PIECING TOGETHER
POVERTY
THE
PUZZLE
PIECING TOGETHER
POVERTY
THE
PUZZLE
© 2018 International Bank for Reconstruction and Development / The World Bank
1818 H Street NW, Washington DC 20433
Telephone: 202-473-1000; Internet: www.worldbank.org
Some rights reserved
1 2 3 4 21 20 19 18
This work is a product of the staff of The World Bank with external contributions. The findings, interpre-
tations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its
Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the
accuracy of the data included in this work. The boundaries, colors, denominations, and other informa-
tion shown on any map in this work do not imply any judgment on the part of The World Bank concern-
ing the legal status of any territory or the endorsement or acceptance of such boundaries.
Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and
immunities of The World Bank, all of which are specifically reserved.
Rights and Permissions
This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO)
http://creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you
are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the
following conditions:
Attribution—Please cite the work as follows: World Bank. 2018. Poverty and Shared Prosperity 2018:
Piecing Together the Poverty Puzzle. Washington, DC: World Bank. License: Creative Commons Attribu-
tion CC BY 3.0 IGO
Translations—If you create a translation of this work, please add the following disclaimer along with
the attribution: This translation was not created by The World Bank and should not be considered an official
World Bank translation. The World Bank shall not be liable for any content or error in this translation.
Adaptations—If you create an adaptation of this work, please add the following disclaimer along with
the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed
in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed
by The World Bank.
Third-party content—The World Bank does not necessarily own each component of the content con-
tained within the work. The World Bank therefore does not warrant that the use of any third-party-
owned individual component or part contained in the work will not infringe on the rights of those third
parties. The risk of claims resulting from such infringement rests solely with you. If you wish to re-use
a component of the work, it is your responsibility to determine whether permission is needed for that
re-use and to obtain permission from the copyright owner. Examples of components can include, but are
not limited to, tables, figures, or images.
All queries on rights and licenses should be addressed to World Bank Publications, The World Bank
Group, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@worldbank.org.
ISBN (paper): 978-1-4648-1330-6
ISBN (electronic): 978-1-4648-1360-3
DOI: 10.1596/978-1-4648-1330-6
Cover design: Patricia Hord.Graphik Design
Library of Congress Cataloging-in-Publication Data has been requested.
Contents
Foreword xi
Acknowledgments xiii
About the Team xv
Abbreviations xix
Overview 1
Staying focused on the poorest 4
Monitoring progress in a growing world 7
Piecing together the poverty puzzle 12
Introduction 15
1 Ending Extreme Poverty: Progress, but Uneven
and Slowing 19
Monitoring extreme poverty: A quarter century of progress 19
Nowcasts and forecasts to 2030 22
Uneven progress: A regional profile of poverty reduction 24
Drilling down: The countries with the most poor 29
Socioeconomic and demographic profile of global poverty 37
Conclusions 39
Annex 1A Historical global and regional poverty estimates 41
Annex 1B Validation check of the 2030 poverty projections 46
Notes 47
2 Shared Prosperity: Mixed Progress 49
Beyond extreme poverty: A focus on the bottom 40 49
Continued progress in most economies though some are falling short 50
The poorest countries have limited information about shared prosperity 54
Growth at the bottom and the top is not always even 55
Who are the bottom 40? 58
Monitoring the twin goals 59
v
Annex 2A Shared prosperity definitions 62
Annex 2B Shared prosperity estimates by economy 63
Notes 66
3 Higher Standards for a Growing World 67
Introduction 67
Higher poverty lines for everyone: US$3.20 and US$5.50 a day 68
Higher lines tailored to country circumstances: Societal poverty 72
Why not simply use national poverty lines? 79
Conclusion 81
Annex 3A Historical global and regional poverty estimates 83
Notes 85
4 Beyond Monetary Poverty 87
Why look beyond monetary poverty? 88
Considerations for constructing multidimensional poverty measures 90
A first global picture 96
A deeper look 103
Conclusion 108
Annex 4A Comparison of indicators used in multidimensional
poverty measures 110
Annex 4B Multidimensional poverty measures: A formalization 111
Annex 4C Statistical tables 113
Notes 120
Spotlight 4.1 National multidimensional poverty indexes 122
5 Inside the Household: Poor Children, Women, and Men 125
Introduction 125
Beyond headship: Gender and age profiles of the global poor 128
Differences in resources and poverty within households 132
An individual perspective on multidimensional poverty 140
Conclusion 144
Annex 5A Technical note: Estimating intrahousehold resource shares 147
Notes 148
Appendix A Data Details 151
Main databases for the report 151
Classification of economies 152
Chapter 1 data and methodology 154
Chapter 2 data and methodology 159
Chapter 3 data and methodology 161
Chapter 4 data and methodology 162
Chapter 5 data and methodology 165
Note 166
References 167
vi POVERTY AND SHARED PROSPERITY 2018
Boxes
1.1 Alignment of the SDGs and the Twin 4.2 Incorporating Aspects of Quality
Goals of the World Bank Group 20 into Multidimensional Poverty
1.2 Chapter 1: Data Overview 21 Measures 95
1.3 India: Issues with the 2015 Poverty 4.3 Chapter 4: Data Overview 97
Estimate and 2030 Forecasts 32 5.1 Differences in Needs and
2.1 The Global Database of Shared Equivalence Scales 127
Prosperity 50 5.2 Chapter 5: Data Overview 128
2.2 Country Stories 53 5.3 Dietary Diversity as an Indicator of
2.3 The Shared Prosperity Premium and Individual-Level Food Security 135
Other Concepts of Inequality 56 5.4 Child Poverty 141
4.1 Early Applications of Multidimensional 5.5 Gender and Socioemotional Dimensions
Poverty Measurement 90 of Poverty: Participatory Studies 145
Figures
O.1 Global Extreme Poverty Rate and 1.10 Household Size and Dependency
Headcount, 1990–2015 2 Ratio in Sub-Saharan Africa 34
O.2 Share of Individuals in 1.11 Under-Five Mortality, Fertility, and
Multidimensional Poverty, circa 2013 5 Population Growth in Sub-Saharan
O.3 Percent of Females and Males Living Africa 35
in Households in Extreme Poverty, 1.12 Concentration of Extreme Poverty
by Age Group, circa 2013 6 in Fragile and Conflict-Affected
O.4 National and Societal Poverty Lines Situations 36
in a Growing World 9 1.13 Fragile Situations Perform Poorly in
O.5 Societal Poverty, Global Estimates, Multiple Constituent Components
1990–2015 9 of Fragility 37
O.6 Contribution to Multidimensional 1A.1 Global Total Consumption Gap of
Poverty, by Dimension, Selected the Extreme Poor, 1990–2015 41
Countries 11 1B.1 Projections to 2015 of Global Extreme
O.7 Gender Gaps, Individual Poverty 46
Multidimensional Poverty, Selected 2.1 Shared Prosperity, 91 Economies,
Countries 12 circa 2010–15 51
1.1 Global Extreme Poverty Rate and 2.2 Shared Prosperity Estimates,
Headcount, 1990–2015 21 91 Economies, by Region, Group,
1.2 Projections to 2030 of Global Extreme and Income 55
Poverty 23 2.3 Correlation between Shared
1.3 Number of Extreme Poor by Region, Prosperity and the Shared Prosperity
1990–2030 25 Premium, 91 Economies 57
1.4 Regional GDP per Capita Growth 2.4 Growth across Deciles of the Income
and Average Growth for the Extreme Distribution, Selected Countries 58
Poor, 1990–2017 25 2.5 Extreme Poverty and the Bottom 40,
1.5 Extreme Poverty Rate by Region and Selected Countries, circa 2015 59
Country, 2015 28 2.6 Mean Income, by Distribution
1.6 Extreme Poverty, Regional and World Decile, Selected Countries, 2015 59
Trends, 1990–2015 29 2.7 Share of Consumption or Income,
1.7 Rate and Number of the Extreme Poor, by Decile, Selected Countries,
by Income Group, 2015 29 circa 2015 60
1.8 Global Distribution of the Extreme 2.8 Shared Prosperity and Changes
Poor by Region and Country, 2015 30 in Extreme Poverty, 91 Economies,
1.9 Projections to 2030 for the Five circa 2010–15 61
Countries with the Most Extreme Poor 2.9 Shared Prosperity among the Poorest
in 2015 31 Economies, circa 2010–15 61
CONTENTS vii
2B.1 The Shared Prosperity Premium, 4.8 Contribution to Multidimensional
91 Economies, by Region or Income Poverty (M), by Dimension,
Classification 66 Selected Countries 106
3.1 Consumption and Income 4.9 The Poor, by Sociodemographic
Distributions, 1990 and 2015 71 Characteristics, Selected Countries 107
3.2 National Poverty Lines and 4C.1 Share of Individuals in Multidimensional
Economic Development 73 Poverty, by Region, circa 2013 119
3.3 Societal Poverty Line 75 5.1 Percent of Females and Males Living
3.4 Societal Poverty, Global Estimates, in Households in Extreme Poverty,
1990–2015 77 by Age Group, circa 2013 130
3.5 Change in the Societal Poverty Line 5.2 Distribution of People Living in
from Growth 78 Households in Extreme Poverty, by Sex
3.6 Societal Poverty and Shared Prosperity and Employment Status, circa 2013 131
in Costa Rica and Ecuador 79 5.3 Distribution of Households in
3.7 Comparing National and Societal Extreme Poverty, by Demographic
Poverty Lines and Rates, Vietnam, Typology, circa 2013 131
1993–2015 81 5.4 Distribution of Households in
4.1 Share of Individuals in Extreme Poverty, by Economic
Multidimensional Poverty, Typology, circa 2013 132
119 Economies, circa 2013 98 5.5 The Gender Gap in Food Consumption
4.2 Share of Individuals Deprived in at over the Life Cycle, China 134
Least a Given Number of Indicators, 5.6 Caloric Shortfalls of Male Heads
119 Economies, circa 2013 100 and Other Household Members,
4.3 Contribution of Indicators to the Bangladesh 136
Adjusted Headcount Measure (M), 5.7 Estimated Consumption Allocation,
119 Economies, circa 2013 102 Men, Women, and Children,
4.4 Difference in the Share of the Poor Bangladesh and Malawi 139
in Rural Areas, Multidimensional 5.8 Individual Poverty Rates, Nuclear
Headcount versus Monetary Households, Bangladesh and Malawi 140
Headcount, 119 Economies, 5.9 Gender Gaps, Education, and
circa 2013 103 Nutrition Deprivation, Selected
4.5 Contribution to Monetary and Countries 142
Multidimensional Poverty, by 5.10 Gender Gaps, Individual
Household Type, 119 Economies, Multidimensional Poverty,
circa 2013 103 Selected Countries 143
4.6 Share of Individuals Deprived in at A.1 Use of Income/Consumption to
Least a Given Number of Indicators, Measure Poverty 157
Selected Countries 105 A.2 Shared Prosperity Indicators Are
4.7 The Headcount Ratio, by Alternative Less Likely in Economies at Lower
Poverty Measures, Selected Countries 106 GDP per Capita 160
Maps
O.1 Shared Prosperity across the World, 2.1 Shared Prosperity across the World,
91 Economies, circa 2010–15 10 91 Economies, circa 2010–15 52
1.1 Extreme Poverty Rate by Country, 4.1 Provincial Poverty Rates, Ecuador 108
2015 27
Tables
O.1 Poverty at Higher Poverty Lines, 1.2 Education and Access to Services
US$3.20 and US$5.50 (2011 PPP) 8 among the Extreme Poor and Nonpoor
1.1 Age and Gender Profile of the Households 39
Extreme Poor, 2015 38
viii POVERTY AND SHARED PROSPERITY 2018
1A.1 Global and Regional Extreme Poverty, 4.6 The Multidimensionally Poor and
1990–2015 42 the Breadth of Deprivation, by Number
1A.2 Extreme Poverty, by Economy, of Deprivations, 119 Economies,
2015 43 circa 2013 101
2.1 Shared Prosperity and Shared 4.7 Regional Contributions to
Prosperity Premium, 91 Economies, Multidimensional Poverty,
119 Economies, circa 2013 101
Summary Table, circa 2010–15 50
4.8 Percent of Individuals Deprived,
B2.3.1 Number of Economies with Top
by Indicator, Selected Countries 104
Incomes Estimated in the World
Inequality Database and in the Poverty 4A.1 Dimensions and Indicators 110
and Shared Prosperity 2018 Report 56 4C.1 People Living in Monetary or
2B.1 Shared Prosperity Estimates, 91 Multidimensional Poverty, by
Economies, circa 2010–15 63 Rural-Urban Areas, 119 Economies,
circa 2013 113
2B.2 Changes in Shared Prosperity,
4C.2 People Living in Monetary or
67 Economies, circa 2008–13 to
Multdimensional Poverty, by
circa 2010–15 65
Household Type, 119 Economies,
2B.3 Changes in the Shared Prosperity circa 2013 113
Premium, 67 Economies,
4C.3 Individuals in Households Deprived
circa 2008–13 to circa 2010–15 65 in Each Indicator, 119 Economies,
3.1 National Poverty Lines, circa 2011 69 circa 2013 114
3.2 Poverty at Higher Poverty Lines, 4C.4 Multidimensional Poverty across
US$3.20 and US$5.50 (2011 PPP) 70 Alternative Measures, 119 Economies,
3.3 Average Societal Poverty Lines, circa 2013 117
by Region and Income Classification, 5.1 Households in Extreme Poverty,
1990–2015 76 Rates and Distribution by Headship,
3.4 Societal Poverty Rates, 1990–2015 77 circa 2013 129
5.2 Recent Data Sets on Individualized
3A.1 Historical Trends, Global Poverty
Consumption 133
Estimates, 1990–2015 83
5.3 Individuals Misclassified by the
3A.2 Historical Trends, Regional Poverty
Household Measure of Caloric
Rates, 1990–2015 84
Availability 136
3A.3 Historical Trends, Regional Number
5.4 Indicators and Dimensions, the
of Extreme Poor, 1990–2015 84 Individual and Household
4.1 Dimensions of Well-Being and Multidimensional Poverty Measure 142
Indicators of Deprivation 93 A.1 Overview of Principal Data Sources
4.2 Indicator Weights: Analysis of Three by Chapter 152
Dimensions 96 A.2 Shared Prosperity Availability across
4.3 Indicator Weights: Analysis of Five Rounds 161
Dimensions 96 A.3 Surveys Used in Chapter 1 and
4.4 People Living in Monetary or Chapter 4 in Cases Where Different
Multidimensional Poverty, Survey Rounds Are Used 163
119 Economies, circa 2013 97 A.4 Household Surveys, Six-Country
4.5 Individuals in Households Deprived Sample 164
in Each Indicator, 119 Economies, A.5 Household Surveys for Case Studies
circa 2013 99 and Sharing Rule Estimates 166
CONTENTS ix
Foreword
Five years ago, the World Bank Group set two overarching goals: to end extreme poverty by
2030, and to promote shared prosperity by boosting the incomes of the bottom 40 percent of
the population in each country.
As this year’s Poverty and Shared Prosperity report documents, the world continues to
make progress toward eliminating poverty. In 2015, approximately one-tenth of the world’s
population lived in extreme poverty—the lowest poverty rate in recorded history. This is an
impressive achievement, considering that in 1990, more than a third of people on earth lived in
extreme poverty. Since we last reported on global poverty two years ago, the number of poor
has diminished by 68 million.
But we cannot take success for granted. Poverty is on the rise in several countries in
Sub-Saharan Africa, as well as in fragile and conflict-affected situations. In many countries,
the bottom 40 percent of the population is getting left behind; in some countries, the living
standard of the poorest 40 percent is actually declining. To reach our goal of bringing extreme
poverty below 3 percent by 2030, the world’s poorest countries must grow at a rate that far
surpasses their historical experience. There is no room for complacency. We must intensify
the effort to promote economic growth in the lagging countries and ensure that the poorest 40
percent of the population benefits more from economic progress.
Reducing extreme poverty to less than 3 percent by 2030 remains a considerable challenge,
and it will continue to be our focus. At the same time, most of the world’s poor now live in
middle-income countries, and our research indicates that those countries tend to have a more
demanding view of poverty. Drawing on national poverty lines, we now also report poverty rates
at two higher thresholds—$3.20 per day and $5.50 per day—which are typical of standards in
lower- and upper-middle-income countries.
These thresholds are a recognition that the concept of poverty itself is dependent on one’s
social circumstances. What is a luxury in one society could be a necessity in another. Even if
minimum physical needs are met, people cannot be said to lead flourishing lives if they are not
able to conduct themselves with dignity in the society in which they live. The societal poverty
rate presented in this report gauges people’s well-being by the standard of their surroundings.
Poverty encompasses a shortfall in income and consumption, but also low educational
achievement, poor health and nutritional outcomes, lack of access to basic services, and
a hazardous living environment. If we hope to tackle poverty “in all its forms everywhere”
as the Sustainable Development Goals call for, we must understand and measure poverty
in all of its manifestations. This report presents results of the World Bank’s first exercise
in multidimensional global poverty measurement to account for multiple and overlapping
components of poverty.
xi
Traditionally, poverty is measured at the household level, but because there is inequality
within households, there are undoubtedly people living in poverty within nonpoor households.
Current data and methods do not permit us to account for inequality within households in most
countries, so a chapter of the report examines select country studies where this accounting
is possible, and it describes how it affects the profile of poverty, including by gender and age.
The twin goals of ending extreme poverty and boosting shared prosperity will continue
to guide our work. The new suite of poverty lines and measures broadens our conception of
poverty. As this report shows, taking such an expansive view only reinforces how far we still
need to go to rid the world of poverty in all of its dimensions.
Jim Yong Kim
President
World Bank Group
xii POVERTY AND SHARED PROSPERITY 2018
Acknowledgments
This report was prepared by a team co-led by Dean Jolliffe and María Ana Lugo. The core
team included Bénédicte Leroy de la Brière, Jed Friedman, Isis Gaddis, Roy Katayama, Daniel
Gerszon Mahler, Mario Negre, David Newhouse, Minh Cong Nguyen, Espen Beer Prydz, Maika
Schmidt, Dhiraj Sharma, and Judy Yang. The extended team included Sabina Alkire, Luis Alberto
Andrés, Paola Buitrago Hernandez, Samuel Freije-Rodríguez, Xavier Godinot, Stephan Klasen,
Rahul Lahoti, Christoph Lakner, Sylvie Lambert, Valérie Lechene, Libbet Loughnan, Carolina
Mejía-Mantilla, Ana María Muñoz Boudet, Rakesh Gupta N. Ramasubbaiah, Raul Santaeulalia-
Llopis, Kenneth Simler, Sharad Tandon, Robert Walker, Alexander Wolf, and Ruoxuan Wu, all
of whom provided key inputs. Karem Nathalia Edwards de Izquierdo, Pamela Gaye Gunio, and
Estella Malayika provided overall support to the report team.
The work has been carried out under the general direction of Francisco H. G. Ferreira, Haishan
Fu, Caren Grown, and Carolina Sánchez-Páramo. The team is also grateful for guidance and
advice from Kaushik Basu, Shantayanan Devarajan, Akihiko Nishio, and Carlos Silva-Jáuregui.
Elizabeth Howton, Mark Felsenthal, and Venkat Gopalakrishnan led the communication and
messaging of the report, with inputs from Indira Chand, Paul Gallagher, Mary Donaldson Lewis,
Mikael Reventar, Victoria Smith, and Divyanshi Wadhwa. Additional support was provided by the
Media and Web & Social Media teams of External and Corporate Relations. Robert Zimmerman,
Honora Mara, and Stuart Grudgings provided editing services. Patricia Katayama, from the World
Bank’s Development Economics unit, was the acquisitions editor. The production of the report
and overview were managed by World Bank Publications, Global Corporate Solutions unit, with
Susan Graham as the production editor and Deborah Appel-Barker as the print coordinator, and
with help from Bruno Bonansea (cartography), Aziz Gokdemir, and Susan Mandel. Patricia Hord
designed the overview booklet and the report covers.
This report would not have been possible without inputs from many different people,
including data inputs from the PovcalNet and Data 4 Goals teams, in particular, Raul Andrés
Castaneda Aguilar, João Pedro Wagner De Azevedo, Shaohua Chen, José Montes, Prem
Sangraula, Nobuo Yoshida, and Qinghua Zhao. Others who helped support this report include
Edouard Al-Dahdah, Aziz Atamanov, Ciro Avitabile, Sophie Charlotte Emi Ayling, M. Abul Kalam
Azad, Leandro Ezequiel Chalela, Urmila Chatterjee, Mickey Chopra, Reno Dewina, Ritika
D’Souza, Patrick Hoang-Vu Eozenou, María Gabriela Farfán Bertrán, Deon Filmer, Tony Henri
Mathias Jany Fujs, Roberta Gatti, María Eugenia Genoni, Michele Gragnolati, Faya Hayati, Ruth
Hill, Talip Kilic, Aart Kraay, Caterina Ruggeri Laderchi, Kihoon Lee, Vasco Molini, Rose Mungai,
Rinku Murgai, Huyen Khanh Nguyen, Nga Thi Viet Nguyen, Gbemisola Oseni Siwatu, Sergio
Olivieri, Utz Pape, Husnul Rizal, Aude-Sophie Rodella, Halsey Rogers, Shwetlena Sabarwal,
xiii
Sarosh Sattar, Prem Sangraula, William Hutchins Seitz, Umar Serajuddin, Hiroki Uematsu,
Aibek Baibagysh Uulu, Malarvizhi Veerappan, Pallavi Vyas, Matthew Wai-Poi, and Alberto
Zezza. The team also benefited from discussions with the following groups within the World
Bank: Poverty and Equity Global Practice—Development Economics Working Group; Water
Supply, Sanitation, and Hygiene; and the Human Capital Project.
The team gratefully acknowledges advice from the peer reviewers: Andrea Brandolini,
José Cuesta, Jesko Hentschel, and Salman Zaidi. The team also appreciates the many helpful
comments received from Junaid Kamal Ahmad, Abdallah Al Dardari, Sabina Alkire, Kathleen
Beegle, Ted Haoquan Chu, James Foster, Caroline Heider, Ejaz Syed Ghani, Alex Gibbs,
Michele Gragnolati, Talip Kilic, Luis Felipe López-Calva, William F. Maloney, Mahmoud Mohieldin,
Samia Msadek, Martin Rama, Nagaraja Rao Harshadeep, Julie Ruel Bergeron, Elizabeth N.
Ruppert Bulmer, Sudhir Shetty, Hans Timmer, Philip Verwimp, and Dominique van de Walle. In
addition, the team gratefully acknowledges help from the many people who have commented
on various drafts of the chapters as well as from those who have provided assistance in the
preparation of this report. And, finally, this report would not have been possible without the
hard work and dedication of the thousands of enumerators and survey respondents around
the world who have graciously shared the details of their lives and the many facets of poverty.
The report is a joint project of the Development Data and Research Groups in the
Development Economics Vice Presidency and the Poverty and Equity Global Practice in the
Equitable Growth, Finance and Institutions Vice Presidency of the World Bank. Financing
from the UK government helped support analytical work on the societal and extreme poverty
measures.
xiv POVERTY AND SHARED PROSPERITY 2018
About the Team
Co-Leads of the Report
Dean Jolliffe is a lead economist in the Development Data Group at the World Bank. He is
a member of the Living Standards Measurement Study team and co-lead of the team that
works on global poverty measurement (PovcalNet). Previously, he worked in the Research
Group and the South Asia region of the World Bank. Prior to joining the World Bank, he
was a research economist with the Economic Research Service of the U.S. Department of
Agriculture, an assistant professor at Charles University Center for Economic Research and
Graduate Education in Prague, an adjunct professor at the Johns Hopkins University School
of Advanced International Studies, an adjunct professor at the Georgetown University Public
Policy Institute, and a postdoctoral fellow at the International Food Policy Research Institute.
Dean holds appointments as a research fellow with the Institute for the Study of Labor, as a
co-opted council member of the International Association for Research in Income and Wealth,
and as a fellow of the Global Labor Organization. He received his PhD in economics from
Princeton University.
María Ana Lugo is a senior economist in the Poverty and Equity Global Practice at the
World Bank. Her current work focuses on issues of poverty and well-being measurement,
multidimensional poverty, economic mobility, inequality of opportunities, and fiscal incidence
analysis, with an emphasis on Latin American countries. She is currently a council member
of the Society for the Study of Economic Inequality. Prior to joining the World Bank, she was
a postdoctoral fellow in economics at the University of Oxford, researcher at the Oxford
Poverty and Human Development Initiative, tutor at Brasenose College (Oxford), researcher
at the Universidad de General Sarmiento in Buenos Aires, and analyst at the Ministry of Social
Development in Argentina. She holds a PhD in economics from the University of Oxford, and a
bachelor’s degree in economics from the Universidad de Buenos Aires, Argentina.
Core Team
Bénédicte Leroy de la Brière is a lead economist with the Social Protection and Jobs Global
Practice at the World Bank. She previously served in the Gender Group, where she contributed
to the development of the gender strategy and coordinated analysis on women’s welfare
and work. Bénédicte’s interests include the design, implementation, and evaluation of social
assistance programs and strategies and their relationship to early childhood development,
women’s empowerment, and household economic resilience. She has worked for the Food
xv
and Agriculture Organization of the United Nations, the International Food Policy Research
Institute, the U.K. Department for International Development, and the government of Brazil,
mostly in Latin America and Africa. Bénédicte holds a PhD in agricultural and resource
economics from the University of California at Berkeley.
Mark Felsenthal is a communications officer in the Development Economics Vice Presidency
of the World Bank, where he does outreach for the Bank’s macroeconomic forecasting and
surveillance, including the flagship Global Economic Prospects report. At the Bank, he has
done communications work on financial inclusion, on operations in East Asia and Pacific, and
for the office of the managing director. Prior to joining the Bank, he covered the White House
and the Federal Reserve for Reuters and was an information officer for UNICEF. He holds an
MS from Columbia University and a BA from Middlebury College.
Jed Friedman is a senior economist in the Development Research Group (Poverty and
Inequality Team) at the World Bank. His research interests include the measurement of well-
being and poverty as well as the evaluation of health and social policies. Jed’s current work
involves investigating the effectiveness of health financing reforms in Kyrgyzstan, Zambia, and
Zimbabwe; the nutritional and development gains from early childhood investment programs in
India and the Philippines; and the incorporation of new approaches to survey-based well-being
measurement in Malawi and Peru. Jed holds a BA in philosophy from Stanford University and
a PhD in economics from the University of Michigan.
Isis Gaddis is a senior economist with the World Bank’s Gender Global Theme Department.
She previously served as a poverty economist for Tanzania, based in Dar es Salaam. Isis co-
authored the 2016 World Bank Africa Region flagship report Poverty in a Rising Africa. Her
main research interest is empirical microeconomics, with a focus on the measurement and
analysis of poverty and inequality, gender, labor, and public service delivery. She holds a PhD
in economics from the University of Göttingen, where she was a member of the development
economics research group from 2006 to 2012.
Venkat Gopalakrishnan is an online communications officer with the Poverty and Equity
Global Practice of the World Bank. He providies communications support, including message
development, dissemination strategies, social media outreach, and media engagement, for
various products and reports across the World Bank. For the past nine years, he has worked
with the World Bank on some of the major international development issues. Previously, he
worked on The Hindu, one of India’s leading English-language daily newspapers, as a senior
copy editor. He has an MBA in finance from St. Joseph’s University in Philadelphia and a
master’s in mass communication and journalism from the University of Madras, India.
Elizabeth Howton is the senior communications officer for the Poverty and Equity Global
Practice. Previously, she worked with the infoDev program, which helps start-up entrepreneurs
in developing countries grow their businesses. Before that, she was the World Bank Group’s
Global Web editor. She joined the World Bank in 2012 as an online communications officer for
the South Asia Region. Her experience prior to the World Bank includes 10 years as an editor at
the San Jose Mercury News in California’s Silicon Valley. She was a Knight Science Journalism
Fellow at the Massachusetts Institute of Technology and earned a bachelor’s degree from
Stanford University and a master’s degree from George Washington University.
xvi POVERTY AND SHARED PROSPERITY 2018
Roy Katayama is a senior economist in the Poverty and Equity Global Practice at the World
Bank. His current work focuses on the design of data collection methods suitable for fragile
situations, iterative beneficiary monitoring, enhanced digital census cartography, and statistical
capacity building in the Central African Republic. During his time at the World Bank, he has
led analytical work on poverty and inequality, poverty measurement, poverty maps, geospatial
analysis of development, welfare impact of shocks, targeting of social safety nets, and
systematic country diagnostics. He has extensive experience working in Sub-Saharan Africa
and previously served as a Peace Corps volunteer in Gabon. He holds an MPA in international
development (MPA/ID) from Harvard University.
Daniel Gerszon Mahler is a Young Professional in the Poverty and Equity Global Practice. His
research deals with the intersection between inequality, welfare measurement, and behavioral
science. Prior to joining the World Bank, Daniel was a visiting fellow in Harvard University’s
Department of Government and worked for the Danish Ministry of Foreign Affairs and the
Danish Ministry for Economic Affairs. Daniel holds a PhD in economics from the University of
Copenhagen.
Mario Negre is a senior researcher at the German Development Institute and a regular
consultant for the World Bank. From 2014 to 2016, he was a senior economist in the World
Bank’s Development Research Group (Poverty and Inequality Team). He has worked at the
European Parliament, first as an adviser to the chairman of the Development Committee and
then for all external relations committees. His fields of specialization are pro-poor growth,
inclusiveness, inequality, and poverty measurement, as well as development cooperation
policy, particularly European. Mario holds a BSc in physics from the University of Barcelona,
an MA in development policies from the University of Bremen, and a PhD in development
economics from the Jawaharlal Nehru University, India.
David Newhouse is a senior economist in the Poverty and Equity Global Practice. Since joining
the practice in 2013, he has led or contributed to the World Bank’s analysis of poverty in
India, Pakistan, and Sri Lanka; the nature of global and child poverty; and the use of satellite
imagery for poverty measurement. He was formerly a labor economist in the Social Protection
and Labor Practice, where he helped lead efforts to analyze the policy response to the 2008
financial crisis. David holds a PhD in economics from Cornell University and has published
numerous journal articles and a book on labor, poverty, health, and education in developing
countries.
Minh Cong Nguyen is a senior data scientist in the Poverty and Equity Global Practice of the
World Bank. His research interests include poverty, inequality, imputation methods, and data
systems. He currently co-leads the Europe and Central Asia Team for Statistical Development
and also co-leads the Data for Goals team. He has worked as a consultant with the Africa Region,
the South Asia Region, the Human Development Network, and the Private Sector Development
Network at the World Bank. Minh has a PhD in economics (applied microeconometrics) from
American University.
Espen Beer Prydz is an economist working on measurement of poverty and inequality with
the World Bank’s Development Data Group. He has previously worked with the World Bank
in Cambodia, Indonesia, and South Sudan on poverty, social protection, and economic policy.
ABOUT THE TEAM xvii
Prior to joining the World Bank, he worked with the Organisation for Economic Co-operation
and Development’s Development Centre and The Abdul Latif Jameel Poverty Action Lab.
Espen is a Norwegian national and holds an MPA in international development (MPA/ID) from
Harvard University and a BSc from the London School of Economics.
Maika Schmidt is a consultant in the Poverty and Equity Global Practice and Development
Research Group (Poverty and Inequality Team). Her research interests are poverty, inequality,
and early childhood development, with a focus on measurement issues, specifically multi-
dimensional indicators. Maika has worked for the Deutsche Gesellschaft für Internationale
Zusammenarbeit. She holds a BSc in economics from the University of Mannheim, a master’s
from the Barcelona Graduate School of Economics, and a master of research from Pompeu
Fabra University, Barcelona. She is currently pursuing her PhD in development economics from
the University of Sussex.
Dhiraj Sharma is an economist in the Poverty and Equity Global Practice. His work focuses on
welfare measurement, poverty diagnostics, and policy analysis. He has led or contributed to
the analysis of poverty in Ghana, Iraq, and Nepal, and he also led impact evaluations in Nepal.
His current work focuses on welfare analysis and statistical capacity building in countries in
the Middle East and North Africa region. His recent work in that region includes research on
the impact of refugee influx on host communities and the factors that shape attitudes toward
refugees. Dhiraj holds a PhD in applied economics from Ohio State University.
Judy Yang is an economist in the Poverty and Equity Global Practice. Prior to this position,
she worked for teams in the Middle East and North Africa Chief Economist’s Office, the Africa
region, and the Enterprise Surveys group. Before joining the World Bank, she worked at the
U.S. Department of Labor. Judy holds a PhD in economics from Georgetown University. Her
research interests include migration, the business environment, household welfare, and
access to finance.
xviii POVERTY AND SHARED PROSPERITY 2018
Abbreviations
CPI consumer price index
FCS fragile and conflict-affected situations
GDP gross domestic product
GDSP Global Database of Shared Prosperity
GMD Global Monitoring Database
GNI gross national income
HFCE household final consumption expenditure
IDA International Development Association
IPL international poverty line
LMIC lower-middle-income country
MDG Millennium Development Goal
MMRP Modified Mixed Reference Period
MPI Multidimensional Poverty Index
MRP Mixed Reference Period
NSS National Sample Survey (India)
OECD Organisation for Economic Co-operation and Development
OPHI Oxford Poverty and Human Development Initiative
PPP purchasing power parity
SDG Sustainable Development Goal
SP shared prosperity
SPL societal poverty line
SPP shared prosperity premium
UMIC upper-middle-income country
UNDP United Nations Development Programme
URP Uniform Reference Period
WDI World Development Indicators
WID World Inequality Database
WIR World Inequality Report
xix
Overview
The world has made remarkable and un- Back in 1990, 36 percent of the world’s
precedented progress in reducing extreme people lived in extreme poverty, defined by
poverty over the past quarter century. In the IPL as consumption (or income) less than
2015, more than a billion fewer people were US$1.90 a day in 2011 purchasing power par-
living in extreme poverty than in 1990. The ity (PPP). By 2015, that share had plunged to
progress has been driven by strong global 10 percent, down from 11.2 percent in 2013.
growth and the rising wealth of many devel- The number of people living in extreme pov-
oping countries, particularly in the world’s erty stood at 736 million in 2015, down from
most populous regions of East Asia and Pa- nearly 2 billion in 1990 (figure O.1).
cific and South Asia. This impressive progress Despite the more sluggish global growth
has brought us closer to achieving the World of recent years, the total count of people in
Bank’s target of reducing extreme poverty to extreme poverty declined by more than 68
less than 3 percent of the world’s population million people between 2013 and 2015—a
by 2030. Half of all countries included in the number roughly equivalent to the population
global poverty counts already have less than 3 of Thailand or the United Kingdom. Tens of
percent of their populations living under the millions of people have escaped poverty every
international poverty line (IPL), which de- year since 1990, reducing the global poverty
fines extreme poverty for global monitoring. rate by an average of 1 percentage point per
Despite this good news, the fight against year between 1990 and 2015.
extreme poverty is far from over—and in Much of the progress in the past quarter
some ways is getting harder. The number of century has been in East Asia and Pacific,
poor worldwide remains unacceptably high, where China’s economic rise has helped lift
and it is increasingly clear that the benefits of millions of people out of extreme poverty.
economic growth have been shared unevenly The countries of this region went from an
across regions and countries. Even as much average poverty rate of 62 percent in 1990
of the world leaves extreme poverty behind, to less than 3 percent in 2015. More recently,
poverty is becoming more entrenched and South Asia has made impressive inroads
harder to root out in certain areas, particu- against extreme poverty, helping to reduce
larly in countries burdened by violent con- the global rate further. The number of poor
flict and weak institutions. Poor households in South Asia dropped to 216 million people
are overwhelmingly located in rural areas, in 2015, compared to half a billion in 1990.
have a large number of children, and suffer These two regions have fared well on the
from a lack of education. World Bank’s other core goal—to increase
They are ill served in essential elements of shared prosperity to ensure that the rela-
well-being such as health care and sanitation, tively poor in societies are participating in
and often are exposed to natural hazards and and benefiting from economic success. This
physical insecurity. goal is measured by monitoring the aver-
1
FIGURE O.1 Global Extreme Poverty Rate and Headcount, 1990–2015
50 1,895 2,000
1,878
45 1,703 1,800
1,729
40 1,610 1,600
35.9
35 33.9 1,352 1,400
30 29.4 28.6 1,223 1,200
Poverty rate (%)
Millions
25 25.7 1,000
963
20.8 804
20 800
18.1 736
15 13.7 600
11.2
10 400
10.0
5 200
0 0
1990 1995 2000 2005 2010 2015
Number of people who live below US$1.90 a day (2011 PPP) (right axis)
Share of people who live below US$1.90 a day (2011 PPP)
Source: Most recent estimates, based on 2015 data using PovcalNet.
Note: PPP = purchasing power parity.
age consumption (or income) growth rate Whereas the average poverty rate for other
of the poorest 40 percent of the population regions was below 13 percent as of 2015, it
(the bottom 40) within each and every coun- stood at about 41 percent in Sub-Saharan Af-
try. On that score, the progress in East Asia rica. Of the world’s 28 poorest countries, 27
and Pacific and South Asia is all the more are in Sub-Saharan Africa, all with poverty
impressive because the economic growth in rates above 30 percent.
those regions is being shared. On average, the In short, extreme poverty is increasingly
consumption (or income) of the bottom 40 becoming a Sub-Saharan African problem.
in these two regions grew by 4.7 percent and Sub-Saharan African countries have struggled
2.6 percent per year, respectively, according to partly because of their high reliance on ex-
the latest estimates for 2010–15. tractive industries that have weaker ties to the
But the huge progress against poverty consumption and income levels of the poor,
in these regions contrasts sharply with the the prevalence of conflict, and their vulner-
much slower pace of poverty reduction in ability to natural disasters such as droughts.
Sub-Saharan Africa. Extreme poverty is be- Despite faster growth in some Sub-Saharan
coming more concentrated there because of African economies, such as Burkina Faso and
the region’s slower rates of growth, problems Rwanda, the region has also struggled to im-
caused by conflict and weak institutions, and prove shared prosperity. The bottom 40 in the
a lack of success in channeling growth into dozen Sub-Saharan African countries cov-
poverty reduction. Sub-Saharan Africa now ered by the indicator saw their consumption
accounts for most of the world’s poor, and— (or income) rise by an average of 1.8 percent
unlike most of the rest of the world—the per year in 2010–15 (slightly below the global
total number of poor there is increasing. The average of 1.9 percent per year). More worry-
number of people living in extreme poverty ing, however, is that the consumption (or in-
in the region has grown from an estimated come) level of the bottom 40 shrank in a third
278 million in 1990 to 413 million in 2015. of those 12 countries.
2 POVERTY AND SHARED PROSPERITY 2018
The stark contrast between Asia and Af- billions of people living above US$1.90, who
rica explains why it is getting harder to re- are still very poor by the standards of their
duce poverty globally. Although overall own societies. Now that extreme poverty
progress against poverty has been steady, continues to be high in some regions while
not all regions have shared in global growth heading down to single digits in most of the
and some are being left behind. As extreme rest of the world, we need to build a more
poverty becomes rarer, there is less scope for complete picture of what is meant by a world
gains to shift to different regions and coun- free of poverty. Certainly, the world could not
tries. With extreme poverty in East Asia and be said to be free of poverty if most countries
Pacific down to 2.3 percent in 2015, for ex- achieve the 3 percent rate while large pock-
ample, the region has little more to give in ets of extreme poverty linger. To have a better
terms of reducing the global rate. A similar understanding of what it means to end pov-
trend is well under way in South Asia. erty, we need more ways of measuring and
The result is a slowdown in overall pov- conceptualizing the problem. We need more
erty reduction that makes it unlikely the pieces of the puzzle to better understand
World Bank’s 2030 target will be met. From what a world free of poverty means.
2013 to 2015, global poverty declined by 0.6 The World Bank’s focus remains on lifting
percentage points per year, well below the people from extreme poverty, and the IPL
25-year average of a percentage point a year. will continue to be a crucial way of monitor-
Our forecasts suggest that the rate of reduc- ing this progress. But we also need to recog-
tion further slowed between 2015 and 2018 nize that societies have not stopped thinking
to less than half a percentage point per year. or caring about poverty even if it has become
Looking ahead to 2030, forecasts indicate much less apparent in its extreme forms.
that the world would need to grow at an un- There is a need to expand our understand-
usually strong pace in order to meet the 3 ing of poverty as a complex, multifaceted
percent target. For example, the target would problem and identify pockets of people who
be met if all countries grow at an average are impoverished but who have remained
annual rate of 6 percent and the consump- unnoticed.
tion (or income) of the bottom 40 grows 2 To do so, we introduce three new pieces
percentage points faster than the average. of the poverty puzzle. The addition of these
Alternatively, the landmark could be reached new ways to measure and conceptualize pov-
if all countries grow at an average pace of 8 erty follows from the recommendations of
percent. But, in either of these scenarios, ex- the Commission on Global Poverty, led by
treme poverty would still be in double digits Professor Sir A. B. Atkinson, to consider com-
in Sub-Saharan Africa by 2030. plementary indicators to the core indicator of
In an alternate scenario where all coun- extreme poverty (in Monitoring Global Pov-
tries grow in line with the average in their erty published by the World Bank in 2017).
region over the last 10 years, our forecasts The new measures recognize that people can
indicate that the global poverty rate would be defined as poor relative to their societies
be above 5 percent in 2030. This “business even at consumption levels well above the
as usual” scenario leads to a bifurcated world US$1.90 level. They also broaden our view
where more than a quarter of the people in of poverty to include elements of basic well-
Sub-Saharan Africa live in extreme poverty being such as access to sanitation and core
whereas it is less than 2 percent in most of health services. Finally, they go beyond the
the rest of the world. household level in a first attempt to measure
These contrasting regional poverty trends poverty as it affects individuals.
have two important implications. First, the These new measures will help both in
primary focus of the international commu- those countries where extreme poverty is
nity’s efforts to eliminate the worst forms currently at very low levels and in countries
of deprivation must remain firmly in Sub- where it is pervasive. Even while maintaining
Saharan Africa and those few other countries a focus on the poorest countries of the world,
elsewhere with very high poverty rates. At the with this broader view we can better un-
same time, we must not forget the plight of derstand the various dimensions of poverty
OVERVIEW 3
globally. And that better understanding can status as the country with the most poor is
provide guidance for policy and help identify ending—Nigeria either already is, or soon
areas of greatest need. will be, the country with the most poor peo-
The new measures can also help us moni- ple. The extreme poverty rate and the num-
tor progress in reducing poverty in a growing ber of poor in South Asia have been steadily
world. Even in those countries where extreme declining and are expected to continue that
deprivation rates are very low, there con- trend. The result of this trend is a shift in pov-
tinue to be significant concerns about pov- erty from South Asia to Sub-Saharan Africa.
erty more broadly defined. Having enough By 2030, the share of the extreme poor liv-
money is critical to living a life free of pov- ing in Sub-Saharan Africa could be as large as
erty, but it is not all that matters. To truly end 87 percent on the basis of historical growth
poverty, we need to better monitor people’s rates. Even if every other country in the world
progress in achieving nonmonetary aspects had zero extreme poverty by 2030, the aver-
of well-being, such as safe drinking water and age rate in Sub-Saharan Africa would have to
access to education. decrease from the 2015 rate of 41 percent to
When it comes to measuring extreme about 17 percent for the global average to be
poverty, the US$1.90 yardstick is used to as- 3 percent. That would require an unprece-
sess how well people are doing relative to the dented annual growth rate for the region.
basic needs in the world’s poorest countries. Stronger economic growth and renewed
But, for people living in countries with higher efforts to resolve violent conflicts will be cru-
overall consumption (or income) levels, there cial to speed up the rate of poverty reduction
is value in monitoring progress with higher in Sub-Saharan Africa and elsewhere. But
poverty lines that reflect the greater needs business as usual will not be enough. More
in a growing world. By using these new lines needs to be done to ensure that growth is in-
and measures in coordination with the ex- clusive, with a stronger focus on raising the
isting measure of extreme poverty—both in productive capacity of the poor.
those countries with high rates of extreme If Sub-Saharan African and other fragile
poverty and those that have nearly vanquished situations are to have a chance of reaching
extreme poverty—we can better monitor the 3 percent goal, not only will their growth
poverty in all countries, in multiple aspects rates have to be high but consumption (or
of life, and for all individuals in every house- income) levels among the bottom 40 in their
hold. This broader monitoring promises to societies will also have to rise at a higher rate.
give us a more nuanced understanding of Yet, in two-thirds of the 13 extremely poor
the nature of poverty in all its forms, so we countries (with poverty rates above 10 per-
can develop better policy tools to tackle the cent) covered by the World Bank’s shared
problem. prosperity indicator, average consumption
(or income) levels of the bottom 40 are grow-
Staying focused on the ing at a slower rate than the global average of
1.9 percent per year. That is a worrying trend
poorest for the poorest economies and conflict-af-
Ending extreme poverty will require a re- fected situations, precisely the countries least
newed focus on Sub-Saharan Africa and likely to reach the 2030 target.
states suffering from weak institutions and A second and crucial worry is that data
conflict. Estimates for 2015 indicate that needed to assess shared prosperity are weak-
India, with 176 million poor people, contin- est in the very countries that most need them
ued to have the highest number of people in to improve. Only 1 in 4 low-income countries
poverty and accounted for nearly a quarter of and 4 of the 35 recognized fragile and conflict-
the global poor. The extreme poverty rate is affected situations have data that allow us to
significantly lower in India relative to the av- monitor shared prosperity over time. Because
erage rate in Sub-Saharan Africa, but because a lack of reliable data is associated with slow
of its large population, India’s total number growth in consumption (or income) for the
of poor is still large. In a sign of change, how- poorest, the situation could even be worse
ever, forecasts for 2018 suggest that India’s than currently observed.
4 POVERTY AND SHARED PROSPERITY 2018
In the fragile situations that are covered by rity. Someone may earn more than US$1.90 a
data, the recent trend is discouraging. After day but still feel poor if lacking access to such
falling sharply between 2005 and 2011, the basic needs. Equally, someone earning less
rate of extreme poverty in these countries than that could be in even direr need without
rose to 35.9 percent in 2015 from a low of clean water to drink or a safe environment
34.4 percent in 2011. The share of the global for his or her family.
poor in these countries has risen steadily This expanded, “multidimensional” view
since 2010 to reach 23 percent in 2015. reveals a world in which poverty is a much
In many low-income countries, the bot- broader, more entrenched problem, under-
tom 40 live on less than US$1.90 a day and lining the importance of investing more in
disproportionately live in rural areas, making human capital. At the global level, the share of
them vulnerable to disruptions caused by the poor according to a multidimensional defini-
climate. Uganda, for example, has suffered tion that includes consumption, education,
significant setbacks in poverty reduction and and access to basic infrastructure is approx-
shared prosperity largely due to droughts imately 50 percent higher than when relying
and pests that affected harvests starting in solely on monetary poverty. In Sub-Saharan
2016. Uganda’s poverty rate rose from 35.9 Africa, more than in any other region, short-
percent in 2012 to 41.6 percent in 2016. Real falls in one dimension go hand in hand with
consumption for its bottom 40 shrank by 2.2 other deficiencies. Low levels of consumption
percent a year. are often accompanied by challenges in non-
As we seek to end poverty, we also need monetary dimensions.
to recognize that being poor is not defined Figure O.2 presents the share of the popula-
just by inadequate consumption or a lack of tion in Sub-Saharan Africa and South Asia that
income. Other aspects of life are critical for are considered multidimensionally deprived
well-being, including education, access to according to consumption (blue oval), edu-
basic infrastructure, health care, and secu- cation for children and adults (orange oval),
FIGURE O.2 Share of Individuals in Multidimensional Poverty, circa 2013
a. Sub-Saharan Africa b. South Asia
Basic infrastructure
2.3 Basic
infrastructure Education
Monetary 0.4
0.7 3.4
12.4 2.8
17 10.9
7
28.2 1.3
Monetary 2.9 0.2
1.4 Education
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database).
Note: The diagram shows the share of population that is multidimensionally poor, and the dimensions they are deprived in. The size of the
ovals is scaled such that they represent the respective proportions in each of the regions. For example, the numbers in the blue oval for
Sub-Saharan Africa add up to 44.9 percent, which is the monetary headcount ratio. Adding up all the numbers for Sub-Saharan Africa
results in 64.3 percent, which is the proportion of people that are multidimensionally deprived. (Numbers may not add to totals because of
rounding.)
OVERVIEW 5
and access to basic infrastructure services in- and conflict with productive activities. This
cluding drinking water, sanitation, and elec- tension is often most pronounced among
tricity (yellow oval). Almost half of the multi- the poorest countries and the poorest groups
dimensional poor in Sub-Saharan Africa (28.2 in society. For example, the average gender
percent out of a total of 64.3 percent multi- gap in poverty rates for 20–34-year-olds in
dimensionally poor) experience simultaneous Sub-Saharan Africa is 7 percentage points,
deprivations in consumption, education, and compared to a global average of 2 percentage
access to some basic infrastructure service. points (figure O.3) and virtually zero in Eu-
This proportion contrasts with other regions, rope and Central Asia.
including South Asia, in which only a quarter There is evidence from studies in sev-
of the multidimensionally poor suffer depri- eral countries that resources are not shared
vations in all three of these dimensions. The equally within poor households, especially
implication is that in Sub-Saharan Africa, the when it comes to more prized consumption
cumulative deprivations reinforce one another items. Evidence also shows complex dynam-
and make it much harder to fight poverty. ics at work within households that go beyond
To build a true picture of poverty as ex- gender and age divides. For example, a wom-
perienced by individuals, we also need to go an’s poverty status may be related to her posi-
beyond the traditional household-level mea- tion as mother versus wife of the household
sures to consider how resources are shared head.
among families. Women and children tend Another way to explore disparities within
to have disproportionately less access to re- the household is to look at how food is shared
sources and basic services, especially in the within families. In Bangladesh, for example,
poorest countries. Women in poorer coun- household survey data reveal that household
tries often withdraw from the labor force and heads—mostly men—have much smaller
lose their earning potential when they reach calorie shortfalls than individuals who are
reproductive age. The gender gap in poverty not household heads. Such differences are in-
rates is largest during the reproductive years visible in standard measures of poverty.
when care and domestic responsibilities, When we estimate individual poverty rates
which are socially assigned to women, overlap on the basis of broader consumption patterns
FIGURE O.3 Percent of Females and Males Living in Households in Extreme Poverty, by Age
Group, circa 2013
25
20
Poverty rate (%)
15
10
5
0
0–4 5–9 10–14 15–19 20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74
Age groups
Males Females
Source: Muñoz Boudet et al. 2018.
Note: Data are from 89 countries.
6 POVERTY AND SHARED PROSPERITY 2018
including nonfood goods, women fare slightly the world relative to the measure of extreme
better than men in Bangladesh. In Malawi, by poverty, which is forecast now to be in single
contrast, women have a significantly higher digits. Nearly half the world (46 percent) lives
poverty rate (73 percent) than men (49 per- on less than US$5.50 per day, a standard that
cent). Children in both countries suffer from defines poverty in a typical upper-middle-
significantly higher poverty rates. income country (table O.1). A quarter of the
We need more comprehensive data to world lives on less than US$3.20 per day.
deepen our understanding of how poverty These higher poverty lines also portray a
affects individuals and to assess how social different regional story of poverty reduction
programs can be better tailored to meet their from the US$1.90 line. The Middle East and
needs. The initial findings of this approach North Africa is a case in point. In 1990, ex-
suggest that current assistance programs risk treme poverty in the region was 6 percent,
missing many poor people who are hidden in and in 2015, it was 5 percent. This discour-
nonpoor households. aging picture of very little progress in reduc-
ing extreme poverty looks different when
Monitoring progress in a examined through the lens of the US$3.20
line. Over this same time period, the coun-
growing world tries of the Middle East and North Africa
As the world grows wealthier and extreme reduced the proportion of people living on
poverty becomes rarer, legitimate questions less than US$3.20 from 27 percent to 16 per-
arise over whether US$1.90 (2011 PPP) is too cent. Important progress in reducing poverty
low to define whether someone is poor in all in this region is hidden when one examines
countries of the world. Even as the number of only extreme poverty. The US$5.50 line, re-
extreme poor declines, many people continue flecting basic needs in upper-middle-income
to live in poverty when measured by stan- countries, presents two distressing findings:
dards that are more appropriate for a wealth- (1) almost half the world lives on less than
ier world. The success in reducing extreme US$5.50 per day, and (2) in the regions of the
poverty allows us to broaden our focus to as- Middle East and North Africa, South Asia,
sess whether such people are also benefitting and Sub-Saharan Africa, despite progress in
from economic development. reducing their poverty rates, more people
Two decades ago, 60 percent of the global were living on less than US$5.50 in 2015 than
population lived in low-income countries. in 1990 due to their growing populations.
By 2015, that had fallen to 9 percent, mean- As we seek a broader understanding of
ing that the majority of people and most of poverty, it is important to recognize that
the world’s poor now live in middle-income what constitutes a basic need can vary de-
countries. To reflect this shift and the rise in pending on a country’s level of consumption
what may constitute basic needs for many or income. In a poorer country, for example,
people, the World Bank now reports on two participating in the job market may require
higher-value poverty lines of US$3.20 and only clothing and food, whereas someone in
US$5.50 per person per day, expressed in a richer society may also need access to the
2011 PPP. The value of these lines is derived internet, transportation, and a cell phone.
from the typical poverty line in lower- and The cost of performing the same function
upper-middle-income countries, respectively, may differ across countries depending on
in the same way that the value of the IPL is their overall level of consumption or income.
derived from the typical poverty line for To assess this type of poverty, the World
some of the poorest countries in the world. Bank is introducing the societal poverty line
These higher-valued poverty lines therefore (SPL) as a complement to its existing lines.
reflect social assessments of what defines The SPL is a combination of the absolute IPL
minimum basic needs in countries at these and a poverty line that is relative to the me-
income levels. dian consumption (or income) level of each
As may be expected, these two standards country. Specifically, it is equal in value to ei-
for measuring poverty portray a less encour- ther the IPL or US$1.00 plus half of daily me-
aging picture of the level of well-being in dian consumption in the country, whichever
OVERVIEW 7
TABLE O.1 Poverty at Higher Poverty Lines, US$3.20 and US$5.50 (2011 PPP)
Poverty rate by Percentage point
region at US$3.20 1990 1999 2008 2013 2015 change, 1990–2015
East Asia and Pacific 85.3 67.1 37.4 17.5 12.5 –72.8
Europe and Central Asia 9.9a 21.1 7.5 5.7 5.4 –4.6
Latin America and the 28.3 27.0 15.7 11.4 10.8 –17.5
Caribbean
Middle East and North 26.8 21.7 16.7 14.4 16.3 –10.5
Africa
South Asia 81.7 76.0a 67.9 53.9 48.6a –33.1
Sub-Saharan Africa 74.9 78.3 72.2 67.8 66.3 –8.6
Rest of the world 0.8 0.8 0.7 0.8 0.9 0.1
World 55.1 50.6 38.2 28.8 26.3 –28.9
Poverty rate by Percentage point
region at US$5.50 1990 1999 2008 2013 2015 change, 1990–2015
East Asia and Pacific 95.2 87.0 63.6 42.4 34.9 –60.3
Europe and Central Asia 25.3a 44.5 17.1 14.1 14.0 –11.3
Latin America and the 48.6 47.0 33.3 27.2 26.4 –22.2
Caribbean
Middle East and North 58.8 54.5 46.6 42.3 42.5 –16.3
Africa
South Asia 95.3 93.1a 89.8 84.2 81.4 a –14
Sub-Saharan Africa 88.5 90.5 88.1 85.4 84.5 –4.1
Rest of the world 1.7 1.3 1.2 1.5 1.5 –0.2
World 67.0 66.8 56.5 48.7 46.0 –21.0
Source: PovcalNet (http://iresearch.worldbank.org/PovcalNet/), World Bank.
Note: PPP = purchasing power parity.
a. The estimate is based on regional population coverage of less than 40 percent. The criteria for estimating survey population coverage is
whether at least one survey used in the reference year estimate was conducted within two years of the reference year.
is greater. This means that, for the poorest of corresponds on average with how all coun-
countries, the value of the SPL will never be tries of the world define being poor.
less than the IPL. But, after a certain point When poverty is defined this way, the
as countries get richer, the value of the SPL number of people who are poor stood at 2.1
will increase as the consumption level of the billion as of 2015, almost three times more
median individual in that country increases. than those living under the US$1.90 level
This increasing value of the SPL corresponds (figure O.5). Strikingly, the number of peo-
with the fact that the value of national pov- ple identified as poor by the SPL has largely
erty lines typically increases as countries stayed the same over the last 25 years even as
grow richer. In fact, the SPL is constructed the number in extreme poverty has plunged.
in such a way that it directly corresponds to The global rate of societal poverty has fallen
the average value of national poverty lines at steadily since 1990, but still at a much slower
different levels of (median) consumption for rate than the decline of extreme poverty. In
each country of the world. Figure O.4 illus- 1990, the rate of societal poverty (45 per-
trates how the value of the societal poverty cent) was about one-fourth greater than the
line (in dark blue) runs through the middle rate of extreme poverty (36 percent). For
of the national poverty lines (in light blue) many low-income countries, societal and ex-
at different levels of median consumption in treme poverty were the same. The economic
each country. In this sense, societal poverty growth of the past quarter century means
provides a global measure of poverty that significantly fewer countries in 2015 have
8 POVERTY AND SHARED PROSPERITY 2018
FIGURE O.4 National and Societal Poverty Lines in a Growing World
20
10
Poverty line (2011 US$ PPP, per day)
5
1.9
1
1 1.9 5 10 20 40
Median consumption (or income) (2011 US$ PPP, per day)
National poverty line Societal poverty line
Source: Based on data and analysis from Jolliffe and Prydz (2016, 2017).
Note: Both axes use log scales. PPP = purchasing power parity.
FIGURE O.5 Societal Poverty, Global Estimates, 1990–2015
a. Poverty rate b. Number of poor
50 2,500
40 2,000
Poverty rate (%)
Millions
30 1,500
20 1,000
10 500
1990 1995 2000 2005 2010 2015 1990 1995 2000 2005 2010 2015
Societal poverty Extreme poverty
Source: Updated analysis from Jolliffe and Prydz (2017).
Note: The international poverty line reflects the extreme poverty rate (in panel a) and the headcount (in panel b) as assessed by the
US$1.90 per day threshold (2011 purchasing power parity). The societal poverty line provides the same information for societal poverty.
an SPL that is the same as their IPL, and the Whereas societal poverty is based on a
rate of societal poverty (28 percent) is almost poverty line that is in part relative to the me-
three times the rate of extreme poverty (10 dian consumption levels across countries, the
percent). shared prosperity measure monitored by the
OVERVIEW 9
MAP O.1 Shared Prosperity across the World, 91 Economies, circa 2010–15
Consumption or income growth among the bottom 40 percent of the population
Sources: GDSP (Global Database of Shared Prosperity) fall 2018 edition.
Note: The map shows annualized growth rates in mean household per capita consumption or income among the poorest 40 percent of the population in each country.
World Bank is similarly relative to how indi- setbacks on the measure even if several econ-
viduals are doing in each and every country. omies in the region, whose bottom 40 suf-
By assessing how the bottom 40 are doing in fered large declines linked to the 2008 finan-
each economy, the World Bank’s measure of cial crisis, are now recovering. This is the case
shared prosperity is relevant to countries of in Estonia, Latvia, and Lithuania, where cur-
all income levels. Overall, the news on shared rent levels of shared prosperity are above 6
prosperity is positive, with almost 80 percent percent a year. The mixed progress on shared
of the countries for which data are available prosperity highlights the need to renew our
showing income growth for the bottom 40 focus on inclusive growth.
(map O.1). But the progress was restrained Shared prosperity and societal poverty
by modest global growth and, despite the capture different aspects of how the relatively
overall improvement, some countries have less well-off are doing in each country. But
experienced slowdowns and even reversals in the two measures are nonetheless linked, as
shared prosperity. an example of two upper-middle-income
Latin America and the Caribbean, for ex- countries—Costa Rica and Ecuador—shows.
ample, saw less growth in shared prosperity Between 2011 and 2016, both countries’
from 2010 to 2015 than in previous years as economies grew at similar rates. But the
its economies cooled amid a downturn in bottom 40 in Ecuador did better than their
global commodity prices. Many countries counterparts in Costa Rica, growing their
in Europe and Central Asia also experienced income by a percentage point more than the
10 POVERTY AND SHARED PROSPERITY 2018
mean in the country. Costa Rica’s bottom 40 greater than those living in monetary poverty.
grew in line with their country’s mean. As a This means that the challenge in securing
result, societal poverty fell faster in Ecuador higher living standards for the population of
than in Costa Rica. South Asia is far more daunting when poverty
Our view of poverty expands again when in all its forms is considered. Although South
we define it not just as a shortage of money Asia is expected to meet the goal of reducing
but also as a lack of basic elements of well- extreme poverty to below 3 percent by 2030,
being. Many countries have made great many people will still be living in unsatisfac-
strides in reducing monetary poverty but still tory conditions if the region does not make
lag in crucial areas—such as basic infrastruc- progress on other components of well-being.
ture, education, and security—that have a The multidimensional approach high-
very real impact on people’s quality of life. In lights how the ways deprivations interact vary
the Middle East and North Africa and Latin widely from country to country. In richer re-
America and the Caribbean, despite the low gions such as Latin America and the Carib-
prevalence of monetary poverty (less than 6 bean, the Middle East and North Africa, and
percent), almost one in seven people lacks East Asia and Pacific, nonmonetary depriva-
adequate sanitation. tions are much less associated with monetary
South Asia is another case in point. De- ones than in other regions. In a sample of six
spite having made progress in poverty re- countries, the multidimensional approach
duction, the region’s shortfalls in education can be extended to include, in addition to
remain high for both adults and children and education and access to basic infrastruc-
are not strongly associated with monetary ture services, two other dimensions: health
poverty. In addition, the number of people and nutrition, and security from crime and
in the region living in households without natural disaster (figure O.6). The higher-
access to an acceptable standard of drinking income countries of Ecuador, Iraq, and Mex-
water, adequate sanitation, or electricity is far ico suffer from higher crime rates and greater
FIGURE O.6 Contribution to Multidimensional Poverty, by Dimension, Selected Countries
100
Contribution to total poverty (%)
80
60
40
20
0
3 5 3 5 3 5 3 5 3 5 3 5
Dimensions
Ecuador Indonesia Iraq Mexico Tanzania Uganda
Monetary Education Services Health Security
Sources: Calculations are based on Ecuador’s Encuesta de Condiciones de Vida 2013–14; Indonesian Family Life Survey, 2014; Iraq
Household Socio-Economic Survey, 2012; Mexican Family Life Survey, 2009–12; Tanzania’s National Panel Survey, 2012–13; Uganda
National Panel Survey 2013–14. See annex 4B for details.
Note: The figure shows the contribution of each dimension to the multidimensional poverty measure based on the dimensional
breakdown method of Alkire et al. 2015.
OVERVIEW 11
FIGURE O.7 Gender Gaps, Individual Multidimensional Poverty, Selected Countries
80
Share who are multidimensionally poor (%)
70
60
50
40
30
20
10
0
Ecuador Indonesia Iraq Mexico Tanzania
Men Women
Source: Klasen and Lahoti forthcoming.
insecurity than the lower-income countries This more nuanced picture highlights new
included in the analysis. In Indonesia, multi- pockets of poverty and can help in formu-
dimensional poverty is largely driven by poor lating policies to address them. For example,
outcomes in children’s health and nutrition. policies to expand infrastructure and social
Including additional dimensions of depri- services should take into account the differ-
vation in our measures of poverty can pro- ent needs of women, children, and men. In
vide valuable insight into how policies can be some regions, improvements in access to ed-
directed to have the most effect on poverty. ucation can particularly help women, who
The profile of the poor can change as we take continue to be held back by gender inequali-
a multidimensional view of poverty. For ex- ties in schooling.
ample, a five-dimension picture of Indonesia
shows that the country may need a stronger
focus on combatting health care depriva-
Piecing together the poverty
tions, whereas efforts in Ecuador may be bet- puzzle
ter directed toward education and security, This report provides a more complete picture
particularly in urban areas. of poverty that reinforces much of the posi-
The multidimensional approach, when tive story revealed by the tremendous prog-
combined with data at the individual level, ress in reducing extreme poverty over the
can also provide new insights into who is last quarter century. But it also uncovers pre-
poor. Applying this approach to five of the viously hidden details about the nature and
six countries reveals that poverty is greater extent of poverty throughout the world. Par-
among women than men, especially in Iraq ticularly distressing findings are that extreme
(figure O.7). Women are revealed as multi- poverty is becoming entrenched in a handful
dimensionally poorer than men in all five of countries and that the pace of poverty re-
countries, and the gender gap may be even duction will soon decelerate significantly.
wider for specific vulnerable groups. Widows, Reaching the target of reducing extreme
for example, are found to be significantly poverty to less than 3 percent by 2030 will
poorer than widowers in all countries except require a redoubling of efforts and greater
Ecuador. focus on those countries where poverty is
12 POVERTY AND SHARED PROSPERITY 2018
the worst. The work of the World Bank will get by 2030 will require more than busi-
continue to focus on monetary poverty with ness as usual: the region will need strong
respect to the IPL; however, truly bringing an and sustained economic growth, signifi-
end to global poverty requires thinking more cant improvements in the living standards
broadly and recognizing the greater complex- of the bottom 40 throughout Sub-Saharan
ity inherent in the concept of poverty around Africa at a scale not seen in recent history,
the world. and substantial investments in people.
Going forward, the World Bank will con-
2. The new measures can enhance policy
tinue its focus on reporting progress toward
dialogue. Welfare monitoring and policy
the twin goals of ending extreme poverty and
dialogue at the country level will continue
boosting shared prosperity. But, to assure that
to be based on national poverty mea-
poverty is also tracked in a relevant manner
sures. Grounded in tools that countries
in countries with very low levels of extreme
already use to monitor progress, the lines
poverty, our regular poverty updates will also
and measures introduced here open new
include progress at the two higher poverty
possibilities for countries to benchmark
lines of US$3.20 and US$5.50 and on the
their performance against relevant com-
new societal poverty line. Likewise, the next
parators using a richer set of instruments.
global poverty update in 2020 will report on This is particularly the case in middle-in-
advances in multidimensional poverty for the come countries, where extreme poverty is
countries where data are available. Between less prevalent, but where the higher pov-
global updates, these new measures will be- erty lines and the new multidimensional
come part of our biannual country reports poverty measure reveal there is still much
on poverty and shared prosperity—Poverty work to be done.
and Equity Briefs.
The use of these new measures for global 3. Data investments are critical. World
poverty monitoring and the findings of the Bank investments in data have helped
report have three important and distinct im- provide a more comprehensive pic-
plications for the work and priorities of the ture of poverty, but there is a need for
World Bank: continued and deeper investment in
data. More and better welfare data are
1. Transformational change is needed in needed to compare poverty across time,
Sub-Saharan Africa and conflict-affected for multiple dimensions, for all indi-
areas. The battle against extreme poverty viduals, and particularly among low-
will be won or lost in Sub-Saharan Africa income and conflict-affected countries.
and fragile and conflict-affected settings. Very few of these countries have shared
Global extreme poverty is increasingly prosperity estimates, and few countries
becoming a Sub-Saharan phenomenon, have data for estimating all dimensions of
and the share of the poor in fragile and poverty. Ensuring that no one is left be-
conflict-affected situations is growing. hind in the fight against extreme poverty
Of all regions, Sub-Saharan Africa has requires that we expand investments in
one of the worst performances in shared country systems and capacity to measure
prosperity and the poor there suffer from and monitor welfare in a timely, compa-
multiple deprivations more than in any rable manner using both traditional and
other region. Reaching the 3 percent tar- newer types of data and methods.
OVERVIEW 13
Introduction
The last 25 years have seen tremendous prog- erty), 27 are in Sub-Saharan Africa, all with
ress toward the goal of ending extreme pov- rates above 30 percent.
erty. The share of the global population living Second, the pace of poverty reduction has
in extreme poverty as measured by the inter- slowed in recent years. Over the 25 years from
national poverty line (IPL, currently valued 1990 to 2015, the global extreme poverty
at US$1.90 in 2011 purchasing power par- rate fell by slightly more than 25 percentage
ity dollars) fell from 35.9 percent in 1990 to points, or an average decline of 1 percentage
11.2 percent in 2013. As noted in this Poverty point a year; however, over the two years be-
and Shared Prosperity report, an additional tween 2013 and 2015, it declined by only 1.2
68 million people were lifted out of extreme points, or 0.6 percentage points a year. One
poverty between 2013 and 2015—the last of the main reasons for the slowdown is the
year for which we have globally comparable growing concentration of extreme poverty in
data—to bring the global rate to a historical Sub-Saharan Africa, where the combination
low of 10 percent. of slower than average economic growth,
However, a more careful look at these often concentrated in capital-intensive sec-
numbers, particularly in recent years, reveals tors, higher than average population growth,
two concerning and interrelated trends. First, low levels of human capital and access to
progress toward the elimination of extreme basic infrastructure, and increased levels of
poverty has been uneven. Whereas in 1990 80 fragility and conflict, has resulted in limited
percent of the extreme poor lived in East Asia progress in poverty reduction and, conse-
and Pacific or South Asia, in 2015 more than quently, the region’s growing number of peo-
half of the global poor resided in Sub-Saharan ple living in extreme poverty.
Africa. The changing regional concentration If economic growth over the next 15 years
of extreme poverty reflects the highly uneven is similar to historical growth patterns, re-
rate of poverty reduction across regions and gional disparities will only become larger over
countries of the world. Four of the six devel- time: forecasts for 2030 put the share of the
oping regions had extreme poverty rates below global extreme poor residing in Sub-Saharan
10 percent in 2015, compared to a rate of over Africa at about 87 percent and extreme pov-
40 percent for Sub-Saharan Africa. Similarly, erty rates in the double digits for many coun-
of the 164 countries for which the World Bank tries in the region. Even in a forecast where
monitors extreme poverty, more than half— countries grow at a rate of 8 percent per
84 countries—had already reached levels year, significantly above historical averages
below 3 percent as of 2015. In contrast, three- in the region and the world, the prevalence
fourths of countries in Sub-Saharan Africa of extreme poverty in Sub-Saharan Africa
had extreme poverty rates above 18 percent in would still be in double digits (13.4 percent),
2015; of the world’s 28 poorest countries (that whereas the average for the rest of the world
is, those with the highest rates of extreme pov- would be close to zero (0.4 percent).
