ESSAYS ON RISK AND SOCIAL VISIBILITY: INSURANCE, ASSET POVERTY

AND INTRAHOUSEHOLD HEALTH INEQUALITY

A Dissertation

Presented to the Faculty of the Graduate School

of Cornell University

in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

by

Jacqueline Vanderpuye-Orgle

January 2008

© 2008 Jacqueline Vanderpuye-Orgle

ESSAYS ON RISK AND SOCIAL VISIBILITY: INSURANCE, ASSET POVERTY

AND INTRAHOUSEHOLD HEALTH INEQUALITY

Jacqueline Vanderpuye-Orgle, Ph.D.

Cornell University 2008

Risk management is intrinsic to the livelihood choices of people in developing

countries. The effects of exposure to risk and attendant risk coping mechanisms could

have dire implications for individual well-being and economic development as a

whole. This study examines access to social insurance and the Pareto efficiency of risk

pooling. It also examines the existence of asset poverty traps and intrahousehold

inverse-U shaped Kuznets curves as well as the implications of access to insurance for

asset dynamics and intrahousehold health inequality.

We identify a distinct subpopulation of socially invisible individuals who tend

to be younger, poorer, engaged in farming, recent arrivals into the village who have

been fostered, and not a member of a major clan. While the socially visible enjoy

Pareto efficient insurance both at the network and village levels, risk pooling fails for

the socially invisible subpopulation. In addition, whereas asset accumulation patterns

of the socially visible exhibit multiple stable dynamic equilibria, the socially invisible

converge towards a unique dynamic stable equilibrium at the lower asset level. The

asset levels at the respective equilibria were generally higher for the socially invisible

as compared to the socially visible, particularly for the unstable dynamic equilibrium.

The latter implies that the socially invisible are more likely to fall into poverty traps

since there is a broader asset domain over which people in this subpopulation would

collapse toward the lower level asset equilibrium.

Multivariate analyses show that shocks increase the likelihood of a socially

invisible individual falling into a poverty trap. On the other hand, shocks have a

significant effect on the likelihood of a socially visible individual falling into a poverty

trap; this is consistent with the observation that the socially visible enjoy complete risk

pooling. The analyses of intrahousehold inequality indicate that whilst the inverted U-

shape may not exist when using aggregated data, it may exist for the socially visible.

In addition the composition of inequality varies by social visibility. Patterns in relative

BMI within the household suggest that the socially invisible may overcompensate for

risk exposure by protecting children in the allocation of limited food and nutritional

resources.

BIOGRAPHICAL SKETCH

Jacqueline Vanderpuye-Orgle was born in Accra, Ghana. She graduated from

the University of Ghana in 1999, with a Bachelor of Science (First Class Honors)

Degree in Agricultural Economics and was awarded the Best Undergraduate

Dissertation (Agriculture) for her thesis: “Econometric Modeling of the Demand and

Supply of Agricultural Credit by the Agricultural Development Bank in Ghana, 1970-

97”. She was one of two recipients of the African University Pre-doctoral Fellowship

in Economics award. She spent a year at Yale University taking courses in Economics,

French and Mathematics. Jacqueline then worked as a Summer Intern with the

Macroeconomics and Public Policy Divisions of the Development Research Group at

the World Bank, after which she went to Cornell University to pursue a Masters in

Applied Economics and a Ph.D. in Policy Analysis and Management.

iii

With love to my mother, Comfort and my siblings,

Yacoba and Bertrand

iv

ACKNOWLEDGEMENTS

I thank the Lord Almighty for His grace and faithfulness. I am greatly indebted

to David Sahn, my committee chairperson, Chris Barrett, Ravi Kanbur and Chris

Udry, my committee members for their advice, guidance and most of all, their

patience and unwavering support throughout my graduate studies.

I extend my gratitude to Ernest Aryeetey and the research staff at the Institute

of Statistical, Social and Economic Research for providing me with an enabling work

environment during my field research. I would like to thank Geysa and the rest of the

administrative staff in the Department of Policy Analysis and Management for their

support. I am grateful to my friends Naa Dei, Abena, Naalamle, Dela, Dola and the

rest of the church family for providing me with support, love and prayers through this

journey.

Finally, I thank my mother, Comfort Vanderpuye-Orgle for inspiring me to

reach for the stars and for believing in me. I thank my sister Yacoba for the sacrifices

she made, my brother Bertrand for his support as well as Kpakpo, Kukua, Papa Kofi

and Nii Akwei for urging me on. I also thank Niilante Amissah for his support,

encouragement and strength. You gave me hope in the challenging moments and for

that I am eternally grateful.

v

TABLE OF CONTENTS

Biographical Sketch iii

Dedications iv

Acknowledgements v

List of Figures ix

List of Tables xiii

Chapter 1 Risk Management, Social Visibility, Asset Poverty and

Intrahousehold Health Inequality: An Overview 1 1.1 Introduction 1 1.1.1 Risk Management 1 1.1.2 Social Networks and Social Visibility 2 1.1.3 Poverty Dynamics and Asset Poverty Traps 5 1.1.4 Intrahousehold Health Inequality 8 1.2 The Data 10 1.3 Dissertation Chapters 11 1.3.1 Risk Management and Social Visibility in Ghana 11 1.3.2 Risk, Asset Poverty and Social Visibility in Ghana 12 1.3.3 Risk, Intrahousehold Health Inequality and Social

Visibility in Ghana 13 1.4 Conclusions 14 References 17 Chapter 2 Risk Management and Social Visibility in Ghana 25

2.1 Introduction 25 2.2 Risk Pooling and Social Visibility: Simple Theoretical Predictions 29 2.3 The Data 32 2.3.1 Individuals’ Social Networks 33 2.3.2 Consumption 34 2.3.3 Shocks 35

vi

2.4 Social Visibility 38 2.5 Risk Pooling Conditional on Social Visibility 46 2.6 Conclusions 55 References 57

Chapter 3 Risk, Asset Poverty and Social Visibility in Ghana 60

3.1 Introduction 60 3.2 The Data 65 3.2.1 Individuals’ Assets 66 3.2.2 Individuals’ Social Visibility 66 3.2.3 Shocks 67

3.2.4 Consumption 67

3.3 Asset Index 68 3.4 Asset Poverty Dynamics 71 3.5 Who is Likely to Remain Poor? 85 3.6 Conclusions 88 References 92

Chapter 4 Risk, Intrahousehold Health Inequality and Social Visibility in Ghana 96

4.1 Introduction 96 4.2 Analytical Framework 100 4.2.1 Bivariate Analysis of Intrahousehold Inequality 100 4.2.2 Determinants of Intrahousehold Health Inequality 102 4.3 The Data 104 4.3.1 Anthropometrics 105

4.3.2 Social Visibility 105 4.3.3 Shocks 106 4.3.4 Other Sample Characteristics 107

vii

4.4 Empirical Model and Results 107 4.4.1 Analysis of Intrahousehold Inequality 107 4.4.2 Determinants of Intrahousehold Inequality 124 4.5 Conclusions 129 References 134

Appendix A Field Notes 139 A.1 Field Activities 139 A.2 Preliminary Visit to Survey Area 140 A.3 Data Collection Trip 142 Appendix B Survey Modules and Interview Schedule 146 Appendix C Questionnaires 152

viii

LIST OF FIGURES

3.1 Nonparametric Estimates of Predicted Asset Index, All Individuals 1997-1998 72 3.2 Nonparametric Estimates of Predicted Asset Index, All Individuals 1997-2004 73 3.3 Nonparametric Estimates of Predicted Asset Index, All Individuals 1998-2004 74 3.4 Nonparametric Estimates of Predicted Asset Index: Social Visibility>0, 1997-1998 75 3.5 Nonparametric Estimates of Predicted Asset Index: Social Visibility=0, 1997-1998 75 3.6 Nonparametric Estimates of Predicted Asset Index: Social Visibility>0, 1997-2004 76 3.7 Nonparametric Estimates of Predicted Asset Index: Social Visibility=0, 1997-2004 77 3.8 Nonparametric Estimates of Predicted Asset Index: Social Visibility>0, 1998-2004 77 3.9 Nonparametric Estimates of Predicted Asset Index: Social Visibility=0, 1998-2004 78 3.10 Asset Index Regression Tree 79 3.11 Nonparametric Estimates of Predicted Asset Index: Social Visibility≥0.793, 1997-1998 81 3.12 Nonparametric Estimates of Predicted Asset Index: Social Visibility=0.793, 1997-1998 81 3.13 Nonparametric Estimates of Predicted Asset Index: Social Visibility≥0.793, 1997-2004 82 3.14 Nonparametric Estimates of Predicted Asset Index: Social Visibility=0.793, 1997-2004 82 3.15 Nonparametric Estimates of Predicted Asset Index: Social Visibility≥0.793, 1998-2004 83

ix

3.16 Nonparametric Estimates of Predicted Asset Index: Social Visibility=0.793, 1998-2004 83 3.17 Nonparametric Estimates of Predicted Asset Index: Social Visibility<0.0955, 1997-1998 84 3.18 Nonparametric Estimates of Predicted Asset Index: 0.0955>Social Visibility≥0.793, 1997-1998 84 4.1 Nonparametric Estimates of Intrahousehold Inequality and Mean

Household BMI, All Individuals 108

4.2 Nonparametric Estimates of Intrahousehold Inequality and Household Expenditure Percentile, All Individuals 108

4.3 Intrahousehold Inequality Regression Tree 109 4.4 Nonparametric Estimates of Intrahousehold Inequality and Mean

Household BMI, Socially Visible Households 110

4.5 Nonparametric Estimates of Intrahousehold Inequality and Household Expenditure Percentile, Socially Visible Households 111

4.6 Nonparametric Estimates of Intrahousehold Inequality and Mean Household BMI, Socially Invisible Households 111

4.7 Nonparametric Estimates of Intrahousehold Inequality and Household

Expenditure Percentile, Socially Invisible Households 112

4.8 Nonparametric Estimates of Male to Child BMI and Mean Household BMI, All Households 115

4.9 Nonparametric Estimates of Male to Child BMI and Household

Expenditure Percentile, All Households 116 4.10 Nonparametric Estimates of Female to Child BMI and Mean Household

BMI, All Households 116 4.11 Nonparametric Estimates of Female to Child BMI and Household

Expenditure Percentile, All Households 117 4.12 Nonparametric Estimates of Male to Female BMI and Mean Household

BMI, All Households 117

x

4.13 Nonparametric Estimates of Male to Female BMI and Household Expenditure Percentile, All Households 118

4.14 Nonparametric Estimates of Male to Child BMI and Mean Household

BMI, Socially Visible Households 118 4.15 Nonparametric Estimates of Male to Child BMI and Household

Expenditure Percentile, Socially Visible Households 119 4.16 Nonparametric Estimates of Female to Child BMI and Mean Household

BMI, Socially Visible Households 119 4.17 Nonparametric Estimates of Female to Child BMI and Household

Expenditure Percentile, Socially Visible Households 120 4.18 Nonparametric Estimates of Male to Female BMI and Mean Household

BMI, Socially Visible Households 120 4.19 Nonparametric Estimates of Male to Female BMI and Household

Expenditure Percentile, Socially Visible Households 121 4.20 Nonparametric Estimates of Male to Child BMI and Mean Household

BMI, Socially Invisible Households 121 4.21 Nonparametric Estimates of Male to Child BMI and Household

Expenditure Percentile, Socially Invisible Households 122 4.22 Nonparametric Estimates of Female to Child BMI and Mean Household

BMI, Socially Invisible Households 122 4.23 Nonparametric Estimates of Female to Child BMI and Household

Expenditure Percentile, Socially Invisible Households 123 4.24 Nonparametric Estimates of Male to Female BMI and Mean Household

BMI, Socially Invisible Households 124 4.25 Nonparametric Estimates of Male to Female BMI and Household

Expenditure Percentile, Socially Invisible Households 125

xi

LIST OF TABLES

2.1 Individual Consumption Expenditures, June 2004- January 2005

(Nominal Cedis Per Month) 36 2.2 Percentage of Individuals Affected by Shocks and Mean Response

Expenditure/ Imputed Value of Damage (Nominal Cedis) 36 2.3 Summary Statistics on Random Matching 39 2.4 Descriptive Statistics: Estimating Social Visibility 40 2.5 Probit Estimation of Social Visibility 42 2.6 Probit Estimation of the Likelihood of Knowing a Random Match 45 2.7 Regression Tests for Full and No Risk Pooling 49 2.8 Risk Pooling for an Individual in Hid Directly Elicited Social Network 53 2.9 Insurance Among Social Networks within an Village 54 2.10 Summary results for Tests for Risk Pooling 55 3.1 Descriptive Statistics: Estimating Livelihood-Weighted Index 69 3.2 Livelihood-Weighted Index Estimation 70 3.3 Asset Index Regression Tree Estimation 80 3.4 Descriptive Statistics: Estimating Likelihood of Poverty Traps 86 3.5 Poverty Trap Probit Estimation, Social Visibility=0 and Social

Visibility>0 87 3.6 Poverty Trap Probit Estimation, Social Visibility≤0.7929 and Social

Visibility>0.7929 89 4.1 Comparisons of Test Points of Intrahousehold Inequality 112 4.2 Parametric Estimates of Intrahousehold Inequality by Mean Household BMI 113

xii

4.3 Parametric Estimates of Intrahousehold Inequality by Household Expenditure Percentile 113

4.4 Decomposition of Overall Inequality (%) 124 4.5 Descriptive Statistics: Intrahousehold Health Inequality Model 127 4.6 Intrahousehold Health Inequality Naïve Regressions 128 4.7 Intrahousehold Health Inequality Reduced Form Model 130 4.8 Intrahousehold Health Inequality First Stage Regression 131 4.9 Intrahousehold Health Inequality Second Stage Regression 132

xiii

1

CHAPTER ONE

Risk Management, Social Visibility, Asset Poverty and Intrahousehold

Health Inequality: An Overview

1.1 Introduction

Risk management is intrinsic to the livelihood choices of people in

developing countries. The effects of exposure to risk and attendant risk coping

mechanisms could have dire implications for individual well-being and

economic development as a whole. What is the extent of risk exposure in

developing societies? In other words, who is insured and who is not? How

does this affect poverty dynamics, particularly asset poverty? What are the

implications for intrahousehold distribution of welfare? The study seeks to

address these issues. To do so, we draw on several threads in the literature.

1.1.1 Risk Management

The first is the literature on risk management. Several risk coping

mechanisms have been identified in the literature. Risk mitigation may be

through savings in cash or kind (Deaton 1992, Fafchamps, et al. 1998, Paxson

and Chaudhuri 1994, Paxson 1992, Rosenzweig and Wolpin 1993), crop and

plot diversification (Deaton 1989, McCloskey 1976, Townsend 1994, Walker

and Ryan 1990, Walker, et al. 1983), investment in human capital or migration

(Paulson 1994). Once adverse shocks hit, they engage in sale of productive

assets (Hoddinott 2004), increasing labor supply (Bardhan 1983, Kochar 1993,

Rose 1994, Townsend 1995), reducing nutritional intake, pulling children out

2

of school and neglecting the health needs of children (Alderman and Paxson

1992, Ellis 1998). 1

Studies have shown that these insurance mechanisms are costly in

terms of efficiency losses (Coate and Ravallion 1993, Morduch 1999, Barrett

2005). In addition, these tools may not be available to some of these

households. Risk sharing with other households may be the only option. In the

absence of formal financial institutions and public provisions for risk

management, households often resort to informal risk sharing arrangements.

1.1.2 Social Networks and Social Visibility

Social networks have been identified as loci of risk sharing. These

networks reflect an individual’s level of social connectedness via extant social

ties; they reflect the extent to which an individual is known by others, in other

words the extent to which an individual is socially visible. These networks are

fostered by kinship ties, ethnicity, geographical proximity, religion, and gender

groups, inter alia (Goldstein 1999, Santos and Barrett 2004, Udry and Conley

2005, De Weerdt 2005). However, empirical tests often reject the provision of

full insurance by these social networks (Deaton 1992, Townsend 1994, Gertler

and Gruber 1997).2

There are two caveats to be noted: First, networks are often considered

as exogenous institutions; usually some clearly identifiable group such as the

whole village or clan. One could argue that the hypothesis of full insurance

will necessarily be rejected since these tests are conducted for exogenously

1 Especially those of girls. 2 Issues pertaining to measurement error, asymmetric information and contract enforcement have been identified as possible reasons (Alderman and Paxson 1992, Murgai et al. 2001).

3

given groups such as the entire village, community, clan or household and

hence may not adequately capture the true domain and scope of risk sharing.

Extant studies on risk pooling have focused on exogenously given

groups such as the households and villages or entire community as against

within the organizations/ networks. Townsend (1994) noted that “village

economies satisfy the explicit or implicit conditions of general equilibrium

modeling, namely that individuals in the entire community can arrange their

institutions and allocations in such a way as to achieve Pareto optimum. Many

families have been present for generations; many contemporary residents live,

eat and work in the village; the villages have their own legal systems replete

with contract enforcement mechanisms; and village residents may have

relatively good information about the ability, effort and outputs of one another.

Moreover, residents of poor, high risk villages have a collective incentive to

come up with good arrangements: the absence of these can be life threatening.”

This indeed provides some justification for the use of villages as units of study

and, a fortiori, endogenous social networks. The village (or clan or household,

for that matter) may not adequately capture the true mapping of the respective

endogenous network links.

“Theory suggest that the formation of insurance links depends on a

myriad of factors related to smooth information flows, norms, trusts, the

ability to punish, discount rates, group size and the potential gains of

cooperation” (De Weerdt 2005). It is expected that individuals with disparate

income trajectories as well as those with markedly varied access to

technologies, risk aversion or consumption needs are likely to pool risk.

Examining the geographic location and various network links between

respondents from a sample in Ghana, Udry and Conley (2004) observed that

4

there are significant overlaps between different networks in a given village –

credit and information links were most closely associated. In addition,

individuals who appear connected in one dimension appear more likely to be

connected along others. There were, however, some individuals in each village

for each network who seem isolated in these graphs. This could be a

misleading result of using ‘ego-centric’ data as concluded by Udry or Conley

(2004) or it could very well be a true reflection of the fact that these

individuals, for one reason or another, have been selected out of these

networks and are thus isolated, at least at the intra-village level. Some marginal

groups of individuals may be excluded from these networks. Evidence showed

that wealth and other socioeconomic status variables matter: the poor have less

dense risk sharing networks than the rich (De Weerdt 2005).

Secondly, most of the extant studies were undertaken at the household

level, with the implicit assumption of a unitary household model -- assuming

perfect substitutability among decision-makers as well as pooling of all

resources within the household. There is ample evidence in the literature

rejecting these assumptions.3 For instance, Goldstein (1999) rejects full risk

pooling within rural households in Ghana. In fact, he observes that husbands

and wives may have different risk sharing groups. In a more recent study,

Duflo and Udry (2004) examined resource allocation and insurance within

households in Cote d’Ivoire. They tested the effects on the types of goods

consumed by the household of different rainfall configurations with identical

impacts on total expenditure; given the types of crops they affect the most.

They rejected complete insurance within the household, even with respect to 3 See Alderman, Chiappori, Haddad, Hoddinott and Kanbur (1995) and Goldstein (1999) for detailed discussions.

5

publicly observed exogenous shocks. In addition they showed that the

expenditure patterns were not consistent with Pareto efficiency.

The first step in trying to assess the impact of risk on the rural poor is

to determine the true domain and scope of insurance available through social

networks. Some marginal groups are excluded from these networks. Evidence

shows that wealth and other socioeconomic status variables matter: the poor

have less dense risk sharing networks and may not have the wherewithal to

make the nominal contributions required by some of these mutual insurance

groups (Dercon 2005, De Weerdt 2005, Santos and Barrett 2006). Goldstein et

al. (2005) concluded that social exclusion of some community members may

result in the incomplete insurance observed in rural communities. Indeed,

extant studies have identified factors affecting network formation as well as the

characteristics of those who are known but left out of these networks– the

socially excluded. However, there is an even more vulnerable subpopulation

who have not been given their due in the development economics discourse –

the socially invisible: those who are not known in the community and are left

out by default. Who are the socially invisible? Who is left out of informal risk

management institutions? For whom is risk pooling Pareto efficient?

1.1.3 Poverty Dynamics and Asset Poverty Traps

Some insurance may be available but not to all. Differential access to

credit and insurance markets has distributional implications not just in terms of

consumption and opportunity cost of foregone profits but also long term

structural poverty. The need for immediate survival may lock the poor into

persistent poverty (Dercon 2003).

6

The poor consist of those who are persistently poor and those who

move in and out of poverty, with the latter being a very large group (Gaiha and

Deolalikar, 1993; Baulch and McCulloch, 2000; Dercon and Krishnan, 2000;

Jalan and Ravallion, 2000). Somewhat surprisingly, mobility analysis shows

that more households move out of poverty than into poverty (Scott, 2000).

Generally, households may be said to move in and out of poverty by crossing a

given poverty line based on some welfare metric: defining a threshold above

which households converge to desirable equilibria and below which they

converge to undesirable equilibria.

Several factors have been identified in the literature as determining

movement in and out of poverty. 4 Using data from Chile, Scott (2000)

estimated income functions for respective periods. He demonstrated that

income is positively related to education of household head, land ownership

and certain geographic locations but was negatively related to household size

and the presence of disabled household members. Using censored quantile

regression, Jalan and Ravallion (2000) identify physical capital, education and

variability in wealth as decisive factors, in Southern China. Smaller and better-

educated households had lower chronic poverty but neither household size nor

education affected the likelihood of transient poverty.

Regarding the direct impact of risk, Dercon and Krishnan (2000),

controlling for shocks with fixed effect regressions, also identify household

specific crop failure, rainfall and seasonal incentives such as changing labor

demand and prices, as factors determining fluctuations in welfare. McCulloch

4 See Dercon and Krishnan (1998), Glewwe and Hall (1998), Gunning et al (2000), Grootaert, Kanbur and Oh (1997), Jayaraman and Lanjouw (2000), Lanjouw and Stern (1991), Maluccio, Haddad and May (2000), Scott (2000), and Walker and Ryan (1990).

7

and Baulch (2000) estimated a reduced form income equation to show that, in

Pakistan, the variability in income is higher the smaller the proportion of adults

in the household, the less educated its male members are, and the more rain-fed

land it owns. They demonstrate, using policy simulations that interventions

that enable households to smooth their incomes over time can give rise to

reductions in transitory poverty. Scott (2000) in assessing the effects of

negative and positive shocks over time concluded that protracted droughts, loss

of employment, theft and sheer ill fate results in impoverishment. Seemingly

transitory shocks can have long-term effects. Serial correlation between shocks

may render structural poverty traps empirically indistinguishable from

persistent poverty (Barrett, 2001). It is expected that a sequence of positive

(negative) shocks may propel some households into rising (falling) welfare

trajectories such that their long-term position in the welfare distribution may be

altered. It is also possible that short- lived negative shocks can propel some

households onto permanently lower trajectories.

Generally fluctuations in income might be quite significant. However,

it is difficult to distinguish between stochastic and structural transitions using

standard poverty measures. The asset-based approach is said to distinguish

transitory poverty which eases with time due to systemic growth from

persistent structural poverty. Under this approach there are nonlinear asset

dynamics with multiple equilibria which include a threshold that distinguishes

households who can be expected to grow out of poverty from those who are

likely to remain trapped in poverty, i.e. those caught in a poverty trap (Carter

and Barrett 2006). These dynamics may be caused, to a large extent, by risk

8

exposure and response, attendant activity choice and returns to endowments

given incomplete financial and insurance markets. 5 However, the empirical

evidence regarding the existence of these dynamics is mixed.

A single negative shock in bad times (with no buffers) can put

households on a different trajectory than if it occurred under a favorable

external environment. Most households resort to informal means of insurance.

Social insurance has been documented as being likely to have a significant

effect on the fortunes of some households (Baulch and Hoddinot, 2000). Given

the prevalence of risk, do asset poverty traps exist? If so for whom, do they

exist? Particularly, how do these asset dynamics vary with access to social

insurance? Are these thresholds the same for everyone or do they vary by

social visibility? How can we identify individuals who are likely to get caught

in asset poverty traps for targeting purposes?

1.1.4 Intrahousehold Health Inequality

Development policy aimed at improving the wellbeing of individuals

through the provision of safety nets or the provision of mechanisms/

opportunities to move people out of structural poverty traps, inter alia, would

be largely ineffective if (i) the distribution of welfare within the household is

unequal or (ii) the individuals are unable to take advantage of these

opportunities due to poor health.

