. 1
( 13)


Poverty Dynamics
This page intentionally left blank
Poverty Dynamics
Interdisciplinary Perspectives

Edited by
Tony Addison, David Hulme, and Ravi Kanbur

Great Clarendon Street, Oxford ox2 6DP
Oxford University Press is a department of the University of Oxford.
It furthers the University™s objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With of¬ces in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
© Oxford University Press 2009
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
First published 2009
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
and you must impose the same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Poverty dynamics : interdisciplinary perspectives / edited by
Tony Addition, David Hulme, and Ravi Kanbur.
p. cm.
ISBN 978“0“19“955754“7
1. Poverty“Measurement. 2. Social isolation. 3. Social structure.
4. Social mobility. 5. Economic policy. I. Addison, Tony.
II. Hulme, David. III. Kanbur, S. M. Ravi.
HC79.P6P6835 2008
339.4 6“dc22 2008035834
Typeset by SPI Publisher Services, Pondicherry, India
Printed in Great Britain
on acid-free paper by
CPI Antony Rowe, Chippenham, Wiltshire

ISBN 978“0“19“955754“7 (Hbk.)
ISBN 978“0“19“955754“4 (Pbk.)

1 3 5 7 9 10 8 6 4 2
David Hulme dedicates this book to his parents
Walter and Jessie Hulme,
for all their loving care and support over the decades
This page intentionally left blank

Research on poverty has never been as vigorous as it is now. The very notion
of poverty”how we conceptualize what it means to be poor”and how we
measure it is generating a rich debate involving anthropologists, economists,
human geographers, political scientists, sociologists, and other social scien-
tists. And whereas previously the subject tended to be split between those who
pursued primarily quantitative approaches to poverty measurement (from
the economics and social statistics traditions) and those pursuing qualita-
tive approaches (from the other social sciences) there is now much greater
interaction and discussion between researchers from different disciplinary
This book re¬‚ects our belief that there are three main fronts on which
progress must be made if we are to dramatically deepen the understanding
of why poverty occurs, and signi¬cantly improve the effectiveness of poverty
reduction policies.
First, poverty research needs to focus on poverty dynamics”over the life
course, across generations, and between different social groups. There is now a
wide acceptance that static analyses have limited explanatory power and may
conceal the processes that are central to the persistence of poverty and/or its
Second, there is a need to move efforts to measure poverty dynamics
beyond mere income and consumption to more multidimensional concepts and
measures of poverty. This is increasingly common in static analyses but is rare
in work on poverty dynamics. This might involve economic assets or, more
ambitiously, using the concepts of human development or well-being.
Third, there is a growing consensus that a thorough understanding of
poverty and poverty reduction requires cross-disciplinary research and the
development of ˜Q-squared methodologies™ (quantitative and qualitative).
Quantitative studies may be representative but by concentrating on what
is readily measurable they risk missing out on crucial variables and ¬nd it
hard to go beyond identifying correlates. Qualitative studies can capture the
complexity of poverty dynamics and the processes that underpin poverty but
they can be dismissed for not being able to explain how their ¬ndings relate
to wider populations. Cross-disciplinary and Q-squared approaches offer the


possibility of combining the strengths of different disciplines and methods to
produce deeper understandings.
In this volume we examine the opportunity for the evolution of innovative,
cross-disciplinary analyses to poverty dynamics by presenting the reader with
the latest thinking from a group of eminent researchers. The book is divided
into three parts.

1. Introduction: this provides an overview and argues that the present con-
versation about poverty dynamics reveals a divide, between economists
and other social scientists, about the concepts and methods that will
push forward the frontier of poverty analysis. However, it also reveals
that there is a strong desire, and increasingly frequent attempts, to bridge
this divide.
2. Poverty Dynamics: Poverty Measurement and Assessment: this focuses on
cutting-edge approaches to analysing poverty dynamics. It covers non-
income poverty measures, assessments of the likelihood of future
poverty, the incorporation of time duration into the FGT poverty mea-
sures, asset-based approaches, and analyses based on life histories and
participatory methods.
3. Explanatory Frameworks for Understanding Poverty Dynamics: this section
focuses upon differing explanatory frameworks. These include politi-
cal economy, social ordering, contemporary social theory, capabilities
approaches, and the intergenerational transmission of poverty.

The overarching aim of the book is to push forward the conceptualization of
poverty dynamics and generate knowledge that will ultimately advance the
goal of poverty reduction.
This book arises from the ongoing research of the Chronic Poverty Research
Centre (CPRC). It derives from a workshop of leading scholars on ˜Con-
cepts and Associated Methods for Analysing Poverty Dynamics™ held at
the University of Manchester, 23“4 October 2006. The workshop was con-
vened by the CPRC at the University™s Institute for Development Policy
and Management (IDPM) and Brooks World Poverty Institute (BWPI). More
than twenty papers were presented and discussed at the workshop (see
<www.chronicpoverty.org> for the original papers and full details). Fourteen
of these papers have been revised for this volume.
Mounting such a workshop and producing a book have involved the efforts
of many organizations and people. Our thanks to CPRC for ¬nancing and
organizing the workshop, and to the UK™s Department for International
Development (DFID) for its ¬nancial support to CPRC. The workshop and
this volume would not have been possible without the efforts and energy of
many people. Our especial thanks to: Denise Redston, the workshop orga-
nizer (who has also coordinated the preparation of this book); Karen Moore


at CPRC for helping to develop the original idea, structure the workshop,
and for reviewing many of the papers; David Clark for advice and support;
CPRC™s theme coordinators (Armando Barrientos, Bob Baulch, Kate Bird, Sam
Hickey, and Andy McKay) for advice and reviewing papers; the participants
at the workshop who provided detailed criticism and encouragement that has
greatly strengthened the collection; and to the many others who reviewed
papers and/or provided practical support.
Finally a word of thanks to the many poor people around the world
who have patiently answered the questions of enumerators and interviewers
and participated in the participatory exercises that have provided the ideas
and data for much of the work in this volume. Their unremitting efforts
to improve their own and their children™s lives serve as an inspiration to
those of us fortunate to be privileged with the task of trying to deepen the
understanding of poverty and poverty reduction. All royalties from this book
are donated to BRAC in Bangladesh to support its programmes which help
some of the world™s poorest people to improve their lives.
Tony Addison, Manchester
David Hulme, Manchester
Ravi Kanbur, Ithaca

This page intentionally left blank

List of Figures xiii
List of Tables xiv
List of Contributors xvi

Part I. Introduction
1. Poverty Dynamics: Measurement and Understanding from an
Interdisciplinary Perspective
Tony Addison, David Hulme, and Ravi Kanbur

Part II. Poverty Dynamics: Poverty Measurement
and Assessment
2. Chronic Poverty and All That: The Measurement of Poverty
Over Time
C©sar Calvo and Stefan Dercon

3. A Class of Chronic Poverty Measures
James E. Foster

4. Measuring Chronic Non-Income Poverty
Isabel Günther and Stephan Klasen

5. The Construction of an Asset Index: Measuring Asset Accumulation
in Ecuador
Caroline Moser and Andrew Felton

6. Looking Forward: Theory-Based Measures of Chronic Poverty
and Vulnerability
Michael R. Carter and Munenobu Ikegami

7. Poverty in Time: Exploring Poverty Dynamics from Life History
Interviews in Bangladesh
Peter Davis


8. Subjective Assessments, Participatory Methods, and Poverty
Dynamics: The Stages of Progress Method
Anirudh Krishna

Part III. Explanatory Frameworks for Understanding
Poverty Dynamics
9. Bringing Politics Back into Poverty Analysis: Why Understanding
of Social Relations Matters More for Policy on Chronic Poverty
than Measurement
John Harriss

10. Poverty Measurement Blues: Beyond ˜Q-Squared™ Approaches to
Understanding Chronic Poverty in South Africa
Andries du Toit

11. When Endowments and Opportunities Don™t Match:
Understanding Chronic Poverty
S. R. Osmani

12. Investments, Bequests, and Public Policy: Intergenerational
Transfers and the Escape from Poverty
Agnes R. Quisumbing

13. Questioning the Power of Resilience: Are Children up to the
Task of Disrupting the Transmission of Poverty?
Jo Boyden and Elizabeth Cooper

14. The Social Distribution of Sanctioned Harm: Thinking through
Chronic Poverty, Durable Poverty, and Destitution
Maia Green

15. Toward an Economic Sociology of Chronic Poverty: Enhancing
the Rigour and Relevance of Social Theory
Michael Woolcock

Index 349

List of Figures

2.1. Illustrative examples of poverty experiences 32
2.2. Examples of poverty experiences from the Ethiopian panel data survey 33
5.1. Difference between regression and PCA 106
5.2. Consumer durables capital density estimates 115
5.3. Household asset accumulation in Guayaquil, Ecuador, 1978“2004 121
5.4. Patterns of housing and consumer durables investment over time by
income group 122
5.5. Trade-off between consumption and kids™ education 123
5.6. Star graphs of household asset portfolios 124
6.1. Assets and livelihood options 131
6.2. The Micawber Frontier (non-stochastic case) 133
6.3. The irreversible consequences of shocks 136
6.4. Vulnerability shifts out the Micawber Frontier 137
6.5. Vulnerability hurts ˜average™ individuals most 138
7.1. Declining smooth: Sukur Ali 166
7.2. Improving smooth: Jehangir 168
7.3. Level saw-tooth: Fuljan”potential for improvement hampered by
dowry and illness 171
7.4. Declining saw-tooth: Amir Hossain 172
7.5. Improving saw-tooth: Rena 175
7.6. Declining single-step: Allauddin”accident and disability 177
7.7. Declining multi-step: Amit 179
10.1. PLAAS and CPRC™s research sites in South Africa (map by John Hall) 235
11.1. Chronic poverty 251
11.2. Varieties of poverty 253
11.3. Varieties of chronic poverty 256

List of Tables

2.1. Poverty episodes 1994 to 2004 (based on ¬ve rounds) 49
2.2. Spearman rank correlation between different poverty measures 50
2.3. Correlates of poverty measures (Tobit model) 52
2.4. Percentage change in poverty index from marginal change in characteristics 53
3.1. Chronic poverty in Argentina: estimated levels for various measures
and durations 72
3.2. Chronic poverty pro¬le for Argentina 73
4.1. Poverty rates and dynamics 89
4.2. Poverty dynamics using ¬tted income poverty rates 90
4.3. Correlation of income and non-income dynamics 91
4.4. Correlation of static income and non-income poverty 92
4.5. Correlation of income and non-income dynamics 93
4.6. Household non-income poverty dynamics 94
4.7. Average household poverty dynamics 95
4.8. Intergenerational chronic poverty (1997) 96
4.9. Chronic poverty as multidimensional poverty 96
5.1. Types of capital, asset categories, and components 112
5.2. Housing stock polychoric PCA coef¬cients 113
5.3. Consumer durables polychoric PCA coef¬cients 114
5.4. Value of educational levels 116
5.5. Financial/productive capital polychoric PCA coef¬cients 118
5.6. Community social capital polychoric PCA coef¬cients 119
6.1. Simulations for archetypical individuals 139
6.2. Simulated chronic poverty measures 143
6.3. Backward- and forward-looking poverty measures for South Africa 148
7.1. Current trajectory direction of all respondents 159
7.2. Ideal typical trajectory patterns 160
7.3. Examples of causes of declining smooth patterns 164
7.4. Reasons for improvement in saw-tooth trajectories 173

List of Tables

7.5. Examples of causes of single-step decline patterns 174
8.1. Stages of Progress and the poverty cut-off 186
8.2. Trends in poverty dynamics over twenty-¬ve years 189
8.3. Principal reasons for falling into poverty 190
8.4. Principal reasons for escaping poverty 192
8.5. Stages of Progress and asset ownership (thirty-six communities in Uganda) 195
8.6. Stages (as recalled) vs. assets actually possessed seven years ago
(sixty-one communities of Rajasthan, India) 196

List of Contributors

Tony Addison is Professor of Development Studies and Executive Director
of the Brooks World Poverty Institute at the University of Manchester, and
Associate Director of the DFID-funded Chronic Poverty Research Centre. His
research focuses on poverty and development, in particular policies towards
chronic poverty reduction in Asia and Africa.
Jo Boyden is a social anthropologist with the Department of International
Development at the University of Oxford. Since 2005 she has been the director
of the Young Lives Project, a longitudinal study of child poverty in Ethiopia,
India, Peru, and Vietnam. Her research has focused on child poverty with
reference to armed con¬‚ict, forced migration, and coping and resilience.
C©sar Calvo is a Departmental Teaching Associate at Oxford University. He
also holds a post at the Universidad de Piura in Peru. His research focuses on
the analysis and measurement of poverty and vulnerability.
Michael R. Carter is Professor of Agricultural and Applied Economics at the
University of Wisconsin-Madison, and directs the BASIS Research Program on
Poverty, Inequality, and Development.
Elizabeth Cooper is a doctoral student in social anthropology at the Univer-
sity of Oxford. Her doctoral research is concerned with children™s, families™,
and institutions™ responses to orphaning in western Kenya. Elizabeth™s previ-
ous research has focused on the situations of children and youth in refugee
camps in Kenya and Uganda.
Peter Davis is Lecturer in International Development at the University of
Bath. His research has focused on poverty dynamics, livelihoods, and welfare
regimes with special reference to Bangladesh.
Stefan Dercon is Professor of Development Economics at the University
of Oxford. His work focuses on applied theory and empirical research on
problems of development. Recent research focuses on themes of poverty, risk,
migration, and institutions, with empirical work focusing on longitudinal
data from Ethiopia, Tanzania, and India.
Andrew Felton is a Ph.D. student in the Public Policy department of the
University of Maryland and a ¬nancial economist at the Federal Deposit

List of Contributors

Insurance Corporation. Previously, he worked as a Senior Research Analyst
at the Brookings Institution.
James E. Foster is Professor of Economics at Vanderbilt University. He has
worked extensively on the measurement of poverty and inequality as well
as literacy, health, education, and human development more generally. He is
currently working on external capabilities and the distribution of well-being
in Mexico.
Maia Green is Professor of Social Anthropology at the University of
Manchester and was formerly Deputy Director of the Global Poverty Research
Group at the Universities of Manchester and Oxford. Her research has focused
on poverty, development, and social institutions in East Africa and witchcraft
in Tanzania.
Isabel Günther ¬nished her Ph.D. in 2007 at the Department of Economics
at the University of Göttingen. Since then she has been a postdoctoral fellow
at Harvard University. Her main research interests are poverty and inequality
analysis as well as applied health and population economics with a regional
focus on sub-Saharan Africa.
John Harriss is Professor of International Studies at Simon Fraser University
in Canada. He has long-standing research interests in the political economy of
development, and in politics and society in South Asia. His early work focused
on the peasant economy and agrarian change, environmental change, labour
markets, and the informal sector. More recently he has engaged in debates on
the concept of social capital and about civil society and politics, and popular
representation, especially with regard to India.
David Hulme is Professor of Development Studies at the University of
Manchester, Associate Director of the Chronic Poverty Research Centre,
and Associate Director of the Brooks World Poverty Institute. He currently
holds a Leverhulme Senior Research Fellowship. His current research interests
focus on integrating time into poverty analysis, the concept and practice
of global poverty reduction, combining quantitative and qualitative research
approaches, understanding poverty dynamics (Bangladesh), the role of social
protection in poverty reduction strategies (global), natural resource manage-
ment and livelihoods (Africa), and micro¬nance.
Munenobu Ikegami is a Ph.D. student at the University of Wisconsin-
Madison. He will join International Livestock Research Institute as a post-
doctoral scientist in summer 2008. He is working on dynamic models of asset
accumulation and their implications for the design and implementation of
social policies.
Ravi Kanbur is T. H. Lee Professor of World Affairs, International Professor of
Applied Economics and Management, and Professor of Economics at Cornell
University. He previously held a number of posts at the World Bank including

List of Contributors

principal adviser to the chief economist. His main areas of research are pub-
lic economics and development economics. He is particularly interested in
bridging the worlds of rigorous analysis and practical policy making. He has
published extensively in leading journals and directed the World Development
Report 2000/01: Attacking Poverty.
Stephan Klasen is a Professor of Development Economics at the University of
Göttingen, Germany. He holds a Ph.D. in economics from Harvard University
and has held positions at the World Bank, King™s College, Cambridge, and
the University of Munich. His research is focused on the measurement and
analysis of poverty and inequality in developing countries.
Anirudh Krishna (Ph.D. in Government, Cornell, 2000; Masters in Econo-
mics, Delhi, 1980) is Associate Professor of Public Policy and Political Science
at Duke University. His research investigates how poor communities and
individuals in developing countries cope with the structural and personal
constraints that result in poverty and powerlessness.
Caroline Moser is Professor of Urban Development and Director of the
Global Urban Research Centre, School of Environment and Development,
University of Manchester. She is also an External Senior Fellow at Brookings
Institution. Previously she was Lead Specialist Social Development, Latin
America and the Caribbean Region, in the World Bank and prior to that a
lecturer at the London School of Economics. She has published widely on
asset accumulation and urban poverty reduction, violence and insecurity,
household vulnerability under structural adjustment, the informal sector, and
gender and development.
S. R. Osmani is Professor of Development Economics at the University of
Ulster. He has published widely on issues related to poverty, employment,
inequality, hunger, famine, nutrition, rights-based approach to development,
and development problems in general, and his publications include Economic
Inequality and Group Welfare, Nutrition and Poverty, Macroeconomics of Poverty
Reduction: The Case Study of Bangladesh, Macroeconomics of Poverty Reduction:
The Case Study of Bhutan, and The Employment Nexus between Growth and
Poverty: An Asian Perspective.
Agnes R. Quisumbing, an economist, is a senior research fellow at the
International Food Policy Research Institute, where she conducts research on
gender, poverty, and economic mobility.
Andries du Toit is Deputy Director of the Programme for Land and Agrarian
Studies at the University of the Western Cape in South Africa. He has worked
on chronic poverty, the sociology of labour relations, and agro-food restruc-
turing in South Africa.
Michael Woolcock is Professor of Social Science and Development Policy,
and Research Director of the Brooks World Poverty Institute, at the University

List of Contributors

of Manchester. His research draws on a range of disciplinary theories and
methods to explore the social dimensions of economic development, in par-
ticular the role that social networks play in the survival and mobility strategies
of the poor, in managing local con¬‚ict, and in shaping the ef¬cacy of legal
and political institutions. He has an MA and Ph.D. in sociology from Brown

This page intentionally left blank
Part I
This page intentionally left blank
Poverty Dynamics
Measurement and Understanding from an
Interdisciplinary Perspective

Tony Addison, David Hulme, and Ravi Kanbur

1.1. Introduction

There are three main fronts on which progress must be made if we are to
deepen our understanding of why poverty occurs, and signi¬cantly improve
the effectiveness of poverty reduction policies. First, poverty research needs
to focus on poverty dynamics”over the life course and across generations.
There is now a wide acceptance that static analyses have limited explanatory
power and may conceal the processes that are central to the persistence of
poverty and/or its elimination. Second, there is a need to move efforts to
measure poverty dynamics beyond mere income and consumption to more
multidimensional concepts and measures of poverty. This is increasingly com-
mon in static analyses but is rare in work on poverty dynamics. This might
involve assets or, more ambitiously, using concepts of human development
or well-being. Third, at the same time there is a growing consensus that
a thorough understanding of poverty and poverty reduction requires cross-
disciplinary research, using the strengths of different disciplines and methods,
and of quantitative and qualitative approaches to poverty analysis.
Thus, we believe that the next frontier in poverty research is at the intersec-
tion of dynamics and cross-disciplinarity. This chapter introduces a signi¬cant
new multidisciplinary collection of studies of poverty dynamics, presenting
the reader with the latest thinking by a group of researchers who are leaders
in their respective disciplines. 1 In this Introduction we set the papers in

The papers in this volume, together with others, were presented at the CPRC Workshop
on ˜Concepts and Methods for Analysing Poverty Dynamics and Chronic Poverty™, held at
the University of Manchester, 23“5 October 2006 (<www.chronicpoverty.org>).

Tony Addison, David Hulme, and Ravi Kanbur

context, beginning in section 1.2 with the issue of how to bring time into
the measurement of poverty and into the analysis of trajectories in and out
of poverty. We then compare qualitative and quantitative approaches and
address the issue of cross-disciplinarity in section 1.3. Section 1.4 presents an
overview of the chapters in the volume. Section 1.5 concludes by highlighting
areas where we believe future research on poverty dynamics should focus.

1.2. Time and Poverty 2

Time is a troubling and ambiguous concept in philosophy and in social analy-
sis. The complexities are apparent in Adam™s (2004) characterization: ˜Time is
lived, experienced, known, theorized, created, regulated, sold and controlled.
It is contextual and historical, embodied and objecti¬ed, abstracted and con-
structed, represented and commodi¬ed™ (Adam, 2004, p. 1).
Set against the notion of time as an abstract relation between the past,
present, and the future, in the tradition of St Augustine and Kierkegaard,
is what Adam (2004, p. 49) calls the ˜clock-time perspective™ of Aristotle,
Newton, Marx, Weber, and Durkheim. This is the dominant conceptualization
in the social sciences, and one that underpins the chapters in this volume. 3
Even within the ˜clock-time™ frame, it is possible to introduce time into the
conceptualization of poverty in one of two ways. The ¬rst of these involves
treating time as an ordinary dimension of well-being and poverty, as in the
World Bank™s Voices of the Poor (see Narayan et al., 2000, esp. pp. 21, 34, 92“3).
In this approach time, or lack of it, is merely another dimension of poverty. A
person is de¬ned as time poor if he or she lacks the necessary time to achieve
things of value, such as adequate sleep and rest, being with family and friends,
or income (see Clark, 2002, ch. 4). In effect, time is viewed as one, of many,
scarce resources (Becker, 1965). This approach ¬ts well into approaches that
emphasize the multidimensionality of well-being and poverty.
The second approach may be identi¬ed with the ˜poverty and well-being
dynamics™ perspective, with a focus on how well-being evolves over time,
what determines this evolution, and how different patterns of evolution
are to be evaluated for policy. This is the approach that characterizes the
chapters in this volume. They provide detailed examples of the ways in which
information about poverty dynamics can be acquired.
r Paneldata methods”this method is considered the most reliable by vir-
tually all quantitative researchers and by many qualitative researchers. It
involves conducting questionnaire surveys or semi-structured interviews
with the same individual or household at different points in time. This
Parts of this section derive from Clark and Hulme (2005).
Bevan (2004) distinguishes three approaches: clocks and calendars, rhythms, and histo-
ries, and considers ways of incorporating rhythms and histories into poverty analysis.

Poverty Dynamics

permits objective data to be collected for key measures and the collection
of information about the ways in which the individual/household explain
the changes that are occurring in their lives. Moser and Felton (this
volume) illustrate this method for both quantitative and qualitative data.
The strengths of this method are its rigour and the comparability of the
data it collects at different points in time. Its disadvantages are its costs
and the signi¬cant delay in analysis that it entails. Additional problems
include interviewee fatigue, matching households in large datasets, and
systematic sample attrition (see earlier).
r One-off indicators”given the dif¬culties of collecting panel data it is logi-
cal to seek to identify ˜one-off™ indicators (i.e. measures collected at a sin-
gle point in time) that provide information about poverty duration. The
most obvious type of indicator for this purpose is ˜nutrition™ oriented”on
the grounds that certain nutritional measures reveal what has been hap-
pening to an individual over an extended period of time. Researchers who
have adopted this approach have favoured child stunting as a measure
indicating that a child has been undernourished for an extensive period of
time and, by implication, that her or his household has been poor for an
extended period of time as it has been unable to provide an adequate diet.
Radhakrishna et al. (2006) have used this method to measure and analyse
chronic poverty in India. The great advantage of this method is that it
permits partial analyses of poverty dynamics for any population for which
an anthropometric survey is available: so, it can be low cost and rapid.
There are, however, severe challenges. These include questions about the
accuracy of data on height and age; the assumption that stunting is
caused by undernourishment, rather than by health problems, cultural
practices, and/or genetic factors; and the dif¬culty of moving beyond
simply identifying factors that correlate with stunting.
r Retrospectivedata: another means of avoiding the costs and delays of
collecting panel data is to ask interviewees to provide data about their
past circumstances at the same time as they are providing data about their
present condition. Many researchers are highly suspicious of this method,
however, because of the well-known problems of recall. These include the
dif¬culty of identifying the exact time that is to be recalled and of remem-
bering what conditions were like at that time. For some indicators”
attending school, formal employment status”data may be reason-
ably accurate. But for others”income, consumption, food availability”
quantitative data is unlikely to be reliable. In addition, over time people
tend to develop selective memories and may be perceived as ˜rewriting™
parts of their lives. 4 As a result, many researchers do not regard this as a

This is not a conscious attempt to lie, but a part of extremely complex psychological and
cognitive processes that are common, but highly varied, across humanity.