15
Reaching the goal of reducing global ex- extreme poverty at the global level is inatten-
treme poverty to less than 3 percent by 2030 tive to how progress is distributed across the
will require that the countries of Sub-Saharan world. The shared prosperity indicator was
Africa realize historically unprecedented and built to ensure the monitoring of progress
sustained economic growth rates. But it will in all countries. Ending poverty and sharing
also require that this growth be highly inclu- in prosperity cannot happen in a satisfactory
sive, not just globally but in every country, be- way if the need for equitable and sustainable
cause a world where extreme poverty is elim- economic development is ignored in certain
inated everywhere except in one region does regions or countries.
not portray a picture of a world free of poverty. To complete this picture of what poverty
Similarly, as the world gets richer and means, we need more information. Just as
progress is made in the battle against ex- one can recognize the picture in a puzzle only
treme poverty, we must not forget that many when enough of the pieces are in place, so
around the world, and particularly in middle- too must there be more pieces of the puzzle
income countries, still live in deprivation, to better bring the state of poverty into full
unable to meet their basic needs, even if their view. A more comprehensive picture helps us
income levels are higher than the IPL. In the understand what meeting the goal requires.
early 1990s, when extreme poverty was perva- The rest of the report introduces three new
sive in most regions of the world, focusing the pieces to the poverty puzzle, broadening the
world’s attention on one core indicator served way poverty is defined and measured. To do
as a galvanizing force for coordinated action. this, the report goes beyond extreme mone-
It was not necessarily a weakness that progress tary poverty to start the process of monitor-
in this indicator could be attained through ing poverty in all its forms. The new lines and
significant improvements in some regions or measures introduced in this report allow one
countries. With the high global prevalence of to better monitor poverty in all countries, in
extreme poverty, a rapid reduction of extreme multiple aspects of life, and for all individuals
poverty was critical. And in this dimension in every household. They also reflect the first
there has been tremendous success. Now that steps taken by the World Bank in respond-
the extreme poverty rate is in single digits (as ing to recommendations from the Atkinson
indicated by the 2018 nowcast) and is becom- Commission on Global Poverty (World Bank
ing increasingly concentrated, finishing the 2017b), and present an evolving view of pov-
job will require constructing a more detailed erty and shared prosperity.
and comprehensive picture of what is meant Chapter 3 expands on the notion intro-
by a world free of poverty. duced with the shared prosperity indicator,
This report builds on the desire to con- that it is important to monitor progress in
struct a more complete picture of what it all countries. The chapter presents two new
means to live in a world free of poverty and sets of monetary poverty lines intended to
in which all prosper. A key point of the report complement the IPL of US$1.90 a day. First,
is that we must broaden our view of poverty. it presents higher poverty lines, at US$3.20
After an update on global extreme poverty and US$5.50 per day, reflecting typical na-
in chapter 1, the remaining chapters of this tional poverty thresholds in middle-income
report can be viewed as expanding our un- countries. In addition, the chapter introduces
derstanding of poverty. Chapter 2 provides a concept of societal poverty that reflects dif-
an update on shared prosperity as measured ferences in the overall level of well-being in
by growth in consumption or income of the each country. The societal poverty line is
bottom 40 percent of the population in each constructed to reflect social and economic
country for the period around 2010–15. One assessments of basic needs in each and every
important reason the concept of monitor- country. It integrates both the idea of mon-
ing shared prosperity was introduced was itoring absolute extreme poverty and the
to expand our view of how to think about more relative notion of ensuring that the less
poverty reduction and growth. Monitoring well-off in each society benefit as that soci-
16 POVERTY AND SHARED PROSPERITY 2018
ety grows. In this way it reflects both abso- equality within households, there undoubt-
lute poverty and the relative notion of shared edly are people living in poverty within
prosperity. nonpoor households, as well as nonpoor
Chapter 4 previews a new multidimen- individuals living in poor households. Chap-
sional poverty measure, which goes beyond ter 5 sheds light on this issue, with a focus
consumption or income poverty by adding on differences by sex and between children
nonmonetary dimensions into the measure. and adults. Current data and methods do
Access to education, health, electricity, water, not permit accounting for inequality within
sanitation, and physical and environmental households in most countries, so the chapter
security are critical for well-being. Because examines select country studies where this
many of these goods cannot be purchased in is possible and describes how this affects the
the market, they are typically not included in global profile of poverty.
the measure of extreme poverty. This work Pieced together, the chapters of this re-
builds on the tradition pioneered by the port provide a more comprehensive picture
United Nations Development Programme of poverty that reinforces much of the posi-
and the Oxford Poverty and Human Develop- tive story revealed by the tremendous prog-
ment Initiative with the Global Multidimen- ress in reducing extreme poverty over the last
sional Poverty Index, and complements it by quarter century. But they also uncover some
placing the monetary measure of well-being previously hidden details about the nature
alongside nonmonetary dimensions. For 119 and extent of poverty throughout the world.
countries, consumption poverty is combined Monetary poverty with respect to the IPL will
with education and access to basic utilities for continue to be the focus of the World Bank’s
circa 2013. In addition, the chapter explores, work. Alarming findings from the forecasts
for only six countries, the addition of dimen- reported in the first chapter are that extreme
sions on health and nutrition and on security poverty appears to be entrenched in a hand-
from crime and natural disaster. Extending ful of countries and that the pace of poverty
and complementing the monetary measure reduction will soon decelerate significantly.
with deprivation in other dimensions gives a The goal of ending extreme poverty as mea-
more comprehensive picture and helps better sured by the IPL itself will require a redou-
understand the interaction among the various bling of efforts and a greater focus on those
dimensions of poverty. countries where poverty is the worst. But, to
Finally, in most countries of the world, truly bring an end to poverty, we now also
poverty is measured at the household level, need to think more broadly and recognize the
implicitly assuming that everyone in a poor greater complexity inherent in the concept of
household is poor. But, because there is in- poverty around the world.
INTRODUCTION 17
Ending Extreme Poverty: 1
Progress, but Uneven
and Slowing
Chapter 1 presents the latest data on global and regional extreme poverty rates using the inter-
national poverty line of US$1.90 in 2011 purchasing power parity dollars. The chapter discusses
the trends, the geographical concentration, and the profile of extreme poverty. It also reflects
on data coverage and methodological issues and their consequences on global estimates.
Extreme poverty declined to 10 percent of the world’s population in 2015, meaning 1
person in every 10 in the world was living in extreme poverty. This rate dropped from nearly
36 percent in 1990, resulting in a world with more than a billion fewer people living in extreme
poverty. Although this progress is remarkable, 10 percent equates to 736 million people still
living in extreme poverty in 2015, and there is evidence that the pace of poverty reduction is
starting to decelerate. There remain significant challenges to reaching the goal of a world free
of poverty. Meeting the global target of reducing extreme poverty to less than 3 percent will
require substantially greater efforts.
Monitoring extreme poverty: billion fewer people lived in extreme poverty
in 2015 than in 1990. Not only are there now
A quarter century of progress
fewer poor people but, on average, the poor
The World Bank is committed to eradicating are also now less poor. In 1990, the average
poverty. The twin goals of ending extreme shortfall between what the poor consumed
poverty and promoting shared prosperity in a and the IPL was 35 percent (of the IPL). This
sustainable manner accord well with the post- shortfall shrank to an average of 31 percent in
2015 development agenda and the Sustain- 2015. The total consumption shortfall of the
able Development Goals (SDGs) to ensure poor (the sum of all consumption shortfalls
that all people can fulfill their potential in of the poor) in 2015 had shrunk to about one-
dignity and equality and in a healthy environ- third of its size from 1990. (For more details
ment (box 1.1). Monitoring global poverty is on the consumption shortfall of the poor, and
critical for tracking progress and identifying the depth and severity of poverty, see annex
areas that require additional policy actions. 1A.) Despite this impressive progress in terms
In 2015, an estimated 736 million people of the declining poverty rate, the number of
were living below the international poverty poor, and the consumption shortfall of the
line (IPL), currently set at US$1.90 in 2011 poor, the number of people living in extreme
purchasing power parity (PPP) dollars. This poverty globally remains unacceptably high.
count of people living in extreme poverty is The World Bank has set a specific target to
down from 1.9 billion people in 1990. Despite help guide the work in eradicating poverty:
the world population increasing by more than reduce the global share of people living in
2 billion people over this period, more than a extreme poverty to less than 3 percent. Over
19
BOX 1.1 Alignment of the SDGs and the Twin Goals of the World Bank Group
On April 20, 2013, the Board of Executive building on the Millenium Development
Directors of the World Bank adopted Goals (MDGs). Ending poverty in all its
two ambitious goals: ending extreme forms and dimensions is the first of the 17
poverty globally and promoting shared SDGs. The General Assembly Resolution
prosperity in every country in a sustainable recognizes that eradicating poverty is
way. Progress toward the first of these the greatest global challenge and an
goals is measured by monitoring the share indispensable requirement for sustainable
of the global population living below the development.
international poverty line. The World Bank The SDGs and the World Bank’s twin
set a target of reducing extreme poverty to goals are aligned. The goals of ending
less than 3 percent by 2030 and to ensure extreme poverty within a generation
continued focus and steady progress and promoting shared prosperity in a
toward the goal, the institution set an sustainable manner accord with the 2030
interim target of 9 percent by 2020. Agenda for Sustainable Development to
The second goal is not defined ensure that all human beings can fulfill
globally, but rather tracks progress at the their potential in dignity and equality and
country level. Progress on the shared in a healthy environment. In contrast to
prosperity goal is measured by the growth the SDGs, the World Bank’s twin goals do
in the average consumption or income not set distinct country-specific targets
expenditure of the poorest 40 percent or targets for the multiple dimensions
of the population (the bottom 40) in a of poverty, equity, and sustainability.
country. This goal is not associated with a However, the World Bank recognizes
target in 2030, but it reflects the aim that that poverty is multidimensional, and
every country should promote the welfare sustainability is critical. The pursuit of
of its least privileged citizens for a more these goals will require the concerted
inclusive and equitable society. effort of all stakeholders. Over the years,
On September 25, 2015, the United the World Bank has collaborated with the
Nations General Assembly adopted the 17 United Nations in nearly every region and
Sustainable Development Goals (SDGs) sector, and its engagement has deepened
and 169 targets as part of the 2030 since the adoption of the MDGs, and now
Agenda for Sustainable Development, with the SDGs.
the last decades, remarkable progress has this trend of steady poverty reduction, the
been made in reducing extreme poverty world is clearly on track to reach the interim
(figure 1.1; see box 1.2 for details on the data poverty target of 9 percent by 2020 set by the
used). The world attained the first Millen- World Bank to monitor progress toward the
nium Development Goal—to cut the 1990 2030 goal.1 Forecasts for 2018 indicate that
poverty rate in half by 2015—six years ahead this target has already been surpassed.
of schedule. With continued reductions, the Reducing extreme poverty to 3 percent by
global poverty rate—the share of the world’s 2030 from 10 percent in 2015 will require an
population living below the IPL—dropped additional 7-percentage-point reduction in
from about 36 percent in 1990 to 10 per- the poverty rate in 15 years. If, over the last 25
cent in 2015, that is, more than a 70 percent years, poverty has steadily declined at 1 per-
reduction. centage point a year, it would seem reasonable
Over the 25 years from 1990 to 2015, the to assume that the world is well on track to re-
global rate of extreme poverty fell by slightly ducing poverty by at least 7 percentage points
more than 25 percentage points, or an average over the next 15 years. The rate of poverty re-
decline of 1 percentage point a year. (Gauged duction could be cut in half to a 1-percentage-
according to today’s population, 1 percent point decline every two years, and the world
equates to about 76 million people.) Given would still reach the 3 percent target.
20 POVERTY AND SHARED PROSPERITY 2018
FIGURE 1.1 Global Extreme Poverty Rate and Headcount, 1990–2015
50 1,895 2,000
1,878
45 1,703 1,800
1,729
40 1,610 1,600
35.9
35 33.9 1,352 1,400
30 29.4 28.6 1,223 1,200
Poverty rate (%)
Millions
25 25.7 1,000
963
20.8 804
20 800
18.1 736
15 13.7 600
11.2
10 400
10.0
5 200
0 0
1990 1995 2000 2005 2010 2015
Number of people who live below $1.90 a day (2011 PPP) (right axis)
Share of people who live below $1.90 a day (2011 PPP)
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: PPP = purchasing power parity.
BOX 1.2 Chapter 1: Data Overview
Data source than consumption has been increasing
The data for this chapter come from over time. The differences between
PovcalNet, which is an online analysis income and consumption measures
tool for global poverty monitoring hosted matter for comparing trends and patterns
by the World Bank (http://iresearch. in poverty. To assure that the income
worldbank.org/PovcalNet). PovcalNet and consumption levels from different
was developed with the purpose of countries are comparable, they need to be
public replication of the World Bank’s expressed in the same unit. To this end,
poverty measures for the IPL. PovcalNet consumer price indexes and purchasing
contains poverty estimates from more power parities are applied. Because
than 1,600 household surveys spanning the frequency and timing of household
164 countries.a Most of the surveys in surveys vary across countries, comparable
PovcalNet are harmonized through the country-level estimates require projecting
Global Monitoring Database, the World the survey data to the reference year
Bank’s repository of household surveys. for which global poverty is expressed,
here 2015. When the timing of surveys
Derivation of country-level estimates does not align with the reference year,
The national poverty rates from household PovcalNet “lines up” the survey estimates
surveys are based on measures of to the reference year.
household consumption or income. In
the current 2015 estimates, about 40 Derivation of regional/global estimates
percent of the countries covered use To arrive at a regional and global
income, but the use of income rather estimate of poverty, population-weighted
(continued)
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 21
BOX 1.2 Chapter 1: Data Overview (continued)
average poverty rates are calculated estimate of the number of poor is the
for each region.b Some countries have product of the population-weighted mean
no household survey data to monitor of the regional poverty rates and the total
poverty. No direct value is imputed for world population.
these countries; rather it is assumed that
the average for the region based on the Further information
countries with data available is the same For further information regarding the data
as the regional average for all countries. sources, geographical regions, data issues,
The number of poor in each region is the and assumptions underlying the global,
product of the region’s poverty rate and regional, and country-level estimates, see
the total regional population. The global appendix A at the end of the report.
a. The term country, used interchangeably with economy, does not imply political independence
but refers to any territory for which authorities report separate social or economic statistics.
b. Population estimates are usually based on national population censuses. Estimates for the years
before and after the census are interpolations or extrapolations based on demographic models
(Source: World Development Indicators).
Despite this optimistic portrait of the which is three years out of date. Why in 2018
path toward the target, there are reasons for is poverty reported for 2015? The global esti-
concern. One reason is the existence of some mates are based on household surveys from
evidence that the rate of poverty reduction 164 countries, and these surveys are carried
has recently slowed. Between 2011 and 2013, out independently, typically by national sta-
extreme poverty declined by 2.5 percentage tistical offices or national planning ministries.
points, but, over the two years between 2013 The surveys are complex and lengthy, requir-
and 2015, it declined by only 1.2 points. Al- ing significant amounts of labor and time
though this change in the rate of poverty to be implemented effectively; and, in most
reduction over these two years should be in- countries, they are not carried out every year.
terpreted with caution because of data chal- Countries implement household surveys that
lenges, it is a first potential signal of change. measure poverty status once every three to
To assess whether this recent change in the five years (Serajuddin et al. 2015). It also takes
path of poverty reduction is an aberration or time to gather, process, and analyze these
a warning sign of what the future holds, fore- data. There is thus frequently a lag between
casts of how extreme poverty may evolve up the completion of the survey fieldwork and
to 2030 can be very informative. Such fore- the publication of the data for the global pov-
casts should be viewed with caution though, erty counts (Independent Evaluation Group
because the factors that affect global poverty 2015). For these reasons, 2015 is the most re-
reduction are complex, and because the fu- cent year for which there are sufficient data to
ture is uncertain. For example, economic estimate a global poverty rate.2 (For details on
growth is a key factor in reducing poverty, how data are shifted forward and backward in
but it can be volatile and difficult to predict. time to produce the 2015 estimate, see appen-
Nonetheless, without forecasts, it is not pos- dix A at the end of the report.)
sible to clarify whether the current trajectory However, if assumptions are made about
is adequate to reach the target. the relationship between economic growth
as observed in national accounts (such as
Nowcasts and forecasts to the real growth in gross domestic product
[GDP]) and in surveys, as well as on popula-
2030 tion projections, it is possible to nowcast the
The current estimate of the global extreme global poverty rate in 2018 and also generate
poverty rate—10 percent—refers to 2015, scenarios about global poverty in 2030.3 To
22 POVERTY AND SHARED PROSPERITY 2018
nowcast poverty in 2018, it is assumed that provides another piece of evidence that there
each household’s welfare grows at a fraction seems to be a significant slowdown in the
of the growth in GDP per capita. Only a frac- rate of global poverty reduction. From 2013
tion of the growth in GDP per capita is passed to 2015, poverty declined by 0.6 percentage
through to the welfare vector because there points per year; this is slower than the 25-year
is a historical divergence between growth average decline of a percentage point per year.
in consumption or income observed in sur- Between 2015 and 2018 the nowcast suggests
veys and the growth observed in national ac- that the rate of poverty reduction has further
counts. The fraction that is passed through to slowed to less than half a point per year.
the welfare vector is based on examining past Projecting global poverty to 2030 is more
data on the average relationship between sur- challenging, but it is possible to consider how
vey means and national accounts data (Prydz, global poverty may evolve under different
Jolliffe, and Serajuddin, forthcoming).4 With scenarios. Four scenarios are considered as
this approach, it is assumed that the scaled described below. The first scenario assumes
growth accrued equally (in proportionate that every country grows at its average growth
terms) to everyone in a country regardless of rate from 2005–15. This growth rate is then
individual income level. If inequality changed used to “grow” the household survey mean
from 2015 to 2018, this assumption will not over time, in a way that does not change the
hold, and poverty will be higher or lower de- level of inequality. This approach makes it pos-
pending on the change in inequality (World sible to move the entire distribution of con-
Bank 2016b; Lakner, Negre, and Prydz 2014). sumption or income forward in time, starting
Under these assumptions, the 2018 nowcast with the 2018 nowcast and moving up to 2030.
for the global extreme poverty rate is 8.6 per- The second scenario is like the first, ex-
cent (figure 1.2). This means that the 2020 in- cept for one difference: the growth rate for
terim target has likely already been achieved. each country is not its historical average, but
One implication of this estimate is that it rather the historical average for its region.
FIGURE 1.2 Projections to 2030 of Global Extreme Poverty
14
12
10
8.6
Poverty rate (%)
8
Growth assumptions
6 Historical country growth
Historical regional growth
4 2 x historical regional growth
3% target
6% annual growth + 2 pp premium
2
0
2012 2015 2018 2021 2024 2027 2030
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/; World Development
Indicators; World Economic Outlook; Global Economic Prospects; Economist Intelligence Unit.
Note: The 2018 nowcast uses realized and projected growth in GDP per capita and household final consumption expenditure per capita
from 2015 to 2018 to grow the 2015 welfare vector. “Historical country (regional) growth” assumes that the annual growth rates countries
(regions) experienced from 2005 to 2015 continue from 2018 to 2030. “6% annual growth + 2 pp premium” assumes that all countries
grow by 6 percent annually from 2018 to 2030, and that the bottom 40 percent on average grow with an additional 2 percentage points
(pp). All assumed growth rates are real, per capita growth.
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 23
For each region, the average annualized real 10-year historic average growth rate (based
growth rate between 2005 and 2015 is esti- on growth from 2000 to 2010) was almost 4
mated and then used as the growth rate for percent, but this was sustained for only a few
each country in the region. The third sce- years and has since declined slightly.
nario is identical to the second but uses twice How can it be that poverty has declined
the historical regional growth averages. These by 25 percentage points over the last 25 years,
three scenarios all assume that inequality in yet the only forecasts that suggest poverty
the country remains unchanged until 2030. will be reduced by 7 percentage points over
The final scenario explores what happens the next 15 years are based on unprecedented
if growth is pro-poor; if the bottom 40 per- growth patterns and rates?
cent on average grows faster than the coun-
try as a whole. This scenario, not anchored to Uneven progress: A regional
any empirical data, assumes that each coun-
try grows by 6 percent annually toward 2030,
profile of poverty reduction
but that the bottom 40 percent, on average, There are several parts to the answer of this
grows by 8 percent annually (while the top question and many of them hinge on the gen-
60 percent grows at 4.7 percent, resulting in eral idea that progress has been uneven, which
the average of 6 percent). Because the bottom is linked to the theme of this report. A slightly
40 percent grows at a rate that is 2 percentage more specific answer to the question posed
points faster than the average, this is referred above is that not all regions have shared in the
to as a shared prosperity premium of 2 per- benefits of the global reduction in poverty.
centage points. In all these scenarios, growth To better understand why the simulations
rates in either GDP per capita or household forecast a challenging path for reaching the
final consumption expenditure (HFCE) per target, it is useful to examine the changing
capita are rescaled to account for the differ- regional profile of poverty that has been
ence between survey means and national ac- brought about by the differing rates of pov-
counts as discussed above.5 erty reduction. Between 1990 and 2015, the
The scenarios based on growth rates that regional profile of poverty has changed sig-
correspond with historical performance of nificantly. In 2015, more than half of the
the countries, or of the average performance global poor resided in Sub-Saharan Africa
of the region do not come close to reaching and more than 85 percent of the poor re-
the target (figure 1.2). Both scenarios suggest sided in either Sub-Saharan Africa or South
global poverty rates in the range of 6 per- Asia (figure 1.3). The remaining 14 percent
cent in 2030. The third scenario, where it is of the global poor, or about 106 million poor
assumed that all countries grow by twice the people, lived in the other four regions or in
average regional growth rate over the past ten high-income economies.7
years, also falls short of the 3 percent target. This is a dramatic shift from 1990, when
This scenario predicts a global extreme pov- over half of the poor were living in East Asia
erty rate of 3.7 percent in 2030. and Pacific. The two regions with the most
This is an alarming finding. poor people in 1990 were East Asia and Pa-
The only scenario where the 3 percent tar- cific and South Asia, which were home to 80
get is met is when a real annual growth rate of percent of the poor. With China’s rapid re-
6 percent and a shared prosperity premium duction of poverty, the concentration of the
of 2 percentage points are assumed.6 The global poor shifted from East Asia and Pa-
most important element of this scenario is cific in the 1990s to South Asia in 2002, and
that Sub-Saharan Africa is assumed to grow then to Sub-Saharan Africa in 2010. In South
steadily at this rate for 12 straight years up Asia, both the poverty rate and number of
through 2030. In considering this scenario, it poor have been steadily declining, but, given
is useful to note that between 2000 and 2015 the sheer size of the populations, the con-
Sub-Saharan Africa has never had a 10-year tribution to global poverty continues to be
average growth rate near 6 percent—let alone high. This contrasts with Sub-Saharan Africa,
8 percent for the bottom 40. The highest av- where the total count of poor people in this
erage growth rate was around 2010, when its region has been increasing, essentially lead-
24 POVERTY AND SHARED PROSPERITY 2018
FIGURE 1.3 Number of Extreme Poor by Region, 1990–2030
2,000
1,800
1,600 World
1,400
1,200
Millions of poor
1,000
800
600
400
200
0
1990 1995 2000 2005 2010 2015 2020 2025 2030
Sub-Saharan Africa Latin America and the Caribbean Rest of the world
South Asia Middle East and North Africa
East Asia and Pacific Europe and Central Asia
Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet/. World Bank, Washington, DC, World Development
Indicators; World Economic Outlook; Global Economic Prospects; Economist Intelligence Unit.
ing to the shifting concentration of poverty FIGURE 1.4 Regional GDP per Capita Growth and Average Growth for
from South Asia to Sub-Saharan Africa. the Extreme Poor, 1990–2017
This pattern is likely to continue in the
coming decade. Simulations show that, as the 10
number of extreme poor continues to decline
in South Asia, the forecasts based on histor-
ical regional performance indicate that there
will be no matching decline in poverty in Sub-
Annual growth (%)
5
Saharan Africa (figure 1.3). In 2030, the share
of the global poor residing in Sub-Saharan
Africa is forecasted to be about 87 percent, if
economic growth over the next 12 years is sim-
ilar to historical growth patterns. (For more 0
details on the simulations, see annex 1B.)
One important reason for the changing
regional concentration of extreme poverty,
and the projected increase in the share of the –5
global poor residing in Sub-Saharan Africa,
1990 1995 2000 2005 2010 2015
is the regional differences in per capita GDP
growth. Focusing on the three regions that East Asia and Pacific (population-weighted growth)
have accounted for the bulk of the poor, the Sub-Saharan Africa (population-weighted growth)
average annual growth rate since 1990 has South Asia (population-weighted growth)
consistently been highest in the East Asia and Global average growth for the extreme poor
Pacific region (between 5 and 10 percent), fol- Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet/. World Bank,
lowed by South Asia, and then Sub-Saharan Washington, DC, World Development Indicators.
Africa. South Asia has maintained an average Note: The orange line reflects the average growth rate as experienced by the population of people in
extreme poverty. It is a weighted average of country growth rates where the weights are the number
growth rate between 5 and 6 percent over the of extreme poor in each country. All curves fit a local polynomial through the annual growth rates to
last decade (figure 1.4). The average growth smooth out year-to-year fluctuations.
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 25
rate in Sub-Saharan Africa has rarely exceeded ing the global poverty count will occur only
5 percent and has decreased in recent years. if progress is primarily made in those coun-
Growth is an important driver of poverty tries where poverty is greatest. This is not
reduction, and, throughout the 1990s and to say that countries with extreme poverty
early 2000s, the vast majority of the poor lived rates below 3 percent cannot make further
in countries with relatively high growth rates. progress. Where there is poverty, there is still
Over the last few years, as the concentration much work to be done. But the core indicator
of poverty has shifted to Sub-Saharan Africa, the World Bank will track up through 2030 is
the majority of the poor now live in countries to reduce the global rate of extreme poverty
with lower-than-average growth rates (figure to less than 3 percent.
1.4). The orange line in figure 1.4 reflects this If the goal is a world free of poverty, why
change because it is a weighted average of is progress monitored toward 3 percent and
country growth rates where the weights are not zero percent? The 3 percent target comes
the number of extreme poor in each coun- from both empirical and conceptual consid-
try. As the concentration of poor moved from erations. Empirically, poverty in some coun-
high-growth to low-growth countries, this tries remains deep, entrenched, and wide-
shift led to a significant deceleration in the spread; and, when the target was initially set, 3
rate at which poverty has been declining. percent was considered an ambitious but fea-
Not only has the growth rate in the coun- sible target (Jolliffe et al. 2015). Conceptually,
tries with the most poor declined in recent however, there is also an important reason for
years but the conversion of growth to poverty setting the target at some level greater than
reduction—the growth elasticity of poverty— zero percent. The purpose of a target is to
has also historically been lower in Sub-Saha- assist in efforts to attain goals. For targets to
ran Africa. Hence, a given growth rate buys help, they need to be credibly measured and
less poverty reduction in Sub-Saharan Africa monitored. The key conceptual concern then
than in most other regions of the world. is that, in general, sample surveys from large
The changing regional concentration of populations cannot measure rare outcomes
extreme poverty reflects the highly uneven well. As countries make progress toward elim-
rate of poverty reduction across countries inating extreme poverty, the accuracy with
of the world. Of the 164 countries for which which samples can measure the increasingly
the World Bank monitors poverty, more than lower rates deteriorates. In particular, sample
half—84 countries—have already reached surveys cannot reliably measure the complete
rates below 3 percent as of 2015. The median eradication of a phenomenon in a popula-
poverty rate of the 164 countries in 2015 is 2.7 tion. In part for this reason, progress is moni-
percent; this median in 2018 is estimated to be tored toward 3 percent, which can be credibly
1.9 percent. This success in having more than measured and is also an ambitious goal.
half the countries of the world with poverty Map 1.1 shows the countries that have
rates below 3 percent is also part of the reason extreme poverty rates in 2015 of less than
why the world is now starting to experience 3 percent and highlights the countries that
a slowdown in the rate of poverty reduction. have reached the interim 9 percent target set
There are now fewer countries than before for 2020. In addition to the 84 countries with
with large populations of poor people. Pre- poverty rates less than 3 percent, there are 23
viously, progress in poverty reduction could countries with poverty rates less than 9 per-
shift over time from one country or region to cent. Two-thirds of the countries have rates
another, but now there is less scope for this. less than 9 percent. Of the remaining one-
The slowdown that is observed at the global third, though, the story is different. In about
level does not mean that poverty reduction is half of these countries, the poverty rate is
declining in every country; however, it does greater than 30 percent; and, in 11 countries,
mean that the number of countries where the poverty rate is greater than 50 percent.
there have been significant declines in the The impressive progress in terms of reducing
number of poor people is shrinking. global poverty to 10 percent masks signifi-
As extreme poverty becomes increasingly cant variation in success at the country level
concentrated, significant progress in reduc- in reducing extreme poverty.
26 POVERTY AND SHARED PROSPERITY 2018
MAP 1.1 Extreme Poverty Rate by Country, 2015
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Map 1.1 also marks countries with ex- 41 percent live below the IPL (figure 1.5).
treme poverty rates between 9 and 18 percent It hasn’t always been like this. In 1990, the
in 2015. This subsample has been created average poverty rate in countries from the
using the simplistic assumption that these East Asia and Pacific region was higher; but,
countries, if they succeed in reducing poverty whereas the rates in these countries quickly
by 1 percentage point a year, will have pov- declined over the years, the decline in the
erty rates less than 3 percent by 2030. There poverty rate in Sub-Saharan Africa was much
are 121 countries with rates at or below 18 slower (figure 1.6). Although the percentage
percent in 2015, and only 43 countries have of poor in Sub-Saharan Africa has slowly de-
extreme poverty rates that are higher than clined, this decline has not been fast enough
this. A closer examination of these countries to counter a growing population—the total
provides more evidence as to why the 2030 population of poor people there has steadily
forecasts indicate that attaining the 3 percent increased from 1990 to 2015 (table 1A.1 in
target will be a hard battle. annex 1A).8 Economic growth and pro-poor
The map reveals that most of the 43 coun- policies in Sub-Saharan Africa over the last
tries with poverty rates above 18 percent 25 years have had anemic effects on reducing
are in Sub-Saharan Africa. Three-fourths poverty. For simulations that use historical
of Sub-Saharan African countries had pov- average growth rates as estimates for future
erty rates above 18 percent in 2015, and, growth, the predicted future path of pov-
of the world’s 28 poorest countries (that is, erty reduction in Sub-Saharan Africa is in-
those with the highest rates of poverty), 27 adequate to bring global extreme poverty to
are in Sub-Saharan Africa, all with poverty below 3 percent.
rates above 30 percent. In 11 countries, all in Although extreme poverty is compara-
Sub-Saharan Africa, more than half the pop- tively much lower in the Middle East and
ulation live in extreme poverty (figure 1.5). North Africa, the rate increased to 5.0 percent
In all regions except for Sub-Saharan Af- in 2015, up from 2.6 percent in 2013, while
rica, the regional average is well below 18 per- the number of poor almost doubled from
cent, whereas in Sub-Saharan Africa about 9.5 million in 2013 to 18.6 million in 2015.
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 27
FIGURE 1.5 Extreme Poverty Rate by Region and Country, 2015
77
Sub-Saharan 70
Africa
65
(41%) 72 78
62
73
55
75
52
58
50
31
East Asia
and Pacific
15 (2%)
South Asiaa
(12%)
Rest of
the world
(1%) 2 14
Europe and
Central Asia
(1%)
Middle East
and North Africa
(5%)
Latin America
and the Caribbean 28
41 (4%)
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: Population-weighted regional average shown in parentheses. Each spike represents a country and all countries within a region are
the same color. Within each region, spikes are numbered with the poverty rate if they have the highest rate within the region or if their
poverty rate is greater than 50 percent.
a. This estimate is based on a regional population coverage less than 40 percent. The criterion for estimating survey population coverage
is whether at least one survey used in the reference year estimate was conducted within two years of the reference year.
These recent estimates should be interpreted ravel, the risks of falling back into economic
with caution because of underlying data deprivation must be managed efficiently and
challenges, but they are, nonetheless, a stark collectively (World Bank 2013). If not, the
reminder that past gains cannot be taken for risks can turn into economic, environmen-
granted. To ensure that progress does not un- tal and political crises, as in the Middle East
28 POVERTY AND SHARED PROSPERITY 2018
and North Africa, where fragility and conflict FIGURE 1.6 Extreme Poverty, Regional and World Trends, 1990–2015
in the region are impacting livelihoods and
manifesting in the recent spike in poverty. East Asia and Pacific
60
Sub-Saharan Africa
Drilling down: The countries 50
with the most poor South Asia
Over time, many of the countries with
Poverty rate (%)
40
high poverty numbers, including Bangla- World
desh, India, Indonesia, Kenya, and Nigeria,
30
have grown their economies out of low-
income-country status and are now middle-
income countries. With this growth, most 20
of the extreme poor have also moved from Latin America and
the Caribbean
being in low-income to being in middle- 10
income countries, and nearly two-thirds of Middle East and North Africa
the world’s poor people now reside in Europe and Central Asia
0
middle-income countries (figure 1.7). How- 1990 1995 2000 2005 2010 2015
ever, as more countries shift from low- to
middle-income status, so does the popula- Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet/. World Bank,
Washington, DC,
tion share. As of 2015, 5.5 billion people lived Note: The regional estimates for Europe and Central Asia in 1990 and South Asia in 1999 and 2015
in middle-income countries as opposed to are based on regional population coverage of less than 40 percent. The criterion for estimating survey
about 640 million in low-income countries, population coverage is whether at least one survey used in the reference year estimate was conducted
within two years of the reference year. Because of the low coverage, these numbers are censored in
explaining why most of the extreme poor— PovcalNet.
over 400 million—now reside in lower-
middle-income countries. As countries de-
velop and per capita GDP increases, poverty FIGURE 1.7 Rate and Headcount of Extreme Poor, by Income Group, 2015
rates tend to fall as economic opportunities
45 450
are expanded. This general trend can be seen
in figure 1.7, with the poverty rate declin- 40 400
ing from 42 percent for low-income coun- 35 350
tries to 14 percent for lower-middle-income
countries, and close to 2 percent for upper- 30 300
Poverty rate (%)
Millions of poor
middle-income countries. This situation is 25 250
promising for continued poverty reduction if
20 200
more poor people can benefit from economic
growth. Conversely, nearly every low-income 15 150
country is in Sub-Saharan Africa (and a few 100
10
countries in other regions, namely Afghan-
istan, Haiti, the Democratic People’s Re- 5 50
public of Korea, and Nepal according to the 0 0
fiscal year 2018 classification), highlighting Low income Lower-middle income Upper-middle income
the need to stimulate and sustain economic Population-weighted poverty rate
growth in low-income countries. Number of poor (right axis)
Drilling down a bit further into the coun-
Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet/. World Bank,
tries that have the largest population of poor Washington, DC,
people, figure 1.8 represents all countries by
the share of the global poor in 2015. Half of
the people living in extreme poverty in 2015 and Nigeria) are the five topping the list of
can be found in just five countries. The most countries with the greatest number of ex-
populous countries in South Asia (Bangla- treme poor. India, with over 170 million
desh and India) and Sub-Saharan Africa poor people in 2015, has the highest num-
(Democratic Republic of Congo, Ethiopia, ber of poor people and accounts for nearly a
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 29
FIGURE 1.8 Global Distribution of the Extreme Poor by Region and Country, 2015
eria
Nig
In
do
ne .
s ia ep
.R
Dem
g o,
Ea d P
n
Co
an
s t ac
Ban
As ific
gla
ia
des
h
pia
Et hio
Tanzania
Sub-Sa
hara
Afric a n Madagas
h As ia ca r
Sout
Keny
a
Mo
ia zamb
I nd iq ue
Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet/. World Bank, Washington, DC.
Note: The inner circle is divided proportionally to each region’s share of the total population living in extreme poverty. The outer circle is
similarly proportionate, but at the country level. The 10 countries with the most extreme poor in the world are listed.
quarter of global poverty. In the South Asia the verge of switching). But the uncertainty
region, four out of five extreme poor reside in about when they have switched or will switch
India. Despite a poverty rate of 13.4 percent, also reflects a series of difficult measurement
India’s large population of 1.3 billion results issues related to global poverty counts. Dis-
in a high number of extreme poor. To achieve cussing some of these issues is useful because
the global poverty goal, progress in poverty it can help convey a sense of the level of
reduction needs to continue in India. (im)precision of the poverty counts, and it
India’s placement as the country with the allows for transparency in the strengths and
most poor people in the world is likely to weaknesses of the data and methods.
change in the near future. In fact, projections In the case of Nigeria, there is one key con-
indicate that Nigeria may already have over- cern with current poverty estimates. Both the
taken India. The uncertainty about whether 2015 estimate and the 2018 nowcast for Nige-
India or Nigeria is currently the country with ria are based on household survey data col-
the most poor people is in part simply be- lected in 2009. To estimate extreme poverty
cause the countries are near a crossing point in 2015 for Nigeria, the survey mean from the
(having either recently switched or being on 2009 data was increased at a rate equal to the
30 POVERTY AND SHARED PROSPERITY 2018
country’s GDP per capita growth rate (which and how one asks has a significant effect on
is estimated annually) and it is assumed that how people respond (Backiny-Yetna, Steele,
the level of inequality was unchanged over and Djima 2017; Beegle et al. 2012; Gibson,
those six years. Similarly, for 2018, the mean Huang, and Rozelle 2003; Jolliffe 2001). Over
is shifted forward on the basis of nine years of the years, changes have been introduced in
growth estimates and assuming unchanged the recall period in the NSS Consumer Ex-
inequality. Although growth measured in penditure Survey, the official instrument for
surveys used for poverty estimation is cor- estimating poverty in India. The extreme
related with growth as measured by national poverty rate for India as reported here is cur-
accounts data such as GDP, there can be size- rently based on an old questionnaire design.
able differences and these differences can With the next NSS data that will be made
have substantial impact on estimated poverty publicly available, it will no longer be possi-
rates. Similarly, if the assumption that the ble to estimate consumption using the same
distribution (or inequality) has not changed questions and the extreme poverty measure
since 2009 is wrong, this too can lead to sub- will be estimated using a new questionnaire
stantial error in the estimated poverty rate design. The 2018 nowcast estimates for India
(Jolliffe et al. 2015). indicate that switching from the old to the
There are two important measurement is- new questionnaire results in a significantly
sues that also temper confidence in the India higher level of total consumption that re-
poverty estimates. The first is similar to the classifies more than 50 million people from
issue for Nigeria. The last round of poverty poor to not poor. Whenever the next round
data available was collected in 2011–12. For of NSS data is released (using the new ques-
India, however, an additional round of the tionnaire), backcasted estimates of poverty in
National Sample Survey (NSS), collected in 2015 will most likely show significantly fewer
2014–15, has the same socioeconomic and people living in extreme poverty (figure 1.9).
demographic information as the 2011–12 For more details on these measurement is-
round, and both provide data on household sues for India, see box 1.3.
expenditures on services and durables. The
2014–15 NSS also contains three additional
FIGURE 1.9 Projections to 2030 for the Five Countries with the Most
schedules with consumption data that were
Extreme Poor in 2015
designed to test the questionnaire design,
but these data are not in the public domain 250
and were not available for analysis. Given
the importance of India to the total poverty
200
count, and the availability of the same so-
cioeconomic, demographic, geographic, and
Millions of poor
limited consumption data at two points in 150
time, a model of consumption was estimated
on the basis of the common variables at these
100
two points in time. The change in the char-
acteristics of the population of India is lev-
eraged to estimate how much consumption 50
increased over time (in a manner that avoids
assuming that inequality did not change). For
the cases of both India and Nigeria, the lack 0
2012 2015 2018 2021 2024 2027 2030
of recent data available for analysis results in
poverty estimates that are almost certainly India URP India MMRP Nigeria
Congo, Dem. Rep. Ethiopia Bangladesh
much less precise than many other estimates
in this report. Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet/. World Bank,
The other measurement issue is that there Washington, DC, World Development Indicators; World Economic Outlook; Global Economic Prospects.
Note: India URP (Uniform Reference Period) relies on poverty estimates and projections based on a
are many different ways to ask survey re- uniform recall period; India MMRP (Modified Mixed Reference Period) relies on poverty estimates and
spondents about their consumption habits, projections based on the modified mixed recall period.
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 31
BOX 1.3 India: Issues with the 2015 Poverty Estimate and 2030 Forecasts
The 2015 estimate, 2018 nowcast, households in the 2011–12 survey areas) to calibrate the growth rate in
and 2030 forecasts for India merit would be given a growth rate of survey mean consumption between
special mention given both the 21 percent, and poverty in 2015 2011–12 and 2014–15. The fraction
importance of India to the global would be estimated using this of growth from national accounts
poverty count and the particularly adjusted welfare vector. Given that is passed through to growth
challenging measurement issues. India’s importance for the global in the survey mean implied by this
One source of the problem arises poverty rate, and the availability of procedure is 55.9 percent for urban
from the fact that only a subset a newer survey (albeit without a India and 73.3 percent for rural
of the 2014–15 survey data was full consumption aggregate), it was India. Earlier projections had used a
released by the government. felt that this extrapolation method pass-through of 57 percent (for both
There are two key issues, the first needed to be cross-validated. urban and rural areas), which was
of which is linked to how survey For this reason, the 2015 based on the observed historical
data from 2011–12 and 2014–15 poverty estimate for India is based relationship between the survey
are used to estimate poverty in on survey-to-survey imputation and national accounts growth
India for 2015. The second issue method to estimate the growth rates (Jolliffe et al. 2015, chapter 1,
is linked to a change in how India rate in HFCE. The method uses the footnote 14; Ravallion 2003).
measures consumption, which 2014–15 National Sample Survey The new method used for India
is the foundation of the poverty (NSS) that collected consumption marks the first time the World
estimate. information on only a small subset Bank is using inputs from survey-
of items but included questions to-survey imputation methods.
2015 poverty estimates for India: on several correlates of household Thus, there can be a variation in the
Imputing consumption consumption like household size, poverty estimate obtained from the
The usual methodology for lining age composition of the household, new method and the conventional
up countries to the reference year caste status, and labor market HFCE-based method. The 2015
(for this report, 2015) is based indicators. In the first step, a model extreme poverty rate for India with
on two assumptions: the survey of the relationship between per the imputation-based growth rate is
mean grows at the same rate as capita household consumption 2.5 percentage points higher than
HFCE or GDP per capita, and there and household characteristics is with the HFCE growth rate (13.4
is no change in the distribution of developed using the NSS data percent versus 10.9 percent).
consumption. These assumptions from 2004–5, 2009–10, and In the coming years, when
may be reasonable when adjusting 2011–12. These surveys have the countries do not have surveys with
over a short period of time, but full consumption questions as well full consumption modules, but have
they become problematic as the as the variables used in the model. other smaller surveys with partial
distance between the survey In the second step, the estimated coverage, similar methods may be
year and the lineup year increases relationship is imposed on the applied to minimize reliance on the
(Jolliffe et al. 2015). 2014–15 data to predict household two assumptions implicit in the
The latest survey with official consumption and poverty status. HFCE approach. Household surveys
poverty estimates for India was See Newhouse and Vyas (2018) with full consumption modules
conducted in 2011–12, so a 2015 for more details on the modeling are undoubtedly the preferred
lineup would imply adjusting exercise. approach, and only in exceptional
the survey forward four years. PovcalNet uses the poverty rates cases will the imputation approach
With an HFCE growth rate of 21 at US$1.90 estimated by Newhouse be relied upon.
percent in India from 2011–12 to and Vyas (2018) (10.0 percent for The new imputation approach
2015, the welfare aggregate for all urban and 16.8 percent for rural implies that the poverty estimate
(continued)
With the cautions in mind that consump- Nigeria is now the country with the most
tion in 2015 for both India and Nigeria is poor people in the world (figure 1.9). When
based on projections, not direct enumera- examining a scenario where the consumption
tions of consumption from recent household measure for India is based on the new ques-
surveys, the nowcast for 2018 suggests that tionnaire rather than the old one, the esti-
32 POVERTY AND SHARED PROSPERITY 2018
BOX 1.3 India: Issues with the 2015 Poverty Estimate and 2030 Forecasts (continued)
for India in 2013 needs to also method under which questions on The choice of method
be updated. It has been revised household expenditure data for all can significantly affect total
from 16.5 percent to 17.8 percent. items were asked for the previous household consumption and
The new estimate is based on an 30-day period. After a series of poverty estimates. The official
average of the estimate from the experiments in the “thin” survey 2004–05 poverty rate for India
2011–12 survey and the 2014–15 rounds from 1994–95 to 1998, the with the URP-based consumption
survey, where, prior to averaging, Mixed Reference Period (MRP) data was 27.5 percent. The
the estimates have been lined up method was introduced in the corresponding figure for the MRP-
to 2013 using the HFCE-based 1999–2000 survey round in which based consumption data was 21.8
approach described above. This expenditure on food, pan, and percent (Government of India
lineup is based on a shorter time tobacco was collected using 7-day 2007). These changes did not,
period where the two assumptions and 30-day recall periods, and the however, affect the estimates
are less problematic. expenditure data for five nonfood of extreme poverty because the
items—clothing, footwear, durable World Bank continued to use
Changes in how consumption data goods, education expenses, and the URP-based aggregate for
are collected: Questionnaire design institutional medical expenses— international poverty monitoring
Recall period affects reported were collected using a 365-day to maintain comparability with
consumption through two main recall period (Deaton and Kozel historical estimates. The poverty
channels: memory decay and 2005). estimates and forecasts for India
telescoping. A longer recall With the 2011–12 round of presented here, based on MMRP
period is better at encompassing the NSS, the Modified Mixed (figure 1.9), similarly indicate a
expenditure on infrequently Reference Period (MMRP) was significant decline in the number of
purchased items, but it can introduced where the recall period poor people. An important caveat,
lead to underreporting if was set at 7 days for perishable however, is that the difference
respondents forget about the items, 365 days for the five low- in the count of extreme poor as
past purchases. Despite lower frequency items, and 30 days for measured by URP and MMRP
average consumption, measured the remaining items (Government dissipates with economic growth.
poverty might be lower under of India, Planning Commission In the most recent “thick” round
the longer recall period because 2014). For the sake of comparability of the NSS Consumer Expenditure
it captures the purchases of low- over time, the World Bank global Survey, India has phased out
frequency items of households in poverty count has been based on the URP as well as the MRP
the lower parts of the distribution. consumption measures derived questions, which means extreme
Short recall periods can mitigate from the URP instrument. With poverty can no longer be tracked
underreporting but can lead to the next NSS Consumption and using the URP-aggregate. The
telescoping, where respondents Expenditure Survey, India is no next update of global poverty will
mistakenly report the consumption longer enumerating consumption likely show a sizeable drop in the
that took place outside of the with the URP. This means that the extreme poverty, both because of
reference period. global poverty count produced by economic growth and because of
Until 1993–94, the consumption the World Bank will soon no longer India’s switch to the MMRP-based
data in India were collected using be based on the URP for India and consumption aggregate.
the Uniform Reference Period (URP) a switch to the MMRP will occur.
mates indicate that Nigeria overtook India in Drilling down: Africa and fragile
2015 as the country with the most poor peo- and conflict-affected situations
ple in the world. These projections are based
on old surveys and strong assumptions, but, In 2002, Sub-Saharan Africa was home to less
if the historically observed patterns in India than a quarter of the world’s extreme poor,
and Nigeria continue, Nigeria either already whereas, in 2015, more extreme poor lived
is or soon will be the country with the most in the region (413 million) than everywhere
people living in extreme poverty. else in the world combined. If this trend con-
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 33
FIGURE 1.10 Household Size and Dependency Ratio in Sub-Saharan Africa
a. Household size b. Dependency ratio
8 160
7 140
6 120
5 100
4 80
3 60
2 40
1 20
0 0
Nonpoor households Poor households Nonpoor households Poor households
Circa 2004 Circa 2011
Source: World Bank Africa Poverty database.
Note: The median years for the base period and the terminal period are 2004 and 2011, respectively. Dependency ratio is the ratio of
dependents (people younger than 15 or older than 64) to the working-age population (ages 15–64).
tinues as the forecasts suggest, extreme pov- the region, the fast rate of population growth
erty will soon become a predominantly Afri- has led to the increase in the total popula-
can phenomenon. An important first step in tion of poor people in Sub-Saharan Africa.
tackling poverty in the region is to better un- These demographic features of the region
derstand the factors associated with poverty will continue to pose a challenge for poverty
in Sub-Saharan Africa. reduction, a point that was anticipated by the
One such factor is the demographic struc- first World Development Report on poverty
ture of the household. In many parts of (World Bank 1990).
the world, the poor generally live in larger A second contributing factor for the slow
households and have more economically decline in extreme poverty in Sub-Saharan
dependent members per working-age adult Africa is that growth in this region has been
(Castaneda et al. 2016). In many regions of less effective in reaching the poor than growth
the world, the ratio of dependent household in other regions. One indicator of this is the
members to working-age adults is declining. region’s low growth elasticity of poverty. For
However, this is not the case in Sub-Saharan every percentage increase in GDP per capita,
Africa. Household surveys from the region poverty in a typical non-African developing
show no appreciable decrease in average country falls by 2 percent, whereas in a typ-
household size or in the dependency ratio ical African country it falls by only 0.7 per-
over the 2000s (figure 1.10). cent (Christiaensen, Chuhan-Pole, and Sanoh
The good news of a declining under-5 2013). There is a caveat to the elasticity com-
mortality rate in Sub-Saharan Africa, and parison—the level of poverty is much higher
elsewhere in the world (figure 1.11, panel a), in Sub-Saharan Africa so a smaller percentage
has combined with a relatively small drop in change in a higher level can still be a signif-
the total fertility rate to keep Sub-Saharan icant reduction in poverty—but the general
Africa’s population growing at a higher rate point is that growth in Sub-Saharan Africa
than that of every other region in the world has been less effective in reducing poverty
(figure 1.11, panels b and c) (Canning, Raja, than elsewhere. Some of the leading explana-
and Yazbeck 2015; Groth and May 2017). Al- tions for this ineffectiveness of growth in re-
though poverty rates have declined slightly in ducing poverty include the overall high levels
34 POVERTY AND SHARED PROSPERITY 2018
FIGURE 1.11 Under-5 Mortality, Fertility, and Population Growth in Sub-Saharan Africa
a. Under-five mortality rate b. Total fertility rate
200 7
180
6
160
Deaths (U5) per 1,000 births
140 5
Births per woman
120 4
100
3
80
60 2
40
1
20
0 0
1990 1995 2000 2005 2010 2015 1990 1995 2000 2005 2010 2015
c. Population growth rate
4
3 Sub-Saharan Africa
Annual growth (%)
East Asia and Pacific
Latin America and the Caribbean
2
South Asia
Europe and Central Asia
1 Middle East and North Africa
World
0
1990 1995 2000 2005 2010 2015
Source: World Development Indicators (http://databank.worldbank.org/data/source/world-development-indicators).
of inequality in several countries and growth After falling sharply between 2005 and 2011,
that is predominantly in capital-intensive sec- the poverty rate has since gone up: in 2015,
tors like natural resource extraction. the poverty rate in 35 economies in FCS was
As the global poverty rate declines, there 35.9 percent, up from a low of 34.4 percent in
is concern that extreme poverty will become 2011. The share of the global poor living in
a phenomenon increasingly associated with FCS has risen steadily since 2010, culminat-
institutional fragility and conflict. It is also ing in 23 percent of all poor people in 2015
the case that most people (54 percent) liv- (figure 1.12, panel b).12
ing in fragile and conflict-affected situations This rise has not come about because pop-
(FCS) in 2015 are in Sub-Saharan Africa.9 ulous countries have joined the ranks of frag-
To see if there is evidence that poverty is al- ile situations, except for a small drop between
ready beginning to pool in FCS, trends in the 2005 and 2008, the share of the world popu-
poverty rate and the share of the global poor lation living in fragile situations has stayed
living in fragile situations are analyzed.10 level through much of the period (figure 1.12,
Figure 1.12, panel a, shows the poverty rate panel c). Were more countries to become
in economies in FCS from 2005 to 2015.11 fragile, the goal of rooting out global poverty
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 35
FIGURE 1.12 Concentration of Extreme Poverty in Fragile and Conflict-Affected Situations
a. Poverty rate b. Share of the global poor
100
50
80
40
Share of poor (%)
Poverty rate (%) 60 81.1 78.5 76.8
30 86.0 84.8
52.4
20 41.6 40
34.4 34.8 35.9
10 20
18.9 21.5 23.2
14.0 15.2
0 0
2005 2008 2011 2013 2015 2005 2008 2011 2013 2015
c. Share of the total population
100
80
Share of population (%)
60
92.4 93.9 93.9 93.0 93.4
40
20
7.6 6.1 6.1 7.0 6.6
0
2005 2008 2011 2013 2015
Fragile and conflict-affected situations
Not fragile and conflict-affected
Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet/. World Bank, Washington, DC. Harmonized List of
Fragile Situations (http://www.worldbank.org/en/topic/fragilityconflictviolence/brief/harmonized-list-of-fragile-situations)
Note: See appendix A for more details on the list of countries in fragile and conflict-affected situations.
would only get more challenging. Panels b and (World Bank 2017a ).13 For illustration, figure
c together also reveal the “poverty burden” 1.13 plots the performance of countries on a
borne by the economies in FCS: they have 6.6 few fundamental indicators of economic and
percent of the global population but 23 per- institutional quality against poverty rates. In
cent of the poor, which is 3.5 times higher than general, there is a negative correlation be-
would be expected if poverty were equally tween poverty rates and the strength of in-
prevalent everywhere. Despite this significant stitutions; countries with high poverty rates
pooling of poverty, these estimates almost cer- have lower financial penetration (panel a;
tainly undercount the extent of poverty in FCS correlation = −0.59), poorer business climate
for several reasons, including technical mea- (panel b; correlation = −0.62), weaker rule of
surement reasons such as missing data on ref- law (panel c; correlation = −0.46), and higher
ugees and displaced persons (see appendix A). perceived corruption (panel d; correlation =
Fragility comprises many elements, and −0.43). Notably, fragile situations (marked in
countries that are in fragile situations are red) are often among the poorest performers
characterized by policy failures and institu- in these areas, falling in the bottom quintile
tional weaknesses in multiple dimensions of the distribution. They must make signifi-
36 POVERTY AND SHARED PROSPERITY 2018
FIGURE 1.13 Fragile Situations Perform Poorly in Multiple Constituent Components of Fragility
a. Financial Inclusion index, 2014 b. Doing Business score, 2015
80 80
Poverty rate (%), 2015
Poverty rate (%), 2015
60 60
40 40
20 20
0 0
0 20 40 60 80 100 30 40 50 60 70 80 90
c. Rule of Law score, 2015 d. Control of Corruption score, 2015
80 80
Poverty rate (%), 2015
Poverty rate (%), 2015
60 60
40 40
20 20
0 0
−2 −1 0 1 2 −2 −1 0 1 2 3
Fragile and conflict-affected situations Not fragile and conflict-affected 20th percentile
Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet; World Bank, Washington, DC. The Global Findex Database (https://globalfindex.worldbank
.org/); Doing Business (http://www.doingbusiness.org/); and Worldwide Governance Indicators (http://info.worldbank.org/governance/wgi/index.aspx#home).
Note: Financial Inclusion Index is the proportion of individuals with a bank account in 2014. Doing Business indicator is the “Distance to Frontier” score for 2014. The 2015 Rule
of Law Indicator and the Control of Corruption Indicator are drawn from the Worldwide Governance Indicators (WGI) project. These indicators are used as guideposts to set the
World Bank’s CPIA ratings (http://pubdocs.worldbank.org/en/600961531149299007/CPIA-Criteria-2017.pdf).
cant progress on several constituent compo- opment programs in proper locations, and
nents of fragility simultaneously to relieve the target the beneficiary population accurately,
constraints to economic growth and poverty it is critically important to know where the
reduction. poor live, what conditions they live in, and
how they earn a living. This description of the
Socioeconomic and poor is frequently done within each country,
informing country dialogue on how best to
demographic profile of
improve the well-being of the less well off in
global poverty society. But researchers and policy makers
To devise an appropriate poverty reduction can also learn a great deal by examining a
strategy, it is not enough to merely know how global profile of the poor. This examination
many people are poor. In order to choose the can aid the international development com-
right poverty reduction policies, place devel- munity to better target poverty alleviation
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 37
programs as well as areas of well-being re- A stronger correlation is observed between
quiring emphasis. poverty and educational achievement. Of the
The profile of the poor is based on har- adults with no education, more than a fifth
monized household surveys from 91 coun- are in poverty. There is a premium to having
tries in the Global Monitoring Database had even some schooling: the poverty rate
(GMD).14 It is an update of a previous pro- more than halves for adults with incomplete
file that was based on the harmonized data primary education, whereas poverty is all
for 89 countries for 2013.15 The sample used but absent among adults with some tertiary
for this profiling covers about 76 percent of education. Given that intergenerational mo-
the world’s population and 86 percent of the bility in education is low in low- and middle-
extreme poor in 2015. The data demands for income countries, there is a danger that this
the global poverty profile are more stringent pattern will carry over to the next generation
than that for the global poverty update. It re- as well (Narayan et al. 2018). Increasing labor
quires harmonization of additional variables productivity in agriculture and improving
like age, gender, education, and sector of human capital to facilitate labor migration
work from diverse household surveys, which into high-productivity sectors and locations
is why the poverty profile is available for only are key to poverty reduction.
a subset of countries and for an earlier date. The fertility rate is usually higher among
Globally, extreme poverty continues to the poor. As a result, poor households are
be disproportionately and overwhelmingly usually large and have many children. There
rural. The poverty rate in rural areas (17.2 are on average 7.7 members and 3.5 children
percent) is more than three times as high as under the age of 14 in the world’s extremely
that in urban areas (5.3 percent); with ap- poor households. Just under a fifth of chil-
proximately 54 percent of the world’s popu- dren under the age of 14 live in poverty, and,
lation, rural areas account for 79 percent of despite representing only about a quarter of
the total poor. Rural poverty is strongly as- the population, they make up more than two-
sociated with the sector of employment; the fifths of the absolute poor (table 1.1). There
extreme poverty rate is higher among agri- is suggestion of increasing concentration of
cultural workers, and they constitute almost poverty among children, with children under
two-thirds of the extreme poor. But nonfarm the age of 14 constituting a marginally larger
employment does not guarantee an escape share of the poor in 2015 (45.7 percent) than
from poverty; a significant share of poor in 2013 (44.2 percent).16 Children who grow
adults in both urban and rural areas is em- up in poverty acquire less human capital be-
ployed in nonagricultural sectors. cause of inadequate or low-quality schooling
and undernutrition. This makes childhood
poverty especially pernicious because it per-
TABLE 1.1 Age and Gender Profile of the Extreme Poor, 2015 petuates intergenerational poverty.
Share of the Share of the The current state of data limits the abil-
Poverty rate (%) poor (%) population (%) ity to understand the prevalence of poverty
Age group by gender and age. Household surveys collect
0–14 19.3 45.7 27.4 information on total household consump-
15–24 11.7 16.9 16.6 tion. They typically do not differentiate how
25–34 9.4 13.0 15.9
35–44 8.7 10.1 13.4
resources are allocated within a household.
45–54 6.4 6.4 11.6 For analytical purposes, it is assumed that all
55–64 5.9 4.2 8.2 household members have equal needs and
65 and up 5.9 3.6 6.9 that total consumption is distributed equally
Total 11.5 100.0 100.0 within a household. The equal distribution
assumption distorts the picture of poverty
Gender
Male 11.7 50.3 49.6
if there is inequality within households. For
Female 11.4 49.7 50.4 example, the profile shows that males and
females are equally likely to be in poverty.
Source: Estimates based on the harmonized household surveys in 91 countries, circa 2015, GMD (Global
Monitoring Database), Global Solution Group on Welfare Measurement and Capacity Building, Poverty Chapter 5 takes up this issue in detail and
and Equity Global Practice, World Bank, Washington, DC. proposes methodological changes in house-
38 POVERTY AND SHARED PROSPERITY 2018
hold surveys to capture the intrahousehold TABLE 1.2 Education and Access to Services among the Extreme
distribution of consumption. In the mean- Poor and Nonpoor Households
time, differences in poverty by gender and Share of households (%)
age will be informed by assuming someone Poor Nonpoor
is poor if he or she lives in a poor household. No adult member has completed primary education 53.1 12.2
The poor lack not just income. Poverty At least one school-age child (up to grade 8) is out of school 22.8 3.4
also materializes as low educational attain- Household does not have access to limited-standard source of 37.0 8.6
ment, poor health and nutrition outcomes, drinking water
exposure to physical insecurity and natu- Household does not have access to limited-standard sanitation 66.8 16.3
ral hazards, and substandard living condi- facilities
tions. Globally, a large share of extreme poor Household does not have access to electricity 67.8 7.1
households has no adult member with pri- Source: Estimates based on the harmonized household surveys in 119 countries, circa 2013, GMD
(Global Monitoring Database), Global Solution Group on Welfare Measurement and Capacity Building,
mary schooling, and in many households at
Poverty and Equity Global Practice, World Bank, Washington, DC.
least one child of school age (up to grade 8)
is out of school (table 1.2). The poor are also
poorly served in essential services like accept-
able standards of drinking water, adequate cline by a percentage point a year the global
sanitation facilities, and electricity (table 1.2). poverty rate declined by half a percentage
Low levels of human capital and poor point a year, the world would still meet the
access to basic services undermine labor 3 percent target. Despite this scope for the
productivity of the poor, often their most pace to significantly slacken, all forecasts for
important source of income, trapping them 2030 considered in this chapter that are based
in income poverty. Increasingly, however, on countries or regions growing in line with
poverty is understood as encompassing more their recent historic performance indicate
than just income. Sufficient education, good that the world will fall well short of the target.
health, a safe living environment, and pro- Part of the explanation for the deceleration
vision of basic services are desired for their in poverty reduction is that not all regions
intrinsic value, beyond their instrumental have shared in the global economic growth
value in raising income. Chapter 4 takes a of the last quarter century, nor have all re-
panoramic view of poverty as the inability gions succeeded in ensuring that the poor
to reach a sufficiency threshold in monetary have fully shared in the benefits of growth.
terms as well as in a wide range of nonmon- Sub-Saharan Africa has had inadequate levels
etary dimensions that directly affect an indi- of growth and inadequate poverty reduction
vidual’s well-being. from growth, and this has resulted in the in-
crease of the total number of people in this
region living in extreme poverty. In 1990, 278
Conclusions million people in Sub-Saharan Africa lived in
Between 1990 and 2015, the world made extreme poverty; by 2015, this increased to an
steady progress toward the target of reduc- estimated 413 million people. Forecasts based
ing the number of people living in extreme on historic average growth rates predict that
poverty to less than 3 percent globally by the number of people living in extreme pov-
2030. The extreme poverty rate dropped on erty in the region will remain above 400 mil-
average 1 percentage point every year, falling lion in 2030.
from 35.9 percent in 1990 to 10.0 percent in A related reason why poverty reduction
2015. As a result of this decline, there were is slowing is that previously progress rested
well over a billion fewer people living in pov- heavily on the success of the countries of
erty despite a global population that had in- East Asia and Pacific and South Asia in re-
creased by more than 2 billion people during ducing the total number of people living in
this period. With the estimated extreme pov- extreme poverty. The countries of East Asia
erty rate at 10 percent in 2015, the target of and Pacific have experienced remarkable
3 percent by 2030 could be attained even if reductions in extreme poverty. In 1990,
the rate of poverty reduction was cut in half. there were 987 million people living in ex-
That is to say, if instead of continuing to de- treme poverty in this region, and this num-
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 39
ber dropped to 47 million people by 2015. erage poverty rates in all regions of the world
On average, each year ended with about 38 except for Africa are below 2 percent; how-
million fewer people living in extreme pov- ever, the forecasted average extreme poverty
erty in the East Asia and Pacific region. But rate for Sub-Saharan Africa is above 25 per-
now, with the prevalence of extreme poverty cent. Even in a forecast based on an assumed
below 3 percent, and the number of poor in real growth rate of 8 percent, the 3 percent
this region contributing only about 6 percent global target is met but extreme poverty in
to the total population of poor, there are few Sub-Saharan Africa is in double digits (13.4
remaining gains to be made in this region in percent).
terms of having a significant effect in reduc- This sort of outcome, where extreme pov-
ing the global poverty rate. erty is eliminated throughout the world ex-
Although there are still many extreme cept in one region where it is in double dig-
poor in South Asia, a similar story will most its certainly does not portray a picture of a
likely soon occur there, and this is good news. world free of poverty. This, then, is one of the
In 1990, more than a half billion people in key messages of this report: it is time to go
South Asia lived in extreme poverty; by 2015, beyond the focus on bringing down the aver-
this dropped to 216 million people. A rela- age global poverty rate to 3 percent to reach
tively large portion of the extreme poor still the goals of eradicating extreme poverty and
live in South Asia, but the forecasts indicate ensure that all share in the benefits of eco-
(combined with the anticipated change in nomic development.
how consumption is measured in India) that A key point of this report is that the view
the total number of poor there is rapidly de- of poverty needs to be broadened. Now that
clining. The success in reducing extreme pov- the extreme poverty rate is less than 3 percent
erty in many regions of the world means that in half the countries of the world and is be-
the majority of the remaining gains in pov- coming increasingly concentrated, finishing
erty reduction must come from the countries the job will require constructing a more de-
of Sub-Saharan Africa. tailed and complete picture of what is meant
The unevenness of the progress in global by a world free of poverty. To do this, the next
poverty reduction brings into focus the rel- chapters in this report go beyond extreme
ative strengths and weaknesses in how prog- income poverty to start the process of moni-
ress toward the goal of a world free of poverty toring poverty in all its forms. New measures
is monitored. In various forecasts assuming introduced in this report allow one to better
that countries continue to grow in line with monitor poverty in all countries, in multiple
their recent performance (or with the aver- aspects of life, and for all individuals in every
age historic growth rate of their region), av- household.
40 POVERTY AND SHARED PROSPERITY 2018
Annex 1A
Historical global and regional
poverty estimates
This annex contains tables of historical pov- FIGURE 1A.1 Global Total Consumption Gap
erty rates at the global, regional, and coun- of the Extreme Poor, 1990–2015
try levels. Poverty rates do not speak to the
1,400 1,276
distribution of consumption (or income)
among the poor, meaning that the poor may 1,200
fare worse in certain countries than in others.
Million of US$ (2011 PPP)
1,000
For this reason, the poverty rates are comple-
mented with other measures of poverty: the 800
poverty gap, the poverty gap divided by the
600
poverty rate, and the squared poverty gap
(Foster, Greer, and Thorbecke 1984). 400 433
The poverty gap measures the average dis-
200
tance to the poverty line, where people above
the poverty line are given a distance of zero. 0
This measure reflects both the share of poor 1990 1995 2000 2005 2010 2015
and the average daily consumption of the Source: PovcalNet (online analysis tool), World Bank, Washington,
poor, but expressed as the average shortfall DC, http://iresearch.worldbank.org/PovcalNet/.
among the entire population. If two countries Note: PPP = purchasing power parity.
have the same poverty rate, but the poor in
the first country have a daily consumption Although both the poverty gap and the pov-
of US$1.50, whereas in the other they have a erty gap divided by the poverty rate are sen-
daily consumption of US$1.80, then the pov- sitive to the average level of consumption (or
erty gap will indicate a higher depth of poverty income) among the poor, they do not account
in the first country. When the poverty gap is for inequality among the poor. The squared
divided by the poverty rate, the resulting num- poverty gap—which is the average squared dis-
ber shows the average distance to the poverty tance to the poverty line, where people above
line, or average consumption shortfall among the poverty line have a distance of zero—is
the poor. If the average consumption shortfall sensitive to inequality among the poor. Sup-
of the poor is 0.25, then poor individuals on pose that two countries have the same poverty
average consume 25 percent less than the value rate, and the poor in both countries on average
of the IPL, or US$1.43 per day ( (1-0.25)*IPL ). consume US$1.50 daily. Suppose further that,
Since 1990, both of these complementary in one of the countries, all the poor consume
measures of poverty have improved. The total US$1.50, whereas the other country has many
consumption gap of the poor (the sum of all people consuming much less. The squared
consumption shortfalls of the poor) shrank poverty gap measures this latter country, with
from US$1,276 million (2011 PPP) in 1990 greater inequality among the poor, as having
to US$433 million (2011 PPP) in 2015 (figure more severe form of poverty.
1A.1). This improvement reflects both that An issue to keep in mind with these com-
the share of people living in extreme poverty plementary poverty measures is that they are
has decreased and that the average income of more sensitive to whether poverty is mea-
the poor has increased over this time interval. sured with consumption or income. Whereas
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 41
poverty estimates based on income can be using income is faring poorly in these com-
zero—and even negative—in a given period plementary measures, and it makes it difficult
because of negative income shocks, they to compare the depth of poverty across coun-
rarely get close to zero when consumption is tries using consumption and income (see ap-
used. This makes it more likely that a country pendix A for more discussion on this).
TABLE 1A.1 Global and Regional Extreme Poverty, 1990–2015
a. Global extreme poverty, 1990–2015
Squared
Year Poverty rate (%) Poverty gap (%) poverty gap Poor (millions) Population (millions)
1990 35.9 12.7 6.1 1,894.8 5,284.9
1993 33.9 11.9 5.8 1,877.5 5,542.9
1996 29.4 9.8 4.7 1,703.2 5,792.6
1999 28.6 9.5 4.5 1,728.6 6,038.1
2002 25.6 8.3 3.9 1,609.9 6,276.8
2005 20.7 6.3 2.9 1,352.2 6,517.0
2008 18.1 5.4 2.4 1,223.2 6,763.7
2011 13.7 4.1 1.9 963.0 7,012.8
2013 11.2 3.4 1.6 804.2 7,182.9
2015 10.0 3.1 1.5 735.9 7,355.2
b. Extreme poverty rates, by region, 1990–2015
Percent
Region 1990 1993 1996 1999 2002 2005 2008 2011 2013 2015
East Asia and Pacific 61.6 54.0 41.1 38.8 29.9 19.1 15.1 8.6 3.6 2.3
Europe and Central Asia 2.9 a 5.0 7.2 7.8 5.9 4.9 2.8 2.1 1.6 1.5
Latin America and the Caribbean 14.2 13.2 13.8 13.6 11.8 9.9 6.9 5.6 4.6 4.1
Middle East and North Africa 6.2 6.7 5.8 3.8 3.2 3.0 2.7 2.7 2.6 5.0
South Asia 47.3 44.9 40.3 39.3 a 38.6 33.7 29.5 19.8 16.2 12.4 a
Sub-Saharan Africa 54.3 58.9 58.2 57.7 56.4 50.7 47.8 45.1 42.5 41.1
Sum of regions 43.1 40.6 35.1 34.0 30.4 24.5 21.3 16.1 13.1 11.6
Rest of the world 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.6 0.6 0.7
World 35.9 33.9 29.4 28.6 25.6 20.7 18.1 13.7 11.2 10.0
c. Number of extreme poor, by region, 1990–2015
Millions
Region 1990 1993 1996 1999 2002 2005 2008 2011 2013 2015
East Asia and Pacific 987.1 902.0 712.9 695.9 552.5 361.6 292.8 169.6 73.1 47.2
Europe and Central Asia 13.3 a 23.4 33.8 36.7 27.6 22.9 13.3 9.8 7.7 7.1
Latin America and the Caribbean 62.6 61.3 67.7 69.7 63.2 54.9 39.9 33.8 28.0 25.9
Middle East and North Africa 14.2 16.6 15.3 10.6 9.4 9.4 8.8 9.2 9.5 18.6
South Asia 535.9 542.1 518.0 534.4 a 554.3 510.4 467.0 328.0 274.5 216.4 a
Sub-Saharan Africa 277.5 327.3 350.7 376.1 398.0 387.7 396.4 406.4 405.1 413.3
Sum of regions 1,890.5 1,872.7 1,698.3 1,723.5 1,605.0 1,346.9 1,218.1 956.9 797.8 728.5
Rest of the world 4.3 4.9 4.9 5.0 4.9 5.3 5.1 6.2 6.4 7.3
World 1,894.8 1,877.5 1,703.2 1,728.6 1,609.9 1,352.2 1,223.2 963.0 804.2 735.9
Source: PovcalNet (http://iresearch.worldbank.org/PovcalNet/), World Bank.
Note: Sum of regions was previously referred to as developing world.
a. This estimate is based on a regional population coverage of less than 40 percent. The criterion for estimating survey population coverage is whether at least one survey used
in the reference year estimate was conducted within two years of the reference year.
42 POVERTY AND SHARED PROSPERITY 2018
TABLE 1A.2 Extreme Poverty, by Economy, 2015
Number of poor Poverty rate Poverty gap Poverty gap/rate
Economy Survey year(s) (millions) (%) (%) (%)
Albania 2012 0.0 0.9 0.2 20.0
Algeria 2011.17 0.1 0.4 0.1 37.1
Angola 2008.5 7.8 27.9 8.7 31.2
Argentina 2014 and 2016 0.3 0.6 0.3 45.3
Armenia 2015 0.1 1.9 0.4 18.8
Australia 2010 0.1 0.5 0.3 62.0
Austria 2015 0.1 0.8 0.5 72.0
Azerbaijan 2005 0.0 0.0 0.0
Bangladesh 2010 and 2016 24.4 15.2 2.8 18.1
Belarus 2015 0.0 0.0 0.0
Belgium 2015 0.0 0.0 0.0
Belize 1999 0.0 13.9 6.0 43.1
Benin 2015 5.2 49.6 22.4 45.1
Bhutan 2012 and 2017 0.0 1.7 0.3 16.3
Bolivia 2015 0.7 6.4 2.8 44.3
Bosnia and Herzegovina 2015 0.0 0.2 0.1 30.0
Botswana 2009.25 0.3 12.8 3.7 29.3
Brazil 2015 6.9 3.4 1.2 34.5
Bulgaria 2014 0.1 1.2 0.5 36.3
Burkina Faso 2014 7.8 42.8 10.8 25.2
Burundi 2013.5 7.6 74.7 32.9 44.0
Cabo Verde 2007.33 0.0 7.2 1.7 23.0
Cameroon 2014 5.2 22.8 7.1 31.3
Canada 2013 0.2 0.5 0.2 32.0
Central African Republic 2008 3.5 77.7 44.0 56.6
Chad 2011 4.8 34.1 13.2 38.7
Chile 2015 0.2 1.3 0.8 58.5
China 2015 10.0 0.7 0.2 21.9
Colombia 2015 2.2 4.5 1.7 38.2
Comoros 2013.5 0.1 18.2 6.5 35.7
Congo, Dem. Rep. 2012.4 55.1 72.3 34.6 47.9
Congo, Rep. 2011 1.7 34.9 13.5 38.7
Costa Rica 2015 0.1 1.5 0.6 38.8
Côte d’Ivoire 2015 6.5 28.2 9.1 32.4
Croatia 2015 0.0 0.8 0.4 46.7
Cyprus 2015 0.0 0.0 0.0
Czech Republic 2015 0.0 0.0 0.0
Denmark 2015 0.0 0.2 0.1 57.1
Djibouti 2013 0.2 18.6 6.3 33.9
Dominican Republic 2015 0.2 1.9 0.5 25.5
Ecuador 2015 0.6 3.4 1.2 35.8
Egypt, Arab Rep. 2015 1.3 1.4 0.2 11.9
El Salvador 2015 0.1 1.9 0.4 20.7
Estonia 2015 0.0 0.5 0.4 78.7
Eswatini 2009.25 0.5 39.0 14.8 37.9
Ethiopia 2010.5 and 2015.5 27.0 27.0 7.7 28.6
Fiji 2013.24 0.0 1.0 0.2 16.7
Finland 2015 0.0 0.0 0.0
France 2015 0.0 0.0 0.0
Gabon 2005 and 2017 0.1 4.1 1.0 24.1
Gambia, The 2010.08 and 2015.31 0.2 11.1 2.5 22.9
Georgia 2015 0.1 4.0 1.0 24.7
Germany 2015 0.0 0.0 0.0
Ghana 2012.8 3.0 10.9 3.1 28.7
Greece 2015 0.2 1.5 0.8 52.7
Guatemala 2014 1.3 7.9 2.3 29.3
Guinea 2012 4.0 33.0 9.4 28.4
Guinea-Bissau 2010 1.2 65.3 29.4 44.9
(continued)
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 43
TABLE 1A.2 Extreme Poverty, by Economy, 2015 (continued)
Number of poor Poverty rate Poverty gap Poverty gap/rate
Economy Survey year(s) (millions) (%) (%) (%)
Guyana 1998 0.1 6.5 1.9 28.9
Haiti 2012 2.5 23.7 7.6 32.1
Honduras 2015 1.4 16.2 5.6 34.9
Hungary 2015 0.0 0.5 0.3 61.2
Iceland 2014 0.0 0.0 0.0
India* 2011.5 175.7 13.4 2.4 17.7
Indonesia 2015 18.5 7.2 1.2 16.6
Iran, Islamic Rep. 2014 0.3 0.4 0.1 16.2
Iraq 2012 0.8 2.2 0.3 14.8
Ireland 2015 0.0 0.2 0.2 95.7
Israel 2012 0.0 0.5 0.3 54.2
Italy 2015 1.2 2.0 1.4 70.5
Jamaica 2004 0.1 1.8 0.4 22.8
Japan 2008 0.3 0.2 0.2 68.2
Jordan 2010.24 0.0 0.2 0.0 16.7
Kazakhstan 2015 0.0 0.0 0.0
Kenya 2005.38 and 2015.67 17.6 37.3 11.9 31.9
Kiribati 2006 0.0 11.8 3.0 25.4
Korea, Rep. 2012 0.1 0.3 0.1 44.0
Kosovo 2015 0.0 0.4 0.1 20.0
Kyrgyz Republic 2015 0.2 2.5 0.5 18.5
Lao PDR 2012.25 0.9 14.0 2.9 20.7
Latvia 2015 0.0 0.7 0.4 47.3
Lebanon 2011.77 0.0 0.0 0.0
Lesotho 2010 1.2 54.8 28.1 51.3
Liberia 2014 1.8 40.2 12.3 30.7
Lithuania 2015 0.0 0.8 0.5 72.0
Luxembourg 2015 0.0 0.2 0.2 95.0
Macedonia, FYR 2014 0.1 5.0 2.4 47.2
Madagascar 2012 18.8 77.5 38.8 50.1
Malawi 2010.23 12.2 69.6 31.7 45.6
Malaysia 2013 and 2015.33 0.0 0.0 0.0
Maldives 2009.5 0.0 4.1 0.8 20.3
Mali 2009.89 8.3 47.8 14.5 30.4
Malta 2015 0.0 0.0 0.0
Mauritania 2014 0.3 6.2 1.5 23.9
Mauritius 2012 0.0 0.4 0.1 17.5
Mexico 2014 and 2016 4.2 3.3 0.8 24.4
Micronesia, Fed. Sts. 2013 0.0 15.4 5.5 35.9
Moldova 2015 0.0 0.0 0.0
Mongolia 2014 and 2016 0.0 0.2 0.0 10.0
Montenegro 2014 0.0 0.0 0.0
Morocco 2013.5 0.3 0.9 0.2 17.4
Mozambique 2014.44 17.4 62.2 27.3 43.8
Myanmar 2015 3.3 6.4 1.5 23.1
Namibia 2009.54 and 2015.27 0.3 13.4 4.5 33.8
Nepal 2010.17 2.0 7.0 1.4 19.8
Netherlands 2015 0.0 0.0 0.0
Nicaragua 2014 0.2 2.9 0.6 22.3
Niger 2014 8.9 44.5 13.5 30.2
Nigeria 2009.83 86.5 47.8 18.6 38.9
Norway 2015 0.0 0.2 0.0 16.7
Pakistan 2013.5 and 2015.5 9.9 5.2 0.7 13.2
Panama 2015 0.1 2.0 0.5 26.8
Papua New Guinea 2009.67 2.3 28.4 10.3 36.3
Paraguay 2015 0.1 1.9 0.4 21.7
Peru 2015 1.1 3.6 1.0 27.3
(continued)
44 POVERTY AND SHARED PROSPERITY 2018
TABLE 1A.2 Extreme Poverty, by Economy, 2015 (continued)
Number of poor Poverty rate Poverty gap Poverty gap/rate
Economy Survey year(s) (millions) (%) (%) (%)
Philippines 2015 8.5 8.3 1.6 18.9
Poland 2015 0.0 0.0 0.0
Portugal 2015 0.1 0.5 0.3 50.0
Romania 2015 1.1 5.7 1.9 33.4
Russian Federation 2015 0.0 0.0 0.0
Rwanda 2013.75 6.0 51.5 17.6 34.2
Samoa 2008 0.0 1.1 0.1 10.5
São Tomé and Príncipe 2010 0.1 26.0 6.2 24.0
Senegal 2011.29 5.3 35.7 11.4 31.9
Serbia 2015 0.0 0.1 0.0 30.0
Seychelles 2013 0.0 1.0 0.4 40.6
Sierra Leone 2011 3.5 48.4 14.8 30.5
Slovak Republic 2015 0.0 0.7 0.3 35.1
Slovenia 2015 0.0 0.0 0.0
Solomon Islands 2013 0.1 24.7 6.7 26.9
South Africa 2014.83 10.4 18.9 6.2 32.8
South Sudan 2009 8.7 73.3 40.0 54.6
Spain 2015 0.5 1.0 0.6 64.6
Sri Lanka 2012.5 and 2016 0.2 0.8 0.1 11.7
St. Lucia 1995 0.1 28.3 9.8 34.6
Sudan 2009 3.0 7.7 2.0 25.8
Suriname 1999 0.1 18.8 14.5 77.0
Sweden 2015 0.0 0.5 0.3 50.0
Switzerland 2015 0.0 0.0 0.0
Syrian Arab Republic 2004 4.0 21.2 4.8 22.4
Tajikistan 2015 0.4 4.8 1.1 21.8
Tanzania 2011.77 21.9 40.7 11.7 28.9
Thailand 2015 0.0 0.0 0.0 33.3
Timor-Leste 2014 0.4 31.2 6.9 22.0
Togo 2015 3.6 49.2 19.9 40.5
Tonga 2009 0.0 1.0 0.2 21.4
Trinidad and Tobago 1992 0.0 0.6 0.2 35.7
Tunisia 2010.41 0.1 0.9 0.2 18.3
Turkey 2015 0.2 0.3 0.1 21.4
Turkmenistan 1998 0.2 2.8 0.4 15.5
Tuvalu 2010 0.0 2.4 0.2 6.8
Uganda 2012.45 and 2016.5 15.8 39.2 12.3 31.2
Ukraine 2015 0.1 0.1 0.0 8.3
United Kingdom 2015 0.1 0.2 0.1 39.1
United States 2013 and 2016 3.7 1.2 1.0 82.8
Uruguay 2015 0.0 0.1 0.0 23.1
Uzbekistan 2003 4.4 14.0 3.8 26.8
Vanuatu 2010 0.0 12.8 3.2 24.7
Venezuela, RB 2006 2.8 8.9 6.6 74.7
Vietnam 2014 and 2016 2.1 2.3 0.4 19.1
West Bank and Gaza 2011 and 2016.75 0.0 0.6 0.1 15.3
Yemen, Rep. 2014 11.0 40.9 12.0 29.3
Zambia 2015 9.3 57.5 29.5 51.3
Zimbabwe 2011 2.5 16.0 3.5 21.5
Source: PovcalNet (http://iresearch.worldbank.org/PovcalNet/), World Bank.
Note: The year column refers to the year of the survey that is used to calculate the 2015 estimate as listed in PovcalNet. Note that for economies that use EU-SILC surveys, the
survey year is backdated by one year to align with the reference period for the income data in the survey (for example, the 2016 survey is listed as 2015). If one year is listed,
and this year is different from 2015, the poverty estimate from the year of the survey has been extrapolated to 2015. If two years are listed, the 2015 estimates are based on
interpolations between these two surveys. For more information on how these interpolations and extrapolations are carried out, see appendix A. The decimal year notation is
used when data are collected over two calendar years. The number before the decimal point refers to the first year of data collection, while the numbers after the decimal point
show the proportion of data collected in the second year. For example, the Algerian survey (2011.17) was conducted in 2011 and 2012, with 17 percent of the data collected in
2012. Pov. rate is the poverty rate, or the percentage of the population living on less than the IPL (international poverty line). Pov. gap is the average consumption shortfall of the
population where the nonpoor have no shortfall (as described above). Pov. gap / pov. rate is the average consumption shortfall of the poor (as described above). * indicates that
the 2015 estimate for India is based on an imputation described in box 1.3.
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 45
Annex 1B
Validation check of the
2030 poverty projections
The poverty projections to 2030 are based FIGURE 1B.1 Projections to 2015 of Global
on several critical assumptions regarding Extreme Poverty
countries’ future growth rates and the nature
of this growth. The global poverty patterns 35
in 2030 may look very different if these as-
sumptions are not met. The soundness of 30
the 2030 forecasts can be assessed indirectly
Poverty rate (%)
by pretending that the poverty levels and 25
growth rates from 2002 to 2015 are unknown
and applying the forecast methodology to 20
2002–15. For example, one can use the coun-
try-specific and regional growth rates from 15
1992 to 2002 to predict poverty rates from
2002 to 2015. With this approach, the 2015 10
forecasts can be benchmarked against the re-
alized poverty levels in 2015. This would help 5
1990 1995 2000 2005 2010 2015
uncover the sensitivity of the assumptions
behind the projections and hence give an in- Projection using average regional growth rates
from 1992 to 2002
dication of the uncertainty surrounding the Projection using average country growth rates
2030 projections. from 1992 to 2002
Using this approach, the global rate of Projection using 6 percent annual growth rate
extreme poverty is predicted to be 13.4 per- Actual trend
cent in 2015—well above the actual rate of Source: PovcalNet (http://iresearch.worldbank.org/PovcalNet/),
10.0 percent (figure 1B.1). This is largely World Bank.
Note: The figure assumes 2002 is the latest year of data and
because the regional growth rates in Sub- applies the forecasting methods used toward 2030 to obtain pov-
Saharan Africa are severely underestimated erty “forecasts” for 2002–15. This can be benchmarked against
using historical growth data.17 The regional realized poverty levels, and hence allows for an assessment of the
soundness of the 2030 projections.
growth rate in GDP per capita in Sub-
Saharan Africa from 1992 to 2002 was 0.7
percent, whereas the actual growth in GDP the actual global poverty rate observed in
per capita from 2002 to 2015 turned out to 2015, the 2002–15 projections would need to
be several percentage points higher. Hence, use annual growth rates in the range of 6 per-
the historical growth rates were not a good cent per year.
indication of the future growth rates, and the Although this example speaks to the in-
projections overestimate the amount of pov- herent uncertainty of making long-term pov-
erty in Sub-Saharan Africa in 2015. In other erty projections, even with an annual growth
regions, such as East Asia and Pacific and rate across the globe of 6 percent until 2030,
South Asia, the projections are very close to the projections still do not predict that the 3
the actual poverty levels in 2015. To match percent target will be met.
46 POVERTY AND SHARED PROSPERITY 2018
Notes
1. The interim target of a poverty rate of 9 per- 0.544 for the Middle East and North Africa,
cent was set by the World Bank Group presi- 0.912 for South Asia, 0.748 for Sub-Saharan
dent at the 2013 Annual Meetings: http://www Africa, and 0.892 for the rest of the world.
.worldbank.org/en/news/speech/2013/10/11 5. GDP rates are used for Sub-Saharan Africa
/world-bank-group-president-jim-yong-kim- and for countries without HFCE growth rates.
speech-annual-meetings-plenary. The same pass-through rates are applied as in
2. Survey coverage is assessed by considering the nowcast. The average regional growth rate
surveys within a two-year window on either is weighted using each country’s population in
side of 2015, that is, surveys conducted be- 2015 as the weight.
tween 2013 and 2017. By this criterion, two- 6. Projections based on a global growth rate of
thirds of the world is covered by a survey for 8 percent and no shared prosperity premium
the 2015 poverty update. The coverage would are nearly identical to the 6 percent growth
be lower for more recent years. and 2 percentage point premium scenario,
3. The core poverty numbers reported in this and thus also get the global rate below 3 per-
chapter are for 2015. These numbers are re- cent by 2030. In general, a mean growth rate
ferred to as estimates. References to a nowcast of x percent combined with a shared pros-
indicate that the poverty rate is a forecasted perity premium of y percent is nearly identi-
estimate up to the current point in time, cal to a growth rate of x + y percent and no
which for this report is 2018. Because this re- shared prosperity premium. Projections using
lies largely on realized growth rates and pop- a 7 percent global growth rate and no shared
ulation figures, it should, in principle, be more prosperity premium, or a 5 percent growth
reliable than a forecast. References to forecasts rate and a 2-percentage-point premium, get
are used when the prediction is more remote very close to the 3 percent target.
in the future, later than the nowcast, typically 7. East Asia and Pacific (6.4 percent), Latin
2030. Forecasts are based on assumed growth America and the Caribbean (3.5 percent), the
rates and predictions of population figures and Middle East and North Africa (2.5 percent),
are estimated with significantly less precision. Europe and Central Asia (1.0 percent), and
4. The growth rates used are from the World the rest of the world (1.0 percent).
Development Indicators. Pass-through rates 8. Some evidence suggests that, if price differ-
are essentially estimated by comparing av- ences within countries are accounted for, the
erage differences between national accounts reduction in Sub-Saharan Africa has been
and household surveys. The mean national greater than the numbers suggested here
consumption or income from each country’s (Beegle et al. 2016). For more information on
household survey is compared with either the impact of price differences within coun-
GDP or household final consumption expen- tries on poverty, see appendix A.
diture (HFCE) from national accounts. HFCE 9. Of the 35 economies in FCS in 2015, 16 (45.7
is the preferred measure in most countries, percent) were in Sub-Saharan Africa. In terms
except in Sub-Saharan Africa where GDP is of population, of the 481.1 million people liv-
used for estimating pass-through factors and ing in FCS, 259.8 million (54.0 percent) were
growth rates. If GDP and HFCE data are un- in Sub-Saharan Africa. More details on how
available, growth forecasts from the Global countries are determined to be in FCS are
Economic Prospects are used. If these are also given in appendix A.
unavailable, growth forecasts from the World 10. The analysis uses a “rolling” roster of fragile
Economic Outlook are used. For Syria, no es- situations, that is, the set of fragile situations
timates are available in these sources. Instead, can change from one year of the analysis to
data from the Economist’s Intelligence Unit are the next.
relied upon. The fraction of GDP/HFCE per 11. This analysis goes back only to 2005 because
capita that is assumed to pass through to the the World Bank classification of fragile situa-
welfare vector is as follows: 0.785 for East Asia tions began that year.
and Pacific, 0.773 for Europe and Central Asia, 12. The aggregate FCS poverty rate is the
0.829 for Latin America and the Caribbean, population-weighted mean of the poverty
ENDING EXTREME POVERTY: PROGRESS, BUT UNEVEN AND SLOWING 47
rates of all economies in FCS. The number of 15. The 2013 profile and the methodological de-
poor is a product of the FCS poverty rate and tails are reported in Castaneda et al. (2016).
the total population living in FCS. This leads 16. The 2013 estimates are based on a different set
to a slightly higher estimate of the total num- of 89 countries (Castaneda et al. 2016). When
ber of poor (744 million versus PovcalNet es- the 2013 profiling is repeated using the same
timate of 736 million), but the discrepancy is 91 countries from 2015, children constitute
inconsequential to the current discussion. 44.9 percent of the poor.
13. The World Bank’s definition of fragility is based 17. Another reason for the discrepancy is that the
on the Country Policy and Institutional Assess- projections used here assume that only a frac-
ment (CPIA), which assesses the conducive- tion of the growth rates observed in national
ness of a country’s policies and institutions to accounts translates into growth in the con-
poverty reduction, sustainable growth, and the sumption aggregate observed in surveys. The
effective use of development assistance. The actual poverty numbers, in contrast, assume
CPIA comprises 16 indicators clustered in four that all growth observed in national accounts
dimensions: economic management, structural translates into growth in the consumption ag-
policies, policies for social inclusion and equity, gregate. This implies that the projections are
and public sector management and institutions. more pessimistic than the actual estimates for
14. Please refer to appendix A for more informa- countries where the 2015 estimate is based on
tion on the GMD. extrapolation.
48 POVERTY AND SHARED PROSPERITY 2018
Shared Prosperity: 2
Mixed Progress
This chapter reports on the progress achieved in promoting shared prosperity, defined as the
growth in the average consumption or income of the poorest 40 percent of the population
(the bottom 40). Introduced as one of two twin goals by the World Bank in 2013 along with
ending extreme poverty, fostering shared prosperity embodies notions of economic growth
and equity.
Shared prosperity is examined by country rather than globally. The latest available data, on
91 economies, paint a mixed albeit moderately positive picture. The bottom 40 were doing
well in most economies for which data are available in about 2010–15. Overall, the incomes
of the bottom 40 grew in 70 of the 91 economies monitored, and, in more than half the
bottom 40 obtained a larger share of the total income. Good performance in shared prosperity
is primarily but not exclusively found in South Asia, East Asia and Pacific, Latin America and
the Caribbean, and the Baltic countries in Northern Europe. However, slow economic progress
is hindering shared prosperity in some regions, particularly in Europe and Central Asia, and
other high-income countries, which experienced negative or low levels of shared prosperity.
More worrying, among the countries with high rates of poverty (most of which are located
in Sub-Saharan Africa), income growth at the bottom has on average been lower than in the
rest of the world. In addition, the picture of shared prosperity among the poorest economies
as well as those in fragile and conflict-affected situations is only partial because data on the
shared prosperity indicator remain limited.
Beyond extreme poverty: ity is shared within each country. Thus, even
in higher-income economies where extreme
A focus on the bottom 40
poverty rates are low, the shared prosperity
Promoting shared prosperity involves ensur- goal is still highly relevant.
ing that the relatively poor in every country To estimate shared prosperity, two com-
are able to participate in and benefit from parable surveys are needed. In this report,
economic success. Progress toward this goal the selected surveys were for circa 2010 and
is monitored through an indicator that mea- circa 2015 (box 2.1). The survey data are used
sures the annualized growth rates in average to calculate changes in consumption or in-
consumption or income among the poorest come. This presents a greater data challenge
40 percent of the population in each country than the calculation of a global poverty rate
(the bottom 40).1 Irrespective of the prev- (chapter 1). Therefore, the set of countries
alence of extreme poverty, this measure is included in the sample is smaller. The shared
meaningful as a gauge of how well prosper- prosperity measure is reported for 91 econ-
49
ulation coverage is lower than in the earlier
report, when it represented 75 percent of the
BOX 2.1 The Global Database of
global population.
Shared Prosperity
Shared prosperity estimates are Continued progress in most
calculated using household surveys
and are presented in the Global
economies though some are
Database of Shared Prosperity (GDSP). falling short
The present report is grounded on the In this sample of 91 economies, the bottom
sixth edition of the GDSP (the fall 2018
40 are mostly doing well. The incomes of
release), which features data on 91
economies circa 2010–15. For details,
the poorest 40 percent were growing in 70 of
please refer to appendix A. the 91 economies circa 2010–15. The simple
average of the annualized consumption or
income growth rate among the bottom 40
was 1.9 percent (table 2.1).
The performance in shared prosperity
omies in which the combined population across the world ranges from an annualized
is 4.6 billion, representing 62 percent of the 8.4 percent decline in income among the bot-
world’s population in 2015. Compared to the tom 40 in Greece to an annualized growth of
previous report with data for circa 2008–13, 9.1 percent in China (see figure 2.1 and map
the number of economies included in the 2.1).2 There are clear regularities in perfor-
present report is higher (91 rather than 83 mance across regions and income groups,
economies). However, given that a few large though with some exceptions. Three groups
countries, such as India, are excluded in this of economies can be identified on the basis of
round because of lack of data, the global pop- their performance in shared prosperity.
TABLE 2.1 Shared Prosperity and Shared Prosperity Premium, 91 Economies, Summary Table, circa 2010–15
SP indicator available Economies, number
Average SP
Population, Number of % of total Growth in SP Premium Average SP Premium
Region millions economies population mean > 0 SP > 0 >0 (%) (p.p)
East Asia and Pacific 2,036.6 8 94.6 7 8 7 4.73 1.33
Europe and Central Asia 487.0 26 89.9 18 20 13 2.22 0.15
Latin America and the Caribbean 626.5 16 87.8 15 16 14 3.19 0.98
Middle East and North Africa 371.6 3 47.8 1 2 2 0.98 1.33
South Asia 1,744.2 4 21.3 4 4 0 2.62 –0.56
Sub-Saharan Africa 1,005.6 12 32.4 9 8 5 1.84 –0.55
Rest of the world 1,083.6 22 71.7 14 12 10 -0.27 –0.33
Fragile and conflict-affected 485.1 4 7.6 2 3 3 2.03 0.80
IDA and Blend 1,539.3 20 42.7 16 17 10 2.16 –0.11
Low income 641.9 7 35.1 6 5 3 2.06 –0.67
Lower-middle income 2,970.0 24 36.1 19 21 13 2.56 0.30
Upper-middle income 2,560.4 28 93.7 21 24 20 2.61 0.77
High income 1,182.9 32 73.6 22 20 15 0.85 –0.20
Total 7,355.2 91 62.1 68 70 51 1.94 0.20
Sources: GDSP (Global Database of Shared Prosperity) fall 2018 edition, http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity; World Bank,
Washington, DC, PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: IDA = International Development Association; Blend = IDA-eligible countries but also creditworthy for some borrowing from the International Bank for Reconstruction and
Development; SP = shared prosperity; the indicator measures growth in the average household per capita consumption or income of the bottom 40. Shared prosperity premium =
the difference in growth in the average consumption or income of the bottom 40 and the mean, in percentage points (p.p.). Population coverage refers to 2015. The list of econ-
omies in fragility and conflict-affected situations is based on data for 2015. The shared prosperity indicator is close to zero (between −0.15 and 0.15 percent) in three countries:
Iceland, Niger, and Romania.
50 POVERTY AND SHARED PROSPERITY 2018
FIGURE 2.1 Shared Prosperity, 91 Economies, circa 2010–15
China Latvia
Malaysia Lithuania
Vietnam Georgia
Thailand East Asia Macedonia, FYR
Indonesia and Pacific Estonia
Philippines Kazakhstan
Mongolia Belarus
Fiji Kosovo
Moldova
Chile Turkey
Nicaragua Poland
Paraguay Tajikistan
Dominican Republic Armenia
El Salvador Russian Federation
Panama Latin America
and the Caribbean Czech Republic Europe and
Brazil Hungary Central Asia
Colombia Kyrgyz Republic
Uruguay Croatia
Peru Bulgaria
Ecuador Romania
Costa Rica Bosnia and Herzegovina
Bolivia Slovak Republic
Honduras Slovenia
Mexico Ukraine
Argentina Serbia
Montenegro
Sri Lanka
Pakistan South Asia Egypt, Arab Rep.
Bhutan Middle East
Iran, Islamic Rep.
Bangladesh and North Africa
West Bank and Gaza
Malta Burkina Faso
Norway Namibia
Sweden Rwanda
Ireland Mauritania
United States Togo Sub-Saharan Africa
Switzerland Ethiopia
Netherlands Mozambique
France Côte d'Ivoire
Belgium Rest of the world Niger
Denmark Zambia
Finland South Africa
United Kingdom Uganda
Iceland
Germany –5 0 5
Canada
Annualized growth in mean
Austria
incomes or consumption (%)
Portugal
Italy
Luxembourg
Spain
Cyprus Bottom 40 (shared prosperity)
Greece
Total population
–5 0 5
Annualized growth in mean
incomes or consumption (%)
Source: GDSP fall 2018 edition, World Bank, Washington, DC, http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity.
Note: The figure shows annualized growth in mean household per capita consumption or income (see annex 2B).
SHARED PROSPERITY: MIXED PROGRESS 51
MAP 2.1 Shared Prosperity across the World, 91 Economies, circa 2010–15
Source: GDSP (Global Database of Shared Prosperity) fall 2018 edition, World Bank, Washington, DC.
Note: The map shows annualized growth in mean household per capita consumption or income (see appendix A).
The first group consists of, by and large, a treme poverty, and the region now consists
large part of the developing world in which of mainly middle-income countries (World
the incomes of those in the bottom 40 are Bank 2018a). The success in South Asia, as
growing, in some cases strongly. This is pri- mentioned in the previous chapter, was more
marily, though not exclusively, the case of recent than in East Asia and Pacific but is still
economies in East Asia and Pacific, South persistent.
Asia, and Latin America and the Caribbean. In many Latin American and Caribbean
On average, the incomes of the bottom 40 in countries, the progress in lifting incomes
these regions grew by 4.7 percent, 2.6 per- of those at the bottom has been widespread
cent, and 3.2 percent per year, respectively since the early 2000s and is still strong despite
(table 2.1). In some cases, such as in various the more recent slowdown. After a decade of
countries in East Asia and Pacific, current strong economic growth and shared prosper-
high levels of shared prosperity represent ity, largely driven by favorable commodity
a continuation of over a quarter century of prices and expanded social protection sys-
strong and broadly shared economic growth tems (Ferreira et al. 2013), regional growth
driven by labor-intensive development com- has slowed since 2012 as international condi-
bined with investment in human capital, tions deteriorated. The economic slowdown
which particularly benefitted the lower part translated into slower poverty reduction and
of the distribution (Birdsall et al. 1993; Com- more sluggish income growth among the
mission on Growth and Development 2008) middle class, particularly in South American
(see box 2.2). This success means that more countries (Calvo-González et al. 2017; World
than a billion people have risen out of ex- Bank 2018b). The income of the bottom 40
52 POVERTY AND SHARED PROSPERITY 2018
BOX 2.2 Country Stories
With contributions from Kenneth Simler, Samuel Freije-Rodriguez, Rakesh Gupta N. Ramasubbaiah, and Carolina Mejia-Mantilla.
Rising East Asia: the dysfunctional wage-setting and skill premiums in several high-
China and Malaysia practices for low-paid workers income European and non-European
As described in chapter 1, the (Del Carpio and Pabon 2014). economies (Katz and Autor 1999;
success of economies in East The increase of minimum wages Goldin and Katz 2007; Katz and
Asia and Pacific in drastically has also been linked to strong Margo 2014; Ganong and Shoag
reducing poverty in the last few reductions in inequality in other 2017; Ridao-Cano and Bodewig
decades is unparalleled. Solid countries such as Brazil (World 2018; Bussolo et al. 2018).
educational foundations and strong Bank 2016a). In contrast, household
export-oriented growth from income growth was lower in Droughts and pests
manufacturing have been some 2013–15, about 6 percent per year, affecting Uganda
of the fundamental growth drivers and almost distribution-neutral. Between 2012 and 2016, Uganda
in the region. The high rates of experienced a setback in terms
Stagnated incomes at the
income growth among the bottom of reducing poverty and boosting
bottom in high-income
40 continue to be observed in the shared prosperity, trends that had
countries
last five years. been observed throughout the
The fast growth of consumption Inequality in the developed world decade leading up to 2012. The
per capita among the bottom 40 has recently been the focus of poverty headcount ratio (under the
in China is supported by faster intensified public debate. Rich international poverty line) increased
growth in rural than in urban evidence using different and new from 35.9 to 41.6 percent, and
household disposable income. For estimation methods and sources consumption for the bottom 40
the period 2013–15, the higher of data on welfare distributions for shrank by 2.15 percent per year,
income dynamism in rural areas Western Europe and the United more than the 0.96 percent per
is driven by household operations States emerging from the last
year decline for the average
(family business or farm incomes), decade suggest that the top 1
consumption. Behind the reversal
which accrue 2.8 percentage points percent are getting a larger share
of fortune were the drought and
(out of 10.1) of disposable income of national income since the 1980s
pests that affected the agricultural
growth in rural households, but only and that the incomes of those at the
sector for the better part of 2016
0.8 percentage points (out of 8.6) bottom 50 percent have remained
and the beginning of 2017. Given
in urban households. This indicates stagnant or even declined (Alvaredo
that households engaged in
that traditional economic activities et al. 2018). In the United States,
agriculture remain highly vulnerable
continue to have a significant for example, estimates suggest that
to weather and price shocks, these
influence in the growth of the the average pre-tax income for this
problems affected the livelihood
rural economy. Higher disposable latter group has stagnated at about
$16,000 (in constant 2014 dollars) of rural households in particular.
income entailed a higher increase in
since 1980 (Piketty et al. 2018). Estimates using panel data show
consumption expenditure in almost
The question of lack of income that the lack of rainfall and low
all consumption items for rural
growth for the median worker (a prices contribute substantially
residents.
In Malaysia, the rapid income comprehensive description can to lower income for Ugandan
growth among the bottom 40 be found in Shambaugh and Nunn agricultural households. A 10
(see figure 2.4) from 2011 to 2018) is complex but has been percent decline in water sufficiency
2015 is fundamentally driven addressed by several studies in the (rainfall) decreases crop income
by extraordinary performance recent literature. Some explanatory by 9.9 percent, while a 10 percent
between 2011 and 2013—when factors focus on the emergence decline in the price of maize or
wages rose sharply and overall of superstar firms that led to beans lowers crop income by 4.5 or
income of the bottom 40 grew increasing monopolistic rents and 9.2 percent, respectively (Hill and
at an annual rate of 12 percent. a declining labor share, which did Mejia-Mantilla 2017). The effects
The timing of the increase in labor not benefit lower-skilled workers are higher for poorer households as
earnings coincides with minimum during this period (Autor et al. 2017; they are more vulnerable to adverse
wage legislation passed in 2012, Barth et al. 2016). Others stress shocks. For these households, a
which introduced minimum the fact that technological change, 10 percent decline in rainfall and
wages for the first time, relevant combined with the educational a 10 percent decline in maize and
to all workers except domestic landscape, has dampened median bean prices result in a 13.4 percent
employees. In part, the minimum wage income growth (and increased and 13.0 percent decline in crop
wage was put in place to address polarization of the wage distribution) income, respectively.
SHARED PROSPERITY: MIXED PROGRESS 53
grew 1.4 percentage points more slowly per 2017; Ridao-Cano and Bodewig 2018; Bus-
year in circa 2010–15 than in circa 2008–13 solo et al. 2018). (See also box 2.2).
(reported in the previous edition of this re- Finally, there is also cause for concern
port) with average annualized rates of 3.2 among some of the poorest economies and
percent compared to 4.6 percent in the pre- those in fragile and conflict-affected situa-
vious period (annex 2B, table 2B.2.). Still, tions. On average, the incomes of the bottom
shared prosperity continued to be high in 40 in Sub-Saharan Africa grew at 1.8 percent
many countries in the region. In Chile, in- per year, a pace slightly lower than in the total
comes of the bottom 40 grew at a rate of 6.0 sample. But this number is the average among
percent per year in 2010–15, driven by soar- economies where incomes of the bottom 40
ing hourly wages and a strong public transfer declined or grew below 1 percentage point
system protecting the most vulnerable. (over a third of African economies) and other
Within this first group of good performers economies in which income growth was
in shared prosperity, the Baltic states—Es- strong, such as Burkina Faso and Rwanda.
tonia, Latvia, and Lithuania—were able to The negative performance in countries with
recover vigorously after the 2008 and 2013 high poverty rates like Uganda and Zambia is
crises. Between 2010 and 2015, incomes of likely related to the poor performance of the
the bottom 40 in these countries grew at agriculture sector, in part due to adverse cli-
rates above 6 percent per year. These coun- mate shocks and pests (see box 2.2). Among
tries were among those that experienced the four conflict-affected economies with avail-
largest gross domestic product declines and able data, two had low or negative income
fiscal deficits during the years of the crisis growth for the bottom 40. Although Côte
(OECD 2012), and implemented large fis- d’Ivoire’s shared prosperity of 0.7 is still low,
cal consolidations programs (Sutherland, it represents a recovery from a decade of po-
Hoeller, and Merola 2012). Starting in 2011, litical and economic crisis. In the Middle East
they experienced some of the strongest eco- and North Africa, the poor performance in
nomic growth recovery relative to other Eu- West Bank and Gaza reflects to a large extent
ropean countries (De Agostini et al. 2015; the economic despair in Gaza, despite prog-
World Bank 2018c). ress in West Bank, which was largely restricted
A second group includes relatively rich to urban areas. A second important source
economies, with low prevalence of extreme of concern among these poor or conflict-
poverty (in the single digit), in which in- affected economies is that their coverage of
comes of the bottom 40 are growing slowly, the shared prosperity indicator is low, an
stagnating, or even losing ground. This is issue highlighted in the next section.
the case of several Eastern and Western Eu-
ropean countries, such as Greece and Spain, The poorest countries have
as well as of other high-income economies,
limited information about
such as the United States. On average, the in-
comes of the bottom 40 in the so-called rest shared prosperity
of the world contracted 0.3 percent per year Of the 164 countries with an available in-
in circa 2010–15. In some countries such as ternational poverty rate, only a quarter
Greece, Portugal, and Spain, the negative of low-income economies and 4 of the 35
performance reflects, to a greater extent, the recognized as being in fragile and conflict-
slow recovery from the European debt crisis affected situations also have a shared prosper-
(IMF 2017; World Bank 2018c). In richer ity indicator.3 As a consequence, the coverage
economies such as the United Kingdom and of shared prosperity in Sub-Saharan Africa
the United States, more structural processes is limited: only 12 of the 45 economies for
that led to the stagnation of incomes at the which poverty estimates are available in the
bottom since the 1980s, or more recently region are included (figure 2.2). In contrast,
in continental European countries such as 84 percent of the high-income economies are
Germany and Poland, which are sometimes represented in the shared prosperity analy-
linked to polarization of wages and regu- sis. Of the 57 countries with extreme poverty
lations (Alvaredo et al. 2018; Piketty et al. rates above 10 percent, only 13 have a shared
54 POVERTY AND SHARED PROSPERITY 2018
prosperity indicator. Two countries that con- FIGURE 2.2 Shared Prosperity Estimates, 91 Economies, by Region,
centrate a high proportion of the world’s Group, and Income
poor, India and Nigeria, are excluded because
they lack comparable data across time. Popu- East Asia and Pacific
lation coverage is also limited among econo- Europe and Central Asia
mies grouped by other World Bank country Latin America and the Caribbean
categories, such as small island nations for Middle East and North Africa
which there is no shared prosperity indicator South Asia
available. Sub-Saharan Africa
Because this round excludes many poorer Rest of the world
countries as well as those in fragile and
conflict-affected situations, the picture on Fragile and conflict affected
shared prosperity for these economies is only IDA and Blend
partial. The computation of the shared pros-
perity measure relies on frequent and com- Low income
parable data collection (appendix A). This is Lower-middle income
often associated with a country’s level of de-
Upper-middle income
velopment because data collection depends
High income
on the capacity of a national statistics office.
Stronger commitments to narrowing the data 0 10 20 30 40 50 60 70
gap are needed if the shared prosperity goal is Number of economies
to be monitored globally in a timely fashion Positive shared prosperity Negative shared prosperity
(Independent Evaluation Group 2017).4 No shared prosperity measure
Sources: GDSP (Global Database of Shared Prosperity) fall 2018 edition, World Bank, Washington, DC,
Growth at the bottom and http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity; PovcalNet
the top is not always even (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: The count is based on the 164 economies on which direct estimates of the poverty rate are avail-
The incomes or consumption of the bot- able through PovcalNet. IDA = International Development Association; Blend = IDA-eligible countries but
also creditworthy for some borrowing from the International Bank for Reconstruction and Development;
tom 40 depend directly on both the average No shared prosperity measure = economies with poverty rates reported in PovcalNet, but insufficient
growth within the economy and the share of data to compute a shared prosperity indicator.
national income that accrues to the bottom
40 (Rosenblatt and McGavock 2013; World
Bank 2016b) (annex 2A). Improvements at prosperity. The number of economies exhib-
the bottom may thus derive from the fact iting a positive premium is less (51) than the
that society in general is doing better—that number showing a positive shared prosperity
is, the tide lifts all boats. Improvements may indicator (70) (table 2.1.). The implication
also arise from progressive shifts in the dis- is that, in almost half the economies moni-
tribution of economic gains (Lakner, Negre, tored, the consumption or income share of
and Prydz 2014, 2015). The shared prosperity the bottom 40 is growing more slowly than
premium represents an effort to capture such the average, suggesting that the distribution
progressive shifts. It is defined as the differ- in these countries is worsening because the
ence between the annual income growth rate bottom 40 are getting a smaller share of total
among the bottom 40 and the annual growth income. Globally, the average shared pros-
rate of the mean in the economy. A positive perity premium is small. The simple average
premium indicates that the incomes or con- across all economies in the sample is 0.2 per-
sumption of the bottom 40 are increasing at centage points.
an above average rate and that the bottom 40 The regions with higher average premi-
are obtaining a larger share of overall con- ums are East Asia and Pacific, the Middle East
sumption or income (see box 2.3 for a com- and North Africa, and Latin America and the
parison with other concepts of inequality Caribbean. In these regions, the incomes of
based on income shares). the bottom 40 grew by 1.3, 1.3, and 1.0 per-
Achieving progress is more elusive in the centage points above the mean, respectively.
shared prosperity premium than in shared These regions also include a larger share of
SHARED PROSPERITY: MIXED PROGRESS 55
BOX 2.3 The Shared Prosperity Premium and Other Concepts of Inequality
The shared prosperity premium distribution to determine whether countries for which the analysis
calculated on the basis of the the rich are becoming richer. is performed differs from WIR.
2010–15 sample shows that, in 51 Although the WIR uses data on
• The absence of the top income
of the 91 economies, the bottom top earners from administrative tax
earners in household surveys.
40 are obtaining a larger share of records only for 10 countries,b this
Often, household surveys tend
total income in their countries. This type of data is currently available
to suffer from nonresponse or
suggests that, in a little more than underreporting at the top of the for 58 countries in the World
half of the economies, inequality distribution. Therefore, to obtain Inequality Database for at least one
has been declining. However, the reliable data on the top earners, year. In the dataset, high-income
perceptions of the public and the studies focusing on the rich, such and upper-middle-income countries
World Inequality Report 2018 (WIR) as the WIR, tend to be based are more represented than low- and
do not seem to agree that within- on tax records, complementing lower-middle-income countries. Of
country inequality is narrowing in a household surveys. Yet, for a the 58 with some information on
majority of countries.a According to large part of the developing top incomes, 32 are also included
the global picture displayed in the world, tax records are not readily in the present chapter. The large
WIR, inequality has been widening available, and thus the present majority of the economies in both
over the past few decades, and the chapter is not able to account datasets (almost 80 percent) are
richest people in each country are for underreporting at the top. upper-middle- and high-income
increasing their share of national The implication is that the economies, in which it was shown
incomes at an alarming pace. analysis from the chapter differs that the progress in terms of the
This mismatch in interpretations from the WIR both because shared prosperity premium was
of inequality trends stems partly consumption or income at the more limited than in the rest of the
from differences in the definition top is not properly accounted for world. Table B2.3.1 compares both
of inequality, as well as from and because the subset of samples.
differences in the supporting data.
TABLE B2.3.1 Number of Economies with Top Incomes
• Inequality at the top versus Estimated in the World Inequality Database and in the
inequality at the bottom. The Poverty and Shared Prosperity Report
shared prosperity premium
focuses on the bottom of the Both WID
Income group and PSPR Only WID Only PSPR
national income distribution as a
High income 18 13 14
gauge of inequality. It reflects an
Upper-middle income 9 3 19
assessment of whether the poor
Lower-middle income 4 6 20
are catching up or falling farther
Low income 1 4 6
behind. Meanwhile, the WIR
Note: PSPR = Poverty and Shared Prosperity (this report); WID = World Inequality
focuses on the top of the income
Database.
a. Several perception-based surveys in East Asia and Pacific indicate that respondents feel income disparities are too large
(World Bank 2018a). For World Inequality Report 2018, see Alvaredo et al. (2018).
b. The WIR uses fiscal and national accounts data to scale up the income distributions to match national income estimates for
a large number of countries. But the distributional information used comes from only 10 countries (Brazil, China, Côte d’Ivoire,
France, Germany, India, Lebanon, Russian Federation, United Kingdom, and United States). These are used to predict income
dynamics in their neighboring countries to obtain regional and global income inequality estimates.
economies with positive shared prosperity In the four South Asian economies included
premiums, with all but one or two in each in the sample, incomes among the bottom
region for which the incomes of the bottom 40 are growing, but at a slower pace than
40 grew at a faster rate than the rest of the the mean. In addition, half the countries in
economy (figure 2B.1). Europe and Central Asia and more than half
In contrast, higher concentrations of in Sub-Saharan Africa have negative shared
shared prosperity premiums close to zero or prosperity premiums. These two regions
negative are found in the other four regions. are unique in that they house the lowest
56 POVERTY AND SHARED PROSPERITY 2018
and most negative shared prosperity pre- FIGURE 2.3 Correlation between Shared Prosperity and the Shared
miums (Armenia, Mozambique, and Zam- Prosperity Premium, 91 Economies
bia), as well as some of the highest premi-
5
ums (Burkina Faso and the Former Yugoslav
Republic of Macedonia). This dichotomous 4
trend in inequality in Sub-Saharan African
Shared prosperity premium (percentage points)
has already been highlighted by Beegle et al. 3
(2016), who find increasing and decreasing
inequality without a clear pattern across 2
economies (that is, no clear association with
1
resource status, income levels, or initial levels
of inequality). 0
Relative to the previous report, the aver-
age shared prosperity premium across all –1
countries was slightly lower in 2010–15 than
–2
in 2008–13 (table 2B.3). Because of the lim-
ited sample coverage in some of the regions, –3
comparisons focus on the three subgroups of
countries for which data coverage is more sta- –4
ble and extensive across the two periods (see
appendix A on comparability across rounds): –5
–10 –8 –6 –4 –2 0 2 4 6 8 10
Europe and Central Asia, Latin America and
the Caribbean, and the rest of the world. The Shared prosperity (%)
decline in the premium was more pronounced Sources: GDSP (Global Database of Shared Prosperity), fall 2018, World Bank, Washington, DC, http://
in Latin America and the Caribbean, suggest- www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity; PovcalNet (online
ing not only that the economic slowdown in analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
this region dampened the performance in
consumption or income growth among the
bottom 40, but also that overall consumption comes of the bottom 40 grew at a more rapid
or income growth was not as equalizing as it rate relative to the average.
had been in the past. This is the case, for ex- If the shared prosperity indicator is neg-
ample, among several South American coun- ative, the shared prosperity premium is al-
tries, such as Peru and Uruguay, in which the most always negative as well (see figure 2.3).
rates of income growth among the bottom Of the 21 economies with negative shared
40 were about 3 percentage points above prosperity indicators, 19 also present nega-
the respective mean in 2008–13, whereas the tive premiums.5 This occurs in Europe and
corresponding gap in 2010–15 was closer to Central Asia, Sub-Saharan Africa, and the
1 percentage point. rest of the industrialized countries (rest of
There is a positive correlation between the world). Greece, Spain, and Zambia are
shared prosperity and the shared prosperity examples shown in figure 2.4, panel b. This
premium (figure 2.3). Of the 91 economies, means not only that incomes among the bot-
49 achieved both a positive shared prosper- tom 40 are shrinking rather than growing,
ity indicator (absolute growth among the but also that the decline is more profound
bottom 40) and a positive shared prosperity among the bottom 40 than across the rest of
premium (relative growth among the bottom the distribution. This result is consistent with
40). This is the case of most countries in Latin the evidence showing that the poor are more
America and the Caribbean and in East Asia highly exposed to downturns and shocks and
and Pacific, but also in 12 of the economies of that policies that safeguard them against such
Europe and Central Asia. As examples, figure risks—safety nets and insurance—can help
2.4, panel a, shows three cases, Latvia, Peru, guarantee that prosperity is shared. Poorer
and the Malaysia, in which incomes grew households are also much more likely to re-
across the entire distribution, whereas the in- duce consumption in response to shocks
SHARED PROSPERITY: MIXED PROGRESS 57
FIGURE 2.4 Growth across Deciles of the Income Distribution, Selected Countries
a. Positive shared prosperity: Positive premium b. Negative shared prosperity: Negative premium
(3 of 49 countries) (3 of 19 countries)
12 6
4
10
2
Annualized growth rate (%)
Annualized growth rate (%)
8 0
–2
6
–4
4 –6
–8
2
–10
0 –12
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Deciles Deciles
Latvia Malaysia Peru Greece Spain Zambia
Sources: GDSP (Global Database of Shared Prosperity), World Bank, Washington, DC, http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity;
PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: The bars illustrate the growth in the mean, by decile. The bottom 40 are in the left bars, in orange and red.
because they are also less likely to maintain more highly developed countries with almost
savings (World Bank 2013). no extreme poverty, children are more likely
to live in relatively more deprived households.
In addition, people in the bottom 40 dif-
Who are the bottom 40? fer significantly across countries. In terms
People in the bottom 40 differ from those liv- of consumption or income, in most low-
ing in the top 60, in terms not only of their income economies, such as Togo and Zam-
income but also of their characteristics. A bia, everyone in the bottom 40 lives on less
closer look at who makes up the bottom 40 in than US$1.90 a day (figure 2.5). In contrast,
a country may offer insights into the groups in more well-developed countries, only a
that are relatively more deprived. It can also small share of the bottom 40 are living in
guide national policy makers in identifying extreme poverty.
problem areas. Differences in income levels among the
Compared with the top 60, people in the bottom 40 across countries reflect not only
bottom 40 live disproportionally in rural the wealth of these economies as a whole
areas and attain less education than the rest of but also how the bottom 40 fare relative to
society. In addition, children are more likely the rest of the population. Although the bot-
to be among the bottom 40 than among the tom 40 in Croatia are consistently doing bet-
top 60. In Côte d’Ivoire, for example, children ter than the bottom 40 in Brazil, the rich in
under 15 years of age constitute about half Brazil are much richer than the top earners
the bottom 40, whereas they make up only in Croatia (figure 2.6). This reflects the fact
a third of the top 60. Similarly, in the Philip- that Brazil is much more unequal than Cro-
pines, children under 15 represent more than atia. The average daily income of the richest
40 percent of the bottom 40 but less than 25 decile in Brazilian society is more than 30
percent of the top 60. This pattern is repeated times higher than the average daily income
across all countries and regions in the current of the poorest decile, whereas the equiva-
sample. Chapter 1 concludes that children are lent ratio in Croatia is 8. Findings are similar
more likely than adults to live in extreme pov- among high-income economies with negligi-
erty. The present chapter finds that, even in ble poverty rates: for example, the bottom 40
58 POVERTY AND SHARED PROSPERITY 2018
FIGURE 2.5 Extreme Poverty and the Bottom FIGURE 2.6 Mean Income, by Distribution Decile, Selected Countries,
40, Selected Countries, circa 2015 2015
Zambia 120
Daily consumption or income per capita
Togo
Niger 100
Burkina Faso
Côte d’Ivoire 80
Honduras
Bangladesh
Bolivia 60
Philippines
Romania 40
Serbia
Mauritania
Tajikistan 20
Colombia
Georgia 0
Mexico 1 2 3 4 5 6 7 8 9 10
Peru Deciles of the income distribution
Brazil
Greece Belgium Brazil Croatia
Dominican Republic United Kingdom Zambia International poverty line
Bulgaria
Kyrgyz Republic Source: PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet/. World Bank,
Costa Rica Washington, DC.
Spain Note: The shaded area indicates the bottom 40. The lines represent the average daily consumption or
Lithuania income per capita by decile, expressed in 2011 purchasing power parity (PPP) U.S. dollars.
Slovak Republic
Hungary
Sri Lanka the bottom 40 receive less than 25 percent
Croatia of the overall income (figure 2.7). In several
United Kingdom Eastern European countries, such as Ukraine,
Belgium
the share is almost 25 percent. At the other
Iran, Islamic Rep.
Turkey extreme is Zambia, where the bottom 40 re-
France ceive less than 10 percent of the pie. Similar,
Uruguay though less extreme, is the situation in several
Ukraine
Latin American countries in which inequality
0 20 40 60 tends to be high.
Share of people living on
less than $1.90 per day (%)
Source: GDSP (Global Database of Shared Prosperity), fall 2018,
Monitoring the twin goals
World Bank, Washington, DC, http://www.worldbank.org/en The joint monitoring of poverty and shared
/topic/poverty/brief/global-database-of-shared-prosperity and
PovcalNet (online analysis tool), World Bank, Washington, DC, prosperity shines a spotlight on the extreme
http://iresearch.worldbank.org/PovcalNet/. poor and the less well-off in each country.
In this way, the most vulnerable can be iden-
in Belgium have higher average incomes than tified no matter the corner of the world in
the United Kingdom, even though the richest which they live and, at the same time, their
10 percent are richer in the United Kingdom progress highlighted. This section addresses
than in Belgium. this progress on the twin goals across the 91
The relative position of the bottom 40— economies for which the shared prosperity
how deprived they are compared with the indicator can be calculated among the 164
rest of the population—also varies largely economies on which the international pov-
across countries. The shared prosperity pre- erty rate is available.
mium captures whether the bottom 40 are There is a strong correlation between the
receiving a larger or smaller share of the twin goals, and most economies are per-
overall pie. How large is this piece of the pie forming well in both poverty reduction and
accruing to the bottom 40 across countries? boosting shared prosperity (figure 2.8, top
In all economies on which data are available, left quadrant). In most of the 91 economies
SHARED PROSPERITY: MIXED PROGRESS 59
FIGURE 2.7 Share of Consumption or Income, by Decile, Selected Countries, circa 2015
Zambia
Honduras
Brazil
Bolivia
Mexico
Peru
Nicaragua
Togo
Philippines
Turkey
Côte d'Ivoire
Uruguay
Romania
Iran, Islamic Rep.
Sri Lanka
Bulgaria
Tajikistan
Niger
United Kingdom
Mauritania
Armenia
Burkina Faso
Mongolia
Bangladesh
Hungary
Belgium
Norway
Ukraine
0 10 20 30 40 50 60 70 80 90 100
Share of total household per capita consumption or income
Decile 1 (poorest) Decile 10 (richest)
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
monitored, if the shared prosperity indicator can be challenging, and economic growth in
is positive, then the poverty rate is falling. Re- these economies does not necessarily align
gionally, circa 2010–15, all countries in East with large welfare improvements among the
Asia and Pacific and in Latin America and poorest in society (Bussolo and López-Calva
the Caribbean enjoyed a reduction in poverty 2014).
and positive shared prosperity. In terms of The risk of failing to reach the goal of
making progress on the twin goals, much can reducing poverty below 3 percent by 2030
be learned from these two regions. is greatest among the economies with ex-
In contrast, some economies have per- treme poverty rates above the global aver-
formed poorly in achieving progress on the age of about 10 percent (figure 2.9). All but
twin goals. In these economies, poverty rates one of these economies are in Sub-Saharan
rose, and the shared prosperity measure was Africa, with the exception being in Central
negative in circa 2010–15 (see figure 2.8, bot- America. Although only a fourth of the ex-
tom right quadrant). Of the 13 economies tremely poor economies are included in the
in this situation, only two also exhibited ini- shared prosperity sample (13 out of 57), an
tially high rates of extreme poverty (South examination of their shared prosperity mea-
Africa and Uganda). The rest are European sure in 2010–15 is not encouraging for many
countries with extremely low international of them.6 Except for a few countries, such as
poverty rates, and the changes in poverty are Burkina Faso, Namibia, and Rwanda, if these
thus also slight. Achieving equitable growth economies are to have a chance of reaching
60 POVERTY AND SHARED PROSPERITY 2018
the 3 percent goal by 2030, growth rates will FIGURE 2.8 Shared Prosperity and Changes in Extreme Poverty,
have to be high and incomes among the bot- 91 Economies, circa 2010–15
tom 40 will have to rise at an even higher rate.
10
Instead, in two-thirds of these countries, av-
Shared prosperity is positive
(the bottom 40 is growing)
erage incomes among the bottom 40 are in- 8
creasing at an annual rate below the global
Annualized growth rates of the bottom 40 (%)
average of 1.9 percent, and, in most of these, 6
consumption growth is slower for the bottom
40 than for the mean in the country. 4
To conclude, although most countries 2
have made progress in shared prosperity, the
results are mixed. This is in part due to the 0
Shared prosperity is negative
fact that in several richer economies incomes
(the bottom 40 is shrinking)
of the bottom 40 are growing slowly or not at –2
all. But there is also cause for concern at the
–4
very bottom—largely in Sub-Saharan Afri-
can and in economies in fragile and conflict- –6
affected situations.
This concern takes two forms: First, data –8
scarcity among the poorest and most fragile
–10
situations continues to be an issue, so cover- –3 –2 –1 0 1 2
age of the shared prosperity measure in these Change in poverty rate (p.p.)
countries is limited. This means that where Poverty is falling Poverty is rising
we need the most light we have the least. Sec- East Asia and Pacific South Asia
ond, where there are data (the 13 countries), Europe and Central Asia Sub-Saharan Africa
progress looks decidedly more mixed than Latin America and the Caribbean Rest of the world
among the middle-income success stories. As Middle East and North Africa
mentioned in chapter 1, reaching the global Sources: GDSP (Global Database of Shared Prosperity), fall 2018, World Bank, Washington, DC,
target of reducing extreme poverty to less http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity; PovcalNet
than 3 percent by 2030 will require greater (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: Changes in poverty are measured as the annual percentage point change in the international
attention to inclusive growth in the world’s poverty rate based on the US$1.90-a-day poverty line. Changes in poverty are measured over the same
poorest countries. period as shared prosperity.
FIGURE 2.9 Shared Prosperity among the Poorest Economies, circa 2010–15
Shared prosperity 2015 Poverty
Economy Type period rate (%)
Mozambique c 2008–14 62.2 Mozambique
Zambia c 2010–15 57.5 Zambia
Rwanda c 2010–13 51.5 Rwanda
Togo c 2011–15 49.2 Togo
Niger c 2011–14 44.5 Niger
Burkina Faso c 2009–14 42.8 Burkina Faso
Uganda c 2012–16 39.2 Uganda
Côte d'Ivoire c 2008–15 28.2 Côte d’Ivoire
Ethiopia c 2010–15 27.0 Ethiopia
South Africa c 2010–14 18.9 South Africa
Honduras i 2011–16 16.2 Honduras
Bangladesh c 2010–16 15.2 Bangladesh
Namibia c 2009–15 13.4 Namibia
Note: The column “Type” denotes whether the data reported are based on –4 –2 0 2 4 6 8
consumption (c) or income (i) data. The 2015 poverty rates have been lined-up to
Annualized growth in consumption or income (%)
2015 using interpolation or extrapolation methods. See appendix A for details.
Bottom 40 (shared prosperity) Total population
Sources: GDSP (Global Database of Shared Prosperity), fall 2018, http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity; World Bank, Wash-
ington, DC, PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
SHARED PROSPERITY: MIXED PROGRESS 61
Annex 2A
Shared prosperity definitions
The definition of shared of overall income that accrues to the bottom
prosperity 40, or both. This can be analytically expressed
as follows:
The shared prosperity measure represents the
annualized growth rate of the mean house- g40 = gmean + gshareB40 , (2A.1)
hold per capita consumption or income of where g40 is the income growth among the
the poorest 40 percent of the population (the bottom 40; gmean is the growth in the mean;
bottom 40), where the bottom 40 are deter- and gshareB40 is the growth in the income share
mined by their rank in household per capita of the bottom 40.
consumption or income. Unlike global and Although not an inequality indicator, the
regional poverty estimates that are popula- second term may be considered as the distri-
tion weighted, global and regional means of butional term that accounts for changes in
shared prosperity are simple averages. This the proportion of total income growth that
is because the shared prosperity indicator is accrues to the bottom 40. This is precisely the
purely national in focus. shared prosperity premium (SPP), which is
obtained by rearranging equation (2A.1) as
follows:
The definition of shared
prosperity premium gshareB40 = g40 – gmean ≡ SPP (2A.2)
The World Bank’s second twin goal, boosting This change in the share, or premium, does
shared prosperity, is sometimes character- not directly measure the inequality in a soci-
ized as a growth indicator and sometimes as ety. But it is a (limited) measure of distribu-
an indicator of inequality. In fact, it is a bit tional changes. If the incomes of the bottom
of both. Growth in the average consumption 40 grow at a rate that is above (or below) av-
(or income) of the bottom 40 can stem from erage, then inequality—at least between the
the rising mean consumption (or income) of bottom 40 and the rest of the distribution—
the overall population, changes in the share will tend to narrow (or widen).
62 POVERTY AND SHARED PROSPERITY 2018
Annex 2B
Shared prosperity estimates
by economy
TABLE 2B.1 Shared Prosperity Estimates, 91 Economies, circa 2010–15
Annualized growth in Mean consumption or income per capitaa
mean consumption or
income per capitaa,b Initial year Most recent year
Total Total Total
Bottom 40 population Bottom 40 Population Bottom 40 Population
Economy Periodc Typed (%) (%) ($ a day, PPP) ($ a day, PPP) ($ a day, PPP) ($ a day, PPP)
Chinaf 2013–15 C 9.11 7.37 3.91 9.46 4.65 10.90
Fiji 2008–13 c 1.17 –0.51 3.33 7.65 3.52 7.47
Indonesia 2015–17 c 4.77 4.79 2.51 5.68 2.75 6.24
Mongolia 2010–16 c 1.86 1.42 4.01 8.05 4.48 8.77
Malaysia 2011–15 i 8.30 5.95 7.89 21.76 11.14 27.95
Philippines 2009–15 i 2.43 1.38 2.38 6.75 2.74 7.33
Thailand 2010–15 c 5.03 3.04 5.67 13.29 7.24 15.43
Vietnam 2010–16 c 5.17 3.75 3.29 7.61 4.46 9.49
Armenia 2011–16 c 2.25 4.58 3.16 5.66 3.53 7.08
Bulgariag 2009–14 i 0.43 2.11 8.15 16.86 8.32 18.72
Bosnia and Herzegovina 2011–15 c –0.45 –0.79 9.51 19.26 9.34 18.65
Belarus 2011–16 c 4.06 3.46 9.40 16.34 11.47 19.37
Czech Republicg 2010–15 i 1.42 1.03 15.98 26.79 17.15 28.20
Estoniag 2010–15 i 6.15 6.62 10.71 21.07 14.44 29.04
Georgia 2011–16 c 6.44 4.32 2.46 5.97 3.36 7.38
Croatiag 2010–15 i 0.47 –0.12 9.28 18.82 9.49 18.71
Hungaryg 2010–15 i 1.19 1.73 10.55 19.57 11.19 21.33
Kazakhstan 2010–15 c 4.09 3.47 5.50 9.58 6.72 11.36
Kyrgyz Republic 2011–16 c 0.59 –0.03 3.07 5.30 3.16 5.29
Kosovo 2012–15 c 3.50 1.57 4.66 8.39 5.17 8.79
Lithuaniag 2010–15 i 6.65 8.10 7.91 16.79 10.91 24.79
Latviag 2010–15 i 7.52 6.47 7.74 16.93 11.11 23.16
Moldova 2011–16 c 2.79 0.39 4.92 9.19 5.65 9.37
Macedonia, FYR 2009–14 I 6.20 1.90 3.36 9.46 4.55 10.42
Montenegro 2009–14 c –2.73 –2.27 8.64 16.27 7.52 14.51
Polandg 2010–15 i 2.52 2.07 11.00 22.29 12.46 24.70
Romaniag 2010–15 i 0.06 1.14 4.25 9.71 4.26 10.27
Russian Federation 2010–15 c 1.62 0.48 9.29 21.84 10.07 22.36
Serbiag 2012–15 i –1.70 –0.88 4.69 12.04 4.45 11.72
Slovak Republicg 2010–15 i –0.62 –0.61 13.17 22.95 12.77 22.25
Sloveniag 2010–15 i –0.78 –0.56 21.12 34.70 20.31 33.74
Tajikistan 2009–15 c 2.30 3.58 2.69 5.13 3.08 6.34
Turkey 2011–16 c 2.53 3.47 6.45 15.73 7.30 18.66
Ukraine 2011–16 c –0.85 –0.69 7.34 11.90 7.03 11.50
Argentinae 2011–16 i 0.15 0.00 8.44 23.25 8.51 23.26
Bolivia 2011–16 i 1.67 1.06 4.07 12.56 4.42 13.24
Brazil 2011–15 i 3.80 2.19 4.77 17.66 5.54 19.25
Chile 2009–15 i 5.97 5.49 5.21 15.69 7.37 21.63
Colombia 2011–16 i 3.49 1.48 3.57 13.27 4.24 14.28
Costa Rica 2011–16 i 2.00 1.95 6.69 21.42 7.39 23.59
Dominican Republic 2011–16 i 4.46 3.53 4.22 12.54 5.24 14.92
Ecuador 2011–16 i 2.95 1.92 4.10 12.26 4.74 13.49
Honduras 2011–16 i 1.17 –1.95 2.15 9.13 2.28 8.28
Mexico 2010–14 i 0.51 0.74 3.88 11.41 3.96 11.75
Nicaragua 2009–14 i 5.64 6.52 2.94 7.90 3.87 10.83
(continued)
SHARED PROSPERITY: MIXED PROGRESS 63
TABLE 2B.1 Shared Prosperity Estimates, 91 Economies, circa 2010–15 (continued)
Annualized growth in Mean consumption or income per capitaa
mean consumption or
income per capitaa,b Initial year Most recent year
Total Total Total
Bottom 40 population Bottom 40 Population Bottom 40 Population
Economy Periodc Typed (%) (%) ($ a day, PPP) ($ a day, PPP) ($ a day, PPP) ($ a day, PPP)
Panama 2011–16 i 4.00 3.89 5.74 20.40 6.98 24.70
Peru 2011–16 i 3.08 2.18 4.11 12.04 4.79 13.41
Paraguay 2011–16 i 4.90 1.65 4.21 15.02 5.35 16.30
El Salvador 2011–16 i 4.08 2.93 3.46 8.86 4.22 10.23
Uruguay 2011–16 i 3.18 1.76 9.10 23.94 10.64 26.13
Egypt, Arab Rep. 2010–12 c 2.58 0.78 2.84 5.17 2.99 5.25
Iran, Islamic Rep. 2009–14 c 1.25 –1.27 6.60 17.42 7.02 16.34
West Bank and Gaza 2011–16 c –0.89 –0.55 5.30 10.84 5.03 10.50
Bangladesh 2010–16 c 1.35 1.54 1.88 3.52 2.03 3.86
Bhutan 2012–17 c 1.63 1.67 3.54 8.08 3.83 8.78
Sri Lanka 2012–16 c 4.80 5.28 3.37 7.51 3.98 8.99
Pakistan 2010–15 c 2.72 4.25 2.28 4.01 2.60 4.94
Burkina Faso 2009–14 c 5.84 2.93 1.04 2.39 1.38 2.76
Côte d’Ivoire 2008–15 c 0.74 –0.22 1.46 3.91 1.53 3.84
Ethiopia 2010–15 c 1.67 4.91 1.48 2.88 1.61 3.66
Mozambique 2008–14 c 1.52 5.36 0.72 1.96 0.78 2.65
Mauritania 2008–14 c 3.17 1.44 2.37 5.27 2.86 5.74
Namibia 2009–15 c 5.73 6.64 1.75 7.79 2.41 11.27
Niger 2011–14 c –0.06 3.26 1.27 2.35 1.27 2.59
Rwanda 2010–13 c 4.82 2.78 0.90 2.43 1.03 2.63
Togo 2011–15 c 2.76 0.82 0.89 2.63 0.99 2.71
Uganda 2012–16 c –2.15 –0.96 1.39 3.32 1.28 3.19
South Africa 2010–14 c –1.34 –1.23 2.12 11.80 1.99 11.11
Zambia 2010–15 c –0.59 2.93 0.68 2.59 0.66 2.99
Austriag 2010–15 i –0.47 –0.28 29.76 56.03 29.07 55.26
Belgiumg 2010–15 i 0.57 0.48 26.73 47.73 27.50 48.89
Canada 2010–13 I –0.24 0.83 27.36 55.97 27.16 57.37
Switzerlandg 2010–15 i 0.98 0.84 31.99 63.63 33.59 66.35
Cyprusg 2010–15 i –4.34 –3.04 27.05 50.63 21.66 43.38
Greeceg 2010–15 i –8.35 –6.98 14.56 31.08 9.41 21.65
Germany 2010–15 I –0.18 0.59 28.13 52.31 27.88 53.88
Denmarkg 2010–15 i 0.57 0.45 28.97 50.77 29.80 51.93
Spaing 2010–15 i –2.16 –1.53 17.74 39.51 15.90 36.58
Finlandg 2010–15 i 0.53 0.17 28.13 48.95 28.89 49.36
Franceg 2010–15 i 0.74 0.21 26.41 52.68 27.40 53.23
United Kingdomg 2010–15 i 0.26 0.11 22.00 46.34 22.29 46.60
Irelandg 2010–15 i 1.69 1.14 22.19 43.74 24.13 46.29
Icelandg 2009–14 i –0.13 –0.47 29.23 51.35 29.04 50.15
Italyg 2010–15 i –2.13 –1.08 19.88 42.44 17.85 40.19
Luxembourgg 2010–15 i –2.14 –0.44 36.83 70.80 33.04 69.24
Maltag 2010–15 i 3.57 3.48 19.49 35.76 23.22 42.43
Netherlandsg 2010–15 i 0.95 0.66 27.90 50.25 29.25 51.92
Norwayg 2010–15 i 2.11 2.95 36.54 61.31 40.57 70.92
Portugalg 2010–15 i –0.87 –0.74 13.11 27.85 12.55 26.84
Swedeng 2010–15 i 1.80 2.40 26.97 47.84 29.49 53.85
United States 2010–16 I 1.31 1.67 24.38 62.43 26.36 68.93
Source: GDSP (Global Database of Shared Prosperity), fall 2018, World Bank, Washington, DC, PovcalNet (online analysis tool), http://iresearch.worldbank.org/PovcalNet.
World Bank, Washington, DC.
Note: PPP = purchasing power parity.
a. Based on real mean per capita consumption or income measured at 2011 Purchasing Power Parity (PPP) using PovcalNet (http://iresearch.worldbank.org/PovcalNet).
b. The annualized growth rate is computed as (Mean in year 2/Mean in year 1)^(1/(Reference year 2 – Reference year 1)) – 1.
c. Refers to the year in which the underlying household survey data were collected and, in cases for which the data collection period bridged two calendar years, the first year
in which data were collected is reported. See appendix A for criteria in selecting shared prosperity periods.
d. Denotes whether the data reported are based on consumption (c) or income (i) data. Capital letters indicate that grouped data were used.
e. Covers urban areas only.
f. See Chen et al. (2018) for details on how the shared prosperity estimate for China is calculated.
g. Source from World Bank (forthcoming). “Living and Leaving. Housing, Mobility and Welfare in the European Union,” World Bank Regional Report.
64 POVERTY AND SHARED PROSPERITY 2018
TABLE 2B.2 Changes in Shared Prosperity, 67 Economies, circa 2008–13 to circa 2010–15
Economies, number Average SP
Higher SP in Lower SP in Average change
Region Total circa 2010–15 circa 2010–15 Circa 2008–13 Circa 2010–15 in SP
East Asia and Pacific 6 5 1 5.82 4.73 –1.09
Europe and Central Asia 22 12 10 1.51 2.41 0.90
Latin America and the Caribbean 14 4 10 4.56 3.21 –1.35
Middle East and North Africa 1 0 1 3.07 1.25 –1.82
South Asia 3 1 2 3.86 3.05 –0.81
Sub-Saharan Africa 1 0 1 4.09 –2.15 –6.24
Rest of the world 20 13 7 –1.10 –0.46 0.64
Total 67 35 32 1.92 1.87 –0.05
Source: GDSP (Global Database of Shared Prosperity), World Bank, Washington, DC, http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity.
Note: SP = shared prosperity; the indicator measures growth in the average consumption or income of the bottom 40. The 2008–13 release refers to the version included in
Poverty and Shared Prosperity 2016 (World Bank 2016b). The 2010–15 release refers to the version used in the present report. Regional and global averages of shared prosper-
ity refer to simple averages across country means.
TABLE 2B.3 Changes in the Shared Prosperity Premium, 67 Economies, circa 2008–13 to circa 2010–15
Economies, number Average SPP
Higher SPP in Lower SPP in Average change
Region Total circa 2010–15 circa 2010–15 Circa 2008–13 Circa 2010–15 in SPP
East Asia and Pacific 6 4 2 0.91 1.10 0.19
Europe and Central Asiaa 22 11 10 0.30 0.21 –0.09
Latin America and the Caribbean 14 4 10 1.51 1.20 –0.31
Middle East and North Africa 1 0 1 4.27 2.52 –1.75
South Asia 3 0 3 0.27 –0.69 –0.96
Sub-Saharan Africa 1 0 1 2.24 –1.19 –3.43
Rest of the world 20 7 13 –0.09 –0.32 –0.23
Total 67 26 40 0.58 0.31 –0.27
Source: GDSP (Global Database of Shared Prosperity), World Bank, Washington, DC, http://www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity.
Note: SPP = shared prosperity premium, which refers to the difference in the consumption or income growth of the bottom 40 and the mean of the country. The 2008–13 release
refers to the version included in Poverty and Shared Prosperity 2016 (World Bank 2016b). The 2010–15 release refers to the version covered in the present report. Regional and
global averages of shared prosperity refer to simple averages across country means.
a. The SPP for FYR Macedonia is the same for both circa 2010–15 and circa 2008–13.
SHARED PROSPERITY: MIXED PROGRESS 65
FIGURE 2B.1 The Shared Prosperity Premium, 91 Economies, by or consumption for measuring poverty and
Region or Income Classification changes over time, see the section on chapter 1
in appendix A. See also boxes 1.1 and 4.4 in
East Asia and Pacific World Bank (2016b).
Europe and Central Asia 2. Estimates for China are based on PovcalNet
Latin America and the Caribbean (see appendix A for further details).
Middle East and North Africa 3. The economies in fragile and conflict-affected
South Asia situations included are Côte d’Ivoire, Kosovo,
Sub-Saharan Africa Togo, and West Bank and Gaza.