Theoretical models have been used in the extant literature to explain

intrahousehold inequality. In addition there is some empirical evidence, albeit

scant, indicating that there may be systemic differences in welfare levels 5 See Azariadis and Stachurski (2005) for a complete review.

9

within the household. However, it is not certain how these differences vary

with overall household well-being.6

Risk exposure may explain some of the proposed patterns in the

distribution of welfare within the household. For instance, the adverse effects

of shocks on health outcomes are borne disproportionately by women and

children within the household (Dercon and Krishnan 2000, Hoddinott and

Kinsey 2000, Dercon and Hoddinott 2003). Some households respond to

shocks by reducing in nutritional intake and neglecting the health needs of

children, especially those of girls (Alderman and Paxson 1992, Ellis

1998).This may have far-reaching effects given the fact that poor child health

has been linked with delays in beginning school, persistent effects on cognitive

capacity, diminished intellectual performance and stature, low work capacity

and lower earnings during adulthood (Glewwe and Jacoby 1995, Glick and

Sahn 1997, Grantham-McGregor et al. 1997, Martorell 1999). Strauss and

Thomas (1998) in discussing nutrition efficiency wages posit that “there may

be some people who are so poor and so unhealthy that they are costly to

employ”. In this case, the offered wage is insufficient to maintain the labor

power of the poor, leading to involuntary unemployment, reduction in

consumption/ nutrient intake with reduction in income (in the absence of

insurance), further loss in physical capacity and productivity, and an even

lesser likelihood of employment. This then has implications for child nutrient

intake and could potentially result in intergenerational persistence in poverty.

Given the effects of risk exposure access to social insurance might

condition intrahousehold health inequality. To what extent do intrahousehold

6 See Haddad and Kanbur (1990), Kanbur and Haddad (1992), Haddad, Kanbur and Bouis (1995) and Sahn and Younger (2007) for complete reviews.

10

inequalities manifest in the data? Are there intrahousehold Kuznets curves?

For whom do they exist? In the event that intrahousehold inequalities exist,

how would social visibility affect the distribution of well-being within the

household and how would the latter vary with overall household well-being?

The study seeks to address the issues raised above, towards a better

understanding of the risk insurance vis-à-vis asset poverty and intrahousehold

heath inequality with the broad objectives of enhancing the welfare of the poor

via improved access to insurance; facilitating movement out of poverty and

improving the odds of policy achieving its desired goal – improving the

wellbeing of individuals. To this end, the following chapters take an in-depth

look at these issues. The next section provides a brief description of the data

used. Section 4 provides a summary of the respective chapters. Section 3

concludes with the major results.

1.2 The Data

The data used in this study are from a rural household survey

undertaken from July 2004 to January 2005. This was the third wave of a panel

data set initiated by Christopher Udry and Markus Goldstein. The research was

conducted in the Akwapim South District (specifically the Nsawam - Aburi

area) in the Eastern Region of Ghana. Farmers in this area have been switching

from the cultivation of maize-cassava intercrop for domestic production to

pineapple cultivation for export since the early 1990s. This transition involves

a significant amount of risk by way of the attendant new agronomic practices

as well as exposure to global price fluctuations, hence the need for insurance.

The transition also provides the potential for marked increase in agricultural

11

incomes and asset accumulation.7 Appendix A provides field notes with details

on the most current wave. Appendix B provides a brief description of the

respective modules and survey schedule. Appendix C provides a copy of the

survey questionnaires.

1.3 Dissertation Chapters

1.3.1 Risk Management and Social Visibility in Ghana

This paper examines access to social insurance and the Pareto

efficiency of risk pooling. We hypothesize that social connectedness is crucial

determinant of access to informal social insurance. The relatively socially

invisible (i.e. those who are not widely known in the community) may get left

out by default, while those who are well known in the community may benefit

from social insurance. We start off by identifying the characteristics of the

socially invisible. We then test for risk pooling within and among social

networks to see if the extent of informal insurance available to individuals in

rural Ghana varies by their social visibility.

The key results are as follows. (i) There exists a small but distinct

subpopulation of socially invisible individuals within the sample villages. The

socially invisible tend to be younger, engaged in farming, recent arrivals into

the village, and not a member of a major clan. (ii) Risk pooling is substantial

for the socially visible; we cannot reject the null hypothesis that individual

shocks do not affect individual consumption and that individual consumption

tracks network and village consumption one-for-one. (iii) On the other hand,

risk pooling fails for the socially invisible subpopulation. We overwhelmingly

7 See Goldstein and Udry (1999) for an in-depth discussion of the historical background of the area and the sampling techniques.

12

reject both the null hypothesis that individual shocks do not affect individual

consumption and the null that individual and network or village consumption

move together one-for-one for those individuals who appear socially invisible.

1.3.2 Risk, Asset Poverty and Social Visibility in Ghana

So far we’ve identified the socially invisible and established that

individuals in this subpopulation are left uninsured. As per the previous

section, residual risk exposure might affect asset dynamics and the existence of

poverty traps. We hypothesize that asset dynamics will vary by social

visibility, with the socially invisible converging to a unique dynamic

equilibrium at a lower asset poverty level whilst the relatively visible

experience multiple stable dynamic asset equilibria. We examine the role of

social visibility in conditioning wealth dynamics by using the theory of

bifurcated asset accumulation strategies to empirically identify poverty traps,

using individual level data to allow for effect of differential social visibility

within the household. We estimate nonparametric regressions for all

individuals in the sample and then disaggregate the sample into socially visible

and socially invisible individuals using a predetermined cut-off as well as an

endogenously identified optimal sample splitting level. We then identify the

respective thresholds in asset dynamics and estimate the likelihood of an

individual falling into poverty traps.

The key results are as follows: (i) Asset trajectories exhibit multiple

dynamic stable equilibria for village-level aggregations of individuals.

(ii) Unpacking the results by social visibility indicates that whereas the assets

of the socially visible exhibit multiple stable dynamic equilibria, the socially

invisible converge towards a unique dynamic stable equilibrium at the lower

13

asset level. In addition, the asset levels at the respective equilibria were

generally higher for the socially invisible as compared to the socially visible,

particularly for the unstable dynamic equilibrium. (iii) Males engaged in non-

farm occupations who have resided in the village for more than one generation

are consistently less likely to fall into poverty traps, conditional on being

socially visible. Socially invisible individuals who belong to a major clan are

less likely to fall into a poverty trap. On the other hand, socially invisible

individuals who suffer a farm shock or theft are more likely to fall into a

poverty trap. It is noteworthy that shocks do not have a significant effect on the

assets of the socially visible. This is consistent with the results described in

Section 1.3.1 which notes the socially visible enjoy full insurance.

1.3.3 Risk, Intrahousehold Health Inequality and Social Visibility in

Ghana

Next, we hypothesize that access to insurance will condition the effect

of shocks on the distribution of health outcomes, specifically intrahousehold

health inequality, and the variation of the latter with overall household well-

being. We assesss whether there is a threshold effect in the existence of

Kuznets curves at the household level. We examine whether Kuznets curves

exist and for whom they exist. We also assess the determinants of

intrahousehold inequality.

The key results are as follows: (i) We find no evidence of the inverted

U-shape when using aggregated data, however there is some indication that the

Kuznets curve may exist for a subsample of the population – in this case the

socially visible.(ii) The composition of total inequality varies markedly by

14

social visibility. Whilst over 75% of overall inequality may be attributed to the

within-group component (i.e. intrahousehold inequality) amongst socially

visible households, a little over 50% of overall inequality may be attributed to

the between-group component amongst socially invisible households. (iii)

Patterns in relative BMI within the household suggest that the socially

invisible may overcompensate for risk exposure by protecting children in the

allocation of limited food and nutritional resources. (iv)Wealth, dependency

ratio, age, social visibility as well experiencing shocks had statistically

significant effects on intrahousehold inequality.

1.4 Conclusions

Economic growth and poverty reduction and have long been the focus

of many development efforts. The environment in which this occurs is fraught

with risk which might, in the medium to long term, thwart these efforts if left

unchecked. We take steps in this directon by assessing access to informal

insurance vis-à-vis social visibility and the implications for asset poverty and

intrahousehold inequality.

We identify a distinct subpopulation of socially invisible individuals

who tend to be younger, poorer, engaged in farming, recent arrivals into the

village who have been fostered, and not a member of a major clan. The results

show that while we cannot reject the null hypothesis that individual shocks do

not affect individual consumption and that individual consumption tracks

network and village consumption one-for-one among the socially visible, risk

pooling fails for the socially invisible subpopulation. In addition, whereas asset

accumulation patterns of the socially visible exhibit multiple stable dynamic

15

equilibria, the socially invisible converge towards a unique dynamic stable

equilibrium at the lower asset level. The asset levels at the respective equilibria

were generally higher for the socially invisible as compared to the socially

visible, particularly for the unstable dynamic equilibrium. The latter implies

that the socially invisible are more likely to fall into poverty traps since there is

a broader asset domain over which people in this subpopulation would collapse

toward the lower level asset equilibrium.

Multivariate analyses of the likelihood of falling into a poverty trap

indicates that males engaged in non-farm occupations who have resided in the

village for more than one generation are consistently less likely to fall into

poverty traps, conditional on being socially visible. Socially invisible

individuals who belong to a major clan are less likely to fall into a poverty

trap. Shocks increase the likelihood of a socially invisible individual falling

into a poverty trap. On the other hand, they do have a significant on the

likelihood of a socially visible individual falling into a poverty trap; this is

consistent with the observation that the socially visible enjoy complete risk

pooling. The analyses of intrahousehold inequality indicate that whilst the

inverted U-shape may not exist when using aggregated data, it may exist for

the socially visible. In addition the composition of inequality varies by social

visibility. Over 75% of overall inequality may be attributed to the within-group

component (i.e. intrahousehold inequality) amongst socially visible

households. On the other hand, a little over 50% of overall inequality may be

16

attributed to the between-group component amongst socially invisible

households.

Patterns in relative BMI within the household suggest that the socially

invisible may overcompensate for risk exposure by protecting children in the

allocation of limited food and nutritional resources. With regards to factors

determining the level of intrahousehold inequality: wealth, dependency ratio,

age, social visibility as well experiencing shocks had statistically significant

effects on intrahousehold inequality.

These findings suggest that external safety net interventions should

target those likely to be socially invisible and not the socially visible persons

for whom informal insurance seems to offer reasonably complete risk pooling.

They also underscore the need for policymakers to focus on structural

determinants of poverty transitions, or the lack thereof, in order to successfully

alleviate poverty in the long run. Finally, the challenges associated with

identifying individuals within the household for targeting purposes lead policy

makers to resort to the second best strategy – focusing on improving overall

household well-being. The existence of a systematic relationship between

intrahousehold inequality and average well-being on the basis of some

identifiable household characteristic such as social visibility should enable

policymakers better tailor interventions to attain the desired goal of improving

individual well-being.

17

REFERENCES

Adato, M., Carter, M. and May, J. (2006). Exploring Poverty Traps and Social Exclusion in South Africa Using Qualitative and Quantitative Data, Journal of Development Studies. 42, 226-47.

Alderman, H., Chiappori, P. A., Haddad, L., Hoddinott, J. and Kanbur, R.

(1995). Unitary Versus Collective Models of the Household: Is It Time to Shift the Burden of Proof? World Bank Research Observer, 10, 1-9.

Alderman, H., and Paxson, C. H. (1992). Do the Poor Insure? A Synthesis of

the Literature on Risk and Consumption in Developing Countries. The World Bank.

Arrow, K. and Fisher, A. (1974). Environmental Preservation, Uncertainty and

Irreversibility. Quarterly Journal of Economics, 312-9. Azariadis, C. and Stachurski, J. (2005). “Poverty traps”. In P. Aghion and S.

Durlauf (Eds.) Handbook of Economic Growth. http://elsa.berkeley.edu/%7Echad/azstach.pdf.

Bardhan, P.K. (1983). Labor-Tying in a Poor Agrarian Economy: A

Theoretical and Empirical Analysis. Quarterly Journal of Economics, 98(3), 501-514.

Barrett, C. (2001). AEM 762 Class Notes. Cornell University, Ithaca,

NewYork. Barrett, C. (2005). Displaced Distortions: Financial Market Failures and

Seemingly Inefficient Resource Allocation. Ithaca, New York: Cornell University.

Baulch, B. and Hoddinot, J. (2000). Economic Mobility and Poverty

Dynamics in Developing Countries. In B. Baulch and J. Hoddinot

18

(Eds.) Economic Mobility and Poverty Dynamics in Developing Countries. London:Frank Cass Publishers.

Biswinger, H. and Rosenzweig, M. (1993). Wealth, Weather Risk and the

Composition and Profitability of Agricultural Investments. Economic Journal, 103, 56-78.

Carter, M. and Barrett, C. (2006). The Economics of Poverty Traps and

Persistent Poverty: An Asset Based Approach. Journal of Development Studies, 2006, 1.

Coate, S. and Ravallion, M. (1993). Reciprocity without Commitment:

Characterization and Performance of Informal Insurance Arrangements. Journal of Development Economics, 40(1), 1-24.

De Weerdt, J. (2005). Risk-Sharing and Endogenous Network Formation. In

S. Dercon (Ed.) Insurance against Poverty. New York: Oxford University Press, 197-216.

Deaton, A. (1989). Saving in Developing Countries: Theory and Review. In S.

Fischer, and D. de Tray (Eds.), Proceedings of the World Bank Annual Conference on Development Economics 1989. Washington, D.C.: International Bank for Reconstruction and Development/ World Bank.

Deaton, A. (1992). Savings and Income Smoothing in Cote D'ivoire. Journal

of African Economies, 1(1), 1-24. Deaton, A. (1992). Household Saving in LDCs: Credit Markets, Insurance and

Welfare. Scandinavian Journal of Economics, 94, 253-73. Dercon, S. (2003). Insurance against Poverty, Helsinki. Finland: United

Nations University -World Institute for Development Economics Research.

19

Dercon, S. (2005). Risk, Poverty and Public Action. In S. Dercon (Ed.) Insurance against Poverty. New York: Oxford University Press, 439-50.

Dercon, S. and Hoddinott, J. (2003). Health, Shocks and Poverty Persistence.

In S. Dercon (Ed.) Insurance against Poverty. New York: Oxford University Press.

Dercon, S. and Krishnan, P. (2000). In Sickness and in Health: Risk Sharing

within Households in Ethiopia. Journal of Political Economy, 108, 688-727.

Duflo, E. and Udry, C. (2004). Intrahousehold Resource Allocation in Cote

D'ivoire: Social Norms, Separate Accounts and Consumption Choices. Mimeo. Department of Economics, MIT and Economic Growth Center, Yale University.

Ellis, F. (1998). Household Strategies and Rural Livelihood Diversification.

Journal of Development Studies, 35(1), 1-38. Fafchamps, M., Udry, C. and Czukas, K. (1998). Drought and Saving in West

Africa: Are Livestock a Buffer Stock? Journal of Development Economics, 55(2), 273-305.

Foster, A. (1995). Prices, Credit Markets and Child Growth in Low-Income

Rural Areas. Economic Journal, 105, 551-70. Gaiha, R. and Deolalikar, A. (1993). Persistent, Expected and Innate Poverty:

Estimates in Semi-Arid Rural India, 1975-1984. Cambridge Journal of Economics, 17, 409-21.

Gertler, P. and Gruber, J. (1997). Insuring Consumption against Illness.

National Bureau of Economic Research Working Paper Series.

20

Glewwe, P. and Hall, G. (1998). Are Some Groups more Vulnerable to Macroeconomic Shocks than Others? Hypothesis Tests Based on Panel Data from Peru. The Journal of Development Studies, 56,181-206.

Glewwe, P. and Jacoby, H. (1995). An Economic Analysis of Delayed Primary

School Enrollment and Childhood Malnutrition in a Low Income Country. Review of Economics and Statistics, 77, 156-69.

Glick, P. and Sahn, D. (1998): Health and Productivity in a Heterogenous

Urban Labor Market. Applied Economics, 30, 203-16. Goldstein, M. (1999). Chop Time, No Friends: Intrahousehold and Individual

Insurance Mechanisms in Southern Ghana. Yale University. Goldstein, M., De Janvry, A. and Sadoulet, E. (2005). Is a Friend in Need a

Friend Indeed? Inclusion and Exclusion in Mutual Insurance Networks in Southern Ghana. In S. Dercon (Ed.) Insurance against Poverty. New York: Oxford University Press.

Goldstein, M. and Udry, C. (1999). Agricultural Innovation and Resource

Management in Ghana. Washington D.C.: International Food Policy Research Institute.

Grantham-McGregor, Walker, S., C., Chang, S. and Powell, C. (1997). Effects

of Early Childhood Supplementation with and without Stimulation on Later Development in Stunted Jamaican Children. American Journal of Clinical Nutrition, 66, 247-53.

Gunning, J. W. Hoddinott, J., Kinsey, B. and Owens, T. (2000). Revisiting

Forever Gained: Income Dynamics in the Resettlement Areas of Zimbabwe, 1993-96. In B. Baulch and J. Hoddinot (Eds.) Economic Mobility and Poverty Dynamics in Developing Countries. London: Frank Cass Publishers.

Haddad, L. and Kanbur, R. (1990). How Serious Is the Neglect of

Intrahousehold Inequality? Economic Journal, 11, 866-81.

21

Haddad, L. Kanbur, R. and Bouis, H. (1993). Intrahousehold Inequality at

Different Welfare Levels: Energy Intake and Energy Expenditure Data from the Philippines. Oxford Bulletin of Economic and Statistics, 57 (3), 389-409.

Hoddinott, J. (2004). Shocks and Their Consequences across and within

Households in Rural Zimbabwe. Mimeo. International Food Policy Research Institute.

Hoddinott, J., and Kinsey, B. (1999). Adult Health in Time of Drought.

Washington D.C.. International Food and Policy Research Institute. Hoddinott, J. and Kinsey, B. (2001). Child Growth in Time of Drought. Oxford

Bulletin of Economics and Statistics, 64, 409-36. Jalan, J. and Ravallion, M. (2000). Is Transient Poverty Different? Evidence

for Rural China. In B. Baulch and J. Hoddinot (Eds.) Economic Mobility and Poverty Dynamics in Developing Countries. London: Frank Cass Publishers.

Jayaraman, R. and Lanjouw, P. (1999). The Evolution of Poverty and

Inequality in Indian Villages. World Bank Economic Observer, 14 (1), 1-30.

Kanbur, R. and Haddad, L. (1992). Is There an Intrahousehold Kuznets Curve?

Some Evidence from the Philippines, Public Finance, 47, 77-93. Kochar, A. (1993). Labor Market Responses to Idiosyncratic Agricultural

Shocks: Empirical Evidence from Rural India. Mimeo. Stanford University.

Lanjouw, P. and Stern, N. (1991). Poverty in Palanpur. World Bank Economic

Observer, 5 (1), 23-55.

22

Lybbert, T. J., Barrett, C. B., Desta, S. and Coppock, D. L. (2004). Stochastic Wealth Dynamics and Risk Management among a Poor Population. The Economic Journal, 114, 750–77.

Maluccio, J., Haddad, L. and May, J. (2000). Social Capital and Household

Welfare in South Africa, 1993-98. In B. Baulch and J. Hoddinot (Eds.) Economic Mobility and Poverty Dynamics in Developing Countries. London: Frank Cass Publishers.

Martorell, R. (1999). The Nature of Child Malnutrition and It's Long-Term

Implications. Food and Nutrition Bulletin, 20, 288-92. McCloskey, D. (1976). English Open Fields as Behavior Towards Risk. In J.A.

Fenkel, and H.G. Johnson (Eds.), Research in Economics History, Volume 1. Connecticut: JAI Press.

McCulloch, N. and Baulch, B. (2000). Simulating the Impact of Policy upon

Chronic and Transitory Poverty in Rural Pakistan China. In B. Baulch and J. Hoddinot (Eds.) Economic Mobility and Poverty Dynamics in Developing Countries. London: Frank Cass Publishers.

Morduch, J. (1999). Between the State and the Market: Can Informal Insurance

Patch the Safety Net? World Bank Research Observer, 14(2), 187-207. Murgai, R., P. Winters, Saduolet, E. and De Janvry, A.(2001). Localized and

Incomplete Mutual Insurance. Journal of Development Economics, 67, 245-74.

Paulson, A.L. (1994). Insurance Motives for Migration: Evidence from

Thailand. Mimeo. Paxson, C. and Chaudhuri, C. (1994). Consumption Smoothing and Income

Seasonality in Rural India. Mimeo.

23

Paxson, C.H. (1992). Using Weather Variability to Estimate the Response of Savings to Transitory Income in Thailand. American Economic Review, 82(1), 15-33.

Rose, E. (1994). Labor Market Responses to Idiosyncratic Agricultural

Shocks: Empirical Evidence from Rural India. Mimeo. University of Washington.

Rosenzweig, M.R. and Wolpin, K.I. (1993). Credit Market Constraints,

Consumption Smoothing, and the Accumulation of Durable Production Assets in Low-Income Countries: Investment in Bullocks in India. Journal of Political Economy, 101(2), 223-244.

Sahn, D. and Stifel, D. (2000). Poverty Comparisons over Time and across

Countries in Africa. World Development, 28, 2123-55. Sahn, D. E. and Younger, S. (2007). Measuring Intra-Household Health

Inequality: Explorations using the Body Mass Index. Cornell University.

Santos, P. and Barrett, C. (2004). Interest and Identity in Network Formation:

Who Do Smallholders seek out for Information in Rural Ghana? Cornell University.

Santos, P. and Barrett, C. (2006). Heterogenous Wealth Dynamics: On the

Roles of Risk and Ability. Cornell University. Santos, P. and Barrett, C. (2006). Informal Insurance in the Presence of

Poverty Traps: Evidence from Ethiopia. Cornell University. Scott, C. (2000). Mixed fortunes: A Study of Poverty Mobility among Small

farm Households in Chile, 1968-86. In B. Baulch and J. Hoddinot (eds.) Economic Mobility and Poverty Dynamics in Developing Countries. London: Frank Cass Publishers.

24

Strauss, J. and Thomas, D. (1998). Health, Nutrition and Economic Development. Journal of Economic Literature, 36 (2), 766-817.

Townsend, R. M. (1994). Risk and Insurance in Village India. Econometrica,

62, 539-91. Townsend, R.M. (1995). Consumption Insurance: An Evaluation of Risk-

Bearing Systems in Low-Income Economies. Journal of Economic Perspectives, 9(3), 83-102.

Udry, C. and Conley, T. (2005). Social Networks in Ghana. In C. Barrett (Ed.)

The Social Economics of Poverty. London: Routeledge. Walker, T.S. and J.G. Ryan (1990). Village and Household Economies in

India's Semi-Arid Tropics. Baltimore and London: Johns Hopkins University Press.

25

CHAPTER TWO

Risk Management and Social Visibility in Ghana

2.1 Introduction

Risk management is crucial to economic advance, indeed to the very

survival, of people in low-income, agrarian countries. In the ideal Arrow-

Debreu world, complete markets with symmetric information would provide

an array of state contingent contracts and all decision-makers in the economy

could make welfare improving exchanges based on each other’s known

preferences and beliefs over states of the world. In this fictional framework all

risks can be addressed with market-based solutions.

In reality, informational asymmetries and covariate risk impede risk

management in general, especially through formal financial institutions. Rural

populations therefore depend heavily on informal institutions for managing

risk in the absence of well-developed insurance markets. 8 But access to

informal insurance is not necessarily uniform. Certain subpopulations may

have superior access to the desirable intertemporal consumption smoothing

made possible by informal insurance mechanisms (Dercon and Krishnan 2000,

Dercon 2005, De Weerdt 2005, Santos and Barrett 2006).

So who is left out of informal risk management institutions? Since

informal institutions are based on endogenously formed social networks among

and within households, for whom is reasonably complete (i.e., Pareto efficient)

8 See Alderman and Paxson (1992), Besley (1995) and Bardhan and Udry (1999) for complete reviews.

26

risk pooling available? To address these issues we draw on several threads in

the literature.

The first thread is the literature on risk pooling and social insurance.

Empirical tests often reject the null hypothesis of full (Arrow-Debreu) social

insurance within rural communities (Deaton 1992, Townsend 1994, Gertler

and Gruber 1997). Issues pertaining to measurement error, asymmetric

information and contract enforcement have been identified as possible reasons

for incomplete risk pooling (Alderman and Paxson 1992, Murgai et al. 2001).

One could argue that the hypothesis of full insurance will necessarily be

rejected because these tests are typically conducted for exogenously given

groups such as the entire village, community, or ethnic group and such

exogenously defined groups might not accurately reflect the true domain and

scope of risk sharing. 9 In addition, most studies of the extent of social

insurance have been undertaken at the household level, implicitly assuming a

unitary household model and thus perfect substitutability among decision-

makers as well as pooling of all resources within the household. Yet there is

ample evidence in the literature rejecting these assumptions, including within

the context of social insurance (Dercon and Krishnan 2000).10

This leads to the second thread of the literature on which we build.

Social networks have long been identified as key loci of informal risk sharing.

Individuals establish networks based on a wide range of individual and mutual

attributes, including but not limited to kinship ties, ethnicity, geographical

proximity, occupation, wealth, religion, and gender (Goldstein 1999, Santos

9 See Goldstein et al. (2005), Santos and Barrett (2006). 10 See Alderman, Chiappori, Haddad, Hoddinott and Kanbur (1995), Doss (1996), Udry (1996), Goldstein (1999) and Duflo and Udry (2004) for evidence on intrahousehold allocation issues.

27

and Barrett 2004, Udry and Conley 2005, De Weerdt 2005, DeWeerdt and

Dercon, 2006, Fafchamps and Gubert 2007). However, some marginal groups

may be less well connected in social networks and may thereby enjoy less

informal insurance access than do wealthier or more powerful members of a

community (Dercon 2002, De Weerdt 2005, DeWeerdt and Dercon 2006,

Santos and Barrett 2006). If social network formation is commonly

asymmetric, in the sense that an agent’s latent demand for a link with another

is a function not just of the absolute social distance between the two, but also

of their ordinal position – e.g., male-female versus female-male, or poorer or

wealthier (Santos and Barrett 2004, DeWeerdt and Dercon, 2006, Fafchamps

and Gubert 2007) – then what appears in the literature as wealth differentiation

in insurance access could be due to differential social visibility based on

individual characteristics, visibility that is correlated with but distinct from

wealth.

We hypothesize, in particular, that social connectedness is the key issue

in access to informal social insurance. Those who are relatively socially

invisible, meaning they are not widely known in the community, may get left

out by default, while those who are well known in the community enjoy the

sort of social insurance widely hypothesized in the literature. If

geographically-defined populations of the sort used in standard empirical

analysis mix the socially visible and the socially invisible, widespread

evidence of partial risk pooling could well mask reasonably complete risk

pooling for those with extensive social networks and negligible risk pooling for

the socially invisible. The practical implication of such a finding would be

straightforward: target external safety net interventions to those likely to be

socially invisible and try not to waste resources on external safety nets for

28

those socially visible persons for whom informal insurance seems to offer

reasonably complete risk pooling.

In this paper we therefore test for risk pooling within and among social

networks, as well as within villages, to see if the extent of informal insurance

available to individuals in rural Ghana varies by their social visibility. The

remainder of the paper proceeds as follows. The next section lays out the

familiar general equilibrium model of risk pooling and uses limiting conditions

on the extent of an agent’s social network to derive exclusionary restrictions

for econometric hypothesis testing. Section 3 then describes the data. Section 4

explores the concept of social invisibility and identifies the characteristics of

those persons we label socially invisible in this sample. Section 5 tests for risk

pooling within and among social networks. Conclusions and policy

recommendations are presented in Section 6.

The key results are as follows. (i) There exists a small but distinct

subpopulation of socially invisible individuals within the sample villages. The

socially invisible tend to be younger, engaged in farming, recent arrivals into

the village, and not a member of a major clan. (ii) Risk pooling is substantial

for the socially visible; we cannot reject the null hypothesis that individual

shocks do not affect individual consumption and that individual consumption

tracks network and village consumption one-for-one. (iii) On the other hand,

risk pooling fails for the socially invisible subpopulation. We overwhelmingly

reject both the null hypothesis that individual shocks do not affect individual

consumption and the null that individual and network or village consumption

move together one-for-one for those individuals who appear socially invisible.

29

2.2 Risk Pooling and Social Visibility: Simple Theoretical Predictions

Shocks are pervasive in most agrarian societies, the study area in

Ghana being no exception. To what extent are shocks insured so that

individuals can smooth consumption and thereby improve intertemporal

welfare? Is risk fully pooled at the village level or at the network level? Does

the extent of risk pooling depend on an individual’s social visibility? To

address these questions, we use a standard model akin to that of Townsend

(1994).

Suppose there are S possible states of the world, each occurring with

objective, constant and commonly known probability πs. Assume individuals

have preferences that are additive across time and over states of nature and

common rates of time preference βt, where λin is the programming weight

associated with individual i in network n. Suppose there are K social networks

in the economy with N members in each network. 11 Let cinst and yinst be

individual i in network n’s consumption and income in state s at time t,

respectively. Then we can denote the Pareto efficient allocation of risk by the

following problem:

(1) ( )⎥⎦

⎤⎢⎣

⎡∑ ∑∑= ==

inst

T

t

S

sst

N

iin cU

instC 0 11max πβλ

subject to

(2) ∑∑==

=N

iinst

N

iinst yc

11

.

The first order condition allocating consumption among two individual

network members, i and j, is given by

11 For now, we abstract from the fact that networks may be interlinked: an individual may belong to more than one network. We explore this possibility below in examining empirically the extent to which risk is spread among social networks.

30

(3) ( ) ( )jnstjninstin cUcU '' λλ =

Suppose each individual’s preferences can be represented by an exponential

utility function

(4) ( ) [ ]instinst ccU αα

−⎥⎦⎤

⎢⎣⎡−= exp1 .

Then substituting (4) into (3), taking logs and rearranging terms, we get

(5) ⎥⎦

⎤⎢⎣

⎡−=

in

jnjnstinst cc

λλ

log .

Then aggregating across all individuals in a network gives

(6) in

N

jjnstinst Ec

Nc += ∑

=1

1

where

(7) ⎥⎦

⎤⎢⎣

⎡−

+= ∑−

=

1

1log

11log1 N

jjninin N

E λλα

which implies that

(8) innstinst cc Ε+=

where nstc is the network mean consumption excluding i and Ein is a constant

that allows for dispersion in consumption levels among network members that

is fixed over time and states of nature. The strong implication of Pareto

efficient risk pooling, as reflected in equation (8), is that contemporaneous own

income, yinst is irrelevant to the determination of individual consumption. We

can decompose agent i’s contemporaneous income in network n into a

permanent component iny and an idiosyncratic component, insty , where the

latter represents a mean zero, i.i.d transitory income shock.. Own income is

thus:

(9) instininst yyy +=

31

Introducing (9) into (8) implies a testable exclusionary restriction for full

insurance:

(10) ( ) innstinstininst cyyc Ε+++= δγβ

Taking first differences in time yields the estimable equation

(11) nstnstinstinsticyc εγβ +Δ+Δ=Δ

Where ∆ is the first difference operator and instε is a mean zero, i.i.d error

term. A test of the full risk pooling hypothesis is then H0: β=0, γ=1 versus HA:

β>0 or γ<1. An analogous test of the full risk pooling hypothesis at village

level emerges from substituting change in mean consumption in village v

(excluding agent i) for that in network n in equation (11 Similarly, the null

hypothesis that distinct social networks within a village pool risk can be

derived by substituting nvstcΔ as the dependent variable and nvstc −Δ for the

second term in equation (11), respectively. Where nvstcΔ is the mean

consumption for individual i’s network and nvstc −Δ is the residual mean for all

other networks in the village excluding individual i’s.

The implication of the above framework for those who are socially

invisible – i.e., for whom N=1 because they are the entirety of their own social

network – is immediately obvious. The summation operators in equations (1)

and (2) drop away, the social allocative efficiency conditions in (3) and (5)

disappear, and the key exclusionary restrictions in equation (11) are thus

turned on their head. For those without recourse to social insurance networks,

there should be no risk pooling, thus H0: β=1, γ=0 versus HA: β<1 or γ>0. This

basic insight that the extent of the social network fundamentally influences the

relation between individual-level income shocks and individual consumption,

i.e., the risk pooling capacity of the individual, implies a need to first define

32

and measure social visibility before testing standard risk pooling hypotheses

because social visibility conditions the appropriate test specification.

2.3 The Data

The data used in this paper are from a rural household survey

undertaken in the Akwapim South District (specifically, the Nsawam - Aburi

area) in the Eastern Region of Ghana from July 2004 to January 2005. This

was the third wave of a panel data set initiated by Chris Udry and Markus

Goldstein and described in detail in Goldstein and Udry (1999). Since the early

1990s farmers in this area have been switching from the cultivation of maize-

cassava intercrop for domestic production to pineapple cultivation for export.

This transition involves a significant amount of risk associated with

transitioning to new agronomic practices and marketing arrangements, as well

as potential disruption of traditional social arrangements.

The original sample was selected using a two-stage procedure. Four

village clusters were purposively selected based on their participation in fruit

and vegetable production as well as their agronomic, market access and

geographic conditions. Sixty married couples (or triples) were then randomly

selected in each village cluster, except for the smallest village cluster, where

all resident couples were included.12 Each individual selected was interviewed

separately. Male enumerators were assigned to male respondents and female

enumerators to female respondents to preserve gender sensitivity and cultural

norms.

12 We loosely refer to village clusters as villages for ease of notation.

33

Three rounds of data were collected at approximately eight week

intervals, rotating between pairs of villages. 13 The first round of data was

collected in September-October 2004. The second round was conducted

October-November 2004. The third round was conducted from late November

2004 through January 2005. Across the four villages, 372 individuals were

surveyed the first round. The second and third rounds had sample sizes of 371

and 350 individuals, respectively. The sample attrition rate was thus 0.3%

between the first and second rounds and 5.9% between the first and third

rounds.14

The subsections that follow offer brief descriptions of the key modules

relevant to our analyses. Other standard variables associated with household

composition, asset holdings, family background, etc. are likewise employed in

the regression analysis in section 5 and described below.

2.3.1 Individuals’ Social Networks

For each respondent, we randomly selected seven individuals in the

sample from the same village (without replacement).15 We then asked each

13 Given budgetary constraints, each enumerator was assigned to one of two villages in pre-assigned pairs. Based on geographic proximity, the first village pair comprised of villages 1 and 4, whereas the second village pair comprised of villages 2 and 3. Interviews were conducted simultaneously in the two villages in a given pair, after which the enumerators moved to the other pair. We spent approximately four weeks in one pair of villages, moved to the next pair for next four weeks, and continued this pattern of rotation for the duration of the study. 14 A simple attrition probit was estimated using robust standard errors, with the dependent variable ATTRIT=1 if individuals were lost between rounds 1 and 3, = 0 otherwise. The estimation results indicate that individuals lost between rounds 1 and 3 were more likely to be younger males whose parents had held village offices. Neither wealth nor the incidence of any of the shocks used in the subsequent analyses was statistically significant in explaining patterns of attrition. 15 Respondents were also non-randomly matched with three other village-specific “focal” individuals identified from the community-studies approach taken in a preliminary field trip as individuals in the villages from whom advice is commonly sought. We focus on the random matches in this study.

34

respondent about their knowledge of the match i.e., “Do you know__?”,

followed by a series of questions about their relationship with each of these

matched individuals: how often they talked with them, and whether or not

he/she could approach the individual to deal with any of a set of specific issues

related to farming and credit. In administering the questionnaire we made a

clear distinction between knowing of someone (i.e., using the Akan translation

of just “having heard of the person”) and actually knowing the person in the

sense of mutual acquaintance. Knowing a random match in this sense is

indicative of an extant social link. This gives us a random sample not only of

individuals but also of prospective social relations, which is the preferred

method of sampling social networks (Santos and Barrett 2007). By design, the

characteristics of these random matches are representative of people with

whom they have (and do not have) extant social links, i.e., it provides an

unbiased representation of the structure of their social networks.

2.3.2 Consumption

Detailed data were collected on purchased food, general family

expenses and personal expenditures by each respondent in the household.16

Even though these expenditure questionnaires were administered at the

individual level, with the head and spouse(s) of head being interviewed

separately regarding contributions made towards purchasing an item,

individual expenditures were not assigned. Hence, we follow Goldstein (1999)

in assigning particular items as purchases for own-consumption: alcoholic

beverages, non-alcoholic pre-packaged beverages, prepared food (from

16 Recall periods varied by expenditure based on the modal frequency of purchase reported in waves 1 and 2. These essentially intra-annual expenditure were converted to nominal monthly rates.

35

kiosks), personal care products, hair cuts, public transport, petrol, car repairs,

newspapers, entertainment, lottery tickets and kola nuts.

Table 2.1 presents the mean and standard deviation of expenditure on

these goods purchased for own-consumption per village and round as well as

the share of total expenditure incurred by each respondent on purchased food,

general family needs and personal items spent on these assigned individual

items in a typical month. On average, respondents spent 393,383 Cedis;

306,422 Cedis and 379,544 Cedis in rounds 1, 2 and 3, respectively.17 These

values account for 16%, 19% and 18% of total expenditures in rounds 1, 2 and

3, respectively. Village 4 had the least individual expenditures.

2.3.3 Shocks

Respondents were asked about a series of prospective idiosyncratic

shocks. We selected four types of shocks: (i) value of damage caused by

general farm problems; (ii) total curative health care expenses; (iii) value of

personal items stolen; and (iv) funeral expenses upon sudden death of family

member. For each of type of shock we asked about the out-of-pocket expense

incurred as a result of that shock or the imputed value of damage experienced.

Table 2.1 presents the frequency of shocks faced by respondents as well as the

nominal mean response expenditures and imputed values of damage caused by

these shocks by round.

Overall, 92% of respondents reported suffering at least one of these

shocks in round 1. 83% and 79% reported experiencing a shock in rounds 2

and 3, respectively. By way of percentage of individuals affected, health

shocks were the most frequent in rounds 1 and 3, with sudden death within the

17 The mean exchange rate over the survey period was roughly 9000 Cedis/ US$1.00.

36

Table 2.1: Individual consumption expenditures, June 2004- January 2005 (nominal Cedis per month)

Table 2.2: Percentage of individuals affected by shocks and mean response expenditure/ imputed value of damage (nominal Cedis)

Village Round 1 Round 2 Round 3

Mean Standard Deviation

Share of Total Expenditure Mean

Standard Deviation

Share of Total Expenditure Mean

Standard Deviation

Share of Total Expenditure

1 378033 361238 0.190 389513 374807 0.244 607648 1952393 0.235 2 615055 1166220 0.189 363188 438698 0.209 373984 333596 0.220 3 442872 1894519 0.117 258599 375004 0.151 359117 1721446 0.131 4 171599 137713 0.126 228730 415352 0.144 188596 210620 0.134

Full Sample 393383 1097788 0.156 306422 404741 0.185 379544 1334326 0.178

Shocks All

Rounds Round 1 Round 2 Round 3

Frequency Frequency Mean

Share of Total

Expenditure Frequency Mean

Share of Total

Expenditure Frequency Mean

Share of Total

Expenditure Farm problems 31.7 52.0 1107214 0.512 22.8 304984 0.238 20.6 200643 0.133 Total Health Expenses 56.7 72.2 1699291 0.580 50.0 379575 0.513 46.0 307786 0.286 Theft of personal item 14.7 19.1 768218 0.405 13.7 381595 0.203 11.0 365600 0.390 Sudden death 50.7 50.5 880862 0.459 55.9 171809 0.125 45.6 228589 0.109 Any/All Shocks 84.6 91.8 2640621 1.093 83.0 492290 0.494 78.7 415992 0.319

36

37

family being the most frequent shock experienced in round 2. Over all three

rounds, 57% of the respondents reported having suffered health problems and

51% had suffered a sudden death in the family. Morbidity and mortality thus

pose a huge financial burden on families in this area. In addition, 32% of the

respondents suffered from at least one of a variety of farm problems, the most

prevalent being infestation by grasscutters (a common rodent in the area). Only

15% of the respondents experienced theft of a personal item. While the

questions were focused on idiosyncratic shocks, covariate risk associated with

rainfall and price patterns did not appear to be the primary concern in these

communities.

Not only were shocks commonplace, they were also very costly. The

total value of losses due to shocks was 2640621 Cedis in round 1, 492290 in

round 2 and 415992 in round 3. These figures correspond to 109%, 49% and

32% of total expenditures incurred by respondents on purchased food, family

and personal items in rounds 1, 2 and 3, respectively. Even though few people

reported experiencing theft of personal items, this shock was associated with

huge losses. This was the most serious shock in rounds 2 and 3, in terms of the

magnitude of imputed value of loss. Health shocks were the most serious in

round 1. By way of share of total expenditure captured by the value of loss,

health shocks were the most important at 58% and 51% in rounds 1 and 2,

respectively, with theft of personal items accounting for 39% in round 3. The

magnitude of these shocks relative to household expenditure levels underscores

that idiosyncratic shocks can imperil the accumulated assets of households if

they have insufficient (formal or informal) insurance. Hence our desire to

understand who is reasonably well insured.

38

2.4 Social Invisibility

Although much discussion of informal insurance implicitly assumes

that everyone participates equally in the social networks that mediate

interhousehold transfers, recent studies find significant intra-village variation

in social connectedness, with a non-trivial share of individuals relatively

isolated from other residents (Santos and Barrett 2006). Historical accounts

from Africa confirm this pattern, emphasizing the correspondence between

extreme poverty and limited social interactions. For example, Iliffe (1987,

p.42) notes that “[t]o be poor is one thing, but to be destitute is quite another,

since it means the person so judged is outside the normal network of social

relations and is consequently without the possibility of successful membership

in ongoing groups, the members of which can help him if he requires it.” Such

observations motivate our hypothesis that risk pooling through social networks

may vary within villages, with poorer individuals generally being more

socially invisible and therefore enjoying less access to informal insurance than

do wealthier, better connected individuals.

We define social invisibility as a condition in which an individual

resides within a community but is not known by some minimal share of other

members of that community. Toward this end, we define social visibility (SV)

index. Let Ni be the number of times that respondent i’s name was drawn in the

random matching process and presented to some other respondent j. Then let

ni=Σjkij for j=1,…,Ni, be the number of times that i was identified by others

when presented as a random match (kij=1 when j knows i, =0 otherwise). We

then estimate the social visibility index, SV, as:

(12) 10 , ≤≤= ii

ii SV

NnSV

39

The descriptive statistics on Ni, ni and SVi are presented in Table 2.3.

Each individual was presented on average 5 times as a random match, out of

which they were known on average by 3.38 respondents. 26 of the respondents

(i.e., 8.39%) were not known by any of their random matches. Clearly they

qualify as socially invisible by this metric. It turns out that, by this measure,

SV=0 proves the statistically optimal maximum index value for denoting a

respondent “socially invisible”, with those with SV>0 classified as socially

visible.18

Table 2.3: Summary statistics on random matching

Notes: * Based on an optimal cut-off level of 0.80 predicted probability of i being known

Table 2.4 presents summary statistics on the individual characteristics

for the socially visible and the socially invisible. Six key facts emerge from

this table. First, the socially invisible are predominantly female (only 23% of

them are male). Second, they are slightly younger than their socially visible

counterparts. Third, a little more than half of them (54%) have been fostered,

i.e., as a child they lived in the care of persons other than their parents and

outside of their homes for at least a year. Fourth, far more of them were the

first generation to reside in the village, 71 percent versus 39 percent among the

18 More precisely, we estimated all the models reported below for different threshold values of SV to separate the socially visible and socially invisible, defining Socially Invisible=1 if SV≤ 0, 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35 and 0.40, with Socially Invisible=0 otherwise. The greatest likelihood value was achieved at SV≤0.

Variable Mean Median Min Max Number of times presented as a match (N) 5.00 5.00 1 15 Number of times known as a match (n) 3.38 3.00 0 13 Social visibility (SV) index – direct matches 0.67 0.75 0 1 SV index – imputed matches* 0.38 0.32 0 1

40

Table 2.4: Descriptive statistics: estimating social visibility

Frequency (%) Variable Definition

Entire Sample

Socially Visible

Socially Invisible

Male =1 if male, 0 otherwise 47.6 49.5 22.9* Age Respondent’s age 45.5

(12.9) 45.9

(13.1) 40.8* (9.4)

Level of Schooling No schooling Primary School Middle School Higher School

=1 if respondent has no schooling, 0 otherwise =1 if primary level, 0 otherwise =1 if middle school or junior secondary school, 0 otherwise =1 if any higher, 0 otherwise

25.2 17.0 50.8 5.7

25.1 16.6 51.2 5.7

25.7 22.9 45.7 5.7

Occupation Farmer Other Unemployed

=1 if farmer, 0 otherwise =1 if trader, artisan, teacher, civil servant, office or health worker, agricultural or non-agricultural labor, 0 otherwise =1 if student/ pupil, unemployed or not in the labor force

78.9 17.8

3.3

78.8 18.9

3.1

80.0 14.3

45.7

Major clan =1 if member of a major clan, 0 otherwise 89.8 90.2 84.3 Herelong =1 if not the first generation to reside in

village, 0 otherwise

59.1 61.4 28.6*

Parents held office =1 if parents holds any village office, 0 otherwise

47.1 48.1 31.3*

Fostered =1 if respondent was fostered, 0 otherwise

53.7 53.6 54.3

Value of non-land wealth

Value of esusu, bonds, pension, jewelry, cash being kept with others, chemicals, seeds, crops and goods to be traded in millions of Cedis

6.7 (9.2)

7.0 (9.5)

3.7 (3.8)

Value of inheritance

Total value of any current and expected land and non-land inheritance in hundred millions of Cedis

92.2 (1656.4

)

99.2 (1718.7)

0.2 (0.6)

Location Village 1 Village 2 Village 3 Village 4

=1 if Village cluster 1, 0 otherwise =1 if Village cluster 2, 0 otherwise =1 if Village cluster 3, 0 otherwise =1 if Village cluster 4, 0 otherwise

25.8 22.9 24.3 27.0

26.7 24.3 21.8 27.2

14.3* 4.3* 57.1*

24.3 Notes: * Differences in means statistically significant at the 5% level. The standard deviations of continuous variables are given in parentheses.

41

socially visible. Fifth, fewer of them had parents who have held village

offices. Finally, the socially invisible are poorer than their socially visible

counterparts, with a mean non-land wealth value of 3.7 million Cedis as

compared to 7.0 million Cedis for the socially visible.

These simple cross-tabulations are reinforced by a multivariate

regression analysis. Let itl =1 be an indicator variable that equals one if

individual i is socially invisible at time t. Let Pr{ itl =1} be the probability that

itl =1 conditional on some individual characteristics, itX . We then estimate

(13) Pr{ itl =1}= )0( >+Λ ititX εβ

by probit regression, where Λ is the normal CDF. We estimate a random

effects model using all three rounds of data, with observations clustered on the

respondent’s identity.

The parameter estimates are presented in Table 2.5. The results

reinforce the descriptive statistical results. Social invisibility is declining in

age and wealth, is significantly less for those whose family has resided in the

village for more than one generation and for those who are not farmers. This

latter result is consistent with Santos and Barrett’s (2004) results using these

same data but a different model; they likewise find that teachers and traders are

more likely to be known and the older are less likely to know the younger

members of community. Belonging to a major clan reduces the likelihood of

being socially invisible since one may establish links with members of ones’

matrikin (Goldstein 1999, De Weerdt 2002, Santos and Barrett 2004, Udry and

Conley 2005). On the other hand, having been fostered increases the likelihood

of being socially invisible. In addition, having education beyond the middle

42

Table 2.5: Probit estimation of social invisibility

Notes: ***, **, * Significant at the 1%, 5% and 10% levels, respectively. Comparison group is a female respondent from village 1 who farms and has a middle school education.

Variables Marginal Effects Standard Error Prob>|z| Dependent Variable: Socially invisible (SV=0) Individual Characteristics Male -0.123 0.226 0.587 Age -0.027* 0.009 0.004 No schooling -0.298 0.252 0.237 Primary school education 0.284 0.215 0.187 Higher than middle school 0.655** 0.329 0.046 Non-farm occupation -0.605*** 0.343 0.077 Unemployed 0.605 0.457 0.185 Assets Value of non-land wealth -0.067** 0.032 0.037 Value of inheritance -0.103 0.106 0.330 Social characteristics Major clan -0.576** 0.230 0.012 Herelong -0.292*** 0.178 0.100 Fostered 0.650** 0.228 0.004 Parents held office -0.077 0.178 0.665 Location Village 2 -0.240 0.344 0.485 Village 3 1.255 0.270 0.000 Village 4 0.532 0.268 0.047 n = 852 Log likelihood = -144.54 Wald χ2(16) = 67.59, p-value = 0.000

43

school level increases the likelihood of being socially invisible. While this

seems counterintuitive, De Weerdt (2002, p.12) found that “households with

educated members tend to lie closer to each other on the network graph.” By

assortative matching, the relatively highly educated are thus more likely to be

linked to each other and since they have fewer peers (i.e., those with higher

than middle school education make up only 7.03% of the population) they

appear less likely to be known by others. Moreover, the relatively highly

educated may have left the village for a number of years in pursuit of

education, thereby interrupting patterns of social interaction that condition

social visibility. Being male, and having parents who held village offices were

consistent in sign with the cross tabulations but not statistically significant,

even at the 10% level.

The definition of social invisibility used thus far relies on reporting by

a very small subsample – from 1 to 15 people – to whom each respondent’s

name was presented as a random match. This may generate small sample

variability in the measure of social visibility. So we repeated the exercise, this

time by estimating each individual’s social connectedness as reflected in the

pattern of kij, the indicator variable reflecting whether j knows i. Now, rather

than using kij directly to estimate SVi as a function of the relatively few

matches to which each individual’s name was presented, Ni, we instead

estimate a probit regression of kij to establish patterns of social visibility and

then use those estimates to predict the probability of i being known by each

sample respondent j.19 Based on an optimally chosen cut-off value for that

probability, we can then estimate SVi over the whole village sample, Nv-1, for

each respondent, substantially increasing the number of prospective social 19 We used the entire roster of individuals in the intra-village samples from wave 2.

44

links over which social visibility is measured for each individual. We follow

Santos and Barrett (2004) in this specification, allowing for asymmetry in who

knows whom by controlling for the direction of prospective differences

between a respondent and her match, not just the algebraic distance between

them. For example, a non-teacher may indicate that she knows a random

match who is a teacher, but the teacher might not remember that respondent.

The probit estimation results are presented in Table 2.6. Respondents

who were both male, had the same level of education, had resided in the

village for more than one generation and belonged to the same clan were more

likely to know each other. Older, poorer respondents who were farmers or

traders were more likely to be known. In addition, females were more likely to

know males, males were more likely to know females and non-farmers were

more likely to know farmers. On the other hand, farmers were less likely to

know non-farmers, respondents who were fostered were less likely to know

those who were not fostered and less likely to know each other.

Based on the results reported in Table 2.6, out-of-sample predictions

were made to determine the likelihood of each individual being known by all

sampled individuals in their village. We then set ijk̂ =1 based on some minimum

threshold probability, using 0.1 point intervals from 0.1 to 0.9. We then

computed iSV∧

≡ΣNv-1 ijk̂ /(Nv-1) for each respondent based on these predicted

probabilities and cut-off value. As before, we used different threshold values to

discretize the resulting ∧

SV continuum into the socially invisible and the

socially visible. For the predicted social visibility index, the maximum

likelihood occurs when we let socially invisible=1 if iSV∧

≤0.25 and 0

otherwise, for ijk̂ =1 if the predicted probability ≥ 0.8. By this definition,

45

Table 2.6: Probit estimation of the likelihood of knowing a random match

Notes:***, **, * Significant at the 1%, 5% and 10% levels, respectively.

Variables Definition of variables (i=respondent, j= -i intra-village sample individual) Marginal Effects Prob>|z| Both male =1 if both i and j are male, 0 otherwise 1.030*** 0.000 Female, male =1 if i is female and j is male, 0 otherwise 0.394*** 0.000 Male, female =1 if i is male and j is female, 0 otherwise 0.200** 0.029 Older =1 if i is older than j, 0 otherwise 0.152** 0.024 Same education =1 if both i and j have the same level of education, 0 otherwise 0.213*** 0.000 Same occupation =1 if both i and j have the same occupation, 0 otherwise -0.119 0.230 Trader, non-trader =1 if only i identifies himself as a trader, 0 otherwise -0.375 0.131 Non-trader, trader =1 if only j identifies himself as a trader, 0 otherwise 0.207 0.121 Farmer, non-farmer =1 if only i identifies himself as a farmer, 0 otherwise -0.527*** 0.000 Non-farmer, farmer =1 if only j identifies himself as a farmer, 0 otherwise 0.542*** 0.000 Same clan =1 if both i and j belong to the same clan, 0 otherwise 0.245*** 0.001 Both herelong =1 if both i and j have resided in the village for more than one generation, 0 otherwise 0.327*** 0.001 Herelong, not-herelong =1 if only i has resided in the village for more than one generation, 0 otherwise 0.012 0.913 Not-herelong, herelong =1 if only j has resided in the village for more than one generation, 0 otherwise 0.061 0.562 Both fostered =1 if both i and j have been fostered, 0 otherwise -0.291** 0.004 Fostered, not-fostered =1 if only i has been fostered, 0 otherwise -0.317** 0.004 Not-fostered, fostered =1 if only j has been fostered, 0 otherwise 0.032 0.762 Poorer =1 if i is poorer than j, 0 otherwise 0.111** 0.054 Age Age of i 0.003 0.323 Value of nonland wealth Value of non-land wealth owned by i, millions of Cedis 0.000 0.433 Farmer =1 if i identifies himself as a farmer, 0 otherwise 0.723*** 0.000 Trader =1 if i identifies himself as a trader, 0 otherwise 0.776** 0.003 Village 2 =1 if i lives in village 2, 0 otherwise 0.187* 0.094 Village 3 =1 if i lives in village 3, 0 otherwise -0.754*** 0.000 Village 4 =1 if i lives in village 4, 0 otherwise -0.199* 0.055 n=3724 Log likelihood = -1952.09 Wald chi2(25) = 342.66 Prob>chi2 = 0.000

45

46

22.79% of the sample was socially invisible, nearly three times as many as

under the stricter SVi=0 criterion used with the direct matching data. The

bottom row of Table 2.3 compares the two socially visibility indices computed

by the direct and probabilistic imputation methods. The predicted social

visibility continuum had a lower mean, 0.38 as compared to 0.67 for the index

based on direct matches. On average, respondents were predicted to be less

visible than previously indicated by the index based on random matches with a

small sub-sample. The two social visibility indices were statistically

significantly correlated at the one percent level with a correlation coefficient of

0.47.

But the real point is that there is considerable variation in social

visibility even within relatively small rural villages. Clearly, not all residents

within these rural Ghanaian villages are equally well known by others within

the community. Some people appear sufficiently infrequently known by others

that one might reasonably term them socially invisible. Does this matter to

individual risk management capacity?

2.5 Risk Pooling Conditional on Social Visibility

If social connectedness fundamentally affects how an individual’s

expenditures vary with the shocks she experiences, as suggested by section 2’s

simple general equilibrium model of risk pooling, then conventional tests for

risk pooling should condition on these measures of social (in)visibility. The

relevant hypotheses differ between the socially visible – who are expected to

pool risk – and the socially invisible – who are not.

47

Given the heterogeneity within these Ghanaian villages in social

connectedness, as demonstrated in section 4, we hypothesize that global tests

of risk pooling may mask heterogeneity in access to risk management through

mutual insurance and related mechanisms. We use the agricultural, health, theft

and mortality shocks discussed earlier as proxies for insty . This requires

adaptation of equation (11), substituting change in the vector of shocks for the

scalar instyΔ variable. The joint null hypothesis of full risk pooling remains

unchanged ( FH0 : β=0, γ=1 versus FAH : β>0 or γ<1). We expect this to apply

only to the socially visible subsample. The no risk pooling null hypothesis has

to be adapted for the use of proxy variables, however. Thus we test NH0 : γ=0

versus NAH : γ>0 and expect the null to apply to the socially invisible subsample

only. The coefficient estimates on the shock variables should also be

statistically significantly different from zero, indicating that individual-level

consumption is directly affected by individual-level shocks, but the magnitudes

of the relevant coefficients are indeterminate given the dummy variable nature

of the shock indicators. Rejection of the former null via an F-test is strong

evidence against full risk pooling, while rejection of the latter null via a t-test

suggests at least partial risk pooling.

In order to generate results that are directly comparable to those in the

existing literature, we first assess the extent to which any individual (visible or

invisible) pools risk with other individuals in the village. Following equation

(11), we regress the period-on-period change in individual private consumption

expenditures on the period-on-period change in farm, health, theft and

mortality shocks as well as the period-on-period change in residual village

average consumption (i.e., excluding person i). We use a fixed effects

estimator, clustering observations on the respondent’s identity, with Huber-

48

White robust standard errors. Table 2.7 shows that individual private

consumption is not statistically significantly related to individual shocks and

tracks village average consumption directly, albeit not one-for-one. A unit

change in the village average consumption corresponds to only a 0.48 change

in individual private consumption. This implies that individual consumption

varies in response to covariate shocks that affect village average consumption,

implying some social insurance. While an F-test rejects the null hypothesis of

full risk pooling, a t-test on the village average consumption also rejects the

null of no risk pooling.

As seems the norm in the existing empirical literature on risk pooling in

village economies, these data support a finding of partial risk pooling but reject

the full insurance hypothesis when one pools all households. Such a finding

could certainly be attributable to any of several well-known and quite plausible

insurance contracting problems related to search, transactions costs,

monitoring and enforcement, etc. (Fafchamps 1992, Murgai et al. 2002).

The contribution of this paper is to offer a different, potentially

complementary explanation of this familiar result. In particular, we

hypothesize that global tests applied to all sample respondents may mask

differences between subpopulations with different degrees of social

connectedness, blending socially visible individuals who enjoy reasonably

complete risk pooling with socially invisible individuals who have little or no

access to risk pooling to manage idiosyncratic shocks.

To explore that hypothesis, we now disaggregate the data into two

subsamples: socially visible and socially invisible individuals. We repeat the

49

Table 2.7: Regression tests for full and no risk pooling Dependent variable: Change in individual private consumption expenditure

All Individuals Visible Individuals Invisible Individuals

Coefficient Robust

Standard Error Coefficient Robust

Standard Error Coefficient Robust

Standard Error change in loss due to farm problems -0.056 0.045 -0.012 0.022 -0.872*** 0.110 change in total health expenses 0.002 0.002 0.001 0.001 0.020*** 0.001 change in theft of personal item 0.029 0.039 -0.008 0.015 -0.979*** 0.099 change in expenses due to sudden death 0.004 0.008 0.002 0.007 -0.614*** 0.077 change in residual village average consumption 0.483*** 0.226 0.564*** 0.268 -0.042 0.123 constant -9691.782 56172.990 -10796.500 67423.300 -39829.600*** 42733.190 Joint test for full risk pooling F(5, 301) =2.10, Prob>F=0.07 F(5, 257) =1.42 , Prob>F=0.22 F(5, 12) =223.6, Prob>F= 0.00 n=649 n=597 n=52 R2=0.010 R2=0.007 R2=0.193 Notes: ***, **, * Significant at the 1% , 5% and 10% levels, respectively.

49

50

previous exercise for each subsample, now regressing change in individual

level consumption for visible individuals on the change in their individual level

shocks and the change in residual village average consumption (i.e., excluding

the respondent) for other visible individuals. We then repeat this regression

using only socially invisible respondents. The results based on the direct

elicitation method for determining social visibility are reported in the second

and third columns in Table 2.7.20

The estimated partial correlation between changes in average

consumption of visible individuals within the village and changes in individual

consumption is more than 10 times greater for socially visible respondents than

for socially invisible persons. Furthermore, the various shocks have no

significant effect on individual private consumption even at the 10% level

among the socially visible subpopulation. Their estimated magnitudes are

smaller than in the pooled regression and much smaller than in the same

regression applied to socially invisible individuals, for whom each shock

variable is significant at the one percent level. Most notably, an F-test fails to

reject the full insurance null hypothesis for the socially visible, even at the

20% level, but overwhelmingly rejects it for the socially invisible. Indeed,

among the socially invisible, we cannot reject the no risk pooling null

hypothesis even at the 20% percent level. Visible individuals appear to achieve

something very close to full risk pooling with other visible individuals in the

20 The results based on the predicted social network structure method – classifying individuals as socially visible if SVi>0.25 for kij=1 if the predicted probability ≥ 0.8, which yielded the greatest likelihood value – are available from the authors by request. These generate qualitatively similar results. But because of the imprecision necessarily introduced by using imputed social networks, we favor the estimates based on directly elicited networks.

51

village while socially invisible persons are left out of these arrangements and

must self-insure against idiosyncratic shocks.21

By an order of magnitude, health and mortality shocks have the least

effect on consumption of the socially invisible, followed by farm and theft

shocks. The difference in the effects of the respective shocks may be attributed

to a number of factors. For instance, tradition places more value on life than

inanimate objects hence a theft, although considered unfortunate, is not viewed

as crucial to survival and hence may not garner the same level of support as a

health shock. Moreover, custom requires that one call on the sick and the

bereaved.22 The latter enforces some generalized reciprocity with regards to

health and mortality shocks. Second, whereas a health shock is readily

observable and verifiable, other than in extreme cases thefts and farm shocks

may not be obvious to all others in the village. People do not assist with shocks

they do not know happened. The directions of change as per the signs on the

coefficients of shocks are mixed. Whereas an increase in the period-on-period

change in mortality, farm and theft shocks is associated with a decrease in the

corresponding individual private consumption expenditures, an increase in the

period-on-period change in health shocks is associated with an increase in

individual private consumption expenditures. Curiously, health shocks appear

slightly overcompensated for by private consumption expenditures.

21 We also tested for full risk pooling of visible individuals with all other individuals in the entire village. The results showed that visible individuals only partially pool risk with the entire village comprising both the socially visible and invisible. This is to be expected since by definition the visible do not have any social connections with the socially invisible in the village. 22 It is customary for a sick person to send a message to friends and family informing them about his/her predicament. In addition, one has to go greet the bereaved and offer them drinks and/or cash towards the organization of the funeral. In the Akan tradition the latter is termed “wo ko bo nsawa” (translated as “going to give a donation towards the funeral”).

52

To this point, we have explored risk pooling at the village and within

the village subpopulation that is similarly (in)visible. Since the data include

information on social network structure, however, we can go one step further

and repeat the risk pooling tests, but now assess the extent to which an

individual pools risk with members of his/her social network by using the

change in average consumption within the individual’s social network, nstcΔ ,

as the key regressor in place of the change in village average consumption. In

using the directly elicited network based on random matching within sample,

this analysis is obviously conditional on being visible since the optimal

threshold for distinguishing the socially visible from the socially invisible

using that measure was SVi=0. We use the mean individual consumption for

the known random matches as a proxy for network average consumption.

The point estimates in Table 2.8, with an intermediate estimate for γ of

0.92 and a standard error of 0.06, indicate that we can reject the no risk pooling

null hypothesis at the one percent level. On the other hand we fail to reject the

full insurance null hypothesis at conventional significance levels. The

individual shock variables’ coefficients are statistically insignificantly different

from zero, with low magnitudes similar to those found in comparing socially

visible individuals against all other socially visible individuals within the

village.23

23 The regression results from the predicted social network structure method were similar. We can reject the no risk pooling null hypothesis at the one percent level, with a γ of 0.80 and a standard error of 0.23. However, we fail to reject the full insurance null hypothesis, with F(5,297) =0.81 and Prob>F=0.5422. The individual shock variables were not statistically significantly different from zero.

53

Table 2.8: Risk pooling for an individual in his directly elicited social network Dependent variable: Change in individual private consumption expenditure

Coefficient Robust Standard

Error change in loss due to farm problems -0.009 0.019 change in total health expenses 0.001 0.002 change in theft of personal item 0.015 0.022 change in expenses due to sudden death 0.005 0.009 change in residual network average consumption 0.920*** 0.059 constant -106935.100 102756.000 Joint test for full risk pooling: F(5, 257)=0.47, Prob>F = 0.80 n=597 R2=0.750

Finally, we assess the extent to which these respective networks pool

risk with other networks in the village. The preceding analysis assumes

implicitly that networks are segmented and thus that there is no spillover of

insurance benefits from one network to another through one or more members

common to multiple networks. However, interlinkages may enable networks

to reinsure each other, such that j’s mutual insurance relation with two

individuals, h and i, who do not know or directly interact with one another,

effectively creates indirect (second-order) risk pooling among h and i. We can

crudely test this hypothesis using the same method by taking the social

network as a unit and regressing the change in each individual’s social network

average consumption on the change in network-level mean farm, health, theft

and mortality shocks – thereby capturing the covariate element of shocks

within the social network – as well as the change in residual average

consumption for all networks in the village (i.e., excluding the current

network). The results are given in Table 2.9.

Network average consumption statistically significantly comoves with

village average consumption. Indeed, the point estimate is strikingly close to

54

one, which would suggest perfect reinsurance if network average consumption

were uncorrelated with average (i.e., covariate) shocks within the network.

However, we do find that network average consumption statistically

significantly covaries with expenditures related to health shocks within the

network. We thus reject both the full and no risk pooling null hypotheses,

although reinsurance among networks appears substantial in economic terms

nonetheless. This only further underscores how social visibility is necessary for

individuals to take advantage of informal risk pooling mechanisms available

through social networks, including the reinsurance apparently available

through network interlinkage.24

Table 2.9: Insurance among social networks within a village Dependent variable: Change in network average consumption expenditure

Coefficient Robust Standard

Error change in loss due to farm problems 0.001 0.013 change in total health expenses 0.002*** 0.001 change in theft of personal item 0.004 0.011 change in expenses due to sudden death 0.004 0.005 change in residual village average consumption 1.021*** 0.244 constant -3083.776 32233.200 Joint test for full risk pooling: F(5, 265) = 28.67, Prob>F = 0.00 n=597 R2=0.055

One important limitation of the foregoing analysis is that the data only

permit us to study intra-village networks. Social networks may certainly cross

village lines. In fact, it may be these ‘weak ties’ that serve the very purpose of 24 The predicted social network structure method yields qualitatively similar results. We reject the full risk pooling null hypotheses. However, the coefficient on the village average consumption was much lower, 0.023 as compared to 1.021 in the direct matching regression. In addition, we can only reject the no risk pooling null hypothesis at the 10% level.

55

spreading risk, e.g., social re-insurance, as in models of marriage markets that

consider risk management incentives in the choice of a spouse for a child

(Rosenzweig and Stark 1989). This remains a topic for future study, as the

present data do not permit exploration of social linkages beyond the village.

Subject to that important caveat, our results corroborate previous studies’

findings, using other data, that there is only partial risk pooling within rural

villages. Unpacking this result by disaggregating the data according to the

social visibility of individual respondents, however, shows that full risk

pooling is achieved by visible individuals with other visible individuals both at

village and network levels, with something very close to full risk pooling (i.e.,

reinsurance) among social networks within the village. On the other hand, the

socially invisible fail to achieve risk pooling at any economically or

statistically significant level. Table 2.10 summarizes these results.

Table 2.10: Summary of results for tests for risk pooling Full Risk Pooling No Risk Pooling Inference An individual in a village Rejected Rejected Partial risk pooling

A visible individual in a village Not rejected Rejected Full risk pooling

An invisible individual in a village Rejected Not rejected No risk pooling

An individual in a network Not rejected Rejected Full risk pooling

A network in a village Rejected Rejected Partial reinsurance

2.6 Conclusions

Although risk management is crucial to rural households in low-income

countries and the development studies literatures in economics and cognate

disciplines are rich with descriptions of informal insurance arrangements, risk

pooling through social networks may not be universally available for the

56

simple reason that not everyone is socially well connected. The simple general

equilibrium theory of risk pooling implies that individuals who are socially

visible within networks will enjoy the consumption smoothing benefits of

mutual insurance while those who are socially invisible will not. If villages

include both types of individuals, then tests of the extent of informal insurance

based on regressions that pool both sorts of individuals can lead to biased,

intermediate results suggesting partial risk pooling, as is typical of the

literature.

This paper identified a minority subpopulation of socially invisible

individuals in rural Ghana who are not widely known by other residents within

their villages. In particular, we find that poorer, younger residents who farm,

do not belong to a major clan, have been fostered and have resided in the

village for only one generation are most likely to be socially invisible.

Estimating now-standard Townsend-style regressions, we obtain the usual,

partial risk pooling result when we fail to separate the sample into socially

invisible and socially visible subpopulations. Once we separate the sample,

however, we cannot reject the full risk pooling null hypothesis for socially

visible individuals, nor can we reject the no risk pooling hypothesis for the

socially invisible. Thus in these data, village-level tests for complete mutual

insurance appear to represent a mixture model that generates misleading results

of universal partial risk pooling when the reality seems more socially

variegated, with a socially invisible minority of the population having little

access to social networks-mediated risk management, while most of the

population enjoys something economically and statistically close to complete

pooling of idiosyncratic risk. Moreover, for those in social networks,

interlinkages among social networks appear to provide quite effective

57

reinsurance against network-level covariate risk that is idiosyncratic within the

village.

In summary, this study corroborates a vast literature that finds many

individuals in rural villages use social networks to effectively insure

themselves against idiosyncratic risk, while also accommodating an oft-

overlooked literature on social exclusion and social invisibility within rural

villages that suggests insurance coverage is likely uneven among individuals.

Within-village variation in social connectedness, like within-village variation

in wealth and other attributes, appears to have a profound effect on risk

management capacity.

In policy terms, empirical evidence of Pareto efficient allocation of

idiosyncratic risk among socially visible members of networks suggests that

given binding budget constraints, interventions should target primarily (i)

village-level (or larger-scale) covariate risk that is inherently uninsurable

through social networks and (ii) idiosyncratic shocks faced by those left out of

these networks (i.e., the socially invisible). This implies a need for careful

identification of who is socially well-connected and who is not, paying

particular attention to the latter subpopulation for the purposes of targeting

interventions that might stitch up the holes in extant social safety nets. This

might be done through direct interventions to provide (quasi-)insurance to

socially invisible persons, or through efforts to improve the social integration

of individuals most likely not to be well-connected socially (e.g., recent

migrants into a community). Given our finding that young farmers who are

relatively recent settlers in a community are most likely to be socially invisible,

this might suggest possibilities involving quasi-insurance built into agricultural

credit, product sales or input delivery contracts with certain demographic

58

subgroups as an indirect means of insuring this subpopulation. Greater effort

needs to be made to identify and reach the socially invisible, lest they fall

through the apparent holes in otherwise well-functioning social safety nets.

57

REFERENCES

Alderman, H. and Paxson, C. H. (1992). Do the Poor Insure? A Synthesis of the

Literature on Risk and Consumption in Developing Countries. Washington DC, World Bank.

Alderman, H. Chiappori, P.A., Haddad, L., Hoddinott, J., and Kanbur, R. (1995).

Unitary Versus Collective Models of the Household: Is it Time to Shift the Burden of Proof? World Bank Research Observer,10, 1-9.

Bardhan, P. and Udry C. (1999). Risk and Insurance in an Agricultural Economy.

Development Economics. Oxford: Oxford University Press. Besley, T. (1995). Nonmarket Institutions for Credit and Risk Sharing in Low Income

Countries. Journal of Economic Perspectives, 9(3), 115-127. De Weerdt, J. (2005). Risk-Sharing and Endogenous Network Formation. In S.

Dercon (Ed.), Insurance Against Poverty. New York: Oxford University Press. De Weerdt, J. and Dercon, S. (2006). Risk-Sharing Networks And Insurance Against

Illness. Journal of Development Economics, 81,337–356. Deaton, A. (1992). Savings and Income Smoothing in Cote D'ivoire. Journal of

African Economies, 1, 1-24. Dercon, S. (2005). Risk, Poverty and Public Action. In S. Dercon (Ed.), Insurance

Against Poverty. New York: Oxford University Press. Dercon, S. and P. Krishnan (2000). In Sickness and In Health: Risk Sharing Within

Households In Rural Ethiopia. Journal of Political Economy, 108(4), 688-727. Doss, C. R. (1996). Intrahousehold Resource Allocation in an Uncertain Environment.

American Journal of Agricultural Economics, 78, 1335-1339.

58

Duflo, E. and Udry, C. (2004). Intrahousehold Resource Allocation in Cote d'Ivoire: Social Norms, Separate Accounts and Consumption Choices: Yale University.

Fafchamps, M. (1992). Solidarity Networks in Preindustrial Societies: Rational

Peasants with a Moral Economy. Economic Development and Cultural Change, 41 (1), 147-174.

Fafchamps, M. and F. Gubert (2007), The formation of risk sharing networks, Journal

of Development Economics, 83(2), 326-350. Gertler, P. and Gruber J. (1997). Insuring Consumption against Illness. National

Bureau of Economic Research Working Paper Series. 6035,1-57. Goldstein, M. (1999). Chop Time, No Friends: Intrahousehold and Individual

Insurance Mechanisms in Southern Ghana. Yale University. Goldstein, M., De Janvry, A. and Sadoulet, E. (2005). Is a Friend in Need a Friend

Indeed? Inclusion and Exclusion in Mutual Insurance Networks in Southern Ghana. In S. Dercon (Ed.), Insurance Against Poverty. New York: Oxford University Press.

Iliffe, J. (1987), The African poor: A history, Cambridge, Cambridge University Press. Murgai, R., Winters, P., Sadoulet E. and De Janvry A. (2002). Localized and

Incomplete Mutual Insurance. Journal of Development Economics, 67, 245-274.

Santos, P. and Barrett, C. (2004). Interest and Identity in Network Formation: Who do

Smallholders Seek out for Information in Rural Ghana?: Cornell University. Santos, P. and Barrett, C. (2006). Informal Insurance in the Presence of Poverty Traps:

Evidence from Ethiopia: Cornell University. Santos, P. and Barrett, C. (2007). Understanding the Formation of Social Networks.

Cornell University.

59

Townsend, R. (1994). Risk and Insurance in Village India. Econometrica, 62(3), 539–91.

Udry, C. (1996). Gender, Agricultural Production and the Theory of the Household.

Journal of Political Economy, 104 (5), 1010-1046. Udry, C. and Conley, T. (2005). Social Networks in Ghana. In C. Barrett (Ed.), The

Social Economics of Poverty. London: Routeledge.

60

CHAPTER THREE

Risk, Asset Poverty and Social Visibility in Ghana

3.1 Introduction

Poverty has long been the focus of policy discussions and empirical research.

Several factors have been identified as determining movement in and out of poverty.

However, using standard poverty measures to distinguish between stochastic and

structural transitions in poverty is difficult. The asset-based approach is said to

distinguish transitory poverty which eases with time due to systemic growth from

persistent structural poverty.

Under this approach there are nonlinear asset dynamics with multiple

equilibria which include a threshold that distinguishes households who can be

expected to grow out of poverty from those who are likely to remain trapped in

poverty, i.e. those caught in a poverty trap (Carter and Barrett 2006). These dynamics

may be caused, to a large extent, by risk exposure and response, attendant activity

choice and returns to endowments given incomplete financial and insurance markets.25

However, the empirical evidence regarding the existence of these dynamics is mixed.

The effect of risk pervasive in most developing countries is exacerbated by the

fact that formal financial markets are essentially missing. The poor often resort to

social networks to insure against residual risk exposure but access to these networks is

not uniform. Social visibility has been identified as a key determinant of access to

social insurance with the socially visible enjoying complete insurance whilst their

socially invisible counterparts are left uninsured (Vanderpuye-Orgle and Barrett

2007). Given the prevalence of risk, do asset poverty traps exist? If so for whom, do

they exist? Particularly, how do these asset dynamics vary with access to social

25 See Azariadis and Stachurski (2005) for a complete review.

61

insurance? Are these thresholds the same for everyone or do they vary by social

visibility? How can we identify individuals who are likely to get caught in asset

poverty traps for targeting purposes?

To address these questions, we draw from two main strands of literature. The

first is the literature on asset dynamics and poverty traps. Extant studies have

identified determinants of economic mobility. 26 Most of these studies use

income/expenditure based poverty measures which can not adequately distinguish

between structural trends and transitory movements. In addition, measurement errors

rife in income or expenditure welfare metrics renders transition variation suspect. The

asset-based approach gets at individuals who are structurally poor and who may stay

poor without policy intervention.27 Carter and Barrett (2006) extend static poverty

indices, developing an approach based on nonlinear dynamics that result in a dynamic

asset poverty line and the possibility of poverty traps. They define a dynamic asset

poverty line which is analogous to the statistic poverty line as the threshold that

distinguishes households who can be expected to grow out of poverty conditional on

asset trajectories from those who are likely to remain trapped in poverty. In essence,

the dynamic asset poverty line is the threshold at which accumulation dynamics

bifurcate leading to multiple dynamic stable equilibria. Households with assets above

this given critical asset level will tend to move ahead towards a high asset equilibrium,

whereas those below this level will fall behind towards a stable dynamic low asset

equilibrium termed a poverty trap.

Several studies have focused on identifying the threshold effect attendant with

multiple dynamic equilibria and identifying poverty traps. The empirical results have

been mixed. On one hand, some studies find no evidence of nonlinear wealth

26 See Baulch and Hoddinnot (2000) for further discussions. 27 See Carter and May (2001), Adato et al. (2006), Barrett et al. (2006), Carter and Barrett (2006) for detailed discussions.

62

dynamics associated with poverty traps (Lokshin and Ravallion 2001, Mckenzie and

Woodruff 2003, Jalan and Ravallion 2004). On the other hand, some studies find that

these dynamics exist (Dercon 1998, Lybbert et al 2004, Adato et al 2006, Santos and

Barrett 2006). For instance, Lybbert et al (2004) find herd dynamics that follow an S-

shaped curve with two stable dynamic equilibria and a threshold (i.e. an unstable

equilibrium). However, it must be noted that even within a given economy, these

dynamics may not exist for everyone. Santos and Barrett (2006) find complex wealth

dynamics with distinct convergence groups defined by individual ability. They found

that individuals with lower ability converge to a unique dynamic equilibrium at a

lower asset level whereas those with higher ability have multiple stable dynamic

wealth equilibria.

Generally, the reasons for asset poverty center on incomplete financial markets

and risk. Carter and Barrett (2006) note that poor households’ limited access to credit,

insurance and savings may encumber their ability to accumulate wealth. Poor

households who neither have access to capital to finance the high costs associated with

higher return options nor have access to insurance to cope against downside risk will

follow lower expected returns thus perpetuating poverty. This leads to the second

strand of literature on the extension of risk, social insurance and social visibility. Risk

is integral to the lives of members of agrarian communities and it may cause poverty

traps in one of several ways. Empirical evidence indicates that risk aversion is

decreasing in wealth thus the poor may forgo considerable income to reduce risk thus

trapping them in poverty. The temporal nature of some risk may also cause the poor to

adopt the quasi-option value, incurring huge losses by delaying irreversible

investments till uncertainty is resolved over time (Rosenzweig and Biswinger 1993).

This may also result in nonadoption of new technology with costly efficiency losses

through foregone output (Feder et al 1985). In addition, limited diversification options

63

conditional on wealth may result in differential rates of return with the poor being

trapped in low return opportunities offering limited accumulation potential and

neglible risk reduction (Barrett 1997, Dercon 1998).

In the absence of well-functioning insurance and credit markets, rural poor

have resorted to informal institutions such as social networks for ex ante and ex poste

responses to shocks. However, access to insurance networks is far from uniform

(Dercon and Krishnan 2000, Dercon 2005, De Weerdt 2005, Santos and Barrett 2006).

Social connectedness is central to access to social insurance. Indeed, social visibility --

the extent to which an individual is widely known in the community-- is a good

predictor of the efficiency of social insurance. Vanderpuye-Orgle and Barrett (2007)

show that the socially visible attain full risk pooling both at the village and network

levels whereas the socially invisible are left uninsured.

Social visibility may thus affect asset dynamics in several ways. Firstly, by

ensuring access to social insurance, social visibility may mitigate the need for physical

and human capital asset decumulation ex-poste of shocks. It may also enhance

productivity by providing ex-ante insurance against shocks associated with the

adoption of new technology, eliminating the need to wait for the temporal resolution

of risk. Secondly, social visibility as characterized by social connectedness may

facilitate social learning and technology adoption through social networks. Individuals

who know and interact with others are more likely to gain access to knowledge about a

high return production process and are thus more likely to switch to this technology

(Conley and Udry, 2005). Thirdly, social visibility may enhance the security of

landholding and thus encourage investments in land fertility. Well-known or well-

connected individuals, who are likely to either hold powerful positions in society or

know others who do, have more secure tenure rights and are more likely to make

investments towards improving land productivity and enjoy higher returns (Gavian

64

and Fafchamps 1996, Goldstein and Udry 2005). Finally, the basic tenets of

microcredit –i.e. the use of group lending and peer monitoring– may imply that

individuals who are not socially visible, with limited connections, are by default

rationed out of these auxiliary credit schemes adopted widely by most non-

governmental agencies. This may impose barriers to entry for diversification vis-à-vis

capital requirements or keep the less socially visible below the minimum project size

required to reap higher returns. The positive wealth gradient in social visibility,

compounded by the positive correlation between returns to assets and social visibility

may result in nonlinear asset dynamics differentiated by social visibility.

We hypothesize that asset dynamics will vary by social visibility: those who

are socially invisible will converge to a unique dynamic equilibrium at a lower asset

poverty level while those who are relatively visible will experience multiple stable

dynamic asset equilibria. We examine the role of social visibility in conditioning

wealth dynamics by using the theory of bifurcated asset accumulation strategies to

empirically identify poverty traps. We estimate nonparametric kernel regressions for

all individuals in the sample and then disaggregate the sample into socially visible and

socially invisible individuals, first using the cut-off in the social visibility continuum

identified by Vanderpuye-Orgle and Barrett (2007) and then using regression trees to

endogenously identify the optimal sample splitting level. We then identify the

respective thresholds in asset dynamics and estimate the likelihood of an individual

falling below the lower poverty thresholds for targeting purposes. We use individual

level data to allow for effect of differential social visibility within the household.

The remainder of the paper is organized as follows. The next section describes

the data used in this paper. Section 3 generates an appropriate multidimensional asset

index. Section 4 then analyzes asset dynamics based on this index whilst Section 5

65

identifies the characteristics of individuals who fall below the lower poverty

thresholds. Conclusions and policy recommendations are presented in Section 6.

The key results indicate that asset trajectories exhibit multiple dynamic stable

equilibria for village-level aggregations of individuals. Unpacking the results by social

visibility indicates that whereas the assets of the socially visible exhibit multiple stable

dynamic equilibria, the socially invisible converge towards a unique dynamic stable

equilibrium at the lower asset level. In addition, the asset levels at the respective

equilibria were generally higher for the socially invisible as compared to the socially

visible. These results are supported by a nonparametric regression tree analyses.

3.2 The Data

The data used in this paper are from a rural household survey undertaken from

July 2004 to January 2005. This was the third wave of a panel data set initiated by

Christopher Udry and Markus Goldstein. The research was conducted in the Akwapim

South District (specifically the Nsawam - Aburi area) in the Eastern Region of Ghana.

Since the early 1990s farmers in this area have been switching from the cultivation of

maize-cassava intercrop for domestic production to pineapple cultivation for export. 28

This transition involves a significant amount of risk by way of the attendant new

agronomic practices as well as exposure to global price fluctuations. It also allows for

marked improvements in asset accumulation given increased incomes associated with

pineapple cultivation.

The original sample was selected in 1996 using a two-stage procedure. Four

village clusters were purposively selected within this area based on their participation

in fruit and vegetable production as well as their array of agronomic, market access

and geographic conditions. Sixty married couples (or triples) were then randomly

28 See Goldstein and Udry (1999) for an in-depth discussion of the historical background of the area and the sampling techniques.

66

selected in each village cluster, except for the smallest village cluster where all

resident couples were interviewed. Male enumerators were assigned to male

respondents and female enumerators to female respondents to preserve gender

sensitivity and cultural norms. In all, 436 individuals were surveyed fifteen times in

the first and second waves. Of these, 372 individuals were surveyed three times in the

third wave. The sample attrition rate from the 1998-2004 interval was thus 14.68%.

Data from all three waves were used. A brief description of the modules relevant to

the analysis in this paper is given below.

3.2.1 Individuals’ Assets

Data were gathered on a variety of assets at the individual level. First we asked

about stores of farm inputs such as chemicals and seeds as well as farm equipment,

machines, motor vehicles and other equipment. We then asked for imputed values or

actual costs of equipment for business outside of farming such as distilling equipment,

commercial vehicles, sewing machines, etc. as well as stocks of trading goods. In

terms of financial assets, we asked about susu savings, expected pot sizes from esusu,

money being held by others, bank deposits, bonds, stocks or other financial assets.29

Information on the approximate current value of jewelry and cloth as well as cash on

hand was gathered.

3.2.2 Individuals’ Social Visibility

For each respondent, we randomly selected seven individuals in the sample

from the same village (without replacement).30 We then asked them whether or no

29 Susu is an informal savings mechanisms where the individual saves money with a local collecting agent to be redeemed at a latter period for an agreed fee. Esusu is similar, in this case the pot of savings is paid out on a rotating schedule amongst a group of savers. 30 Respondents were also non-randomly matched with three other village-specific “focal” individuals identified from the community-studies approach taken in a preliminary field trip as individuals in the villages from whom advice is commonly sought. We focus on the random matches in this study.

67

they knew the match. In administering the questionnaire we were sure to make the

distinction between knowing of someone (i.e. using the Akan translation of just

“having heard of the person”) and actually knowing the person. Knowing a random

match in this sense is indicative of an extant social link. This gives us a random

sample of the individuals with which they have social links. The characteristics of

these random matches, by sampling design would be representative of people with

whom they have extant social links i.e., their social networks. Even though we do not

have the overall shape or size of the network, we know about the type of people they

establish social links with. Vanderpuye-Orgle and Barrett (2007) defined social

visibility as the ratio of the number of times an individual is known when presented as

a random match to the total number of times one is presented as a random match.

3.2.3 Shocks

We use data on the realization of shocks as captured by the incidence of the

following: (i) damage caused by general farm problems; (ii) illnesses (iii) theft of

personal item stolen; (iv) contributions towards organization of funeral upon sudden

death of family member.

3.2.4 Consumption

Detailed data were collected on purchased food, general family expenses and

personal expenditures by each respondent in the household. 31 Even though these

expenditure questionnaires were administered at the individual level, with the head

and spouse(s) of head being interviewed separately regarding contributions made

towards purchasing an item, individual expenditures were not assigned. Hence, we

follow Goldstein (1999) in assigning particular items as purchases for own-

31 Recall periods varied by expenditure based on the modal frequency of purchase reported in waves 1 and 2. These essentially intra-annual expenditure were converted to nominal monthly rates.

68

consumption: alcoholic beverages, non-alcoholic pre-packaged beverages, prepared

food (from kiosks), personal care products, hair cuts, public transport, petrol, car

repairs, newspapers, entertainment, lottery tickets and kola nuts.

3.3 Asset Index

An analysis of asset dynamics requires a measure that appropriately aggregates

the assets of an individual at any given point in time and allows for

multidimensionality. Thus, we follow Adato, Carter and May (2006) in defining a

livelihood-weighted asset index. This index aggregates the respective economics

assets of an individual at time t, given by a vector At, into a uni-dimensional index

L(At).

Carter and May (2001) define an asset poverty line as the level of assets

required to produce an expected living standard equal to the poverty line. The asset

poverty line was derived by regressing livelihood of individual i at time t, lit, on a

vector of k assets for individual i at time t, Aikt.32

lit =βAit + εit (1)

where lit= t

it

PLc (2)

cit is private individual consumption expenditure and PLt is a proxy of the value of the

individual subsistence needs, in this case the universal $1/day poverty line.33 Ait is a

32 The use of the term livelihood in reference to the ratio of some income proxy to the poverty line is simply aimed at shifting the focus from money-metric welfare indicators to a broader notion of income generating strategies, entitlements and vulnerabilities associated with these strategies (Carter and May, 1999) 33 We are unable to appropriately deflate the asset values because we do not have corresponding community level prices for the assets. However, we minimally control for this by including year dummies in the subsequent regression.

69

vector of number of years of schooling, value of aggregate non-land wealth and value

of land inheritance. Table 3.1 presents the descriptive statistics for the respective

waves.

Table 3.1: Descriptive statistics: Estimating livelihood-weighted asset index

Mean Variable Definition Entire

Sample Year 1997 Year 1998 Year 2004

Consumption expenditures

Individual private consumption expenditures

128156 (608106)

19798 (27545)

26166 (24603)

393383 (1097788)

Livelihood metric

Individual consumption expenditures normalized by the poverty line

0.6 (2.1)

0.3 (0.4)

0.3 (0.3)

1.3 (3.7)

Years of schooling

Number of years of schooling

6.8 (4.7)

6.9 (4.8)

7.0 (4.8)

6.4 (4.4)

Value of non-land wealth

Value of crops, chemicals, cash, jewelry, goods to be traded and business equipment in millions of Cedis

0.8 (3.4)

0.1 (0.2)

0.2 (0.5)

2.1 (5.5)

Value of land inheritance

Total value of land inheritance in hundred millions of Cedis

2.5 (5.8)

1.3 (1.0)

1.5 (1.0)

4.5 (9.5)

Notes: Standard deviations given in parentheses

Equation 1 was estimated using a fixed effects regression, with

heteroskedasticity robust standard errors and observations clustered on the individual.

We included a polynomial expansion of the respective assets as well as interaction

terms to allow the marginal returns on these assets to vary with the level of the asset as

well as the magnitudes of other assets. The regression results in Table 3.2 indicate

that an additional year of schooling is associated with a 5.4% increase in the asset

index.

70

Table 3.2: Livelihood-weighted asset index estimation

Variables Coefficient Robust Standard

Error Dependent Variable: Livelihood-weighted asset index Assets Years of schooling, S 0.054** 0.024 Value of non-land wealth, W 0.637*** 0.051 Value of land inheritance, I 0.059** 0.025 Asset Interactions S2 0.000 0.002 W2 0.012*** 0.001 I2 -0.001*** 0.001 SxW -0.102*** 0.005 SxI -0.001 0.002 WxI -0.021*** 0.006 SxWxI 0.003*** 0.001 Other Factors Village 2 0.128 0.121 Village 3 0.000 0.117 Village 4 -0.143 0.115 Year 1998 0.038 0.096 Year 2004 0.750*** 0.120 n = 1038

R2 = 0.636 Notes: ***, **, * Significant at the 1%, 5% and 10% levels, respectively.

71

A unit increase in the value of land inheritance is associated with a 5.9% increase in

the asset index whereas a unit increase in non-land wealth is associated with 63.7%

increase in the index. The second order polynomial of the non-land wealth as well as

the interaction between education wealth and land inheritance have positive effects on

the asset index while the second order polynomial of land inheritance and the

interactions between schooling and wealth and wealth and inheritance are associated

with a decrease in the asset index. As expected, the year 2004 is associated with an

increase in the asset index.

The significance of the most of the coefficients and the overall goodness of fit

(R2 statistic) of 0.636 indicate that the model adequately estimates the weighted asset

index. It must be noted that the respective coefficients give the marginal contribution

of the respective assets to livelihood. We estimate the fitted value of the regression

function in order to generate an asset index with assets weighted by their marginal

contributions to livelihood. The latter is termed the livelihood-weighted asset index,

Lit, and is expressed in poverty line units (PLUs) such that a value of 1 means the

bundle predicts the poverty level of material wellbeing and 0.5 means the assets

predict livelihood at half the poverty line (Adato, Carter and May 2006). We use this

index as a basis for exploring asset dynamics in the next section.

3.4 Asset Poverty Dynamics

Asset dynamics were estimated using a nonparametric method to allow for

non-linearities in asset trajectories. We used a Nadaraya-Watson regression with an

Epanechnikov kernel to estimate the bivariate relationship between Lit and Lit+1,

estimating the asset dynamics for 1997-1998, 1997-2004 and 1998-2004 respectively.

72

We selected the optimal bandwidth based on Silverman’s rule of thumb as determined

by the Bounds of Stata package (Beresteanu and Manski 2000) 34 . The dynamic

equilibria are identified by the asset levels at which Lit = Lit+1.

We first estimate the regressions for all individuals in the sample. Figure 3.1 shows

the nonparametric estimates of the predicted livelihood-weighted asset index for the

1997-1998 period with the corresponding confidence band shown by the grayed out

area. A 45-degree line was overlaid to identify the points at which Li1997 = Li1998. The

horizontal axis measures initial period assets, Li1997, and the vertical axis measures

later period assets, Li1998. The asset dynamics display the expected S-shaped trajectory.

The critical asset threshold (the Micawber threshold) occurs at 0.76 PLUs. Thus,

individuals at with asset levels at 76% of the poverty line are likely to decumulate

assets, tending towards the lower poverty trap equilibrium of 0.48 PLUs. Whereas

those above this threshold may move ahead towards the upper asset equilibrium of

0.86 PLUs. However, it is uncertain if this equilibrium exists since the lower

confidence band does not cross the equilibrium at this point.

0.2

.4.6

.81

1998

Ass

et In

dex (P

overty Line Units)

0 .2 .4 .6 .8 11997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.065) Figure 3.1. Nonparametric estimates of predicted asset index: All individuals, 1997-1998 34 See Silverman (1986) for detailed discussions.

0.48 0.73 0.86

Poverty Trap

Micawber Threshold

Expected Asset Dynamics Asset Indext+1= Asset Indext

73

Figure 3.2 replicates the analyses for the 1997-2004 interval. The results are similar

in terms of the general shape of the curve as well as the location of multiple equilibria.

The lower asset poverty equilibrium is again at 0.48 PLUs. However, the unstable

critical threshold is at lower level, 0.61 PLUs and both confidence bands cross the

equilibrium. The upper asset equilibrium is also at a lower level, 0.72 PLUs. Similarly,

the asset dynamics for the 1998-2004 interval display the S-shaped trajectory with

multiple equilibria (Figure 3.3). It is interesting to note that the equilibria are also

lower than that of the 1997-1998 equilibria. The poverty trap occurs at 0.36 PLUs as

compared to 0.48 PLUs. Even though it is not certain if this difference is statistically

significant, these results may indicate a temporal component to the existence of asset

poverty, albeit nonlinear. This may seem counterintuitive to entire concept of a stable

equilibrium, on the face of it. However it should be noted that it does not negate the

existence of equilibria in the limits, yet it is plausible that the given level of equilibria

identified at any point in time may be some nonlinear function of the length of the t to

t+1 interval.

0.2

.4.6

.81

2004

Ass

et In

dex (P

over

ty L

ine

Units)

0 .2 .4 .6 .8 11997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.053) Figure 3.2. Nonparametric estimates of predicted asset index: All individuals, 1997-2004

Poverty Trap

Micawber Threshold

Expected Asset Dynamics Asset Indext+1= Asset Indext

0.48 0.61 0.72

74

0.2

.4.6

.81

2004

Ass

et In

dex

(Pov

erty

Lin

e Uni

ts)

0 .2 .4 .6 .81998 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.047) Figure 3.3. Nonparametric estimates of predicted asset index: All individuals, 1998-2004

Given variations in social visibility, these single (dominant) trajectories per

given period may distort the equilibrium levels for respective groups of individuals.

We disaggregate the data and estimate the trajectories for socially visible versus

socially invisible individuals. We first use the social visibility cut-off identified by

Vanderpuye-Orgle and Barrett (2007). We identify individuals as being socially

invisible if social visibility is equal to zero and socially visible if social visibility is

greater than zero. Figure 3.4 shows the nonparametric estimates of the predicted

livelihood-weighted asset index for the socially visible for 1997-1998. The graph is

identical to that of the aggregated data since they are the dominant subgroup.

However, the lower poverty equilibrium is slightly less at 0.46 PLUs. In addition, the

lower confidence band crosses the 45-degree line at the upper asset equilibrium. On

the other hand, the socially invisible have a much higher critical asset threshold of

0.81 PLUs as well as a higher poverty trap equilibrium of 0.51 PLUs (Figure 3.5).

Perhaps the most notable difference is the absence of the upper asset equilibrium for

Expected Asset Dynamics Asset Indext+1= Asset Indext

Micawber Threshold Poverty Trap

0.36 0.51 0.62

75

0.2

.4.6

.81

1998

Ass

et In

dex

(Pov

erty

Lin

e Uni

ts)

0 .2 .4 .6 .8 11997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.063) Figure 3.4. Nonparametric estimates of predicted asset index: Social Visibility>0,

1997-1998

0.2

.4.6

.81

1998

Ass

et In

dex (P

over

ty L

ine

Uni

ts)

0 .2 .4 .6 .81997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.054) Figure 3.5. Nonparametric estimates of predicted asset index: Social Visibility=0,

1997-1998

Expected Asset Dynamics Asset Indext+1= Asset Indext

Micawber Threshold Poverty Trap

0.46 0.73 0.86

Expected Asset Dynamics Asset Indext+1= Asset Indext

Poverty Trap

Micawber Threshold

0.51 0.81

76

the socially invisible. In essence, whereas the assets of the socially visible exhibit

multiple stable dynamic equilibria, the socially invisible converge towards a unique

dynamic stable equilibrium at the lower asset level, with the latter being slightly

higher for the socially invisible. The results for 1997-2004 are similar. Figures 3.6 and

3.7 indicate that the socially invisible have a unique stable equilibrium whilst the

socially visible have multiple equilibria. In addition, both the Micawber threshold and

lower asset poverty traps occur at a slightly higher asset levels for the socially

invisible as compared to the socially visible. Figures 3.8 and 3.9 show similar results

for 1998-2004.

0.2

.4.6

.81

2004

Ass

et In

dex

(Pov

erty

Lin

e U

nits

)

0 .2 .4 .6 .8 11997 Asset Index (Poverty Line units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.043) Figure 3.6. Nonparametric estimates of predicted asset index: Social Visibility>0,

1997-2004

Poverty Trap

Micawber Threshold

0.50 0.61 073

Expected Asset Dynamics Asset Indext+1= Asset Indext

77

0.2

.4.6

.820

04 A

sset

Inde

x (P

over

ty L

ine

Uni

ts)

0 .2 .4 .6 .81997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.060) Figure 3.7. Nonparametric estimates of predicted asset index: Social Visibility=0,

1997-2004

0.2

.4.6

.81

2004

Ass

et In

dex

(Pov

erty

Lin

e U

nits

)

0 .2 .4 .6 .8 11998 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.039) Figure 3.8. Nonparametric estimates of predicted asset index: Social Visibility>0,

1998-2004

Poverty Trap Micawber Threshold

0.51 0.68

Expected Asset Dynamics Asset Indext+1= Asset Indext

Poverty Trap

Micawber Threshold

0.38 0.50 0.62

Expected Asset Dynamics Asset Indext+1= Asset Indext

78

0.2

.4.6

.820

04 A

sset

Inde

x (P

over

ty L

ine

Uni

ts)

0 .2 .4 .6 .81998 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.064) Figure 3.9. Nonparametric estimates of predicted asset index: Social Visibility=0,

1998-2004

These results are consistent with our hypothesis -- asset dynamics vary by

social visibility. However the actual levels of the estimated dynamic equilibria do not

seem to differ much in terms of magnitude, particularly at the lower stable

equilibrium. In essence the estimated proportions of people caught in the poverty trap

do not differ much between the subpopulations. On other hand, the unstable

equilibrium is consistently higher for the socially invisible. This indicates a broader

asset domain over which people in this subpopulation would collapse toward the lower

level asset equilibrium, i.e. the poverty trap.

Next, we use regression trees to identify the optimal sample split based on

homogenous subsets, eliminating the exogenous determination of the threshold

variable and cut-off levels. This is nonparametric procedure generates sample splits to

Expected Asset Dynamics Asset Indext+1= Asset Indext

0.39 0.55

79

maximize the fit of piecewise linear regression functions.35 We use the Generalized,

Unbiased Interaction Detection and Estimation (GUIDE) algorithm described in Loh

(2002).

The asset index regression tree is presented in Figure 3.10. The numbered

circles represent terminal nodes that contain different subsamples. The empty circles

show the splitting criteria with the threshold variables and values. At each node, an

observation goes the left if the value is smaller than the threshold level, otherwise it

goes to the right. The first sample splitting variable is social visibility with a threshold

level of 0.7929. Within the subsample of those with social visibility above the 0.7929,

another split occurs on the basis of education. Within the subsample of those with

social visibility below this level, a further split occurs at 0.0955. These splits are

consistent with the existence of multiple dynamic equilibria for the socially visible. It

must be noted that the threshold level of 0.0955 is close to the level identified in

Vanderpuye-Orgle and Barrett (2007) – there are only two individuals with social

visibility between zero and 0.0955. The lack of further splits for the subsample with

social visibility less than or equal to 0.0955 is also consistent with the existence of a

unique dynamic stable equilibrium for the socially invisible.

visib≤ 0.7929

visib≤ 0.0955

4

0.1282

5

1.15

noschool= 0

6

5.45

7

5.66

GUIDE piecewise multiple linear least-squares model. At each intermediate node, acase goes to the left child node if and only if the condition is satisfied. Number initalics beneath a leaf is the sample mean of livelihood.

Figure 3.10. Asset index regression tree

35 See Hardle (1990) for a brief description of regression trees.

80

Table 3.3 presents the asset index regression models associated with the

respective terminal nodes as well as the number of observations in each subsample.

The first sample split at the 0.7929 identifies 995 individuals as socially invisible and

252 individuals as socially visible. We replicate the poverty dynamics analysis using

this threshold. Figure 3.11 shows the social visible experience multiple dynamic stable

equilibria for 1997-1998. In addition, the socially invisible also experience multiple

dynamic stable equilibria (Figure 3.12). This is to be expected since this definition of

the socially invisible lumps more observations towards the bottom of the continuum

and includes fairly well connected individuals.36 Nonetheless, the equilibria for the

socially invisible occur at higher asset levels.

Table 3.3: Asset index regression tree estimation

36 It should be noted that the statistical significance of the splits identified by the regression trees is uncertain.

Variable Coefficient t-

statistic Coefficient t-

statistic Coefficient t-

statistic Coefficient t-

statistic

Terminal Node Visibility≤0.0955 Visibility>0.0955

Visibility>0.7929, No school=0

Visibility>0.7929, No school=1

Social Visibility 12.69***

26.76 0.26*** 4.76 0.34* 1.59 1.00 1.13

Age 0.01** 2.28 0.01 0.62 0.01** 2.28 0.01** 2.05 Value of land inheritance 0.01*** 3.30 0.01** -2.53 0.01 0.80 0.01 0.56 Value of non-land wealth -0.02** -1.96 -0.01 -0.87 0.03** 2.67 0.55*** 13.92 Farmer -0.03** -2.29 0.05** 2.27 0.18*** 3.22 0.55** 2.60 Male 0.01 0.63 0.01 0.49 -0.05 -1.23 0.26* 1.65 Year 1998 0.01 1.42 0.02*** 5.23 0.01 1.20 0.02 0.14 Year 2004 0.01* 1.90 0.06*** 16.08 0.02*** 3.40 0.01 0.21 No school -0.09*** -7.78 -0.07*** -2.81 -- -- -- -- n = 224 n = 771 n = 160 n = 92

R2 = 0.811 R2 = 0.299 R2 = 0.257 R2 =0.731

81

0.2

.4.6

.81

1998

Ass

et In

dex (P

over

ty L

ine

Units)

0 .2 .4 .6 .8 11997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.064) Figure 3.11. Nonparametric estimates of predicted asset index: Social Visibility≥ 0.793, 1997-1998

0.2

.4.6

.81

1998

Ass

et In

dex

(Pov

erty

Lin

e Uni

ts)

0 .2 .4 .6 .8 11997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.045) Figure 3.12. Nonparametric estimates of predicted asset index: Social Visibility< 0.793, 1997-1998

Poverty Trap

Poverty Trap Micawber Threshold

0.53 0.69 0.86

Expected Asset Dynamics Asset Indext+1= Asset Indext

Expected Asset Dynamics Asset Indext+1= Asset Indext

Micawber Threshold

0.43 0.68 0.84

82

Figures 3.13 and 3.14 yield similar results for 1997-2004. Figures 3.15 and

3.16 also yield similar results for 1998-2004. Hence, using a high social visibility cut-

off which lumps individuals who are fairly well connected (i.e. those who are known

about 80% of the time) to the bottom results in both socially visible and invisible

experiencing multiple dynamic stable equilibria. However, the socially invisible have

consistently higher asset equilibria.

0.2

.4.6

.8

2004

Ass

et In

dex (P

overty Line Units)

0 .2 .4 .6 .81997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.053) Figure 3.13. Nonparametric estimates of predicted asset index: Social Visibility≥ 0.793, 1997-2004

0.2

.4.6

.81

2004

Ass

et In

dex (P

over

ty Line Units)

0 .2 .4 .6 .8 11997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.041) Figure 3.14. Nonparametric estimates of predicted asset index: Social Visibility< 0.793, 1997-2004

Micawber Threshold

Poverty Trap

0.48 0.59 0.68

Expected Asset Dynamics Asset Indext+1= Asset Indext

Micawber Threshold

Poverty Trap

0.50 0.61 0.70

Expected Asset Dynamics Asset Indext+1= Asset Indext

83

0.2

.4.6

.820

04 A

sset

Inde

x (P

over

ty L

ine

Uni

ts)

0 .2 .4 .6 .81998 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.053) Figure 3.15. Nonparametric estimates of predicted asset index: Social Visibility≥ 0.793, 1998-2004

0.2

.4.6

.81

2004

Ass

et In

dex

(Pov

erty

Lin

e U

nits

)

0 .2 .4 .6 .8 11998 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.040) Figure 3.16. Nonparametric estimates of predicted asset index: Social Visibility< 0.793, 1998-2004

Micawber Threshold

Poverty Trap

0.38 0.47 0.61

Expected Asset Dynamics Asset Indext+1= Asset Indext

Micawber Threshold Poverty Trap

0.36 0.51 0.62

Expected Asset Dynamics Asset Indext+1= Asset Indext

84

Finally, we split the subsample with social visibility below 0.7929 at the

0.0955 threshold to check the robustness of our original results. Figures 3.17 and 3.18

show those with social visibility below 0.0955 have a unique stable equilibrium whilst

those with socially visibility between 0.0955-0.7929 have multiple dynamic equilibria

for 1997-1998. The results for 1997-2004 and 1998-2004 are similar.37 Thus the initial

results are consistent with the regression tree analysis.

0.2

.4.6

.81

1998

Ass

et In

dex (P

over

ty Line Units)

0 .2 .4 .6 .81997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.063) Figure 3.17. Nonparametric estimates of predicted asset index: Social Visibility< 0.0955, 1997-1998

0.2

.4.6

.81

2004

Ass

et In

dex (P

over

ty Line Units)

0 .2 .4 .6 .8 11997 Asset Index (Poverty Line Units)

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.065) Figure 3.18. Nonparametric estimates of predicted asset index: 0.0955>Social Visibility≥0.793, 1997-1998 37 The graphs are available upon request from authors.

Poverty Trap

0.54

Expected Asset Dynamics Asset Indext+1= Asset Indext

Micawber Threshold

0.73

Poverty Trap

Expected Asset Dynamics Asset Indext+1= Asset Indext

Micawber Threshold

0.72 0.49 0.85

85

3.5 Who is Likely to Remain Poor?

The previous section establishes the existence of multiple dynamic equilibria

and poverty traps. It is evident that the asset thresholds vary by social visibility. Even

within these two subpopulations of socially visible and invisible individuals, some

individuals may be more likely to fall into poverty traps. We now turn to identifying

the characteristics of those who are likely to fall into poverty traps and remain the

structurally poor. It should be noted that our focus here is not on establishing causality

rather we seek to provide benchmarks for targeting purposes.

Based on the nonparametric kernel regression estimates in the previous section

we identify individuals who are at/ below the lower poverty threshold – i.e. those

caught in the poverty trap. We generated a categorical variable, poverty trap using the

respective results for the socially visible and invisible in the 1997-1998, 1997-2004

and 1998-2004 periods.

Let Γit=1 be an indicator variable that equals one if individual i is caught in the

poverty trap equilibrium at time t. Let Pr{ itl =1} be the probability that Γit =1

conditional on some individual characteristics, itX . We then estimate

Pr{ Γit =1}= )0( >+Λ ititX εβ (3)

by probit regression, where Λ is the normal CDF. We estimate a random effects

model, with observations clustered on the respondent’s identity.

The parameter estimates presented in Table 3.4 presents the summary statistics

of the variables used in the analysis. Table 3.5 presents the probit estimation of the

likelihood of falling into a poverty trap using the social visibility cut-off identified by

Vanderpuye-Orgle and Barrett (2007). We’ll focus on the results that are robust across

all three time intervals. Conditional on being socially visible, males engaged in non-

farm occupations who have resided in the village for more than one generation are

86

Table 3.4: Descriptive statistics: Estimating the likelihood of Poverty Traps Frequency (%) Variable Definition

1997-1998

1997-2004

1998-2004

Poverty Trap (visible) =1 if the livelihood-weighted index is at or below the Poverty trap threshold

for socially visible individuals, 0 otherwise

37.5 39.7 37.8

Poverty Trap (invisible) =1 if the livelihood-weighted index is at or below the Poverty trap threshold for socially invisible individuals, 0 otherwise

40.7 40.1 38.9

Poverty Trapregression tree(visible) =1 if the livelihood-weighted index is at or below the Poverty trap threshold for socially visible individuals using regression tree sample split, 0 otherwise

33.1 38.68 37.6

Poverty Trap regression tree(invisible) =1 if the livelihood-weighted index is at or below the Poverty trap threshold or socially invisible individuals using regression tree sample split, 0 otherwise

42.1 39.68 33.64

Male =1 if male, 0 otherwise 73.2 47.4 47.2 Age Respondent’s age 40.8

(14.3) 43.0

(13.2) 42.9

(11.5) Occupation Farmer Other Unemployed

=1 if farmer, 0 otherwise =1 if trader, artisan, teacher, civil servant, office or health worker, agricultural or non-agricultural labor, 0 otherwise =1 if student/ pupil, unemployed or not in the labor force

71.1 17.2

11.6

69.5 19.3

11.2

77.5 14.2

10.3 Major clan =1 if member of a major clan, 0

otherwise 68.1 79.4 79.8

Herelong =1 if not the first generation to reside in village, 0 otherwise

82.9 74.2 72.6

Fostered =1 if respondent was fostered, 0 otherwise

67.4 64.7 62.2

Shocks Farm problems Health shock Sudden death Theft of personal item

=1 if experienced problems on farm, 0 otherwise =1 if experienced a health shock, 0 otherwise =1 if experienced sudden death in the family, 0 otherwise =1 if experienced theft of personal item, 0 otherwise

72.7

52.3 76.5

22.1

69.57

48.17 56.10

15.24

62.3

60.8 65.6

28.1

Location Village 1 Village 2 Village 3 Village 4

=1 if Village cluster 1, 0 otherwise =1 if Village cluster 2, 0 otherwise =1 if Village cluster 3, 0 otherwise =1 if Village cluster 4, 0 otherwise

25.9 21.8 28.2 24.1

26.7 23.4 24.8 25.2

26.7 23.4 24.8 25.2

Notes: The standard deviations of continuous variables are given in parentheses.

Table 3.5: Poverty trap probit estimation, social visibility=0 and social visibility>0

Notes: ***, **, * Significant at the 1%, 5% and 10% levels, respectively.

Variables 1997-1998 1997-2004 1998-2004 Visible Invisible Visible Invisible Visible Invisible

Marginal Effects

Prob>|z| Marginal Effects

Prob>|z| Marginal Effects

Prob>|z| Marginal Effects

Prob>|z| Marginal Effects

Prob>|z| Marginal Effects

Prob>|z|

Dependent Variable: Poverty Trap Individual Characteristics

Male -0.267* 0.077 0.519 0.435 -0.270** 0.031 1.845*** 0.009 -0.226* 0.077 1.146** 0.032

Age 0.000 0.944 -0.015 0.511 -0.016*** 0.000 -0.020 0.428 -0.012** 0.011 -0.007 0.704 Non-farm occupation -0.543** 0.012 -1.667** 0.037 -0.521*** 0.001 -0.413 0.598 -0.454* 0.009 -1.591** 0.037 Unemployed -0.139 0.779 -0.209 0.834 -0.237 0.516 -0.246 0.860 -0.379 0.359 3.333** 0.037 Major clan -0.134 0.370 -1.259** 0.050 -0.770*** 0.000 -1.640** 0.013 -0.533*** 0.000 -1.449** 0.022 Herelong -0.408** 0.024 -0.243 0.619 -0.509*** 0.001 0.746 0.244 -0.461*** 0.002 0.323 0.535 Fostered 0.185 0.234 2.229*** 0.001 0.383*** 0.003 1.795** 0.020 0.031 0.813 0.685 0.190 Shocks

Farm problems 0.534 0.385 0.448*** 0.002 0.891 0.561

3.000*** 0.003 0.095 0.497

2.134*** 0.001

Health shock -0.088 0.516 -0.682 0.232 -0.097 0.409 -0.830 0.227 -0.093 0.442 -0.011 0.985

Sudden death 0.196 0.250 -0.494 0.496 -0.188 0.238 1.932*** 0.037 0.069 0.691 0.452 0.472

Theft of personal item 0.125 0.450 0.934*** 0.067 0.070 0.620 1.415* 0.069 0.139 0.346 1.208* 0.060 Location Village 2 -0.495** 0.024 -0.171 0.776 -0.278 0.102 -0.729 0.395 -0.483*** 0.010 -0.711 0.273 Village 3 0.378** 0.036 -1.773 0.115 0.402 0.015 -2.379** 0.028 0.209 0.202 -0.694 0.357 Village 4 0.205 0.324 0.760 0.201 0.249 0.135 1.600 0.150 0.504*** 0.003 0.909 0.151 n = 644 n = 156 n = 612 n = 103 n = 594 n = 133 Log likelihood = -231.85 Log likelihood = -25.64 Log likelihood = -331.74 Log likelihood = -36.67 Log likelihood = 309.22 Log likelihood = -32.72 Wald χ2(14) = 213.08 Wald χ2(14) = 32.67 Wald χ2(14) = 144.12 Wald χ2(14) = 32.21 Wald χ2(14) = 68.16 Wald χ2(14) = 37.78 p-value = 0.000 p-value = 0.000 p-value = 0.000 p-value = 0.000 p-value = 0.000 p-value = 0.000

87

88

consistently less likely to fall into poverty traps, across the 1997-1998, 1997-2004 and

1998-2004 periods. Contemporaneous shocks do not significantly affect asset

dynamics for the socially visible.

Conditional on being socially invisible, individuals who belong to a major clan

are less likely to fall into a poverty trap. Conversely, socially invisible individuals who

experience farm shocks or thefts are more likely to fall into a trap. These results are

robust across the 1997-1998, 1997-2004 and 1998-2004 periods. We replicated the

analysis using the first sample split from the regression tree analysis. Table 3.6 shows

the results are generally mixed for the respective periods. Focusing on the temporally

robust results indicate that being male was associated with a decreased likelihood of

falling into poverty traps, conditional on being socially visible whereas socially

invisible individuals who experience farm shocks are more likely to fall into poverty

traps.

3.6. Conclusions

Poverty alleviation has been central to policy debates for decades. Some

milestones have been reached vis-à-vis poverty favorable indices; however a lot

remains to be done, particularly with regards improving the welfare of those who may

be persistently poor. The asset-based approach allows analysts to differentiate between

stochastic and structural poverty transitions.

This paper complements the extant literature in exploring the existence of

multiple dynamic asset equilibria and poverty traps. We contribute to this body of

work by examining the role of social visibility in conditioning asset dynamics. Our

results indicate that asset trajectories exhibit multiple dynamic stable equilibria for

village-level aggregations of individuals. The general shape of the graph vis-à-vis the

Table 3.6: Poverty Trap regression tree probit estimation, social visibility≤0.7929 and social visibility>0.7929

Variables 1997-1998 1997-2004 1998-2004 Visible Invisible Visible Invisible Visible Invisible

Marginal Effects

Prob>|z| Marginal Effects

Prob>|z| Marginal Effects

Prob>|z| Marginal Effects

Prob>|z| Marginal Effects

Prob>|z| Marginal Effects

Prob>|z|

Dependent Variable: Poverty Trap regression tree Individual Characteristics Male -0.852*** 0.007 0.022 0.889 -1.380*** 0.000 0.007 0.964 -1.224*** 0.000 0.073 0.670 Age 0.008 0.538 0.001 0.919 -0.011 0.378 -0.021*** 0.000 -0.006 0.623 -0.006 0.279 Non-farm occupation -1.582** 0.047 -0.592*** 0.005 -0.434 0.342 -0.315** 0.066 -0.554 0.352 -0.395* 0.071 Unemployed 0.219 0.536 -0.159 0.712 2.430*** 0.001 0.473 0.211 1.622** 0.025 0.609 0.177 Major clan 0.553 0.343 -0.101 0.528 -0.808* 0.062 -0.822*** 0.000 -0.963* 0.075 -0.598*** 0.001 Herelong -0.095 0.850 -0.054 0.729 -0.879* 0.089 0.163 0.269 -0.271 0.533 -0.037 0.835 Fostered 0.088 0.827 0.602*** 0.000 0.276 0.397 0.358** 0.015 0.232 0.503 0.434** 0.013 Shocks

Farm problems 0.102 0.075 0.662*** 0.001 0.722 0.821 1.111*** 0.005 0.049 0.879 1.892*** 0.001

Health shock -0.070 0.822 -0.263** 0.082 -0.520 0.121 -0.679 0.732 -0.457 0.178 -0.404** 0.013 Sudden death 0.092 0.785 -0.017 0.935 -0.008 0.979 0.690*** 0.000 0.030 0.933 0.630 0.734 Theft of personal item 0.269 0.510 0.030 0.861 0.276 0.464 -0.315** 0.049 0.564 0.155 -0.331** 0.097 Location Village 2 1.169 0.303 0.164 0.413 0.533 0.361 -0.334* 0.069 0.560 0.456 -0.857*** 0.001 Village 3 0.068 0.952 0.346* 0.098 1.665** 0.006 0.024 0.898 1.328* 0.080 -0.068 0.759 Village 4 0.431 0.696 0.187 0.377 0.231 0.706 0.368* 0.050 0.899 0.236 0.298 0.157 n = 167 n = 633 n = 180 n = 535 n = 165 n =562 Log likelihood = -51.23 Log likelihood = -200.76 Log likelihood = -56.64 Log likelihood = -249.81 Log likelihood = -48.94 Log likelihood = -172.48 Wald χ2(14) = 42.99 Wald χ2(14) = 106.92 Wald χ2(14) = 50.37 Wald χ2(14) = 150.81 Wald χ2(14) = 38.87 Wald χ2(14) = 161.80 p-value = 0.000 p-value = 0.000 p-value = 0.000 p-value = 0.000 p-value = 0.000 p-value = 0.000

89

90

presence of multiple equilibria was robust across respective time intervals. However,

the actual asset levels identified as equilibria varied non-linearly with the length of the

time interval.

Unpacking the results by social visibility indicates that whereas the assets of

the socially visible exhibit multiple stable dynamic equilibria, the socially invisible

converge towards a unique dynamic stable equilibrium at the lower asset level. In

addition, the asset levels at the respective equilibria were generally higher for the

socially invisible as compared to the socially visible, particularly for the unstable

dynamic equilibrium. The latter implies that the socially invisible are more likely to

fall into poverty traps since there is a broader asset domain over which people in this

subpopulation would collapse toward the lower level asset equilibrium. However the

estimated proportions of people caught in the poverty trap do not differ much between

the subpopulations. These results are supported by a nonparametric regression tree

analyses.

We also undertake a multivariate regression analyses to identify the

characteristics of the socially visible and socially invisible individuals who are likely

to fall into poverty traps. Focusing on results that are robust across the three time

intervals, males engaged in non-farm occupations who have resided in the village for

more than one generation are consistently less likely to fall into poverty traps,

conditional on being socially visible. On the other hand, socially invisible individuals

who belong to a major clan are less likely to fall into a poverty trap. On the other

hand, socially invisible individuals who suffer a farm shock or theft are more likely to

fall into a poverty trap.

These results underscore the need for policymakers to focus on structural

determinants of poverty transitions, or the lack thereof, in order to successfully

alleviate poverty in the long run. Given limited resources there is a need to tailor

91

social interventions to individual characteristics, in this case social visibility.

Specifically there is a need to target younger, female residents who farm, do not

belong to a major clan and have resided in the village for only one generation with

interventions to hoist them out of structural poverty.

92

REFERENCES

Adato, M., Carter, M. and May, J. (2006). Exploring Poverty Traps and Social

Exclusion in South Africa Using Qualitative and Quantitative Data. Journal of Development Studies, 42, 226-47.

Azariadis, C. and Stachurski, J. (2005). “Poverty traps”. In P. Aghion and S. Durlauf

(Eds.) Handbook of Economic Growth. http://elsa.berkeley.edu/%7Echad/ azstach.pdf.

Baulch, B. and Hoddinot, J. (2000). Economic Mobility and Poverty Dynamics in

Developing Countries. In B. Baulch, and J. Hoddinot (Eds.), Economic Mobility and Poverty Dynamics in Developing Countries. London: Frank Cass Publishers.

Barrett, C. B. (1997). Food Marketing Liberalization and Trader Entry: Evidence from

Madagascar. World Development. 25 (5), 763-777. Barrett, C. B., Marenya, P., McPeak, J.G., Minten, B., Murithi, F. M., Oluch-Kosura,

W., Place, F., Randrianarosoa, J. C., Rasambianarivo, J. and Wangila, J., (2006). Welfare Dynamics in Rural Kenya and Madagascar. Journal Development Studies. 42 (1), 248-277.

Beresteanu, A. and Manski, C. F. (2000). Bounds for STATA: Draft Version 1.0.

Northwestern University. Rosenzweig, M. and Biswinger, H. (1993). Wealth, Weather Risk and the

Composition and Profitability of Agricultural Investments. Economic Journal, 103, 56-78.

Carter, M. and Barrett, C. (2006). The Economics of Poverty Traps and Persistent

Poverty: An Asset-Based Approach. Journal of Development Studies, 42 (2), 178-199.

Carter, M. and May, J. (1999). Poverty, Livelihood and Class in Rural South Africa.

World Development, 27 (1), 1-20.

93

Carter, M. and May, J. (2001). One Kind of Freedom: Poverty Dynamics in Post-

Apartheid South Africa. World Development, 29, 1987-2006. Conley, T. G. and Udry, C. R. (2005). Learning about a New Technology: Pineapple

in Ghana. Yale University. De Weerdt, J. (2005). Risk-Sharing and Endogenous Network Formation. In S.

Dercon (Ed.), Insurance against Poverty. New York: Oxford University Press, 197-216.

Dercon, S. (1998). Wealth, Risk and Activity Choice: Cattle. Journal of Development

Economics, 55, 1-42. Dercon, S. (2005). Risk, Poverty and Public Action. In S. Dercon (Ed.), Insurance

against Poverty. 439-450. New York: Oxford University Press. Dercon, S. and P. Krishnan (2000). In Sickness and In Health: Risk Sharing Within

Households In Rural Ethiopia. Journal of Political Economy 108(4): 688-727. Fafchamps, C. and Gavian, S. (1996). Land Tenure and Allocative Efficiency in

Niger. American Journal of Agricultural Economics, 78: 460-471, Feder, G., Just, R. E. and Zilberman, D. (1985). Adoption of Agricultural Innovations

in Developing Countries: A Survey. Economic Development and Cultural Change, 33 (2) 255-298.

Gertler, P. and Gruber, J. (1997). Insuring Consumption against Illness. National

Bureau of Economic Research Working Paper Series. Goldstein, M. (1999). Chop Time, No Friends: Intrahousehold and Individual

Insurance Mechanisms in Southern Ghana. Yale University. Goldstein, M., De Janvry, A. and Sadoulet, E.(2005). Is a Friend in Need a Friend

Indeed? Inclusion and Exclusion in Mutual Insurance Networks in Southern

94

Ghana. In S. Dercon (Ed.), Insurance against Poverty. New York: Oxford University Press.

Goldstein, M. and Udry, C. (1999). Agricultural Innovation and Resource

Management in Ghana. Washington D.C.: International Food Policy Research Institute.

Hardle, W. (1990). Applied Nonparametric Regression. Cambridge: Cambridge

University Press.

Jalan, J. and Ravallion, M. (2004). Household Income Dynamics in Rural China. In S. Dercon (Ed.), Insurance against Poverty. New York: Oxford University Press, 108-124.

Loh, W. (2002). “Regression Trees With Unbiased Variable Selection and Interaction

Detection”. Statistica Sinica, 12, 361-386. Lokshin, M. and Ravallion, M.(2001). Household Income Dynamics in Two

Transition Economics. World Bank. Lybbert, T., Barrett, C. B., Desta, S. and Coppock, D. L. (2004). Stochastic Wealth

Dynamics and Risk Management among a Poor Population. Economic Journal, 114 (498), 750-777.

Mckenzie, D. and Woodruff, C. (2003). Do Entry Costs Provide an Empirical Basis

for Poverty Traps? Evidence from Mexican Microenterprises. BREAD Working Paper No. 020.

Santos, P. and Barrett, C. (2004). Interest and Identity in Network Formation: Who Do

Smallholders Seek out for Information in Rural Ghana? Cornell University. Santos, P., and Barrett, C. (2006). Heterogeneous Wealth Dynamics: On the Roles of

Risk and Ability. Cornell University. Silverman, B. (1986). Density Estimation for Statistics and Data Analysis, London:

Chapman & Hall.

95

Udry, C. and Conley, T. (2005). Social Networks in Ghana. In C. Barrett (Ed.), The

Social Economics of Poverty. London: Routeledge. Vanderpuye-Orgle, J. and Barrett, C. (2007). Risk Management and Social Visibility

in Ghana. Cornell University. Zimmerman, F. and Carter, M. (2003). Asset Smoothing, Consumption Smoothing

and the Reproduction of Inequality under Risk and Subsistence Constraints. Journal of Development Economics, 71, 233-60.

96

CHAPTER FOUR

Risk, Intrahousehold Health Inequality and Social Visibility in Ghana

4.1 Introduction

If the goal of development policy is to improve the well-being of individuals

then potential intrahousehold disparities in the welfare of individuals should be taken

into account when targeting households for interventions. To this extent, policy

discussions are rife with debates on the implications of different patterns of

intrahousehold resource allocation and intrahousehold inequality as well as questions

regarding whether or not the household is the appropriate unit of empirical analyses

and policy design. Theoretical models have been used in the extant literature to

explain intrahousehold inequality. In addition there is some empirical evidence, albeit

scant, indicating that there may be systemic differences in welfare levels within the

household. However, it is not certain how these differences vary with overall

household well-being.38 To what extent do intrahousehold inequalities manifest in the

data? Are there intrahousehold Kuznets curves? For whom do they exist?

Risk exposure may explain some of the proposed patterns in the distribution of

welfare within the household. For instance, the adverse effects of shocks on health

outcomes are borne disproportionately by women and children within the household

(Dercon and Krishnan 2000, Hoddinott and Kinsey 2000, Dercon and Hoddinott

2003). Some households respond to shocks by reducing in nutritional intake and

neglecting the health needs of children, especially those of girls (Alderman and

Paxson 1992, Ellis 1998). The effects of risk may be exacerbated by the fact that

formal insurance and credit markets are essentially missing in most parts of

38 See Haddad and Kanbur (1990), Kanbur and Haddad (1992), Haddad, Kanbur and Bouis (1995) and Sahn and Younger (2007) for complete reviews.

97

developing nations. The rural poor often resort to social networks for coping with risk

but access to these networks is disparate (Goldstein 1999, Santos and Barrett 2004,

Udry and Conley 2004, De Weerdt 2005, Goldstein et al 2005). Social connectedness

as proxied by social visibility has been found to be a key factor in determining access

to social insurance (Vanderpuye-Orgle and Barrett 2007). Nonetheless it is not certain

if access to social insurance would condition intrahousehold inequality. In the event

that intrahousehold inequalities exist, how would social visibility affect the

distribution of well-being within the household and how would the latter vary with

overall household well-being?

In addressing these issues, we draw on several strands in the literature. The

first focuses on the extensions of inequality, intrahousehold resource allocation and

intrahousehold inequality. Kuznets (1955) hypothesized that the relationship between

inequality and average well-being is characterized by an inverted U-shaped curve;

with inequality increasing as welfare increases and then declining at some higher

level. This inverse U-shaped curve is often referred to as the Kuznets curve. There is

ample empirical evidence to support this hypothesis using time series and cross-

sectional data at the national level.39 Analogous theoretical relationships have been put

forward using models of intrahousehold allocation. Using the unitary household

model, intrahousehold inequality could reflect an efficient arrangement designed to

improve aggregate welfare. It may reflect different distributions of energy-intensive

activities such that households maximizing a common objective function allocate

marginal calories to men in order to enhance productivity (Pitt, Rosenzweig and

Hassan 1990). Using cooperative and noncooperative bargaining models, Kanbur and

Haddad (1990) showed that bargaining models predict Kuznets’ inverse-U relationship

as a result of interactions between the effects of increases in the total resources being

39 See Anand and Kanbur for a complete review.

98

bargained over and changes in bargaining positions. However the empirical evidence

supporting the intrahousehold Kuznets curve has been mixed.

Variations in these results may be attributed in part to challenges in identifying

a suitable econometric procedure and welfare metric. Haddad and Kanbur (1990) find

tentative empirical support for an inverse-U relationship using caloric intake data

standardized by age, sex and physiological need requirements however they did not

fully account for gender and age differentiation in energy intensity and activities.

Kanbur and Haddad (1992) using activity-pattern information to construct energy

expenditures for each household member concluded that the continuity requirements

of the spline model may be driving the exhibited continuous inverse-U shape. Haddad,

Kanbur and Bouis (1995) improving both variable construction and variable choice by

adjusting for energy expenditures and using total household expenditures per capita as

a measure of overall household welfare found no evidence of the inverse-U shape.

Similarly, Sahn and Younger (2007) using the body mass index (BMI) as a summary

measure of caloric consumption net of needs found no evidence of the intrahousehold

Kuznets curve.

These studies generally estimate a single (dominant) curve which aggregates

all households within a given sample (or country). One could argue that there might be

an endogenous threshold within a given sample, above which households may exhibit

a Kuznets curve and below which no evidence would be found. For instance upon

disaggregating the data, Sahn and Younger (2007) note that there is some indication

that these curves might exist for better off countries – there was a hint of down-turn in

inequality at very high levels of mean BMI.40 It may be that these curves exist only for

a subset of the population. The key issue may be to empirically identify these subsets.

40 The small number of observations and the large standard errors at the tails did not allow for the validation of this hypothesis.

99

In this paper, we explore the effect of differences in risk exposure and the attending

compensating behavior.

This leads to the second strand literature on the effects of shocks directly on

health outcomes and indirectly through the budget constraint and consumption or

nutrient intake. Principal results that emerge indicate that (i) uninsured shocks

adversely affect household health outcomes and (ii) these effects are not equitably

distributed within the household (Behrman and Deolalikar 1990, Alderman and

Paxson 1992, Ellis 1998, Gertler and Gruber 2002, Dercon and Krishnan 2000,

Hoddinott and Kinsey 2000, Dercon and Hoddinott 2003). For instance, some families

use the body as ‘a store of wealth’– feasting when prices are low in anticipation of a

shock and fasting when prices are high (Dercon and Hoddinott 2003). The effects of

shocks have been associated with significant fluctuations in body mass with

individuals within poorer household, especially women, being unable to smooth their

consumption over time and within the household (Dercon and Krishnan 2000,

Hoddinott and Kinsey 2000, Dercon and Hoddinott 2003). Access to social insurance

may moderate the effect of shocks on the distribution of health, particularly for the

liquidity constrained with no little or no private insurance.

The third strand of literature is the extension of risk management and social

insurance. Social networks have been identified as a locus of insurance in lieu of

formal financial and insurance markets. These networks are fostered by kinship ties,

ethnicity, geographical proximity, religion, and gender groups, inter alia (Goldstein

1999, Santos and Barrett 2004, Udry and Conley 2004, De Weerdt 2005). However,

some marginal groups are excluded from these networks (Dercon 2002, De Weerdt

2005, Santos and Barrett, 2006). Social connectedness has been found to be a crucial

element to social insurance. Conceptualizing exclusion on the basis of social visibility,

Vanderpuye-Orgle and Barrett (2007) showed that wealthier, older residents who

100

belong to a major clan, have resided in the village for more than one generation and

are not engaged in farming are more likely to be socially visible. In addition, the

socially visible attained full risk pooling both at the social network and village levels

whereas the socially invisible were left uninsured. In essence, social visibility is an

indicator of the extent to which an individual may be insured from risk through social

networks and this may affect the nature of intrahousehold inequality.

This paper complements existing literature that use non-income measures to

examine intrahousehold inequality. It builds on the extant analyses by assessing

whether there is a threshold effect in the existence of Kuznets curves at the household

level. The goal of the paper is twofold: (i) to assess whether Kuznets curves exist and

for whom they exist (ii) to assess the determinants of intrahousehold inequality. We

hypothesize that access to insurance as proxied by social visibility will condition the

effect of shocks on the distribution of health outcomes and the variation of the latter

with overall household well-being.

The remainder of the paper is organized as follows. The next section presents

the analytical framework. Section 4.3 describes the data used in this paper. Section 4.4

presents the specification of the empirical model and results and Section 4.5 concludes

with a summary of the results and policy implications.

4.2 Analytical Framework

4.2.1 Bivariate Analysis of Intrahousehold Inequality

First, we identify an appropriate measure of inequality. We follow Haddad,

Kanbur and Bouis (1995) and Sahn and Younger (2007) in adapting the Generalized

Entropy (GE) class of measures to estimate the level of intrahousehold health

inequality. Where GE is given as follows:

101

yi is the value of the welfare index for individual i, μ is the mean value for the

household, f (yi) represents the population share of the individual i in the household, in

this case the inverse of the household size, and K is the number of households.

We use the mean logarithm deviation (i.e. GE with c=0). This is particularly

sensitive to changes at the bottom of the distribution. It must be noted that the GE

class of entropy measures is additively decomposable into within-group and between-

group components.

For K exogenously given households the GE can be decomposed as following:

The first term on the right-hand-side of the above equation is the within-group

inequality and second term is the between-group inequality.

Where

( )GE w I I e egg

K

g k k= +=∑

11 1μ μ,...,

w

f c

f c

f c

g

gg

c

gg

g

=

⎧

⎨

⎪⎪⎪⎪⎪⎪⎪⎪

⎩

⎪⎪⎪⎪⎪⎪⎪⎪

⎛

⎝⎜

⎞

⎠⎟ ≠

⎛

⎝⎜

⎞

⎠⎟ =

=

μμ

μμ

0 1

1

0

,

( )

( )

( )

GE

f yy

c

f yy y

c

f yy

c

ii

c

i

K

ii i

i

K

ii

i

K

=

⎛⎝⎜

⎞⎠⎟ −

⎡

⎣⎢⎢

⎤

⎦⎥⎥

≠

⎛⎝⎜

⎞⎠⎟

⎛⎝⎜

⎞⎠⎟ =

⎛⎝⎜

⎞⎠⎟ =

⎧

⎨

⎪⎪⎪⎪

⎩

⎪⎪⎪⎪

=

=

=

∑

∑

∑

μ

μ μ

μ

1 0 1

1

0

1

1

1

,

log

log

102

Ig is the inequality in the gth household, μg is the mean of the gth household, eg is a

vector of 1's of the length ng, and ng is the population of the gth household.

Next, we pick the Body Mass Index (BMI), Quetelet Index, as the health

outcome to be analyzed. The BMI is measured as the individual’s weight in kilograms

divided by squared height in meters and is a reliable measure of nutrient intake and

health which also captures chronic energy deficiency (James, Ferro-Luzzi and

Waterlow 1988). Although the BMI captures only consumption relating to food and

health, it adequately reflects intrahousehold allocation of resources relative to need. It

is also easy to measure fairly accurately at the individual level as compared to

consumption and income.41 We standardize the BMI for age and gender using the Epi

info software made available by the Center for Disease Control. Using a 20-year old

female as the reference group we generate the following:

BMI=F-1a*,g* (Fa,g(bmi)) (4)

where BMI is the standardized BMI, F is the distribution function of BMI in the WHO

reference population for age a and gender g, bmi is the individual’s body mass index,

a* is 20years and g* is female. We use this as a basis for estimating the Kuznets

curve.

4.2.2 Determinants of Intrahousehold Health Inequality

Our analytical approach focuses on the determinants of health outcomes. We

draw from the theory on health demand functions using a model akin to that of

Thomas, Lavy and Strauss (1996) and Foster (1995).

Suppose households maximize intertemporally additive expected utility:

Ui=v(Ui1,Ui2,...,UiT) for t=1,…, T (5)

41 See Sahn and Younger (2007) for a comparison of the BMI to other welfare metrics.

103

Where utility at time t which is a concave function dependent on the

household’s consumption of goods (cit), health outcome (hit) and some preference

shifters based on household characteristics such as life cycle position, education,

occupation, gender (Xit) and is given as follows:

Uit = uit (cit, hit; Xit) (6)

Suppose households maximize expected utility subject to constraints by which

inputs such as nutrients generate health outcomes via a health production function Hit.

Where health in period t is a function of previous health outcome (Hit-1), physical

inputs such as nutrients (git), energy expenditure (Lit), environmental factors such as

sanitation (Eit) which may have a direct impact on health outcomes conditional on

inputs, unobserved household characteristics such as inherent health or immunities (εi)

as well as observed household characteristics such as life cycle position, household

size and dependency ratio (Xit) The latter would also include social visibility which

would affect health outcomes by mitigating the need for either ex ante responses to

shocks such as using the body as a store of wealth or ex poste responses such as

reducing nutritional intake of some household members. The health production

function is thus given as follows:

Hit=h(Hit-1, git, Lit, Xit, Eit, εi) (7)

In addition, the household’s budget constraint at time t is given as follows:

Wit+1=Wit (1+rt) + yit(Hit-1, wit, Lit ,sit) - ptcit (8)

Where Wit and Wit+1 is wealth in period t and t+1, respectively, yit is the income

level in period t, wit and sit are levels of income-generating assets and transitory

income shocks faced in period t, ptcit is the expenditure on consumption bundle ci at

prices pt and rt is the interest rate. The constrained maximization problem yields a set

of necessary conditions which when solved lead to reduced form goods and adult

health demand functions. The latter is given as follows:

104

Hit = ft (Hit-1, Wit, Xit, Eit, εi) (9)

Thus health status in period t is function of previous period’s health outcome

(Hit-1), wealth (Wit), a vector of household characteristics including social visibility

(Xit), environmental factors (Eit) and unobserved household characteristics (εi). The

main implication of this framework is that social visibility should have a statistically

significant effect in determining health outcomes.

4.3 The Data

The data used in this paper are from a rural household survey undertaken from

July 2004 to January 2005. This was the third wave of a panel data set initiated by

Christopher Udry and Markus Goldstein. The research was conducted in the Akwapim

South District (specifically the Nsawam - Aburi area) in the Eastern Region of Ghana.

Since the early 1990s farmers in this area have been switching from the cultivation of

maize-cassava intercrop for domestic production to pineapple cultivation for export.

This transition involves a significant amount of risk by way of the attendant new

agronomic practices as well as exposure to global price fluctuations, hence the need

for insurance.42

The original sample was selected using a two-stage procedure. Four village

clusters were purposively selected within this area based on their participation in fruit

and vegetable production as well as their array of agronomic, market access and

geographic conditions. Sixty married couples (or triples) were then randomly selected

in each village cluster, except for the smallest village cluster where all resident

couples were interviewed. Male enumerators were assigned to male respondents and

42 See Goldstein and Udry (1999) for an in-depth discussion of the historical background of the area and the sampling techniques.

105

female enumerators to female respondents to preserve gender sensitivity and cultural

norms.

Three rounds of data were collected at approximately eight week intervals with

372 individuals in the first. 371 and 350 individuals were surveyed in the second and

third rounds, respectively. We use data from the first and third rounds for during

which anthropometric information was collected. The sample attrition rate between the

first and third rounds was 5.9%. A brief description of the modules relevant to the

analysis in this paper is given below.

4.3.1 Anthropometrics

Anthropometric data was collected on all members of the households in rounds

one and three, respectively. We collected height and weights of each individual using

standard procedures uniform across all enumerators in the study. The ideal would have

been to use stadiometers and medical scales however we didn’t have enough of them,

for as long as we needed and they were too heavy to be carried to each household.

Instead we used regular tape measures for measuring heights: the individuals were

made to stand against a wall, we then marked the top of their heads on the wall and

measured the distance from the base of the wall to the mark using the tape. We used

bathroom scales for the weights: we first set the scale on a flat concrete surface, reset

it to zero using no parallax as the standard and had the individuals stand upright on it.

4.3.2 Social Visibility

For each respondent, we randomly selected seven individuals in the sample

from the same village (without replacement). 43 We then asked them a series of

43 Respondents were also non-randomly matched with three other village-specific “focal” individuals identified from the community-studies approach taken in a preliminary field trip as individuals in the villages from whom advice is commonly sought. We focus on the random matches in this study.

106

questions about each match, preceded by their knowledge of the match i.e., “Do you

know__?”. In administering the questionnaire we were sure to make the distinction

between knowing of someone (i.e. using the Akan translation of just “having heard of

the person”) and actually knowing the person. Knowing a random match in this sense

is indicative of an extant social link. This gives us a random sample of the individuals

with which they have social links. The characteristics of these random matches, by

sampling design would be representative of people with whom they have extant social

links i.e., their social networks. Even though we do not have the overall shape or size

of the network, we know about the type of people they establish social links with.

Using this data, Vanderpuye-Orgle and Barrett (2007) estimate a continuum of

social visibility as the ratio of the number of times the respondent is known by others

when presented as a random match to the number of times the respondent’s name was

drawn in the random matching process. There were some differences, albeit highly

correlated, in social visibility within the household. We use the mean value to

represent the household’s level of social visibility.

4.3.3 Shocks

To measure shocks to income we focus on the causes of income shocks as

against the actual changes in income. Respondents were asked about a series of

morbidity, mortality, on-farm problems and theft of personal items. It must be noted, it

is likely that these self-reported shocks are likely to be correlated with wealth. For

instance, the wealthier are more likely to have items that may be stolen. In addition,

morbidity may be endogenous to BMI. In this paper we use data on the following: (i)

whether or not they experience farm problems, (ii) whether or not they made a

contribution towards organizing the funeral of a close relative who died suddenly.

107

These shocks were highly correlated within the household thus we use data on shocks

faced by the household head.

4.3.4 Other Sample Characteristics

This above was supplemented with data from the household roster, assets and

family background questionnaires. We used data from the first and third round of this

wave for which there is corresponding anthropometric information.

4. 4 Empirical Model and Results

4.4.1 Analysis of Intrahousehold Inequality

To assess the existence of the Kuznets curves, we estimate a bivariate

nonparametric relationship between the mean household BMI (i.e. our measure of

household well-being) and the mean log deviation of the standardized BMI using a

Nadaraya-Watson regression with an Epanechnikov kernel overlaid by a quadratic fit

of the regression equation. We selected the optimal bandwidth based on Silverman’s

rule of thumb as determined by the Bounds of Stata package (Beresteanu and Manski

2000)44. We also replicated the analysis using mean per capita household expenditure

for purposes of comparison.

Figure 4.1 shows the graph for all households in the sample. The pattern

indicates that intrahousehold inequality increases as the mean household BMI

increases. This pattern is not consistent with the inverted U-shaped Kuznets curve

however it is consistent withthe results from Sahn and Younger(2007) which indicates

an increasing relationship between intrahousehold inequality and mean household

well-being. On the other hand the graph using the household expenditure percentile

shows some curvature, albeit very slight (Figure 4.2).

44 See Silverman (1986) for detailed discussions.

108

0.0

2.0

4.0

6M

ean

Log

Dev

iatio

n

16 18 20 22 24 26Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.5311784864565047) Figure 4.1. Nonparametric estimates of intrahousehold inequality and mean household

BMI, all households

-.4-.2

0.2

.4M

ean

Log

Dev

iatio

n

.4 .45 .5 .55 .6Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0023179840992232) Figure 4.2. Nonparametric estimates of intrahousehold inequality and household

expenditure percentile, all households

109

To determine whether the Kuznets curve exists for some and not for others, we

disaggregated the data into socially visible and socially invisible households. We used

the mean social visibility index of the household instead of the head of the

household’s to allow for the fact that the heads and respective spouses may have

different levels of social connectedness. We used regression trees to identify the

optimal sample split based on homogenous subsets thus eliminating the exogenous

determination of the threshold variable and cut-off levels. This nonparametric

procedure generates sample splits to maximize the fit of piecewise linear regression

functions.45 We used the Generalized, Unbiased Interaction Detection and Estimation

(GUIDE) algorithm described in Loh (2002) to estimate the regression trees.

The regression tree for the mean log deviation is presented in Figure 4.3. The

numbered circles represent terminal nodes that contain different subsamples. The

empty circles show the splitting criteria with the threshold variables and values. At

each node the, an observation goes the left if the value is smaller than the threshold

level, otherwise it goes to the right. The only sample splitting variable is social

visibility with a threshold level of 0.4875.

visib≤ 0.4875

2

2.97E-03

3

0.0315

GUIDE piecewise multiple linear least-squares model. At each intermediate node, acase goes to the left child node if and only if the condition is satisfied. Number initalics beneath a leaf is the sample mean of theilmeasu.

Figure 4.3: Intrahousehold inequality regression tree 45 See Hardle (1990) for a brief description of regression trees.

110

The graph showing the bivariate nonparametric relationship of mean household

BMI against the mean log deviation of the standardized BMI for socially visible

households in Figure 4.4 maps a delicate inverted-U shape, indicative of Kuznets

curve. This is consistent with our hypothesis. Given access to social insurance, the

socially visible manage to reduce disparities in health outcomes as the mean

household BMI, our non-income welfare metric, increases beyond a certain level.

However, this result is not robust to alternative measures of welfare. The pattern of

intrahousehold inequality with increasing per capita expenditure is essentially flat

(Figure 4.5).On the other hand, intrahouseholds inequality increases with mean

household BMI for the socially invisible households (Figure 4.6). Figure 4.7 shows

the pattern of intrahousehold inequality is essentially flat relationship with increasing

per capita expenditure.

.01

.02

.03

.04

.05

Mea

n Lo

g D

evia

tion

16 18 20 22 24 26Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.5566200442016269) Figure 4.4. Nonparametric estimates of intrahousehold inequality and mean household

BMI, socially visible households

111

-.1-.0

50

.05

.1M

ean

Log

Dev

iatio

n

.35 .4 .45 .5 .55 .6Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0045625218129007) Figure 4.5. Nonparametric estimates of intrahousehold inequality and household

expenditure percentile, socially visible households

-.02

-.01

0.0

1.0

2.0

3M

ean

Log

Dev

iatio

n

18 20 22 24 26Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.7109861203703327) Figure 4.6. Nonparametric estimates of intrahousehold inequality and mean household

BMI, socially invisible households

112

-.2-.1

0.1

.2

Mea

n Lo

g Dev

iatio

n

.38 .4 .42 .44 .46Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0020687822185208) Figure 4.7. Nonparametric estimates of intrahousehold inequality and household

expenditure percentile, socially invisible households

Pairwise statistical comparisons of intrahousehold inequality at the 10th, 50th

and 90th percentile of mean household BMI and per capita expenditure are consistent

with these results (Table 4.1). The parametric estimates presented in Table 4.2 and

4.3 also provide qualitative support of these results vis-à-vis the signs of the

coefficients however the latter are not statistically significant.

Table 4.1: Comparisons of test points of intrahousehold inequality

Mean BMI Expenditure per capita

10th vs. 50th percentile

50th vs. 90th percentile

10th vs. 50th percentile

50th vs. 90th percentile

All Individuals I I I* D* Visible Individuals I D* I* D* Invisible Individuals I* I* I D

113

Table 4.2: Parametric estimates of intrahousehold inequality by mean household BMI

Notes: ***, **, * statistically significant at the 1%, 5% and 10% levels Coefficients multiplied by 100

Table 4.3: Parametric estimates of intrahousehold inequality by per capita expenditure

Notes: ***, **, * statistically significant at the 1%, 5% and 10% levels Coefficients multiplied by 100

Coefficient

Robust Standard Error Coefficient

Robust Standard Error Coefficient

Robust Standard Error

Dependent Variable: Log Mean Deviation All Individuals Socially Visible Socially Invisible Mean BMI 0.605 0.017 0.566 0.008 0.595 0.013 Mean BMI squared 0.016 0.000 -0.017 0.001 0.014 0.001 Constant -2.915 0.173 -1.127 0.082 -5.832 0.135 n =238 n =196 n =42 R2 = 0.009 R2 = 0.042 R2 = 0.041

Coefficient

Robust Standard Error Coefficient

Robust Standard Error Coefficient

Robust Standard Error

Dependent Variable: Log Mean Deviation All Individuals Socially Visible Socially Invisible Household Expenditure Percentile -0.161 1.806 -14.986 1.326 -31.469 2.247 Household Expenditure Percentile squared 0.036 1.843 -4.312 1.48 26.8388 2.629 Constant 0.088 0.437 8.296 0.302 8.792 0.483 n =238 n =196 n =42 R2 =0.028 R2 =0.072 R2 =0.062

114

The extant literature highlights the differential effect of shocks on women and

children household. To explore the extent to which this manifests in relative BMI

ratios over different ranges of welfare, we generate the ratios of male to child BMI,

female to child BMI and male to female BMI. First we estimate the kernel regression

for all households. The graphs for the male to child and female to child BMI ratios are

striking similar. Figures 4.8 and 4.10 show a convex relationship between the

respective ratios and mean household BMI. The high ratio at low mean household

BMI could be indicative of an allocation based on labor productivity in times of

relative scarcity, the need for this strategy diminishes as welfare increases. The high

ratio at low mean household BMI could reflect cultural tendency for people in the

survey areas to put on some weight to reflect their wealth and status in society, with

adults gaining weight much faster than children. The latter is consistent with the

increasing ratios with per capita expenditures in Figures 4.9 and 4.11. Similarly, the

male to female ratio displays a convex pattern with increasing mean household BMI

(Figure 4.12), although the relationship with expenditure is flat (Figure 4.13).

Disaggregating the sample, the socially visible households have identical patterns of

male to child BMI, female to child BMI and male to female BMI as the entire sample

(Figures 4.14-4.19). The male to child BMI and female to child BMI ratios remain

essentially flat with increasing mean household BMI and per capita expenditure for

the socially invisible households (Figures 4.20-4.23). Even though, they are face

significant uninsured risk there is no indication that socially invisible households

discriminate against children. It may be that they overcompensate for risk exposure by

protecting children in the allocation of limited nutritional and health inputs that affect

body mass. This overcompensating behavior may also be gleaned from Sahn and

Younger’s (2007) observation that at the children’s body mass is highest relative to

both male and female adults among the most resource constrained. They note that

115

“among poor households where command over resources, including food is most

limited, there is an attempt to protect the most vulnerable members of the household,

the young children, from food and related stresses that contribute to low BMI”( Sahn

and Younger 2007, p 13). On the other hand, the male to female ratio increases with

increasing welfare and then declines (Figures 4.24 and 4.25).

The different patterns of intrahousehold inequality for socially visible and

socially invisible households suggest that the level of inequality in the respective

households might differ systematically. We decompose overall inequality into within

and between-group components to assess the share of overall inequality explained by

intrahousehold inequality. Table 4.4 shows that over 75% of total inequality in the

entire sample and among socially visible households may be attributed to

intrahousehold inequality, whereas a little over 50% of total inequality among socially

invisible households may be attributed to inequality between households.

11.5

22.5

3

Male

to C

hild B

MI R

atio

18 20 22 24Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.5411384864562047) Figure 4.8. Nonparametric estimates of male to child BMI and mean household BMI, all households

116

1.2

1.4

1.6

1.8

22.

2M

ale

to C

hild

BM

I Rat

io

.35 .36 .37 .38 .39 .4Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0064632280128579) Figure 4.9. Nonparametric estimates of male to child BMI and household expenditure percentile, all households

11.

52

2.5

3Fe

mal

e to

Chi

ld B

MI R

atio

18 20 22 24Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.5411384864562047) Figure 4.10. Nonparametric estimates of female to child BMI and mean household BMI, all households

117

1.2

1.4

1.6

1.8

22.

2Fe

mal

e to

Chi

ld B

MI R

atio

.35 .36 .37 .38 .39 .4Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0064632280128579) Figure 4.11. Nonparametric estimates of female to child BMI and household expenditure percentile, all households

.98

.99

11.

011.

02M

ale

to F

emal

e BM

I Rat

io

16 18 20 22 24 26Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.5411384864562047) Figure 4.12. Nonparametric estimates of male to female BMI and mean household BMI, all households

118

.9.9

51

1.05

Mal

e to

Fem

ale

BM

I Rat

io

.3 .35 .4 .45 .5Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0064632280128579) Figure 4.13. Nonparametric estimates of male to female BMI and household

expenditure percentile, all households

.51

1.5

22.

53

Mal

e to

Chi

ld B

MI R

atio

18 20 22 24Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.5566200442016269) Figure 4.14. Nonparametric estimates of male to child BMI and mean household BMI, socially visible households

119

.51

1.5

22.

53

Mal

e to

Chi

ld B

MI R

atio

.35 .36 .37 .38 .39 .4Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0040055326519716) Figure 4.15. Nonparametric estimates of male to child BMI and household expenditure percentile, socially visible households

.51

1.5

22.

53

Fem

ale

to C

hild

BM

I Rat

io

18 20 22 24Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.5566200442016269) Figure 4.16. Nonparametric estimates of female to child BMI and mean household BMI, socially visible households

120

.51

1.5

22.

53

Fem

ale

to C

hild

BM

I Rat

io

.35 .36 .37 .38 .39 .4Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0040055326519716) Figure 4.17. Nonparametric estimates of female to child BMI and household expenditure percentile, socially visible households

.98

.99

11.

011.

021.

03M

ale

to F

emal

e BM

I Rat

io

16 18 20 22 24 26Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.5566200442016269) Figure 4.18. Nonparametric estimates of male to female BMI and mean household BMI, socially visible households

121

.85

.9.9

51

1.05

Mal

e to

Fem

ale

BM

I Rat

io

.3 .35 .4 .45 .5Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0040055326519716) Figure 4.19. Nonparametric estimates of male to female BMI and household expenditure percentile, socially visible households

-50

510

Mal

e to

Chi

ld B

MI R

atio

18 20 22 24Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.6129779981451228) Figure 4.20. Nonparametric estimates of male to child BMI and mean household BMI, socially invisible households

122

02

46

Mal

e to

Chi

ld B

MI R

atio

.36 .37 .38 .39Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0057232964924509) Figure 4.21. Nonparametric estimates of male to child BMI and household expenditure percentile, socially visible households

-50

510

Fem

ale

to C

hild

BM

I Rat

io

18 20 22 24Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.6129779981451228) Figure 4.22. Nonparametric estimates of female to child BMI and mean household BMI, socially invisible households

123

02

46

Fem

ale

to C

hild

BM

I Rat

io

.36 .37 .38 .39Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0057232964924509) Figure 4.23. Nonparametric estimates of female to child BMI and household expenditure percentile, socially visible households

.98

.99

11.

011.

02M

ale

to F

emal

e B

MI R

atio

18 20 22 24 26Mean Household BMI

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.6129779981451228) Figure 4.24. Nonparametric estimates of male to female BMI and mean household BMI, socially invisible households

124

.97

.98

.99

11.

01M

ale

to F

emal

e BM

I Rat

io

.34 .36 .38 .4Household Expenditure Percentile

Nadaraya-Watson estimates using Epanechnikov kernel with Silverman’s optimal bandwidth (h=0.0057232964924509) Figure 4.25. Nonparametric estimates of male to female BMI and household expenditure percentile, socially visible households

Table 4.4: Decomposition of overall inequality (%)

4.4.2 Determinants of Intrahousehold Inequality

So far we have examined bivariate relationships between intrahousehold

inequality and welfare, using mean household BMI and per capita expenditure. This

allows us to assess how intrahousehold inequality varies with welfare. However, it

doesn’t illuminate on the underlying factors driving the observed levels of

intrahousehold inequality. We use the theory on household health demand functions to

GE(-1) GE(0) GE(1) GE(2) Within Between Within Between Within Between Within Between All Individuals 85.311 14.689 80.895 19.137 79.060 20.940 78.421 21.509 Socially Visible 85.067 14.955 80.116 19.884 78.097 21.903 77.387 22.509 Socially Invisible 48.043 51.957 45.321 54.679 43.011 56.989 40.941 58.554

125

examine these factors. Recall that we seek to estimate the following health demand

function:

Hit = ft (Hit-1, Wit, Xit, Eit, εi) (10)

We write an estimable version of equation (5) as

Hit =βH Ht-1 + βw Wit +βX Xit +βissit + +βiVVi+ εi (11)

Suppose that the household derives utility from reducing intrahousehold

inequality in health, ceteris paribus. The we can estimate an analogous model where

the health outcome at the household level is intrahousehold inequality represented by

the log mean deviation. Where Wit is a vector of the household’s stock of income-

generating assets, Xit is a vector of household characteristics including social visibility,

and sit are transitory shocks that affect intrahousehold inequality through the budget

constraint and Vi is the village dummy for household i.

It must be noted that social visibility is itself a function of the individual’s

characteristics. 46 We use the instrumental variables approach to deal with the

endogeneity of social visibility. In order to avoid a finite sample bias which is a

function of the number of instruments, we select two variables which are strong

predictors of the endogenous variable. We use data on whether or not an belong to a

major clan and the whether or not the respondent’s parents held a village office as

instruments for social visibility. These variables affect social visibility but are not

correlated with BMI. We estimate the following system of structural equations.

Socially Visibility = α0 + α1age + α2 gender +α3 education + α4occupation +

α 5wealth + α6major clan + α7parent’s office +ε (12)

Mean Log Deviation BMI = γ0 + γ1age + γ2 gender+ γ3 education + γ4occupation +

γ5wealth + γ6dependency ratio+ γ7household size + γ8shock +

γ9 Social Visibility + γ9village+ v (13)

46 See Vanderpuye-Orgle and Barrett (2007) for a detailed description.

126

The Necessary (Order) Condition for identification states that “For an equation to

be identified, the total number of variables excluded from it but included in the other

equations must be at least as great as the number of equations in the system less than

one” (Koutsoyiannis 1977). Let K be the total number of exogenous and endogenous

variables in the model. Suppose M is the total number of exogenous and endogenous

variables in a particular equation and G is the total number of endogenous variables.

Then the Order Condition is satisfied if (K-M) ≥ (G-1), where inequality denotes

overidentification. From equations 12 and 13, K=17, G=2, M=8 for equations 12and

M=9 for equation 13. Thus the Order Condition is satisfied for equations 12 and 13,

with both equations being overidentified.

We start off by estimating a naive regression which ignores the endogeneity of

social visibility. Table 4.5 presents the summary statistics of variables estimated in

this model. We estimate two equations, i.e. with and without shocks, using a random

effect generalized least squares estimation procedure and a panel household-level data,

with observations clustered within the household. The results are presented in Table

4.6.

First, overlooking the endogeneity of social invisibility, the results indicate that

wealthy households have lower mean log deviation BMI indices. Conversely socially

invisible households with high dependency ratios have higher intrahousehold

inequality. Controlling for current period shocks yields similar results; in addition

households with older heads have higher intrahousehold inequality. Suffering either a

farm or mortality shock is associated with an increase in mean log deviation BMI.

127

Table 4.5: Descriptive statistics: intrahousehold health inequality model

Variable Definition Frequency (%) or Mean

Log mean deviation BMI

Household log mean deviation using BMI as a welfare metric 0.02 (0.05)

Socially Invisible =1 if mean social visibility is less than or equal to 0.48750, 0 otherwise

11.54

Age Age of household head 47.24 (13.34)

Head’s wealth Household head’s non-land wealth in millions of Cedis 8.01 (17.9)

Spouse’s wealth Spouse’s non-land wealth in millions of Cedis 1.04 (3.02)

No education =1 if household head has no formal education, 0 otherwise

16.53

Farmer =1 if household head farmer, 0 otherwise

88.84

Dependency Ratio Ratio of number of individuals under age 18 and above age 64 to those between ages 18-64

1.24 (2.05)

Household size Number of individuals in household 9.14 (4.20)

Major clan =1 if household head belongs to a major clan, 0 otherwise

79.17

Parent’s office =1 if household head parents held village offices, 0 otherwise

73.72

Farm shock =1 if suffered farm shock in round 0 otherwise

39.10

Mortality shock =1 if suffered mortality shock 0 otherwise

62.50

Location Village 1 Village 2 Village 3 Village 4

=1 if Village cluster 1, 0 otherwise =1 if Village cluster 2, 0 otherwise =1 if Village cluster 3, 0 otherwise =1 if Village cluster 4, 0 otherwise

24.42 20.16 28.42 27.00

128

Table 4.6: Intrahousehold health inequality naïve regressions

Coefficients Standard

error Coefficients Standard

error Dependent Variable: Household’s log mean deviation Household characteristics Head: age 0.004 0.003 0.004* 0.002 Head: no education -0.014 0.010 -0.011 0.009 Head: farmer -0.011 0.009 -0.012 0.009 Head's Wealth -0.003*** 0.001 -0.003** 0.001 Spouse's Wealth -0.002 0.007 -0.003 0.007 Socially Invisible 0.033*** 0.012 0.023* 0.012 Shocks Farm shock -- -- 0.009* 0.005 Mortalitity shock -- -- 0.026*** 0.008 Location Village 2 0.018* 0.010 0.012 0.010 Village 3 0.027*** 0.010 0.041*** 0.011 Village 4 0.000 0.009 0.016 0.010 Other factors Dependency ratio 0.009* 0.002 0.008* 0.002 Household size -0.001 0.001 -0.001 0.001

n =238 n =238 Wald χ2(11) = 61.89 Wald χ2(13) = 79.61

p-value = 0.00 p-value = 0.00 Notes: ***, **, * Significant at the 1%, 5% and 10% levels, respectively.

Comparison group is a household with a head from village 1 who is not a farmer has some education and is socially visible.

129

Next, we estimate the reduced form model with parent’s office and major clan

as instruments for social visibility. The results presented in Table 4.7 indicate that

wealthy households with uneducated heads who belong to a major clan and have

parents who held village offices have lower mean log deviation BMI indices.

Households with older heads high and dependency ratios have higher intrahousehold

inequality. Controlling for current period shocks, yields similar results. In this case,

suffering a mortality shock increases intrahousehold inequality as well.

The first stage regression of the structural model presented in Table 4.8

indicates that wealthy households with younger heads who belong to a major clan and

have parents who held village offices are less likely to be in social invisible. This is

consistent with the results of Vanderpuye-Orgle and Barrett (2007). The second stage

regression presented in Table 4.9 shows that being socially invisible with a high

dependency ratio increases intrahousehold inequality. This is robust to the

specification with shocks. In addition, experiencing a mortality shock is associated

with an increase in mean log deviation BMI .

4.5 Conclusions

This paper assesses the effects of social visibility on the existence of Kuznets

curves and the determinants of intrahousehold inequality. The results indicate that

whilst the inverted U-shape may not exist when using aggregated data, it may exist for

a subsample of the population – in this case the socially visible. In addition the

composition of inequality varies by social visibility. Over 75% of overall inequality

may be attributed to the within-group component (i.e. intrahousehold inequality)

amongst socially visible households. On the other hand, a little over 50% of overall

inequality may be attributed to the between-group component amongst socially

invisible households.

130

Table 4.7: Intrahousehold health inequality reduced form model

Notes: ***, **, * Significant at the 1%, 5% and 10% levels, respectively. Comparison group is a household with a head from village 1 who is not a farmer has some education , does not from a major clan, has some education and has parents who did not hold village offices.

Coefficients Standard

error Coefficients Standard

error Dependent Variable: Household’s log mean deviation Household Characteristics Head: age 0.005* 0.003 0.005** 0.002 Head: no education -0.018* 0.009 -0.014 0.009 Head: farmer -0.006 0.009 -0.007 0.009 Head's Wealth -0.003** 0.002 -0.003** 0.002 Spouse's Wealth 0.000 0.001 0.000 0.001 Parent’s office -0.026*** 0.008 -0.027*** 0.008 Major clan -0.029** 0.012 -0.026** 0.012 Shocks Farm shock -- -- 0.006 0.005 Mortality shock -- -- 0.029*** 0.007 Location Village 2 0.020* 0.010 0.013 0.010 Village 3 0.023** 0.010 0.039*** 0.011 Village 4 0.002 0.010 0.019* 0.010 Other factors Dependency ratio 0.010*** 0.001 0.008*** 0.001 Household size -0.001 0.001 -0.001 0.001

n =238 n =238

Wald χ2(12) = 76.33 Wald χ2(14) = 98.07

p-value = 0.000 p-value = 0.000

131

Table 4.8: Intrahousehold health inequality first stage regression

Notes: ***, **, * Significant at the 1%, 5% and 10% levels, respectively. F-test of instruments=249.68, p-value = 0.00 Comparison group is a household with a head from village 1 who is not a farmer has some education, does not from a major clan, has some education and has parents who did not hold village offices.

Coefficient Standard error Dependent Variable: Socially Invisible Individual Characteristics Head: age -0.0030*** 0.001Head: no education 0.022 0.032Head: farmer 0.023 0.031Head's Wealth 0.008* 0.005Spouse's Wealth -0.001 0.002Location Village 2 0.004 0.034Village 3 -0.003 0.035Village 4 0.057* 0.048Instrumental Variables Parent’s office -0.106*** 0.029Major clan -0.482*** 0.035n =238 Wald χ2(10) = 363.21 p-value = 0.00

132

Table 4.9: Intrahousehold health inequality second stage regression

Coefficients Standard

error Coefficients Standard

error Dependent Variable: Household’s log mean deviation Individual characteristics Head: age 0.002 0.003 0.002 0.003Head: no education -0.011 0.009 -0.007 0.009Head: farmer -0.006 0.009 -0.007 0.009Head's Wealth -0.004 0.001 -0.004 0.001Spouse's Wealth 0.000 0.000 0.007 0.007Socially Invisible 0.271*** 0.063 0.263*** 0.064Shocks Farm shock -- -- 0.005 0.005Mortalitity shock -- -- 0.029*** 0.007Location Village 2 0.028* 0.011 0.021 0.010Village 3 0.016 0.010 0.032*** 0.011Village 4 -0.002 0.010 0.015 0.010Other factors Dependency ratio 0.010*** 0.001 0.008*** 0.001Household size -0.001 0.001 -0.001 0.001

n =238 n =238

Wald χ2(11) = 68.74 Wald χ2(13) = 85.63 p-value = 0.00 p-value = 0.00

Notes: ***, **, * Significant at the 1%, 5% and 10% levels, respectively. Comparison group is a household with a head from village 1 who is not a farmer has some education and is socially visible.

133

Patterns in relative BMI within the household suggest that the socially invisible

may overcompensate for risk exposure by protecting children in the allocation of

limited food and nutritional resources. With regards to factors determining the level of

intrahousehold inequality: wealth, dependency ratio, age, social visibility as well

experiencing shocks had statistically significant effects on intrahousehold inequality.

By way of policy implications, the challenges associated with identifying

individuals within the household for targeting purposes lead policy makers to resort to

the second best strategy – focusing on improving overall household well-being. The

existence of a systematic relationship between intrahousehold inequality and average

well-being on the basis of some identifiable household characteristic such as social

visibility should enable policymakers better tailor interventions to attain the desired

goal of improving individual well-being.

134

REFERENCES

Alderman, H., Chiappori, P.A., Haddad, L., Hoddinott, J. and Kanbur, R. (1995).

Unitary Versus Collective Models of the Household: Is it Time to Shift the Burden of Proof? World Bank Research Observer. 10: 1-9.

Alderman, H. and C. H. Paxson (1992). Do the Poor Insure? A Synthesis of the

Literature on Risk and Consumption in Developing Countries, The World Bank.

Anand, S. and Kanbur, R. (1993). The Kuznets Process and the Inequality-

Development Relationship, Journal of Development Economics, 40, 25-52. Beresteanu, A. and Manski, C. F. (2000). Bounds for STATA: Draft Version 1.0.

Northwestern University. Behrman, J. R. and Deolalikar, A. (1988). Health and Nutrition. In H. Chenery and

T.N. Srinivasan (Eds.) Handbook of Development Economics, 1. Amsterdam: North-Holland Press, 631-711.

De Weerdt, J. (2005). Risk-Sharing and Endogenous Network Formation. In S.

Dercon (Ed.) Insurance against Poverty. New York: Oxford University Press, 197-216.

Deaton, A. (1992). Household Saving in LDCs: Credit Markets, Insurance and

Welfare, Scandinavian Journal of Economics, 94, 253-73. Dercon, S. and Hoddinott, J. (2003). Health, Shocks and Poverty Persistence. In S.

Dercon (Ed.) Insurance against Poverty. New York: Oxford University Press. Dercon, S. and Krishnan, P. (2000). In Sickness and in Health: Risk Sharing within

Households in Ethiopia, Journal of Political Economy, 108, 688-727. Doss, C. R. (1996). Intrahousehold Resource Allocation in an Uncertain Environment.

American Journal of Agricultural Economics, 78, 1335-1339.

135

Duflo, E. and Udry, C. (2004). Intrahousehold Resource Allocation in Cote d'Ivoire: Social Norms, Separate Accounts and Consumption Choices.

Ellis, F. (1998). Household Strategies and Rural Livelihood Diversification. Journal of

Development Studies, 35(1), 1-38. Gertler, P. and Gruber, J. (1997). Insuring Consumption against Illness, National

Bureau of Economic Research Working Paper Series. Goldstein, M. (1999). Chop Time, No Friends: Intrahousehold and Individual

Insurance Mechanisms in Southern Ghana. