETS (CONTINUED)

214

FINANCIAL ASSETS (CONTINUED)

FA-11. FA-12.
Is one of the joint lenders your spouse/partner? SELECT THE METHOD WHICH WAS USED TO INTERVIEW THE
RESPONDENT.

RESPONSE CODES:
YES..........1
NO...........2
ALONE.......................................1
WITH ADULT FEMALES PRESENT..................2
Identifying whether the respondent jointly lent the money with
WITH ADULT MALES PRESENT....................3
FA-11.
his/her spouse or partner allows for analysis of patterns of FA-12.
lending between spouses/partners. Other patterns of lending WITH ADULTS MIXED SEX PRESENT...............4
Is one of the joint lenders your spouse/partner? SELECT THE METHOD
WITH CHILDREN WHICH WAS USED TO INTERVIEW THE
PRESENT.......................5
are also possible, such as between siblings or a parent and
an adult child, and countries that are interested in identifying RESPONDENT.
WITH ADULTS MIXED SEX AND CHILDREN PRESENT..6
these patterns are encouraged to ask, in place of Qs 10 and
11, “Who are the joint lenders, including household members
and non-household members?” The personal IDs assigned to RESPONSE CODES:
YES..........1
household members in the household roster should be
NO...........2
recorded for each household member who jointly lent the ALONE.......................................1
(Reason(s) interview
WITH ADULT not administered
FEMALES with the respondent(s) alone should
PRESENT..................2
money andwhether
Identifying each non-household
the respondentmember who the
jointly lent jointly lent with
money the
money beWITH
explained
ADULTin the survey
MALES officer remarks section).
PRESENT....................3
his/her should
spousebe or assigned a non-household
partner allows member
for analysis of patternsIDofcode
(e.g. 100).
lending between spouses/partners. Other patterns of lending WITH ADULTS MIXED SEX PRESENT...............4
are also possible, such as between siblings or a parent and WITH CHILDREN PRESENT.......................5
an adult child, and countries that are interested in identifying WITH ADULTS MIXED SEX AND CHILDREN PRESENT..6
these patterns are encouraged to ask, in place of Qs 10 and
11, “Who are the joint lenders, including household members
and non-household members?” The personal IDs assigned to
household members in the household roster should be
recorded for each household member who jointly lent the (Reason(s) interview not administered with the respondent(s) alone should
money and each non-household member who jointly lent the
money should be assigned a non-household member ID code be explained in the survey officer remarks section).
(e.g. 100).
 Model questionnaire 215

LIABILITIES
LIA-1. Do you owe money to any person or institution?
YES......................1
NO.......................2>>END OF QUESTIONNAIRE
REFUSES TO RESPOND.......3>>END OF QUESTIONNAIRE

This is the screening question to determine whether the respondent should complete the module. If the respondent does not owe any money, including lines
of credit, the module should be skipped as it is only administered to the respondent if she/he self-reports having any debts.

LIA-2. LIA-3. LIA-4. LIA-5.

Who is the person or institution that What was the main purpose for borrowing the Is anyone jointly responsible for How many other persons are
you owe? money? paying back the debt with you, jointly responsible for paying
including household members and back the debt?
non-household members?
TO BUY PRINCIPAL DWELLING.......1
FAMILY.....................1 TO BUY AGRICULTURAL LAND........2
FRIEND.....................2 TO BUY NON-AGRICULTURAL LAND....3
EMPLOYER...................3 YES............1
TO PAY EDUCATION EXPENSES.......4
LANDLORD...................4 NO.............2>>LIA-7
TO PAY HEALTH EXPENSES..........5
BANK.......................5 TO PAY FOR FOOD.................6
D PRIVATE MONEYLENDER.......6 TO PURCHASE A VEHICLE...........7
SHOPKEEPER................7 TO PURCHASE OTHER CONSUMER
E OTHER(SPECIFY)............96 DURABLES........................8
B TO PAY FOR CEREMONIAL EXPENSES
T (WEDDINGS, FUNERALS, ETC).......9
OTHER(SPECIFY).................96
N
O
This question measures the main Because the person who If the respondent reports
This question measures the
purpose for which the respondent borrowed the money may that she or he is
source of the respondent’s
borrowed the money. NSOs will need not be the same person who responsible for paying
debt. NSOs will need to
to customize the response categories is responsible for paying back the debt with one or
customize the response
according to their country context and back the debt, the question more persons, information
categories according to their
based on whether they wish to should be phrased as on the number of joint
country context.
develop net wealth estimates by asset, suggested, rather than as borrowers is needed for
as discussed above. “Who jointly borrowed the calculation of the net
money with you?” gender wealth gap, as
discussed in part four of
these guidelines.