Rest of the world
4. As of August 8, 2018, the World Bank consid-
ered that 83 economies exhibited moderate or
extreme data deprivation. Data deprivation
Fragile and conflict-affected
occurs if a country conducts fewer than two
household surveys in a 10-year period (Sera-
Low income juddin et al. 2015). Recognizing that the poor-
Lower-middle income est countries are more data challenged, the
Upper-middle income World Bank pledged in 2015 to help the poorest
High income countries improve the frequency of data collec-
tion to one household survey every three years.
0 10 20 30 40 50 60
5. A positive premium occurs in association with
Number of economies
a negative shared prosperity indicator in only
Positive premium Negative premium No shared prosperity measure
two cases, namely, Bosnia and Herzegovina and
Sources: GDSP (Global Database of Shared Prosperity), fall 2018, World Bank, Washington, DC, http:// Iceland. In these countries, the entire growth
www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity; PovcalNet (online distribution is negative, shared prosperity is
analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/. also negative though close to zero, and incomes
Note: The count is based on the 164 economies on which PovcalNet includes direct estimates of pov-
erty. Premium refers to the shared prosperity premium. “Positive premium” indicates that the income of among the top 60 are declining even more rap-
the bottom 40 grew at a faster rate than the average. “Negative premium” indicates that the incomes idly than the incomes of the bottom 40.
of the bottom 40 grew at a slower rate than the average in the country. “No shared prosperity measure” 6. The sample of economies in which shared
indicates that a poverty rate is reported in PovcalNet for the economy, but that the data are inadequate
for computing shared prosperity. prosperity can be measured in circa 2010–15
(13 of the 57 countries with poverty rates above
10 percent) is small, but similar conclusions
would be reached if older time spells for shared
Notes prosperity are considered—thus increasing the
1. Survey income and consumption are used coverage among economies with poverty rates
herein as equivalent aggregates. The assump- above 10 percent. Taking this expanded sample,
tion that they can be used interchangeably is a in the five countries with the highest level of
requirement of the global poverty and shared poverty at the US$1.90 a day poverty line, none
prosperity exercise given that country data of which is included in the present round on
are often available on only one or the other. shared prosperity, four have a negative shared
For more on the implications of using income prosperity and all have a negative premium.
66 POVERTY AND SHARED PROSPERITY 2018
Higher Standards 3
for a Growing World
This chapter presents two new sets of monetary poverty lines intended to complement the
international poverty line (IPL) of US$1.90 a day. First, two higher poverty lines, at US$3.20
and US$5.50 per day, are presented, reflecting typical national poverty thresholds in middle-
income countries. Second, the chapter introduces a global societal poverty line (SPL) reflecting
how monetary definitions of poverty at the national level vary with the overall income in a
society. The SPL counts individuals as poor if they are living either on less than the IPL or on
less than US$1.00 a day plus half the median value of consumption or income of their nation.
The two sets of complementary poverty lines enrich our understanding of global monetary
poverty. They reveal that global poverty rates are higher and being reduced more slowly than
is indicated by assessments using the IPL. Although only 10 percent of the world population
was living on less than US$1.90 per person per day in 2015, a quarter of the world was living
on less than US$3.20 per person per day, and close to half the world was living on less than
US$5.50 per person per day. The societal poverty rate declined by about a third between 1990
and 2015, dropping from approximately 45 percent to 28 percent. The chapter shows that the
elimination of monetary poverty, more broadly defined, is still a distant goal.
Introduction
In 2013, the World Bank set a target of re- in each country of how much someone needs
ducing extreme poverty as assessed by the to meet basic needs and live a life free of pov-
international poverty line (IPL) to less than erty. These national poverty lines came from
3 percent of the global population by 2030. some of the poorest countries in the world,
A frequent and important question posed and the US$1.90 value was an average of na-
when monitoring progress toward the goal tional poverty lines from 15 of these very poor
of ending poverty is whether the IPL, cur- countries (Ferreira et al. 2016). The inference
rently valued at US$1.90 in 2011 purchasing is that, if US$1.90 defines the cost of basic
power parity (PPP) U.S. dollars, is too severe needs in some of the poorest countries of the
a threshold for defining whether someone world, then it can be viewed as an absolute
is poor or not. Or, is US$1.90 per day really minimum threshold for defining poverty in all
enough to live a life free of extreme poverty? countries. This approach for setting the IPL is
One element of the answer involves exam- therefore guided by decisions made in some of
ining the reason this amount was initially se- the poorest countries of the world and, in this
lected. The value of the IPL was derived from way, respectful of national values and choices.
a set of national poverty lines—lines that re- In addition to reflecting national values
flected social and economic assessments made and choices, the IPL also has the desirable
67
attribute that it is fixed in real terms over time of the global population. By 2015, however,
and across countries. The value of the line only 9 percent of the global population was
will be regularly adjusted to reflect changing living in low-income countries (Fantom and
prices over time so that it maintains a con- Serajuddin 2016). Because most of the ex-
stant value through 2030 in each country of treme poor are now living in middle-income
the world. Fixing the real value of the IPL in countries, and most of the total population
this way ensures that the 3 percent by 2030 is in middle- and high-income countries, the
target will not be shifted to make it easier or use of average assessments of basic needs in
more difficult to reach. low-income countries is gradually becoming
Additionally, the value of the IPL is con- less relevant in many countries of the world.
verted into local currencies using the 2011 To address this concern in part, the World
PPP index to lock in corresponding amounts Bank has introduced a new set of poverty
of each local currency that can purchase lines that are higher in value and more rele-
approximately the same amount of basic vant to current economic conditions. Look-
goods within each country. Uniformity in ing beyond the IPL helps us better under-
purchasing power across countries is desir- stand what poverty means in different parts
able because it guarantees that the yardstick of the world. This chapter discusses two ways
of material well-being used in each country in which the World Bank will now also report
is comparable with the yardsticks used in on poverty, by assessing complementary pov-
all other countries. The comparable value erty lines that will help guide efforts to de-
of the line makes certain that, if individuals liver on the broader objective of establishing
are identified as poor in one country because a world free of poverty.
they are not able to acquire a basic bundle of
goods, they would also be identified as poor
in other countries if unable to purchase a Higher poverty lines for
similarly valued bundle of goods. everyone: US$3.20 and
“Measurable, time-bound goals are crucial US$5.50 a day
to focusing our work,” explains World Bank
President Jim Yong Kim (2016). The decision Although maintaining the value of the IPL
to fix the purchasing power of the IPL over fixed in real terms is essential to monitoring
time (up through 2030), and over all coun- progress toward achieving the 2030 poverty
tries of the world, ensures that the goal line target, recognizing that how countries and the
for this time-bound target is not changed. global community define poverty and basic
All of these attributes of the IPL have been needs can change is also imperative. “The ne-
persuasive in helping the global community cessities of life are not fixed” argues Townsend
reach agreement around the poverty goal. (1979, 915). “They are continuously being
The success of the IPL in fostering coordina- adapted and augmented as changes take place
tion in the international community on the in society and its products.”
issue of poverty is evident in the incorpora- To address the concern that the value of
tion of the IPL in first the Millennium Devel- the IPL could be viewed as too extreme for
opment Goals (MDGs) and now the Sustain- much of the world or that the necessities
able Development Goals (SDGs).1 of life are greater now than previously, the
Although the World Bank will continue to World Bank also uses poverty lines that are
focus on the 3 percent target as assessed by higher in value. The values of these lines have
the IPL, there are, nonetheless, reasonable been identified in a manner similar to the
concerns with the current valuation of the IPL, that is, they reflect social and economic
IPL. One source of concern is simply that, assessments made by governments; however,
when those national poverty lines were con- the assessments are more recent, and they are
structed for the 15 poor countries, 60 percent also produced in countries that are, on av-
of the global population was living in low- erage, richer than those upon which the IPL
income countries. The average value of these is based.
national poverty lines was meaningful for the These complementary lines reflect typical
vast majority of the poor and a large portion poverty assessments in lower-middle-income
68 POVERTY AND SHARED PROSPERITY 2018
countries (LMICs) and upper-middle-income TABLE 3.1 National Poverty Lines, circa 2011
countries (UMICs) in recent years.2 Specifi- Economy, income classification Median Mean
cally, the lines are the median values of LMIC
Low income 1.90 2.20
and UMIC national poverty lines in about Lower-middle income 3.20 3.90
2011 (Jolliffe and Prydz 2016). The value Upper-middle income 5.50 5.60
of the poverty line based on assessments High income 21.70 21.20
of needs in LMICs is US$3.20 per person
Source: Jolliffe and Prydz 2016.
per day expressed in 2011 PPP U.S. dollars, Note: Values are rounded to nearest 0.10. Economies are classified on the basis of official World Bank
whereas the value of the line based on typi- income classifications, which rely on measures of per capita gross national income. Estimates are based
cal basic needs in UMICs is US$5.50 (table on national poverty lines in 126 economies. The selected poverty line for each economy is the one that is
closest in time to 2011.
3.1). Although these lines may sometimes
be referred to as LMIC and UMIC lines, this
does not mean that, for example, the LMIC Table 3.2 shows the change since 1990 in
line can be applied only in the case of LMICs. the proportion of people living on less than
The two poverty lines simply offer higher val- US$3.20 or less than US$5.50 a day. The find-
ues that reflect assessments of basic needs in ings illustrated in the table suggest that the
these two groups of countries. (The values success in reducing extreme poverty has not
are based on a large database of harmonized been completely matched by reductions in the
national poverty lines in about 2011; see ap- relative size of the population living on less
pendix A for details.) than these higher-valued lines. Like the MDG
As with the IPL, the intention is that the of halving extreme poverty as measured by the
value of these LMIC and UMIC lines will IPL, the proportion of people living on less
remain fixed in real terms, thereby allowing than US$3.20 a day was also halved between
poverty reduction to be monitored also at 1990 and 2015. However, in contrast to the
higher global poverty lines.3 Because they are MDG, which was met about six years ahead
complementary lines based on more recent of schedule, the proportion of people living
social assessments of basic needs, the lines on less than US$3.20 was only halved by 2014,
will maintain greater relevance as poverty re- five years after the MDG target was reached.
duction is monitored over the next 15 years. Measured according to the US$5.50 line, the
The decision to use social assessments from success in improving the well-being of people
middle-income countries also reflects the living in poverty must be additionally tem-
overall growth in the global economy. Using pered. In 1990, approximately two-thirds of
LMIC and UMIC median national poverty the population of the world was living on less
lines as the basis for the complementary lines than US$5.50 a day. By 2015, this proportion
means that these new lines better reflect the had fallen, but it had not been halved. Slightly
situations in countries that are home to most less than half (46 percent) of the world was
of the global population and most of the still living on less than US$5.50 a day in 2015.
global poor. Figure 3.1, panel a, illustrates why the rate
Chapter 1 in this report shows the tre- at which extreme poverty is being reduced is
mendous progress the world has made in re- not matched by reductions in the share of the
ducing extreme poverty as measured by the world population living on less than US$3.20
IPL. As one remarkable example, target 1.A or US$5.50. In 1990, there was a concentra-
of MDG 1, to cut the poverty rate of 1990 in tion of people who were consuming just less
half by 2015, was reached approximately six than the US$1.90 threshold, as revealed by the
years ahead of schedule. This is true whether distribution peaking to the left of this value.4
we examine the global poverty rate or the Although one-third of the world’s population
global poverty rate less several high-income consumed less than US$1.90, most of those
countries. This extraordinary success allows people consumed at rates between US$1.00
us to broaden our focus to ensure that those and US$1.90. Economic development shifted
people who may not be poor as measured by the distribution to the right, moving the
the IPL, but who struggle nonetheless to sat- hump over the US$1.90 threshold, leading to a
isfy their basic needs, also benefit from eco- rapid reduction in the number of people con-
nomic development. suming less than US$1.90. In contrast, panel
HIGHER STANDARDS FOR A GROWING WORLD 69
TABLE 3.2 Poverty at Higher Poverty Lines, US$3.20 and US$5.50 (2011 PPP)
a. Poverty rate by region at US$3.20 (%)
Percentage point
Region(s) 1990 1999 2008 2013 2015 change, 1990–2015
East Asia and Pacific 85.3 67.1 37.4 17.5 12.5 −72.8
Europe and Central Asia 9.9a 21.1 7.5 5.7 5.4 −4.6
Latin America and the Caribbean 28.3 27 15.7 11.4 10.8 −17.5
Middle East and North Africa 26.8 21.7 16.7 14.4 16.3 −10.5
South Asia 81.7 76a 67.9 53.9 48.6a −33.1
Sub-Saharan Africa 74.9 78.3 72.2 67.8 66.3 −8.6
Sum of regions 66.4 60.1 45 33.7 30.7 −35.7
Rest of the world 0.8 0.8 0.7 0.8 0.9 0.1
World 55.1 50.6 38.2 28.8 26.3 −28.9
b. Poverty rate by region at US$5.50 (%)
Percentage point
Region(s) 1990 1999 2008 2013 2015 change, 1990–2015
East Asia and Pacific 95.2 87 63.6 42.4 34.9 −60.3
Europe and Central Asia 25.3a 44.5 17.1 14.1 14 −11.3
Latin America and the Caribbean 48.6 47 33.3 27.2 26.4 −22.2
Middle East and North Africa 58.8 54.5 46.6 42.3 42.5 −16.3
South Asia 95.3 93.1a 89.8 84.2 81.4a −14
Sub-Saharan Africa 88.5 90.5 88.1 85.4 84.5 −4.1
Sum of regions 80.5 79.3 66.5 57 53.7 −26.7
Rest of the world 1.7 1.3 1.2 1.5 1.5 −0.2
World 67 66.8 56.5 48.7 46 −21
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: The criteria for estimating survey population coverage is whether at least one survey used in the reference year estimate
was conducted within two years of the reference year. PPP = purchasing power parity.
a. This estimate is based on less than 40 percent of regional population coverage.
a shows that a significantly smaller share of with significantly fewer people now living
people was living on more than US$1.90 but below the $1.90 threshold, future growth will
less than US$3.20. So the economic growth not lift as many people past this threshold
that led to a rapid reduction in extreme pov- as previously experienced. Thus, the reduc-
erty could not carry as many people above tion in extreme poverty will be tempered, al-
the US$3.20 threshold. This narrative is sim- though the potential for progress in reducing
ilar in the case of the US$5.50 line: economic the share of the world’s population living on
growth carried significantly fewer people past less than US$5.50 a day will be significant.
the US$5.50 threshold. This reinforces the conclusion in chapter 1
The global distribution of consumption that the slowdown in the rate of decline of
for 2015 offers useful insights into what one extreme poverty will likely continue.
may expect in the near future (as illustrated In addition to providing insight on the po-
by the histogram in figure 3.1, panel b). In tential for global poverty reduction in the near-
2015, the peak in the consumption distribu- term future, these higher lines also present
tion had shifted to the right and is now be- clear regional differences in the profile of the
tween US$3.20 and US$5.50. Only about 10 people living in extreme poverty or nearly so.
percent of the global population is still living The countries in East Asia and Pacific not only
on less than US$1.90 a day. An implication of had the largest reductions in extreme poverty,
this is that growth in the near future will shift but they also experienced the largest reductions
the distribution further to the right, leading in the proportion of people living on less than
to a rapid reduction in the share of people US$3.20 and US$5.50 (figure 3.1, panels c and
living on less than US$5.50 a day. However, d). Between 1990 and 2015, the proportion of
70 POVERTY AND SHARED PROSPERITY 2018
FIGURE 3.1 Consumption and Income Distributions, 1990 and 2015
a. World, 1990 b. World, 2015
1,500 1,500
Population (millions)
Population (millions)
1,000 1,000
500 500
0 0
1− 1
−1
.5
60
60
1.9 .9
1.9 .9
.5
20 0
40 0
80 80
80 80
20 0
40 0
10 0
10 0
3.2 3.2
5.5 .5
60
5.5 .5
60
3.2 .2
−
−2
−4
−2
−4
−1
−1
<0
1
1
<0
0.5
0.5
−5
−3
−5
>1
>1
−
−
−1
−1
1−
−
Consumption/income per day (2011 US$ PPP) Consumption/income per day (2011 US$ PPP)
c. East Asia and Pacific, 1990 d. East Asia and Pacific, 2015
800 800
Population (millions)
600 Population (millions) 600
400 400
200 200
0 0
−1
60
60
.5
1.9 .9
1.9 .9
−1
20 0
40 0
80 80
20 0
40 0
80 80
10 0
10 0
3.2 .2
5.5 .5
60
3.2 .2
5.5 .5
60
.5
−2
−4
−2
−4
−1
−1
<0
1
1
0.5
−3
−5
−3
−5
<0
>1
>1
0.5
−
−
−1
−1
1−
1−
Consumption/income per day (2011 US$ PPP) Consumption/income per day (2011 US$ PPP)
e. South Asia, 1990 f. South Asia, 2015a
600 600
Population (millions)
Population (millions)
400 400
200 200
0 0
−1
60
60
1.9 .9
1.9 .9
20 0
40 0
80 80
20 0
40 0
80 80
1− 1
.5
10 0
10 0
3.2 3.2
5.5 .5
60
3.2 .2
5.5 .5
60
.5
−2
−4
−
−2
−4
−1
−1
1
1
<0
0.5
−5
−3
−5
>1
>1
<0
−
−
0.5
−1
−1
1−
−
Consumption/income per day (2011 US$ PPP) Consumption/income per day (2011 US$ PPP)
g. Sub-Saharan Africa, 1990 h. Sub-Saharan Africa, 2015
250 250
Population (millions)
Population (millions)
200 200
150 150
100 100
50 50
0 0
−1
.5
60
60
1.9 .9
1.9 .9
−1
80 80
20 0
40 0
20 0
40 0
80 80
10 0
10 0
60
3.2 .2
5.5 .5
3.2 .2
5.5 .5
60
.5
−2
−4
−2
−4
−1
−1
<0
0.5
1
1
−3
−5
−3
−5
>1
>1
<0
0.5
−
−
−1
−1
1−
1−
Consumption/income per day (2011 US$ PPP) Consumption/income per day (2011 US$ PPP)
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: Bins were purposely selected to highlight US$1.90, US$3.20, and US$5.50 poverty lines. The size of the selected bins produces a
histogram that approximates the shape of the estimated density function of the log of income/consumption.
a. This estimate is based on less than 40 percent of regional population coverage.
HIGHER STANDARDS FOR A GROWING WORLD 71
people living on less than each of these three Higher lines tailored to
thresholds declined by nearly 60 percentage
country circumstances:
points. This can be seen in panels c and d in
the large rightward shift of the distribution Societal poverty
between 1990 and 2015. This massive prog- The second set of complementary poverty
ress over every threshold was experienced only lines the World Bank is now reporting are tai-
in East Asia and Pacific. In the other regions, lored to the specific levels of economic devel-
progress in reducing poverty at the various opment of each country and are designed to
thresholds has been much more modest. measure societal poverty. The introduction
Figure 3.1, panel e, reveals that in South Asia of this measure is in direct response to rec-
the peak of the consumption distribution was ommendations of the Commission on Global
slightly below US$1.90 in 1990. By 2015, most Poverty, led by Professor Sir A. B. Atkinson,
people now lived on more than US$1.90 but to “introduce a societal head count measure
less than US$3.20 (figure 3.1, panel f). There of global consumption poverty that takes
was a large decline—35 percentage points—in account, above an appropriate level, of the
the share of people living on less than US$1.90. standard of living in the country in question,
There was also a decline (60 percent) in the thus combining fixed and relative elements of
number of people living below US$1.90 (table poverty” (World Bank 2017, xxi).
1A.1). The story for South Asia changes, how- A key attribute of the IPL is that it is con-
ever, when we examine the US$3.20 poverty verted into local currencies using the 2011
threshold. The percentage of the total pop- PPP U.S. dollars to ensure that the value of
ulation living below this threshold declined the line reflects approximately the same pur-
substantially over this time, but because of chasing power in all countries (see earlier
a growing population, the number of people discussion). If an individual who is able to
living on less than US$3.20 declined by only 8 buy US$2.00 worth of goods in one country
percent over this 25-year period. In contrast to each day is not considered poor, then an in-
East Asia where the peak of the distribution es- dividual who is able to consume at that same
sentially shifted past the US$5.50 threshold, in level in another country will also not be poor.
South Asia the peak of the distribution of con- Everyone is assessed by the same standard re-
sumption essentially shifted from just below gardless of where they live. This guiding prin-
US$1.90 to just below US$3.20. ciple of the monitoring of extreme poverty
In the case of Sub-Saharan Africa (fig- ensures that the material well-being of people
ure 3.1, panels g and h), the distribution has can be assessed and compared meaningfully
shifted rightward only very slightly. Although across the world.
chapter 1 reported that extreme poverty Although ensuring equality in the yard-
declined by 13 percentage points in Sub- stick of poverty is desirable, there are some
Saharan Africa between 1990 and 2015, panel trade-offs in making this choice. One trade-
d reveals that the peak of the consumption off in particular helped guide the World Bank
distribution was essentially around US$1.90 toward the development of a new comple-
in both 1990 and 2015. The decline in the mentary poverty line, the societal poverty
prevalence of extreme poverty coincided with line (SPL). Fixing the value of the line in
nearly a 50 percent increase in the number of constant PPP terms across all countries en-
people living in extreme poverty during this sures that the bundle of goods that can be
time period. Overall, the population of Sub- purchased is the same. As economies grow,
Saharan Africa nearly doubled in this time however, this bundle is becoming a less use-
period, with one of the largest increases in ful indicator of basic needs in many places.
population being for those living on less than For example, in 2015, the extreme poverty
US$3.20 and more than US$1.90. Economic rate was less than 3 percent in more than half
growth slightly outpaced population growth the 164 countries in which the World Bank
resulting in a distribution of consumption monitors extreme poverty; and the majority
that shifted only slightly to the right but grew of the world no longer lives in low-income
significantly larger, reflecting the near dou- economies. For many countries, the social
bling of the population.
72 POVERTY AND SHARED PROSPERITY 2018
relevance of the IPL has lessened over time on a relative notion of the poverty line re-
as their economies have grown. This is largely volves around the fact that participation in
due to the observance that needs change as society with dignity may require more goods
the world becomes richer (Townsend 1979). in a richer country than in a poorer coun-
A very closely related point is that, as try. Social participation might thus be more
countries grow richer, uniformity in the con- closely related to the concept of meeting
sumption bundle may not result in the same basic needs in the poorest of countries, but
level of well-being everywhere. Carrying out in richer countries the ability to participate in
basic functions of life might require more society might be costlier.
goods in some countries than in others, and This conceptual point, that the very defini-
fixing the consumption bundle could result tion of basic needs in terms of goods and ser-
in unequal assessment of people across the vices may vary across countries, appears to be
world in terms of their ability to function empirically supported. Figure 3.2 shows that
in society in a socially acceptable manner. there is significant variation across countries
Another way to express this is that ensuring in how basic needs are defined, as expressed
equality across countries in terms of carrying in national poverty lines. The analysis in the
out the same basic functions of life in each figure is based on 699 estimated national
society may result in a poverty line that takes poverty lines—all of which are expressed in
different monetary values (Sen 1983). For ex- comparable purchasing power terms. It re-
ample, participating in the labor market may veals a strong positive correlation between the
be viewed as a minimal social functioning; median level of consumption in each country
the cost of this functioning, however, may and the assessment of basic needs. Analysis
require only clothing and food in a poor soci- on a different set of national poverty lines has
ety, whereas in a richer society it may require similarly shown that the values of absolute
access to the internet, transportation, and a national poverty lines range across countries
cell phone, in addition to clothing and food. from US$0.63 a day to more than US$9.00 a
Another example that more directly builds day (in 2005 PPP U.S. dollars) and that higher
FIGURE 3.2 National Poverty Lines and Economic Development
a. Level scale b. Log scale
20 20
Poverty line (2011 US$ PPP)
Poverty line (2011 US$ PPP)
10
15
5
10
2
5
1
0 0
0 10 20 30 0 1 2 5 10 20 40
Median consumption/income per day Median consumption/income per day
(2011 US$ PPP) (2011 US$ PPP)
Full sample 2011 sample Predicted 10th and 90th percentiles
Source: Jolliffe and Prydz 2016.
Note: Both panels plot 699 harmonized national poverty lines. Dark dots indicate the 104 poverty lines that are closest to 2011 (one
unique line for each country), excluding lines prior to 2000. Both panels plot the same data. Panel a plots the lines on actual values. Panel
b plots these same values, but the axis values of the plots are log transformations. Lines in panel b are predicted (conditional bivariate)
10th and 90th percentile lines. All axis values are expressed in 2011 purchasing power parity (PPP) U.S. dollars.
HIGHER STANDARDS FOR A GROWING WORLD 73
poverty lines correspond to relatively more of median consumption (or income) per day
well-off economies (Ravallion 2010). in that country. If US$1.00 plus half the me-
This finding is not merely a cross- dian consumption is less than the IPL, then
sectional association. If the definition of pov- the SPL is equal to the IPL. In many countries,
erty changes as countries grow richer on aver- this value is greater than US$1.90, and this
age, national poverty lines should be changing greater value then becomes the SPL. More for-
in real terms over time. This is indeed what mally, the SPL adopted by the World Bank is
is observed. A few specific examples follow. In calculated in 2011 PPP U.S. dollars as follows:
2011, the government of India raised the real
SPL = max (US$1.90, US$1.00
value of the urban poverty line by more that
+ 0.5 ןmedian consumption).7 (3.1)
40 percent, increasing it from Rs 33 to Rs 47
per person per day. The change in rural pov- For example, in a country in which the
erty lines was significantly less, about 19 per- median consumption per person is US$1.60
cent, increasing from Rs 27 to Rs 32. At about per day, the IPL is greater than US$1.00
this time, China increased the real value of the plus half of US$1.60, so the value of the SPL
rural poverty line by more than 75 percent is US$1.90.8 Alternatively, in a country in
(Addison and Niño-Zarazúa 2012). Many which the median consumption is US$3.00
governments have increased the real value per day, the SPL is US$2.50 (US$1.00 +
of national poverty lines in recognition that 0.5 ןUS$3.00). In defining societal poverty
their economies have grown so significantly in this way, Jolliffe and Prydz (2017), build
that the concept of basic needs has changed on the important contributions of Atkinson
fundamentally. After 15 years of keeping the and Bourguignon (2001), Chen and Raval-
real value of the national poverty line con- lion (2013), Foster (1998), and Ravallion and
stant, the government of Nepal raised the real Chen (2011).
value of its poverty line in 2011 (CBS 2012). By this definition, societal poverty rep-
Similarly, the government of Jordan increased resents a combination of extreme poverty,
the real value of the poverty line by about 10 which is fixed in value for everyone, and a
percent in 2011 (Jolliffe and Serajuddin 2018; relative dimension of well-being that differs
World Bank 2009).5 Absolute national poverty in every country depending on the median
lines are behaving like relative poverty lines in level of consumption in that country. Figure
that they are becoming higher as countries get 3.3 illustrates how the SPL changes as the
richer. “It can be agreed that a sustained in- median consumption in a country increases.
crease in average living standards is likely to In countries with low median consumption
lead eventually to more generous perceptions (less than US$1.80 per person per day), a rise
of what ‘poverty’ means in a given society,” in median consumption does not change the
notes Ravallion (1998, 29). SPL. Indeed, the SPL has the same value as
the IPL in all countries with median con-
sumption at less than US$1.80. However, as
Characteristics of the societal
countries with median consumption at more
poverty line
than US$1.80 become richer, and the median
To reflect this viewpoint, the World Bank consumption increases, the value of the SPL
will now initiate reporting on societal pov- also rises. The climbing cost of social partic-
erty, which is based on a poverty line that ipation as the economy grows is reflected in
is adjusted for the median level of well- the positive slope of the line.
being in each country.6 First, according to The slope of one-half, the rate at which
the definition of societal poverty used by the the SPL is rising as countries become richer,
World Bank, individuals living in extreme comes from the empirical association ob-
poverty as measured by the IPL are also suf- served between national poverty lines and
fering from societal poverty. Second, the new different measures of overall consumption
measure considers that individuals are suffer- in society. It indicates that, on average, the
ing from societal poverty if they are living on national poverty lines are increasing at a
less than US$1.00 a day plus half of the value rate equal to half the median consumption
74 POVERTY AND SHARED PROSPERITY 2018
FIGURE 3.3 Societal Poverty Line for country poverty rates.10 Similarly, Eu-
ropean countries typically set national pov-
erty thresholds at 50 percent or 60 percent
of median disposable household income
Societal poverty line
(Vecchi 2015). The gradient of 50 percent
coincides with SDG indicator 10.2.1 on in-
Slope = 0.5 equality, namely, the proportion of peo-
ple living below 50 percent of the median
income, by sex, age, and disability status.11
Similarly, the intercept of US$1.00 per
person per day in 2011 PPP U.S. dollars cor-
$1.90 responds in value with some relevant empir-
ical findings. Ravallion (2016) estimates an
$1.00
empirical lower bound on consumption in
part to address the issue of how to monitor
the concept of leaving no one behind. His
0 analysis indicates that the value of this con-
Median national consumption (or income) per day sumption floor is US$0.67 in 2005 PPP U.S.
Source: Jolliffe and Prydz 2017. dollars, which is US$1.00 after conversion to
Note: The lower bound is equal to the international poverty line, 2011 PPP.12 There are also analyses that aim
which is currently valued at US$1.90 in 2011 purchasing power to estimate minimum biological needs—a
parity U.S. dollars. The slope is equal to 0.5. The intercept is
US$1.00. The kink point in the figure is at a median national con- concept that differs significantly from socially
sumption or income of US$1.80. acceptable ways of meeting basic needs. The
value of these minimum needs tends to be
about US$1.00 (Lindgren 2015).13
in the countries. The slope of one-half and The SPL is estimated by first extracting the
the intercept of US$1.00 are the values that median level of daily per capita consumption
most closely fit the data provided by the na- (or income) for each national distribution
tional poverty lines and overall consumption from PovcalNet, then following the formula
in each country. This observed relationship in equation (3.1) to derive a set of country-
between national poverty lines and national specific values of the SPL.14 If this value is
well-being determines the formula for mea- greater than US$1.90, the SPL is passed to
suring societal poverty.9 In an important PovcalNet, which reports the poverty rate as-
sense, the SPL and the IPL share the same sociated with this line. This rate is the societal
empirical underpinning. Both are anchored poverty rate. (If the SPL < US$1.90, then so-
in the distribution of national poverty lines, cietal poverty is simply the same as extreme
which represent countries’ own judgements poverty estimated in chapter 1.)
of what poverty means for them. Whereas the By design, the SPL rises with growth. The
IPL focuses narrowly—and deliberately—on population-weighted average SPL across all
the choices of some of the poorest countries, countries increased from US$5.30 in 1990 to
the SPL is built on information from across about US$6.90 in 2015, reflecting the steady,
the whole range of levels of development. global growth in real median consumption
In addition to fitting the data well, the during that time. The SPL growth rate has
slope coefficient of half the median is widely been much stronger in higher-income coun-
used by many countries and organizations as tries. Among today’s UMICs, the mean SPL
a measure of relative poverty and inclusion. nearly doubled over the same time period, ris-
In the academic literature on poverty, this ing from US$3.00 in 1990 to US$5.80 in 2015.
slope has been a subject of discussion for a In contrast, the average SPL only slightly in-
long time, and, in policy, the Organisation creased in value in low-income countries over
for Economic Co-operation and Develop- this period—in large part because of changes
ment uses 50 percent of median household in country composition of these income
income as the headline poverty indicator categories.
HIGHER STANDARDS FOR A GROWING WORLD 75
TABLE 3.3 Average Societal Poverty Lines, by Region and Income Classification, 1990–2015
Percentage point
a. Region(s) 1990 1999 2008 2013 2015 change, 1990–2015
East Asia and Pacific 2.0 2.2 3.2 4.3 4.8 2.8
Europe and Central Asia 5.9a 4.4 7.1 7.8 7.6 1.8
Latin America and the Caribbean 3.9 4.1 5.2 6.1 6.1 2.2
Middle East and North Africa 3.6 3.8 4.5 4.7 4.6 1.0
South Asia 2.0 2.1a 2.2 2.5 2.6a 0.6
Sub-Saharan Africa 2.1 2.1 2.2 2.3 2.3 0.2
Sum of regions 2.7 2.7 3.4 4.0 4.1 1.4
Rest of the world 17.8 19.8 22.1 22.0 22.8 5.0
World 5.3 5.4 6.2 6.7 6.9 1.6
Percentage point
b. Income group 1990 1999 2008 2013 2015 change, 1990–2015
Low income 2.1 2.1 2.1 2.2 2.2 0.1
Lower-middle income 2.2 2.2 2.5 2.8 2.9 0.7
Upper-middle income 3.0 3.0 4.4 5.4 5.8 2.8
High income 16.4 18.2 20.4 20.5 21.2 4.8
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: The table presents (population-weighted) average of the value of country societal poverty lines, evaluated at US$1.00 + 50 per-
cent ןmedian consumption (or income) with a lower bound of US$1.90 (2011 PPP). Current (2018) World Bank income classifications have
been used. The criteria for estimating survey population coverage is whether at least one survey used in the reference year estimate was
conducted within two years of the reference year. PPP = purchasing power parity.
a. This estimate is based on less than 40 percent of regional population coverage.
Table 3.3 reveals significant differences in rate of societal poverty, as measured by the
the pattern of the regional growth of the SPL. SPL. It also displays the count and rate of abso-
For example, the mean SPL in South Asia, lute extreme poverty as measured by the IPL of
East Asia and Pacific, and Sub-Saharan Africa US$1.90 a day. The first striking aspect of the
in 1990 was just slightly higher than the IPL of figure is that, although the total count of peo-
US$1.90. Because of strong economic growth ple living in extreme poverty has declined rap-
in East Asia and Pacific, the mean line more idly, the number of people who are identified
than doubled, to US$4.80 per day in 2015. In as societally poor has largely stayed the same
contrast, in Sub-Saharan Africa, which has ex- over the 25 years, between 1990 and 2015.
perienced much weaker overall growth, there In contrast, the share of the global popu-
has been little change in the value of the SPL, lation that is societally poor has fallen steadily
increasing only by $0.20 since 1990. since 1990, but at a much slower pace than the
decline in extreme poverty (figure 3.4, panel
a). This divergence in the rate of decline am-
Profile of societal poverty
plifies the distinction between the two mea-
Global counts of extreme poverty are based sures. Table 3.4 shows that, in 1990, the societal
on data from PovcalNet (described in appen- poverty rate, at 44.5 percent, was estimated
dix A), and so too are the estimates of societal at about 9 percentage points higher than the
poverty presented in this chapter.15 Using the extreme poverty rate (35.9 percent, as seen in
country-specific SPL and following the same figure 3.4, panel a). By 2015, the gap between
aggregation and lining-up methods as in the societal and extreme poverty, in terms of the
case of the extreme poverty estimates reported percentage point difference (18.4), had more
in chapter 1, the estimated societal poverty than doubled. In a growing global economy,
headcount was approximately 2.1 billion peo- this divergence is an expected outcome, and
ple in 2015.16 This is almost three times more the magnitude of the change in the difference
than the global count of people living on less in the rates over the decades highlights the
than US$1.90 a day, which was estimated at distinction in the informational content in
approximately 736 million in 2015. Figure 3.4 these measures. In the 1980s and early 1990s,
displays the change in both the count and the the societal poverty rate and the extreme pov-
76 POVERTY AND SHARED PROSPERITY 2018
FIGURE 3.4 Societal and Extreme Poverty, Global Estimates, 1990–2015
a. Poverty rate b. Number of poor
50 2,500
40 2,000
Poverty rate (%)
Millions
30 1,500
20 1,000
10 500
1990 1995 2000 2005 2010 2015 1990 1995 2000 2005 2010 2015
Societal poverty Extreme poverty
Note: Panel a shows the rate of extreme poverty based on the international poverty line (US$1.90, 2011 PPP) and societal poverty based
on the societal poverty line. Panel b shows the corresponding number of people who are poor by both lines. PPP = purchasing power parity.
erty rate were largely similar concepts because Similar to the case of regional profiles of
most of the world population was living in absolute poverty, Sub-Saharan Africa stands
countries with low median national consump- out because of the substantially higher rates of
tion, whereby the IPL and the SPL were either societal poverty. Although the societal poverty
identical or close in value. They largely por- rate has declined 9 percentage points over the
trayed the same picture of poverty. But now, as last 25 years in Sub-Saharan Africa, the overall
countries have grown richer, and median con- rate is still almost half the population, 49 per-
sumption is above US$1.80 in many countries cent, in 2015. In contrast, societal poverty had
of the world, the SPL is capturing significantly dropped 38 percentage points in the East Asia
more information about the distributional as- and Pacific region, reducing by more than half
pects of growth. the rate of 63.4 percent in 1990, to 25.1 per-
TABLE 3.4 Societal Poverty Rates, 1990–2015
Percent
Percentage point
a. Region(s) 1990 1999 2008 2013 2015 change, 1990–2015
East Asia and Pacific 63.4 46.6 34.7 27.2 25.1 −38.3
Europe and Central Asia 22.2a 27.0 19.4 17.7 17.3 −4.9
Latin America and the Caribbean 33.9 34.0 29.4 27.5 26.9 −7.0
Middle East and North Africa 28.6 26.6 23.7 21.5 22.9 −5.7
South Asia 51.0 46.9a 42.0 35.4 32.9a −18.0
Sub-Saharan Africa 57.9 61.2 53.3 49.9 49.0 −9.0
Sum of regions 50.6 44.3 37.0 31.9 30.6 −20.0
Rest of the world 15.5 15.2 15.4 16.0 16.0 0.5
World 44.5 39.7 33.7 29.6 28.4 −16.1
Percentage point
b. Income group 1990 1999 2008 2013 2015 change, 1990–2015
Low income 63.6 65.0 55.6 51.4 51.2 −12.3
Lower-middle income 50.5 46.7 40.3 34.9 32.9 −17.6
Upper-middle income 50.8 39.7 30.4 24.7 23.5 −27.3
High income 15.8 15.8 15.9 16.4 16.3 0.5
Source: PovcalNet (online analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
Note: World Bank income classifications are current as of 2018. Change is measured in percentage points (pp). “Sum of regions” was
previously referred to as “developing world” for which PovcalNet monitors poverty.
a. The criteria for estimating survey population coverage is whether at least one survey used in the reference year estimate was
conducted within two years of the reference year.
HIGHER STANDARDS FOR A GROWING WORLD 77
cent in 2015. All developing regions have seen of the SPL increases in percentage terms at
an overall decline in societal poverty rates a rate that is slower than the percentage in-
since 1990, especially during the 2000s. In crease in economic growth. This means that,
contrast societal poverty has been stubbornly if median consumption doubles, the SPL in-
static, at about 16.0 percent in aggregate, in creases, but by an amount less than double.
the mainly high-income countries in the “rest Because the percent increase in the SPL
of the world” category, though remaining will always be less than the percent increase
lower than in all the developing regions. in median consumption, distribution-
A similar pattern emerges in the lower neutral growth will reduce societal poverty.
half of table 3.4, which presents societal pov- By construction, the percentage increase in
erty rates by country income classifications. the SPL in response to a percentage increase
Countries are shown in their income classifi- in median consumption differs among rich
cation as of 2018. So a country identified as a countries relative to poor countries. For the
UMIC in 2018 was not necessarily a UMIC in poorest countries, among which median con-
1990. It might have grown economically into sumption is less than US$1.80 a day, growth
that classification, and this happened often. in median consumption does not change
Partly for this reason, the largest declines in the value of the SPL. If a country’s median
societal poverty occurred among UMICs. consumption grows sufficiently and crosses
The countries classified as UMICs in 2018 the US$1.80 kink point, then the SPL will
had realized some of the highest economic increase slightly (see figure 3.1). Figure 3.5
growth rates over the preceding 25 years. shows that for a typical country that has
The analysis of societal poverty by in- reached high-income status, that is, median
come classification confounds two issues. consumption around US$40 a day, the SPL
Economic growth is an important engine of rises at a percentage rate that is nearly equal
poverty reduction, but growth alone is a less to the percentage increase in median con-
effective vehicle for reducing societal poverty sumption. For the richest of countries, dou-
if a country is already in the higher-income bling median consumption nearly doubles
category. This is because societal poverty is the value of the SPL. In contrast, increasing
a hybrid concept that mixes elements of ab- the median consumption for countries whose
solute and relative poverty (Foster 1998). An median consumption is less than US$1.80
implication of this hybrid concept (more has no effect on the SPL if the SPL has less
specifically, the lower bound at the IPL and value than the IPL.
the positive intercept at one) is that the value An alternative way to interpret this is that,
among low-income countries, improvements
FIGURE 3.5 Change in the Societal Poverty Line from Growth in societal poverty are highly correlated with
improvements in extreme poverty; in fact,
1.0 they are identical in the poorest countries.
Among high-income countries, the shared
0.8 prosperity premium is highly correlated
with reductions in societal poverty. Positive
0.6 shared prosperity, combined with a shared
Elasticity
prosperity premium, indicates that a country
0.4 is growing and that the poorest in the coun-
try are benefitting more from this growth.
0.2 In high-income countries, this is precisely
what is needed to reduce societal poverty.
0 In this way, societal poverty combines infor-
1 2 5 10 20 40 mation about reductions in extreme poverty
Median income/consumption (2011 US$ PPP, log scale) (discussed in chapter 1) and the notions of
Societal poverty line, elasticity Fixed share of societal poverty line shared prosperity and the shared prosperity
premium (discussed in chapter 2).
Note: Vertical lines indicate the average national median consumption or income in 2013 for World
Bank income classification groupings (from left to right): low-income (US$2.1/day), lower-middle-income Figure 3.6 illustrates this by displaying
(US$3.7), upper-middle-income (US$9.3), and high-income (US$40) countries. PPP = purchasing power parity. the case of two UMICs, Costa Rica and Ec-
78 POVERTY AND SHARED PROSPERITY 2018
FIGURE 3.6 Societal Poverty and Shared shared prosperity premium and the reduc-
Prosperity in Costa Rica and Ecuador tion in societal poverty, but at a lower level
(about 0.4). Improvement in societal pov-
28 erty in low-income countries is driven much
more by reductions in extreme poverty.
Because societal poverty is a hybrid of abso-
27 lute and relative poverty concepts, it provides
a natural bridge between the dual goals of re-
Poverty rate (%)
ducing extreme poverty and increasing shared
prosperity. Among the poorest countries, the
26
value of the SPL is primarily determined by
the IPL, and policies that promote reductions
in extreme poverty will be the same as policies
25 that reduce societal poverty. As countries be-
come wealthier, the SPL is increasingly deter-
mined by the relative component of the pov-
24 erty line, which means that policies that focus
2011 2012 2013 2014 2015 2016 on raising the shared prosperity premium—
the difference between the growth rate of the
Costa Rica Ecuador
bottom 40 and the average growth rate in a
Note: The figure shows the decline in societal poverty for Ecuador country—will be more effective in reducing
and Costa Rica over a time period where both countries had sim- societal poverty than policies that simply pro-
ilar levels of economic growth. Societal poverty declined by more
in Ecuador because the poor shared to a much larger extent in the mote growth in overall national income.
economic growth.
Why not simply use national
uador. Between 2011 and 2016, both coun-
tries exhibited comparable overall economic poverty lines?
growth. The average annual growth in survey The social and economic assessments made
consumption was 1.95 percent in Costa Rica by governments in setting national poverty
and 1.92 percent in Ecuador. However, the lines underpin essentially all global poverty
level of shared prosperity during this period lines, including the IPL, the higher lines of
was greater in Ecuador than in Costa Rica. US$3.20 and US$5.50 (based on the me-
In Costa Rica, growth among the bottom dian national poverty lines in LMICs and
40 percent of the income distribution (the UMICs), and now the SPL.17 Despite the im-
bottom 40) was essentially the same as the portance of using assessments of basic needs
growth in mean consumption. In contrast, undertaken by countries, this report reflects
the bottom 40 grew a full percentage point a purposeful decision not to allow these as-
more than the mean in Ecuador, resulting in sessments alone to completely determine the
a shared prosperity premium. Although the value of the SPL. An assumption underlying
level of growth was the same, the decline in the SPL is that the cost of social participation
societal poverty was greater in Ecuador over rises with the level of economic development
the period because of the difference in shared (as evidenced by the positive income gradient
prosperity. An examination across all UMICs of national poverty lines), but does not vary
and high-income countries for which data across countries at the same income.18
are available on shared prosperity reveals a This differs greatly from a proposal that
strong correlation (equal to 0.6) between each and every national poverty line should
the shared prosperity premium and the re- be used as a global SPL (Gentilini and
duction in societal poverty. Improvement in Sumner 2012). Such a definition of societal
societal poverty in UMICs and high-income poverty would certainly show respect for the
countries requires economic growth in which judgment of the government of each coun-
the poor disproportionately share. An exam- try, but it would suffer from the problem that
ination of LMICs and low-income countries countries with the same level of median con-
likewise indicates a correlation between the sumption could have different assessments
HIGHER STANDARDS FOR A GROWING WORLD 79
of basic needs. The premise of global soci- In addition, the use of national poverty
etal poverty is that it captures the idea that lines to count societal poverty is also prob-
participation in society becomes costlier as lematic over time. As societies prosper, the
countries become richer and that it is also real value of the threshold used to determine
meant to serve as a tool for global poverty who is considered poor tends to increase. In
monitoring. This latter element, that the SPL poorer countries, this is typically a stepwise
is a global poverty line, means that it should process. A poverty line is held static in real
allow comparisons across countries or over terms for several years or even several decades,
time. The use of national poverty lines as the and then it is revised and held static again for
SPL is problematic on both these counts. a long time. The length of time between the
National poverty lines do not rise strictly revisions depends on the country and the rate
in parallel with economic development, nor of growth experienced. The World Bank’s SPL
are they fixed in value as is the IPL. Figure aims to capture how national poverty lines
3.2 shows that there are many cases in which evolve as countries grow and thus provide a
a country may exhibit higher median con- consistently defined measure of poverty that
sumption than some other country but have mirrors how societies typically measure pov-
a lower national poverty line. There are also erty. The global SPL is derived from a global
many cases in which countries at the same relationship between overall economic devel-
level of economic development rely on vastly opment and observed national poverty lines
different assessments of basic needs. If one across societies, and this averaging over all
were to construct a global SPL based on the countries helps improve comparability. An
sum of national poverty lines, then two peo- example from Vietnam follows.
ple who consume at the same level and living In 1993, the General Statistics Office of
in countries at the same level of economic Vietnam set a national poverty line that
development might be treated differently in would reflect basic needs at the time. The line
the global aggregation of societal poverty. An was equivalent to approximately US$2.05 a
awkward implication of the use of national day at 2011 PPP U.S. dollars, which was kept
poverty lines directly, without any averaging, roughly constant in real purchasing power
is that the global aggregation based on this until 2010.19 Between 1992 and 2008, living
rule would embody a counterintuitive social standards improved twofold, and poverty
judgement that someone who is poor in one measured at the 1993 line fell from 58.0 per-
country may not be identified as poor if his or cent to 14.5 percent. When a new survey was
her well-being were assessed in a richer coun- conducted in 2010, a fresh welfare measure
try with a lower national poverty line. and poverty methodology were developed to
Figure 3.2, panel b, also includes predicted capture living standards and poverty more
lines at the 90th and 10th percentiles from the effectively and reflect current basic needs.
bivariate (quantile) regression of the poverty The new poverty line was set at a value equiv-
line on median consumption. These predicted alent to approximately US$3.50 a day at 2011
lines have similar slopes, and the ratio of these PPP, with a corresponding estimated poverty
lines in levels is approximately 2 over the en- rate of 21 percent.
tire range. This suggests that, at any given level Figure 3.7 shows how the value of the na-
of national well-being, the range in values of tional poverty line, SPLs, and correspond-
national poverty lines is large. The most gen- ing headcount ratio have evolved in Viet-
erous line is consistently about twice as large nam. The SPL in 1993 was US$1.92 a day,
as the least generous line. This result is prob- only slightly below the national threshold of
lematic for the proposal to construct a global US$2.05. When the economy grew rapidly
count of the poor that treats the poverty line in the early 2000s, the value of the SPL rose.
of each country as the relevant threshold. Al- In 2010, when the new poverty line was set,
lowing for such significant differences in the the SPL was US$3.80, a little above the na-
definition of basic needs across countries that tional poverty line; for the latest survey, it was
are essentially at the same level of well-being US$4.90. Whereas the national poverty line
is inconsistent with the idea that needs may is fixed in intervals, and goes up in discrete
rise as economic development expands. steps, the SPL has risen more smoothly, fol-
80 POVERTY AND SHARED PROSPERITY 2018
FIGURE 3.7 Comparing National and Societal Poverty Lines and Rates, Vietnam, 1993–2015
a. Comparison of poverty lines b. Poverty rate
5
70
4 60
Poverty line (2011 US$ PPP)
50
3
Percent
40
2 30
20
1
10
0 0
93
95
97
99
01
03
05
07
09
11
13
15
93
95
97
99
01
03
05
07
09
11
13
15
19
19
19
19
20
20
20
20
20
20
20
20
19
19
19
19
20
20
20
20
20
20
20
20
Vietnam 1993 Vietnam 2010 Societal poverty
Note: PPP = purchasing power parity.
lowing the average trend of the national pov- US$3.20 per person per day, and slightly less
erty line. In 2009, prior to the large increase than half of the world’s population is living on
in the national poverty line, the SPL defini- less than US$5.50. The introduction of these
tion of basic needs was much closer to the yet lines is motivated primarily by noting that
to be determined national poverty line defi- the world has grown richer, and now most
nition of basic needs in 2010 than to the defi- of the extreme poor no longer live in low-
nition in 1993. Because the SPL was smoothly income countries but rather are in middle-
updated as the country prospered, the 2009 income countries. The relevance of an IPL
SPL was likely a better reflection of the social based on national poverty lines from low-
assessment of basic needs at that point than income countries has gradually diminished
the existing definition based on the 1993 na- with time. The motivation for these new
tional poverty line value. higher lines could just as easily be made by
recognizing that it is difficult to precisely
identify thresholds and legitimate to have
Conclusion differing views on what defines basic needs
This chapter discusses two new sets of pov- (Atkinson 1987). The higher lines can help
erty lines that the World Bank will use to re- address this concern.
port on global poverty, and that are intended There are a couple of key takeaways from
to complement the monitoring of poverty as these higher poverty lines. First, the rate of
measured with respect to the IPL. One set has the reduction in extreme poverty in recent
complementary poverty lines that are fixed decades has not been matched by a similarly
at values greater than the IPL. These lines paced reduction in the share of people living
reflect typical assessments of basic needs, on less than US$3.20 or US$5.50. More than
as measured in national poverty lines, for a 80 percent of the population of South Asia
set of LMICs and UMICs and are valued at and Sub-Saharan Africa still live on less than
US$3.20 and US$5.50 (2011 PPP). The basic US$5.50 a day. Second, a large share of the
descriptive statistics of the fixed poverty lines world’s population is living on slightly less
are quite striking. As chapter 1 describes, 10 than US$5.50. A reasonable expectation is
percent of the population is living on less that, if it continues, global economic growth
than US$1.90. This chapter highlights that will produce a rapid reduction in the count
one-fourth of the world is living on less than of people below this threshold.
HIGHER STANDARDS FOR A GROWING WORLD 81
The other new poverty line that the World the highest rate of all regions in 1990 (63.4
Bank is now reporting is the SPL, which is a percent) to one of the lower rates (25.1 per-
mixture of the fixed-in-value IPL and a line cent) in 2015. This impressive performance
that rises in value with median consumption in reducing societal poverty was driven in
in a country. According to this line, individ- large part by the extraordinary success in
uals are considered poor if they are living eradicating extreme poverty.
either on less than the IPL or on a dollar a The focus of monitoring poverty reduc-
day, plus 50 percent of median consumption tion will continue to be on the progress in
in their country of residence. The decision bringing extreme poverty below 3 percent,
to anchor the SPL in a median measure of but it is clear that this measure of poverty
well-being fits the data well (as assessed by is becoming less helpful in the majority of
regressions of national poverty lines on con- countries, which already exhibit rates near
sumption) and corresponds to existing defi- zero. Even though extreme poverty rates may
nitions of relative poverty in many countries. be well below 3 percent in many countries,
The proposed SPL is also relevant to SDG this does not mean that poverty is no lon-
target 10.2 aimed at the social, economic, and ger a problem in these societies. The higher
political inclusion of all. The indicator asso- poverty lines, set in accord with typical na-
ciated with this target is the share of people tional poverty lines from countries classified
living on less than 50 percent of the median as lower-middle and upper-middle income,
income. Although the focus of this SDG is on provide useful guides for monitoring prog-
reducing inequality and improving inclusion, ress on the basis of lines that are fixed in real
it overlaps with the idea of monitoring soci- terms over time. For middle-income coun-
etal poverty. As countries grow, societal pov- tries, these are useful markers for measuring
erty provides information on the extent to progress that aligns with the definition of
which the poor share in the growth. basic needs in middle-income countries from
The rate of decline in societal poverty 2011. For lower-income countries, they could
has been slower than the rate of decline in be viewed as markers for more aspirational
extreme poverty. This is to be expected: the targets in poverty reduction.
value of the SPL rises as the economy grows. Similarly, the measure of societal poverty
Societal poverty has declined by about a third provides a global tool to measure poverty in
across the world, dropping from approxi- accord with how countries assess changing
mately 44.5 percent to 28.4 percent between standards of basic needs; however, in contrast
1990 and 2015. The reduction in extreme to the US$3.20 and US$5.50 lines, the real
poverty was about twice this rate, declining value of these lines changes over time as the
by about 72 percent, dropping from 35.9 country grows richer. Although the SPL can
percent to approximately 10.0 percent. In change in real terms over time, it is constant
the 1990s, when extreme poverty was more in value across countries that are at the same
widespread, the difference between societal level of median consumption or income. Be-
poverty and extreme poverty was relatively cause the SPL is constructed to reflect, on
modest. In 2015, the societal poverty rate was average, national poverty lines at different
almost three times larger than the extreme levels of median consumption or income, it
poverty rate. The continued decline in ex- provides a useful measure of global poverty
treme poverty will likely lead to greater diver- that aligns well with national assessments of
gence in the informational content of these poverty. Keeping the IPL fixed is highly desir-
two measures. able because it allows the progress toward an
Another useful takeaway from the exam- unmoving target to be monitored, but, as the
ination of societal poverty is the differential world advances toward the eradication of ex-
performance across regions. Most regions treme poverty, the US$1.90 poverty line will
experienced a fairly modest reduction in the become increasingly less relevant in many
prevalence of societal poverty. The exceptions countries. In contrast, because the SPL yard-
were the economies of East Asia and Pacific. stick is explicitly a function of the well-being
Societal poverty was cut by more than half of each country, it is, by construction, rele-
there between 1990 and 2015, declining from vant for all countries over time.
82 POVERTY AND SHARED PROSPERITY 2018
Annex 3A
Historical global and
regional poverty estimates
TABLE 3A.1 Historical Trends, Global Poverty Estimates, 1990–2015
a. US$3.20 (2011 PPP) Poverty
Squared Population
Year Poverty rate (%) Poverty gap (%) poverty gap Poor (millions) (millions)
1990 55.1 26.6 15.5 2,914.0 5,284.9
1993 54.4 25.6 14.7 3,013.4 5,542.9
1996 51.7 22.8 12.7 2,993.8 5,792.6
1999 50.6 22.3 12.4 3,056.1 6,038.1
2002 47.2 20.2 11.0 2,962.7 6,276.8
2005 42.2 16.9 8.8 2,753.3 6,517.0
2008 38.2 14.9 7.7 2,586.9 6,763.7
2011 32.8 12.1 6.0 2,298.8 7,012.8
2013 28.8 10.2 5.0 2,071.7 7,182.9
2015 26.3 9.2 4.6 1,932.7 7,355.2
b. US$5.50 (2011 PPP) Poverty
Squared Population
Year Poverty rate (%) Poverty gap (%) poverty gap Poor (millions) (millions)
1990 67.0 41.5 28.8 3,540.5 5,284.9
1993 67.9 40.9 28.0 3,761.2 5,542.9
1996 67.3 38.7 25.6 3,900.0 5,792.6
1999 66.8 38.1 25.1 4,035.2 6,038.1
2002 64.0 35.6 23.0 4,018.2 6,276.8
2005 60.4 31.9 19.9 3,939.4 6,517.0
2008 56.5 29.0 17.8 3,823.7 6,763.7
2011 52.2 25.3 15.0 3,662.3 7,012.8
2013 48.7 22.6 13.1 3,498.3 7,182.9
2015 46.0 20.9 12.0 3,386.5 7,355.2
Source: PovcalNet (http://iresearch.worldbank.org/PovcalNet/), World Bank.
Note: PPP = purchasing power parity.
HIGHER STANDARDS FOR A GROWING WORLD 83
TABLE 3A.2 Historical Trends, Regional Poverty Rates, 1990–2015
Percent
a. US$3.20 (2011 PPP) Poverty rates
Region 1990 1993 1996 1999 2002 2005 2008 2011 2013 2015
East Asia and Pacific 85.3 79.7 70.6 67.1 57.2 45.4 37.4 26.5 17.5 12.5
Europe and Central Asia 9.9a 15.1 19.2 21.1 14.9 11.8 7.5 6.6 5.7 5.4
Latin America and the Caribbean 28.3 27.1 27.7 27.0 24.9 21.4 15.7 13.1 11.4 10.8
Middle East and North Africa 26.8 28.9 28.0 21.7 19.6 18.8 16.7 14.9 14.4 16.3
South Asia 81.7 80.4 77.3 76.0a 75.5 71.5 67.9 58.9 53.9 48.6a
Sub-Saharan Africa 74.9 78.2 78.0 78.3 78.2 74.8 72.2 70.1 67.8 66.3
Sum of regions 66.4 65.1 61.6 60.1 55.9 49.9 45.0 38.5 33.7 30.7
Rest of the world 0.8 0.8 0.7 0.8 0.7 0.7 0.7 0.8 0.8 0.9
World 55.1 54.4 51.7 50.6 47.2 42.2 38.2 32.8 28.8 26.3
b. US$5.50 (2011 PPP) Poverty rates
Region 1990 1993 1996 1999 2002 2005 2008 2011 2013 2015
East Asia and Pacific 95.2 93.2 89.3 87.0 79.9 71.7 63.6 52.3 42.4 34.9
Europe and Central Asia 25.3a 35.9 41.2 44.5 34.5 26.5 17.1 15.4 14.1 14.0
Latin America and the Caribbean 48.6 48.0 48.2 47.0 45.1 40.9 33.3 29.6 27.2 26.4
Middle East and North Africa 58.8 59.4 59.6 54.5 51.4 49.5 46.6 43.0 42.3 42.5
South Asia 95.3 95.0 93.9 93.1a 92.8 91.0 89.8 86.4 84.2 81.4a
Sub-Saharan Africa 88.5 90.4 90.2 90.5 90.9 89.9 88.1 86.9 85.4 84.5
Sum of regions 80.5 81.2 80.2 79.3 75.7 71.3 66.5 61.2 57.0 53.7
Rest of the world 1.7 1.6 1.4 1.3 1.3 1.2 1.2 1.4 1.5 1.5
World 67.0 67.9 67.3 66.8 64.0 60.4 56.5 52.2 48.7 46.0
Source: PovcalNet (http://iresearch.worldbank.org/PovcalNet/), World Bank.
Note: The criteria for estimating survey population coverage is whether at least one survey used in the reference year estimate was conducted within two years of the
reference year. “Sum of regions” was previously referred to as “developing world.” PPP = purchasing power parity.
a. This estimate is based on less than 40 percent of regional population coverage.
TABLE 3A.3 Historical Trends, Regional Number of Extreme Poor, 1990–2015
Millions
a. Number of poor at US$3.20 (2011 PPP)
Region 1990 1993 1996 1999 2002 2005 2008 2011 2013 2015
East Asia and Pacific 1,366.5 1,332.1 1,224.7 1,205.4 1,057.1 859.5 723.8 524.0 352.2 254.2
Europe and Central Asia 46.1a 70.8 90.4 99.4 70.2 55.4 35.6 31.6 27.7 26.2
Latin America and the Caribbean 124.5 125.9 135.7 138.4 133.0 118.8 90.8 78.3 70.0 67.5
Middle East and North Africa 61.5 71.2 73.4 60.4 57.4 58.2 54.6 51.2 51.5 60.6
South Asia 925.3 971.5 992.5 1,034.4a 1,085.5 1,081.5 1,075.8 973.5 916.0 847.2a
Sub-Saharan Africa 383.2 434.7 470.0 510.5 552.3 572.5 599.1 631.8 645.4 667.0
Sum of regions 2,907.1 3,006.2 2,986.7 3,048.6 2,955.5 2,745.9 2,579.6 2,290.3 2,062.8 1,922.9
Rest of the world 6.8 7.2 7.1 7.5 7.2 7.4 7.3 8.5 8.9 9.8
World 2,914.0 3,013.4 2,993.8 3,056.1 2,962.7 2,753.3 2,586.9 2,298.8 2,071.7 1,932.7
b. Number of poor at US$5.50 (2011 PPP)
Region 1990 1993 1996 1999 2002 2005 2008 2011 2013 2015
East Asia and Pacific 1,525.3 1,557.7 1,550.2 1,562.2 1,476.0 1,357.5 1,231.0 1,035.2 851.7 710.4
Europe and Central Asia 117.3a 168.5 194.0 209.7 161.8 124.4 81.0 73.7 67.8 68.2
Latin America and the Caribbean 214.4 223.1 235.8 240.8 241.1 227.6 192.5 177.2 166.9 165.4
Middle East and North Africa 135.1 146.4 156.3 151.6 150.9 152.9 151.9 148.3 151.7 157.9
South Asia 1,080.1 1,148.5 1,206.7 1,267.6a 1,334.1 1,377.0 1,423.1 1,429.6 1,431.0 1,419.0a
Sub-Saharan Africa 452.8 502.6 543.5 590.3 641.5 687.4 731.7 783.4 813.1 849.5
Sum of regions 3,525.0 3,746.8 3,886.5 4,022.2 4,005.4 3,926.9 3,811.2 3,647.4 3,482.2 3,370.3
Rest of the world 15.5 14.4 13.5 13.0 12.9 12.6 12.5 15.0 16.1 16.1
World 3,540.5 3,761.2 3,900.0 4,035.2 4,018.2 3,939.4 3,823.7 3,662.3 3,498.3 3,386.5
Source: PovcalNet (http://iresearch.worldbank.org/PovcalNet/), World Bank.
Note: The criteria for estimating survey population coverage is whether at least one survey used in the reference year estimate was conducted within two years of the
reference year. “Sum of regions” was previously referred to as “developing world.” PPP = purchasing power parity.
a. This estimate is based on less than 40 percent of regional population coverage.
84 POVERTY AND SHARED PROSPERITY 2018
Notes
1. Target 1.A of the MDGs is to halve, between “a ‘societal’ head count measure of global con-
1990 and 2015, the proportion of people sumption poverty.”
whose income is less than one dollar a day. The 7. In the relatively small number of countries
indicator for monitoring progress in achiev- in which extreme poverty is assessed using
ing the target was fixed at the proportion of income rather than consumption, the SPL is
people living on less than the World Bank IPL similarly defined in terms of income instead
of US$1.25 a day (in 2005 PPP values). Sim- of consumption.
ilarly, target 1.1 of the SDGs, to be achieved 8. If median consumption is US$1.60, then
by 2030, is to eradicate extreme poverty for all US$1.00 + half of US$1.60 is US$1.80. This
people everywhere, measured as people living value is less than the IPL of US$1.90; so, in
on less than $1.90 a day, the IPL. See Millen- this case, the SPL is set at the lower bound,
nium Development Goals Indicators (data- US$1.90.
base), Development Indicators Unit, Statistics 9. For a detailed discussion of the fit of the SPL
Division, United Nations, New York, http:// with national poverty lines and how this fit
mdgs.un.org/unsd/mdg/Host.aspx?Content= compares with other candidate specifications,
Indicators%2fOfficialList.htm; “Sustainable see Jolliffe and Prydz (2017).
Development Goals: 17 Goals to Transform 10. See Fuchs (1967); “Poverty Rate” (indicator),
Our World,” United Nations, New York, http:// Organisation for Economic Co-operation and
www.un.org/sustainabledevelopment/. Development, Paris (accessed January 26,
2. The World Bank classification of countries 2017), https://doi.org/10.1787/0fe1315d-en.
according to regions and income groups 11. For details on each of the 17 SDGs, including
is followed here. For details on income metadata and indicators, see “Compilation of
classification, see Fantom and Serajuddin Metadata for the Proposed Global Indicators
(2016). For the World Bank regions, see for the Review of the 2030 Agenda for Sus-
“Select a Region,” in “Where We Work,” tainable Development,” Inter-agency Expert
http://www.worldbank.org/en/country. Group on SDG Indicators, Statistics Division,
3. There may be different interpretations of what Department of Economic and Social Affairs,
“fixed in real terms” means. Here it means that United Nations, New York. http://unstats.un
the lines are converted to domestic currency .org/sdgs/iaeg-sdgs/metadata-compilation/.
in 2011 prices, using the 2011 PPP conversion The decision that the cost of social partici-
factors, and are thereafter adjusted over time pation is increasing in median consumption
by the main domestic consumer price index rather than, say, average consumption is dis-
used in each country. cussed in detail in Jolliffe and Prydz (2017)
4. The bin sizes of the consumption distributions and is consistent with arguments made by
have been selected to correspond to key thresh- Aaberge and Atkinson (2013), Birdsall and
olds at US$1.90, US$3.20, and US$5.50. The Meyer (2015), and Stiglitz, Sen, and Fitoussi
statement then about most people consuming (2010) that the median is a better represen-
just less than US$1.90 is affected by the selected tation of the material well-being of a country
bin sizes. But an estimated density function of relative to the mean and is also a simple way of
the log of consumption closely corresponds to capturing distributional aspects of well-being.
the shape of the histogram displayed. 12. See Ferreira et al. (2016) for a discussion on
5. For more examples of countries that have inflating 2005 PPP values into 2011 PPP
changed the value of their national pov- values. They assert that, on average, US$1.90
erty lines, see the online appendix of Jolliffe in 2011 PPP U.S. dollars maintains the same
and Prydz (2016), at https://static-content purchasing power as US$1.25 in 2005 PPP for
.springer.com/esm/art%3A10.1007%2Fs10888 the set of 15 poor countries that determine the
-016-9327-5/MediaObjects/10888_2016_9327 IPL. They also demonstrate that this inflation
_MOESM1_ESM.pdf. rate of about 52 percent maintains an average
6. The motivation for referring to the line as the purchasing power for essentially all countries
SPL is drawn from the World Bank (2017, in the PovcalNet database for which they esti-
xxi), which recommends the introduction of mate poverty (and have measures of PPP in
HIGHER STANDARDS FOR A GROWING WORLD 85
both years). Inflating US$0.67 by 52 percent data used in the rest of this report. See Pov-
results in US$1.01. Furthermore, direct rees- calNet (online analysis tool), World Bank,
timation of Ravallion’s (2016) consumption http://iresearch.worldbank.org/PovcalNet/.
floor using 2011 PPP gives a value of US$1.00 16. Household survey data do not exist for every
at 2011 PPP. country in every year, but all global poverty es-
13. Similarly, Allen (2017, table 11) estimates the timates are for a specific year. To overcome the
lowest cost of a diet consisting of 2,100 calo- data gaps, survey data are projected forward
ries per day with 50 grams of protein and 34 and, sometimes, backcast to produce country
grams of protein across several countries. The poverty rates for each year. For an overview of
lowest value he estimates is US$0.98 in 2011 the methods, see Ferreira et al. (2016); Jolliffe
PPP terms for Zimbabwe. et al. (2015).
14. See PovcalNet (online analysis tool), World 17. The idea that national poverty lines represent
Bank, Washington, DC. http://iresearch.world social assessments of minimum needs has
bank.org/PovcalNet/. The estimates cited here been a motivating argument behind the use
were produced from the version of PovcalNet of the IPL for many years. Ravallion, Datt,
updated on October 1, 2016. China, India, and and van de Walle (1991) and the World Bank
Indonesia have separate rural and urban dis- (1990) interpret national poverty lines in
tributions in PovcalNet, and no national me- some of the poorest countries as representa-
dian is readily available. For these countries, tive of absolute minimum needs and use them
the national median is derived by combining in calculating the dollar-a-day IPL.
the rural and urban population-weighted 18. The claim is not being made that this report
distributions available in PovcalNet and esti- empirically disentangles whether the rising
mating the median of the joint national dis- value of national poverty lines reflects the
tribution. The resulting national median is growing cost of social participation (as is as-
used in defining the SPL for these countries. sumed here) or simply reflects a definition of
For high-income countries, the alignment basic needs that is more generous, resulting in
of the surveys closest to the reference years greater utility. For a discussion of this iden-
is replicated using National Accounts data, tification challenge, see Ravallion and Chen
the method in the PovcalNet reference-year (2017).
aggregation. 19. The 1993 value was estimated from the na-
15. The profile of societal poverty presented tional headcount ratio and an internation-
here is based on estimates from PovcalNet as ally harmonized welfare vector, following the
of September 2018, the same version of the method of Jolliffe and Prydz (2016).
86 POVERTY AND SHARED PROSPERITY 2018
Beyond Monetary Poverty 4
This chapter reports on the results of the World Bank’s first exercise of multidimensional global
poverty measurement. Information on consumption or income is the traditional basis for the
World Bank’s poverty estimates, including the estimates reported in chapters 1–3. However,
in many settings, important aspects of well-being, such as access to quality health care or a
secure community, are not captured by standard monetary measures. To address this concern,
an established tradition of multidimensional poverty measurement measures these nonmone-
tary dimensions directly and aggregates them into an index. The United Nations Development
Programme’s Multidimensional Poverty Index (Global MPI), produced in conjunction with the
Oxford Poverty and Human Development Initiative, is a foremost example of such a multi-
dimensional poverty measure. The analysis in this chapter complements the Global MPI by
placing the monetary measure of well-being alongside nonmonetary dimensions. By doing so,
this chapter explores the share of the deprived population that is missed by a sole reliance on
monetary poverty as well as the extent to which monetary and nonmonetary deprivations are
jointly presented across different contexts.
The first exercise provides a global picture using comparable data across 119 economies
for circa 2013 (representing 45 percent of the world’s population) combining consumption or
income with measures of education and access to basic infrastructure services. Accounting
for these aspects of well-being alters the perception of global poverty. The share of poor
increases by 50 percent—from 12 percent living below the international poverty line to 18
percent deprived in at least one of the three dimensions of well-being. Across this sample,
only a small minority of the poor is deprived in only one dimension: more than a third of the
poor suffer simultaneous deprivations in all three dimensions. More than in any other region
of the world, in Sub-Saharan Africa shortfalls in one dimension occur alongside deprivations in
other dimensions. In South Asia, the relatively high incidence of deprivations in education and
sanitation imply that poverty rates could be more than twice as high when these nonmonetary
dimensions are added.
A second complementary exercise for a smaller set of countries (six) explores the inclusion
of two additional nonmonetary dimensions. When measures of health and household security
(the risk of experiencing crime or a natural disaster) are included alongside the previous three
dimensions, the profile of the poor changes. In most countries, the share of the poor living in
female-headed households is greater than when the nonmonetary dimensions are excluded
and, in some countries, the poor also have a significantly higher presence in urban areas.
87
Why look beyond monetary various goods taking their relative prices into
account, these relative prices serve as natural
poverty?
weights with which to aggregate those quan-
Consider the following hypothetical exam- tities consumed.1 That is why they form the
ple. Two families have the same income, say basis for the first three chapters in this report.