Tony Addison, David Hulme, and Ravi Kanbur

credible method for collecting quantitative, cardinal data. It is used exten-
sively to collect qualitative data, often to triangulate other data, and/or to
understand the ways in which people subjectively interpret change over
time. In recent years this method has become popular as a component
of participatory poverty assessments (PPAs). Arguably, the group-based
methods used in PPAs make data more reliable as interviewees debate each
other™s recall and researchers can triangulate data between groups.

Until the late 1980s the main ways in which time featured in poverty analysis
was in terms of poverty trends, seasonality, the timing of experiences, and his-
torical accounts of poverty. Poverty trends commonly contrasted headcounts
of poverty across a population at two (or more) different times. However,
comparing poverty trends in this sense does not tell us whether individuals or
households are persistently poor or if they typically move into and/or out of
poverty over time (see Hulme and Shepherd, 2003; Carter and Barrett, 2006;
Hulme, 2006). For example, Lawson, McKay, and Okidi (2003) record that
between 1992 and 1999 consumption poverty in Uganda fell by about 20 per
cent as the headcount rate fell from 55.7 per cent to 35.2 per cent. How-
ever, moving beyond conventional static poverty analysis by looking at the
dynamics of poverty (i.e. what actually happened to individual households
over time) provides a richer picture. Almost 30 per cent of poor households
in 1992 managed to move out of poverty by 1999, but around 10 per cent
of non-poor households fell into poverty. About 19 per cent of households
that were poor in 1992 remained poor in 1999 (ibid. 7 and table 1). Rather
than getting the false impression that life has improved for everyone we gain
a nuanced understanding of the ups and downs of welfare status.
The seasonality (or timing) of income, consumption, and access to food has
been another focus with particular interest in the annual cycles of relative
plenty and food shortage/hunger that occur in many rural areas (Chambers,
Longhurst, and Pacey, 1981; Chambers, 1983). The signi¬cance of speci¬c
poverty experiences at certain times in the life course has also been high-
lighted with a particular focus on lack of access to food/nutrition for pregnant
women and education for children. A lack of access to nutrition, basic health
services, or education in early life (foetal and infant) can have irreversible
effects on the physical stature and cognitive ability of people (Loury, 1981;
Strauss and Thomas, 1998). Historical accounts of poverty”seeking to lay out
and interpret the main experiences and events in a chronological order”also
continued (Hufton, 1974; Haswell, 1975; Geremek, 1994), although Iliffe™s
(1987) work moved things forward through its contrast of structural and
conjunctural poverty in Africa which went beyond the static poverty analyses
typical of his era.
Since the late 1980s there has been growing interest in examining the
duration of poverty. Economists initially led the way through studies of

Poverty Dynamics

transitory and chronic poverty, poverty dynamics, and patterns of poverty
spells (Bane and Ellwood, 1986; Gaiha, 1988). While these studies have helped
to put duration on the research agenda, their narrow focus on income or
consumption poverty means that they have, at best, only tangentially linked
up with the conceptual advances promoted by Amartya Sen and others (e.g.
Sen, 1985, 1999). 5 This pattern has continued and Hulme and McKay (2008)
report that out of the 28 panel datasets available on developing countries,
26 assess the standard of living in terms of income or consumption and for
23 of these datasets they are the only poverty measures available. Baulch and
Masset (2003) have produced one of the few studies that broaden panel dataset
analysis to human development measures.
We believe that the duration aspect of time merits particular attention for
four main reasons. First, there is a simple logic that says if x has experienced
the same forms and depths of poverty as y, but for a much longer period,
then a moral concern with helping the more disadvantaged requires that x
be prioritized and supported as she or he has experienced more deprivation
than y. 6 Second, a failure to analyse the distribution of spells in poverty
in a population is likely to lead to weak analyses of ˜why™ people are poor
and, potentially, to weak policies. For example, hypothetically two different
countries might have the same scores for the headcount, depth, and severity
of poverty. Apparently, poverty in both of these countries is similar. However,
in the ¬rst country poverty is largely transitory and is a phenomenon that
many of its population experience but only for short durations. In the other,
most of the population are non-poor but a minority are trapped in poverty
for most or all of their lives. In the former country policies need to help those
experiencing short spells of poverty”unemployment insurance and bene¬ts,
reskilling, microcredit, temporary social safety nets, health services. In the
latter, deeper structural problems must be addressed”inclusion of the poor
in access to health and education services, asset redistribution, tackling social
exclusion, and regional infrastructural development.
Third, recent important work (Barrett, 2005; Carter and Barrett, 2006) has
revealed the linkages between the depth of poverty, in terms of material and
social assets, and duration with a focus on household-level poverty traps. The
assumption behind this work is that low levels of assets lead to persistent
poverty (at least in the absence of ¬nancial markets and safety nets), but
a conceptualization is needed that will also permit an analysis of the ways
in which the duration of poverty leads to depleted asset levels. Finally, the

Alkire (2002) and Clark (2002, 2006a) discuss much of the relevant literature.
In effect this is arguing that the breadths, depths, and durations of the deprivations x
and y experience should be multiplied and thus x will score a higher level of deprivation
than y. If this computation were pursued it would be necessary to decide whether duration
was computed as absolute time or relative time, i.e. the proportion of x and y™s lives spent in

Tony Addison, David Hulme, and Ravi Kanbur

duration of time spent in poverty has important implications for individual
or household future strategies. This is in terms of physical and cognitive capa-
bilities and the ways in which past experience shapes the agency (motivation,
preferences, and understandings) of people.

1.3. Methods and Disciplines

Over the past decade, there has been growing interaction between two strands
of, or two approaches to, poverty analysis in developing countries”the qual-
itative and the quantitative. Interaction between these two approaches has
been forced to some extent by the strengthening (in some cases mandated)
requirement by development agencies to expand the traditional quantitative
base of their poverty assessments with a qualitative component. The best-
known cases of this trend are the World Bank™s Poverty Assessments. But
other agencies such as the United Kingdom Department for International
Development have also encouraged and often insisted on the incorporation
of qualitative methods in poverty analysis and development analysis more
generally. While ˜mixed methods™ frameworks have of course been present in
the literature outside of development, and in the academic literature more
generally, it is undoubtedly true that the degree of interest in such methods
for poverty analysis in developing countries has heightened considerably in
the last ten years. There is now a website dedicated to such analysis and a
series of conferences attest to the growing body of work in this area. 7
What exactly is meant by a ˜quantitative™ versus a ˜qualitative™ approach
in the context of poverty analysis? From the discussions reported in Kanbur
(2003), the following are among the key elements characterizing analyses that
the literature recognizes as falling into the ˜quantitative™ category:
r Theinformation base comes from statistically representative income/
expenditure type household surveys (which may also have a wide range
of modules covering other aspects of well-being and activity).
r The questionnaire in these surveys is of a ˜¬xed response™ type, with little
scope for unstructured discussion on the issues.
r Statistical/econometric analysis is carried out to investigate and test
r ˜Neoclassical homo-economicus™ theorizing underlies the development of
hypotheses, interpretation of results, and understanding of causality.

For example, conferences at Cornell in 2001 and at Toronto in 2004, which led to the
publications Kanbur (2003) and Kanbur and Shaffer (2007a), and the conference in Hanoi,
2007. Details are available at <www.q-squared.ca>.

Poverty Dynamics

Similarly, the following seem to be some of the key characteristics of analyses
that fall into the ˜qualitative™ category:
r Unstructured interviews, the outcomes from which are then analysed
with textual analysis methods.
r Related to the above, use of interviews to develop ˜life histories™ of indi-
r Participatory Poverty Analysis, where a community as a whole is helped
to discuss, to de¬ne, and to identify poverty.
r Ethnography, involving immersion of the analyst into the community in
question over a signi¬cant length of time to get a deeper understanding
of the context.
r Related to all of the above, anthropological and sociological theorizing to
understand results and discuss causality.

Three further points can be made on the above characterization. First, notice
that while the quantitative category is relatively uniform, the qualitative cate-
gory is relatively diverse. The unifying (homogenizing) force of the economic
method is felt in the former, while the latter is a battleground across disci-
plines and indeed within disciplines such as anthropology. 8 Second, some
analyses do combine elements of both, and are on a continuum between
the qualitative and the quantitative, rather than being strictly one or the
other. Thus the qualitative“quantitative distinction might best be viewed as a
tendency rather than as a discrete divide. Third, the qualitative“quantitative
divide to some extent aligns with, and to some extent cuts across, disciplinary
divides in poverty analysis, especially as between economics and the other
social sciences.
The advantages and disadvantages of the two types of approaches are
becoming better understood and are well illustrated by Adato™s (2007) mixed-
methods study on assessing conditional cash transfers. The quantitative part
of the appraisal was statistically representative, and addressed econometrically
the dif¬culties in attributing causality to the programme from ˜before and
after™ or ˜with and without™ comparisons. Moreover, it does appear that, at
least to some extent, policy makers tend to put greater weight on statistically
representative ˜large sample™ assessments than on a small number of case
studies. It is now generally accepted that the quantitative assessments of
Mexico™s conditional cash transfer programme played a key role in convinc-
ing a new administration to continue a programme started by the previous

The case for cross-disciplinary research on poverty in the social sciences is discussed at
length elsewhere (Hulme and Toye, 2006). See also Harris (2002), Kanbur (2002), and Clark

Tony Addison, David Hulme, and Ravi Kanbur

However, Adato™s (2007) assessments from the qualitative approach throw
up key issues which policy makers and analysts ignore at their peril. For
example, while the quantitative assessments have generally praised these
programmes for being well targeted to bene¬ciary groups (low ˜leakage™), and
indeed have recommended the tightening up of monitoring to reduce what
leakage there is, the qualitative assessments reveal a great deal of incom-
prehension and resentment on the ground by those who are left out of the
bene¬ciary group, when they see their near neighbours being included. Thus,
whatever the ˜objective™ criteria laid out at the centre and developed through
quantitative surveys and analysis, what is important is the meaning ascribed
to those criteria on the ground. The tensions caused by such factors, identi¬ed
as being serious in the qualitative assessment, could undermine support for
the programme.
This suggests that qualitative approaches are better suited to emphasizing
deeper processes, and the context generating the outcomes revealed by the
study. This is clearly relevant for understanding, and also for the local-level
implementation of policy. That quantitative studies do not (or cannot) do
this is in part the burden of the critique advanced by Harris (this volume) and
du Toit (this volume), who criticize not only quantitative approaches but also
the related economic approaches to measurement and understanding. On the
other hand, whether a phenomenon is widespread, or perhaps only locally
relevant, is better addressed by studies in the quantitative tradition. Statistical
analysis on representative samples is also better suited, for example, to going
beyond ˜before and after™ or ˜with and without™ comparisons of policy or other
events, as revealed by interviews with individuals, no matter how context
The bene¬ts of combining quantitative and qualitative approaches are thus
not to be doubted, and are revealed in a large number of recent studies. 9
Further, as Harriss (2002, p. 494) says, ˜disciplines need to be saved from
themselves™. Effective cross-disciplinarity seeks to capture the ˜productive™
aspects of disciplinarity which ˜produces the conditions for the accumulation
of knowledge and deepening of understanding™ while avoiding the ˜constrain-
ing™ effects of disciplinarity which can lead ˜to the point where it limits
thought . . . and even [becomes] repressive™ (Harriss, 2002, pp. 487“8).
However, this is not to say that there are no problems. While conducting
studies side by side, or making quantitative studies a little more qualitative
(for example, by conducting a participatory appraisal prior to designing the
survey questionnaire, or by adding an unstructured portion at the end of a
questionnaire), or by making the qualitative studies a little more qualitative
(for example, by choosing the sites for the qualitative assessment on the basis
of a national sampling frame, or by generating numerical values from coding

See Kanbur and Shaffer (2007a) and <www.q-squared.ca>.

Poverty Dynamics

of the unstructured interviews), there remain fundamental issues of discipline
and epistemology that will not simply go away. Kanbur and Shaffer (2007b)
identify some deep philosophical issues about different conceptions of the
nature of knowledge in different disciplinary traditions that are bound to
bedevil ˜deep integration™ of the different approaches. For example, it is not
entirely clear that national-level policy making is well served by community-
level measures of poverty which are based on community perceptions of what
it means to be poor. These practical issues also have their roots in whether
poverty can and should be identi¬ed ˜objectively™ by ˜brute data™, or whether it
is inherently to do with intersubjective meanings. Kanbur and Shaffer (2007b)
come out strongly in favour of mixed methods, but caution that there are
pitfalls that we should be aware of.
The above discussion applies to poverty analysis in general. Consider now
an application of the above discussion to poverty dynamics in particular,
and especially to the chapters in this volume. As discussed in the previous
section, time adds novel and irreducible dimensions to the conceptualization,
measurement, and understanding of poverty. For example, the economic
theory of poverty measurement is very well developed for the static case.
Going back at least as far as Sen™s (1976) classic exposition, axioms have
been proposed to capture basic intuitions on what constitutes ˜poverty™ and
˜higher poverty™, and poverty measures that satisfy these axioms have been
described which are now the workhorse of empirical poverty analysis in the
quantitative tradition (for example, the famous FGT measure, Foster, Greer,
and Thorbecke, 1984). But all this is for the static case. The introduction of
time into the economic theory of poverty measurement is relatively recent,
and the chapters by Foster (this volume) and Calvo and Dercon (this volume)
represent the state of the art. The issues that arise hinge on how to aggregate
individual poverty experiences over time, in conjunction with aggregation
across individuals into a poverty measure for the society as a whole. De¬ning
and separating out risk, vulnerability, transient poverty, and chronic poverty
are the concerns of the current economic literature on poverty measurement
(for example the chapters mentioned above and also Günther and Klasen (this
volume) and Carter and Ikegami (this volume).
However, there are signi¬cant conceptual, methodological, and empirical
questions that face the standard economic approach. Empirically, to imple-
ment any of these measures we need surveys of panels of individuals or
households who are followed over time. If the object is to take a medium-
term perspective on time, and especially if we wish to take a longer, inter-
generational or dynastic, perspective, then panels of twenty years or more
are needed by de¬nition. There has been a recent ¬‚owering of panel dataset
collection in a few developing countries. Effective use of this information for
analysis to poverty and well-being dynamics is well illustrated by the review
in Quisumbing (this volume). However, a majority of developing countries

Tony Addison, David Hulme, and Ravi Kanbur

do not have panel data at all, certainly not of the national representative
variety. And no countries have comparable panels over twenty years or more.
Quantitative panel-based analysis on poverty dynamics, therefore, is largely
an analysis of fairly short-run ¬‚uctuations in well-being and poverty, for the
small number of countries that have them.
One way to obtain information about the past when we do not have
actual intertemporal panels is to ask people about their past and record and
utilize this information. This is often done in quantitative analysis (see papers
referred to in Quisumbing, this volume). The method, of probing people about
their past, is related to the life history method in qualitative poverty analysis,
as exempli¬ed by Davis (this volume). Each individual is engaged in a semi-
structured discussion about their life course. The objective is not only to ¬nd
out about the trajectory of well-being, but also the causes underlying it”as
seen by the individual. Some quantitative information can be collected, on
incomes, purchase of assets, value of dowry, etc. But the main focus is on the
narrative and interpretation of the narrative. The ˜stages of progress™ approach
of Krishna (this volume) is also a backward-looking self-assessment, but this
operates at the community level, and there is a stronger push towards present-
ing at least some numerical indicators of changes over time. While there is an
interesting discussion to be had on the relative strengths of individual-focused
versus community-based histories, it is clear that both share the feature of
semi-structured interviews of the qualitative approach, as distinct from the
(largely) ¬xed-response questionnaire method of the quantitative approach.
The contextual detail emerging from the narratives is not something that is
intended to be replicated in standard panel survey instruments. Moreover,
especially if the panel-based survey is, say, every few years (which is the case
for most panels in developing countries), then (apart from the ˜attrition bias™
from people leaving the sample, which quantitative analysts are well aware of)
major twists and turns in the life course will be missed in the panel (except
to the extent that they are re¬‚ected in the next snapshot of the household or
the individual several years later). However, such events can be picked up in
a life history discourse, and put to good analytical use, as is shown in Davis
(this volume).
The chapter by Moser and Felton (this volume) is an interesting amalgam of
the qualitative and quantitative approaches. It combines relatively standard
quantitative information (suf¬cient to allow econometric regressions to be
run) with ethnographic detail and long-term engagement with the commu-
nities studied”over twenty years, in fact. There are of course many anthro-
pologists who have had similar long-term engagement with small numbers
of communities (sometimes only one). But it is unlikely that information
they have collected can be fed directly into quantitative type analysis”nor
would they wish it to be. However, one possibility is to do for analysis of
poverty dynamics what Ostrom (1990) and her colleagues did for analysis of