L01

L02

L03

…
 LIABILITIES (CONTINUED)

216

This is the screening question to determine whether the respondent should complete the module. If the respondent does not owe
of credit, the module should be skipped as it is only administered to the respondent if she/he self
LIABILITIES (CONTINUED)

LIA-6. LIA-7. LIA-8.

Is one of these joint borrowers your spouse or partner? What is the remaining amount SELECT THE METHOD WHICH WAS USED TO INTERVIEW
to be repaid on the debt? THE RESPONDENT.

This is the screening question to determine whether the respondent should complete the module. If the respondent does not owe
of credit, the module should be skipped as it is only administered to the respondent if she/he self

YES........1
NO.........2 RESPONSE CODES:

LIA-6. LIA-7. ALONE.......................................1

LIA-8.
Identifying whether the respondent is jointly WITH ADULT FEMALES PRESENT..................2
Is one of these joint
responsible borrowers
for the your
debt with spouse
his/her or partner?
spouse or partner WhatThisis question
the remaining amount SELECT THE METHOD
WITH ADULT WHICH WAS USED TO INTERVIEW
MALES PRESENT....................3
D
allows for analysis of debt patterns between spouses/ to be repaid on
measures the debt?
the THE RESPONDENT.
WITH ADULTS MIXED SEX PRESENT...............4
E
partners. Other patterns of joint debt-taking are also outstanding amount WITH CHILDREN PRESENT.......................5
B to be repaid on the WITH ADULTS MIXED SEX AND CHILDREN PRESENT..6
possible, such as between siblings or a parent and an
T adult child, and countries that are interested in debt and should
identifying these patterns are encouraged to ask, in include principal plus
YES........1
N place of LIA-5 and LIA-6, “Who are the joint borrowers
NO.........2 interest. A response RESPONSE CODES:
O of this debt, including household members and non- category for “refuses
household members?” The personal IDs assigned to to respond” should ALONE.......................................1
Identifying (Reason(s) interview not administered with the respondent(s)
household whether
membersthe in respondent
the householdis jointly
roster should be also be included. WITH ADULT FEMALES PRESENT..................2
alone should be MALES
explained in the survey officer remarks section).
responsible This question
D recorded forfor thehousehold
each debt with his/her
member spouse
who isora partner
joint WITH ADULT PRESENT....................3
allows forand
analysis measures the WITH ADULTS MIXED SEX PRESENT...............4
E borrower each of debt patterns member
non-household betweenwho spouses/
is also
partners. Other patterns of joint debt-taking outstanding amount WITH CHILDREN PRESENT.......................5
B responsible for paying back the debt shouldare
be also
possible, such as between member
siblings or to be repaid on the WITH ADULTS MIXED SEX AND CHILDREN PRESENT..6
T assigned a non-household IDacode
parent and
(e.g. an
100).
adult child, and countries that are interested in debt and should
identifying these patterns are encouraged to ask, in include principal plus
N place of LIA-5 and LIA-6, “Who are the joint borrowers interest. A response
O of this debt, including household members and non- category for “refuses
household members?” The personal IDs assigned to to respond” should
(Reason(s) interview not administered with the respondent(s)
household members in the household roster should be also be included.
alone should be explained in the survey officer remarks section).
L01 recorded for each household member who is a joint
borrower and each non-household member who is also
responsible for paying back the debt should be
assigned a non-household member ID code (e.g. 100).
L02

L03

L01
…

L02

L03

…
 Model questionnaire 217

DECISION-MAKING This module on within-household decision-making was included in the EDGE pilot survey in South Africa, in order to analyse the relationship between
women’s and men’s asset ownership and household decision-making. Questions in this module have to asked at the individual level.
DM-1 DM-2 DM-3 DM-4 DM-5
Who usually decides how your Who usually decides how your Who usually makes decisions about Who usually makes decisions about Who usually makes decisions about
income will be used? spouse/ partner's income will be your health care, including visits to making major household visits to your family or relatives?
used? health facilities or practitioners? purchases? (Example: car, house,
etc.)
RESPONDENT..............1 RESPONDENT..............1
SPOUSE/PARTNER..........2 SPOUSE/PARTNER..........2
RESPONDENT AND RESPONDENT AND
SPOUSE/PARTNER SPOUSE/PARTNER
JOINTLY...............3 RESPONDENT..............1 RESPONDENT.............1 RESPONDENT.............1
JOINTLY...............3
PARENTS OF SPOUSE/PARTNER..........2 SPOUSE/PARTNER.........2 SPOUSE/PARTNER.........2
PARENTS OF
SPOUSE/PARTNER........4 RESPONDENT AND RESPONDENT AND RESPONDENT AND
SPOUSE/PARTNER........4
OTHER RELATIVES.........5 SPOUSE/PARTNER SPOUSE/PARTNER SPOUSE/PARTNER
OTHER RELATIVES.........5
NOT APPLICABLE, JOINTLY...............3 JOINTLY..............3 JOINTLY..............3
NOT APPLICABLE,
I DON'T HAVE A PARENTS OF PARENTS OF PARENTS OF
I DON'T EARN
SPOUSE/PARTNER........6 SPOUSE/PARTNER........4 RESPONDENT...........4 RESPONDENT...........4
AN INCOME..............6
NOT APPLICABLE, PARENTS OF PARENTS OF
OTHER (SPECIFY)........96
SPOUSE/PARTNER HAS SPOUSE/PARTNER.......5 SPOUSE/PARTNER.......5
NO INCOME.............7 OTHER RELATIVES........6 OTHER RELATIVES........6
OTHER (SPECIFY)........96

DECISION-MAKING (CONTINUED)

DM-6 DM-7 DM-8 DM-9

In general, is it okay for a man In your opinion, is a husband/ boyfriend Have you ever been hit, slapped, SELECT THE METHOD WHICH WAS USED TO INTERVIEW THE
to hit a woman? justified in hitting or beating his wife/ kicked or been physically hurt in any RESPONDENT.
girlfriend in the following situations: way by a current or former partner/
(a) If she goes out without telling him? spouse?
(b) If she neglects the children?
RESPONSE CODES:
(c) If she argues with him?
(d) If she refuses sex? ALONE.......................................1
(e) If she burns food or does not cook? WITH ADULT FEMALES PRESENT..................2
WITH ADULT MALES PRESENT....................3
WITH ADULTS MIXED SEX PRESENT...............4
WITH CHILDREN PRESENT.......................5
YES IN THE LAST 12 MNTHS....1 WITH ADULTS MIXED SEX AND CHILDREN PRESENT..6
YES MORE THAN 12 MNTHS AGO..2
NEVER.......................3
YES.................1 YES.................1 REFUSE TO ANSWER...........97
NO..................2 NO..................2
REFUSE TO ANSWER...97 REFUSE TO ANSWER...97

(Reason(s) interview not administered with the respondent(s) alone

should be explained in the survey officer remarks section).
a)
b)
c)
d)
e)
218

END OF QUESTIONNAIRE
End-1
ENUMERATOR: ENTER RESPONSE CODE FOR COMPLETION STATUS OF INDIVIDUAL QUESTIONNAIRE:

COMPLETED..............................11
PARTIALLY COMPLETED....................12>>end of interview
NON-CONTACT............................21>>end of interview
NOT INTERVIEWED, REFUSAL...............22>>end of interview
OTHER NON-RESPONSE.....................23>>end of interview
UNOCCUPIED DWELLING....................31>>end of interview
RESIDENTS OF DWELLING ABSENT...........32>>end of interview
DEMOLISHED.............................33>>end of interview
NEW DWELLING UNDER CONSTRUCTION........34>>end of interview
STATUS CHANGED.........................35>>end of interview
LISTING ERROR..........................36>>end of interview
NOT INTERVIEWED, NON-HOUSEHOLD MEMBER..37>>end of interview

(Reason(s) for interview not fully completed or not administered should be explained.)

End-2. MANNER IN WHICH INDIVIDUAL INTERVIEW CONDUCTED. SIMULTANEOUSLY.....................1

SEQUENTIALLY.......................2
OTHER..............................3
NOT APPLICABLE, ONLY 1 INDIVIDUAL INTERVIEW
IN THIS HOUSEHOLD........4

End-3. ENUMERATOR: DID YOU HAVE TO REVISIT THE HOUSEHOLD, IN ORDER TO INTERVIEW THE RESPONDENT?

YES........1
NO.........2 >> End-5

End-4. ENUMERATOR: INDICATE THE NUMBER OF REVISITS YOU MADE TO THE HOUSEHOLD, IN ORDER TO INTERVIEW RESPONDENT.

YES........1
End-5. ENUMERATOR: DO YOU HAVE ANY REMARKS REGARDING THIS INDIVIDUAL QUESTIONNAIRE? NO.........2

PLEASE WRITE DOWN THE REMARKS INDICATING THE SECTION AND QUESTION NUMBER TO WHICH THEY APPLY.
 219

References

Allendorf, Keera (2007). Do women’s land rights promote empowerment and child health in
Nepal? World Development, vol. 35, No. 11, pp. 1975–1988.
American Association for Public Opinion Research (2016). Standard Definitions: Final Disposi-
tions of Case Codes and Outcome Rates for Surveys. Chicago. Available at www.aapor.
org/AAPOR_Main/media/publications/Standard-Definitions20169theditionfinal.pdf.
Antonopoulos, Rania and Maria Floro (2005). Asset ownership along gender lines: evidence
from Thailand. Economics Working Paper No. 418. Annandale-on-Hudson, New York:
Levy Economics Institute, Bard College.
Bardasi, Elena, and others (2011). Do labor statistics depend on how and to whom the questions
are asked? Results from a survey experiment in Tanzania. The World Bank Economic
Review, vol. 25, No. 3, pp. 418-447. 
Bhattacharyya, Manasi, Arjun Bedi and Amrita Chhachhi (2011). Marital violence and women’s
employment and property status: evidence from North Indian villages. World Develop-
ment, vol. 39, No. 9, pp. 1676–1689.
Bomuhangi, Allan, Cheryl Doss and Ruth Meinzen-Dick (2011). Who owns the land? Perspec-
tives from rural Ugandans and implications for land acquisitions. Discussion Paper
01136. Washington, D.C.: International Food Policy Research Institute.
Canadian International Development Agency (1997). Guide to Gender-Sensitive Indicators.
Quebec: Hull.
Carter, Michael and Christopher Barrett (2006). The economics of poverty traps and persis-
tent poverty: an asset-based approach. Journal of Development Studies, vol. 42, No. 2,
pp. 178–199.
Chen, Joyce and LaPorchia Collins (2014). Let’s talk about the money: spousal communica-
tions, expenditures and farm production. American Journal of Agricultural Economics,
vol. 96, No. 5, pp. 1272–1290.
Clarke, Stephen (2007). Improving the quality of EU farm registers. Paper presented at the
Seminar on Registers in Statistics, Helsinki, 21–23 May.
Davies, James and others (2008). The world distribution of household wealth. Personal Wealth
from a Global Perspective. Oxford: Oxford University Press.
Deere, Carmen Diana and Cheryl Doss (2006). The gender asset gap: what do we know and why
does it matter? Feminist Economics, vol. 12, Nos. 1–2, pp. 1–50. Available at http://doi.
org/10.1080/13545700500508056.
Deere, Carmen Diana, Gina Alvarado and Jennifer Twyman (2012). Gender inequality in asset
ownership in Latin America: female owners vs. household heads. Development and
Change, vol. 43, No. 2, pp. 505–530.
Deere, Carmen and Jennifer Twyman (2012). Asset ownership and egalitarian decision making
in dual-headed households in Ecuador. Review of Radical Political Economics, vol. 44,
No. 3, pp. 313–320.
Deere, Carmen Diana and Magdalena León (2003). The gender asset gap: land in Latin America.
World Development, vol. 31, No. 6, pp. 925–947.
220

Doss, Cheryl, Ruth Meinzen-Dick and Allan Bomuhangi (2013). Who owns the land? Perspec-
tives from rural Ugandans and implications for large-scale land acquisitions. Feminist
Economics, vol. 20, No. 1, pp. 76–100. Available at http://doi.org/10.1080/13545701.20
13.855320.
Doss, Cheryl and others (2013). Gender Inequalities in Ownership and Control of Land in Africa:
Myth versus Reality. Washington, D.C.: International Food Policy Research Institute.
Available at http://ebrary.ifpri.org/cdm/ref/collection/p15738coll2/id/127957.
Doss, Cheryl and others (2013). Lessons from the field: implementing individual asset surveys
in Ecuador, Ghana, India and Uganda. Journal of Economic Inequality, vol. 11, No. 2,
pp. 249–265.
Doss, Cheryl and others (2013). Measuring personal wealth in developing countries: interview-
ing men and women about asset values. Gender Asset Gap Project Working Paper Series,
No.15. Bangalore: Centre of Public Policy, Indian Institute of Management. Available
at www.researchgate.net/publication/270278605_Measuring_Personal_Wealth_in_
Developing_Countries_Interviewing_Men_and_Women_about_Asset_Values.
Doss, Cheryl and others (2014). The gender asset and wealth gaps. Development, vol. 57,
Nos. 3–4, pp. 400–409. Available at http://doi.org/10.1057/dev.2015.10.
Doss, Cheryl and others (2018). Do men and women estimate property values differently? World
Development, vol. 107, issue C, pp. 75–78.
European Commission, International Monetary Fund, Organization for Economic Coopera-
tion and Development, United Nations and World Bank (2009). System of National
Accounts 2008. New York.
Eurostat, International Labour Organization, International Monetary Fund, Organization for
Economic Cooperation and Development, United Nations, Economic Commission
for Europe and World Bank (2013). Handbook on Residential Property Prices Indices
(RPPIs). Brussels.
Eurostat (2016). Agricultural land prices and rents data for the European Union. Research
Paper. Brussels.
Eurostat (2017). Private households by type, tenure status and NUTS 2 region, based on 2011
population censuses (extracted in August 2017).
Fisher, Monica, Jeffrey Reimer and Edward Carr (2009). Who should be interviewed in surveys
of household income? World Development, vol. 38, No. 7, pp. 966–973.
Food and Agriculture Organization of the United Nations (2002). Land tenure and rural devel-
opment. FAO Land Tenure Studies 3. Rome.
Food and Agriculture Organization of the United Nations (2003). Multilingual Thesaurus on
Land Tenure. Rome.
Food and Agriculture Organization of the United Nations (2005). A System of Integrated Agri-
cultural Censuses and Surveys, Volume 1: World Programme for the Census of Agricul-
ture 2010. Rome.
Food and Agriculture Organization of the United Nations (2017). World Programme for the
Census of Agriculture 2020, vol. 1, Programme, Concepts and Definitions. Rome.
Food and Agriculture Organization of the United Nations (2017). World Programme for the
Census of Agriculture 2020, vol. 2, Operational Guidelines. Rome.
Food and Agriculture Organization of the United Nations (forthcoming). Measuring individu-
als’ rights to land: an integrated approach to data collection for SDG indicators 1.4.2
and 5.a.1.
Food and Agriculture Organization of the United Nations and the World Bank (2014). Gender
Disaggregated Data—Western Balkans. Statistical Reports 2005–2013.
 References 221