US$3.00 per person per day. However, only It is why poverty has typically been defined
one family has access to adequate water, sani- in terms of whether a household’s income
tation, and electricity, whereas the other lives reaches or surpasses a monetary threshold, the
in an area lacking the necessary infrastruc- poverty line, which represents the minimum
ture for basic services, such as a power grid or amount needed to purchase a sufficient quan-
water mains. Members of this second family tity of essential goods and services.
will still consume water and use energy for Yet the point of the example is that
lighting and cooking, but they may have to monetary-based measures do not encompass
spend hours per week fetching water from all aspects of human well-being. One reason
a well, or pay higher prices to obtain lower- for this is that not all goods and services that
quality water from a truck. For sanitation, matter to people are obtained exclusively
they may use a private or communal latrine, through markets. Consequently, the prices
without the convenience or hygiene benefits necessary to cost these goods and services ei-
of a sewerage connection. And with no ac- ther do not exist or do not accurately reflect
cess to an electricity grid, the second family’s their true consumption value (World Bank
choice set for lighting and power options is 2017b). Common examples of nonmarket
severely reduced. Both households will spend goods without prices are public goods such
some of their US$3.00 per person per day to as a clean environment and a secure commu-
meet their energy and water needs. Yet, be- nity. Examples of goods with prices that often
cause their choice sets (including the prices do not reflect true consumption value include
they face) are so different, the differences in those that require large public investments
their living standards arising from the access to make them available—the provision of a
that the first family enjoys are not captured power grid is often necessary before a house-
by a monetary measure of poverty alone. hold can access electricity. Other core services
The first family clearly enjoys a higher stan- at least partially provided through systems
dard of living than the second, but a welfare supported by direct government spending
judgment that considers only their incomes include health care and education. General
will pronounce them equally well-off. This is government health expenditure accounts for
an example of when public action—or lack more than half of total global health expen-
thereof—can directly affect the well-being diture. Likewise, governments on average
of households by expanding—or not—their spend the equivalent of nearly 5 percent of
choice sets in ways that incomes and prices fail the gross domestic product (GDP) of their
to fully internalize. It is possible that, under economies on education. The presence of
a broader assessment of poverty, the second such goods renders the traditional monetary
family might be considered poor or deprived, welfare measure incomplete with respect to a
even though its daily income is above the in- variety of core aspects of well-being.
ternational poverty line of US$1.90 per day. This chapter presents a broader picture of
To be clear: Income (or consumption ex- well-being than that found in chapters 1–3,
penditures valued at prevailing market prices) by considering a notion of poverty that rec-
is hugely important for human well-being. ognizes the centrality of the monetary mea-
Indeed, income and consumption are the sure, but looks to complement it by explicitly
workhorse metrics of individual welfare in treating access to key nonmarket goods as
economic analysis. They summarize a house- separate dimensions of well-being. Specif-
hold’s capacity to purchase multiple goods and ically, the chapter previews a multidimen-
services that are crucial for well-being, such sional poverty measure derived from stan-
as food, clothing, and shelter. And they do so dardized data for 119 economies that provide
with one remarkable property: because con- a global picture for circa 2013. The multidi-
sumers choose the quantities they consume of mensional measure is anchored on consump-
88 POVERTY AND SHARED PROSPERITY 2018
tion or income as one dimension of welfare, and national level (box 4.1). The capability
and includes several direct measures of access framework inspired the development of the
to education and utilities (such as electricity, first global efforts to measure poverty multi-
water, and sanitation) as additional dimen- dimensionally. These were carried out by the
sions. Although this multidimensional mea- United Nations Development Programme
sure has wide country coverage, it still lacks (UNDP), through the Human Poverty Index
information on other important dimensions in the late 1990s (UNDP 1997) and, more re-
of well-being including health care and nu- cently, through the Global Multidimensional
trition, as well as security from crime and Poverty Index (Global MPI), introduced
natural disasters. Consequently, in a more in the 2010 Human Development Report
exploratory manner, the chapter extends the (UNDP 2010), developed with the Oxford
analysis by adding these dimensions for a Poverty and Human Development Initiative
smaller subset of countries for which infor- (OPHI), and reported annually for over 100
mation for all these dimensions can be cap- countries. At the country level, an increasing
tured within the same household survey. number of governments are choosing to ex-
The two exercises—one with broad coun- pand or complement their poverty measures
try coverage, but fewer dimensions than one with multidimensional indicators (see spot-
would ideally like, and the other with a rela- light 4.1 at the end of this chapter). The ef-
tively extensive set of dimensions, but available forts of the UNDP, OPHI, and most govern-
only as a pilot for a few countries—represent ments build on influential research by Sabina
the World Bank’s first steps toward including Alkire and James Foster (see, for example,
multidimensional poverty indicators in the Alkire and Foster 2011).
set of complementary indicators of global The efforts here are also indebted to these
poverty, as suggested by the Commission on previous efforts by other researchers, gov-
Global Poverty (World Bank 2017b). Going ernments, and international institutions. In
forward, the World Bank will monitor prog- addition, they follow on the World Develop-
ress on multidimensional poverty at the global ment Report (WDR) 2000/01 Attacking Pov-
level using the three-dimensional measures erty (World Bank 2001), which recognized the
presented in this chapter, while continuing its many dimensions of poverty and considered
efforts to incorporate the dimensions missing deprivations in education and health alongside
from the global analysis for future rounds. income in its analysis of the evolution of pov-
This approach adopts a living standards erty. The present report goes beyond the WDR
perspective, in that each dimension is valued 2000/01 by taking advantage of richer house-
instrumentally, that is, each dimension rep- hold-level data that combine monetary and
resents the ability to command goods and nonmonetary indicators to present deprivation
services that households value for other ends in each domain as well as measures that aggre-
(in other words, consuming or owning these gate these different deprivations. This proposal
commodities allows for the satisfaction of follows from the recommendations of the
different needs and wants). But it is also con- Commission on Global Poverty, led by Profes-
sistent with the capability framework, which sor Sir A. B. Atkinson, to consider complemen-
calls for expanding the evaluative space for tary indicators to monetary poverty “where a
assessing welfare (Sen 1987). The capability dashboard approach is proposed as part of the
approach advocates for a broader perspective Complementary Indicators, . . . together with a
to capture the “plurality of different features measure of the extent of overlapping depriva-
of our lives and concerns” (Sen 2009, 233). In tions” (World Bank 2017b, 100).
this approach people have varying abilities to The present exercise is also related to the
convert resources into the opportunity to be Sustainable Development Goals (SDGs) es-
and do what they most value—that is, into tablished by the United Nations in 2015,
what Sen terms “capabilities.” which include a call for governments to re-
Of course, measuring poverty multi- port on their progress in improving the na-
dimensionally is not a new endeavor. In- tional multidimensional poverty indicator
deed, multidimensional poverty measures (Indicator 1.2.2 of SDG 1, end poverty in
have become widespread both at the global all its forms everywhere).2 The focus of this
BEYOND MONETARY POVERTY 89
BOX 4.1 Early Applications of Multidimensional Poverty Measurement
The approach followed in this children, housing conditions, access the Human Poverty Index,
chapter builds on previous to basic services, and the economic which appeared in the Human
applications of the multidimensional capacity of household members. Development Reports from 1997
poverty concept. There is a long The basic needs indicators are to 2009 measuring country-
history of assessing the deprivation generally calculated using census level aggregate deprivations in
of individuals by combining multiple data. health, education, and standard of
components of well-being. Inspired The Mexican government living. The Global MPI combines
by empirical studies in the 1970s has taken a lead in adopting a 10 indicators grouped in three
and early 1980s, various European multidimensional approach in the dimensions, namely, education,
countries have been measuring official poverty measure. Following health, and standard of living, and
the share of the population that a comprehensive consultative identifies each person as poor or
is deprived in a select number of process initiated in 2006, and nonpoor according to how many
socially perceived necessities as a grounded on a human rights deprivations they face (Alkire and
core indicator of social exclusion.a perspective, the government, Santos 2010; Alkire et al. 2015).
In many of these cases, such as since 2010, has measured poverty This work has been adapted and
in Ireland, the United Kingdom, as the share of the population adopted by many developing
and, later, the European Union, the that is deprived simultaneously in countries (see spotlight 4.1).
assessment of multiple deprivations monetary terms and in at least one The 2018 edition of the Global
combines income poverty with of six social indicators reflecting MPI includes 105 countries,
the counting of these material core social rights. These indicators with a population coverage of 75
deprivations.b Since the 1980s, cover gaps in education, access to percent of the global population
many countries in Latin America health services, access to social (OPHI 2018). A comparison of the
have complemented monetary security, access to basic residential indicators included in the Global
poverty measures developed services, housing conditions, and MPI, as well as the Mexican
through household surveys with access to food (CONEVAL 2010). poverty measure and (selected
an indicator of unsatisfied basic Since 2010, OPHI and the indicators) for Europe 2020 and the
needs that counts the number of UNDP have been computing multidimensional poverty measures
deprivations in several indicators, the Global MPI for over 100 presented in the chapter, is found in
including school enrollments among countries. The Global MPI replaced annex 4A.
a. The Level-of-Living Survey in Sweden and Townsend (1979) and Mack and Lansley (1985) in the United Kingdom are considered
pioneers in Europe in this approach. Excellent reviews on early applications include Aaberge and Brandolini (2015) and Alkire et al.
(2015). For the Swedish survey, see LNU (The Swedish Level-of-Living Survey) (database), Swedish Institute for Social Research,
Stockholm University, Stockholm, https://www.sofi.su.se/english/2.17851/research/three-research-units/lnu-level-of-living.
b. In Ireland, “consistent poverty” is measured as the population share that is both income poor and deprived in two or more
essential items. In the United Kingdom, a similar approach has been used since 2010 to measure child poverty. In the European
Union, the Europe 2020 poverty and social exclusion headline indicator combines income poverty (the at-risk-of-poverty rate),
household quasi-joblessness, and severe material deprivation (lacking at least four of nine items that are considered fundamental
to enjoying an adequate standard of living). See Atkinson et al. (2002); Marlier et al. (2007).
chapter, on steps to develop a useful global harmonization, several key insights emerge
multidimensional poverty measure, should from the analysis.
not be taken as a preference for such a global
measure over possibly richer country-level Considerations
measures when assessing national progress. for constructing
The requirement of a global multidimen-
multidimensional poverty
sional poverty measure for standardized
household indicators across many countries measures
necessarily limits indicator choice to the rel- This is the initial step by the World Bank to
atively few that are consistently measured. expand the space of assessment beyond the
Nonetheless, despite this constraint of data monetary to explicitly include access to non-
90 POVERTY AND SHARED PROSPERITY 2018
market goods and services that are essen- former is not available) captures people’s
tial for well-being. In addition to a measure access to certain crucial goods and ser-
based on economic resources, it incorporates vices, including food, clothing, and shelter.
a core set of indicators for nonmonetary di- The consumption measure uses market
mensions and presents results on the extent prices to aggregate across the various con-
to which these deprivations arise and overlap. sumption goods.3 Market prices reflect the
Furthermore, it presents summary measures ability of people to purchase goods and
that combine the information into a single services, while allowing for variation in
index, the multidimensional poverty head- individual preferences. Other aspects of
count ratio. well-being on which prices are not avail-
Broadening the poverty measure to in- able or are arguably not a good representa-
corporate additional directly measured com- tion of value should therefore complement
ponents involves two steps. First, one must monetary poverty. Public goods as well as
select the dimensions, the indicators, and private goods that are heavily subsidized
the respective sufficiency thresholds for each are cases in which prices either do not exist
indicator. For example, in the case of the ed- or, if they do exist, do not closely represent
ucational dimension, one possible indicator the household’s valuation of the good.
could be school enrollment for the school-
age children in the household, and the suf- • Relevance. The indicators included should
ficiency threshold is that all children are in be relevant in that they are widely ac-
school (and therefore every household mem- knowledged to represent essential aspects
ber is considered deprived if at least one child of well-being. Indicator thresholds should
is not enrolled). To consider the existence of reflect minimum basic needs, comparable
multiple deprivations occurring in the case with the US$1.90 per person per day pov-
of a same individual, all indicators need to be erty lines. The SDGs and other similar ini-
observed or inferred for the same individual, tiatives provide useful guidance.
typically from the same data source. Second, • Data availability. Indicators should ideally
the information on each dimension is then be derived from the same data source (typ-
aggregated into one index. Summary indexes ically a household survey). One of the key
can be applied to generate rankings across features of the multidimensional approach
population groups and countries, while ac- is that it can be used to assess the extent
knowledging the multiplicity of deprivations. to which deprivation in one dimension is
This section briefly discusses the proposed related to deprivation in other dimensions
choices in each of these two stages. for the same individual. However, because
of the requirement about data sources, the
Selected dimensions and choice of the dimensions and indicators to
indicators be included will ultimately be shaped by
the availability of meaningful data.
The selection of the dimensions and indica-
tors relevant to the measurement of standards • Parsimony. The multidimensional mea-
of living is never simple. Possessing a clear sure should be parsimonious. It should
conceptual framework to advise this process involve only a small number of judiciously
is therefore fundamental. The approach to selected dimensions to lend prominence to
the selection of the nonmonetary indicators multidimensionality, while ensuring suffi-
is guided by the idea that poverty, at least in cient population coverage.
part, represents an inability to reach a min-
imum standard of material well-being com- Because of data limitations, there exists a
prising both market and nonmarket goods. trade-off between the number of dimensions
The choice of dimensions is informed by (measured by harmonized indicators) that
the following core principles: can be included in the multidimensional pov-
erty measure and the number of countries
• Centrality of private consumption. Pri- that can be included in the analysis. For ex-
vate consumption (or income, when the ample, comprehensive assessments of health
BEYOND MONETARY POVERTY 91
services and health outcomes are rarely avail- Most often, this is a measure of school
able in the same household survey that also enrollment (among children and youth
contains the lengthy questionnaires typically of school age) or educational attainment
necessary to measure consumption poverty. (among adults). The education dimension
For this reason, the chapter conducts two here similarly has these two components.
complementary exercises. To get a global These indicators are available for many
picture, the next section presents an analy- countries and are standardized in recent
sis including a large number of economies surveys across 119 economies.
(119, covering 45 percent of the world’s
3. Access to basic infrastructure. The third
population) and includes three dimensions,
dimension encompasses access to key ser-
including consumption, represented by six
vices that often require large-scale public
indicators. The second exercise uses data for a
investments to make them widely avail-
much smaller set of countries (six) to explore
able. Access to electricity and a certain
the impact of adding two additional dimen-
standard of drinking water and sanitation
sions. The analysis that follows should be un-
are critical for economic activity and sur-
derstood as an initial exploration to generate
vival (related to SDGs 6 and 7). Although
a consistent, conceptually robust, and prac-
many individuals pay for the provision of
tical proposal for expanding current poverty
these services (through utility bills or oth-
measurement methods to include other non-
erwise), the choice set available to users
monetary dimensions of well-being.
(and their prices) depends to a large ex-
The five well-being dimensions consid- tent on the initial investments that gov-
ered in this chapter are the following: ernments have made on electricity grids
and water and sewer networks. This pub-
1. Monetary well-being. The first dimen- lic action often determines the price and
sion is the monetary measure of well-be- quality of the service provided.4 For the
ing that the World Bank uses as its prin- 119-economy sample, indicators can be
cipal poverty measure: the consumption standardized across multipurpose house-
or income per person per day, valued at hold surveys to reflect wider definitions
2011 purchasing power parity (PPP) U.S. of “at least limited” drinking water and
dollars, that is available to the individuals “at least limited” sanitation used in the
in the household (SDG target 1.1). This is SDG monitoring, whereas, for the smaller
the well-being measure and threshold fea- six-country sample, the chosen indicator
tured in chapter 1 of this report. The di- applies a more stringent definition also
mension encompasses the range of goods used under the SDG framework of access
and services that can be purchased at mar- to “at least basic” water and sanitation.5
ket prices. The sufficiency threshold is the
international poverty line, currently set at 4. Health and nutrition. Health is widely
US$1.90 per person per day. Individuals considered a core dimension of well-
living in households in which per capita being. It is the focus of SDG 3: ensure
income falls below this cutoff are consid- healthy lives and promote well-being for
ered deprived in the monetary dimension all at all ages. As in other cases, health care
of well-being. is typically not supplied entirely through
the market or valued entirely at market
2. Education. Although education may be prices. The empirical challenge of in-
available through private or public institu- cluding this dimension for a large set of
tions, provision among a large share of the countries limits the feasibility of investi-
population is fully or partially subsidized gating health and nutrition meaningfully
in most countries. The price that families in the 119-economy sample. However,
must pay therefore does not adequately for a smaller selection of countries, one
represent the value of the service. Indexes may analyze indicators of access to for-
of multidimensional poverty typically mal health care services as well as direct
include at least one indicator of access individual assessments of nutrition. Four
to formal education (related to SDG 4). indicators are included in the health and
92 POVERTY AND SHARED PROSPERITY 2018
nutrition dimension: facility-based birth incidence of crime at the household level
delivery, vaccination among children, as well as the threat of crime, often defined
the incidence of child stunting, and un- by the incidence of crime in the commu-
dernourishment among adult women. nity. The six-country study includes this
Whereas nutrition is intimately linked to indicator. In addition, this dimension in-
food consumption—and thus can be ar- corporates a measure of the prevalence
gued to be already indirectly included in of natural disasters that severely affected
monetary poverty—stunting and malnu- households’ well-being beyond short-
trition are also reflective of exposure to term losses in consumption. Although
illness and lack of nutritional knowledge information on the incidence of natural
as well as possible unequal access of re- disasters is sometimes captured in shock
sources within households. modules in household surveys—such as
in the six-country study analyzed in this
5. Household security. A final dimension chapter—other environmental qualities
considers the risks to which households essential for a good life, such as air free of
are exposed and for which insurance or pollution, are most often not included and
mitigation programs, where they exist, thus cannot be incorporated at this stage.6
are often partially or fully supplied by the
government. One of the basic functions of Table 4.1 illustrates the individual indica-
government is to ensure that the daily lives tors. Appendix A contains technical details
of the population are free of the fear of on indicator definitions.
exposure to violence and crime. Although One limitation of the approach followed
few living standard–type surveys ade- in this chapter is that it relies on indicators
quately cover the relevant issues, some do that are readily available in standard house-
contain questions designed to measure the hold surveys. For many of the dimensions
TABLE 4.1 Dimensions of Well-Being and Indicators of Deprivation
Dimensions Three dimensions (119 economies) Five dimensions (6 countries)
Monetary Daily consumption or income is less than US$1.90 Daily consumption or income is less than US$1.90 per person
poverty per person
Education At least one school-age child up to the age of grade 8 At least one school-age child up to the age of grade 8 is not enrolled in school
is not enrolled in school
No adult in the household (age of grade 9 or above) No adult in the household (age of grade 9 or above) has completed primary
has completed primary education education
Access to basic The household lacks access to limited-standard The household lacks access to a basic-standard drinking water (“limited-standard”
infrastructure drinking water with an added criterion of the source being within a round trip time of 30 minutes)
The household lacks access to limited-standard The household lacks access to basic-standard sanitation (“limited-standard” with
sanitation an added criterion of the facility for the exclusive use of the household)
The household has no access to electricity The household has no access to electricity
Health and Any woman age 15–49 with a live birth in the last 36 months did not deliver at a
nutrition health facilitya
Any child age 12–59 months did not receive DPT3 vaccinationa
Any child age 0–59 months is stunted (HAZ < −2)
Any woman age 15–49 is undernourished (BMI < 18.5)
Security The household has been subject to crime in the previous 12 months or lives in a
community in which crime is prevalent
The household has been affected by a natural disaster (including flooding, drought,
earthquake) in the previous 12 months
Note: BMI < 18.5 = body mass index below 18.5 (underweight); DPT3 = diphtheria-pertussis-tetanus vaccine; HAZ < −2 = the height-for-age Z-score is below −2, that is, more
than two standard deviations below the reference population mean. “Limited-standard” drinking water is drinking water that comes from an improved source (for example,
piped, borehole, protected dug well, rainwater, or delivered water). “Limited-standard” sanitation means using improved sanitation facilities (for example, flush/pour flush to
piped sewer system, septic tank, or a composting latrine).
a. If the indicator is not applicable, for example if the household includes no women who gave birth in the previous 36 months, the household is classified as deprived if the
relevant deprivation rates in the subregion of residence are sufficiently high. Specifically, the deprivation threshold is set such that the share of individuals in nonapplicable house-
holds that are classified as deprived equals the national share of deprived individuals in applicable households who actually experienced a recent birth or have a child under age 6.
BEYOND MONETARY POVERTY 93
considered, relevant information on the im- combine household information on well-
portant aspect of service quality is sometimes being across dimensions into a single num-
available in specialized surveys, but not in ber. Such indicators facilitate comparisons
standard household surveys that also record across countries and time, especially if the
other data on well-being. Essential infor- extent of deprivation within countries varies
mation on quality thus cannot be used for across dimensions under consideration.
various indicators here (box 4.2). If this in- Any aggregation of indicators into a single
formation becomes available through multi- index invariably involves a decision on how
purpose household surveys in the future or if each of the indicators is to be weighted. There
a method can be developed to apply relevant are various approaches to the selection of
administrative data at a sufficiently granular weights, including those stipulated by policy
level, then subsequent measures of multi- makers and those that are based on a poll of
dimensional well-being may reflect quality the preferences among the target population
more accurately. (Decancq and Lugo 2013). Although there are
One dimension often featured in multidi- advantages and disadvantages to each of the
mensional well-being indexes, but not con- methods, the approach chosen here follows
sidered here, is employment in a stable, dig- standard practice in the field. Dimensions are
nified job. Employment may matter beyond weighted equally, and within each dimension
the monetary benefits individuals derive each indicator is also equally weighted. The
from it because jobs can give people a sense result is that each indicator has a different
of self-esteem and help them stay connected weight depending on the number of elements
with society. An unstable employment con- within its dimension. Weights must also ad-
tract could be detrimental to well-being be- just as the number of considered dimensions
cause of the financial and other risks associ- changes, as illustrated in tables 4.2 and 4.3,
ated with such jobs. Employment is not part where the number of dimensions rises from
of the multidimensional poverty measure three to five.7
presented here for two reasons. First, many of The main summary measure presented in
the frequently used indicators of employment the chapter is the multidimensional poverty
in high-income countries, such as unemploy- headcount ratio, denoted by H. This index
ment and wage employment, are not as rel- describes the share of people who are consid-
evant in low-income countries, which have ered multidimensionally deprived and par-
very different labor market structures (Lugo allels the headcount measure used for global
2007). Second, whatever relevant indicators poverty monitoring (the poverty rate). Indi-
of employment exist, these indicators are not viduals are considered multidimensionally
available or not sufficiently harmonized in deprived if they fall short of the threshold
the different surveys considered here. in at least one dimension or in a combina-
tion of indicators equivalent in weight to a
full dimension. In other words, in the three-
Aggregating multiple indicators
dimension exercise, households will be con-
into a single index
sidered poor if they are deprived in indica-
Each of the five dimensions discussed above tors whose weight adds up to 1/3 or more.
is considered fundamental to well-being, Analogously, in the five-dimension exercise,
even if other, equally important aspects of the weights on all deprivations must add up
living standards are missing. They are im- to 1/5 or more for a household to be clas-
portant not only separately but also in the sified as poor. For example, in the three-
way they are often present or absent together. dimension case, every person who lives in a
The chapter therefore examines the share of household without access to water and sani-
people deprived according to each separate tation and with a child who does not attend
indicator, along with measures that capture school is considered multidimensionally de-
the degree to which these deprivations arise prived, whereas members of another house-
together by counting the number of depriva- hold may be deprived because the household
tions that individuals experience. In addition, income does not meet basic needs. The index
the chapter presents summary indicators that is thus a simple expression of an approach
94 POVERTY AND SHARED PROSPERITY 2018
BOX 4.2 Incorporating Aspects of Quality into Multidimensional Poverty Measures
The measure of multidimensional that could be included in standard Organization–United Nations
poverty considered in this chapter household surveys, a possible Children’s Fund Joint Monitoring
does not contain sufficient solution may involve national or Programme for Water Supply,
information to thoroughly assess subnational indicators of learning Sanitation and Hygiene (JMP)
household well-being in all major outcomes. Recently, the World developed an operational model
dimensions, especially as it Bank has harmonized data gathered for monitoring SDG 6, on safely
relates to the quality of services through international educational managed drinking water, sanitation,
utilized. Although such information testing programs—such as the and hygiene.b Safely managed
sometimes becomes available Latin American Laboratory for drinking water sources are basic
through specialized surveys, these Assessment of the Quality of drinking water sources located in
specialized surveys often do not Education, the Program for the the household, available as needed,
include all relevant dimensions Analysis of Education Systems and compliant with standards
of poverty. Therefore, the data of Confemen, the Program for on fecal and chemical content.
requirement is too large for International Student Assessment, Similarly, safely managed sanitation
multidimensional poverty indicators the Southern and Eastern Africa services are basic sanitation
to be accurately and consistently Consortium for Monitoring facilities that are not shared and
estimated across countries. In Educational Quality, and the Trends through which excreta are safely
practice, this means that the in International Mathematics disposed in situ or transported and
indicators of multidimensional and Science Study—to allow treated off-site.
poverty considered here are for comparable indicators of Measures of quality could
restricted to reporting on the learning to be computed across improve the indicator on electricity.
access of households to services, countries.a These data are core In many countries, households
but not the quality of these to the newly designed Human may have access to electricity,
services. Going forward, additional Capital Index (HCI) that the World but, because of frequent power
efforts are needed to collect richer Bank is presenting as part of the outages, the service is unreliable.
data that include both access and Human Capital Project (World This ought to be incorporated so
quality of services. Bank 2019). The HCI is a measure the indicator captures the benefits
Ensuring inclusive, equitable of human capital, designed as an derived from the electricity rather
education of high quality is indicator of each country’s future than only a binary measure of
one of the core SDGs. Access labor productivity, going beyond access. Likewise, the quality of
to education is considered a years of schooling. Specifically, maternal care could be incorporated
fundamental right, but it needs the HCI combines, for each into the indicator on the births at
to lead to “relevant and effective country, information on the level of health facilities. Many pregnant
learning outcomes” (SDG target education adjusted for quality and women may deliver at facilities,
4.1). An ideal indicator of education indicators of health status (stunting but the conditions of the facilities
in a multidimensional poverty and mortality) (Kraay 2018). and the expertise of the people
index ought to be the attainment The core drinking water and assisting the delivery can vary
by individuals of a basic level of sanitation indicators of SDG 6.1 greatly. Accurate data on the quality
learning capability (World Bank and 6.2 focus on the concept of of the facilities and the skills of
2018d). Although indicators that safely managed, which contains the staff assisting in the deliveries
account for learning outcomes a quality dimension that is not would improve the accuracy of the
are rare and might prove difficult captured in the indicators described health service indicator.
to calculate through questions in this chapter. The World Health
a. See LLECE (Latin American Laboratory for Assessment of the Quality of Education), Regional Bureau for Education in
Latin America and the Caribbean, United Nations Educational, Scientific and Cultural Organization, Santiago, Chile;
http://www.unesco.org/new/en/santiago/education/education-assessment-llece/;PASEC (Program for the Analysis of Education
Systems of Confemen) (database), PASEC and Conference of the Ministers of Education of French-Speaking Countries, Dakar,
Senegal, http://www.pasec.confemen.org/donnees/; PISA (Programme for International Student Assessment) (database),
Organisation for Economic Co-operation and Development, Paris, http://www.oecd.org/pisa/pisaproducts/; SACMEQ
(Southern and Eastern Africa Consortium for Monitoring Educational Quality) (database), SACMEQ, Gaborone, Botswana,
http://www.sacmeq.org/ReadingMathScores; TIMSS (Trends in International Mathematics and Science Study) (database),
International Association for the Evaluation of Educational Achievement, Amsterdam, http://www.iea.nl/timss.
b. See JMP (WHO/UNICEF Joint Monitoring Programme for Water Supply, Sanitation and Hygiene) (database), United Nations
Children’s Fund, New York; World Health Organization, Geneva, https://washdata.org/data.
BEYOND MONETARY POVERTY 95
TABLE 4.2 Indicator Weights: Analysis of TABLE 4.3 Indicator Weights: Analysis of
Three Dimensions Five Dimensions
Three dimensions Weights Five dimensions Weights
Income per capita 1/3 Income per capita 1/5
Child school enrollment 1/6 Child school enrollment 1/10
Adult school attainment 1/6 Adult school attainment 1/10
Limited-standard drinking water 1/9 Basic-standard drinking water 1/15
Limited-standard sanitation 1/9 Basic-standard sanitation 1/15
Electricity 1/9 Electricity 1/15
Coverage of key health services 1/10
Malnourishment (child and adult) 1/10
whereby the number of deprivations that Incidence of crime 1/10
people suffer are counted (Atkinson 2003). Incidence of natural disaster 1/10
The chapter also presents two alterna-
tive multidimensional poverty indexes (see
annex 4B for a formalization of the mea- in table 4.1, namely educational attainment
sures). The first one, the adjusted head- among adults and access to limited-standard
count measure M, combines the incidence sanitation. Considering these two indicators
of poverty H with the average breadth of alongside monetary poverty and using a sam-
deprivation suffered by each poor person, ple of 119 economies for circa 2013 (on data,
as proposed by Alkire and Foster (2011). see box 4.3.), the exercise finds 12 percent of
In addition, the chapter uses a measure the people to be monetarily poor, but, among
that penalizes for the compounding ef- them, only one individual in five is deprived
fect of multiple deprivations experienced only in the monetary dimension.8 The rest of
by the same household (Chakravarty and the 12 percent are deprived at least in either
D’Ambrosio 2006; Datt, forthcoming). As educational attainment or access to limited-
a result, if a household is deprived in any standard drinking water, with 5 percent of
two indicators, its deprivation will be con- individuals experiencing deprivations in all
sidered greater than the sum of the depri- three dimensions. At the same time, many
vations of two other households each only individuals are not monetarily poor but are
deprived on a single indicator. The measure deprived in other aspects of well-being.
is referred to as the distribution-sensitive This observation raises several questions:
multidimensional measure, denoted by D. How does our view of global poverty change
By incorporating information of the extent if poverty is defined as insufficiency not only
of deprivation suffered by individuals, both in monetary resources but also in a range of
these measures bring valuable elements to nonmonetary attributes that directly affect
the analysis. Although the three measures people’s well-being? Who are the new poor?
(H, M, and D) are presented in the chapter, In how many ways are they deprived? How do
precedence is given to the multidimensional different regions fare if a wide-angle view of
poverty headcount ratio H because it is the poverty is considered? Insights into the dif-
closest analogue to the monetary poverty ferential prevalence, nature, and distribution
headcount ratio, used to monitor the first of of multidimensional poverty in contrast to
the World Bank’s twin goals (see chapter 1 of monetary poverty can be important for the
this report). formulation of effective poverty reduction
policies. Highlighting the additional depriva-
tions experienced by the extreme poor sen-
A first global picture sitizes policy makers to the importance of
Expanding a poverty measure to include improving those aspects of human welfare
nonmonetary aspects brings into focus not captured by the monetary measure alone.
deprivations that may otherwise remain hid- This is even more important as more people
den. For example, consider a slight extension leave extreme poverty behind because a siz-
of the monetary poverty measure: the addi- able share of the non-income-poor popula-
tion of only two of the indicators described tion experiences other deprivations.
96 POVERTY AND SHARED PROSPERITY 2018
Table 4.4 describes the share of people who
are poor because of either monetary depriva-
BOX 4.3 Chapter 4: Data Overview
tion or multidimensional poverty as defined
by the three dimensions and six indicators il- This chapter relies on information from the harmonized
lustrated in table 4.1. The indicators cover the household surveys in the Global Monitoring Database (GMD) for
dimensions of monetary poverty, education circa 2013. Surveys have been included in the multidimensional
(two indicators), and access to basic infrastruc- poverty analysis if they satisfy the following criteria:
ture (three indicators). Approximately one • They include a monetary welfare measure (income or
individual in eight (11.8 percent) in the 119- expenditure) and indicators on education and basic
economy sample in circa 2013 lives in a house- infrastructure access that may be used to construct a
hold experiencing monetary poverty, whereas multidimensional poverty measure.
almost one person in five (18.3 percent) lives • The surveys were conducted within three years of 2013, that
in a multidimensionally deprived household.9 is, from 2010 to 2016.
The multidimensional measure yields a more The extreme poverty rate (headcount ratio) reported in this
expansive view of poverty by counting as poor chapter cannot be compared to the information presented in
any individual with a cumulative deprivation chapter 1 for practical and methodological reasons. For more
above the critical threshold of 1/3. details, see appendix A.
The monetary poverty measure presented
in chapter 1 outlines a bipolar world, with
Africa on one end (a high poverty rate) and
all the other regions, South Asia included, A different image of the world emerges
on the other end (a relatively low poverty through the multidimensional lens. The
rate). The separation of Sub-Saharan Africa poverty rate in Sub-Saharan Africa contin-
from the other regions is seen more clearly ues to be worryingly high, with almost two
when looking at the poverty trends over the in three individuals (64.3 percent) living in
last 25 years. East Asia and Pacific, South multidimensional poverty in circa 2013. This
Asia, and Sub-Saharan Africa all started is an increase of 40 percent from an already
with a relatively high poverty rate in 1990; high monetary poverty rate of 44.9 percent.
however, while poverty declined rapidly in South Asia, however, changes even more dra-
the first two regions, the decline was much matically. In South Asia, more than twice as
slower in Sub-Saharan Africa. Consequently, many people are multidimensionally poor as
Sub-Saharan Africa today comprises most monetarily poor (table 4.4).
of the world’s poor. If the trend contin- This raises important questions about
ues, by 2030 the extreme poor will almost the success of poverty reduction in South
exclusively be in this region. Asia. The challenge in securing higher living
TABLE 4.4 People Living in Monetary or Multidimensional Poverty, 119 Economies, circa 2013
Monetary Multidimensional
Headcount Share of the poor Headcount Share of the poor Number of Population
Region ratio (%) ratio (H) (%) economies coverage (%)
East Asia and Pacific 5.3 8.1 7.5 7.3 13 28.9
Europe and Central Asia 0.3 0.4 1.1 0.8 17 90.0
Latin America and the Caribbean 3.9 5.7 6.1 5.8 17 91.5
Middle East and North Africa 3.2 2.2 5.9 2.6 9 72.1
South Asia 11.9 12.3 26.6 17.7 5 23.0
Sub-Saharan Africa 44.9 70.9 64.3 65.4 29 60.7
Rest of the world 0.5 0.5 0.5 0.3 29 39.6
Total 11.8 100.0 18.3 100.0 119 45.0
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global Solution Group on Welfare Measure-
ment and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: The reported multidimensional headcount ratio is estimated on the basis of three dimensions—monetary, education, and basic infrastructure access, as defined in table
4.1—and an overall poverty cutoff of one-third of the weighted deprivations. The data are derived from household surveys conducted in about 2013 (+/−3 years). Because of the
unavailability or incomparability of data, analysis does not include all countries. The last column shows the percentage of regional or global populations covered by the surveys.
Percentages may not sum to 100 because of rounding.
BEYOND MONETARY POVERTY 97
standards for the population of South Asia is FIGURE 4.1 Share of Individuals in
more daunting when poverty in all its forms Multidimensional Poverty, 119 Economies,
is considered. Although South Asia is ex- circa 2013
pected to meet the goal of reducing extreme
Basic infrastructure
poverty below 3 percent by 2030, many peo-
ple will still be living in unsatisfactory con- 0.7
ditions if no progress is made in the other
components of well-being.
3.3
It is apparent from table 4.4 that the multi-
dimensional poverty headcount is always
higher than the monetary poverty headcount. 5.2
This regularity arises because of the relative
importance assigned to each component
and the stipulated overall poverty threshold 6.6
that determines if a household is considered
multidimensionally poor. If a household is
deprived in at least one dimension, then the
members are considered multidimension- 0.6 0.6
1.3
ally poor. Because the monetary dimension Monetary Education
is measured using only one indicator, any-
one who is income poor is automatically also Source: Estimates based on the harmonized household surveys in
poor under the broader poverty concept. The 119 economies, circa 2013, GMD (Global Monitoring Database),
Global Solution Group on Welfare Measurement and Capacity
difference between the headcounts therefore Building, Poverty and Equity Global Practice, World Bank, Wash-
hinges on those individuals among whom ington, DC.
the privation is a result of a shortfall in the Note: The diagram shows the share of population that is multi-
dimensionally poor, and the dimensions they are deprived in. For
nonmonetary dimensions of life despite their example, the numbers in the blue oval add up to 11.8 percent,
ability to command sufficient financial re- which is the monetary headcount. Adding up all numbers in the
sources to cross the monetary poverty thresh- figure results in 18.3 percent, which is the proportion of people
that are multidimensionally deprived.
old. These households would be deemed
nonpoor under the narrower poverty con-
cept on the basis of insufficiency in monetary cause of the relatively low correlation in depri-
resources, leaving policy makers with an un- vations across dimensions. In these countries,
duly optimistic assessment of poverty from a a household that is deprived in education at-
multidimensional perspective. tainment has a high probability of being de-
The underlying structure of the depriva- prived in school enrollment as well, making
tion experienced by the multidimensionally its members multidimensionally poor. But the
poor is depicted in figure 4.1. There is a large correlation between the monetary dimension
degree of overlap between dimensions. Only and the education indicators is weak, which
a small minority of the multidimensionally means the same households are not deprived
poor are deprived in only one dimension, in the monetary dimension. This adds new
whereas more than a third are simultaneously households to the count of the poor.
deprived in all three dimensions. The over- Because the difference in poverty incidence
lap is highest in Sub-Saharan Africa (annex according to the two measures is the result
4C, figure 4C.1). A larger overlap between of cumulative nonmonetary deprivations, it
dimensions indicates a larger extent of in- is natural to inquire about the components
terdependence, which implies that policy in- most responsible for the difference. Table
terventions targeted exclusively toward one 4.5 presents the poverty headcount ratio at
dimension may not reduce multidimensional US$1.90 a day as well as the deprivation rate
poverty and therefore a multipronged ap- associated with each of the five nonmonetary
proach might be required. indicators. Despite having made progress in
Going from monetary to multidimen- poverty reduction, the countries included in
sional poverty, the poverty rate more than the sample for South Asia still are highly de-
doubles in the five South Asian countries be- prived in the education dimension. An issue
98 POVERTY AND SHARED PROSPERITY 2018
TABLE 4.5 Individuals in Households Deprived in Each Indicator, 119 Economies, circa 2013
Educational Educational Drinking
Monetary attainment enrollment Electricity Sanitation water
Region (%) (%) (%) (%) (%) (%)
East Asia and Pacific 5.3 7.5 3.2 4.5 14.0 11.3
Europe and Central Asia 0.3 0.9 5.6 0.5 6.8 2.6
Latin America and the Caribbean 3.9 12.2 2.7 3.3 15.6 6.4
Middle East and North Africa 3.2 11.1 7.9 3.8 14.6 4.2
South Asia 11.9 31.6 22.6 23.8 39.5 7.0
Sub-Saharan Africa 44.9 46.2 20.8 64.8 61.9 33.9
Rest of the world 0.5 1.2 0.0 0.0 0.6 0.0
Total 11.8 17.0 9.0 15.9 23.8 10.9
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global Solution Group on Welfare Measure-
ment and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: The definition of the indicators and the deprivation thresholds are as follows: Monetary poverty: a household is deprived if income or expenditure, in 2011 purchasing
power parity U.S. dollars, is less than US$1.90 per person per day. Educational attainment: a household is deprived if no adult (grade 9 equivalent age or above) has completed
primary education. Educational enrollment: a household is deprived if at least one child (grade 8 equivalent age or below) is not enrolled in school. Electricity: a household is
deprived if it does not have access to electricity. Sanitation: a household is deprived if it does not have access to even a limited standard of sanitation. Drinking water: a house-
hold is deprived if it does not have access to even a limited standard of drinking water. The data reported refer to the share of people living in households deprived according to
each indicator.
of apparent global concern is poor sanitation: rate, but Pakistan’s level of deprivation in
approximately a quarter of the population in education attainment and enrollment is far
the 119-economy sample lives in households higher than that of Vietnam (Table 4C.4).
lacking access to even a limited standard of These countries typify the development ex-
sanitation. The populations in regions with perience of the two regions. Expansion in
low monetary poverty like East Asia and Pa- access to education preceded or was contem-
cific, Latin America and the Caribbean, and poraneous with the growth in income in East
the Middle East and North Africa suffer a san- Asia, whereas despite rising incomes human
itation deprivation rate several times as high development has lagged in South Asia (World
as that in the monetary dimension. Globally, Bank 2018d). Iraq experiences the highest
almost one individual in six is not connected deprivation in the education dimension, and
to electricity. Yet this is overwhelmingly a it is one of the few countries where school
South Asian and Sub-Saharan African phe- enrollment outcome is worse than education
nomenon: approximately one South Asian in attainment. Over the last 15 years, access to
four and two Sub-Saharan Africans in three schooling in Iraq has been disrupted because
lack electricity at home. of conflict, which is a reminder that progress
An examination of deprivation rates, one cannot be taken for granted, especially in
indicator at a time, generally confirms that fragile and conflict-affected situations.
the regional ranking for any one indicator is The examination of indicator deprivation
consistent with the others. Regions more de- rates does not reveal information about the
prived in one indicator are highly likely to be simultaneity of deprivations. To consider this
more deprived in other indicators. However, aspect, other tools are needed. One of the
there are anomalies. For example, the Europe simplest approaches involves counting the
and Central Asia region shows the lowest in- number of indicators in which people are de-
cidence of monetary poverty; however, the prived contemporaneously. Figure 4.2 shows
share of people deprived in school enroll- the shares of individuals deprived according
ment in the region is higher than in both the to the maximum of six indicators. Approxi-
East Asia and Pacific and the Latin America mately 60 percent of people in the 119 econ-
and Caribbean regions. omies are not deprived in any of the six indi-
Important insights on the pattern of de- cators. More than 80 percent of Sub-Saharan
velopment can be gleaned from country Africans exhibit at least one deprivation, but
outcomes as well. For example, Pakistan and a smaller share of South Asians (65.6 per-
Vietnam both have a low absolute poverty cent) experience at least one deprivation; as
BEYOND MONETARY POVERTY 99
FIGURE 4.2 Share of Individuals Deprived in at Least a Given Number The adjusted headcount measure M defined
of Indicators, 119 Economies, circa 2013 in the previous section is sensitive to both
the incidence and breadth of multidimen-
100
sional poverty. If a poor household becomes
90 deprived in additional elements, the changes
are registered by the measure—something
80
that will not be captured by the headcount H.
Share of population (%)
70 The adjusted headcount measure, however,
does not take into account the deprivations
60 of households deemed to be multidimension-
50 ally nonpoor. This can ignore a substantial
portion of deprivation. Of the total popula-
40 tion in the sample, 15.5 percent is deprived
30
in only one indicator and another 8.2 per-
cent deprived in two indicators (table 4.6). A
20 subset of these households is not identified
as multidimensionally poor because their
10
total weighted deprivation does not cross the
0 poverty threshold of one-third. In fact, most
1 2 3 4 5 6 individuals experiencing one deprivation
Number of indicators deprived in and two-thirds of individuals experiencing
East Asia and Pacific Rest of the world two deprivations are not multidimensionally
Europe and Central Asia South Asia poor. They face an average of 0.13 and 0.25
Latin America and the Caribbean Sub-Saharan Africa weighted deprivations, respectively, which is
Middle East and North Africa Total missed by the intensity-sensitive measure.
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD
The picture of poverty can shift yet again
(Global Monitoring Database), Global Solution Group on Welfare Measurement and Capacity Building, under the distribution-sensitive measure D,
Poverty and Equity Global Practice, World Bank, Washington, DC. the third measure, because it differs from the
adjusted headcount measure in two crucial
ways. Unlike the adjusted headcount mea-
the number of deprivations rises, a large gap sure, the distribution-sensitive measure is
opens between South Asia and Sub-Saharan not associated with a prespecified poverty
Africa. Whereas 20.5 percent of South Asia’s threshold so it counts deprivations experi-
population is deprived in three or more in- enced by all households. Second, it penalizes
dicators, 55.1 percent of Africans are so de- compounding deprivations such that poverty
prived. On the shares experiencing four or is higher when one household experiences
more deprivations, South Asia catches up to two deprivations than when two households
the world at large. Thus, in addition to the experience one deprivation each.
relatively larger share of Sub-Saharan Afri- The regional estimates for multidimen-
cans who are deprived in each dimension, sional headcount, adjusted headcount, and
Sub-Saharan Africans suffer from a greater distribution-sensitive measures are presented
average number of deprivations than people in table 4.7. Because the scales of the two
elsewhere. measures do not lend themselves to easy
comparison, the focus is on the regional con-
tribution to global poverty under each ap-
Incorporating breadth of poverty
proach. Moving from multidimensional pov-
into the measurement erty headcount (H) to the intensity-sensitive
Summarizing the information on the num- measure (M), the concentration of poverty
ber of deprivations into a single index proves shifts further to Africa. This shift is driven by
useful in making comparisons across popula- the breadth of deprivation in Sub-Saharan
tions and across time. Aggregate multidimen- Africa, which is twice as high as in South Asia
sional poverty measures provide an easy way and several times higher than in other re-
to rank countries and monitor their progress. gions of the world (table 4.7).
100 POVERTY AND SHARED PROSPERITY 2018
TABLE 4.6 The Multidimensionally Poor and the Breadth of Deprivation, by Number of Deprivations, 119 Economies,
circa 2013
Multidimensional poverty status Breadth of deprivation
Number of Share of the population
deprivations (%) Nonpoor (%) Poor (%) Nonpoor Poor
0 62.0 62.0 0.0 0.00 n.a.
1 15.5 14.1 1.4 0.13 0.33
2 8.2 5.7 2.5 0.25 0.43
3 6.0 0.0 6.0 n.a. 0.48
4 4.8 0.0 4.8 n.a. 0.65
5 2.8 0.0 2.8 n.a. 0.83
6 0.7 0.0 0.7 n.a. 1.00
Total 100.0 81.7 18.3 0.04 0.58
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global Solution Group on Welfare Measure-
ment and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: A household is multidimensionally poor if it is deprived in more than a third of weighted deprivations. Breadth of deprivation refers to the average number of deprivations
relative to the total number of indicators. It varies from 0 to 1, where 1 represents a person deprived in all six indicators. The shares may not sum to 100 because of rounding.
n.a. = not applicable.
TABLE 4.7 Regional Contributions to Multidimensional Poverty, 119 Economies, circa 2013
Multidimensional Adjusted headcount Distribution-sensitive
headcount (H) measure (M) measure (D)
Breadth of Share of the
Region deprivation population (%) H Contribution (%) M Contribution (%) D Contribution (%)
East Asia and Pacific 0.07 17.8 7.5 7.3 0.03 5.8 0.02 5.5
Europe and Central Asia 0.02 13.3 1.1 0.8 0.00 0.5 0.01 0.9
Latin America and the Caribbean 0.07 17.4 6.1 5.8 0.03 4.7 0.02 5.1
Middle East and North Africa 0.06 8.1 5.9 2.6 0.03 2.1 0.02 2.2
South Asia 0.21 12.1 26.6 17.7 0.14 15.9 0.09 15.2
Sub-Saharan Africa 0.44 18.6 64.3 65.4 0.40 70.8 0.29 70.9
Rest of the world 0.00 12.7 0.5 0.3 0.00 0.2 0.00 0.2
Total 0.14 100.0 18.3 100.0 0.11 100.0 0.07 100.0
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global Solution Group on Welfare Measure-
ment and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: Breadth of deprivation refers to the average number of deprivations relative to the total number of indicators. It varies from 0 to 1, where 1 represents a person deprived
in all six indicators.
The distribution of global poverty is An appealing feature of the adjusted head-
subject to two countervailing effects when count measure M is that the overall measure
going from the intensity-sensitive measure can be easily decomposed into the relative
(M) to the distribution-sensitive measure contribution of each indicator. Such de-
(D). Counting all deprivations pushes the compositions matter for understanding the
distribution of poverty to regions that have drivers of multidimensional poverty, and the
few multidimensionally poor but many who sectors that ought to be given priority in the
suffer from at least one deprivation. At the design of poverty-alleviating policies. If the
same time, assigning more importance to poverty rate is high because of income in-
compounding deprivations pulls it toward sufficiency, a focus on economic growth or
regions with high breadth of deprivation. income support is appropriate; but, if edu-
The first effect more than offsets the second cation or access to utilities plays a dominant
in Europe and Central Asia, Latin America role in multidimensional poverty, invest-
and the Caribbean, and the Middle East and ments in the corresponding sectors may yield
North Africa, resulting in a slightly higher the highest returns to poverty reduction.
contribution of these regions to global pov- In high-income countries, multidimen-
erty under D than under M (table 4.7). sional poverty, though extremely low, almost
BEYOND MONETARY POVERTY 101
FIGURE 4.3 Contribution of Indicators to the Adjusted Headcount Measure (M), 119
Economies, circa 2013
Total
Sub-Saharan Africa
South Asia
Rest of the world
Middle East and North Africa
Latin America and the Caribbean
Europe and Central Asia
East Asia and Pacific
0 20 40 60 80 100
Percent
Monetary Educational attainment Educational enrollment
Electricity Limited-standard sanitation Limited-standard water
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global
Solution Group on Welfare Measurement and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
exclusively arises because of insufficient in- poverty is predominantly a rural phenome-
come given the near-universal access to edu- non: 45.8 percent of the total sample popula-
cation and infrastructure services (figure 4.3). tion is rural, but 81.3 percent of the monetary
For the multidimensionally poor in Europe poor are living in rural areas (annex 4C, table
and Central Asia, access to electricity is a much 4C.1). If poverty is considered more broadly
more important driver of poverty than else- with the multidimensional lens, the distribu-
where. The comparison across Sub-Saharan tion of poverty tilts even more toward rural
Africa and South Asia reveals how the underly- areas. Thus, 83.5 percent of the multidimen-
ing structure of deprivations differs across the sionally poor are rural dwellers, implying that,
two regions. In South Asia, the education di- relative to urban households, rural house-
mension has a disproportionate contribution holds suffer cumulatively more deprivations
to poverty (46 percent), whereas the contribu- in access to education and essential utilities.
tion of monetary poverty is relatively low (24.6 The most pronounced shifts of poverty to-
percent). In Sub-Saharan Africa, the services ward rural areas are observed in East Asia and
(39.7) and the monetary (36.1) dimensions Pacific and in Latin America and the Carib-
contribute the most to multidimensional pov- bean (figure 4.4). In these regions, the shift in
erty, and the education dimension contributes the composition is largely driven by depriva-
the least (24.2 percent). This may suggest a dif- tions in limited-standard sanitation and adult
ferent policy focus in the two regions. The pri- educational attainment. In contrast, poverty
ority in these South Asian countries should be becomes more urban in the Middle East and
wider access to education whereas expansion North Africa and South Asia, suggesting that
of basic infrastructure services will have the urban residents in these regions, although not
strongest impact in Sub-Saharan Africa. monetarily poor, experience deprivations in
some of these additional aspects of life.
With respect to household composition,
Who are the monetarily and
households with children are overrepresented
multidimensionally poor?
among both the monetary poor and the mul-
As the definition of poverty broadens to in- tidimensionally poor, regardless of the gender
clude additional aspects of deprivation, the or number of adults in the household (figure
composition of the poor changes. Monetary 4.5; also annex 4C, table 4C.2).10 The shift
102 POVERTY AND SHARED PROSPERITY 2018
from an exclusively monetary approach to a FIGURE 4.4 Difference in the Share of the Poor in Rural Areas,
multidimensional account of poverty does Multidimensional Headcount vs. Monetary Headcount, 119
not substantially change the demographic Economies, circa 2013
composition of the poor, though house-
10
holds with only one adult woman (with or
without children) represent a slightly larger 8
share in the latter case (8.8 percent compared
Percentage point difference
with 8.1 percent). All indicators included in 6
this chapter are measured at the household
level and thus do not distinguish differences 4
within households. The estimates also assume
2
that resources are distributed equally within
a household, that all household members 0
have similar needs, and that there are no scale
economies in larger households. Assessing –2
individual well-being requires measuring
intrahousehold resource allocation and the –4
East Asia Europe Latin Middle Rest South Sub- Total
needs of each household member. Chapter 5 and and America East and of the Asia Saharan
investigates methods that estimate individual Pacific Central and the North world Africa
Asia Caribbean Africa
well-being from underlying household data.11
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD
(Global Monitoring Database), Global Solution Group on Welfare Measurement and Capacity Building,
A deeper look Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: The lines indicate the difference in percentage points of the rural share of the poor when com-
Extending monetary poverty by including paring multidimensional and monetary poverty. A positive value indicates that the rural share of the poor
measures of access to education and basic in- is greater with the multidimensional measure.
frastructure services changes the understand-
ing of poverty. However, even this extension FIGURE 4.5 Contribution to Monetary and Multidimensional Poverty,
to three dimensions fails to capture other by Household Type, 119 Economies, circa 2013
key dimensions of well-being. This section
augments multidimensional poverty by also 100
including measures of access to health care 90
services and lack of security. The analysis is
carried out on six countries for which in- 80
formation on households from a single data 70
source is available. This exercise is exploratory
in nature and the numbers presented might 60
Percent
diverge from recent official sources (and even 50
from the analysis performed in the previous
40
section) because in all but one country the
analysis is based on different household sur- 30
veys than the one used for calculating mone-
20
tary poverty. Instead, it uses surveys that are
comprehensive enough to include the addi- 10
tional dimensions. The purpose of the exer-
0
cise is to illustrate the gains and insights that Population Monetary poor Multidimensional poor
could emerge if this information was avail-
Only children Two adults, with child
able for a larger set of countries.
Only seniors One adult male, with no children
Accounting for the two extra dimensions Multiple adults, with no children One adult male, with children
of well-being further enhances the under- Multiple adults, with children One adult female, with no children
standing of poverty. The proportion of peo- Two adults, without child One adult female, with children
ple identified as poor under the expanded
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD
definition is higher than with the three- (Global Monitoring Database), Global Solution Group on Welfare Measurement and Capacity Building,
dimensional measure, suggesting that the Poverty and Equity Global Practice, World Bank, Washington, DC.
BEYOND MONETARY POVERTY 103
share of individuals who are unnoticed by tively balanced view of how countries might
monetary poverty measures could be even fare after the multidimensional poverty mea-
higher than reported in the previous section. sure is extended.12
Including health and security can also shift Summary analysis of the data reveals that
the common understanding of who the poor deprivation rates vary greatly by country (table
are and where they are located. Specifically, 4.8). Monetary poverty ranges from 2 percent
acknowledging deprivations along these two in Ecuador to 44 percent in Tanzania.13 Only
dimensions reveals that a larger share of the 1 percent of the population does not have ac-
poor live in female-headed households and, cess to electricity in Ecuador, Indonesia, and
in several cases, shifts poverty back toward Iraq, whereas the same measure is as high as
urban areas. 87 percent in Uganda. The countries also ex-
hibit different deprivation rates in the newly
added dimensions. More than 43 percent of
The six-country sample
individuals in Tanzania live in households
The extended measure of poverty is com- where at least one child is stunted, whereas
puted for six countries—Ecuador, Indonesia, the same deprivation rate for Mexico is 15
Iraq, Mexico, Tanzania, and Uganda—and percent. The country ranking on the crime
covers the years 2009–14 (see appendix A for indicator is nearly the reverse of the rankings
details on the surveys used). These countries on the other indicators. The upper-middle-
have primarily been chosen on the basis of income countries in the sample—Ecuador,
data availability. In each of these countries, Iraq, and Mexico—suffer from high crime
a household survey has been conducted re- rates and substantial insecurity in comparison
cently that collected information relevant to with the low-income countries, Tanzania and
the five dimensions of poverty in a compa- Uganda. The share of individuals affected by
rable manner. The six countries include low- a natural disaster also differs markedly across
income, lower-middle-income, and upper- the six countries. Uganda stands out as the
middle-income countries, as well as all World least well performing country; there, nearly
Bank regions except Europe and Central Asia a third of the population was affected by a
and South Asia. They therefore offer a rela- drought in the year leading up to the survey.
TABLE 4.8 Share of Individuals Deprived, by Indicator, Selected Countries
Percent
Dimension Indicator Ecuador Indonesia Iraq Mexico Tanzania Uganda
Monetary poverty Daily consumption < $1.9 2.0 3.5 2.5 9.2 43.6 35.8
Education Any school-aged child is not enrolled in school 2.2 3.6 26.0 10.4 32.2 15.4
No adult has completed primary education 4.8 5.3 12.6 5.3 13.9 26.1
Access to basic No access to basic-standard drinking water 11.3 19.0 13.4 3.7 54.6 54.0
infrastructure No access to basic-standard sanitation 14.1 26.6 13.5 19.4 74.5 77.0
No access to electricity 1.2 0.8 0.7 4.3 79.7 87.2
Health No facility delivery 6.8 16.6 11.7 4.6 36.7 30.8
No DPT3 vaccination 3.6 33.6 — 11.9 — 8.4
Any child is stunted 25.7 41.8 40.5 15.0 43.4 40.7
Any female is malnourished 3.5 10.5 6.0 5.3 13.6 —
Security Experienced or in threat of crime 33.0 6.9 21.1 16.4 1.8 5.1
Affected by natural disaster 2.9 0.9 3.0 0.1 5.6 32.3
Source: Calculations based on Ecuador’s Encuesta de Condiciones de Vida 2013–14; Indonesian Family Life Survey, 2014; Iraq Household Socio-Economic Survey, 2012; Mexican
Family Life Survey, 2009–12; Tanzania’s National Panel Survey, 2012–13; Uganda National Panel Survey 2013–14. See appendix A for details.
Note: Monetary poverty rates might differ from recent official estimates because, in all cases except for Iraq, this exploratory analysis is based on different household surveys
than the ones used to calculate official monetary poverty, as reported in chapter 1 and earlier in this chapter. When an indicator is not available for the particular country,
weights are shifted to the other indicators in the dimension. A household has access to a basic-standard drinking water if its drinking water comes from an improved source
(for example, piped, borehole, protected dug well, rainwater, or delivered water) within a round trip time of 30 minutes. A household has access to basic-standard sanitation if
it is using improved sanitation facilities (for example, flush/pour flush to piped sewer system, septic tank, or a composting latrine) and the facility is for the exclusive use of the
household. — = not available; DPT3 = diphtheria-pertussis-tetanus vaccine.
104 POVERTY AND SHARED PROSPERITY 2018
With the addition of health and security FIGURE 4.6 Share of Individuals Deprived in at Least a Given Number
indicators, the share of individuals deprived of Indicators, Selected Countries
in at least one indicator is troublingly high
100
(figure 4.6). In Tanzania and Uganda, as
many as 95 percent of the population is de-
prived in at least one indicator. Even in the
80
Share of the population (%)
top-performing countries, Ecuador and
Mexico, more than half the population is
deprived in at least 1 of the 12 indicators. If
a household is considered worthy of atten- 60
tion when it is deprived in any of the rele-
vant indicators, then monetary poverty and
even multidimensional poverty measures in 40
three dimensions fail to capture many house-
holds. The number of deprivations people
experience declines rapidly as the deprived 20
indicator count increases, and virtually no
one is deprived in all 12 indicators (or 11
or 10) in any country. Yet the decline occurs 0
more quickly in some countries than in oth- 1 2 3 4 5 6 7 8 9 10 11 12
ers. In Tanzania and Uganda, about half of Number of indicators deprived in
the population is deprived in five indicators, Ecuador Indonesia Iraq
highlighting the compounded disadvantages Mexico Tanzania Uganda
many households suffer in these countries. Source: Calculations based on Ecuador’s Encuesta de Condiciones de Vida 2013–14; Indonesian Family
Life Survey, 2014; Iraq Household Socio-Economic Survey, 2012; Mexican Family Life Survey, 2009–12;
Tanzania’s National Panel Survey, 2012–13; Uganda National Panel Survey 2013–14. See appendix A for
Comparing alternative measures details.
of poverty
Because of the frequency of cumulative more stringent definitions in the services di-
deprivations, headcount ratios rise several- mension, or the correlational structure link-
fold in some countries if one shifts from ing the various dimensions. The last reason
monetary poverty to the multidimensional may be less apparent, but it is conceptually
poverty measure in five dimensions (figure important: if households deprived in any
4.7). In Iraq, 2.5 percent of the population of the added dimensions were already de-
are counted among the monetary poor; 10.4 prived according to the three-dimension
percent are poor if three dimensions are measures, implying that the correlation be-
considered (with a cutoff of one-third); and tween the deprivations are high, then adding
28.4 percent are poor if five dimensions are new dimensions need not raise the poverty
considered (with a cut-off of one-fifth). Pov- headcount rates. Conversely, if the new di-
erty rates climb by an average 41 percent if mensions are uncorrelated or, especially, neg-
the five-dimension measure is used instead of atively correlated with deprivation according
the three-dimension measure. Clearly, as the to the three-dimension measure, then the ad-
poverty measure becomes more comprehen- dition of the new dimensions may lead to an
sive and deprivation in a single dimension upward surge in poverty rates. Similar to the
(or indicators whose weights add up to that three-dimension multidimensional measure
of a single dimension) continues to define above, decompositions of the adjusted head-
poverty, the count of individuals living in count ratios (M) can be used to unpack how
poverty rises. much the different dimensions contribute to
The headcount ratios mask the dimen- poverty in each of the countries studied.
sions and indicators driving the rise in pov- The addition of the health and security
erty rates, and those dimensions and indica- dimensions to the three-dimension measure
tors vary across countries. The increase may shifts the drivers of poverty in several coun-
be caused by any of the added dimensions, tries (figure 4.8). Measured in three dimen-
BEYOND MONETARY POVERTY 105
FIGURE 4.7 The Headcount Ratio, by Alternative Poverty Measures, Selected Countries
80 76.0
71.9
70
63.3
60
54.8
Share of population (%) 50
43.6
40 35.8
30 28.4
20 16.1 16.2
13.3
10.3 10.4 9.2
10 5.9 7.3
2.0 3.5 2.5
0
Ecuador Indonesia Iraq Mexico Tanzania Uganda
Monetary poverty Multidimensional poverty (3 dimensions) Multidimensional poverty (5 dimensions)
Source: Calculations based on Ecuador’s Encuesta de Condiciones de Vida 2013–14; Indonesian Family Life Survey, 2014; Iraq Household
Socio-Economic Survey, 2012; Mexican Family Life Survey, 2009–12; Tanzania’s National Panel Survey, 2012–13; Uganda National Panel
Survey 2013–14. See appendix A for details.
Note: The figure shows the share of the population that is considered poor under three different definitions of poverty. Monetary poverty =
individuals living on less than US$1.90 a day. Multidimensional poverty (three dimensions) = individuals deprived in at least 33 percent of
the (weighted) indicators according to the multidimensional headcount measure; the dimensions considered are monetary poverty, edu-
cation and access to basic infrastructure. Multidimensional poverty (five dimensions) = individuals deprived in at least 20 percent of the
(weighted) indicators according to the multidimensional headcount measure and considering all five dimensions. Each dimension in the
three-dimension measure is weighted 0.33. Each dimension in the five-dimension measure is weighted 0.20. In the multidimension mea-
sures, each indicator is weighted equally within dimensions. Monetary poverty rates might differ from recent official estimates because,
in all cases except for Iraq, this exploratory analysis is based on different household surveys than the ones used to calculate official mon-
etary poverty, as reported in chapter 1 and earlier in this chapter.
FIGURE 4.8 Contribution to Multidimensional Poverty (M), by Dimension, Selected Countries
100
Contribution to total poverty (%)
80
60
40
20
0
3 5 3 5 3 5 3 5 3 5 3 5
Dimensions
Ecuador Indonesia Iraq Mexico Tanzania Uganda
Monetary Education Basic infrastructure Health Security
Source: Calculations based on Ecuador’s Encuesta de Condiciones de Vida 2013–14; Indonesian Family Life Survey, 2014; Iraq Household
Socio-Economic Survey, 2012; Mexican Family Life Survey, 2009–12; Tanzania’s National Panel Survey, 2012–13; Uganda National Panel
Survey 2013–14. See appendix A for details.
Note: The figure shows the contribution of each dimension to the adjusted-headcount ratio M based on the dimensional breakdown
method of Alkire et al. (2015).
106 POVERTY AND SHARED PROSPERITY 2018
sions, deprivations in the education dimen- of the households that suffer from crime do
sion are behind two-thirds of the headcount not experience other deprivations, and hence
ratio in Iraq. If the five-dimension measure do not meet the criteria for classification
is used, the role of educational deprivations among the poor. Consequently, security con-
decreases noticeably, and the two extra di- tributes only modestly to multidimensional
mensions are behind roughly half the poverty poverty in Mexico. In Tanzania and Uganda,
headcount. Particularly, health deprivations health care deprivations are positively cor-
emerge as an area with large contributions to related with monetary poverty, education
poverty in Iraq. In contrast, in Tanzania and deprivations, and deprivation in services.
Uganda, the two new dimensions account Yet, because many households already meet
for only 20 percent of poverty; and, in both the cutoff for classification among the poor
the three-dimension measure and the five- without adding the health care dimension,
dimension measure, monetary poverty and the dimension does not contribute much to
lack of access to basic infrastructure services the ranks of the poor.
are the major contributors to poverty.
These effects are partially driven by the
Poverty profiling with five
extent to which the deprivations tend to ap-
dimensions of well-being
pear together, and the number of depriva-
tions experienced by households. In Ecuador The correlational structure between the di-
and Mexico, monetary poverty and threat of mensions of well-being and their association
crime are negatively correlated, implying that with population characteristics may change
the two indicators capture different types the composition of the poor and the corre-
of households; households that suffer from sponding policy actions needed to reduce
monetary poverty are less likely to suffer from poverty. In Ecuador and Iraq, where the
deprivations associated with crime relative to contribution to poverty from the security
households that do not suffer from monetary dimension is relatively large, many of the
poverty. When deprivations linked to crime individuals suffering from threats of crime
are included in the measure of multidimen- reside in urban centers. As a result, the share
sional poverty, many new households may be of the poor who reside in urban areas in Iraq
added to the ranks of the poor, which is the rises from 31 percent to 44 percent if the
case in Ecuador. In the case of Mexico, many focus shifts from monetary poverty to five-
FIGURE 4.9 The Poor, by Sociodemographic Characteristics, Selected Countries
a. Poor living in urban areas b. Poor living in female-headed households
60 14
50 12
Share of poor (%)
Share of poor (%)
10
40
8
30
6
20
4
10 2
0 0
or
ia
q
ico
ia
a
or
ia
q
o
ia
a
Ira
d
Ira
ic
nd
es
an
es
an
d
ad
an
ex
ex
ua
a
on
on
nz
nz
u
Ug
Ug
M
M
Ec
Ec
Ta
Ta
Ind
Ind
Monetary poor Monetary poor
Multidimensionally poor (5 dimensions) Multidimensionally poor (5 dimensions)
Source: Calculations based on Ecuador’s Encuesta de Condiciones de Vida 2013–14; Indonesian Family Life Survey, 2014; Iraq Household
Socio-Economic Survey, 2012; Mexican Family Life Survey, 2009–12; Tanzania’s National Panel Survey, 2012–13; Uganda National Panel
Survey 2013–14. See appendix A for details.
BEYOND MONETARY POVERTY 107
MAP 4.1 Provincial Poverty Rates, Ecuador
a. Monetary measure b. Three-dimension measure c. Five-dimension measure
PA C I FI C PACIFIC PAC I F I C
OCEAN OCE AN OC E AN
QUITO QUITO QUITO
Poverty rate (%)
> 35
35
30
25
20
15
10
5
0
IBRD 43980 | OCTOBER 2018
Source: Calculations based on Ecuador’s Encuesta de Condiciones de Vida 2013–14. See appendix A for details.
dimension poverty, and similarly from 18 within a country. In Ecuador, for example,
percent to 37 percent in Ecuador (figure 4.9). the thinly populated province of Pastaza is
In contrast, in Mexico, Tanzania and Uganda, one of several eastern provinces with high
where the security addition had a relatively poverty rates according to the monetary pov-
small contribution to total poverty, urban erty measure, but it has an exceptionally high
poverty rates change only marginally in re- poverty rate according to the extended mul-
sponse to the addition of more dimensions. tidimensional poverty measure (map 4.1).
In Indonesia, where health deprivations Similar changes occur in other countries,
make up the greatest contribution to poverty, suggesting that the geographical concentra-
the share of poor in urban areas decreases, tion of poverty shifts if more dimensions are
suggesting that lack of health care primarily considered. This may have important impli-
is germane to rural areas. cations for policies aiming to eliminate the
Adding more dimensions also highlights pockets of poverty and for the allocation of
differences in the types of households con- resources across regions within a country.
sidered poor. If the five-dimension measure
is used instead of the monetary poverty mea-
sure, the share of the poor living in female-
Conclusion
headed households, defined as households in Monetary poverty is the World Bank’s work-
which the only adult is a woman, increases in horse measure to assess progress in poverty
all six countries in the sample except Tanza- reduction across the world. This chapter ex-
nia. In Indonesia, the shift causes the poverty amines the effects of extending the measure of
rate among individuals in female-headed poverty by adding nonmonetary dimensions
households to rise from less than the average in an attempt to broaden the measurement of
rate to more than the average rate, hence tar- well-being. The analysis should be viewed as
geting female-headed households becomes a starting point for a deeper investigation of
an important means to combat poverty. the measurement of poverty that recognizes
As the composition of poverty changes, that many dimensions of well-being are not
so does the spatial concentration of poverty all readily available through markets.
108 POVERTY AND SHARED PROSPERITY 2018
In addition to income and consumption, human suffering. Although this appreciation
up to four other dimensions of poverty are is not new or original, elevating additional
included in the analysis, represented by a aspects of well-being to the same level as
total of 12 indicators of well-being. Although consumption or income poverty can high-
there are many other valuable indicators that light the relevance of those aspects in com-
could have been included in the portfolio of parison to an exclusive focus on monetary
nonmonetary indicators, the selected indica- poverty.
tors satisfy explicit principles, including the Going forward, the World Bank will mon-
centrality of private consumption, data avail- itor progress on multidimensional poverty
ability and parsimony. using the three-dimension poverty head-
The consideration of access to education count presented in this chapter. However, the
and basic infrastructure alongside income, in empirical challenges of a multidimensional
a sample of 119 economies for circa 2013 re- poverty measure, especially at the global
veals that about a third of those that are mul- level, are great. The analysis described in this
tidimensionally deprived are not captured by chapter relies heavily on available data for the
monetary poverty. The most prevalent depri- various components of well-being. The data
vation is access to adequate sanitation, which on 119 economies had to have been stan-
is associated with higher deprivation rates dardized so indicators on education and util-
than income. In the exploratory analysis for ities could be examined alongside consump-
six countries in which indicators of health, tion. However, household consumption or
nutrition, and security are added to the anal- income surveys often lack adequate informa-
ysis of poverty, new aspects of deprivation tion on many key aspects of well-being, such
are uncovered. In some cases, the incidence as health, nutrition, and security. Thus, the
of crime or the threat of crime is weakly or extended analysis on additional dimensions
even negatively associated with monetary of poverty was restricted to six countries.
poverty. This implies that the characteristics These exercises are also suboptimal because
of the poor shift as the definition of poverty information on the quality of the related ser-
is broadened to include security. For several vices is missing. Richer datasets harmonized
countries, a larger share of the multidimen- with respect to the measurement of essential
sional poor live in urban areas and in fe- service access and quality are needed. This
male-headed households. appeal does not necessarily mean that already
A growing toolbox for the assessment of lengthy household survey questionnaires
well-being enhances the understanding of should be lengthened further. Where possi-
poverty. In some regions, deprivations in one ble, alternative information sources, such as
dimension are accompanied by deprivations administrative data or vital statistics, can be
in other dimensions, whereas this is not the combined with survey data at relatively little
case for other regions. This has important additional cost in order to broaden the un-
implications for policies aimed at reducing derstanding of well-being.
BEYOND MONETARY POVERTY 109
Annex 4A
Comparison of indicators used in
multidimensional poverty measures
TABLE 4A.1 Dimensions and Indicators
World Bank EU Social Indicators
(3 and 5 dimensions) UNDP–OPHI MPI Mexico (selected)
Monetary (and Consumption or income Income below national Income below 60% of median
living standard) below $1.90 well-being threshold national equivalized income
Housing Housing
Assets Assets
Basic Electricity Electricity Electricity
infrastructure Drinking water Drinking water Drinking water
Sanitation Sanitation Sanitation
Cooking fuel Cooking fuel
Education Adult school attainment Adult school attainment Complete level of education Early school-leavers (ages 18–24)
(years of schooling)
Child school enrollment Child school attendance School attendance
Health and Coverage of vaccination Child mortality Infant mortality
nutrition Coverage of birth attendance Coverage of health service
Nutrition (children and adults) Nutrition (children and adults) Access to food
Life expectancy
Self-reported unmet need for
health care
Security Incidence of crime
Incidence of natural disasters
Employment Access to social security Jobless households
Employment of older workers
Sources: OPHI 2018; and World Bank 2017b.
Note: Indicators in blue reflect those that are included in the World Bank’s multidimensional poverty measure for five dimensions. EU = European Union; MPI = Multidimensional
Poverty Index; OPHI = Oxford Poverty and Human Development Initiative; UNDP = United Nations Development Programme.
110 POVERTY AND SHARED PROSPERITY 2018
Annex 4B
Multidimensional poverty measures:
A formalization
The adjusted headcount measure M was developed by Alkire and Foster (2011), as a special
case of the Alkire–Foster family of multidimensional poverty measures. One of the main char-
acteristics of the measure is that it uses a dual cutoff. The first cutoff is the specific sufficiency
threshold for each dimension. The second cutoff is often identified by the parameter k and rep-
resents the number of (weighted) deprivations needed before an individual may be considered
multidimensionally deprived. The deprivations among individuals who are poor in at least k
dimensions are aggregated for an entire society as follows:
L G N
4IJ
) \6 - 4] S [ f[ 2J ^ Q _ (IJ g ( \+I V -] (4B.1)
. 5J
I=: J=:
where yij is the achievement of person i on dimension j ; z j is the sufficiency threshold for di-
mension j ; Iij is a dimension-specific indicator function that takes the value of 1 if yij < zj and
0 otherwise; α is a parameter of the measure’s sensitivity to the depth of poverty; and I(ci ≥ k)
is a poverty indicator function that equals 1 if the number of (weighted) dimensions in which
the individual is deprived is at least equal to the parameter k. The measure M(α, k ; y) is de-
composable across population groups, which can facilitate a regional analysis and is useful for
targeting. It also satisfies several desirable properties, including dimensional breakdown, which
is useful to understand the contribution of each dimension to overall poverty.
The most common application of the measure involves setting α equal to zero. This
special case is known as the adjusted headcount ratio (M), and is defined as the share of
multidimensionally poor households multiplied by the average number of deprivations expe-
rienced by the multidimensionally poor. This case is used more frequently because, in many
applications, some indicators are categorical, and thus higher values of α are not appropriate.
This measure can be seen as more appealing than the multidimensional headcount H because
it incorporates information on the breadth of poverty. The special case included in the pres-
ent chapter is as follows:
For α = 0, k = 1 3
, then
L G
) S [ d[ 2J (IJ e ( X+I V Z S ' R % (4B.2)
.
I=: J=:
where H is the multidimensional headcount rate, that is, the share of individuals who are
multidimensionally deprived, and A is the average number of deprivations among those in-
dividuals who are multidimensionally deprived. This chapter first reports H as a summary
measure across countries and regions and then H ϫ A.
BEYOND MONETARY POVERTY 111
Datt (forthcoming) proposes an alternative family of multidimensional poverty measures,
known as the distribution-sensitive multidimensional poverty measures. The measure pro-
posed does not make use of a dual cutoff, recognizing the essentiality of every deprivation.
Every deprivation is counted toward the measurement of poverty even if a person is deprived
in a single indicator with low weight. In addition, the measure penalizes for any compounding
effect of deprivations characterized by parameter β . The larger the value of β , the higher the
weight it places on the cumulative deprivations.
O
L G N
4IJ
for 6 V
)\6 7 4] S [ f[ 2J ^ Q _ (IJ g ; ,/06
V 77U
U (4B.3)
. 5J
I=: J=:
Although M(α, β, y) is sensitive to the intensity of deprivation suffered by individuals, it does
not satisfy the dimensional breakdown (unlike the previous measure).
If some of the indicators are binary, as in the case of this chapter, α is set at 0, and
M(α, β, y) coincides with the measure of social exclusion presented by Chakravarty and
D’Ambrosio (2006). The measure is also a member of the M-gamma class of indicators pre-
sented in Alkire and Foster (2016). The measure used in the chapter is defined as follows:
;
L G
& S )\ 4] S [ f[ 2J (J W4IJ T 5J Yg (4B.4)
.
I=: J=:
112 POVERTY AND SHARED PROSPERITY 2018
Annex 4C
Statistical tables
TABLE 4C.1 People Living in Monetary or Multidimensional Poverty, by Rural-Urban Areas, 119 Economies, circa 2013
Monetary Multidimensional
headcount ratio (%) headcount ratio (%)
Rural
share of total Rural share Rural share Economies Population
Region population (%) Rural Urban of the poor Rural Urban of the poor (number) coverage (%)
East Asia and Pacific 55.7 6.5 3.9 67.8 10.2 4.2 75.5 13 28.9
Europe and Central Asia 33.5 0.5 0.2 52.7 1.8 0.8 52.2 17 90.0
Latin America and the Caribbean 21.0 11.2 1.9 61.0 19.9 2.5 68.2 17 91.5
Middle East and North Africa 43.6 6.4 0.9 84.8 11.5 1.9 82.2 9 72.1
South Asia 70.6 15.2 3.9 90.3 33.3 10.5 88.4 5 23.0
Sub-Saharan Africa 67.0 55.9 22.6 83.4 81.8 28.8 85.2 29 60.7
Rest of the world 24.6 0.6 0.4 30.7 0.6 0.4 30.7 29 39.6
Total 45.8 21.0 4.1 81.3 33.6 5.6 83.5 119 45.0
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global Solution Group on Welfare Measure-
ment and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: Location of residence is missing for 1 percent of the total sample.
TABLE 4C.2 People Living in Monetary or Multdimensional Poverty, by Household Type, 119 Economies, circa 2013
Monetary poverty Multidimensional poverty
Population
Profile share (%) Headcount ratio (%) Share of the poor (%) Headcount ratio (%) Share of the poor (%)
One adult female, with child 4.4 20.3 7.63 32.7 7.92
One adult female, without child 2.8 1.8 0.42 5.5 0.84
One adult male, with child 0.8 17.1 1.09 30.3 1.25
One adult male, without child 2.4 1.7 0.35 5.5 0.74
Two adults, with child 37.3 14.9 47.48 23.5 48.29
Two adults, without child 7.7 1.6 1.04 4.4 1.86
Multiple adults, with child 31.1 15.1 39.92 21.4 36.59
Multiple adults, without child 9.7 1.8 1.53 3.0 1.6
Only seniors 3.9 1.4 0.45 3.9 0.83
Only children 0.0 24.9 0.08 38.7 0.08
Total 100.0 11.7 100.0 18.2 100.0
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global Solution Group on Welfare Measure-
ment and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: The monetary and multidimensional poverty rates in this table differ slightly from those reported in tables 4.5, 4.6, and 4.7 because household type cannot be constructed
for 0.4 percent of the sample because of missing information.
BEYOND MONETARY POVERTY 113
Economy-level estimates supply and sanitation). Despite best efforts to
harmonize country-specific questionnaires to
The estimates in tables 4C.3 and 4C.4 are the standard definition, there could be some
derived from household surveys included in discrepancies with measures reported else-
the GMD, for circa 2013. The surveys collect where. Therefore, the estimates must be taken
information on total household income or as the best possible estimate under the strin-
expenditure for monetary poverty estima- gent data requirement of joint observation of
tion, as well as information on a host of other monetary as well as nonmonetary dimensions
topics, including education enrollment, adult of well-being. Finally, both education indica-
education attainment, and access to basic in- tors are household-level indicators (that is, the
frastructure services, which permits the con- number of individuals living in a household in
struction of the multidimensional poverty which one child is not attending school). This
measure. However, there is large heteroge- means that the table of each country’s educa-
neity in how the questions are worded, how tional deprivations presented in the chapter
detailed the response choices are, and how cannot be directly compared to official esti-
closely they match the standard definitions mates from the United Nations Educational,
of access (for example, as defined by the Joint Scientific and Cultural Organization, which
Monitoring Programme for drinking water are based on individual-level indicators.
TABLE 4C.3 Individuals in Households Deprived in Each Indicator, 119 Economies, circa 2013
Deprivation rate (share of population)
Education Education Drinking
Monetary attainment enrollment Electricity Sanitation water
Country Year (%) (%) (%) (%) (%) (%)
Albania 2012 1.06 2.27 6.73 0.50 3.04 0.34
Argentina 2014 0.74 1.31 0.98 0.00 0.72 0.05
Armenia 2013 2.24 0.09 1.89 0.50 10.64 0.06
Austria 2013 0.34 0.00 0.00 0.00 0.96 0.00
Bangladesh 2010 19.63 29.14 16.83 43.63 48.32 3.75
Belarus 2013 0.00 0.79 0.00 — 10.60 0.00
Belgium 2013 0.14 1.41 0.00 0.00 2.10 0.00
Benin 2015 49.55 61.61 25.45 69.02 70.67 26.87
Bhutan 2012 2.17 49.96 6.53 12.90 33.76 1.73
Bolivia 2014 5.80 18.33 4.39 9.60 32.58 7.88
Bosnia and Herzegovina 2015 0.20 1.70 30.77 0.04 1.55 8.79
Brazil 2014 2.76 19.50 0.94 0.33 24.67 4.13
Bulgaria 2013 1.77 0.94 0.00 0.00 20.64 0.00
Burundi 2013 71.73 66.34 18.89 93.07 94.31 18.87
Cameroon 2014 23.83 24.39 15.94 1.20 38.93 23.20
Chad 2011 38.43 67.86 5.87 95.51 92.68 56.05
Chile 2013 0.92 4.64 0.39 0.33 0.47 0.91
Colombia 2014 5.03 6.71 3.03 2.39 9.70 5.04
Congo, Dem. Rep. 2012 77.08 28.73 26.94 85.50 70.08 47.90
Congo, Rep. 2011 36.96 13.39 2.25 40.90 47.29 20.23
Costa Rica 2014 1.45 4.43 1.01 0.57 1.92 0.39
Croatia 2013 0.75 0.26 0.00 0.00 1.93 0.00
Cyprus 2013 0.05 1.81 0.00 0.00 1.03 0.00
Czech Republic 2013 0.05 0.01 0.00 0.00 0.70 0.00
Côte d’Ivoire 2015 28.21 53.15 25.57 37.42 59.47 23.28
Denmark 2013 0.31 1.40 0.00 0.00 0.48 0.00
Djibouti 2012 18.32 32.98 4.05 45.41 37.33 9.25
Dominican Republic 2013 2.37 17.03 1.64 1.37 4.28 27.28
Ecuador 2014 2.63 4.74 1.97 0.96 5.57 4.28
Egypt, Arab Rep. 2012 2.29 14.15 7.16 0.28 11.43 1.08
El Salvador 2014 2.97 28.66 5.54 4.49 2.63 1.30
Estonia 2013 0.86 0.07 0.00 0.00 7.21 0.00
(continued)
114 POVERTY AND SHARED PROSPERITY 2018
TABLE 4C.3 Individuals in Households Deprived in Each Indicator, 119 Economies, circa 2013
(continued)
Deprivation rate (share of population)
Education Education Drinking
Monetary attainment enrollment Electricity Sanitation water
Country Year (%) (%) (%) (%) (%) (%)
Ethiopia 2010 33.56 72.39 37.98 82.53 95.56 50.55
Fiji 2013 1.37 0.79 1.43 10.00 7.15 8.27
Finland 2013 0.09 0.76 0.00 0.00 0.61 0.00
France 2013 0.06 1.04 0.00 0.00 0.47 0.00
Gambia, The 2010 25.08 20.12 10.59 67.10 16.44 11.37
Georgia 2013 6.88 0.19 0.97 0.00 6.04 0.18
Germany 2011 0.04 0.00 0.00 0.00 0.92 0.00
Ghana 2012 12.05 17.08 8.56 33.55 77.09 13.65
Greece 2013 0.96 2.63 0.00 0.00 0.48 0.00
Guatemala 2014 8.65 24.85 18.35 16.52 46.72 8.45
Guinea 2012 35.27 53.73 7.70 0.00 65.66 31.25
Guinea-Bissau 2010 67.08 44.12 5.81 97.09 65.77 36.35
Haiti 2012 23.49 23.18 9.00 64.31 68.80 33.50
Honduras 2013 17.32 14.05 16.56 13.92 19.55 9.85
Hungary 2013 0.11 0.01 0.00 0.00 5.33 0.00
Iceland 2013 0.05 0.06 0.00 0.00 0.00 0.00
Indonesia 2016 6.49 4.99 1.74 2.38 16.51 10.68
Iran, Islamic Rep. 2013 0.11 4.49 1.38 0.12 — 2.44
Iraq 2012 2.46 13.55 22.69 0.66 0.95 10.01
Ireland 2013 0.65 0.76 0.00 0.00 0.12 —
Italy 2013 1.37 1.29 0.00 0.00 0.76 0.00
Jordan 2010 0.12 1.83 2.99 0.00 0.00 0.24
Kazakhstan 2013 0.02 0.00 0.00 0.00 0.01 0.38
Kosovo 2013 0.29 0.73 59.40 0.24 — 2.67
Kyrgyz Republic 2013 3.26 0.21 0.00 5.29 0.50 10.67
Lao PDR 2012 22.75 13.45 14.45 11.13 32.10 44.34
Latvia 2013 1.14 0.77 0.00 0.00 14.68 0.00
Lebanon 2011 0.00 9.24 2.25 0.94 — 0.86
Lesotho 2010 59.65 21.25 10.52 83.63 — —
Liberia 2014 38.61 40.56 2.83 95.67 53.39 19.13
Lithuania 2013 0.71 0.55 0.00 0.00 12.48 —
Luxembourg 2013 0.09 0.64 0.00 0.00 0.08 0.00
Madagascar 2012 77.63 82.46 34.63 27.98 89.48 58.84
Malawi 2010 71.38 47.72 3.12 5.14 26.43 19.41
Malta 2013 0.03 0.52 0.00 0.00 0.09 0.00
Mauritania 2014 5.97 54.26 8.31 62.54 49.30 23.54
Mexico 2012 3.93 6.08 2.76 0.81 4.43 7.41
Micronesia, Fed. Sts. 2013 15.96 8.75 27.99 23.63 19.06 4.97
Moldova 2013 0.08 0.23 0.62 0.09 0.00 28.86
Mongolia 2016 0.50 5.96 3.16 0.17 9.56 12.82
Montenegro 2013 1.04 13.00 37.73 0.99 12.35 4.75
Mozambique 2014 62.90 54.91 33.31 72.76 71.29 40.77
Myanmar 2015 6.36 17.75 13.70 16.20 20.12 29.43
Nepal 2010 14.99 28.56 9.51 31.47 47.32 16.78
Netherlands 2013 0.09 0.58 0.00 0.00 0.01 0.00
Nicaragua 2014 3.24 14.11 8.06 19.98 42.74 12.49
Niger 2014 44.54 70.58 11.71 87.03 83.74 48.54
Norway 2013 0.13 0.78 0.00 0.00 0.00 0.00
Pakistan 2013 6.07 37.09 31.65 8.13 35.11 7.90
Paraguay 2014 2.41 7.76 3.14 0.92 10.95 6.67
Peru 2014 3.72 5.83 1.09 6.80 8.98 14.37
Philippines 2015 6.58 4.52 4.40 9.13 6.78 10.61
Poland 2015 0.00 1.16 2.63 0.00 2.92 0.57
Portugal 2013 0.86 3.57 0.00 0.00 0.95 0.00
(continued)
BEYOND MONETARY POVERTY 115
TABLE 4C.3 Individuals in Households Deprived in Each Indicator, 119 Economies, circa 2013
(continued)
Deprivation rate (share of population)
Education Education Drinking
Monetary attainment enrollment Electricity Sanitation water
Country Year (%) (%) (%) (%) (%) (%)
Romania 2013 0.00 0.29 4.78 1.63 33.43 29.76
Russian Federation 2013 0.01 0.02 1.12 1.01 1.28 0.00
Rwanda 2013 59.49 37.54 4.34 80.55 15.03 25.58
São Tomé and Príncipe 2010 32.28 26.74 17.52 — 60.87 7.03
Senegal 2011 37.98 41.16 6.41 47.05 28.70 18.21
Serbia 2013 0.29 4.28 0.00 0.07 5.03 0.29
Seychelles 2013 1.06 94.93 0.00 — — 8.70
Sierra Leone 2011 52.21 42.49 0.99 0.00 51.03 30.35
Slovak Republic 2013 0.28 0.01 0.00 0.00 1.29 0.00
Slovenia 2013 0.02 0.00 0.00 0.00 0.56 0.00
Solomon Islands 2013 25.14 11.40 13.54 53.83 58.52 25.46
South Africa 2014 18.85 2.26 1.54 4.09 4.17 6.38
Spain 2013 1.16 4.02 0.00 0.00 0.13 0.00
Sri Lanka 2016 0.73 3.78 4.01 2.47 1.15 11.02
Sweden 2013 0.64 0.89 0.00 0.00 0.00 0.00
Switzerland 2013 0.04 0.00 0.00 0.00 0.16 0.00
Tajikistan 2015 4.81 0.31 22.49 2.03 3.51 26.31
Tanzania 2011 49.09 60.61 26.47 84.28 40.79 31.77
Thailand 2013 0.04 15.07 0.67 0.15 0.26 2.68
Timor-Leste 2014 30.31 21.20 0.31 27.20 48.60 22.10
Togo 2015 49.15 26.57 2.32 — 51.82 40.63
Tunisia 2010 1.99 22.55 3.05 0.53 33.32 5.48
Turkey 2013 0.33 3.21 4.22 0.00 2.86 0.68
Tuvalu 2010 3.26 4.54 6.09 9.20 11.54 0.03
Uganda 2012 35.86 47.91 18.48 91.12 72.06 25.97
Ukraine 2013 0.00 0.50 28.94 0.00 27.10 0.00
United Kingdom 2013 0.16 0.48 0.00 0.00 0.40 0.00
Uruguay 2014 0.11 3.04 1.25 0.25 1.52 0.15
Vanuatu 2010 13.15 18.48 14.63 55.93 45.63 19.13
Vietnam 2014 2.64 5.85 1.29 0.89 19.84 7.09
West Bank and Gaza 2011 0.20 3.23 5.49 0.32 1.38 1.95
Yemen, Rep. 2014 18.82 15.95 15.74 33.89 42.53 14.02
Zambia 2015 57.50 24.37 30.37 69.21 59.80 30.67
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global
Solution Group on Welfare Measurement and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: The definition of the indicators and the deprivation thresholds are as follows. Monetary poverty: a household is deprived if income
or expenditure, in 2011 purchasing power parity U.S. dollars, is less than US$1.90 per person per day. Educational attainment: a house-
hold is deprived if no adult (grade 9 equivalent age or above) has completed primary education. Educational enrollment: a household is
deprived if at least one child (grade 8 equivalent age or below) is not enrolled in school. Electricity: a household is deprived if it does not
have access to electricity. Sanitation: a household is deprived if it does not have access to even a limited standard of sanitation. Drinking
water: a household is deprived if it does not have access to even a limited standard of drinking water. The data reported refer to the
share of people living in households deprived according to each indicator. — = not available.
116 POVERTY AND SHARED PROSPERITY 2018
TABLE 4C.4 Multidimensional Poverty across Alternative Measures, 119 Economies, circa 2013
Number of deprivations (share of population)
Multidimensional Adjusted Distribution-
headcount (H) headcount sensitive 0 1 2 3 4 5 6
Country (%) measure (M) measure (D) (%) (%) (%) (%) (%) (%) (%)
Albania 1.21 0.005 0.005 87.7 11.0 1.0 0.1 0.1 0.0 0.0
Argentina 0.77 0.003 0.002 96.4 3.4 0.2 0.0 0.0 0.0 0.0
Armenia 2.24 0.008 0.005 85.4 13.9 0.7 0.0 0.0 0.0 0.0
Austria 0.34 0.001 0.000 98.7 1.3 0.0 0.0 0.0 0.0 0.0
Bangladesh 32.22 0.178 0.121 27.9 23.9 21.3 15.6 8.3 2.7 0.2
Belarus 8.56 0.029 0.011 89.2 10.4 0.4 0.0 0.0 0.0 0.0
Belgium 0.14 0.000 0.001 96.5 3.4 0.1 0.0 0.0 0.0 0.0
Benin 71.74 0.462 0.332 10.4 11.2 13.1 20.7 23.9 16.2 4.5
Bhutan 11.09 0.050 0.046 35.0 33.8 22.4 7.0 1.7 0.1 0.0
Bolivia 11.88 0.060 0.044 57.8 20.6 11.6 6.3 3.0 0.7 0.1
Bosnia and Herzegovina 1.05 0.005 0.013 61.5 34.5 3.6 0.3 0.1 0.0 0.0
Brazil 4.59 0.021 0.021 62.6 26.0 8.6 2.2 0.5 0.0 0.0
Bulgaria 1.77 0.008 0.006 78.7 19.3 1.8 0.1 0.0 0.0 0.0
Burundi 86.54 0.589 0.429 3.0 2.9 8.6 23.6 40.7 18.0 3.3
Cameroon 36.60 0.210 0.142 42.5 21.2 15.3 11.5 7.0 2.5 0.0
Chad 85.47 0.496 0.323 1.5 3.6 9.8 28.4 36.9 18.2 1.5
Chile 1.01 0.004 0.003 93.0 6.5 0.4 0.1 0.0 0.0 0.0
Colombia 6.54 0.028 0.018 78.6 14.4 4.4 1.7 0.7 0.2 0.0
Congo, Dem. Rep. 83.16 0.560 0.404 7.7 7.1 10.6 21.1 28.6 19.4 5.6
Congo, Rep. 42.64 0.228 0.141 27.0 28.5 23.6 15.4 5.0 0.5 0.0
Costa Rica 1.70 0.007 0.005 91.9 7.0 0.8 0.2 0.1 0.0 0.0
Croatia 0.75 0.003 0.001 97.1 2.8 0.1 0.0 0.0 0.0 0.0
Cyprus 0.05 0.000 0.001 97.4 2.3 0.3 0.0 0.0 0.0 0.0
Czech Republic 0.05 0.000 0.000 99.2 0.7 0.0 0.0 0.0 0.0 0.0
Côte d’Ivoire 49.89 0.294 0.206 16.9 19.0 20.7 18.9 14.5 8.1 1.9
Denmark 0.31 0.001 0.001 97.9 1.9 0.1 0.0 0.0 0.0 0.0
Djibouti 27.86 0.162 0.115 35.9 23.5 16.2 11.6 7.4 5.4 0.1
Dominican Republic 5.24 0.023 0.021 62.3 25.2 9.5 2.4 0.6 0.1 0.0
Ecuador 3.34 0.014 0.009 85.0 11.2 2.7 0.8 0.2 0.0 0.0
Egypt, Arab Rep. 4.12 0.017 0.015 70.6 23.5 5.1 0.7 0.1 0.0 0.0
El Salvador 6.50 0.028 0.022 66.2 24.8 6.8 1.9 0.3 0.1 0.0
Estonia 0.86 0.003 0.002 92.0 7.9 0.1 0.0 0.0 0.0 0.0
Ethiopia 82.18 0.523 0.372 2.6 7.3 11.0 18.4 28.1 24.0 8.6
Fiji 2.38 0.009 0.008 78.2 15.9 4.8 1.0 0.1 0.0 0.0
Finland 0.09 0.000 0.000 98.6 1.4 0.0 0.0 0.0 0.0 0.0
France 0.06 0.000 0.000 98.4 1.5 0.0 0.0 0.0 0.0 0.0
Gambia, The 30.83 0.161 0.104 21.5 31.7 26.8 15.3 3.9 0.8 0.0
Georgia 6.89 0.024 0.009 86.6 12.6 0.8 0.0 0.0 0.0 0.0
Germany 0.04 0.000 0.000 99.0 1.0 0.0 0.0 0.0 0.0 0.0
Ghana 26.02 0.137 0.095 19.6 36.9 19.7 13.6 7.2 2.6 0.5
Greece 0.96 0.003 0.002 96.0 3.8 0.1 0.0 0.0 0.0 0.0
Guatemala 21.56 0.110 0.075 39.4 24.3 18.3 11.1 5.2 1.6 0.2
Guinea 46.20 0.257 0.168 19.0 18.3 26.3 24.0 11.4 0.9 0.0
Guinea-Bissau 79.70 0.495 0.334 1.0 11.0 18.4 25.5 29.7 13.1 1.2
Haiti 43.90 0.248 0.169 15.0 22.1 21.3 19.0 14.1 6.9 1.6
Honduras 22.48 0.118 0.075 53.9 22.4 10.8 6.8 4.1 1.6 0.3
Hungary 0.11 0.000 0.001 94.6 5.4 0.0 0.0 0.0 0.0 0.0
Iceland 0.05 0.000 0.000 99.9 0.1 0.0 0.0 0.0 0.0 0.0
Indonesia 8.03 0.034 0.021 70.6 19.7 6.9 2.0 0.6 0.2 0.0
Iran, Islamic Rep. 0.70 0.002 0.003 92.1 7.2 0.6 0.0 0.0 0.0 0.0
Iraq 7.26 0.031 0.024 63.6 25.8 8.0 1.9 0.6 0.1 0.0
Ireland 0.65 0.002 0.001 98.5 1.5 0.0 0.0 0.0 0.0 0.0
Italy 1.37 0.005 0.002 96.6 3.3 0.0 0.0 0.0 0.0 0.0
Jordan 0.33 0.001 0.002 95.1 4.6 0.3 0.0 0.0 0.0 0.0
Kazakhstan 0.02 0.000 0.000 99.6 0.4 0.0 0.0 0.0 0.0 0.0
Kosovo 2.31 0.008 0.020 39.2 58.6 2.0 0.2 0.1 0.0 0.0
(continued)
BEYOND MONETARY POVERTY 117
TABLE 4C.4 Multidimensional Poverty across Alternative Measures, 119 Economies, circa 2013 (continued)
Number of deprivations (share of population)
Multidimensional Adjusted Distribution-
headcount (H) headcount sensitive 0 1 2 3 4 5 6
Country (%) measure (M) measure (D) (%) (%) (%) (%) (%) (%) (%)
Kyrgyz Republic 3.26 0.012 0.007 81.8 16.6 1.6 0.0 0.0 0.0 0.0
Lao PDR 28.77 0.151 0.099 24.0 41.4 17.3 10.4 4.4 1.9 0.6
Latvia 1.14 0.005 0.004 84.3 14.9 0.8 0.0 0.0 0.0 0.0
Lebanon 0.76 0.003 0.004 87.5 11.8 0.8 0.0 0.0 0.0 0.0
Lesotho 90.88 0.529 0.342 8.3 28.7 44.9 16.2 2.0 0.0 0.0
Liberia 53.24 0.312 0.211 2.6 24.2 25.3 22.9 18.8 6.2 0.1
Lithuania 1.06 0.004 0.005 86.8 12.6 0.6 0.0 0.0 0.0 0.0
Luxembourg 0.09 0.000 0.000 99.2 0.8 0.0 0.0 0.0 0.0 0.0
Madagascar 85.35 0.669 0.560 2.5 7.1 13.3 19.7 31.5 23.5 2.4
Malawi 75.07 0.385 0.216 16.4 27.6 32.3 19.2 4.4 0.2 0.0
Malta 0.03 0.000 0.000 99.4 0.6 0.0 0.0 0.0 0.0 0.0
Mauritania 43.25 0.206 0.119 21.0 18.6 18.8 22.1 16.0 3.3 0.1
Mexico 4.59 0.019 0.012 80.6 14.5 3.9 0.7 0.2 0.0 0.0
Micronesia, Fed. Sts. 20.84 0.103 0.067 42.4 30.4 16.0 7.3 3.4 0.5 0.0
Moldova 0.08 0.000 0.004 70.6 28.8 0.5 0.0 0.0 0.0 0.0
Mongolia 1.27 0.005 0.008 75.5 17.6 6.2 0.7 0.0 0.0 0.0
Montenegro 9.48 0.041 0.032 51.8 34.5 8.2 3.4 1.4 0.6 0.0
Mozambique 76.61 0.543 0.419 12.6 8.6 8.9 14.2 22.3 22.4 10.9
Myanmar 15.32 0.078 0.057 44.1 28.2 14.5 8.2 3.3 1.4 0.2
Nepal 28.17 0.150 0.102 32.1 24.8 19.3 13.9 6.9 2.4 0.7
Netherlands 0.09 0.000 0.000 99.3 0.7 0.0 0.0 0.0 0.0 0.0
Nicaragua 15.00 0.071 0.048 45.9 28.9 11.7 7.8 3.5 1.9 0.2
Niger 79.16 0.500 0.348 6.2 6.3 9.1 18.6 34.6 23.5 1.7
Norway 0.13 0.000 0.000 99.1 0.9 0.0 0.0 0.0 0.0 0.0
Pakistan 24.38 0.119 0.079 35.2 28.8 19.1 10.6 4.5 1.6 0.1
Paraguay 3.92 0.018 0.014 77.3 15.9 4.9 1.6 0.3 0.1 0.0
Peru 6.15 0.027 0.018 74.3 15.2 6.8 2.7 0.8 0.1 0.0
Philippines 8.70 0.040 0.025 74.2 15.7 5.8 2.7 1.2 0.3 0.0
Poland 0.08 0.000 0.002 93.5 5.9 0.6 0.0 0.0 0.0 0.0
Portugal 0.86 0.003 0.002 94.8 5.0 0.2 0.0 0.0 0.0 0.0
Romania 3.49 0.013 0.019 62.8 8.3 25.4 3.2 0.3 0.0 0.0
Russian Federation 0.99 0.003 0.002 96.6 3.4 0.0 0.0 0.0 0.0 0.0
Rwanda 63.13 0.362 0.226 12.9 15.7 27.5 27.0 13.6 3.1 0.2
São Tomé and Príncipe 51.11 0.247 0.139 20.2 33.6 30.7 12.5 2.7 0.2 0.0
Senegal 46.69 0.268 0.174 31.1 17.2 16.7 16.5 13.3 4.7 0.4
Serbia 0.38 0.002 0.003 91.0 8.2 0.7 0.1 0.0 0.0 0.0
Seychelles 9.69 0.048 0.048 4.6 86.2 9.2 0.1 0.0 0.0 0.0
Sierra Leone 56.91 0.298 0.173 21.5 22.1 24.6 21.8 10.0 0.1 0.0
Slovak Republic 0.28 0.001 0.000 98.5 1.5 0.0 0.0 0.0 0.0 0.0
Slovenia 0.02 0.000 0.000 99.4 0.6 0.0 0.0 0.0 0.0 0.0
Solomon Islands 37.62 0.193 0.124 14.2 24.4 33.0 18.3 7.6 2.3 0.1
South Africa 19.21 0.073 0.031 70.6 22.9 5.4 1.0 0.0 0.0 0.0
Spain 1.16 0.004 0.003 94.8 5.1 0.1 0.0 0.0 0.0 0.0
Sri Lanka 1.12 0.005 0.006 80.1 17.1 2.3 0.4 0.1 0.0 0.0
Sweden 0.64 0.002 0.001 98.6 1.3 0.1 0.0 0.0 0.0 0.0
Switzerland 0.04 0.000 0.000 99.8 0.2 0.0 0.0 0.0 0.0 0.0
Tajikistan 5.43 0.024 0.024 54.2 34.4 9.7 1.5 0.3 0.1 0.0
Tanzania 67.19 0.434 0.315 8.2 12.0 18.2 23.9 19.8 13.4 4.4
Thailand 0.28 0.001 0.005 82.2 16.9 0.8 0.1 0.0 0.0 0.0
Timor-Leste 39.49 0.192 0.111 28.6 25.3 22.6 15.8 6.5 1.1 0.0
Togo 60.66 0.343 0.216 26.0 19.8 22.4 22.0 9.8 0.2 0.0
Tunisia 4.16 0.020 0.024 52.6 31.0 13.5 2.6 0.3 0.0 0.0
Turkey 0.62 0.002 0.004 90.1 8.6 1.0 0.2 0.0 0.0 0.0
Tuvalu 3.88 0.015 0.012 70.7 24.4 4.3 0.5 0.0 0.0 0.0
Uganda 65.03 0.384 0.261 5.8 9.9 21.3 27.9 22.6 10.8 1.8
Ukraine 0.07 0.000 0.015 51.5 40.6 8.0 0.0 0.0 0.0 0.0
(continued)
118 POVERTY AND SHARED PROSPERITY 2018
TABLE 4C.4 Multidimensional Poverty across Alternative Measures, 119 Economies, circa 2013 (continued)
Number of deprivations (share of population)
Multidimensional Adjusted Distribution-
headcount (H) headcount sensitive 0 1 2 3 4 5 6
Country (%) measure (M) measure (D) (%) (%) (%) (%) (%) (%) (%)
United Kingdom 0.16 0.001 0.000 99.0 1.0 0.0 0.0 0.0 0.0 0.0
Uruguay 0.19 0.001 0.002 94.1 5.6 0.3 0.1 0.0 0.0 0.0
Vanuatu 33.12 0.154 0.095 22.4 25.3 26.5 17.4 6.7 1.6 0.3
Vietnam 3.78 0.019 0.016 74.8 16.3 6.3 1.9 0.6 0.1 0.0
West Bank and Gaza 0.49 0.002 0.004 88.4 10.6 0.9 0.1 0.0 0.0 0.0
Yemen, Rep. 29.98 0.154 0.098 34.0 24.5 18.2 14.6 7.1 1.4 0.1
Zambia 63.69 0.432 0.318 19.9 11.4 10.6 17.1 21.5 14.8 4.7
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global Solution Group on Welfare Measure-
ment and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
FIGURE 4C.1 Share of Individuals in Multidimensional Poverty, by Region, circa 2013
a. Sub-Saharan Africa
Basic infrastructure
Monetary 2.3
2.9
12.4
17
28.2
0.2
1.4 Education
b. South Asia c. East Asia and Pacific d. Latin America
and the Caribbean
Basic infrastructure
Basic
infrastructure Education 0.5
Monetary 1.1
0.4
0.7 3.4 2 1.7
1.5 1.1 0.4 Basic
1 0.5 0.2 infrastructure
2.8 0.3 0.1
Education 1.2 Education
10.9 Monetary
2
Monetary
7
e. Middle East and North Africa f. Europe and Central Asia
Basic infrastructure Basic infrastructure
1.3 0.2
0.6
1.4
1 Monetary 0.1
1.1
0.2 0.04 0.2
0.6 0.5 1.1 0.02
Monetary Education Education
Source: Estimates based on the harmonized household surveys in 119 economies, circa 2013, GMD (Global Monitoring Database), Global
Solution Group on Welfare Measurement and Capacity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
Note: The diagrams show the fraction of the regional population that is multidimensionally poor, and the dimensions the poor are
deprived in.
BEYOND MONETARY POVERTY 119
Notes
1. Economists describe this result formally by explains how the core SDG drinking water
saying that utility-maximizing consumers and sanitation indicators focus on a concept
will, in choosing their consumption bundles, of “safely managed,” but there are relatively
end up equating their marginal rates of sub- few datasets available with all necessary criteria
stitution (a ratio of marginal utilities) to the (and data sources beyond household surveys
relative prices. Hence, so long as markets func- are needed for some aspects of safely man-
tion well, relative prices are natural weights aged sanitation services). SDG monitoring also
with which to aggregate goods and services. uses the less-stringent concepts of “limited”
Where markets work less well, the case for and “basic” access adopted in this report, for
adding separate dimensions of well-being is which data availability is higher, and because
stronger (Ravallion 2011; Ferreira and Lugo of the strong relevance of the concepts glob-
2013). In addition, even when there are prices ally. SDG “limited-standard” drinking water is
for some of these dimensions, such as school drinking water that comes from an improved
fees for private schooling, these might not be source (for example, piped, borehole, pro-
reflective of the “value of a poorly maintained tected dug well, rainwater, delivered water).
village school without a regular teacher. The SDG “basic-standard” drinking water has an
implications of debt owed to a landlord may added criterion of being within a roundtrip
not be captured by ruling interest rates. The time of 30 minutes. SDG “limited-standard”
value of health services depends on the cir- sanitation means using improved sanitation
cumstances of the individual and household” facilities (for example, flush/pour flush to
(World Bank 2017b, 155). piped sewer system, septic tank, a compost-
2. See Sustainable Development Knowledge Plat- ing latrine). SDG “basic-standard” sanitation
form (database), Department of Economic has an added criterion of being for the exclu-
and Social Affairs, United Nations, New York, sive use of the household (these concepts are
https://sustainabledevelopment.un.org/. reflected in table 4.1). Thus “safely managed”
3. So long as markets work reasonably well, is a subset of “basic-standard,” which is a sub-
prices—the weights for the quantity of goods set of “limited-standard,” with each additional
and services consumed—bear a very close criterion meaning fewer datasets currently
relationship to the marginal contribution of available for analysis. Graphs at the World
those goods to well-being. In technical terms, Bank Atlas http://datatopics.worldbank.org
the ratio of two prices equals their marginal /sdgatlas/SDG-06-clean-water-and-sanitation
rate of substitution between the two goods. .html make this clearer, and at the website of
When externalities or other imperfections the WHO/UNICEF JMP, custodian agency for
distort the market price, then a shadow price monitoring these indicators globally: https://
can be used in principle to value a good. How- washdata.org/monitoring.
ever, the information required to estimate an 6. The quality of the environment in which the
accurate shadow price is high, and frequently individual resides matters greatly for well-
shadow prices cannot be estimated with much being. Although environmental degradation
accuracy. Typically, when there is no adequate can be partially offset through market pur-
comparator, or the distortion is too great, one chases, such as flood insurance, these sorts
moves to add the good or service in question of goods and services are not widely available
as a separate dimension. and, in any case, often only partly alleviate the
4. Hentschel and Lanjouw (2000) distinguish physical, mental, and health costs imposed
three reasons for the price of public utilities when environmental disasters strike.
to vary across consumers: rationed markets, 7. Not all indicators are applicable to every
public subsidies, and increasing marginal household. For example, not every household
tariff rates. The authors present a method to has a child below the school age for grade 8
impute the value of consumption of basic (necessary for the school enrollment indica-
utilities irrespective of the source of water, tor). In these cases, the weight for the missing
cooking fuel, or electricity to be incorporated indicator is shifted to other indicators within
into the consumption aggregate. At present, the dimension so that each dimensional
data are not available at a large scale across weight is unchanged. The same process occurs
countries and thus cannot be implemented. if the information on an indicator for a house-
5. “At least limited” and “at least basic” drinking hold is missing, even if the indicator is appli-
water and sanitation reflect the key concepts cable. Because of this reweighting process, few
of SDG monitoring in this exercise. Box 4.2 households are ignored because of missing
120 POVERTY AND SHARED PROSPERITY 2018
data. Indeed, only households on which infor- survey from 2014, is lower than the lined-up
mation is missing on all the indicators that estimate in 2015 of 41 percent.
constitute a dimension are not considered in 9. These figures may not be representative of the
the analysis. entire region because of incomplete popula-
8. The share of monetary poor differs from tion coverage. The coverage in East Asia and
the numbers presented in chapter 1. This is Pacific and South Asia is particularly low (see
primarily because the estimates presented last column of table 4.4.): China and India
in chapter 4 are from around 2013, and not are not a part of this exercise because of data
“lined up” to 2015, as is the case in chapter 1 availability.
(see appendix A for details on why the num- 10. See the demographic composition typology
bers presented in this chapter differ slightly, proposed in Muñoz Boudet et al. (2018) for
and for how survey estimates are lined up to the methodology followed.
a reference year). The 2015 lined-up head- 11. Some studies show that households with chil-
count ratio for the 119 economies covered dren will likely appear more disadvantaged
here is 11.2 rather than 11.8 percent. This under a child-specific multidimensional pov-
difference mostly reflects that countries had erty measure (Hjelm et al. 2016).
positive growth rates from 2013 to 2015, and 12. The Indonesian survey is not nationally rep-
hence reduced poverty during that time. For resentative. The sample is representative of
example, the headcount ratio presented here about 83 percent of the population and cov-
for the Lao People’s Democratic Republic, ers 13 of the 27 provinces in the country. The
which is based on a survey from 2012, is near provinces excluded tend to be less developed.
23 percent, whereas the 2015 lined-up esti- 13. The surveys are not necessarily the same sur-
mate for Lao PDR is 14 percent. Conversely, veys used for official national poverty esti-
the recent crisis unfolding in the Republic of mates. The monetary poverty rate cited here
Yemen means that the headcount ratio of 19 may therefore vary somewhat from official
percent presented in this chapter based on a estimates.
BEYOND MONETARY POVERTY 121
SPOTLIGHT 4.1
National Multidimensional
Poverty Indexes
Prepared by Sabina Alkire, Oxford Poverty and Human Development Initiative
National Multidimensional Poverty Indexes (MPIs) are increasingly being adopted as official
permanent poverty statistics, usually standing alongside and complementing national mone-
tary poverty statistics (Alkire and Foster 2011; Alkire et al. 2015; UNECE 2017; OPHI 2018).
Updated usually every one to two years, national MPIs are used to shape and energize effective
policy actions. They are reported against SDG Indicator 1.2.2.
The Atkinson Commission report Monitoring Global Poverty placed great emphasis on na-
tional poverty estimates, both monetary and multidimensional, given their central relevance
to national policy and public debate. In the case of poverty measurement, the report advocated
jointly reporting the global and national poverty measures in national poverty reports. The
report also scrutinized the dimensions and indicators covered by national MPIs, and observed
that their data requirements are modest: most require 38–70 questions compared to 450 or
more survey questions for monetary poverty measures (World Bank 2017b, 172).
But how do national MPIs differ from the global MPI that the United Nations Develop-
ment Programme and Oxford Poverty and Human Development Initiative have published
since 2010 (UNDP 2010; Alkire and Santos 2014) or from the ones presented in this chapter?
The difference is similar to the difference between the US$1.90 per day measure of extreme
poverty globally and national monetary poverty measures. That is, whereas the global poverty
measures are computed in a standardized format for every country, national poverty mea-
sures are tailored to the contours of poverty and the policy priorities of each context. Also, na-
tional measures are computed by national statistical offices, using national survey data. Thus,
national MPIs may have different dimensions and indicators; their deprivation cutoffs may
reflect the aspirations, context, or national plan of the country; and the weights and poverty
cutoff are set so as to identify poverty according to national definitions. Nearly all national
MPIs cover health, education, and living standards. The Atkinson report recommended six
nonmonetary dimensions for the global MPI including employment, and many national MPIs
already include a dimension on work.
National MPIs cannot be compared across countries precisely because the definitions differ.
However, the great advantage of national MPIs is that they can be—and indeed are being—
used to guide policy in powerfully practical ways. In particular, the following are the main uses
to date of national MPIs by the increasing community of countries that use these indexes.
• Complement monetary poverty. The national MPI makes visible a set of nonmonetary
deprivations. The value added is seen by exploring mismatches. For example, Chile’s MPI
made visible situations of high poverty in Atacama, a low-monetary poverty region in the
country. Bhutan (2017) found that the district of Gasa, which had lowest monetary poverty,
had the highest MPI because of missing services and infrastructure.
122 POVERTY AND SHARED PROSPERITY 2018
• Ease communication. The headcount ratio of monetary poverty, which is widely used, is
always compared to the headcount ratio of multidimensional poverty. And the national MPI
often accords with participatory work showing how people are poor. El Salvador’s MPI was
informed by a study on poverty “as Viewed by Its Protagonists” linking poor people’s voices
to each indicator of the MPI (UNDP 2014).
• Monitor trends. In every country, the national MPI tracks the trend of multidimensional
poverty over time, nationally and by rural-urban regions, subnational regions, and social
groups, providing a rigorous overview of progress.
• Allocate resources. The national MPI is regularly used to shape both sectoral and regional
budget allocation across regions and across sectors. For example, Bhutan’s district allocation
formula uses the MPI.
• Leave no one behind. The national MPI is disaggregated by population subgroups to see
who is the poorest. Changes over time are reported across subgroups, to establish whether
the poorest regions are catching up—or whether their progress is slower than less poor re-
gions, so that the poorest regions are gradually being left behind. For example, Pakistan’s
poorest district, Musakhel, reduced poverty the fastest over the period 2005–15.
• Target households. The national MPI structure is used to identify which poor people are to
be recipients of certain benefits. This is usually done using a different data source: a census,
partial census, or eligibility applications. For example, Costa Rica targets households ac-
cording to their deprivation scores on the national MPI.
• Coordinate policies. The national MPI is used as a management tool to coordinate poli-
cies across sectors and across levels of government, and to design and monitor integrated,
multisectoral policies that bridge silos. Although practices vary, this measure-to-manage ap-
proach is mainly used when data are updated every one to two years. For example, Colombia
has a Ministerial Round Table chaired by the President, which meets regularly to accelerate
progress in reducing its MPI—which is updated annually.
• Be transparent. Many countries post the computational files required to replicate their of-
ficial national MPI online. For example, Mexico’s CONEVAL both launches its MPI and
posts online tables two weeks after cleaned data are received. In many cases, methodological
notes, data tables, microdata, and presentations are also online, so citizens can easily learn
and interact.
Thus, national MPIs provide a headline and high-resolution information panel on subna-
tional conditions across population groups and across the joint distribution of deprivations in
different dimensions of poverty. Although most cannot be compared cross-nationally,a they
do complement official national monetary poverty statistics by providing policy-relevant evi-
dence on poverty in other forms and dimensions.
Details of national MPIs and of their policy applications are available on the website of
the Multidimensional Poverty Peer Network, a South–South Network that convenes countries
using or designing or exploring national MPIs (see www.mppn.org).
a. Nepal adopted the global MPI, with slight adaptation, as its national MPI in 2018, partly in order to
avail such comparisons.
BEYOND MONETARY POVERTY 123
Inside the Household: 5
Poor Children,
Women, and Men
The aim of this chapter is to enter the household to try and answer an apparently simple
question: how many children, women, and men are poor? The common approach assigns all
individuals within a household to the same poverty status as the household. However, this
masks potential differences in poverty among household members. Ignoring these decreases
the effectiveness of common approaches to targeting poverty reduction interventions and the
take-up of these interventions because they do not address the needs and constraints of the
poorest individuals.
The chapter begins with an analysis of global poverty data, including comparisons between
male- and female-headed households, and introducing alternative household classifications
related to the number of adults and income earners. Despite maintaining the concept of
poverty based on the household, the analysis provides insights into sex and age differences
among the poor. Next, evidence is presented on intrahousehold differences in resource allo-
cation, first, by relying on a few datasets that provide information on consumption among
individuals and, second, by applying models of intrahousehold resource allocation. A broader
exploration of adult poverty follows according to the multidimensional approach introduced in
chapter 4 but including individual-level information on educational attainment and body mass
index.
The accumulated evidence of numerous studies and data sources suggests that women
and children are often disproportionately affected by poverty albeit with considerable varia-
tion across countries and across types of households. Sex differences in poverty are largest
during the reproductive years, when care and domestic responsibilities, which are socially
assigned to women, overlap and conflict with productive activities. This tension is often most
pronounced among the poorest countries and the poorest groups in society.
Introduction
How many women are poor? How many households in which they live. This masks
poor children are there? These seem straight- differences in poverty among the individuals
forward questions, but there are no straight- within the same household.
forward answers. Most poverty measures, In the absence of poverty data on individ-
including most of those presented earlier in uals, perceptions about differences in pov-
this report, refer to households. Individuals erty by sex and age are rarely supported by
are typically classified as poor or nonpoor evidence. Consider, for example, the widely
in accordance with the poverty status of the cited claim that 70 percent of the world’s
125
poor are women. There is a solid consensus in retrofitting household-level data to the
that the empirical data do not substantiate individual. Advancing our understanding of
this claim and that the statistic is false (Chant the poverty of individuals requires a renewed
2008; Green 2010; Greenberg 2014; Quisum- emphasis on data collection and investments
bing, Haddad, and Peña 2001; Sánchez- in survey data collection methodologies fo-
Páramo and Muñoz Boudet 2018). A com- cused on the individual.
mon lens on the gender dimension of pov- More reliable poverty estimates on indi-
erty is the difference between female- and viduals would facilitate a better understand-
male-headed households. The concept of ing of the characteristics of poverty and its
household head is, however, often ill-defined intergenerational transmission, the interven-
and may even be misleading, for example, if tions appropriate for different types of indi-
vulnerable widows and more affluent single viduals, and the more effective targeting of
women are lumped under a single category social protection and broader development
of female-headed households and then used programs. Such programs often rely on ap-
as a proxy for women in general (Bradshaw, proaches targeted to households but may fail
Chant, and Linneker 2017; Grown 2010, to reach potentially poor beneficiaries if many
2014; Milazzo and van de Walle 2017). of these live in households not identified as
Drawing on new work conducted for this poor (Brown, Ravallion, and van de Walle,
report, and a review of the existing literature, forthcoming).
this chapter revisits what we know about the Measuring the monetary poverty of in-
poverty of individuals, with a focus on differ- dividuals requires two main pieces of infor-
ences by sex and between children and adults. mation. The first is information on how total
Child poverty, though related to the poverty household resources are allocated among
of women, is a distinct issue. This chapter household members. This is an intuitive idea,
considers both because they are the two di- but one vexed with theoretical and practical
mensions prioritized for the disaggregation of challenges. Data on the food consumption of
the global poverty figure (World Bank 2017b, individuals are difficult to collect whenever
114). The accumulated evidence from many household members consume meals together.
studies and data sources suggests that women Other consumption items, such as housing or
and children are often disproportionately consumer durables, are shared among house-
affected by poverty, but with considerable hold members and often cannot be allocated
variation across countries. Sex differences in to specific individuals even in principle. Be-
poverty are largest during the reproductive cause of these and other challenges, living
years when, because of social norms, women standards surveys, the main data source for
face strong trade-offs between reproductive measurements of monetary poverty, typi-
care and domestic responsibilities on the cally collect most data on the consumption of
one hand and productive activities on the households as a single entity. Poverty analysis
other hand. The tension is often most pro- thus remains fixed on the household. The sec-
nounced in the poorest countries and among ond key ingredient is information on the ways
the poorest groups in society. In addition, basic needs differ across household members,
women’s intrahousehold bargaining power for example, by sex and age, and across house-
and poverty appear to be related to their po- holds of different sizes and compositions to
sition within the household, for example, as assess whether differences in resources trans-
the first or more junior wife of the principal late into differences in well-being and poverty.
male, his mother, and so on. This underscores Even though not the primary focus of this
that gender, age, and status within the house- chapter, the need to measure the poverty of
hold are interrelated dimensions, which can individuals highlights the need to revisit the
be difficult to disentangle. broader issue of equivalence scales (box 5.1).
A secondary objective of the chapter is to This chapter highlights various methods
test the boundaries on methods for identify- that can be used to measure poverty among
ing the poor, whether they live in poor house- individuals and explore the effects of gen-
holds or not, and to highlight the challenges der and age differences on poverty data. The
126 POVERTY AND SHARED PROSPERITY 2018
BOX 5.1 Differences in Needs and Equivalence Scales
Global poverty estimates use data national poverty assessments in associated with children (Folbre,
on consumption or income per both developing and high-income Murray-Close, and Suh 2018). In
capita to measure poverty. Similarly, countries, including member addition, adjusting consumption
the international poverty line, which countries of the Organisation or income by an equivalence scale
is anchored on the average cost of for Economic Co-operation and requires recalibrating the poverty
meeting basic needs in the poorest Development (OECD), routinely use line (Ravallion 2015). Central to
societies, is expressed in per capita equivalence scales. The failure to this recalibration is the choice
terms. This per capita approach account for equivalence scales will of a household with “reference
assumes that needs do not vary overestimate poverty in regions demographics,” which may also vary
across the members of households where households are large and from country to country. The use of
and that there are no economies contain lots of children, such as Sub- a per capita scale in global poverty
of scale in larger households. Both Saharan Africa, compared to regions monitoring therefore imposes
assumptions are subject to criticism. where households are small and comparability across countries and is
Caloric needs vary by sex, age, contain few children, such as Europe also transparent and easy to explain
physical activity (often related to and Central Asia and to some extent no matter how problematic it may be
occupation), and so on and are thus East Asia and Pacific and Latin in the details (Ferreira et al. 2016).
not the same across all household America and the Caribbean. The question of how to adjust
members. For example, a person The main problem with adopting for differences in needs arises
engaged in heavy agricultural work an equivalence scale approach in even more prominently once the
typically requires more calories global poverty monitoring is that focus of the analysis moves inside
than an office worker. Likewise, there is no consensus on what the household. A comparison of
shared household public goods may the best scale for this purpose inequality in consumption between
represent an advantage for larger would be across a wide range of adults and children or between men
households even at the same level countries. For example, nutrition- and women remains incomplete
of per capita consumption. One based equivalence scales, which if we do not also consider
way to adjust for such differences account for differences in needs differences in needs between
in household size and composition by sex and age and are used in these groups. (See also the section
is to use equivalence scales, the many low-income countries, may on “Differences in resources and
discussion of which goes back be less appropriate in higher- poverty within households” in which
to the seminal work by Engel income countries where food all the country studies have adopted
(1895) and Rothbarth (1943) (see constitutes a smaller relative share some variant of an equivalence
Coulter, Cowell, and Jenkins 1992; in total consumption. Similarly, scale.) Measuring the poverty of
Deaton 1997). Equivalence scales the economies of scale in shared individuals would require not only
approximate the consumption needs goods may be offset by the estimating intrahousehold resource
of a household of a given size and greater need for health care and allocation but also adjusting for
demographic composition relative education expenditures (Abdu the differences in needs among
to a reference household (usually and Delamonica 2017) and the individuals living in the same
a household consisting of a single failure to value nonmarket (time household and between households
adult, or a single adult male). Many and resources) expenditures of different sizes.
starting point, in the next section, titled “Be- household. This assumption is inadequate
yond headship: Gender and age profiles of for a clear understanding of the differences
the global poor,” is the monetary poverty within households and biases country pov-
estimates introduced in chapter 1, which erty rates and the demographic profile of
represent the current state of play in global poverty if there are systematic differences by
poverty monitoring. In comparing per capita sex and age in the household. Despite these
household consumption against the interna- limitations, even the current data can provide
tional poverty line, which is also expressed meaningful though incomplete insights into
in per capita terms, this approach assumes sex and age differences in poverty if the anal-
that resources are shared equally and that ysis probes more deeply than comparisons
needs are the same across all members of a of female- and male-headed households to
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 127
explore differences by household composi- Beyond headship: Gender and
tion and over the life cycle. age profiles of the global poor
The subsequent section of the chapter,
titled “Differences in resources and poverty This section analyzes data from the Global
within households” presents evidence on in- Monitoring Database (GMD), which is a col-
trahousehold differences in resource alloca- lection of globally harmonized household
tion, thus relaxing the assumption of equal survey data the World Bank uses to monitor
sharing among household members. A few global poverty and shared prosperity (box
specialized datasets provide information, 5.2).1 The global poverty figures rely on a
for at least some aspects of consumption, concept of poverty based on the household
on how much is allocated to whom within (though expressed in per capita terms) and
the household. Invoking assumptions about classify individuals as poor or nonpoor ac-
household behavior and equivalence scales, cording to the poverty status of the house-
a growing academic literature provides es- holds in which they live. Although this ap-
timates on resource allocation across indi- proach cannot reveal differences in poverty
vidual household members on the basis of within households, innovative ways to analyze
(largely) household-level data. the data can reveal meaningful, though in-
In the penultimate section, the chapter complete, information on sex and age differ-
describes a broader examination of well- ences, which are explored in this section.
being and poverty among adult household This section shows that, although the pro-
members based on the multidimensional ap- portion of women and men living in poor
proach introduced in chapter 4. Straightfor- households is similar on aggregate, the pro-
ward documentation on gender differences portions vary by women’s and men’s marital
in nonmonetary dimensions of well-being status, the presence of children and depen-
may be derived from data collected on indi- dents in their households, whether or when
viduals, rather than households. An example they join the labor market, and their respon-
is education, for which indicators of educa- sibilities within the family. Children and other
tional attainment have been used for many dependents are an important factor of vul-
years to compare achievements and depri- nerability, particularly among women during
vations between women and men. Likewise, their reproductive years. Care responsibili-
anthropometric data, such as weight, height, ties, especially borne by women, are greatest
and the body mass index (BMI), which are during those years in the life cycle that tend
commonly used to measure malnutrition, also to be the best for income generation. Re-
refer to individuals, not households. These lying on the economic activity of more adults,
data are used to provide perspective on multi- both women and men, helps shield the house-
dimensional poverty among individuals. hold against poverty, though doing so requires
BOX 5.2 Chapter 5: Data Overview
This section relies on information from quality concerns in the economic
the harmonized sample of 104 household participation variables, 18 countries were
surveys for 89 countries in the 2013 Global dropped for the analysis of employment
Monitoring Database (GMD).a Additional and economic typology of households.
labor data from the International Income Because of low coverage in the Middle
Distribution dataset were merged for 17 East and North Africa (4.1 percent), the
countries in Sub-Saharan Africa (Muñoz results from this region are not presented.
Boudet et al. 2018). Because of remaining
a. GMD (Global Monitoring Database), Global Solution Group on Welfare Measurement and Capac-
ity Building, Poverty and Equity Global Practice, World Bank, Washington, DC.
128 POVERTY AND SHARED PROSPERITY 2018
quality and affordable care services for chil- groups, and might privilege one sex over the
dren, the sick, and the elderly. Formal school- other. Globally, self-reported female-headed
ing is also a strong protective factor against households account for 23 percent of all
poverty, especially for women. Interventions households, but only 16 percent of poor
aimed at reducing poverty need to consider households. Although this shows that the
both household structure and individual char- poverty rate is lower among these households
acteristics to increase their chances of success. than among male-headed households, we can
The rates of women and men living in learn little else (table 5.1).
poor households are similar in the 89-country
dataset used here (12.8 percent and 12.3 per- Poverty by age
cent, respectively2). These poverty rates vary
across regions, but gender differences are only Nearly one child in five3 lives in a poor house-
statistically significant in South Asia. World- hold. Children are twice as likely as adults
wide, this translates to 104 women in poor to live in poor households. This primarily
households for every 100 men. In South Asia, reflects the fact that the poor tend to live in
the corresponding comparison is 109 women large households with more children. Chil-
for every 100 men. These differences become dren are the poorest across all regions, but
starker at specific ages. the patterns vary by region. For example, in
Sub-Saharan Africa, 49.3 percent of girls and
49.5 percent of boys live in poor households
Beyond headship
and boys represent a slightly larger share
Many global and country-level analyses of (51 percent) of poor children than girls do.
poverty compare female- and male-headed Differences with other age groups are even
households to highlight sex differences in starker: boys and girls under 15 years of age
poverty. However, the concept of the female- are 10 percentage points more likely to live
head is often difficult to interpret. First, it in a poor household than their young adult
typically combines women who have never (ages 15–24) counterparts, and girls are 17.2
married, women who are widowed or di- percentage points more likely than females
vorced, and some women who are married. A above 60. In contrast, in South Asia, girls
related concern is that the headship concept are poorer than boys (22.2 and 20.1 percent,
risks conflating gender gaps with differences respectively) and slightly more numerous
caused by demographic composition. For than boys among the poor (50.5 percent),
example, many female-headed households but differences in poverty rates between chil-
contain children but not adult males, whereas dren and older adults—although sizable—
most male-headed households contain adult are smaller than in other regions.
women and children. Second, self-reported The rates of women and men who are
household headship reflects social norms and living in poor households decline sharply as
views about who is understood as the head of children reach adulthood, and they tend to
the household, for example, the main bread- stabilize after women and men reach 50 years
winner, the main decision maker, the oldest of age. Starting in their early 20s and up to age
man, and so on. These norms may vary across 34, women are 2 percentage points more likely
countries, within countries, or across income than men to live in poor households, which
TABLE 5.1 Households in Extreme Poverty, Rates and Distribution by Headship, circa 2013
Percent
Share of poor
Poverty rate households Share of total households
Female-headed households 5.8 16.4 23.5
Male-headed households 9.0 83.6 76.5
All households 8.2 100.0 100.0
Source: Muñoz Boudet et al. 2018.
Note: Data are from 89 countries.
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 129
FIGURE 5.1 Percent of Females and Males Living in Households in Extreme Poverty, by Age
Group, circa 2013
25
20
Poverty rate (%)
15
10
5
0
0–4 5–9 10–14 15–19 20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74
Age groups
Males Females
Source: Muñoz Boudet et al. 2018.
Note: Data are from 89 countries.
is a significant, sizable difference (figure 5.1). capacity (proxied by employment status) of
In this age group, an average of 120 women individuals. This allows for a closer look at
are living in poor households for every 100 how these characteristics build on the age
men. This gender gap coincides with the peak and sex differences.
productive and reproductive ages of men and Formal schooling is inversely correlated
women, and can be related to factors such as with poverty among both women and men.
household formation4 and income genera- Of the poor population ages 15 or above, 41
tion for both men and women, and the im- percent have no education. Women represent
plications of such processes on their welfare. 62.3 percent of the poor population ages 15
It is well documented that female labor force or above with no schooling, but only 36.9
participation declines during women’s repro- percent of the poor with tertiary schooling.
ductive years, particularly if they have young The share of women living in poor house-
children (Aguero and Marks 2008; Cruces and holds diminishes strongly with schooling.
Galiani 2007; Goldin and Katz 2002). Among The association between employment and
the 20–34 age group, the gender gap in pov- poverty varies by sex and type of employ-
erty rates ranges from 0.12 percentage points ment. In the prime productive years, between
in Europe and Central Asia to 7.1 percent- 25 and 54 years of age, women represent 86
age points in Sub-Saharan Africa. In this age percent of those out of the labor force and
group, the gaps are wider in the poorest coun- 60 percent of those who are unpaid work-
tries, especially the 17 countries with overall ers. In poor households, while most men
poverty rates above 35 percent, that is, Haiti are paid workers or self-employed, over
and 16 Sub-Saharan African countries. half of women are not in the labor force.
Globally, 40 percent of poor men are self-
employed, compared with only 19 percent of
Schooling, the labor market,
women (figure 5.2). In Sub-Saharan Africa
and gender differences
and South Asia, self-employment is closely
Household surveys collect information on associated with poverty for men, but slightly
educational attainment and income-earning less so for women.
130 POVERTY AND SHARED PROSPERITY 2018
Household structure and FIGURE 5.2 Distribution of People Living in Households in Extreme
gender differences Poverty, by Sex and Employment Status, circa 2013
The analysis demonstrates that household 100
10.3
composition, particularly the presence of de-
80 12.4
pendents and the type of earners, influences 50.1
gender differences in poverty over the life 60
Percent
cycle. Building on the framework introduced 40.2
in Grown and Valodia (2010), this subsection 40 17.8
illustrates two ways to classify households: a
demographic typology and an economic one. 20
34.9
19.0
The demographic typology is based on the
11.7
adult composition of the household, start- 0
ing with the age and sex of the adults (18–64 Men Women
years) who live in the household and distin- Paid worker Self-employed Unpaid worker
guishing separate categories for the elderly or Unemployed Out of labor force
seniors (ages 65 years or above) and children
Source: Muñoz Boudet et al. 2018.
(under age 18). The economic typology is Note: Data are from 71 countries. Ages are 25–54.
based on the presence and sex of all earners
in the household and of the dependents who
depend on the income of the earners. Earn- married or cohabiting—with children account
ers are defined as any individuals ages 15 or for the largest share of poor households (figure
above who are engaged in any economic ac- 5.3). They are overrepresented among the poor,
tivity for pay or profit.5 Dependents here in- representing 31 percent of all households but
clude nonearners ages 18–64 (unpaid family accounting for 42 percent of poor households.
workers, and those that are unemployed or Adult-couple households with children and
not in the labor force) and traditional depen- other adults, that is, extended family house-
dents (children and seniors). holds, which represent 17 percent of all house-
Within the lens of the household demo- holds, account for the second-largest share
graphic typology, adult-couple households— among poor households (28 percent), and
consisting of two adults of opposite sex who are they are also overrepresented among the poor.
FIGURE 5.3 Distribution of Households in Extreme Poverty, by Demographic Typology, circa 2013
Other adults One adult, female
combinations with with children,
children 7.2% (4.6%) 6.1% (3.2%)
Other adults One adult only, Senior(s)
combinations only, no
no children
(other than children,
2.8% (11.2%)
a couple) 2.1% (6.2%)
without
children, Multiple adults, Adult
3.9% only female couple, no
Adult couple with children and (13.6%) with children children
Adult couple with children, 41.5% (30.6%) other adults, 28.2% (17.1%) 2.4%, (1.1%) 1.9%, (8.2%)
Source: Muñoz Boudet et al. 2018.
Note: The percentages in the cells refer to the share of the type among poor households; the numbers in parentheses refer to the share of the typology among all households.
The figure shows typologies that represent at least 2 percent of either poor or all households. Data are from 89 countries.
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 131
FIGURE 5.4 Distribution of Households in Extreme Poverty, by Economic Typology, circa 2013
Female earner
with children
Head couple earner with children and nonearner
only, 8.2% (10.2%) 4.9% (2.4%)
Female
Children and nonearner, 14.2% (4.9%)
Multiple earners earner with
Head couple earner with nonearner, children only,
with children and 2.7% (4.9%) 2.4% (1.3%)
nonearner,
4.7% (4.9%) Male
earner
Senior nonearner,
with
2.3% (6.7%)
non-
Multiple earners earner
Multiple earners with children and with children only, Adults, all nonearners, 1.9%
Male earner with children and nonearner, 36.2% (21.0%) nonearner, 10.9% (7.1%) 3.2% (2.6%) 2.1% (4.4%) (6.0%)
Source: Muñoz Boudet et al. 2018.
Note: The percentages in the cells refer to the share of the type among poor households; the numbers in parentheses refer to the share of the typology among all households.
The figure shows typologies that represent at least 2 percent of either poor or all households. Data are from 71 countries.
Meanwhile, adult-couple households without in households with children and nonearners
children are less likely to be poor (8 percent of (42 percent in households where there is only
all households; 2 percent of the poor). Other a male earner and 15 percent in households
types of households gain relevance depend- with multiple earners).
ing on the region. Adult woman households
with children are disproportionately rep-
Differences in resources and
resented among the poor in Latin America
and the Caribbean and in Sub-Saharan Af- poverty within households
rica. Three poor women in four live in adult- The previous section summarizes what the
couple households with children only or with data used to monitor global poverty reveal
other adults, and this proportion increases to about gender and age differences in the pro-
four poor women in five for the 20–34 years file of poverty, while maintaining the (gen-
age group. erally implicit) assumption that resources
The analysis of poverty using the eco- under the per capita measure are shared
nomic typology confirms that households equally. A more comprehensive measurement
with traditional dependents (children below of gender and age differences in the profile of
15 or seniors) fare less well than households poverty requires a relaxation in the assump-
without dependents (figure 5.4). In most tion of equal sharing to consider intrahouse-
cases, the presence of a nonearner, age 18–64, hold differences in resource allocation.
also raises the poverty rate. Households with Measuring intrahousehold inequalities in
no earners, combined with the presence of resource allocation and poverty in household
children, are the household type most at risk surveys is not an easy task. Accurate data on
of poverty (14 percent of the poor while they food consumption across individuals in a
account for less than 5 percent of house- household are difficult to collect whenever
holds), followed by households with a single household members cook together and share
woman earner and dependents (5 percent of meals. Some household surveys collect such
the poor and 2 percent of the population) data using a 24-hour recall or direct obser-
and households with a male earner only, a vation (weighting, measuring by resident
nonearner and children (36 percent of the enumerators), but these methods are time-
poor while they account for 21 percent of the consuming and intrusive. Other consump-
population). Poor women are concentrated tion items, such as housing, are de facto
132 POVERTY AND SHARED PROSPERITY 2018
public goods within the household that are suggest that total inequality in calorie ade-
shared among household members and can- quacy among individuals is significantly un-
not be allocated to specific individuals even derestimated, by 30 to 40 percent, if inequal-
in principle (Case and Deaton 2002; Klasen ity within households is ignored.
2007). The following section reports findings More recent data collection efforts in Af-
from four recent country surveys that collect rica and Asia have allowed a fresh look at
data on consumption among individuals. intrahousehold differences in poverty across
These case studies are then complemented various contexts and social settings (De Vreyer
by model-based estimates of poverty in two and Lambert 2017 on Senegal; D’Souza and
countries. Modeling allows the resource Tandon 2018 on Bangladesh; Mercier and
shares of men, women, and children to be es- Verwimp 2017 on Burundi; Santaeulàlia-
timated over the entire consumption basket Llopis and Zheng 2017 on China).6 Though
even though individual consumption data these studies individualize only a few separate
are only available on a few items, thus provid- components of consumption (table 5.2), they
ing a more complete picture of intrahouse- reveal interesting differences in resource allo-
hold resource sharing. cation among women, men, and children.
The evidence in this section shows that in-
trahousehold differences in consumption and
Individual-level data on
poverty are widespread. In most cases, women
consumption
and children are allocated a smaller share of
Starting in the 1980s, a few specialized studies the households’ resources than men.7 Intra-
have collected data on consumption among household inequalities in resource allocation
individuals, often with a focus on food appear to be more pronounced for nonfood
(Behrman and Deolalikar 1990; Haddad, items than for core food items, hinting at a
Hoddinott, and Alderman 1997; Haddad and degree of solidarity within families. Similar
Kanbur 1990; Pitt, Rosenzweig, and Hassan to the previous section, we find evidence of
1990). An early example of this literature is complex dynamics within households, linked
the work of Haddad and Kanbur (1990) who to life cycle and status that extend beyond
investigate intrahousehold inequality in food simple gender or age divides. For example, in-
consumption in the Philippines through the trahousehold bargaining power and poverty
lens of calorie adequacy, that is, calorie intake among women are related to their relation-
relative to standardized calorie requirements ship with the principal male, such as first ver-
by age, sex, and pregnancy status. These data sus second wife, or mother versus wife.
TABLE 5.2 Recent Datasets on Individualized Consumption
Country Survey Year(s) Representativeness Items individualized and data collection method
Bangladesh Bangladesh 2011–12, 2015 National (rural) Food (24-hour recall by the woman in charge of cooking)
Integrated
Household
Survey 1, 2
Burundi Panel Priority 2012 The 2012 wave is a Food and clothing (respondents were asked to specify the share of
Survey follow-up of a 1998 household expenditures going to the main adult man, woman, sons,
nationally representative daughters, and other household members)
survey
China China Health 1989, 1991, 1993, Select provinces Food, alcohol, and cigarettes (24-hour recall over three days, plus
and Nutrition 1997, 2000, 2004, household food inventory)
Survey 2006, 2009, 2011,
2015
Senegal Poverty and 2006–07, National Most consumption is captured at the cell level (for example, clothing,
Family Structure 2010–12 mobile phones, transport, food outside the home); food consumed at home
Survey is individualized based on accounts about which meals are shared and
estimates of the preparation costs
Note: The italicized years are used in the case studies.
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 133
China Typical household-level data miss a sub-
In China, intrahousehold and gender dy- stantial portion of inequality among indi-
namics over the past 20 years have evolved viduals. A comparison of an individual-level
against the backdrop of rapid economic and measure of extended food consumption to a
demographic change. The China Health and household-level measure, where the latter is
Nutrition Survey data allow the computation normalized for differences in household de-
of an individual measure of extended food mographic composition using equivalence
consumption, which includes all food items scales highlights this clearly. In the rural
as well as alcohol and tobacco (Santaeulàlia- (urban) subsamples, household consumption
Llopis and Zheng 2017).8 In 1991, extended per adult equivalent misses about 41 percent
food consumption was twice as high among (38 percent) of individual inequality. This is
men as among women, and, by 2009, this ratio again driven primarily by individual inequality
had risen to 2.3. This gender gap is, however, in the consumption of alcohol, tobacco, coffee,
largely accounted for by four items—tea, cof- and tea. Core food consumption inequality
fee, alcohol, and tobacco—that are consumed among small children ages 0–5 is about twice
disproportionately by men and may reflect as high as the inequality among adults.
different degrees of control over resources
or social norms about acceptable behavior Burundi
for men and women. Excluding these items Burundi is one of the poorest countries in
gives a narrower measure of core food con- Africa, with a legacy of conflict and violence.
sumption, according to which consumption Mercier and Verwimp (2017) use a household
is about 12 percent greater among men than survey conducted in 2012 that asked mostly fe-
among women, a ratio that has remained male respondents to specify how categories of
close to constant and could reflect differences consumption goods were allocated within the
in caloric need between men and women. household to examine intrahousehold con-
Analysis over the life cycle shows that the sumption inequality.9 The data show a gender
gender gap in extended food consumption gap in clothing and food expenditures (the
starts to emerge at about age 15 and peaks latter less pronounced) that benefits women.
between the ages of 45 and 55, after which it Among children, the consumption shares of
declines sharply (figure 5.5). In contrast, the food and clothing appear to be balanced be-
gender gap in core food consumption peaks tween boys and girls. The large share of miss-
much earlier, at around age 17–18, and stays ing values in item groups other than food and
at a similar level until age 50. clothing prevents additional analysis.
FIGURE 5.5 The Gender Gap in Food Consumption over the Life Cycle, China
a. Extended food consumption b. Core food consumption
6,000 3.5 2,000 3.5
Annual consumption (US$)
Annual consumption (US$)
3.0 1,800 3.0
5,000
2.5 2.5
Gender gap
Gender gap
4,000 1,600
2.0 2.0
3,000 1,400
1.5 1.5
1,200
2,000
1.0 1.0
1,000
1,000 0.5 0.5
0 5 10 15 20 25 30 35 40 45 50 55 60 65 0 5 10 15 20 25 30 35 40 45 50 55 60 65
Age (years) Age (years)
Men Women Gender gap (right axis) No gender gap
Source: Based on Santaeulàlia-Llopis and Zheng 2017 and their supplementary material.
Note: The gender gap is the ratio of male-to-female consumption, based on a regression with age dummies and time fixed effects (pooling data from 1989, 1991, 1993, 1997,
2000, 2004, 2006, and 2009).
134 POVERTY AND SHARED PROSPERITY 2018
Assuming equal sharing among siblings ual consumption depends more on a person’s
of the same sex, irrespective of age, one may position within the household than on age.
use the reported resource shares for food and In Burundi, unlike in the other countries
clothing to compute a partially individual- discussed in this section, women appear to
ized measure of consumption. Taking into be less poor than men. This highlights the
account differences in caloric needs by sex context specificity of intrahousehold distri-
and age through the use of equivalence scales bution rules. However, another potential ex-
yields poverty rates of 65 percent among planation for the higher consumption shares
men, 56 percent among women, and 77 per- among women may be that women overes-
cent among children. Because of the dispro- timate their consumption relative to that of
portionate incidence of child poverty, chil- their husbands, for example because of inter-
dren make up 68 percent of the hidden poor, nalized social norms or because they are not
that is, poor individuals living in nonpoor aware of some components of consumption
households, significantly more than their among their husbands, such as food con-
share in the sample population (56 percent). sumed away from home. Relying on one
Mirroring the results from Senegal below, the (female) respondent who reports about other
age effect becomes insignificant if the analy- members’ consumption (see also box 5.3 for
sis controls for the household member’s sta- alternative measures of food security) may
tus within the family, suggesting that individ- generate some measurement error.
BOX 5.3 Dietary Diversity as an Indicator of Individual-Level Food Security
The four case studies show in question. Some measures Although the dietary diversity of
intrahousehold inequalities in additionally account for how often mothers and their young children
the consumption of calories and a given food (or items from a given tends to be strongly correlated,
nutrients, a pattern also found food group) is consumed. Common children often consume fewer
to varying degrees in Ethiopia metrics for dietary diversity are food groups than their mothers
(Coates et al. 2017), India Household and Individual Dietary (Amugsi, Mittelmark, and Oduro
(Fledderjohann et al. 2014), Nepal Diversity Scores (Maxwell, Vaitla, 2015; Nguyen et al. 2013). In
(Harris-Fry et al. 2018), and South and Coates 2014), which count the Bangladesh, even more food
East Asia (Bühler, Hartje, and number of food groups consumed secure households have poor infant
Grote 2018). A double burden over a 24-hour recall period by diets (Owais et al. 2016). Among
of malnutrition—simultaneous the whole household (reflects children in Nepal, older children
presence of undernourished the household economic ability have better dietary diversity,
and overweight individuals—is to access a variety of foods) or but their diets are more likely to
occurring in many households individual members (reflects dietary deteriorate when the household
and countries, for example, in quality and nutrient adequacy experiences a negative shock.
middle-income countries, stunted [Arimond et al. 2010; Moursi et al. Younger children have less diverse
children living with obese mothers 2008; Savy et al. 2005; Torheim but more stable diets (Finaret et al.
(Aitsi-Selmi 2015). et al. 2004]). 2018). In India, children’s diets vary
An alternative to the collection Individual-level dietary diversity by age and sex, with girls’ diets
of individual food consumption indicators are strongly correlated less diverse than boys’—especially
could be dietary diversity. It is with the three common measures in adolescence (Aurino 2017).
routinely collected for vulnerable of child undernutrition: stunting, In sum, individual-level dietary
individuals (infants and their wasting, and underweight diversity metrics are a promising
mothers) in household health (Arimond and Ruel 2004; approach to assess individual food
surveys, but less frequently Chandrasekhar et al. 2017; Headey security (Bühler, Hartje, and Grote
collected for individuals in and Ecker 2013; Mallard et al. 2016; 2018; Headey and Ecker 2013;
household consumption surveys. Rah et al. 2010; Ruel 2003). Across Leroy et al. 2015). Adding these
Dietary diversity indicators countries, even a very simple questions to existing household
capture the number of food dietary diversity measure is better consumption surveys could
items or groups consumed, at predicting malnutrition than provide an alternative source of
often weighted according to the calorie deprivation (Headey and information about differences
nutritional importance of the food Ecker 2013). within households.
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 135
FIGURE 5.6 Caloric Shortfalls of Male Heads and Other Household ries and micronutrients are classified as ad-
Members, Bangladesh equately nourished, and those who do not
are classified as undernourished. Similarly,
0.0006 a household is adequately nourished if the
total household caloric availability exceeds
the sum of the individual dietary require-
ments. The analysis reveals that male heads
0.0004
have much smaller caloric and micronutri-
Density
ent shortfalls than other household members
(figure 5.6).
0.0002 These differences lead to the misclassifica-
tion of individuals relative to their household
status, that is, undernourished individuals
in adequately nourished households or ade-
0 quately nourished individuals in undernour-
−6,000 −4,000 −2,000 0 2,000 4,000 ished households. Overall, the proportion of
Caloric shortfall
misclassification varies between 18 percent
and 30 percent according to the type of mem-
Male heads Other household members
ber (first row of table 5.3) but in adequately
Source: D’Souza and Tandon 2018. nourished households, 55 percent of boys
and 47 percent of girls are undernourished
Bangladesh (whereas only 22 percent of heads and 9 per-
A significant portion of the population in cent of spouses are undernourished, third row
Bangladesh is undernourished in terms of of table 5.3).
calories and specific micronutrients. Studies
have also repeatedly demonstrated inequi- Senegal
table intrahousehold resource distribution. The household structure in Senegal, as in
D’Souza and Tandon (2018) use the Bangla- other West African countries, is complex be-
desh Integrated Household Survey to explore cause of polygamy and the frequent presence
intrahousehold differences in undernourish- of foster children. This offers opportunities
ment.10 The analysis draws on data of 3,060 to explore intrahousehold inequality within
rural households with male heads who are extended families. The 2006/07 Poverty and
married and whose spouses are present, but Family Structure Survey, described in De
without pregnant or lactating women. In- Vreyer et al. (2008), can be used to construct
dividual shortfalls from minimum dietary a relatively individualized measure of con-
energy requirements are computed. Individ- sumption and poverty status. To reflect intra-
uals who meet these requirements in calo- household structure and resource allocation
TABLE 5.3 Individuals Misclassified by the Household Measure of Caloric Availability
Measure Male heads Spouses Boys Girls Other adults
All households
Share 0.24 0.18 0.30 0.28 0.22
Number 3,060 3,060 2,462 2,342 1,722
Adequately nourished households
Share 0.22 0.09 0.55 0.47 0.15
Number 1,901 1,901 1,257 1,207 1,207
Undernourished households
Share 0.26 0.32 0.05 0.09 0.39
Number 1,159 1,159 1,205 1,135 515
Source: D’Souza and Tandon 2018.
Note: Shares = population-weighted means of undernourished individuals in adequately nourished households and
adequately nourished individuals in undernourished households. Number = observations.
136 POVERTY AND SHARED PROSPERITY 2018
more accurately, each household is divided ated with the household-reported male head.
into cells whereby the household-reported The same is true for sisters versus brothers.
head forms a cell with unaccompanied de- Cells headed by women in a leviratic union—
pendent members; each wife of the head and that is widows who “remarried” their former
her children and any other dependents then husband’s brother or other male relative—
form separate cells, as do other adults with have a higher probability of being poor, at
dependents, for example, married brothers. an odd ratio of 1.4 relative to women in their
This cell structure is characteristic of house- first marriage, but the difference is not sta-
holds in Senegal. tistically significant (De Vreyer and Lambert
The cell consumption data show that in- 2017 and their supplementary material).
trahousehold inequality accounts for almost
14 percent of total consumption inequality Taken as a whole, these studies give an idea
in Senegal, driven largely by intrahousehold of the potential misclassification of individ-
inequality in nonfood consumption. About uals with respect to households’ poverty clas-
13 percent of the poor live in nonpoor house- sification: many poor individuals do not live
holds and are hence invisible in standard in poor households. In addition, they point
measures of poverty (De Vreyer and Lambert out complex relationships between sex, age,
2017). There are also important gender dif- and status within the household, especially in
ferences. Per capita consumption is 33 per- nonnuclear households, making it difficult to
cent greater among cells headed by a man disentangle those effects. Furthermore, there
than among those headed by a woman, and are potentially complex interactions between
this difference is statistically significant. This the way the data were collected (for example,
pro-male-headed cell gap in consumption single or multiple respondents in the house-
narrows if the analysis controls for education hold, direct enumerator observation), the
because literacy and numeracy outcomes variable analyzed (caloric intake, food con-
are worse among women than among men. sumption, total consumption), and the level of
The remaining gender difference appears to disaggregation (individual-level analysis, cells/
be mainly attributable to the greater depen- subgroups of household members, or broad
dency ratio in female-headed cells because categories such as children/women/men).
children are ascribed to their mother’s cell
(and not their father’s) if the mother is pres- Estimating individual
ent in the household (De Vreyer and Lambert consumption from
2017 and their supplementary material).
household-level data
The social roles ascribed to women imply
that their position in the household and their Collecting data on individual-level consump-
marital status are much more strongly asso- tion is costly and not always feasible in the
ciated with their material well-being than is context of large-scale household surveys.
the case for men. The mothers and daughters Even specialized datasets, such as the ones
of the household-reported male head, and, presented earlier in this section, tend to in-
to a lesser extent, his junior wives tend to be dividualize only some components of the
found in the least favored positions in the overall consumption basket and thus provide
household, whereas no equivalent consump- a partial picture of sharing within house-
tion penalty exists among fathers and sons. holds. Moreover, basing our understanding
Widowed women, whether remarried or not, of intrahousehold differences in well-being
are also particularly vulnerable. These gender and poverty on differences in the consump-
differences in per capita consumption extend tion of specific consumption items is prob-
to poverty. A cell headed by a daughter of the lematic if preferences over those items differ
household-reported male head is 2.5 times between household members. For example,
more likely to be poor than the cell associated even if men disproportionately consume
with the household head, whereas there is no alcohol and tobacco, women might spend
significant difference in poverty status be- more on other items so that any subset of
tween cells headed by sons and those associ- items cannot provide the full picture (Tian,
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 137
Yu, and Klasen 2018). An alternative ap- imposes strong assumptions on the ways in
proach is to model intrahousehold resource which households and individuals behave,
allocation on the basis of the observed behav- and those assumptions are open to criti-
ior of the household and a structural model cism (Basu 2006; Cuesta 2006; Doss 1996;
that describes the preferences of household Sen 1990; Udry 1996; World Bank 2017b).
members and how they make decisions (for For example, this literature is largely based
example, the collective household model pi- on the standard assumption of utility max-
oneered by Chiappori 1988, 1992). Armed imization and does not consider alternative
with this structural model, and exploiting the explanations of human behavior. Likewise,
fact that many household surveys collect con- the collective model assumes that all house-
sumption data of one or two items in a way hold decisions are efficient—in other words,
that can be “assigned” to individuals, demand whatever decision the household takes, no al-
functions can be estimated that allow for ternative decision would have been preferred
teasing out how resources are shared inside by all its members. This rules out inefficient
the household even if data on consumption bargaining outcomes, whereby households
of most items are collected at the household may get trapped in situations where at least
level (see annex 5A for further details). This one household member could be made bet-
approach has two main advantages. First, it ter off without making the others worse off
allows an estimation of the resource shares (see Basu 2006; World Bank 2017b). Because
of women, men, and children over the entire of these assumptions, and additional econo-
consumption basket and therefore provides metric challenges in estimating the sharing
a more complete picture of the allocation rules empirically, model-based estimations of
of resources within households. Second, be- individual resource shares warrant additional
cause the data requirements are modest, this validation and sensitivity analysis before they
approach could open the door to estimating can be used in routine poverty monitoring.
individual-level poverty in many countries, As a first step in this direction, we use the
beyond the select few case studies discussed model proposed by Dunbar, Lewbel, and
in the previous section. A small but growing Pendakur (2013) to estimate consistent in-
literature uses model-based estimates of in- trahousehold differences in resource allo-
trahousehold resource allocation to explore cation and poverty in nuclear households
differences in poverty between women and in two countries (Bangladesh and Malawi).
men or between adults and children in devel- The model has the advantage that it is con-
oping countries.11 siderably less complex than previous ap-
Estimating individual poverty in this proaches, which enhances transparency and
way requires that at least some parts of the makes estimating individual resource shares
household consumption basket can be as- across countries more feasible using the same
signed to individuals. In other words, one method (see annex 5A). Figure 5.7 shows esti-
observes who within the household con- mates of resource shares in Bangladesh (pool-
sumes what—either because the underly- ing data for 2011/12 and 2015), with either
ing household survey disaggregates items food or clothing as the assignable good, and in
in such a way (for example, men’s clothing, Malawi in 2004/05 and 2010/11, with clothing
women’s clothing, and children’s clothing), as the assignable good.12 The horizontal axis
or because the survey asks respondents to gives the percentage of household resources,
assign an item to specific household mem- both the point estimate and the confidence
bers. These data requirements are modest. In interval, that are allocated to an individual of
fact, most studies rely on a single assignable type j living in a household of type s, holding
good, typically clothing, that is disaggregated the other household characteristics fixed at
among men, women, and children in many their mean. On the vertical axis are the types
standard household surveys. However, the of individuals and household sizes. The share
underlying structural model estimates the of household resources that goes to children
resource shares of men, women, and chil- has been divided by the number of children.
dren over the entire consumption basket. The The results on Bangladesh in figure 5.7,
flip side of this is that the structural model panel a, which use food as the assignable
138 POVERTY AND SHARED PROSPERITY 2018
FIGURE 5.7 Estimated Consumption Allocation, Men, Women, and Children, Bangladesh and Malawi
a. Bangladesh, 2012–15 (food) b. Bangladesh, 2012–15 (clothing)
Man in 1-child hh Man in 1-child hh
Man in 2-child hh Man in 2-child hh
Man in 3-child hh Man in 3-child hh
Man in 4-child hh Man in 4-child hh
Woman in 1-child hh Woman in 1-child hh
Woman in 2-child hh Woman in 2-child hh
Woman in 3-child hh Woman in 3-child hh
Woman in 4-child hh Woman in 4-child hh
Child in 1-child hh Child in 1-child hh
Child in 2-child hh Child in 2-child hh
Child in 3-child hh Child in 3-child hh
Child in 4-child hh Child in 4-child hh
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Resource share (%) Resource share (%)
c. Malawi, 2004–5 (clothing) d. Malawi, 2010–11 (clothing)
Man in 1-child hh Man in 1-child hh
Man in 2-child hh Man in 2-child hh
Man in 3-child hh Man in 3-child hh
Man in 4-child hh Man in 4-child hh
Woman in 1-child hh Woman in 1-child hh
Woman in 2-child hh Woman in 2-child hh
Woman in 3-child hh Woman in 3-child hh
Woman in 4-child hh Woman in 4-child hh
Child in 1-child hh Child in 1-child hh
Child in 2-child hh Child in 2-child hh
Child in 3-child hh Child in 3-child hh
Child in 4-child hh Child in 4-child hh
0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80
Resource share (%) Resource share (%)
Source: Gaddis et al., forthcoming.
Note: The horizontal axis gives the percentage of household (hh) resources, both the point estimate and the confidence interval, that are allocated to an individual of type j
living in a household of type s, holding the other household characteristics fixed at their mean. On the vertical axis are the types of individuals and household sizes. The share of
household resources that goes to children has been divided by the number of children. hh = household.
good, show that, in households with one or with two children, 29 percent in households
two children, men receive about 37 percent of with three children, and 26 percent in house-
the resources. The share of resources going to holds with four children. Among the children,
men is smaller in households with three chil- an only child receives, on average, about 21
dren (31 percent) and in households with four percent of the resources. In households with
children (27 percent). In households with one multiple children, each child receives between
child, women’s resource shares are larger than 12 percent and 14 percent of the resources.
those of men (42 percent), but their resource The broad patterns in resource allocation
shares decline more steeply as the number of for Bangladesh are similar if one uses cloth-
children increases, to 35 percent in households ing as the assignable good (figure 5.7, panel
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 139
b), which lends credibility to the estimation ferences in results underscore the need to fur-
method.13 However, the precision is much ther explore the robustness of model-based es-
greater with food, presumably because of timates of intrahousehold resource allocation.
food’s larger share in household consump- In Malawi in 2004/05 (figure 5.7, panel
tion (33 percent versus 3 percent). Moreover, c), one finds that the share of household re-
in households with more than one child, sources going to men does not vary with the
the resource shares of women are somewhat number of children. It is greater than the
smaller, and the resource shares of children share of resources going to women, though
are larger if the estimation is based on food. the confidence intervals overlap. The share of
These estimates suggest inequalities in resources going to women also does not vary
the way resources are shared among house- significantly with the number of children.
hold members, particularly between adults The share of resources going to each child is
and children. However, unlike the nutrition- not significantly different in households with
centered Bangladesh case study presented ear- one, two, or three children, but it is smaller
lier, the estimates in this section do not sug- when there is a fourth child. Focusing on the
gest that women fare worse than men. One confidence intervals together with the point
explanation for this divergence could be that estimates, the results on Malawi in 2010/11
D’ Souza and Tandon (2018) use a measure are qualitatively similar (figure 5.7, panel d)
of needs; another is that we are looking at a apart from the fact that the resource share of
different sample—nuclear households here, men is greater in households with one child
compared with all couple-households, exclud- than in households with more children.14
ing pregnant and lactating women, in D’Souza One may use the resource shares to esti-
and Tandon (2018). Yet another explanation is mate poverty rates among men, women, and
that, per definition, the approach used in this children, depending on the size of the rele-
section uses information on the assignable vant household. This requires additional as-
good to estimate individual-level resource al- sumptions about household economies of
location over the entire consumption basket, scale and the relative needs of children. The
beyond just food and nutrition. Still, these dif- estimates here follow Dunbar, Lewbel, and
Pendakur (2013) in relying on an equivalence
scale used by the OECD. Figure 5.8 summa-
FIGURE 5.8 Individual Rates of Poverty, Nuclear Households, rizes the information on Bangladesh (using
Bangladesh and Malawi the more precise estimates based on food
100
as the assignable good) and on Malawi (using
the latest available survey). In both countries,
90 88 the estimated poverty rates are significantly
80 higher among children than among adults.
73 The model estimates that women are poorer
70 68
than men in Malawi, but not in Bangladesh.
Poverty rate (%)
60 However, these results only apply to nuclear
49 households. These make up the largest share
50 46 45 of poor households globally but are often less
40 poor than extended multigenerational house-
30
holds (see the previous section).
20
An individual perspective on
10 multidimensional poverty
0 The chapter now builds on the multidi-
Bangladesh, 2012–15 Malawi, 2010–11
mensional approach described in chapter 4,
Men Women Children which captured deprivations in education,
health and nutrition, access to services, and
Source: Gaddis et al., forthcoming.
Note: Based on estimated resource shares in figure 5.7, panel a, Bangladesh, using food as the assign- security, in addition to monetary poverty.
able good; panel d, Malawi, using the 2010–11 data. Bringing the multidimensional approach to
140 POVERTY AND SHARED PROSPERITY 2018
individuals takes advantage of the fact that, in Data on five countries—Ecuador, Indo-
most household surveys, in contrast to con- nesia, Iraq, Mexico, and Tanzania—are used
sumption expenditures, nonmonetary indi- to exemplify how one might apply the multi-
cators in a few key dimensions of well-being, dimensional poverty measure to the indi-
such as education and nutrition, are often vidual.15 The focus is on adults (18+ years)
collected on an individual basis. For example, because some of the indicators are not di-
educational attainment is often lower among rectly valid for infants and young children,
adult women than among adult men because such as educational attainment or the BMI,
of past gender gaps in school enrollments, and because a multidimensional measure of
and these differences within the household child poverty should consider child-specific
can be captured by a measure of multidimen- vulnerabilities (box 5.4).
sional poverty among individual household The analysis uses the same five dimensions
members. of multidimensional poverty as the country
The multidimensional poverty measure in- case studies in chapter 4.16 The datasets have
troduced in chapter 4 combines monetary and been selected on the basis of the availabil-
nonmonetary dimensions of well-being, but ity of information on individuals, but the
it relies on households as the unit of analysis. surveys provide information only about in-
By way of illustration, consider the dimension dividual deprivations in the education and
of education. The measure retroactively col- health-nutrition dimensions. The individual
lapses the information about the educational multidimensional poverty measure considers
attainment of individual household members adults deprived in the education dimension if
into an indicator for the household, whereby they have not completed primary schooling,
the household is deprived if no adult member and they are considered deprived in the nu-
has completed primary education. Like the trition indicator of the health and nutrition
monetary poverty estimates in chapter 1, the dimension if they are undernourished (table
household multidimensional poverty mea- 5.4). The other dimensions—monetary pov-
sure in chapter 4 cannot provide insights into erty, access to services, and security—and the
differences within households. health indicator of the health and nutrition
BOX 5.4 Child Poverty
Children growing up in extreme as adults, and the transmission and face immediate threats
poverty require special attention. of poverty down the generations, such as gender-based violence,
They experience poverty differently including through early marriage. recruitment as child soldiers, and
than adults, and their needs and Beyond this sad and avoidable discrimination in the provision of
vulnerabilities change rapidly in impact on human life and potential, basic services. Irregular migration,
ways that are foreign to adults neglecting children fails to build the displacement, and trafficking create
(Abdu and Delamonica 2017). human capital the world needs for multiple dangers for children; girls,
Key dimensions of poverty sustained economic prosperity. especially, are disadvantaged
among children include health, The numbers are stark: because of gender inequalities.
information, nutrition, education, Children are more than twice Children living in poverty
water, sanitation, and housing. as likely as adults to be living in often experience stress, anger,
Poverty causes poor children to poor households (the results are frustration, sadness, and
miss out on a good start in life. robust to the use of 32 different hopelessness because of the
The consequences of inadequate equivalence scales, and the repeated instances of discrimination
nutrition, deficient early stimulation youngest children are the least well and social exclusion they encounter,
and learning, and exposure to off [Newhouse, Suárez-Becerra, which may lead them to drop out of
stress and shame last a lifetime. and Evans 2017]). More than school, lose friends, and become
They lead to stunted development, half (58 percent) of the children exposed to risks that more well off
low capacity in the skills needed for in fragile and conflict-affected children and adults never have to
work, restrained future productivity situations live in poor households face (Save the Children 2016).
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 141
TABLE 5.4 Indicators and Dimensions, the Individual and Household Multidimensional Poverty Measure
Deprived if Weight
Dimension Individuals Households (%)
Monetary poverty Daily consumption per capita < US$1.90 20
No adult has completed primary school
Education Adult has not completed primary school 20
Any school-aged child is not attending school
Health and nutrition Any woman (ages 15–49) experiencing a live birth in the previous 36 months did not deliver at a facility 20a
Any child (ages 12–59 months) did not receive a DPT3 vaccination
Any woman (ages 15–49) is undernourished (BMI < 18.5)
Adult undernourished (BMI < 18.5)
Any child (ages 0–59 months) is stunted
Access to services No access to an improved source of water within a round trip distance of 30 minutes 20
No access to improved sanitation facilities for use exclusively by the household
No access to electricity
Security Household has been negatively affected by crime in the previous 12 months or lives in an area where more than 20% of 20
households have been negatively affected by crime
Note: Dimensions on which data on individuals are available are shaded gray. BMI < 18.5 = body mass index below 18.5 (underweight); DPT3 = diphtheria-pertussis-tetanus
vaccine.
a. Health and nutrition each has a weight of 10 percent.
dimension may be analyzed meaningfully in those dimensions that can be measured
only among households with the existing among individuals. Nonetheless, even a par-
data. Thus, the multidimensional poverty tially individualized multidimensional pov-
measure is de facto only partially individ- erty measure reveals that multidimensional
ualized; only 30 percent of deprivations are poverty is greater among women than among
measured among individuals. This is a clear men in the countries under examination,
limitation because one must fall back on the driven by women’s disadvantaged position in
assumption of equal sharing among house- educational attainment.
hold members in the other indicators and di- Figure 5.9 shows the share of men and
mensions (70 percent), and this dilutes what- women who are deprived in the two indica-
ever intrahousehold inequality one may find tors on which data on individuals are avail-
FIGURE 5.9 Gender Gaps, Education and Nutrition Deprivation, Selected Countries
a. Education b. Nutrition
50 50
45 45
Share deprived in education (%)
Share deprived in nutrition (%)
40 40
35 35
30 30
25 25
20 20
15 15
10 10
5 5
0 0
Ecuador Indonesia Iraq Mexico Tanzania Ecuador Indonesia Iraq Mexico Tanzania
Men, household measure Women, household measure Men, individual measure Women, individual measure
Source: Klasen and Lahoti, forthcoming.
142 POVERTY AND SHARED PROSPERITY 2018
able: education and nutrition. For each coun- vation and poverty using a household-level
try and indicator, deprivation rates among approach (Klasen and Lahoti 2016).
men and women are compared through two The share of men and women who are
approaches: one relying on the household, multidimensionally poor, measured across
whereby all household members are assigned individuals, is shown in figure 5.10. Multidi-
the same deprivation status, and the other mensional poverty is more prevalent among
relying on the individual, measuring individ- women than among men in all countries,
ual deprivations directly.17 with the largest gender gap in Iraq (54 percent
In education (figure 5.9, panel a), the versus 38 percent). Klasen and Lahoti (forth-
household approach reveals some gender coming) show that a significant gender gap
differences in education deprivation that in multidimensional poverty is also found in
tend to disadvantage women, showing that India.
women are more likely than men to live in a These gender gaps may even be wider
household where no adult has completed pri- among the most vulnerable groups. For exam-
mary school. These gender differences, which ple, in all countries but Ecuador, widows are
are muted under the household approach, significantly more likely to be multidimen-
are amplified if the data on individuals are sionally poor than widowers, and the gender
used. In the five countries under examina- gap ranges from 8 percentage points in Iraq to
tion, women are much more likely to be de- 19 percentage points in Mexico (Klasen and
prived in education than men if deprivations Lahoti, forthcoming). This highlights widow-
are measured across individuals, especially in hood as an important vulnerability factor
Iraq (a gap of 19 percentage points). In ad- among women, which is not revealed in the
dition to these gender gaps, an individual, household multidimensional poverty mea-
whether a man or a woman, is more likely sure (Djuikom and van de Walle 2018).
to be considered deprived in education if The gender gaps illustrated in figure 5.10
the measure of deprivation is applied across are probably still an underestimation of the
individuals. This reflects the fact that the true extent of gender inequality in multidi-
household education indicator is defined in mensional poverty. Because of data limita-
an expansive way, that is, all household mem- tions, even the individual multidimensional
bers are considered nondeprived if any adult poverty measure individualizes only some
in the household has completed primary
school, irrespective of who in the household
FIGURE 5.10 Gender Gaps, Individual Multidimensional Poverty,
benefited from education and whether there
Selected Countries
is any systematic gender bias. (Klasen and La-
hoti 2016 show that defining deprivation in 80
this way will lead to an underestimation of
Share who are multidimensionally poor (%)
deprivation and poverty rates using a house- 70
hold-level approach because typically many
60
deprived individuals live in households where
one member has the required education.) 50
In terms of nutrition (figure 5.9, panel
b), gender gaps are small, even if measured 40
with reference to individuals, and they do
not show a consistent pattern.18 Unlike the 30
case of education, a person is less likely to be
20
considered deprived in nutrition under the
individual approach than under the house- 10
hold approach. This is because the household
nutrition indicator is defined restrictively, 0
that is, all household members are considered Ecuador Indonesia Iraq Mexico Tanzania
deprived if any adult in the household is un- Men Women
dernourished, which will overestimate depri- Source: Klasen and Lahoti, forthcoming.
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 143
of the dimensions in which one may expect on simultaneous activities (watching a child
to find variations within households and while selling at the market) also hide the pro-
systematic gender differences. As discussed found effect these differences have for labor
in the previous section, intrahousehold in- force participation decisions, types of jobs,
equalities in consumption may disadvantage and hours spent working for pay or profit.
women and children. But, because none of Participatory poverty research often shows
the datasets used here allows estimates of re- that, although insufficient financial means
source allocation across individuals, the in- are central to the experience of destitution
dividual multidimensional poverty measure among poor people, they are interlocked with
must fall back to reliance on (unsatisfactory) other dimensions, such as voicelessness, so-
assumptions about equal sharing associated cial exclusion, shame, exposure to violence,
with the monetary poverty dimension. Sim- lack of access to basic infrastructure and ser-
ilarly, other studies have shown a gender di- vices, lack of education, poor physical and
mension in access to services. For example, mental health, and illness. Box 5.5 summa-
the individual deprivation measure, a new rizes findings from recent and ongoing par-
gender-sensitive multidimensional measure ticipatory analysis of poverty (Narayan et al.
of poverty, illustrates how men and women 2000a; Walker and Godinot 2018).19
are affected differently by lack of access to
services because of social norms assigning
Conclusion
domestic work to women (Hunt 2017; IDM
2017). A more refined individual multidi- This chapter starts with a question: How
mensional poverty measure would also cap- many women and children are poor? De-
ture women’s and men’s exposure to all forms spite the conceptual challenges in answering
of violence under the security dimension. this question and the data limitations, accu-
Some forms of violence, particularly gender- mulating evidence using different methods
based violence and especially intimate part- and data sources confirms the existence of a
ner violence, are more frequently experi- pattern of consumption inequality between
enced by women than by men (Stöckl et al. children and adults and between women and
2013; UBOS and ICF International 2017). In men in the household. The results suggest
contrast, men are more susceptible to recruit- that women are disproportionately affected
ment in gangs and armed groups. An individ- by poverty. Likewise, the global poverty data
ual measure of exposure to violence could re- and country studies show that children are
veal such differences within households and poorer than adults, which is partly driven by
lean toward greater intrahousehold variation demographic patterns of fertility and house-
in the multidimensional poverty measure. hold formation. However, the picture of how
Another direction for expanding the in- much poorer children are in relation to adults
dividual multidimensional poverty measure is sensitive to assumptions about the relative
along gender lines would be to broaden the needs of children, which requires further
set of dimensions, to include time use and investigation beyond the scope of this chap-
socioemotional dimensions of poverty. As ter. In several countries, households seem to
discussed earlier, patterns of time use are very share basic food items somewhat equitably,
different between men and women, especially but inequality among gender lines is stronger
in the presence of children. Many studies for more prized consumption items.
(World Bank 2011; Bardasi and Wodon 2010; These general patterns mask contextual
Blackden and Wodon 2006; Rubiano Matu- variation related to the position of individu-
levich and Viollaz 2018) show the persistent als in the life cycle (marital status and parent-
gap between time spent in market and non- hood), their status within the household (the
market activities, with women consistently sons, first wife, or mother of a man who is the
spending twice as much time as men in the household head hold higher relative status
latter (household chores, child and elderly than his daughters or more junior wives), and
care) and often having less leisure time. their human capital and position in the labor
Data limitations on the actual distribution of market (schooling and employment status).
time between care and household chores and Because of gendered social norms that view
144 POVERTY AND SHARED PROSPERITY 2018
BOX 5.5 Gender and Socioemotional Dimensions of Poverty: Participatory Studies
The World Bank (2017b) recognizes Provisional findings indicate role. Whereas women may face
that in-depth consultation with that, while lack of financial sexual exploitation and gender-
people experiencing poverty is resources and the inability to based violence, especially as
essential to an understanding of meet basic needs are central, domestic workers, men face
the true nature of the multifaceted both women and men frequently exploitation and discrimination
phenomenon of poverty. The associate these needs with their as casual laborers. Children find
Voices of the Poor reports (Narayan direct consequences in terms themselves socially excluded at
et al. 2000a, 2000b) highlight of physical and mental health. school, singled out if they are
the importance of nonmonetary Shame, fear, depression, worry, unable to afford the totem items
dimensions, access to services, and anger emerge as integral of their peers. They are often
and gender norms. Under the components of the experience of embarrassed to invite friends home
strain of vast social, economic, poverty. Poverty is also relational. to their substandard housing.
and political transformation, poor As a group, people living in In rural areas, people living in
household members reflect on the poverty experience oppression, poverty may lack basic social and
contradiction between purported exploitation, humiliation, and the infrastructure service provision
gender roles—homemaker for denial of rights, including the locally, whereas, in cities, point of
women and breadwinner for denial of rights to health care and use charges deny them access.
men—and the reality of women education. As individuals, they Gender roles imply that lack of
performing income-earning tasks, experience social isolation, stigma, proximate clean water affects more
which increases their time poverty. and discrimination. Beyond their the time and lives of women (and
Under stress, men are more likely intrinsic importance, these factors children) who are responsible for
to abuse alcohol, and domestic also contribute to a lack of social fetching it, cooking, and cleaning.
violence spreads. All these factors and political voice and to relative Stigma is more contagious in rural
affect children negatively. powerlessness, all often resulting settings, afflicting all members of
Following the same approach, in social exclusion. extended families, than in urban
people living in extreme poverty Both women and men areas, where social life is more
in Bangladesh, Bolivia, France, emphasize these dimensions, individualized. Although poverty is
Tanzania, the United Kingdom, but they experience them pain, people experiencing it often
and the United States are leading differently. Gender roles mean demonstrate resourcefulness;
research with the International that women feel stress and stigma they acquire knowledge and skills
Movement ATD Fourth World and in the context of care and family that could be useful to others, and
Oxford University to understand responsibilities under tightly they feel they have a positive and
the dimensions of poverty that constrained domestic budgets. valuable contribution to offer to
matter most in their lives (Walker Men can feel emasculated if they society.
and Godinot 2018). cannot fulfill their breadwinning
unpaid work as a female prerogative, women Gender gaps are also pervasive in other
face a strong trade-off between reproductive key components of welfare. Although gender
and productive functions, and mothers who gaps in school enrollments have narrowed
do not work for pay are especially likely to significantly over the past decades (and in
live in poor households. Adult couples with some countries reversed), adult women
dependent children or other nonearners around the world continue to be disadvan-
ages 18–64 in the household are overly rep- taged in educational attainment because
resented among the poor. These gender gaps of past (and sometimes present) gender in-
in poverty are stronger in Sub-Saharan Af- equalities in access to schooling. Participatory
rica and South Asia; within countries, these research also highlights gender differences in
inequities seem stronger among the poorest, the socioemotional dimensions of poverty.
which has strong implications for reaching Advancing our understanding of poverty
the twin goals, reducing poverty and sharing among individuals requires a renewed em-
prosperity. phasis on individual-level data collection.
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 145
This chapter has touched upon various data ising but need to be put to the test in addi-
gaps limiting our understanding of individual tional validation studies and extended to
poverty. Three broad directions for data col- more complex household structures (beyond
lection and methodological survey research nuclear households). Specialized data collec-
emerge from this discussion. First, although tions and participatory research could help to
full individual-level consumption data col- test some of the key assumptions underlying
lection remains infeasible for most living these methods and explore the sensitivity of
standards surveys, there may be some scope results to alternative assumptions. Further
to collect partially individualized consump- investigations of how relative needs and pref-
tion data. This could take the form either of erences differ inside the household would
fielding an individual-level module to a sub- allow for a better understanding of whether
set of households or of identifying a subset an unequal resource allocation translates into
of consumption items (beyond clothing) that differences in well-being and poverty.
can signal inequalities within households The findings of this chapter have import-
and that can be collected for individuals (or, ant implications for policies and interven-
at the minimum, for men, women, girls, and tions to alleviate poverty and enhance shared
boys) in a reliable and cost-effective way. prosperity. Given the importance of maternal
Advancing this type of data collection would health and education for the formation of
facilitate the application of the collective children’s human capital in many contexts,
model to estimate intrahousehold resource better understanding intrahousehold poverty
shares. Second, expanding individual-level could help design more effective interventions
data collection on nonmonetary dimensions, to weaken its intergenerational transmission.
such as time use, violence, access to services Understanding differences in poverty levels
and assets, and some of the socioemotional between different household members is im-
dimensions highlighted by participatory re- portant for the effective targeting of poverty
search, would allow for the advance of multi- reduction programs. At present, commonly
dimensional measures of individual poverty used household targeting of social assistance
and analysis of the intersectionality of depri- programs may miss a significant share of the
vations. Third, additional methodological poor: those people hidden in overall nonpoor
research is needed to shed light on the differ- households. Understanding how gender and
ence, in terms of accuracy and cost, between age affect the demand for basic services is key
self- and proxy-reporting for data referring to to making sure that interventions to expand
individuals. The marginal cost of individual- basic infrastructure and social services ad-
level data collection is strongly influenced dress the differentiated needs and constraints
by whether survey enumerators need to in- of the poorest. Factoring in the potential
terview multiple household members (thus impacts of interventions on time use would
allowing for repeat visits to the household), benefit women disproportionately. Finally,
which has major implications for survey op- better understanding of the socioemotional
erations. Existing research highlights the im- dimensions of poverty would help increase
portance of respondent selection for data on the take-up of programs and strengthen their
assets and labor (on assets: Kilic and Moylan design and implementation by lifting rele-
2016; Doss, Kieran, and Kilic 2017; on labor: vant social and psychological barriers and de-
Bardasi et al. 2011; Dammert and Galdo creasing stigma. As more poverty alleviation
2013), but similar investigations would be programs focus on productive inclusion, the
useful for other dimensions of living stan- success of active and enabling policies that
dards and welfare, including consumption. stress agency and entrepreneurial initiative
In terms of research, recent advances in also depends on fostering the mindset that
the application of the collective bargaining help poor people and society recognize their
model to household survey data are prom- potential.
146 POVERTY AND SHARED PROSPERITY 2018
Annex 5A
Technical note: Estimating
intrahousehold resource shares
The basic approach by singles and by couples, with assumptions
on economies of scale for the public goods.
Most studies estimating intrahousehold re- An alternative route, which is followed in
source shares are based on the collective this chapter, is to use information on the
household model (Chiappori 1988, 1992). consumption of assignable goods, that is,
The collective model recognizes that house- goods that are consumed only by one type
hold members have their individual prefer- of individual in the household. For assign-
ences and assumes Pareto efficiency, that is, able goods, the household’s consumption
whatever decision the household takes, no al- is also the consumption of the individual,
ternative decision would have been preferred so that the household’s budget share for an
by all its members. In this model, it is as if assignable good (observed) is equal to the
each household member (that is, woman, product of an individual’s resource share by
man, or child) is allocated a fraction of the the budget share the individual would choose
household’s total resources (that individual’s subject to his or her own shadow budget
resource share), which the individual then constraint (both unobserved). The estimates
allocates according to his or her own prefer- presented in this section, which are based on
ences. Each household member determines the approach proposed by Dunbar, Lewbel,
his or her demand for each consumption and Pendakur (2013), make some further
item by maximizing his or her utility func- assumptions of similarity of certain aspects
tion, subject to the individual’s resource of preferences.20 The resource shares are
constraint (that is, determined by resource identified from the observation of the bud-
share) and a vector of shadow prices. These get shares of assignable goods (see below for
shadow prices are equivalent to market prices details).
for private goods, but lower than market
prices for goods that are shared by multiple The model underlying
household members. (Bourguignon and Chi-
appori 1992; Browning et al 1994; Chiappori
individual resource shares
and Meghir 2015.) There are two routes to Households are supposed to be composed
recover individual resource shares from ob- of one adult man, one adult woman, and s
served household expenditures. One is to children. Each household member is of type
assume that preferences of adults in couples j, where j = m, f, c for the adult man, adult
are no different from preferences of singles. woman, and children, respectively. Following
Consumption by adults in couples is then de- Dunbar, Lewbel, and Pendakur (2013), the
duced from the observation of consumption demand system can be written as follows:
*KM S 8KM \5]\6KM \5] P 79 \8KM \5]3]]
*HM S 8HM \5]\6HM \5] P 79 \8HM \5]3]] (5A.1)
8FM \5]
*FM S 8FM \5] b6FM \5] P 79 ` 3ac
1
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 147
where Wj,s is the household budget share of in this section points to differences not fully
member j’s assignable good in a household accounted by those.
with s children; hj,s(z) is the resource share 8. The China Health and Nutrition Survey is a
of household member of type j in a house- panel dataset that has tracked food consump-
hold with s children; x is the household’s total tion among individuals in about 6,800 house-
nondurable expenditure; and z is a set of so- holds in nine provinces since the early 1990s.
ciodemographic characteristics of the house- The survey records the quantity (in grams)
hold. The last equation in the demand system of a variety of food items, including alcohol
(5.1) gives the household budget share of the and tobacco, that each household member
children’s assignable good. The children are consumed at and between meals, at home
jointly treated as one member of the house- and away from home, during three days at a
hold; this requires the simplifying assump- level of detail suitable for nutritional analysis.
tion that resources are shared equally among Local prices are used to compute a monetary
the children. measure of consumption.
The term in parentheses in each equation 9. The Burundi survey included a module on
of the demand system (5.1)—aj,s(z) + b0 individual consumption, which asked a sin-
ln(hj,s(z)x)—is referred to as j ’s latent bud- gle respondent, a woman household mem-
get share (for j = m, f, and the corresponding ber considered responsible for the household
term for children). The latent budget share is budget, to specify the share of household
linear in the log of individual resources. consumption dedicated to five groups of in-
dividuals: the main adult man, the main adult
woman, the sons, the daughters, and all other
Notes household members. In about two-thirds of
1. This section draws on Muñoz Boudet et al. households, the woman respondents reported
(2018). that they were the wives of the household
2. These rates are higher than the rates in chap- heads whereas, in the remaining third, they
ter 1 because they are based on a subset of reported that they headed the households.
countries and household surveys (see box 10. The Bangladesh Integrated Household Survey
5.2). Corresponding rates for the 2015 GMD was conducted between December 2011 and
data are 11.4 and 11.7 percent for women and March 2012. It covered 5,000 households and
men, respectively. was representative of rural Bangladesh. The
3. In 2015, 19.3 percent of those ages 0–14 lived survey recorded individual food consump-
in poor households. tion, in grams, for over 300 food items for
4. Average age at marriage by country was 25 every household member during the previous
years for women (minimum 17.2 and max- 24 hours, as reported by the woman in charge
imum 33.8 years) and 28.4 years for men of cooking and serving.
(minimum 21.7 and maximum 36.5 years) 11. See Bargain, Donni, and Kwenda (2014) on
(World Marriage Data 2015 using the latest Côte d’Ivoire; Bargain, Kwenda, and Ntuli
data for 2013). (2017) on South Africa; Bargain, Lacroix,
5. Farmers are considered earners, even if they and Tiberti (2018) on Bangladesh; Belete
produce mostly for subsistence purposes, un- (2018) on Ethiopia; Brown, Calvi, and Pen-
less they are classified in the survey as unpaid glase (2018) on Bangladesh; Cuesta (2006) on
family workers. Chile; Dunbar, Lewbel, and Pendakur (2013)
6. To the best of our knowledge, these are the on Malawi.
few relatively recent datasets that collect 12. The results are based on pooling the Bangla-
consumption data with the level of detail desh Integrated Household Survey 2011–12
necessary for intrahousehold analysis and a and 2015 and on using the Malawi Integrated
significant geographical coverage. Other ex- Household Survey 2004–05 and 2010–11.
isting datasets are either limited in geographic 13. See Bargain, Lacroix, and Tiberti (2018) for a
scope, are outdated, or can only assign a small similar validation study.
proportion of consumption to individuals. 14. The resource shares are estimated less pre-
7. Although these smaller shares may reflect dif- cisely in Malawi than in Bangladesh, even in
ferences in needs or preferences, the evidence comparisons with resource shares estimated
148 POVERTY AND SHARED PROSPERITY 2018
on the basis of expenditures on clothing. This indicator (any woman [ages 15–49] in the
may arise because of differences in sample size household is undernourished) and the indi-
(4,149 households in Bangladesh against 3,045 vidual indicator (the adult is undernourished).
in Malawi in 2004/05). The additional estima- 18. In addition, most surveys are characterized by
tion of resource shares in Tanzania based on numerous missing values for nutrition among
pooling the 2012–13 and 2014–15 datasets individuals, which reduces the reliability of
did not yield interpretable results. The sample this indicator. This is because household sur-
size was considerably smaller, with only 1,552 vey protocols typically allow for only a limited
observations, which may explain why the esti- number of revisits to each household. House-
mation results were inconclusive. hold members who are not at home during
15. Details on the datasets used are presented in the first visit and subsequent revisits are not
chapter 4. This section does not include a dis- measured.
cussion of Uganda, because anthropometric 19. In this ongoing work to gain insight on the
information is not available on adults in that dimensions of poverty in six countries, each
country. national team of 10–15 people is responsible
16. Following chapter 4, the individual multi- for the local design, execution, and analysis
dimensional poverty measure gives equal of the research. Each team includes people
weight to each dimension (0.2), and all in- who are poor, but also academics and prac-
dicators within a dimension are weighted titioners who provide services or advocate for
equally. The only exception is the health and the poor. Outreach is undertaken among peo-
nutrition dimension; the two subdimensions ple of working age, the elderly, and children,
(health, nutrition) are weighted equally. For all of whom participate in detailed moderated
the Alkire-Foster (2011) measure, α = 0 is discussion, first, within peer groups of people
used, and a household classified multidimen- with similar experiences and, then, in mixed
sionally poor if it is deprived in at least 0.2 of groups that explore relationships across di-
the weighted indicators (k = 0.2). The results mensions and seek consensual conclusions.
are qualitatively similar for different parame- 20. The first is that Engel curves for the assign-
ters of the Alkire-Foster (2011) measure and able good have the same shape across house-
for the Datt (forthcoming) measure. hold members. The second is that preferences
17. In education, the approach compares the share are similar across household types, where
of adults deprived according to the household household types are differentiated by the
indicator (no adult has completed primary number of children living in the household.
school) and the individual indicator (the adult These assumptions can be used in isolation
has not completed primary school). In nu- or jointly (as done here) to identify the share
trition, the approach compares the share of of resources accruing to each member of the
adults deprived according to the household household.
INSIDE THE HOUSEHOLD: POOR CHILDREN, WOMEN, AND MEN 149
Appendix A
Data Details
The poverty and shared prosperity measures and regional estimates and to facilitate com-
and supporting analysis in this report are parisons across countries, PovcalNet aligns
based on household surveys from around the the surveys to specific reference years (for
world. Because the variables available in the additional details, see the chapter 1 section
household surveys differ across countries and of this appendix). This report is based on the
years, the country coverage varies from chap- September 2018 vintage of PovcalNet. The
ter to chapter according to the data require- PovcalNet poverty measures are used for the
ments for the analysis. As the data require- analysis of global poverty at the IPL in chap-
ments become more demanding, the subset ter 1 and for the analysis of poverty at higher
of countries that can meet them decreases. poverty lines in chapter 3 (table A.1).
Thus, the same country coverage is not possi-
ble across all five chapters of this report.
Global Database of
This data appendix first provides an over-
view of the main data sources for this report
Shared Prosperity
along with country classification definitions The Global Database of Shared Prosperity
applicable throughout the report. In the sub- (GDSP) includes the most recent figures on
sequent sections, chapter-specific data and annualized consumption or income growth
methodological issues, such as survey se- of the bottom 40 percent of the popula-
lection criteria, definitions, additional data tion (the bottom 40) and related indicators
sources, and key measurement issues are de- over similar time periods and intervals. All
scribed separately for each of the five chapters. numbers were vetted by an internal Techni-
cal Working Group. This report is based on
Main databases for the the sixth edition of the GDSP (the fall 2018
release), which features data on 91 econo-
report mies circa 2010–15 (http://www.worldbank
.org/en/topic/poverty/brief/global-database
PovcalNet
-of-shared-prosperity). The harmonized sur-
PovcalNet is an online tool for global pov- veys for the GDSP are all sourced from the
erty monitoring hosted by the World Bank Global Monitoring Database (see below). The
(http://iresearch.worldbank.org/PovcalNet). GDSP is the main data source for the shared
PovcalNet was developed with the purpose prosperity analysis presented in chapter 2 of
of public replication of the World Bank’s this report (see table A.1).
poverty measures at the international pov-
erty line (IPL). PovcalNet contains poverty
Global Monitoring Database
estimates from more than 1,600 household
surveys spanning 164 economies and over 40 The Global Monitoring Database (GMD) is
years, from 1977 to 2017. To produce global the World Bank’s repository of multitopic
151
TABLE A.1 Overview of Principal Data Sources by Chapter
Global Monitoring Global Database of
Database PovcalNet Shared Prosperity
Chapter 1: Ending Extreme Poverty Fall 2018 release, Fall 2018 release,
data from circa 2015 data from 1977–2017
Chapter 2: Shared Prosperity Fall 2018 release, Fall 2018 release,
data from circa 2010–15 data from circa 2010–15
Chapter 3: Higher Standards for a Fall 2018 release,
Growing World data from 1977–2017
Chapter 4: Beyond Monetary Poverty Fall 2017 release,
data from circa 2013
Chapter 5: Inside the Household Fall 2016 release,
data from circa 2013
income and expenditure household surveys By income
used to monitor global poverty and shared
prosperity.1 As of June 2018, the GMD con- The World Bank updates annually the income
tains more than 1,100 household surveys classification of economies. The income clas-
conducted in 156 economies. For a few econ- sification used in this report is based on the
omies, the welfare aggregate of the GMD World Bank’s 2018 fiscal year classifications.
spans up to 46 years, from 1971 to 2017, According to fiscal 2018 definitions, low-
whereas for most other economies, coverage income economies are defined as those with a
is significantly less. The household survey gross national income (GNI) per capita, cal-
data are typically collected by national statis- culated using the World Bank Atlas method,
tical offices in each country, and then com- of US$1,005 or less in 2016; lower-middle-
piled, processed, and vetted for inclusion in income economies are those with a GNI per
the GMD by the World Bank’s internal Tech- capita between US$1,006 and 3,955; upper-
nical Working Group. Selected variables have middle-income economies are those with
been harmonized to the extent possible such a GNI per capita between US$3,956 and
that levels and trends in poverty and other 12,235; and high-income economies are
key sociodemographic attributes can be rea- those with a GNI per capita of US$12,236 or
sonably compared across and within coun- more. The list of economies by income and
tries over time. The GMD’s harmonized mi- lending classification is available at https://
crodata are used in PovcalNet and the GDSP. datahelpdesk.worldbank.org/knowledgebase
In this report, the GMD is used for the /articles/906519-world-bank-country-and
global poverty profile in chapter 1, the multi- -lending-groups.
dimensional poverty measures in chapter 4,
and the individual poverty measures in chap- By geographical region
ter 5. Whereas chapter 1 uses the latest version
of the GMD, analyses in chapters 4 and 5 are In this report, the six geographical regions
based on previous versions (see table A.1). comprise (1) low- and middle-income econ-
omies, and (2) economies eligible to receive
loans from the World Bank (such as Chile)
Classification of economies and recently graduated economies (such
The economy classifications by income level, as Estonia). The aggregate for the six geo-
geographical region, and fragile and conflict- graphical regions is reported as the “sum of
affected situation are described in this sec- regions,” which in previous publications was
tion. The term country, used interchange- often referred to as the “developing world.”
ably with economy, does not imply political The economies excluded from the six geo-
independence but refers to any territory for graphical regions (as defined above), mostly
which authorities report separate social or high-income economies, are grouped in a
economic statistics. category called “rest of the world” in this
152 POVERTY AND SHARED PROSPERITY 2018
report. This group was often referred to as Madagascar; Malawi; Mali; Mauritania; Mau-
“other high-income” or “industrialized econ- ritius; Mozambique; Namibia; Niger; Nige-
omies” in previous publications. ria; Rwanda; São Tomé and Príncipe; Sene-
The economies in each of the six regions gal; Seychelles; Sierra Leone; Somalia; South
and the “rest of the world” category are listed Africa; South Sudan; Sudan; Tanzania; Togo;
below. Uganda; Zambia; Zimbabwe.
East Asia and Pacific: American Samoa; Rest of the world: Andorra; Antigua and
Cambodia; China; Fiji; Indonesia; Kiribati; Barbuda; Aruba; Australia; Austria; The Ba-
Democratic People’s Republic of Korea; Lao hamas; Bahrain; Belgium; Bermuda; British
People’s Democratic Republic; Malaysia; Mar- Virgin Islands; Brunei Darussalam; Canada;
shall Islands; Federated States of Micronesia; Cayman Islands; Channel Islands; Curaçao;
Mongolia; Myanmar; Northern Mariana Is- Cyprus; Denmark; Faroe Islands; Finland;
lands; Palau; Papua New Guinea; Philippines; France; French Polynesia; Germany; Gibral-
Samoa; Solomon Islands; Thailand; Timor- tar; Greece; Greenland; Guam; Hong Kong
Leste; Tonga; Tuvalu; Vanuatu; Vietnam. SAR, China; Iceland; Ireland; Isle of Man; Is-
Europe and Central Asia: Albania; Ar- rael; Italy; Japan; Republic of Korea; Kuwait;
menia; Azerbaijan; Belarus; Bosnia and Her- Liechtenstein; Luxembourg; Macao SAR,
zegovina; Bulgaria; Croatia; Czech Repub- China; Malta; Monaco; Nauru; Netherlands;
lic; Estonia; Georgia; Hungary; Kazakhstan; New Caledonia; New Zealand; Norway; Por-
Kosovo; Kyrgyz Republic; Latvia; Lithuania; tugal; Puerto Rico; Qatar; San Marino; Saudi
former Yugoslav Republic of Macedonia; Arabia; Singapore; Sint Maarten (Dutch
Moldova; Montenegro; Poland; Romania; part); Spain; St. Martin (French part); Swe-
Russian Federation; Serbia; Slovak Republic; den; Switzerland; Taiwan, China; Turks and
Slovenia; Tajikistan; Turkey; Turkmenistan; Caicos Islands; United Arab Emirates; United
Ukraine; Uzbekistan. Kingdom; United States; U.S. Virgin Islands.
Latin America and the Caribbean:
Argentina; Barbados; Belize; Bolivia; Brazil; By fragile and conflict-affected
Chile; Colombia; Costa Rica; Cuba; Domi-
situation
nica; Dominican Republic; Ecuador; El Sal-
vador; Grenada; Guatemala; Guyana; Haiti; Economies with fragile situations are primar-
Honduras; Jamaica; Mexico; Nicaragua; ily International Development Association–
Panama; Paraguay; Peru; St. Kitts and Nevis; eligible countries and nonmember or in-
St. Lucia; St. Vincent and the Grenadines; active countries and territories with a 3.2
Suriname; Trinidad and Tobago; Uruguay; or lower harmonized average of the World
República Bolivariana de Venezuela. Bank’s Country Policy and Institutional As-
Middle East and North Africa: Algeria; sessment (CPIA) rating and the correspond-
Djibouti; Arab Republic of Egypt; Islamic Re- ing rating by a regional development bank,
public of Iran; Iraq; Jordan; Lebanon; Libya; or with a United Nations or regional peace-
Morocco; Oman; Syrian Arab Republic; Tuni- building and political mission (for example
sia; West Bank and Gaza; Republic of Yemen. by the African Union, European Union, or
South Asia: Afghanistan; Bangladesh; Organization of American States) or peace-
Bhutan; India; Maldives; Nepal; Pakistan; Sri keeping mission (for example, by the Afri-
Lanka. can Union, European Union, North Atlan-
Sub-Saharan Africa: Angola; Benin; tic Treaty Organization, or Organization of
Botswana; Burkina Faso; Burundi; Cabo American States) during the last three years.
Verde; Cameroon; Central African Repub- The group excludes World Bank countries
lic; Chad; Comoros; Democratic Republic (for which the CPIA scores are not publicly
of Congo; Republic of Congo; Côte d’Ivoire; disclosed) unless they have a peacekeeping or
Equatorial Guinea; Eritrea; Eswatini; Ethi- political/peacebuilding mission. This defini-
opia; Gabon; The Gambia; Ghana; Guinea; tion is pursuant to an agreement between the
Guinea-Bissau; Kenya; Lesotho; Liberia; World Bank and other multilateral develop-
APPENDIX A: DATA DETAILS 153
ment banks at the start of the International putation estimates for India are not counted
Development Association 15 round in 2007. toward the 40 percent, which means the South
The World Bank releases annually the Asia coverage for 2015 is below the threshold.
Harmonized List of Fragile Situations. The The recent availability of additional survey
first such list was compiled in fiscal 2006 data has filled gaps in the regional poverty
and has gone through a series of changes trend for the Middle East and North Africa.
in terms of classification from the Low- In the 2016 edition of the Poverty and Shared
Income Countries Under Stress List (2006– Prosperity Report, the estimates for the Mid-
09), to the Fragile States List (2010), to the dle East and North Africa region were not re-
current Harmonized List of Fragile Situa- ported for 1999, 2002, and after 2008 because
tions (2011–15). The concept and the list of low population coverage of the data. In
have evolved as the World Bank’s under- the current edition, regional estimates for the
standing of the development challenges in Middle East and North Africa are reported
countries affected by violence and instabil- for all years.
ity has matured. The lists of economies by
year are available at http://www.worldbank India
.org/en/topic/fragilityconflictviolence/brief
/harmonized-list-of-fragile-situations. Although the most recent round of National
Sample Survey (NSS) data that the Govern-
ment of India uses for poverty estimation
Chapter 1 data and was collected in 2011–12, a subsequent round
methodology of the NSS was collected in 2014–15. This
The World Bank now reports global and re- more recent survey collects socioeconomic
gional poverty estimates every two years, co- and demographic information similar to the
inciding with the publication of the Poverty 2011–12 NSS and earlier NSS rounds. But
and Shared Prosperity report. Up until 2008, the 2014–15 NSS cannot be used for direct
the frequency of the global estimates was poverty estimation because the consumption
every three years. Because new surveys be- data on only a small subset of items have been
come available and existing survey and aux- released. Given the importance of India to the
iliary data are sometimes updated, the global global poverty count, and the unique situa-
and regional estimates are revised regularly. tion of having common socioeconomic and
The 2018 edition of global poverty esti- demographic data in the 2014–15 NSS (and
mates is based on the most recent data avail- found in earlier NSSs), a model of consump-
able. This section explains notable changes tion has been estimated on the basis of the
since the 2016 edition of global poverty esti- common socioeconomic, demographic, and
mates, discusses some key measurement is- geographic characteristics of the population
sues, and describes the auxiliary data, includ- (Newhouse and Vyas 2018). This allows for
ing purchasing power parity (PPP) conversion an estimate of poverty at the IPL for India in
factors, consumer price indexes (CPIs), popu- 2014–15, which is then lined up to 2015 and
lation data, and national accounts data. used as the poverty estimate for India in chap-
ter 1 (for details on the lineup method, see the
section “Estimating global and regional pov-
Household survey data for
erty: The ‘lineup,’” below). For further details
poverty monitoring on the consumption model for India, see box
Poverty rates for a region are marked with a 1.3 in chapter 1.
note if the available household surveys cover
less than 40 percent of the population in the Auxiliary data: PPP, CPI,
region. The criterion for estimating survey
population, and national accounts
population coverage is whether at least one
survey used in the reference year estimate was PPP conversion factors. The poverty esti-
conducted within two years of the reference mates for all countries are based on con-
year. For the purpose of this chapter, the im- sumption PPPs from the 2011 round of data
154 POVERTY AND SHARED PROSPERITY 2018
collection coordinated by the International The CPI, population, and national ac-
Comparison Program. The PPP conversion counts data used for the latest global esti-
factors include benchmark countries where mates are available on the PovcalNet site
actual price surveys were conducted, and (http://iresearch.worldbank.org/PovcalNet
regression-based PPP estimates where such /Data.aspx). For additional details on recent
surveys were not conducted or not appro- changes and data updates, see the What’s
priate for poverty measurement. Since the New notes of the Global Poverty Monitoring
2016 edition of the Poverty and Shared Pros- Technical Notes (http://iresearch.worldbank
perity Report, the 2011 PPP conversion fac- .org/PovcalNet/whatIsNew.aspx).
tors for Egypt, Iraq, Jordan, Lao PDR, and
the Republic of Yemen have been revised Estimating global and regional
(Atamanov, Jolliffe, and Prydz 2018). poverty: The “lineup”
CPI. The primary source of CPI data used
for global poverty measurement is the In- Because the household surveys necessary to
ternational Monetary Fund’s International measure poverty are conducted in different
Finance Statistics (IFS) monthly series. Pre- years and at varying frequencies across coun-
viously, the World Development Indica- tries, producing global and regional poverty
tors (WDI) annual series were used. When estimates entails bringing each of the country-
monthly IFS series are not available or not level poverty estimates to a common reference
appropriate for poverty monitoring, other or “lineup” year. For countries with surveys
sources are used. China and India use rural available in the reference year, the direct es-
and urban CPIs provided by the national sta- timates of poverty from the surveys are used.
tistical offices, six countries use national se- For other countries, the poverty estimates
ries provided by the national statistical offices are imputed for the reference year using the
(the Islamic Republic of Iran, Iraq, Kenya, country’s recent household survey data and
Maldives, Nicaragua, and República Boli- real growth rates from national accounts data.
variana de Venezuela), and five countries use The procedures for doing this depend on the
CPIs implied from the surveys (Bangladesh, survey years available for the country.
Ghana, Lao PDR, Malawi, and Tajikistan). A When a survey is available only prior to
more detailed description of CPIs used for the reference year, the consumption (or in-
global poverty monitoring is available in Lak- come) vector from the latest survey is extrap-
ner et al. (2018). olated forward to the reference year using real
Population. The primary source of pop- growth rates of per capita GDP (or HFCE)
ulation data is the December 2017 version of obtained from national accounts. Each ob-
the WDI. For additional details see Chen et al. servation in the welfare distribution is multi-
(2018). plied by the growth rate in per capita GDP
National accounts. The primary source (or HFCE) between the reference year and
of per capita gross domestic product (GDP) the time of the survey. Poverty measures
and household final consumption expendi- can then be estimated for the reference year.
ture (HFCE) data is the December 2017 ver- This procedure assumes distribution-neutral
sion of the WDI. Per capita GDP is used for growth—that is, no change in inequality—
countries in Sub-Saharan Africa and in coun- and that the growth in national accounts is
tries for which HFCE is not available. Every- fully transmitted to growth in household
where else, per capita HFCE is used. A more consumption or income. If the only available
detailed description of the national accounts surveys are after the reference year, a similar
data used for global poverty monitoring will approach is applied to extrapolate backward.
be available on the PovcalNet website. For When surveys are available both before
nowcasts, growth projections for recent years and after the reference year, information
are taken from the World Bank’s Global Eco- from both surveys is used to interpolate pov-
nomic Prospects, and from the International erty. In these cases, the welfare vectors (that
Monetary Fund’s World Economic Outlook, is, per capita consumption or income) from
when the former is unavailable. the two surveys are both lined up to the ref-
APPENDIX A: DATA DETAILS 155
erence year using growth rates of per capita consistent with a “truly global” approach to
GDP (or HFCE). After this, the poverty rate poverty measurement (World Bank 2017b,
is calculated for each of the two lined-up 47). The Commission therefore advised the
surveys and then averaged, with each point inclusion of all economies in the global pov-
weighted by the relative distance of the sur- erty measures. For further discussion, see
vey year to the reference year. The surveys Ferreira, Lakner, and Sanchez (2017).
are lined up to the reference year using two
different interpolation methods. The default Key poverty measurement
method is applied when the growth in the
issues
survey mean between the two surveys is of
the same sign as the real growth in per capita There are many technical details on how global
GDP (or HFCE) from the first survey to the poverty is measured. Ferreira et al. (2016) pro-
reference year, and from the reference year to vide a good overview of many of these issues,
the second survey. With this default method, particularly concerning the valuation of the
the growth in welfare from the time of the most recent IPL at US$1.90 in 2011 PPPs. For
survey to the reference year is proportional a more in-depth discussion of select measure-
to the relative growth in per capita GDP (or ment and data issues, see also Jolliffe et al.
HFCE) over the same period. The first step (2015). Two key measurement concerns are
entails imputing the survey mean at the refer- discussed below. These two areas are currently
ence year using the following formula: being examined, and potential methodss for
improvement are being considered.
, (A.1)
Consumption- and income-based
measures of well-being
where tr indicates the reference year, t1 indi-
National poverty rates are based on measures
cates the time of the first survey, t2 indicates
of consumption or income. Countries typ-
the time of the second survey (such that t2 >
ically choose the measure that can be more
tr > t1), and m indicates the survey mean at
accurately measured while balancing con-
the specified time. Upon computing mt r, each
cerns about respondent burden. On the one
element of the welfare vector from the first
hand, consumption measures of poverty re-
survey is grown or shrunk by the rate , quire a wide range of questions and are thus
more time consuming. Income measures, on
while each element of the welfare vector from
the other hand, are difficult to obtain when
the second survey is grown or shrunk by the
a large fraction of the population works in
the informal sector or is self-employed, and
rate
. The alternative method involves
income data are not collected for tax pur-
extrapolating the consumption vector to the poses. This is frequently the case in poorer
reference year for each of the two surveys countries, which therefore often opt for
using the real growth rates of per capita GDP using consumption (figure A.1). None of the
(or HFCE). The mechanics of the extrapola- low-income countries uses income, but this
tion and interpolation are described in more share increases to 10 percent, 40 percent, and
detail in box 6.4 in Jolliffe et al. (2015). 97 percent for lower-middle-, upper-middle-,
and high-income countries, respectively. As
living standards have improved, so has the
A truly global approach to
share of countries using income-based mea-
poverty measurement
sures of poverty, and it will likely continue to
All economies are now included in the global do so (figure A.1).
poverty estimates. Previously, the practice Both approaches to measuring poverty
was to assume that economies in the “rest of have advantages and disadvantages. The con-
the world” category have zero extreme pov- sumption approach is arguably more con-
erty. As pointed out in the Commission on nected to economic welfare. Whereas income
Global Poverty report, this assumption is in- is valuable because it allows individuals to
156 POVERTY AND SHARED PROSPERITY 2018
FIGURE A.1 Use of Income/Consumption to Measure Poverty
a. By income group, 2015
97.4 b. Over time
100
40
Percentage of economies using income
Percentage of economies using income
80
60 35
40.4
40
30
20
9.6
0
0 25
Low Lower- Upper- High
income middle middle income 1999 2002 2005 2008 2011 2014
income income
Source: PovcalNet, World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/.
purchase goods, consumption is valuable for rather than consumption. For a given poverty
its own sake. Income measures of poverty also rate, poor households also tend to be further
suffer from the disadvantage that incomes below the poverty line when income is used.
might be very low—even negative—in a given This is explained by the earlier point about
period. Negative incomes are often not an ac- very low incomes: whereas it is plausible that
curate depiction of the well-being of a house- households have a zero income in a given time
hold, so currently negative values are being period, subsistence requires a minimum level
discarded. This is particularly relevant for of consumption, which is strictly above zero.
self-employed individuals who tend to experi- The differences also matter for nowcasting
ence large income shocks at greater frequen- and making poverty projections for the fu-
cies. At a theoretical level, consumption will ture. Typically, such projections are made by
likely be smoothed to safeguard against such assuming a fixed growth rate of household
shocks, preventing consumption-based mea- consumption/income over time. If some
sures of poverty from being as vulnerable to households have zero income or a negative
large shocks as income-based measures. A income, then, regardless of how large growth
household that has managed to save sufficient rates are assumed to be, those households will
resources may not suffer greatly from a nega- never be projected to move out of poverty.
tive income shock. Consumption-based mea-
sures of poverty, conversely, are often more Accounting for spatial price
time demanding, require detailed price data, differences across and within
and often post fieldwork adjustments, such as countries
rent imputations, which can matter greatly Welfare is measured by aggregating a house-
for the final poverty estimates. Income mea- hold’s total value of consumption or total
sures need not rely on more than a handful of income over a defined time period and then
questions and can, at times, be verified from dividing by household size. When converted at
other sources. market exchange rates, US$100 can buy differ-
The differences between income and con- ent quantities and qualities of goods and ser-
sumption measures matter for comparing vices in say Nigeria than in the United States.
trends and patterns in poverty. Given that in- When comparing poverty rates across coun-
comes can be very low and negative, poverty tries, local currencies are converted to PPP
rates are typically higher when income is used dollars to account for differences in the pur-
APPENDIX A: DATA DETAILS 157
chasing power across countries, ensuring that well-being relative to other countries at the
a dollar can purchase approximately the same same nominal level of average consumption.
bundle of goods and services across countries. Much work is yet to be done to assure that
Important differences in price levels also similar practices are applied in various coun-
appear within countries. Suppose a house- tries. Ferreira et al. (2016) contains more in-
hold pays $1.00 for a kilo of rice in an urban formation on the methods applied in differ-
center, whereas a rural household in the same ent countries.
country pays only $0.50 for a similar quality
and amount of rice. Assume more generally Data for global and regional
that prices for all goods are twice as high in
poverty profiles
urban areas. If both households consumed
the same quantity of goods, and if one were The global poverty profile for 2015 in chapter
to assess poverty on the basis of the self- 1 is an update of the global profile of the poor
reported value of goods and services con- first reported in Castaneda et al. (2016) for
sumed without accounting for these price 2013. The methodological details of poverty
differences, one would conclude that the rural profiling are presented in the original paper.
household in this scenario is poorer than the The current exercise uses the 2018 vintage of
urban household. From a welfare perspective, the GMD, covering 91 economies and more
however, both households are consuming than 5.6 billion people, and lines up the
the same items and are at approximately the survey-based poverty estimates to 2015. The
same level of well-being. To properly com- exercise also uses recent population projec-
pare the welfare levels of the two households, tions from the United Nations Department
one would need to account for the differences of Economic and Social Affairs to adjust (that
in price levels that the two households face. is, post-stratify) the sampling weights to the
This example highlights the importance “lineup” year.
of spatial price adjustments within countries. For the Sub-Saharan Africa regional pov-
If certain households are deemed poorer erty profile, the analysis of demographic
solely because they face different price lev- characteristics presented in this section
els, then policy responses to poverty within builds on the harmonized 24-country data
countries may be misinformed. Because price from the book Poverty in a Rising Africa. The
differences can vary greatly within a country, book examines the trends in poverty and in-
accounting for regional price differences can equality in Sub-Saharan Africa using com-
have vast implications for subnational pro- parable surveys (Beegle et al. 2016). Of the
files of poverty, allocation of resources, and 148 surveys conducted in 48 Sub-Saharan
the design of poverty reduction strategies. As African countries between 1990 and 2012,
national poverty is falling in many countries two or more surveys were comparable in only
around the world, it is becoming increasingly 27 of 48 countries, and the data were avail-
important to correctly identify the remaining able for 24 of the 27 countries. The current
areas where poverty reduction lags. Without analysis adds Burundi (2006 and 2013) and
spatial price adjustments, a national poverty Seychelles (2006 and 2013); uses more re-
line could overestimate poverty in areas with cent data for Cameroon (2014), Côte d’Ivoire
low prices, typically rural areas, and underes- (2015), Madagascar (2012), Rwanda (2013),
timate poverty in areas with high prices, typ- and Togo (2015); and drops Mauritius, re-
ically urban areas. sulting in a 25-country sample with a slightly
Current measurement practices comprise different compostion. For the set of countries
a wide range of methods to account for dif- and surveys included in the present analysis,
ferences in the cost of living across regions, or the median year for the base period is 2004
across rural and urban areas. Some countries and the median year for the terminal period
peg prices to the price level of the capital re- is 2011. The countries represent 73 percent
gion, or a large city. With this approach, the of the total population of Sub-Saharan Af-
mean of the spatially adjusted welfare aggre- rica in 2015, and the average poverty rates
gate is larger than the mean without adjust- for the two periods are 59.7 and 47.7 per-
ments, essentially inflating the overall level of cent, respectively. These figures are different
158 POVERTY AND SHARED PROSPERITY 2018
from but close to the poverty rates for Sub- Because of the low prevalence of refugees
Saharan Africa around the same time—56.9 in general and their concentration in dense
percent in 2002 and 44.9 percent in 2011 from geographical pockets, it might be difficult to
PovcalNet. The discrepancy arises because draw a nationally representative sample using
PovcalNet includes a wider range of surveys. conventional sampling methods. Refugees
and internally displaced persons are highly
Missing data on forcibly mobile, especially when the crisis is unfold-
ing, which complicates the survey effort.
displaced persons
Even when the displaced households can be
Worldwide, it is estimated that there are located, the nonresponse rate might be high
nearly 70 million people in 2017 who have because of their wariness of divulging per-
been forcibly displaced because of persecu- sonal information. The problem with non-
tion, conflict, and generalized violence. Over response can become more severe when the
the last 10 years, the number of forcibly dis- survey needs to interview vulnerable popula-
placed persons has increased by more than 50 tions like women (for example, for birth his-
percent (UNHCR 2018). As the number of tory) and children (for example, for anthro-
forcibly displaced persons—refugees, asylum pometric measures).
seekers, and internally displaced persons— In sum, socioeconomic surveys on dis-
continues to increase, it becomes essential to placed persons are marked with incomplete
measure their welfare for an accurate moni- coverage, unrepresentative samples, and pos-
toring of global poverty. However, there are sibly larger-than-usual sampling and non-
many challenges in monitoring the welfare of sampling errors, which results in an under-
the displaced persons. Many countries do not estimate of the level of global poverty and an
count refugees as part of the usual resident undercount of the number of poor. To im-
population in the population census, and prove the ability to get a complete picture of
the census enumeration often excludes refu- the poverty situation in the world, and to un-
gee camps and temporary reception centers derstand how policy can affect the well-being
where refugees are housed. The exclusion of of displaced persons, a first step is to ensure
refugees from the population census implies that they are included in population censuses
they are not a part of the sampling frame and the national sample surveys of the coun-
used in household surveys. Similarly, typical try of their residence.
sample designs for household surveys used
for poverty measurement explicitly exclude Chapter 2 data and
people living in institutions or camps and
without an address.
methodology
Administrative registration databases
Welfare aggregate
maintained by government agencies or inter-
national organizations like the United Nations The mean of the bottom 40 within each
High Commissioner for Refugees are not well country refers to the average household per
integrated into the data systems of national capita consumption or income among this
statistical offices throughout the world, nor segment of the population. The choice of
do these data correspond well with definitions consumption or income depends on the data
in household surveys. For example, the unit available for each economy, and in most cases
of record in administrative databases is typi- is consistent with the welfare aggregate used
cally a case (for example, border crossing that to measure poverty (see annex 2B, table 2B.1).
can occur multiple times for an individual) or For China, shared prosperity is estimated
application, which does not match the defini- by PovcalNet using grouped data. Because
tion of a household, the unit of analysis for grouped data are provided separately for
sample surveys. This difference makes admin- urban and rural populations, the bottom 40
istrative databases challenging to use as sam- percent of the national population must be
pling frames of the population of displaced estimated. The bottom 40 are identified using
persons (Expert Group on Refugee and Inter- the national poverty gap and choosing a pov-
nally Displaced Persons Statistics 2018). erty line that corresponds to the threshold
APPENDIX A: DATA DETAILS 159
consumption level of the national bottom 40 prosperity estimate because of stricter data
percent. PovcalNet uses a parametric Lorenz requirements. Economies are included in the
curve fitted to grouped data, with an adjust- fall 2018 edition of the GDSP if the following
ment for differences in price levels between requirements are met:
urban and rural areas, and urban–rural pop-
• Two relevant household surveys have been
ulation shares from the WDI. Because shared
conducted and have yielded comparable
prosperity is estimated using grouped data
data.
for China, it is approximate and may differ
from using official microdata (see Chen et al. • Among comparable surveys, one must be
2018 for details). conducted within two years of 2010, and
In countries in Europe and Central Asia the other within two years of 2015. For
using household per capita income as the example, the Solomon Islands cannot be
welfare aggregate, households with nega- included because, although two rounds
tive incomes are included when estimating of a comparable household survey have
shared prosperity. been conducted (in 2005 and 2013), 2005
is more than two years from 2010. China is
an exception to this rule because a survey
Surveys used to calculate shared
break between 2012 and 2013 means that
prosperity
surveys conducted around 2010 and 2015
Among the 164 economies with a poverty are not comparable. The shared prosperity
estimate, significantly fewer have a shared period used for China is 2013–15.
• The period between the selected initial and
end years should range between three and
FIGURE A.2 Shared Prosperity Indicators Are Less Likely in seven years. For example, a shared pros-
Economies at Lower GDP per Capita perity period of 2012–13 meets the second
80 selection criterion but would not be al-
lowed because it does not meet this third
requirement.
• In cases where multiple surveys can fulfill
60
these criteria, the most recent survey years
Extreme poverty rate (%)
are typically chosen.
40 Factors affecting the inclusion of
economies in the GDSP
The computation of the shared prosperity
20 measure relies on frequent data collection,
which may depend on the capacity of a na-
tional statistics office—often related to the
level of development of a country. Among the
0 107 economies with a poverty rate below 10
percent in 2015 measured by the IPL, 78 also
0
0
0
00
00
0
00
00
00
00
00
10
20
50
00
1,0
5, 0
,0
,0
,0
,0
0,0
have a shared prosperity estimate for 2010–15
2,
10
20
50
0
20
10
GDP per capita (logarithmic scale) (figure A.2). Meanwhile, among 57 economies
Economies with a poverty and shared prosperity indicator with a poverty rate at more than 10 percent,
Economies with a poverty indicator but no shared prosperity indicator only 13 have a shared prosperity indicator.
Population coverage is also limited among
Sources: GDSP (Global Database of Shared Prosperity), fall 2018, World Bank, Washington, DC, http://
www.worldbank.org/en/topic/poverty/brief/global-database-of-shared-prosperity; PovcalNet (online economies grouped by other World Bank
analysis tool), World Bank, Washington, DC, http://iresearch.worldbank.org/PovcalNet/; WDI (World country categories, such as vulnerable, poor,
Development Indicators) (database), World Bank, Washington, DC, http://data.worldbank.org/products or small nations. For example, a shared pros-
/wdi.
Note: Based on data on 164 economies in PovcalNet associated with direct estimates of poverty. perity measure is not available on any of the
Poverty rates are based on the PovcalNet 2015 lineup. 15 small island nations.
160 POVERTY AND SHARED PROSPERITY 2018
Comparison of shared prosperity omies with data updates were mainly in Eu-
across rounds of GDSP rope and Central Asia, Latin America and the
Caribbean, and other high-income countries
Comparing the performance in shared pros- (the rest of the world). Therefore, only in
perity across rounds has limitations. The cur- these regions can trends in shared prosperity
rent release of the GDSP includes 91 econo- be reliably examined. At the other extreme,
mies. Since the circa 2008–13 GDSP used in new household survey data in the Middle
Poverty and Shared Prosperity 2016 (World East and North Africa and in Sub-Saharan
Bank 2016b), 19 countries have been added, Africa are scarcer, and shared prosperity esti-
and 10 countries removed because they no mates were updated in only one economy per
longer fulfill the data requirements (table region following the publication of Poverty
A.2). Of the 72 economies occurring in both and Shared Prosperity 2016.
rounds, the shared prosperity measure has
not been updated in five—Mexico, Mon-
tenegro, Nicaragua, Rwanda, and Togo— Chapter 3 data and
because no new surveys have become avail- methodology
able or, in the case of Mexico, because of a
break in the survey series. A comparison of Poverty rates at higher poverty
shared prosperity indicators can be carried lines
out in 67 economies across rounds.
The country sample changed across the The poverty estimates at the higher poverty
releases of the GDSP for two main reasons: lines presented in chapter 3 are extracted from
PovcalNet. See the discussion in the chapter 1
1. Data requirements were met in one round
section of this appendix for details on house-
but not in the next because appropriate
hold surveys, auxiliary data, and measure-
data within the established time frame
ment issues. For India, the poverty estimates
were not available or because of a lack of
are extrapolated using 2011–12 survey data
data comparability. Between circa 2008–
and the pass-through rate described in box 1.3
13 and circa 2010–15, 10 countries were
in chapter 1. Poverty rates at the societal pov-
removed for these reasons.
erty line are also estimated from PovcalNet.
2. Some countries that did not previously
meet data requirements do so now. Be-
tween circa 2008–13 and circa 2010–15, 19 Database of harmonized
countries were added for this reason. This national poverty lines
occurs when countries collect new house- A database of harmonized national poverty
hold surveys, following a long gap. lines is used to derive the societal poverty
Despite these challenges, the fall 2018 line presented in chapter 3. Jolliffe and Prydz
GDSP contains updated values of shared (2016) construct a set of national poverty
prosperity for three-quarters of the sample lines by combining national poverty rates
(67 economies) used in Poverty and Shared from national sources, reported in the World
Prosperity 2016 (World Bank 2016b). Econ- Bank’s databases, with corresponding con-
sumption and income distributions from
PovcalNet used for international poverty
TABLE A.2 Shared Prosperity Availability estimates. Because the consumption and in-
across Rounds come distributions used are all expressed in
Number of per capita PPP terms, the estimated national
GDSP round economies poverty lines are all expressed in comparable
Circa 2008–13 82 per capita PPP dollars. The national poverty
Removed 10 lines are harmonized in terms of the unit
Added 19
of measure in the sense that they are all ex-
Circa 2010–15 91
pressed in per capita terms.
Circa 2008–13 and circa 2010–15 72 Following this approach, rather than col-
With updated shared prosperity measure 67
lecting publicly reported poverty lines, al-
APPENDIX A: DATA DETAILS 161
lows for a substantial increase of the set of indicators and maintaining cross-country
countries for which we have national poverty comparability.
thresholds. This approach also results in a se-
Most of the surveys used in the analysis
ries of historic and current poverty lines that
were conducted during 2012–14 (88 coun-
allows one to subset on a specific year cor-
tries). No household income and expenditure
responding to the most recent International
survey data were available for the populous
Comparison Program reference year (for ex-
African countries of Nigeria and Sudan in
ample, 2011).
the 2010–16 period, which explains the low
Subsetting on national poverty lines
regional population coverage in Sub-Saharan
closest to 2011 both provides recent socio-
Africa (see table 4.4). The population cover-
economic assessments of basic needs and
age for the rest of the world category is small
reduces the reliance on CPI data for lining
because of limited coverage in the GMD. Be-
up the poverty lines to a common year. The
cause of the selection criteria above, the set of
larger database contains 864 harmonized
countries differs from that in chapter 1.
national poverty lines. The analysis of the
circa-2011 national poverty lines for the
lower-middle-income and upper-middle- Differences from chapter 1
income country lines is based on a subsample poverty estimates
of 126 lines; and the estimation of the soci- The extreme poverty rates (headcount ratios)
etal poverty line, discussed in this chapter, reported in this chapter cannot be compared
is based on a subsample of 699 harmonized to the information presented in chapter 1 for
national poverty lines. For more details on three practical and methodological reasons.
the construction of the database of harmo- First, if a survey was available for a country
nized national poverty lines, see Jolliffe and in both 2013 and 2015, the 2013 data are used
Prydz (2016); and for discussion of the data in this chapter to minimize the overall dis-
underlying the estimation of the societal pov- persion in survey years. Second, to examine
erty line, see Jolliffe and Prydz (2017). For a the simultaneous incidence of deprivations,
discussion of the precision of these harmo- only unit-record data are used in this chapter,
nized lines, see the online appendix to their which limits the number of countries consid-
paper at https://static-content.springer.com ered. In contrast, grouped data also enter into
/esm/art%3A10.1007%2Fs10888-016-9327-5 the estimation of the global poverty rate re-
/MediaObjects/10888_2016_9327_MOESM1_ ported in chapter 1 if unit-record data are un-
ESM.pdf. available. China is a notable example where
only grouped data are available. This explains
Chapter 4 data and the low population coverage of the East Asia
methodology and Pacific region in this chapter. Third,
PovcalNet relies on recent surveys to impute
Chapter 4 uses data from the harmonized the headcount ratio for the lineup year, 2015,
household surveys from the 2017 edition of assuming distribution-neutral growth. These
the GMD. Surveys have been included in the adjustments are not made in this chapter be-
multidimensional poverty analysis if they cause the lineup process cannot be applied to
satisfy the following criteria: the other indicators of well-being. A full list
• They include a monetary welfare measure of the countries for which different surveys
(consumption or income) and indicators are used in chapter 1 (for the 2015 estimates)
on education and service access that may and chapter 4 is included in table A.3.
be used to construct a multidimensional
poverty measure. Six-country sample
• The surveys were conducted within three The extended multidimensional analyses cov-
years of 2013, that is, from 2010 to 2016. ering five dimensions of poverty are based on
The circa 2013 restriction strikes a balance the household surveys for the six countries in
between maximizing country coverage of table A.4. Except for Iraq, the surveys are not
162 POVERTY AND SHARED PROSPERITY 2018
TABLE A.3 Surveys Used in Chapter 1 and Chapter 4 in Cases Where Different Survey Rounds
Are Used
Survey(s) used in chapter 1
Economy Survey used in chapter 4 for extreme poverty
Argentina EPHC 2014 EPHC 2014 and EPHC 2016
Armenia ILCS 2013 ILCS 2015
Austria EU-SILC 2014 EU-SILC 2016
Bangladesh HIES 2010 HIES 2010 and HIES 2016
Belarus HHS 2013 HHS 2015
Belgium EU-SILC 2014 EU-SILC 2016
Bhutan BLSS 2012 BLSS 2012 and BLSS 2017
Bolivia EH 2014 EH 2015
Brazil PNAD 2014 PNAD 2015
Chile CASEN 2013 CASEN 2015
Colombia GEIH 2014 GEIH 2015
Costa Rica ENAHO 2014 ENAHO 2015
Croatia EU-SILC 2014 EU-SILC 2016
Cyprus EU-SILC 2014 EU-SILC 2016
Czech Republic EU-SILC 2014 EU-SILC 2016
Denmark EU-SILC 2014 EU-SILC 2016
Dominican Republic ENFT 2013 ENFT 2015
Ecuador ENEMDU 2014 ENEMDU 2015
Egypt, Arab Rep. HIECS 2012 HIECS 2015
El Salvador EHPM 2014 EHPM 2015
Estonia EU-SILC 2014 EU-SILC 2016
Ethiopia HICES 2010 HICES 2010 & HICES 2015
Finland EU-SILC 2014 EU-SILC 2016
France EU-SILC 2014 EU-SILC 2016
Gambia, The IHS 2010 IHS 2010 and IHS 2015
Georgia HIS 2013 HIS 2015
Germany EU-SILC 2012 EU-SILC 2016
Greece EU-SILC 2014 EU-SILC 2016
Honduras EPHPM 2013 EPHPM 2015
Hungary EU-SILC 2014 EU-SILC 2016
Indonesia SUSENAS 2016 SUSENAS 2015
Iran, Islamic Rep. HEIS 2013 HEIS 2014
Ireland EU-SILC 2014 EU-SILC 2016
Italy EU-SILC 2014 EU-SILC 2016
Kazakhstan HBS 2013 HBS 2015
Kosovo HBS 2013 HBS 2015
Kyrgyz Republic KIHS 2013 KIHS 2015
Latvia EU-SILC 2014 EU-SILC 2016
Lithuania EU-SILC 2014 EU-SILC 2016
Luxembourg EU-SILC 2014 EU-SILC 2016
Malta EU-SILC 2014 EU-SILC 2016
Mexico ENIGH 2012 ENIGH 2014 and ENIGH 2016
Moldova HBS 2013 HBS 2015
Mongolia HSES 2016 HSES 2014 and HSES 2016
Montenegro HBS 2013 HBS 2014
Netherlands EU-SILC 2014 EU-SILC 2016
Norway EU-SILC 2014 EU-SILC 2016
Pakistan PSLM 2013 PSLM 2013 and PSLM 2015
Paraguay EPH 2014 EPH 2015
Peru ENAHO 2014 ENAHO 2015
Portugal EU-SILC 2014 EU-SILC 2016
Romania HBS 2013 EU-SILC 2016
Russian Federation HBS 2013 HBS 2015
Serbia HBS 2013 HBS 2015
Slovak Republic EU-SILC 2014 EU-SILC 2016
Slovenia EU-SILC 2014 EU-SILC 2016
(continued)
APPENDIX A: DATA DETAILS 163
TABLE A.3 Surveys Used in Chapter 1 and Chapter 4 in Cases Where Different Survey Rounds
Are Used (continued)
Survey(s) used in chapter 1
Economy Survey used in chapter 4 for extreme poverty
Spain EU-SILC 2014 EU-SILC 2016
Sri Lanka HIES 2016 HIES 2012 and HIES 2016
Sweden EU-SILC 2014 EU-SILC 2016
Switzerland EU-SILC 2014 EU-SILC 2016
Thailand SES 2013 SES 2015
Turkey HICES 2013 HICES 2015
Uganda UNHS 2012 UNHS 2012 and UNHS 2016
Ukraine HLCS 2013 HLCS 2015
United Kingdom EU-SILC 2014 EU-SILC 2016
Uruguay ECH 2014 ECH 2015
Vietnam VHLSS 2014 VHLSS 2014 and VHLSS 2016
West Bank and Gaza PECS 2011 PECS 2011 and PECS 2016
Source: GMD (Global Monitoring Database), Global Solution Group on Welfare Measurement and Capacity Building, Poverty and Equity
Global Practice, World Bank, Washington, DC.
Note: Only economies where different survey rounds are used for chapter 4 and the 2015 poverty estimates of chapter 1 are listed. For
economies where EU-SILC is used, the income data is from the year prior to the survey. For example, the EU-SILC 2016 survey uses data
from 2015. Romania is the only economy where both the survey year and the type of survey differ from chapter 1 to chapter 4.
the same surveys used for official national applicable, and the deprivation in the edu-
poverty estimates. Therefore, the monetary cation dimension is measured solely using
poverty headcount ratios cited in this section the adult school attainment indicator.
may vary from official estimates.
• Adult school attainment. Individuals are
considered deprived if no adult (at or
above the age one is normally at when
Definitions of indicators
attending the ninth grade) in the house-
Monetary poverty hold has completed primary education.
• Income per capita. A person is considered
Access to basic infrastructure
deprived if the household consumption or
income per person per day falls below the • Electricity. A person is considered de-
IPL, currently set at US$1.90 in 2011 PPPs. prived if the household has no access to
electrification from any source, that is, grid
Education electricity or self-generation.
• Child school enrollment. Individuals • Limited-standard drinking water. A per-
are considered deprived if they live in a son is considered deprived if the household
household in which at least one school- has no access to even a limited standard of
aged child up to the age of grade 8 is not drinking water. For a selection of coun-
enrolled in school. If a household has no tries, a variation closer to the Sustainable
child up to this age, this indicator is not Development Goals’ safely managed drink-
ing water concept is available: a house-
TABLE A.4 Household Surveys, Six-Country Sample hold is considered deprived if it has no
Country Year Survey access to basic drinking water (a limited-
standard source that is within a round-trip
Ecuador 2013–14 Encuesta de Condiciones de Vida
Indonesia 2014 Indonesian Family Life Survey
time of 30 minutes). For more informa-
Iraq 2012 Iraq Household Socio-Economic Survey
tion, see https://washdata.org/monitoring.
Mexico 2009–12 Mexican Family Life Survey • Limited-standard sanitation. A person is
Tanzania 2012–13 National Panel Survey considered deprived if the household has no
Uganda 2013–14 Uganda National Panel Survey access to even a limited standard of sanita-
164 POVERTY AND SHARED PROSPERITY 2018
tion facilities, that is, a sanitation facility that cable households who actually experienced
hygienically separates excreta from human a recent birth or have a child younger than 6
contact. For a selection of countries, exclu- years.
sivity of the facility is also taken into con-
sideration. In those countries, a household Security
is considered deprived if it lacks a limited-
• Incidence of crime. A person is considered
standard facility that is used only by mem-
deprived if anyone in the household has
bers of the same household. The addition
experienced crime in the previous year or
of this criterion to “limited” is called “basic-
lives in a neighborhood where at least 20
standard” sanitation. For more information,
percent of households contain at least one
see https://washdata.org/monitoring.
individual who experienced crime in the
previous year.
Health and nutrition
• Incidence of natural disaster. Individuals
• Birth delivery. A person is considered de-
are considered deprived if their household
prived if any woman in the household be-
has experienced a severe shock (a loss of
tween the ages of 15 to 49 has given birth
income, property, or livestock) because of
(live) in the previous 36 months, and the
drought, flooding, earthquake, or other
delivery did not occur in a formal facility.
natural disaster in the previous 12 months.
• Vaccination. A person is considered de-
prived if the household has any child be-
tween the ages of 12 to 59 months who has Chapter 5 data and
not received a third diphtheria-pertussis- methodology
tetanus vaccination.
This section uses the harmonized household
• Child stunting. A person is considered surveys from the 2016 release, circa 2013
deprived if the household has any child data, edition of the GMD. Even though GMD
between the ages of 0 to 59 months who data for circa 2013 was used for chapters 4
is stunted (the height-for-age Z-score is and 5, the set of countries covered differs
below −2, that is, more than two standard because different variables are required for
deviations below the reference population the analysis. The combined sample of the
median). data used in chapter 5 contains records rep-
resenting 5.2 billion individuals in 89 coun-
• Undernourishment. A person is consid-
tries, with estimates of poverty figures lined
ered deprived if any woman between the
up—that is, extrapolated—to 2013 and then
ages of 15 to 49 in the household is under-
updated to 2016. The data include welfare
nourished (her body mass index is below
aggregates based on a money metric, either
18.5 [underweight]).
household per capita consumption or in-
The measure of access to formal health come, depending on the concept used in
care is not applicable to all households be- each country (see chapter 1 discussion above
cause a significant share of households have for details). Nearly 83 percent of the sample
not experienced a birth in the previous three originates in middle-income countries. East
years or do not have a child younger than 5 Asia and Pacific and South Asia account for
years. For such households, access to health nearly two-thirds of the sample. The GMD
services is approximated by the share of indi- sample has a high regional coverage of de-
viduals in applicable households in the same veloping countries in East Asia and Pacific,
community who are observed to be deprived. South Asia, Latin America and the Carib-
The deprivation threshold for the rate of bean, and Europe and Central Asia (above
health service access is set such that the share 87 percent) and partial coverage of Sub-
of individuals in nonapplicable households Saharan Africa (74 percent). Additional
that are classified as deprived equals the na- labor data from the International Income
tional share of deprived individuals in appli- Distribution dataset were merged for 17
APPENDIX A: DATA DETAILS 165
TABLE A.5 Household Surveys for Case Studies and Sharing Rule Estimates
Country Survey Year(s)
Case studies
Bangladesh Bangladesh Integrated Household Survey 2011–12
China China Health and Nutrition Survey 1989, 1991, 1993, 1997, 2000, 2004, 2006, 2009
Senegal Poverty and Family Structure Survey 2006–07
Burundi Panel Priority Survey 2012
Sharing rule estimates
Bangladesh Bangladesh Integrated Household Survey 2011–12, 2015
Malawi Malawi Integrated Household Survey 2004–05, 2010–11
Tanzania National Panel Survey 2012–13, 2014–15
countries in Sub-Saharan Africa (Muñoz An individual perspective on
Boudet et al. 2018). Because of remaining multidimensional poverty
quality concerns in the economic participa-
tion variables, 18 countries were dropped for This section uses the same household surveys
the economic typology of households. Be- that were used in the six-country sample in
cause of low coverage in the Middle East and chapter 4 (see table A.4), except Uganda is
North Africa (4.1 percent), the results from excluded because the survey did not collect
this region are not presented. anthropometric information for adults.
Differences in resources and Note
poverty within households 1. GMD (Global Monitoring Database), Global
This section draws on the household surveys Solutions Group on Welfare Measurement and
in table A.5. Capacity Building, Poverty and Equity Global
Practice, World Bank, Washington, DC.
166 POVERTY AND SHARED PROSPERITY 2018
References
Aaberge, Rolf, and Andrea Brandolini. 2015. Alkire, Sabina, and Maria Emma Santos. 2010.
“Multidimensional Poverty and Inequality.” “Acute Multidimensional Poverty: A New
Handbook of Income Distribution (2). Edited Index for Developing Countries.” OPHI Work-
by Anthony B. Atkinson and Francois Bour- ing Papers 38, University of Oxford, Oxford.
guignon. Oxford: Elsevier. ———. 2014. “Measuring Acute Poverty in the
Aaberge, Rolf, and Anthony B. Atkinson. 2013. Developing World: Robustness and Scope of
“The Median as Watershed.” Discussion Papers, the Multidimensional Poverty Index.” World
Research Department, Statistics Norway, Oslo. Development 59: 251–74.
Abdu, Maryam, and Enrique Delamonica. 2017. Allen, Robert C. 2017. “Absolute Poverty: When
“Multidimensional Child Poverty: From Com- Necessity Displaces Desire.” American Eco-
plex Weighting to Simple Representation.” So- nomic Review 107 (12): 3690–721. https://doi
cial Indicators Research 136 (3): 881–905. .org/10.1257/aer.20161080.
Addison, Tony, and Miguel Niño-Zarazúa. 2012. Alvaredo, Facundo, Lucas Chancel, Thomas
“Redefining Poverty in China and India.” Piketty, Emmanuel Saez, and Gabriel Zuc-
United Nations University, Tokyo. man. 2018. World Inequality Report 2018.
Aguero, Jorge M., and Mindy S. Marks. 2008. Paris: World Inequality Lab, Paris School of
“Motherhood and Female Labor Force Par- Economics.
ticipation: Evidence from Infertility Shocks.” Amugsi, Dickson Abanimi, Maurice B. Mittel-
American Economic Review 98 (2): 500–04. mark, and Abraham Oduro. 2015. “Association
Aitsi-Selmi, Amina. 2015. “Households with a between Maternal and Child Dietary Diversity:
Stunted Child and Obese Mother: Trends and An Analysis of the Ghana Demographic and
Child Feeding Practices in a Middle-Income Health Survey.” PloS ONE 10 (8): e0136748.
Country, 1992–2008.” Maternal and Child Arimond, Mary, and Marie T. Ruel. 2004. “Dietary
Health Journal 19 (6): 1284–91. Diversity Is Associated with Child Nutritional
Alkire, Sabina, and James Foster. 2011. “Counting Status: Evidence from 11 Demographic and
and Multidimensional Poverty Measurement.” Health Surveys.” The Journal of Nutrition 134
Journal of Public Economics 95 (7): 476–87. (10): 2579–85.
———. 2016. “Dimensional and Distributional Arimond, Mary, Doris Wiesmann, Elodie Bec-
Contributions to Multidimensional Poverty.” quey, Alicia Carriquiry, Melissa C. Daniels,
OPHI Working Papers 100, University of Ox- Megan Deitchler, Nadia Fanou-Fogny, Maria L.
ford, Oxford. https://ophi.org.uk/dimensional Joseph, Gina Kennedy, Yves Martin-Prevel,
-and-distributional-contributions-to-multi and Liv Elin Torheim. 2010. “Simple Food
dimensional-poverty/. Group Diversity Indicators Predict Micronu-
Alkire, Sabina, José Manuel Roche, Paola Ballon, trient Adequacy of Women’s Diets in 5 Diverse,
James Foster, Maria Emma Santos, and Suman Resource-Poor Settings.” The Journal of Nutri-
Seth. 2015. Multidimensional Poverty Measure- tion 140 (11): 2059S–69S.
ment and Analysis. Oxford: Oxford University Atamanov, Aziz, Dean Jolliffe, and Espen Beer
Press. Prydz. 2018. “Predicting 2011 PPP Private
167
Consumption Factors for Problematic Coun- Bargain, Olivier, Guy Lacroix, and Luca Tiberti.
tries.” Global Poverty Monitoring Technical 2018. “Validating the Collective Model of
Note, World Bank, Washington, DC. Household Consumption Using Direct Evi-
Atkinson, Anthony B. 1987. “On the Measurement dence on Sharing.” IZA Discussion Paper
of Poverty.” Econometrica 55 (4): 749. https:// 11653 (July), Institute of Labor Economics,
doi.org/10.2307/1911028. Bonn, Germany.
———. 2003. “Multidimensional Deprivation: Barth, Erling, Alex Bryson, James C. Davis, and
Contrasting Social Welfare and Counting Ap- Richard Freeman. 2016. “It’s Where You Work:
proaches.” The Journal of Economic Inequality Increases in the Dispersion of Earnings Across
1 (1): 51–65. Establishments and Individuals in the United
Atkinson, Anthony B., and Francois Bourguignon. States.” Journal of Labor Economics 34(2):
2001. “Poverty and Inclusion from a World 67–97.
Perspective.” Governance, Equity, and Global Basu, Kaushik. 2006. “Gender and Say: A Model of
Markets. Edited by Joseph E. Stiglitz and Pierre- Household Behavior with Endogenously De-
Alain Muet. Oxford: Oxford University Press. termined Balance of Power.” Economic Journal
Atkinson, Anthony B., Bea Cantillon, Eric Marlier, 116 (511): 558–80.
and Brian Nolan. 2002. Social Indicators: The Beegle, Kathleen, Luc Christiaensen, Andrew
EU and Social Inclusion. Oxford: Oxford Uni- Dabalen, and Isis Gaddis. 2016. Poverty in a
versity Press. Rising Africa. Washington, DC: World Bank.
Aurino, Elisabetta. 2017. “Do Boys Eat Better Beegle, Kathleen, Joachim De Weerdt, Jed
than Girls in India? Longitudinal Evidence Friedman, and John Gibson. 2012. “Methods
on Dietary Diversity and Food Consumption of Household Consumption Measurement
Disparities among Children and Adolescents.” through Surveys: Experimental Results from
Economics & Human Biology 25: 99–111. Tanzania.” Journal of Development Economics
Autor, David, David Dorn, Lawrence F. Katz, Chris- 98 (1): 3–18.
tina Patterson, and John Van Reenen. 2017. Behrman, Jere R., and Anil B. Deolalikar. 1990.
“Concentrating on the Fall of the Labor Share.” “The Intrahousehold Demand for Nutrients in
American Economic Review 107 (5): 180–85. Rural South India: Individual Estimates, Fixed
Backiny-Yetna, Prospère, Diane Steele, and Ismael Effects, and Permanent Income.” Journal of
Yacoubou Djima. 2017. “The Impact of House- Human Resources 25 (4): 665–96.
hold Food Consumption Data Collection Belete, Getachew Yirga. 2018. “Children’s Re-
Methods on Poverty and Inequality Measures sources and Poverty in Single-Mother and
in Niger.” Food Policy 72: 7–19. Male-Headed Households: A Collective Con-
Bardasi, Elena, Kathleen Beegle, Andrew Dillon, sumption Evidence from Ethiopia.” Working
and Pieter Serneels. 2011. “Do Labor Statistics paper, Department of Economics, University
Depend on How and to Whom the Questions of Verona, Verona.
Are Asked? Results from a Survey Experiment Bhutan (Royal Government of Bhutan National
in Tanzania.” World Bank Economic Review 25 Statistics Bureau and Oxford Poverty and
(3): 418–47. Human Development Initiative). 2017. Bhutan
Bardasi, Elena, and Quentin Wodon. 2010. “Work- Multidimensional Poverty Index 2017. Bhutan.
ing Long Hours and Having No Choice: Time Birdsall, Nancy, José Edgardo Campos, Chang-
Poverty in Guinea.” Feminist Economics 16 (3): Shik Kim, W. Max Corden, Howard Pack, John
45–78. Page, and Joseph Stiglitz. 1993. The East Asian
Bargain, Olivier, Olivier Donni, and Prudence Miracle: Economic Growth and Public Policy.
Kwenda. 2014. “Intrahousehold Distribu- New York: Oxford University Press.
tion and Poverty: Evidence from Côte d’Ivo- Birdsall, Nancy, and Christian J. Meyer. 2015. “The
ire.” Journal of Development Economics 107 Median Is the Message: A Good Enough Mea-
(March): 262–76. sure of Material Wellbeing and Shared Devel-
Bargain, Olivier, Prudence Kwenda, and Miracle opment Progress.” Global Policy 6(4): 343–57.
Ntuli. 2017. “Gender Bias and the Intrahouse- Blackden, Mark C., and Quentin Wodon. 2006.
hold Distribution of Resources: Evidence from “Gender, Time Use, and Poverty in Sub-
African Nuclear Households in South Africa.” Saharan Africa.” Working Paper 73, World
Journal of African Economies 27 (2): 201–26. Bank, Washington, DC.
168 POVERTY AND SHARED PROSPERITY 2018
Bourguignon, François, and Pierre-André Chiap- Pedro Azevedo. 2016. “Who Are the Poor in the
pori. 1992. “Collective Models of Household Developing World?” Policy Research Working
Behavior: An Introduction.” European Eco- Paper 7844, World Bank, Washington, DC.
nomic Review 36 (2–3): 355–64. Chakravarty, Satya R., and Conchita D’Ambrosio.
Bradshaw, Sarah, Sylvia Chant, and Brian Lin- 2006. “The Measurement of Social Exclusion.”
neker. 2017. “Gender and Poverty: What We Review of Income and Wealth 52 (3): 377–98.
Know, Don’t Know, and Need to Know for Chandrasekhar, S., Víctor M. Aguayo, Vandana
Agenda 2030.” Gender, Place, and Culture 24 Krishna, and Rajlakshmi Nair. 2017. “House-
(12): 1667–88. hold Food Insecurity and Children’s Dietary
Brown, Caitlin, Rossella Calvi, and Jacob Penglase. Diversity and Nutrition in India. Evidence
2018. “Sharing the Pie: Undernutrition, Intra- from the Comprehensive Nutrition Survey in
Household Allocation, and Poverty.” http://dx Maharashtra.” Maternal & Child Nutrition 13:
.doi.org/10.2139/ssrn.3199129. e12447.
Brown, Caitlin, Martin Ravallion, and Dominique Chant, Sylvia. 2008. “The ‘Feminisation of Pov-
van de Walle. Forthcoming. “Are Poor Individ- erty’ and the ‘Feminisation’ of Anti-Poverty
uals Mainly Found in Poor Households? Evi- Programmes: Room for Revision?” Journal of
dence Using Nutrition Data for Africa.” Review Development Studies 44 (2): 165–97.
of Economics and Statistics. Chen, Shaohua, Dean Mitchell Jolliffe, Christoph
Browning, Martin, François Bourguignon, Lakner, Kihoon Lee, Daniel Gerszon Mahler,
Pierre-André Chiappori, and Valérie Lechene. Rose Mungai, Minh Cong Nguyen, Espen Beer
1994. “Income and Outcomes: A Structural Prydz, Prem Sangraula, Dhiraj Sharma, Judy
Model of Intrahousehold Allocation.” Journal Yang, and Qinghua Zhao. 2018. “September
of Political Economy 102 (6): 1067–96. 2018 Povcalnet Update: What’s New.” Global
Bühler, Dorothee, Rebecca Hartje, and Ulrike Poverty Monitoring Technical Note, no. 2. World
Grote. 2018. “Matching Food Security and Mal- Bank, Washington, DC.
nutrition Indicators: Evidence from Southeast Chen, Shaohua, and Martin Ravallion. 2013. “More
Asia.” Agricultural Economics 49 (4): 481–95. Relatively-Poor People in a Less Absolutely-
Bussolo, Maurizio, María E. Dávalos, Vito Peragine, Poor World?” Review of Income and Wealth
and Ramya Sundaram. 2018. Toward a New So- 59(1): 1–28.
cial Contract: Taking On Distributional Tensions Chiappori, Pierre-André. 1988. “Rational House-
in Europe and Central Asia. Europe and Central hold Labor Supply.” Econometrica 56 (1): 63–90.
Asia Studies. Washington, DC: World Bank. ———. 1992. “Collective Labor Supply and Wel-
Bussolo, Maurizio, and Luis F. López-Calva. 2014. fare.” Journal of Political Economy 100 (3):
Shared Prosperity: Paving the Way in Europe and 437–67.
Central Asia. Europe and Central Asia Studies. Chiappori, Pierre-André, and Costas Meghir.
Washington, DC: World Bank. 2015. “Intra-household Inequality.” Handbook
Calvo-González, Oscar, R. Andrés Castañeda, of Income Distribution 2B: 1369–1418.
Maria Gabriela Farfán, Germán Reyes, and Christiaensen, Luc, Punam Chuhan-Pole, and
Liliana D. Sousa. 2017. “How Is the Slowdown Aly Sanoh. 2013. “Africa’s Growth, Poverty
Affecting Households in Latin America and the and Inequality Nexus—Fostering Shared Pro-
Caribbean?” Policy Research Working Paper sperity.” Working paper, World Bank, Washing-
7948, World Bank, Washington, DC. ton, DC.
Canning, David, Sangeeta Raja, and Abdo S. Yaz- Coates, Jennifer, Beatrice Lorge Rogers, Alexander
beck. 2015. Africa’s Demographic Transition: Blau, Jacqueline Lauer, and Alemzewed Roba.
Dividend or Disaster? Washington, DC: World 2017. “Filling a Dietary Data Gap? Validation
Bank. of the Adult Male Equivalent Method of Es-
Case, Anne, and Angus Deaton. 2002. “Consump- timating Individual Nutrient Intakes from
tion, Health, Gender, and Poverty.” Working Household-level Data in Ethiopia and Bangla-
paper (July), Research Program in Develop- desh.” Food Policy 72: 27–42.
ment Studies, Princeton University, Prince- Commission on Growth and Development.
ton, NJ. 2008. The Growth Report: Strategies for Sus-
Castaneda, Andres, Dung Doan, David Newhouse, tained Growth and Inclusive Development.
Minh Cong Nguyen, Hiroki Uematsu, and Joao Washington, DC: World Bank.
REFERENCES 169
CONEVAL. 2010. “Methodology for Multidimen- Del Carpio, Ximena, and Pabon, Laura. 2014. Min-
sional Poverty Measurement in Mexico.” Con- imum Wage Policy: Lessons with a Focus on the
sejo Nacional de Evaluación de la Política de ASEAN Region. World Bank, Washington, DC.
Desarrollo Social, Mexico City. Djuikom, Marie Albertine, and Dominique van
Coulter, Fiona A. E., Frank A. Cowell, and Stephen de Walle. 2018. “Marital Shocks and Women’s
P. Jenkins. 1992. “Differences in Needs and Welfare in Africa.” Policy Research Working
Assessment of Income Distribution.” Bulletin Paper 8306, World Bank, Washington, DC.
of Economic Research 44 (2): 77–124. Doss, Cheryl R. 1996. “Testing among Models of
Cruces, Guillermo, and Sebastian Galiani. 2007. Intra-Household Allocation.” World Develop-
“Fertility and Female Labor Supply in Latin ment 24 (10): 1597–1609.
America: New Causal Evidence.” Labour Eco- Doss, Cheryl, Caitlin Kieran, and Talip Kilic. 2017.
nomics 14 (3): 565–73. “Measuring Ownership, Control, and Use of
Cuesta, José. 2006. “The Distributive Conse- Assets.” Policy Research Working Paper 8146,
quences of Machismo: A Simulation Analysis World Bank, Washington, DC.
of Intra-Household Discrimination.” Journal D’Souza, Anna E., and Sharad Alan Tandon. 2018.
of International Development 18 (8): 1065–80. “Intra-household Nutritional Inequities in
De Agostini, Paola, Alari Paulus, and Iva Valen- Rural Bangladesh.” Economic Development
tinova Tasseva. 2016. “The Effect of Changes
and Cultural Change. https://www.journals
in Tax-Benefit Policies on the Income Distri-
.uchicago.edu/doi/10.1086/698311.
bution in 2008–2015. ” EUROMOD Working
Dunbar, Geoffrey R., Arthur Lewbel, and Krishna
Papers EM6/16, EUROMOD at the Institute
Pendakur. 2013. “Children’s Resources in Col-
for Social and Economic Research.
lective Households: Identification, Estima-
Dammert, Ana C., and Jose Galdo. 2013. “Child
tion, and an Application to Child Poverty in
Labor Variation by Type of Respondent: Evi-
Malawi.” American Economic Review 103 (1):
dence from a Large-Scale Study.” World Devel-
438–71.
opment 53: 207–20.
Engel, Ernst. 1895. “Die Lebenskosten Belgischer
Datt, Gaurav. Forthcoming. “Distribution-
Arbeiter-Familien Früher und Jetzt.” Interna-
Sensitive Multidimensional Poverty Measures.”
tional Statistical Institute Bulletin 9: 1–74.
The World Bank Economic Review.
Expert Group on Refugee and Internally Displaced
De Vreyer, Philippe, and Sylvie Lambert. 2017. “By
Ignoring Intra-household Inequality Do We Persons Statistics. 2018. “International Recom-
Underestimate the Extent of Poverty?” Docu- mendations on Refugees Statistics.”
ment de Travail DT/2017–05, Unité Mixte de Fantom, Neil, and Serajuddin, Umar. 2016. “The
Recherche DIAL (Développement, Institutions World Bank’s Classification of Countries by
et Mondialisation), Institut de Recherche pour Income.” Policy Research Working Paper 7528,
le Développement, Université Paris-Dauphine, World Bank, Washington, DC.
Paris. Ferreira, Francisco H. G., Shaohua Chen, Andrew
De Vreyer, Philippe, Sylvie Lambert, Abla Safir, Dabalen, Yuri Dikhanov, Nada Hamadeh, Dean
and Momar Ballé Sylla. 2008. “Pauvreté et Jolliffe, Ambar Narayan, Espen Beer Prydz,
Structure Familiale: Pourquoi une nouvelle Ana Revenga, Prem Sangraula, Umar Sera-
enquête?” Statéco 102: 55–70. juddin, and Nobuo Yoshida. 2016. “A Global
Deaton, Angus. 1997. The Analysis of Household Count of the Extreme Poor in 2012: Data Is-
Surveys. A Microeconometric Approach to De- sues, Methodology and Initial Results.” The
velopment Policy. Baltimore, MD: John Hop- Journal of Economic Inequality 14(2): 141–72.
kins University Press. Ferreira, Francisco H. G., Christoph Lakner, and
Deaton, Angus, and Valerie Kozel. 2005. “Data and Carolina Sanchez. 2017. “The 2017 Global Pov-
Dogma: The Great Indian Poverty Debate.” The erty Update from the World Bank.” Let’s Talk
World Bank Research Observer 20 (2): 177–99. Development (blog), October 16. http://blogs
Decancq, Koen, and María Ana Lugo. 2013. .worldbank.org/developmenttalk/2017-global
“Weights in Multidimensional Indices of Well- -poverty-update-world-bank.
being: An Overview.” Econometric Reviews 32 Ferreira, Francisco H. G., and Maria Ana Lugo.
(1): 7–34. 2013. “Multidimensional Poverty Analysis:
170 POVERTY AND SHARED PROSPERITY 2018
Looking for a Middle Ground.” The World come and Expenditure Survey.” Review of In-
Bank Research Observer 28 (2): 220–35. come and Wealth 49 (1): 53–68.
Ferreira, Francisco, H. G., Julián Messina, Jamele Goldin, Claudia, and Lawrence F. Katz. 2002.
Rigolini, Luis F. López-Calva, María Ana Lugo, “The Power of the Pill: Oral Contraceptives
and Renos Vakis. 2013. Economic Mobility and and Women’s Career and Marriage Decisions.”
the Rise of the Latin American Middle Class. Journal of Political Economy 110 (4): 730–70.
Washington, DC: World Bank. ———. 2007. “The Race Between Education
Finaret, Amelia B., Laurie C. Miller, Neena Joshi, and Technology: The Evolution of US Educa-
Shubh Mahato, Mahendra N. Lohani, John tional Wage Differentials, 1890 to 2005.” NBER
Drozdowsky, and Beatrice Lorge Rogers. 2018. Working Paper 12984, National Bureau of Eco-
“Longitudinal Analysis of the Intrahousehold nomic Research, Cambridge, MA.
Distribution of Foods in Rural Nepal: Relative Government of India. 2007. “Poverty Estimates
Variability of Child Dietary Quality across Age for 2004–05.” New Delhi: Press Information
and Sex Cohorts.” Food Policy 79: 101–10. Bureau.
Fledderjohann, Jasmine, Sutapa Agrawal, Suku- Government of India, Planning Commission.
mar Vellakkal, Sanjay Basu, Oona Campbell, 2014. “Report of the Expert Group to Review
Pat Doyle, Shah Ebrahim, and David Stuckler. the Methodology for Measurement of Pov-
2014. “Do Girls Have a Nutritional Disadvan- erty.” Government of India, New Delhi. http://
tage Compared with Boys? Breastfeeding, Food planningcommission.nic.in/reports/genrep
Consumption, and Mortality among Indian /pov_rep0707.pdf.
Siblings.” European Journal of Public Health 24 Government of Nepal, Central Bureau of Sta-
(S2): cku162–060. tistics. 2012. “Poverty in Nepal: 2010–2011.”
Folbre, Nancy, Marta Murray-Close, and Jooyeoun Central Bureau of Statistics, Kathmandu.
Suh. 2018. “Equivalence Scales for Extended http://cbs.gov.np/sectoral_statistics/poverty
Income in the U.S.” Review of Economics of the /poverty_nepal_2010-11.
Household 16: 189–227. Green, Duncan. 2010. “Are Women Really 70% of
Foster, James E. 1998. “Absolute versus Relative the World’s Poor? How Do We Know?” From
Poverty.” The American Economic Review 88 Poverty to Power (blog), February 3. https://
(2): 335–41. oxfamblogs.org/fp2p/are-women-really-70-of
Foster, James, Joel Greer, and Erik Thorbecke. -the-worlds-poor-how-do-we-know/.
1984. “A Class of Decomposable Poverty Mea- Greenberg, Jon. 2014. “Carly Fiorina: 70% of World’s
sures.” Econometrica 52 (3): 761–66. Poor Are Women.” Punditfact (blog), January 15.
Fuchs, Victor. 1967. “Redefining Poverty and Re- http://www.politifact.com/punditfact/statements
distributing Income.” The Public Interest 14 (8): /2014/jan/15/carly-fiorina/carly-fiorina-70
88–95. -worlds-poor-are-women/.
Gaddis, Isis, Talip Kilic, Valérie Lechene, and Al- Groth, Hand, and John May. 2017. Africa’s Popu-
exander Wolf. Forthcoming. “Operationalizing lation: In Search of a Demographic Dividend.
Estimates of Intrahousehold Resource Alloca- Cham: Springer International Publishing AG.
tion for Poverty Analysis: A 3-Country Study.” Grown, Caren. 2010. “Taxation and Gender Equal-
Policy Research Working Paper Series, World ity: A Conceptual Framework.” In Taxation
Bank, Washington, DC. and Gender Equity: A Comparative Analysis of
Ganong, Peter, and Daniel Shoag. 2017. “Why Has Direct and Indirect Taxes in Developing and De-
Regional Income Convergence in the US De- veloped Countries, edited by Caren Grown and
clined?” Journal of Urban Economics 102: 76–90. Imraan Valodia, 1–22. Routledge International
Gentilini, Ugo, and Andy Sumner. 2012. “What Do Studies in Money and Banking Series. New
National Poverty Lines Tell Us about Global York: Routledge.
Poverty?” IDS Working Paper 392, Institute ———. 2014. “Missing Women: Gender and the
for Development Studies, University of Sussex, Extreme Poverty Debate.” Learning Lab, United
Colchester. States Agency for International Development,
Gibson, John, Jikun Huang, and Scott Rozelle. Washington, DC. https://usaidlearninglab.org
2003. “Improving Estimates of Inequality and /library/missing-women-gender-and-extreme
Poverty from Urban China’s Household In- -poverty-debate.
REFERENCES 171
Grown, Caren, and Imraan Valodia. 2010. Taxa- IMF (Internatonal Monetary Fund). 2017. Europe
tion and Gender Equity: A Comparative Anal- Regional Economic Outlook: Europe Hitting Its
ysis of Direct and Indirect Taxes in Developing Stride. Washington, DC: International Mone-
and Developed Countries. New York: Routledge. tary Fund.
Haddad, Lawrence James, John F. Hoddinott, Independent Evaluation Group. 2015. The Poverty
and Harold Alderman. 1997. Intrahousehold Focus of Country Programs: Lessons from World
Resource Allocation in Developing Countries: Bank Experience. Washington, DC: World Bank.
Models, Methods, and Policy. Washington, DC: ———. 2017. Growth for the Bottom 40 Percent:
International Food Policy Research Institute; The World Bank Group’s Support for Shared
Baltimore, MD: John Hopkins University Press. Prosperity. Washington, DC: World Bank.
Haddad, Lawrence James, and Ravi Kanbur. 1990. Jolliffe, Dean. 2001. “Measuring Absolute and
“How Serious Is the Neglect of Intra-house- Relative Poverty: The Sensitivity of Estimated
hold Inequality?” Economic Journal 100 (402): Household Consumption to Survey Design.”
866–81. Journal of Economic and Social Measurement 27
Harris-Fry, Helen A., Puskar Paudel, Niva Shres- (1-2): 1–23.
tha, Tom Harrisson, B. James Beard, Sonali Jolliffe, Dean, Peter Lanjouw, Shaohua Chen, Aart
Jha, Bhim P. Shrestha, Dharma S. Manandhar, Kraay, Christian Meyer, Mario Negre, Espen
Anthony Costello, Mario Cortina-Borja, and Beer Prydz, Renos Vakis, and Kyla Wethli. 2015.
Naomi Saville. 2018. “Status and Determinants A Measured Approach to Ending Poverty and
of Intra-household Food Allocation in Rural
Boosting Shared Prosperity: Concepts, Data, and
Nepal.” European Journal of Clinical Nutrition,
the Twin Goals. Policy Research Report. Wash-
January 22.
ington, DC: World Bank.
Headey, Derek, and Olivier Ecker. 2013. “Rethink-
Jolliffe, Dean, and Espen Beer Prydz. 2016. “Esti-
ing the Measurement of Food Security: From
mating International Poverty Lines from Com-
First Principles to Best Practice.” Food Security
parable National Thresholds.” The Journal of
5 (3): 327–43.
Economic Inequality 14 (2): 185–98.
Hentschel, Jesko, and Peter Lanjouw. 2000.
———. 2017. “Societal Poverty: A Relative and
“Household Welfare Measurement and the
Relevant Measure.” Policy Research Working
Pricing of Basic Services.” Journal of Interna-
Paper Series, 8073, World Bank, Washington,
tional Development, 12 (1): 13–27.
DC. http://hdl.handle.net/10986/26845.
Hill, Ruth, and Mejia-Mantilla, Carolina. 2017.
Jolliffe, Dean, and Umar Serajuddin. 2018. “Non-
“With a Little Help: Shocks, Agricultural In-
comparable Poverty Comparisons.” Journal of
come, and Welfare in Uganda.” Policy Research
Working Paper 7935, World Bank, Washington, Development Studies 54 (3): 523–36.
DC. Katz, Lawrence, and David Autor. 1999. “Changes
Hjelm, Lisa, Lucia Ferrone, Sudhanshu Handa, in the Wage Structure and Earnings Inequal-
and Yekaterina Chzhen. 2016. “Comparing ity.” In Handbook of Labor Economics 3A, edited
Approaches to the Measurement of Multidi- by Orley Ashenfelter and David Card. Amster-
mensional Child Poverty.” Office of Research— dam: Elsevier.
Innocenti Working Paper 29, UNICEF, Katz, Lawrence, and Robert A. Margo. 2014. “Tech-
Florence. nical Change and the Relative Demand for
Hunt, Janet. 2017. “Individual Deprivation Mea- Skilled Labor: The United States in Historical
sure: Methodology Update 2017.” Australian Perspective.” In Human Capital in History: The
National University, Canberra; International American Record, edited by Leah Platt Boustan,
Women’s Development Agency, Melbourne. Carola Frydman, and Robert A. Margo, 15–57.
IDM (Individual Deprivation Measure). 2017. Chicago: University of Chicago Press.
“Exploring Multidimensional Poverty in Fiji: Kilic, Talip, and Heather Moylan. 2016. “Meth-
Key Findings from a Study Using the IDM.” odological Experiment on Measuring Asset
Australian National University, Canberra; In- Ownership from A Gender Perspective
ternational Women’s Development Agency, (MEXA).” Technical Report, Living Stan-
Melbourne. http://www.individualdeprivation dards Measurement Study Unit, Development
measure.org/wp-content/uploads/IDM_6Pager_ Data Group, World Bank, Washington, DC.
V4.pdf. http://siteresources.worldbank.org/INTLSMS
172 POVERTY AND SHARED PROSPERITY 2018
/Resources/3358986-1423600559701/MEXA_ Lugo, María Ana. 2007. “Employment: A Proposal
Technical_Report.pdf. for Internationally Comparable Indicators.”
Kim, Jim Yong. 2016. “Speech by World Bank Pres- Oxford Development Studies 35 (4): 361–78.
ident Jim Yong Kim: The World Bank Group’s Mack, Joanna, and Stewart Lansley. 1985. Poor
Mission: To End Extreme Poverty.” World Britain. London: Allen & Unwin.
Bank, Washington, DC. Mallard, Simonette R., Lisa A. Houghton, Suzanne
Klasen, Stephan. 2007. “Gender-Related Indicators Filteau, Molly Chisenga, Joshua Siame, Lack-
of Well-Being.” In Human Well-Being: Concept son Kasonka, Anne Mullen, and Rosalind S.
and Measurement, edited by Mark McGillivray, Gibson. 2016. “Micronutrient Adequacy and
167–92. London: Palgrave Macmillan. Dietary Diversity Exert Positive and Distinct
Klasen, Stephan, and Rahul Lahoti. 2016. “How Effects on Linear Growth in Urban Zambian
Serious Is the Neglect of Intra-Household In- Infants.” The Journal of Nutrition 146 (10):
equality in Multi-dimensional Poverty Indi- 2093–101.
ces?” Discussion Paper 200, Courant Research Marlier, Eric, Anthony B. Atkinson, Bea Cantil-
Centre, Georg-August-Universität Göttingen, lon, and Brian Nolan. 2007. The EU and Social
Göttingen, Germany. Inclusion: Facing the Challenges. Bristol, UK:
———. Forthcoming. “An Individual-based Policy Press at the University of Bristol.
Multidimensional Poverty Assessment: An Ap- Maxwell, Daniel, Bapu Vaitla, and Jennifer Coates.
plication to Six Developing Countries.” Policy 2014. “How Do Indicators of Household Food
Research Working Paper Series, World Bank, Insecurity Measure Up? An Empirical Com-
Washington, DC. parison from Ethiopia.” Food Policy 47: 107–16.
Kraay, Aart. 2018. “Proposed Methodology for Mercier, Marion, and Philip Verwimp. 2017. “Are
a World Bank Human Capital Index.” Policy We Counting All the Poor? Accounting for the
Research Working Paper 8593, World Bank, Intra-household Allocation of Consumption
Washington, DC. in Burundi.” Journal of Demographic Economics
Lakner, Christoph, Daniel Gerszon Mahler, Minh 83 (3): 307–27.
Cong Nguyen, João Pedro Azevedo, Shaohua Milazzo, Annamaria, and Dominique van de
Chen, Dean Jolliffe, Espen Beer Prydz, and Walle. 2017. “Women Left Behind? Poverty and
Prem Sangraula. 2018. “Consumer Price In- Headship in Africa.” Demography 54: 1119–45.
dices Used in Global Poverty Measurement.” Moursi, Mourad M., Mary Arimond, Kathryn G.
Global Poverty Monitoring Technical Note, no. 4. Dewey, Serge Treche, Marie T. Ruel, and Fran-
World Bank, Washington, DC. cis Delpeuch. 2008. “Dietary Diversity is a
Lakner, Christoph, Mario Negre, and Espen Beer Good Predictor of the Micronutrient Density
Prydz. 2014. “Twinning the Goals: How Can of the Diet of 6- to 23-Month-Old Children in
Shared Prosperity Help to Reduce Global Pov- Madagascar.” The Journal of Nutrition 138 (12):
erty?” Policy Research Working Paper 7106, 2448–53.
World Bank, Washington, DC. Muñoz Boudet, Ana Maria, Paola Buitrago, Ben-
———. 2015. “The Role of Inclusive Growth in edicte Leroy de la Briere, David Newhouse,
Ending Extreme Poverty.” Paper presented at Eliana Rubiano Matulevich, Kinnon Scott, and
the Sixth Meeting of the Society for the Study Pablo Suarez-Becerra. 2018. “Gender Differ-
of Economic Inequality, University of Luxem- ences in Poverty and Household Composition
bourg, Luxembourg, July 13–15. through the Life-Cycle: A Global Perspective.”
Leroy, Jef L., Marie Ruel, Edward A. Frongillo, Policy Research Working Paper 8360, World
Jody Harris, and Terri J. Ballard. 2015. “Mea- Bank, Washington, DC.
suring the Food Access Dimension of Food Narayan, Ambar, Roy Van der Weide, Alexandru
Security: A Critical Review and Mapping of Cojocaru, Christoph Lakner, Silvia Redaelli,
Indicators.” Food and Nutrition Bulletin 36 (2): Daniel Gerszon Mahler, Rakesh Gupta N.
167–95. Ramasubbaiah, and Stefan Thewissen. 2018.
Lindgren, Mattias. 2015. “The Elusive Quest for Fair Progress?: Economic Mobility across Gen-
the Subsistence Line. How Much Does the Cost erations around the World. Equity and Develop-
of Survival Vary between Populations?” Com- ment. Washington, DC: World Bank.
parative Institutional Analysis Working Paper Narayan, Deepa, Raj Patel, Kai Schafft, Anne
2015 (1), Lund University, Lund. Rademacher, and Sarah Koch-Schulte. 2000a.
REFERENCES 173
Voices of the Poor: Can Anyone Hear Us? Wash- Using National Accounts and Surveys.” Policy
ington, DC: World Bank; New York: Oxford Research Working Paper Series, World Bank,
University Press. http://documents.worldbank Washington, DC.
.org/curated/en/131441468779067441/Voices Quisumbing, Agnes R., Lawrence Haddad, and
-of-the-poor-can-anyone-hear-us. Christine Peña. 2001. “Are Women Overrepre-
Narayan, Deepa, Robert Chambers, Meera K. Shah, sented among the Poor? An Analysis of Poverty
and Patti Petesh. 2000b. Voices of the Poor: in 10 Developing Countries.” Journal of Devel-
Crying Out for Change? Washington, DC: World opment Economics 66: 225–69.
Bank; New York: Oxford University Press. Rah, Jee H., Nasima Akhter, Richard D. Semba,
Newhouse, David, Pablo Suárez-Becerra, and Saskia De Pee, Martin W. Bloem, A. A. Camp-
Martin Evans. 2017. “New Global Estimates of bell, Regina Moench-Pfanner, K. Sun, Jane
Child Poverty and Their Sensitivity to Alter- Badham, and Klaus Kraemer. 2010. “Low Di-
native Equivalence Scales.” Economics Letters etary Diversity Is a Predictor of Child Stunting
157(August): 125–8. in Rural Bangladesh.” European Journal of Clin-
Newhouse, David, and Pallavi Vyas. 2018. ical Nutrition 64 (12): 1393.
“Nowcasting Poverty in India for 2014–15: Ravallion, Martin. 1998. “Poverty Lines in Theory
A Survey to Survey Imputation Approach.” and Practice.” LSMS Working Paper 133, World
World Bank, Washington, DC. Bank, Washington, DC. http://documents.
Nguyen, Phuong H., Rasmi Avula, Marie T. Ruel, worldbank.org/curated/en/916871468766156
Kuntal K. Saha, Disha Ali, Lan Mai Tran, Ed- 239/pdf/multi-page.pdf.
ward A. Frongillo, Purnima Menon, and Rahul ———. 2003. “Measuring Aggregate Welfare in
Rawat. 2013. “Maternal and Child Dietary Di- Developing Countries: How Well Do National
versity Are Associated in Bangladesh, Vietnam, Accounts and Surveys Agree?” Review of Eco-
and Ethiopia.” The Journal of Nutrition 143 (7): nomics and Statistics 85 (3): 645–52. https://
1176–83. doi.org/10.1162/003465303322369786.
OECD (Organisation for Economic Co-operation ———. 2010. “Poverty Lines across the World.”
and Development). 2012. Restoring Public Fi- Policy Research Working Paper 5284, World
nances, 2012 Update. Paris: OECD Publishing. Bank, Washington, DC.
OPHI (Oxford Poverty and Human Development ———. 2011. “On Multidimensional Indices of
Initiative). 2018. Global Multidimensional Pov- Poverty.” The Journal of Economic Inequality 9
erty Index 2018: The Most Detailed Picture to (2): 235–48.
Date of the World’s Poorest People. Oxford Uni- ———. 2015. “On Testing the Scale Sensitivity
versity Press. of Poverty Measures.” Economics Letters 137:
Owais, Aatekah, David G. Kleinbaum, Parminder 88–90.
S. Suchdev, A. S. G. Faruque, Sumon K. Das, ———. 2016. “Toward Better Global Poverty
Benjamin Schwartz, and Aryeh D. Stein. 2016. Measures.” The Journal of Economic Inequality
“Household Food Security and Infant Feeding 14 (2): 227–48.
Practices in Rural Bangladesh.” Public Health Ravallion, Martin, and Shaohua Chen. 2011.
Nutrition 19 (10): 1875–81. “Weakly Relative Poverty.” Review of Economics
Piketty, Thomas, Emmanuel Saez, and Gabriel and Statistics 93 (4): 1251–61.
Zucman. 2017. “Distributional National Ac- ———. 2017. “Welfare-Consistent Global Pov-
counts: Methods and Estimates for the United erty Measures.” NBER Working Paper 23739,
States. ” The Quarterly Journal of Economics 133 National Bureau of Economic Research, Cam-
(2): 553–609. bridge, MA.
Pitt, Mark M., Mark R. Rosenzweig, and Md. Ravallion, Martin, Shaohua Chen, and Prem
Nazmul Hassan. 1990. “Productivity, Health, Sangraula. 2009. “Dollar a Day Revisited.”
and Inequality in the Intrahousehold Distribu- The World Bank Economic Review 23 (2):
tion of Food in Low-Income Countries.” Amer- 163–84.
ican Economic Review 80 (5): 1139–56. Ravallion, Martin, Gaurav Datt, and Dominique
Prydz, Espen Beer, Dean Jolliffe, and Umar Ser- van de Walle. 1991. “Quantifying Absolute
ajuddin. Forthcoming. “Mind the Gap: Dis- Poverty in the Developing World.” Review of
parities in Assessments of Living Standards Income and Wealth 37 (4): 345–61.
174 POVERTY AND SHARED PROSPERITY 2018
Ridao-Cano, Cristobal, and Bodewig, Christian. Sen, Amartya. 1983. “Poor, Relatively Speaking.”
2018. Growing United: Upgrading Europe’s Con- Oxford Economic Papers 35 (2): 153–69.
vergence Machine. Washington, DC: World ———. 1987. The Standard of Living: Tanner Lec-
Bank. tures in Human Values. Cambridge: Cambridge
Roemling, Cornelia, and Matin Qaim. 2013. “Dual University Press.
Burden Households and Intra-household Nu- ———. 1990. “Gender and Cooperative Con-
tritional Inequality in Indonesia.” Economics & flicts.” In Persistent Inequalities: Women and
Human Biology 11 (4): 563–73. World Development, edited by Irene Tinker,
Rosenblatt, David, and Tamara J. McGavock. 2013. 123–49. New York: Oxford University Press.
“A Note on the Simple Algebra of the Shared ———. 2009. The Idea of Justice. Cambridge, MA:
Prosperity Indicator.” Policy Research Working The Belknap Press of Harvard University Press.
Paper 6645, World Bank, Washington, DC. Serajuddin, Umar, Hiroki Uematsu, Christina
Rothbarth, Erwin. 1943. “Note on a Method of Wieser, Nobuo Yoshida, and Andrew Dabalen.
Determining Equivalent Income for Fami- 2015. “Data Deprivation: Another Deprivation
lies of Different Composition.” In War-Time to End.” Policy Research Working Paper 7252,
Pattern of Saving and Expenditure, edited by World Bank, Washington, DC.
Charles Madge. Cambridge: Cambridge Uni- Shambaugh, Jay, and Ryan Nunn, eds. 2018. Revi-
versity Press. talizing Wage Growth: Policies to Get American
Rubiano Matulevich, Eliana, and Mariana Vio- Workers a Raise. The Hamilton Project. Wash-
llaz. 2018. “Gender Differences in Time Use: ington DC: Brookings.
Allocating Time Between the Market and the Stiglitz, Joseph, Amartya Sen, and Jean-Paul Fi-
Household.” Unpublished manuscript. toussi. 2010. “Report by the Commission on
Ruel, Marie T. 2003. “Operationalizing Dietary Di- the Measurement of Economic Performance
versity: A Review of Measurement Issues and and Social Progress.” Council on Foreign Rela-
Research Priorities.” The Journal of Nutrition tions, Washington, DC.
133 (11): 3911S–26S. Stöckl, Heidi, Karen Devries, Alexandra Rotstein,
Sánchez-Páramo, Carolina, and Ana María Naeemah Abrahams, Jacquelyn Campbell,
Muñoz Boudet. 2018. “No, 70% of the World’s Charlotte Watts, and Claudia Garcia Moreno.
Poor Aren’t Women, but That Doesn’t Mean 2013. “The Global Prevalence of Intimate Part-
Poverty Isn’t Sexist.” Let’s Talk Development ner Homicide: A Systematic Review.” Lancet
(blog), March 8. http://blogs.worldbank.org 382 (9895): 859–65.
/developmenttalk/no-70-world-s-poor-aren-t Sutherland, Douglas, Peter Hoeller, and Rossana
-women-doesn-t-mean-poverty-isn-t-sexist. Merola. 2012. “Fiscal Consolidation: How
Santaeulàlia-Llopis, Raül, and Yu Zheng. 2017. Much, How Fast, and by What Means?” Eco-
“Why Is Food Consumption Inequality Un- nomic Policy Paper 1, Organisation for Eco-
derestimated? A Story of Vices and Children.” nomic Co-operation and Development, Paris.
Working Paper 969, Barcelona Graduate Tian, Xu, Xiaohua Yu, and Stephan Klasen. 2018.
School of Economics, Barcelona. http://r-santa “Gender Discrimination in China Revisited:
eulalia.net/pdfs/Scales-short.pdf. A Perspective from Family Welfare.” Journal of
Save the Children. 2016. Child Poverty: What Chinese Economic and Business Studies 16 (1):
Drives It and What It Means to Children across 95–115.
the World. London: Save the Children UK. Torheim, Liv E., Fatimata Ouattara, Modibo M.
https://resourcecentre.savethechildren.net Diarra, F. D. Thiam, Ingrid Barikmo, Anne
/library/child-poverty-what-drives-it-and- Hatløy, and Arne Oshaug. 2004. “Nutrient Ad-
what-it-means-children-across-world. equacy and Dietary Diversity in Rural Mali:
Savy, Mathilde, Yves Martin-Prével, Prosper Association and Determinants.” European Jour-
Sawadogo, Yves Kameli, and Francis Del- nal of Clinical Nutrition 58 (4): 594.
peuch. 2005. “Use of Variety/Diversity Scores Townsend, Peter. 1979. Poverty in the United King-
for Diet Quality Measurement: Relation with dom: A Survey of Household Resources and
Nutritional Status of Women in a Rural Area Standards of Living. Harmondsworth: Penguin.
in Burkina Faso.” European Journal of Clinical UBOS (Uganda Bureau of Statistics) and ICF In-
Nutrition 59 (5): 703. ternational. 2017. “Uganda Demographic and
REFERENCES 175
Health Survey 2016: Key Indicators Report.” ———. 2011. World Development Report 2012:
UBOS, Kampala, Uganda; ICF International, Gender Equality and Development. Washing-
Rockville, MD. ton, DC: World Bank.
Udry, Christopher. 1996. “Gender, Agricultural ———. 2013. World Development Report 2014:
Production, and the Theory of the Household.” Risk and Opportunity—Managing Risk for De-
Journal of Political Economy 104 (5): 1010–45. velopment. Washington, DC: World Bank.
UNDP (United Nations Development Pro- ———. 2015. Global Monitoring Report
gramme). 1997. Human Development Report 2014/2015: Ending Poverty and Sharing Pros-
1997: Human Development to Eradicate Pov- perity. Washington, DC: World Bank.
erty. New York: Oxford University Press. ———. 2016a. Brazil Systematic Country Diagnos-
———. 2010. Human Development Report 2010: tic: Retaking the Path to Inclusion, Growth, and
The Real Wealth of Nations. New York: Palgrave Sustainability. Washington, DC: World Bank.
Macmillan. ———. 2016b. Poverty and Shared Prosperity
———. 2014. “Poverty in El Salvador: As Viewed 2016: Taking on Inequality. Washington, DC:
World Bank.
by Its Protagonists.” UNDP, San Salvador.
———. 2017a. World Development Report 2017:
UNECE (United Nations Economic Commission
Governance and the Law. Washington, DC:
for Europe). 2017. “Guide on Poverty Measure-
World Bank.
ment.” United Nations, New York.
———. 2017b. Monitoring Global Poverty: Report
UNHCR (United Nations High Commissioner
of the Commission on Global Poverty. Washing-
for Refugees). 2018. “Global Trends: Forced
ton, DC: World Bank.
Displacement in 2017.” Geneva. http://www
———. 2018a. Riding the Wave: An East Asian
.unhcr.org/statistics/unhcrstats/5b27be547
Miracle for the 21st Century. World Bank East
/unhcr-global-trends-2017.html.
Asia and Pacific Regional Report. Washington,
Vecchi, Giovanni. 2015. “Counting the Poor: New DC: World Bank.
Thinking about European Poverty Measures ———. 2018b. Stagnant Poverty Reduction in
and Lessons for the United States.” Economica Latin America. Washington, DC: World Bank.
82 (327): 586–87. ———. 2018c. Growing United. Washington, DC:
Walker, Robert, and Xavier Godinot. 2018. “An World Bank.
International Participatory Research on the ———. 2018d. World Development Report 2018:
Dimensions of Poverty.” ATD Fourth World Learning to Realize Education’s Promise. Wash-
International, Pierrelaye, France. ington, DC: World Bank.
World Bank. 1990. World Development Report ———. 2019. World Development Report 2019:
1990: Poverty. Washington, DC: World Bank. The Changing Nature of Work. Washington,
———. 2001. World Development Report DC: World Bank.
2000/2001: Attacking Poverty. Washington, DC: ———. Forthcoming. “Living and Leaving: Hous-
World Bank. ing, Mobility, and Welfare in the European
———. 2009. “Jordan—Poverty Update.” Report Union.” World Bank Regional Report, World
47951, World Bank, Washington, DC. Bank, Washington, DC.
176 POVERTY AND SHARED PROSPERITY 2018
ECO-AUDIT
Environmental Benefits Statement
The World Bank Group is committed to reducing its environmental footprint. In
support of this commitment, we leverage electronic publishing options and print-
on-demand technology, which is located in regional hubs worldwide. Together,
these initiatives enable print runs to be lowered and shipping distances decreased,
resulting in reduced paper consumption, chemical use, greenhouse gas emissions,
and waste.
We follow the recommended standards for paper use set by the Green Press
Initiative. The majority of our books are printed on Forest Stewardship Council
(FSC)–certified paper, with nearly all containing 50–100 percent recycled con-
tent. The recycled fiber in our book paper is either unbleached or bleached using
totally chlorine-free (TCF), processed chlorine–free (PCF), or enhanced elemental
chlorine–free (EECF) processes.
More information about the Bank’s environmental philosophy can be found at
http://www.worldbank.org/corporateresponsibility.