Poverty Dynamics

the commons, namely build a bridge between qualitative and quantitative
analysis by conducting a textual analysis of the reports and using coding
to generate quantitative measures for further analysis from different perspec-
tives. We leave this as a suggestion and an open question.
Finally, it should be recognized that, unlike in the static case, the com-
bination of qualitative and quantitative approaches, and indeed cross-
disciplinarity, in a single study, or in studying the same speci¬c problem, is
relatively rare. The papers in Kanbur and Shaffer (2007a) bear ample testimony
to how far things have come in the static case. The chapters in this volume,
however, show how far we have to go in poverty dynamics in advancing
mixed-methods approaches. As a collectivity the chapters do highlight the
bene¬ts from combining quantitative and qualitative approaches, as discussed
above. However, except for Moser and Felton, Boyden and Cooper, and Wool-
cock, the chapters are largely in one tradition or the other. It is to be hoped
that the lead given by these chapters, and by recent papers such as Baulch and
Davis (2007) and Lawson (2007), and the bene¬ts of combination shown by
bringing the two traditions together in this volume, will continue in poverty
dynamics the trend that is already well under way in the static analysis of

1.4. Poverty Dynamics: Measurement and Understanding

This volume is divided into three parts. After Part I, which consists of this
introduction and overview, Part II explores poverty measurement and assess-
ment, with a focus on cutting-edge approaches to incorporating poverty
dynamics, using both quantitative and qualitative methods. Part III focuses
upon differing explanatory frameworks for understanding poverty dynamics.
Introducing time into poverty measurement and analysis is a major chal-
lenge, which researchers in developing countries have only begun to really
address over the last decade. Dividing the past into discrete time periods
(˜spells™) for the purpose of measuring living standards is a well-established
practice, often accompanied by analysis of poverty mobility using tools such
as ˜poverty transition matrices™ applied to individuals or groups (Baulch and
Hoddinott, 2000). More ambitious are efforts to develop a single intertemporal
measure of poverty to summarize different poverty paths; the best known
is Ravallion™s chronic poverty measure which uses the average poverty level
(using the FGT poverty measures over the entire period for which (con-
sumption) data is available). However, to derive satisfactory intertemporal
measures we must be very clear about what underlying assumptions are being
made. In particular, should we treat all spells of poverty equally and if not
then how should they be weighted? How do we incorporate risk into the
measure (a big concern of the poor, constantly voiced in surveys), especially

Tony Addison, David Hulme, and Ravi Kanbur

when we are concerned to project poverty forward and individuals cannot be
assumed to have perfect foresight? Much effort has gone into incorporating
vulnerability”the unpredictability and riskiness found in the lives of the
poor”into static poverty measures, but the effort is only just beginning with
dynamic measures (Elbers and Gunning, 2006). Finally, different individuals
and groups will experience different patterns of spells of poverty and non-
poverty; how is this information to be combined?
In Chapter 2 C©sar Calvo and Stefan Dercon argue that existing approaches
have not been explicit enough about their underlying assumptions, and they
set themselves the task of deriving a number of axioms which satisfactory
measures should possess. The axiomatic approach is valuable because it forces
us to be explicit about our values. Thus, do different time periods carry equal
weight? To what extent can a period in poverty be compensated by future
higher income (a key question for assessing the poverty impact of policy
reforms that often generate short-term adjustment costs with the promise of
long-term gains)? And to what extent are you the same person across time (a
question raised by the philosophy of identity)? Calvo and Dercon illustrate
their discussion with a panel of Ethiopian household data, ¬nding substantial
differences between static and intertemporal poverty measures.
One of Calvo and Dercon™s theoretical propositions is likely to be contro-
versial: they reject the notion of time discounting which prevails in other areas
of economics when intertemporal welfare effects are being compared (in the
cost“bene¬t analysis of environmental impacts, for example). Instead, they
appeal to the principle of ˜universalism™ which argues strongly for valuing
distress equally whatever the time period in which it has occurred”a principle
that is used by Anand and Hanson (1997) to reject the use of time discounting
in deriving intertemporal measures of health status. Some may feel that this
goes too far; there is by no means unanimity among health economists as
regards the use of time discounting and there are strong proponents for it
(see Smith and Gravelle, 2000). But those who favour discounting poverty
(as with health) must consider a major dif¬culty: what is to be the rate of
discount? And if it varies across countries (because of differences in rates of
time preference, in their turn in¬‚uenced by cross-country variation in life
expectancies) does this undermine the comparison of intertemporal poverty
across countries? In summary, in seeking to clarify the theoretical basis of
intertemporal poverty measures, Calvo and Dercon open a Pandora™s box of
important issues for future theoretical and empirical research.
The FGT measure (Foster, Greer, and Thorbecke, 1984) has been the most
widely used poverty measure of the last two decades but it takes no account
of duration. Since the length of time in poverty negatively affects outcomes,
especially for children (see Chapter 13 by Boyden and Cooper) this is clearly
a very important missing dimension of poverty measurement. Yet ¬lling this
gap raises major conceptual issues. In Chapter 3 James Foster takes up the

Poverty Dynamics

challenge of introducing time into the measurement of chronic poverty,
speci¬cally by incorporating the duration of time spent in poverty into the
FGT poverty measures. He creates a measure which obeys a number of crucial
axioms and conditions (such as the need for the measure to be subgroup
decomposable) with two cut-off points de¬ning the chronically poor: a stan-
dard (absolute) poverty line and a duration line. As with Calvo and Dercon,
an earlier period in poverty is given the same weight as a later period (i.e.
no time discounting is used). Foster reports on an application of this new
poverty measure to a panel for Argentina, with the duration-adjusted FGT
measure yielding a signi¬cantly different estimate of poverty (with a large
variation in spatial chronic poverty). Foster notes one criticism of this new
measure: it is con¬ned to income. The next step is to create multidimensional,
duration-adjusted measures of chronic poverty, but this is an exceptionally
demanding task (not least in making commensurate the different dimensions
of well-being to construct a single measure). Notwithstanding this remaining
challenge, Foster™s duration-adjusted FGT measure is work that promises to
revolutionize the measurement of poverty dynamics in the way that the
original FGT measures revolutionized static poverty measurement.
Part II offers a spectrum of different dimensions of well-being and poverty.
Chapter 4 analyses the dynamics of non-income poverty measures which are
as important as those of income and consumption measures of changes in
poverty status. Working within a capabilities framework, Isabel Günther and
Stephan Klasen analyse nutrition, health, and education poverty indicators
for Vietnamese panel data, selecting households with at least two generations
present. They argue that non-income indicators can be as good (and some-
times better) as income at capturing intergenerational poverty transmission:
income tells only part of the story. Vietnam is especially relevant since the
economy is experiencing fast growth and structural change. There has been
a sharp decline in income poverty, but nutrition and health indicators show
fewer households escaping from poverty (overall there is a lower correlation
between non-income and income measures than one would expect). Günther
and Klasen ¬nd intergenerational education poverty remains particularly
strong: many households with low education among the older generation also
have low education among the young.
In Chapter 5 Caroline Moser and Andrew Felton apply a principal com-
ponents analysis to panel data from urban Ecuador (collected over 1978“
2004) and construct an asset index to measure asset accumulation. They
inductively construct the index on the basis of longitudinal anthropolog-
ical research (rather than building an index and then applying it to the
data), a methodology they term ˜narrative econometrics™. Moser and Felton
argue that it is imperative to understand the social context of assets and
how they vary in their importance; simply plugging assets into an index
is highly unsatisfactory. Their chosen assets are: physical capital (including

Tony Addison, David Hulme, and Ravi Kanbur

housing); ¬nancial/productive capital; human capital and social capital (nat-
ural capital is not included as this is an urban study). Different asset indices
deploy different weighting methodologies and the three most common are:
weighting by asset prices (but these are dif¬cult to obtain and it is hard to
impute a price for non-marketed assets); equal weights (which has obvious
problems since it assumes all assets have equal value (a computer and a
horse for example); and principal components analysis (using correlations to
estimate the underlying unobservable variable, following Filmer and Pritchett,
2001). Moser and Felton adopt the latter. The distribution of each type of
capital is then calculated over different points in time to highlight asset
Crucially the importance of assets can vary over time due to structural
changes in the economy as well as economic policy which affect the returns
to speci¬c assets (asset indices can be used to identify the effect of macro-
economic shocks). Thus Guayaquil has seen large changes in labour demand
due to globalization; imports of cheap Chinese-made goods have reduced
the demand for artisanal male skills which provided a reasonable income in
the 1970s. And the shift from community-based services to market-provided
services (the result of privatization) is showing up in changes in social capital
at the community level.
Assets are a long-running theme in the poverty debate from the 1970s
paradigm of ˜redistribution of growth™ (Chenery et al., 1974) through to the
WDR-2000 and WDR-2005 policy discussions, and in livelihood approaches
to poverty analysis (Scoones, 1998; Ellis, 2000). Assets (stocks) generate income
and consumption ¬‚ows (and stocks may be more easily measured than ¬‚ows);
they enable households to withstand shocks (within limits); the level and
composition of assets determines whether a household is in a poverty trap
(and its chances of escape); and helping the poor to build assets (including
human capital) has policy traction”although there is much debate about
which assets are the most important in the many livelihood contexts that
the poor face (Hulme, Moore, and Shepherd, 2001). Fundamentally, assets
bring the production dimension into poverty measurement, adding to the
income, consumption, and human development dimensions (and telling us
more about how levels of these latter three dimensions arise in households).
The methodology of assets-based approaches has become increasingly
sophisticated (and is (panel) data intensive), particularly in incorporating time
into the formal models to address a key question: who among the currently
poor are likely to be poor in the future? Dynamics are therefore centre stage
in this approach, with a theory of poverty traps underlying empirical appli-
cations (Buera, 2005; Carter and Barrett, 2006). Drawing upon this recent
literature, and in a model applied to data from KwaZulu-Natal, Michael Carter
and Munenobu Ikegami (Chapter 6) introduce new theory-based measures of
chronic poverty and vulnerability and illustrate their feasibility using South

Poverty Dynamics

African data. They identify three types of poor people each with different
future prospects: (i) the low-skilled with few livelihood possibilities who are
in a low-level equilibrium trap (the Economically Disabled); (ii) a middle-ability
group that will move either up or down the income scale depending upon
their initial asset level (the Multiple Equilibrium Poor); and (iii) a high-ability
group who can move out of poverty given enough time (the Upwardly Mobile).
Forward-looking measures of poverty are then derived. In the FGT measure
poverty is measured using an income gap, but it is possible to see poverty as
an asset gap as well, and this is what Carter and Ikegami do, to calculate the
percentage of people who will stay poor under different assumptions of asset
dynamics. Asset shocks are then simulated in this model, with individuals
reacting to the risk of shocks by, for example, being unwilling to forgo present
consumption in order to accumulate assets that they may well lose. Different
policy recommendations are developed for each group. The economically
disabled are candidates for social protection while the middle-ability group
need protection to reduce their risk and asset transfers to put them over the
asset threshold and to give them a ¬ghting chance of exiting poverty.
The ¬nal two chapters in Part II focus on the measurement and assessment
of poverty through subjective approaches. Peter Davis™s chapter examines the
role of individual/household life history methods in assessing poverty dynam-
ics, while Anirudh Krishna uses participatory methods to assess changes in
poverty and well-being at the community level.
By providing contextual and historical detail, life histories constitute a
valuable complement to quantitative approaches. Peter Davis (Chapter 7)
demonstrates their ability to reveal phenomena concealed by other methods,
including: events with multiple causation; ˜last-straw™ threshold effects (the
culmination of a series of adverse trends); outcomes based on the ordering of
a sequence of events; and events associated with household breakdown which
tend to be masked in household survey approaches. For rural Bangladesh,
Davis constructs household resource pro¬les before conducting the life history
interviews and seeks out a high level of historical and contextual detail (both
Davis and Krishna use ˜referencing™”mapping events and changes at the
household level to a template of easily recalled national events). Davis ¬nds
that most improvements tend to only happen gradually, whereas declines are
often more sudden (in this Davis™s work links to that of Paul Farmer, 2005,
on the structural violence that poor people face) and he develops trajectory
patterns: saw-tooth patterns in trajectories are more common among the poor
than smooth paths. Davis also deploys the methodology of ˜fuzzy sets™ in
identifying chronic poverty. However, the very richness of life histories means
that the number of cases studied is generally small, limiting generalization
across larger populations.
Whereas Peter Davis focuses on one country (Bangladesh), Anirudh Krishna
(Chapter 8) tracks households in ¬ve countries: four developing countries and

Tony Addison, David Hulme, and Ravi Kanbur

the United States. He aims to capture poverty dynamics through a ˜stages
of progress™ methodology. This has seven steps: (i) get together representa-
tive community group; (ii) discuss the objectives of the exercise; (iii) de¬ne
poverty collectively in terms of stages of progress, then ask the question: if
a poor household gets a bit more money what do they do with it? (typically
they specify food for the family as their ¬rst priority); (iv) de¬ne ˜x years ago™
in terms of a well-known signifying event; (v) list all village households, and
then ask about each household™s stage at the present time and x years ago;
(vi) categorize all present-day households into chronically poor or not; and
then (vii) take a random sample within each category to ascertain reasons for
change or stability. To cross-check the reliability of the method the researchers
share the results with key informants, before leaving the community, to see
whether they agree with the ¬ndings.
Krishna ¬nds that health and healthcare expenses were a primary event in
the descent into poverty (41 per cent of cases in North Carolina and 88 per
cent in Gujarat, India). Other reasons were more context speci¬c: funerals and
marriage (important in four countries), debt (important in India), drought,
and loss of land (Uganda and Peru). Among the reasons for successful escape
from poverty, interviewees cited a supplemental income source (mainly city-
based informal sector) as the most important.
As will have become clear, many of the chapters in Part II, although primar-
ily about measurement, also address the understanding of poverty dynamics.
Indeed, in some of the chapters measurement is a route into understanding,
so a simple division between the two is not possible. Similarly, the chapters
in Part III, although primarily about understanding, also broach questions of
Part III begins with two chapters that offer a critique of measurement.
In Chapter 9 Harriss argues that most poverty research is working with
a model of knowledge from the natural sciences: that is to say, there are
objective facts to be discovered; methods for uncovering these facts improve
over time as better techniques are discovered and employed; and, predictive
theories that can be universally applied across all societies will eventually
emerge. But this approach is doomed to disappointment, argues Harriss, for
the focus is on measurement and on the characteristics of individuals and
households with very little attention to the structural processes that move
people in and out of poverty. Numerous studies identify the same set of
factors (assets, household characteristics, demographics) as being associated
with poverty dynamics, yet these are proximate factors only. This supposedly
˜value-neutral™ approach depoliticizes poverty. Harriss highlights similarities
between the new asset-based approaches and research during the 1970s on
agrarian differentiation and class formation”although the social context is
much less explicit in the former (a consequence of the household being the
primary unit of analysis). Thus a key but outstanding question is why the

Poverty Dynamics

poor come to have so few assets and the role of wealthy elites in blocking
their asset-accumulation strategies (including historical and contemporary
expropriation). Clearly, there is considerable scope for qualitative research to
inform quantitative data collection in this area.
In Chapter 10 Andries du Toit emphasizes the need to engage with the
structural dimensions of persistent poverty and therefore with social relations,
agency, culture, and subjectivity. He illustrates his argument with examples
from South Africa. While welcoming the recent dialogue between quantitative
and qualitative research, he emphasizes the need to go beyond the positivist
assumptions underlying econometric approaches which at their worst consti-
tute a ˜mystifying narrative™ of what poverty means and how we come to
understand it. Drawing on the work of James Scott (1998) and others, du
Toit argues that the process of abstraction in poverty measurement results
in a decontextualization of poverty; certain information (that which can be
standardized and quanti¬ed) is given preference in building a narrative of the
poor and the processes that result in impoverishment. In South Africa, govern-
ment of¬cials have become ¬xated on ¬nding unambiguous and quanti¬able
systems of indicators of structural vulnerability, to the detriment of really
understanding the role of national and local history and power relations.
By focusing on what is readily measurable at the individual and household
level, measurement approaches neglect culture, identity, agency, and social
structure that are central to creating wealth and poverty (see Chambers, 1983;
Bevan, 2004) and the policy conclusions do not connect to the realities of
poor societies.
The next two chapters offer attempts at understanding poverty dynam-
ics within a recognizably economics/quantitative framework. In Chapter 11
Siddiqur Osmani develops a dynamic approach to capabilities; people may
develop or lose speci¬c capabilities over time, and their opportunities are
often changing as economic change favours some skills, and downgrades
others. Poverty traps for both households and individuals then result from a
mismatch between the structure of endowments and the structure of opportu-
nities. Osmani contrasts the roles of level and structure of assets/endowments
in explaining chronic poverty. Chronic poverty has an inherent time dimen-
sion, but the analysis to date is insuf¬ciently explicit”for example how
long do people have to be poor to be categorized as chronically poor? Most
discussion adopts a backward-looking approach, whereas in Osmani™s view
we need to be more forward looking”someone is in a poverty trap indef-
initely unless something changes for the better. Since even a chronically
poor person can move above the poverty line, the key point is that for most
of the time a chronically poor person is below the poverty line”unable to
accumulate to get out during their working lifetime. He develops a de¬nition
of chronic poverty with expected income as its core, with expected income
in turn conditional on the expected accumulation of assets over time as well

Tony Addison, David Hulme, and Ravi Kanbur

as initial exogenous circumstances. If that conditional expected income lies
below the poverty line then that person is chronically poor. With limited
endowments a person can be chronically poor without being caught in a
poverty trap (for the fortunate their income may be on a time path to move
them out of poverty even if they are chronically poor at present). For policy
it is then essential to look at the pattern of growth and not just its rate, for
the former restructures the pattern of opportunities, devaluing some initial
asset investments while raising the returns on others (as will economic policy
change, for example market liberalization). Targeted interventions to improv-
ing endowments and putting people on upward accumulation paths out of
chronic poverty must take account of the changing pattern of growth. Assets
are also socially constructed (a theme echoed by Maia Green) and a mismatch
between endowments and opportunities can arise when social relations, not
just the economy, change.
A key assumption in existing models is that individuals cannot borrow
against their future earnings to build present assets (which in turn yield higher
(future) income ¬‚ows) and must save instead. A threshold of initial assets
exists below which accumulation through saving is not a viable strategy for
moving out of poverty and, with a binding credit constraint, the household
cannot become a successful entrepreneur”even if it has the skills and knowl-
edge to do so. 10
Conceptually, many different types of asset have been identi¬ed: natural
capital, physical capital, human capital, social capital, and ¬nancial capi-
tal, with further re¬nements within each (for example Hulme, Moore, and
Shepherd (2001) divide social capital into socio-cultural and socio-political
assets). But data on human capital and physical capital are the most readily
available, and in Chapter 12 Agnes Quisumbing focuses on these in an ana-
lytical survey of how intergenerational asset transfers can create (or block off)
escape routes from poverty. The poor are typically constrained in their ability
to trade off present for future consumption (exacerbated by credit constraints)
and an inability to invest in human capital persists across generations (there
is plenty of evidence from the Philippines that the children of parents with
little schooling and/or assets have lower school participation, and the children
of credit-constrained households are shorter than those of unconstrained
Quisumbing argues that context matters greatly in determining which
assets work best for poverty reduction. Thus in Ghana more land is better
for increasing women™s income than more education given the low returns to

There is a growing literature on modelling credit constraints; using United States data,
Buera (2006) ¬nds that the welfare cost of such constraints is signi¬cant (about 6% of the
household™s lifetime consumption), and there is clearly much scope for applying these tools
to simulate the impact of micro¬nance on future poverty trajectories.

Poverty Dynamics

female schooling in rural labour markets. If asset accumulation takes time and
is dif¬cult for the poor, then assets at marriage largely determine lifetime pros-
perity. The marriage market therefore plays a central role, and evidence from
Ethiopia shows that assortative mating increases inequality and reduces social
mobility (due to intergenerational transfers at marriage)”thereby continuing
social strati¬cation from one generation to the next. For the poor to transfer
assets across generations they must ¬rst accumulate them; hence the need to
strengthen property rights, reduce the initial costs of acquiring capital, and
provide savings instruments (and provide mechanisms to maintain the poor
people™s asset base in the face of shocks). More mechanisms for human capital
investment by the credit constrained are essential (Mexico™s PROGRESA is a
An alternative disciplinary approach is presented by Jo Boyden and
Elizabeth Cooper (Chapter 13) to address the concept of ˜resilience™ in research
and practice concerning children™s poverty and the life course and intergener-
ational transmission of poverty. ˜Resilience™ means the strategies that people
use to cope with adversities, such as income poverty or violent con¬‚ict. For
children much attention has been paid to the issue of whether they can in
some way overcome initial disadvantages. Unfortunately children are more
susceptible to the effects of poverty than adults, particularly to the effect of
under-nutrition. Boyden and Cooper argue that while super¬cially attractive
the resilience concept has not yet proved to be a useful tool for poverty
research. Resilience lacks a satisfactory de¬nition, it is impossible to observe
directly, and indeed the concept disguises multivariate phenomena. Thus the
correlation between inputs (mother™s education, for example) and outputs
(child health, for example) are derived from the analysis of datasets that cover
many different parental and community characteristics. In short, research in
this area has been highly mechanistic (prematurely identifying direct cause
and effect), thereby failing to take account of moderating forces. Moreover,
what is often taken for granted in the policy debate is not borne out by
recent research; for example, current research challenges assumptions about
the foundational role of the family in child development. Static models of
human development often underpin the conventional wisdom on the effects
of deprivation in early childhood, whereas more dynamic approaches are
called for in which child-development trajectories are constantly modi¬ed
(implying that it is better to speak in terms of probabilities). Boyden and
Cooper argue that much more attention must be given to the interaction
of genetic and environmental impacts on poverty, as well as the structural
In much of the analysis of assets there have been attempts to understand
the social context that gives assets their value, a point emphasized by Moser
and Felton (this volume) and further developed in Chapter 14 by Maia

Tony Addison, David Hulme, and Ravi Kanbur

Green who argues that the ˜mystery of capital™ lies in social relationships;
hence entitlements do not exist in the abstract but within networks of moral
relationships. The latter determine what different categories of people can
expect. Most importantly, these categories can shift radically. Building on
Barbara Harris-White™s (2005) work on social exclusion, Green argues that
social ordering sanctions harm to some, but not to others, illustrating this
point with an examination of witchcraft in contemporary Africa which is used
to change relationships within and between families (including control over
assets and the value attached to them). In Green™s view the concept of chronic
poverty usefully highlights a situation but does not really explain it, tending
to yield frameworks that are far from local conceptions of poverty, and local
concerns. To get deeper insights we need to develop the idea of durable
poverty (based on deprivation) rather than chronic poverty, for the former
concept is better able to handle the institutional factors that keep people
The idea of the multidimensionality of poverty is now ¬rmly embedded
in the policy discourse, and we have already discussed non-income poverty
dynamics in the contribution by Günther and Klasen to this volume. Yet there
is still much to do. In Chapter 15, Michael Woolcock highlights how the need
for a broader social theory of chronic poverty must look to systems of social
relations, rules, and meaning. Thus understanding how groups are de¬ned
is key to a better understanding of the social relations that underlie chronic
poverty (a point also made by Maia Green). Rules systems, which constitute
everything from constitutions and contracts to languages and social norms,
can lie at the heart of ˜legal inequality traps™ that condemn people to chronic
poverty. A better understanding of meaning systems (how people make sense
of what happens in the world and to them) is essential to deepen our
knowledge of chronic poverty since groups can sometimes subvert practices

. 1
( 13)