Frankenberg, Elizabeth (2000). Community and price data. Designing Household Survey Ques-
tionnaires for Developing Countries: Lessons from Fifteen Years of the Living Standards
Measurement Study, vol. 1, Margaret Grosh and Paul Glewwe, eds. Washington, D.C.:
World Bank.
Gaziano, Cecilie (2005). Comparative analysis of within-household respondent selection tech-
niques. Public Opinion Quarterly, vol. 69, No. 1, pp. 124–157.
Gillespie, Stuart and Suneetha Kadiyala (2005). HIV/AIDS and food and nutrition security:
interactions and response. American Journal of Agricultural Economics, vol. 87, No. 5,
pp. 1282–1288.
Grabe, Shelly (2010). Promoting gender equality: the role of ideology, power, and control in the
link between land ownership and violence in Nicaragua. Analyses of Social Issues and
Public Policy, vol. 10, No. 1, pp. 146–170.
Grosh, Margaret and Paul Glewwe (2000). Designing Household Survey Questionnaires for
Developing Countries: Lessons from Fifteen Years of the Living Standards Measurement
Study, vol. 1. Washington, D.C.: World Bank.
Groves, Robert (1989). Survey Errors and Survey Costs. Hoboken, New Jersey: John Wiley and
Sons.
Groves, Robert and others (2009). Survey methodology, 2nd ed. Hoboken, New Jersey: John
Wiley and Sons.
Hallman, Kelly (2000). Mother-father resource control, marriage payments, and girl-boy health
in rural Bangladesh. Discussion Paper 93. Washington, D.C.: Food Consumption and
Nutrition Division, International Food Policy Research Institute.
Household Finance and Consumption Network (2016). The Household Finance and Consump-
tion Survey: results from the second wave. ECB Statistics Paper Series, No. 18. Frankfurt
am Main: European Central Bank.
Hussmanns, Ralf, Farhad Mehran and Vijay Verma (1990). Surveys of Economically Active
Population, Employment, Unemployment and Underemployment: An ILO Manual on
Concepts and Methods. Geneva: International Labour Office.
International Labour Organization (2018), Employment by sector—ILO modelled
estimates. Available at www.ilo.org/ilostat/faces/wcnav_defaultSelection?_
afrLoop=704653335497001&_afrWindowMode=0&_afrWindowId=ww2zuc2wi_1#!%
40%40%3F_afrWindowId%3Dww2zuc2wi_1%26_afrLoop%3D704653335497001%26_
afrWindowMode%3D0%26_adf.ctrl-state%3Dww2zuc2wi_74.
Jacobs, Krista, and Aslihan Kes (2015). The ambiguity of joint asset ownership: cautionary tales
from Uganda and South Africa. Feminist Economics, vol. 21, No. 3, pp. 23–55.
Kilic, Talip, and Heather Moylan (2016). Methodological Experiment on Measuring Asset Own-
ership from a Gender Perspective (MEXA): Technical Report. Washington, D.C.: World
Bank.
Kish, Leslie (1949). A procedure for objective respondent selection within the household. Jour-
nal of the American Statistical Association, vol. 44, No. 247, pp. 380–387.
Kish, Leslie (1965). Survey Sampling. New York: John Wiley and Sons.
Meinzen-Dick, Ruth, and Rajendra Pradhan (2002). Legal pluralism and property rights. CAPRi
Working Paper. Washington, D.C.: International Food Policy Research Institute.
Meinzen-Dick, Ruth, and others (2014). The gender asset gap and its implications for agricul-
tural and rural development. Gender in Agriculture: Closing the Knowledge Gap, Agnes
Quisumbing and others, eds. Rome: Food and Agriculture Organization of the United
Nations; Dordrecht, Netherlands: Springer Science and Business Media B.V.
Minnesota Population Center (2017). Integrated Public Use Microdata Series, International,
data set version 6.5. Minneapolis: University of Minnesota. Available at https://doi.
org/10.18128/D020.V6.5.
222

Muñoz, Juan (2008). A guide for data management of household surveys. Household Sample
Surveys in Developing and Transition Countries. Studies in Methods, Series F, No. 96.
United Nations publication, Sales No. E.05.XVII.6.
Neciu, Adriana (2013). Approaches to measuring asset ownership and control in agricultural
censuses and surveys. Paper prepared for the EDGE project. Rome: Food and Agricul-
ture Organization of the United Nations.
Organization for Economic Cooperation and Development (2013). Guidelines for Micro Statis-
tics on Household Wealth. Paris.
Organization for Economic Cooperation and Development (2013). Framework for Statistics on
the Distribution of Household Income, Consumption and Wealth. Paris.
Panda, Pradeep and Bina Agarwal (2005). Marital violence, human development and women’s
property status in India. World Development, vol. 33, No. 5, pp. 823–850.
Panda, Pradeep and others (2006). Property Ownership and Inheritance Right of Women for
Social Protection—the South Asia Experience. Washington, D.C.: International Center
for Research on Women.
Petterson, Hans and Pedro Luis do Nascimento Silva (2005). Analysis of design effects for sur-
veys in developing countries. Household Sample Surveys in Developing and Transition
Countries. New York: United Nations.
Quisumbing, Agnes and John Maluccio John (2003). Resources at marriage and intrahousehold
allocation: evidence from Bangladesh, Ethiopia, Indonesia, and South Africa”. Oxford
Bulletin of Economics and Statistics, vol. 65, No. 3, pp. 283–327.
Statistics South Africa (2012). Census 2011 Municipal Report—KwaZulu-Natal. Pretoria.
United Nations (1984). Handbook of Household Surveys, Revised edition, Studies in Methods,
Series F, No. 31. Sales No. E.83.XVII.13.
United Nations (2008). Designing Household Survey Samples: Practical Guidelines, Studies in
Methods, Series F, No. 98. Sales No. E.06.XVII.13.
United Nations (2014). System of Environmental-Economic Accounting 2012—Central Frame-
work, Series F, No. 109. Sales No. E.12.XVII.12.
United Nations (2014). Principles and Recommendations for Population and Housing Censuses,
Revision 3. Statistical papers, Series M, No. 67/Rev. 3. Sales No. E.15.XVII.10.
United Nations (2015). The World’s Women 2015: Trends and Statistics. Series K, No. 20. Sales
No. E.15.XVII.8.
United Nations, Statistics Division, United Nations Entity for Gender Equality and the Empow-
erment of Women and National Statistics and Geographic Institute of Mexico (2016).
Assessing Mexico’s pilot survey on measuring individual level asset ownership and
entrepreneurship from a gender perspective. EDGE final report. New York.
United Nations (forthcoming). Guidelines on Use of Electronic Data Collection Technologies in
Censuses.
United Nations, Statistical Commission and Economic Commission for Europe (2000). Ter-
minology on Statistical Metadata, Sales No. E.00.II.E.21. Available at www.unece.org/
fileadmin/DAM/stats/publications/53metadaterminology.pdf.
United Nations, Economic Commission for Europe (2011). Using Administrative and Secondary
Sources for Official Statistic: A Handbook of Principles and Practices. Geneva.
United Nations, Economic Commission for Europe (2015). Conference of European Statisticians
Recommendations for the 2020 Censuses of Population and Housing. Geneva.
United Nations, Inter-agency and Expert Group on MDG Indicators (2013). Lessons learned
from MDG monitoring from a statistical perspective: report of the task team on lessons
learned from MDG monitoring of the IAEG-MDG.
 References 223

Vaessen, Martin and others (2008). The demographic and health surveys. Household Sample
Surveys in Developing and Transition Countries. Studies in Methods, Series F, No. 96.
United Nations publication, Sales No. E.05.XVII.6.
Valliant, Richard, Jill Dever and Frauke Kreuter (2013). Practical Tools for Designing and
Weighting Survey Samples. New York: Springer.
World Bank (2015). Survey Solutions Interviewer Manual. Washington, D.C. Available at http://
siteresources.worldbank.org/INTCOMPTOOLS/Resources/8213623-1380598436379/
Interviewer_manual.pdf.
World Bank (2015). Survey Solutions Questionnaire Designer-User’s Guide. Washing-
ton, D.C. Available at http://siteresources.worldbank.org/INTCOMPTOOLS/
Resources/8213623-1380598436379/designer.pdf.
World Bank (2015). Survey Solutions Supervisor Manual. Washington, D.C. Available at http://
siteresources.worldbank.org/INTCOMPTOOLS/Resources/8213623-1380598436379/
Supervisor_manual.pdf.
Yansaneh, Ibrahim (2008). Overview of sample design issues for household surveys in develop-
ing and transition countries. Household Sample Surveys in Developing and Transition
Countries. Studies in Methods, Series F, No. 96. United Nations publication, Sales No.
E.05.XVII.6.
18-00223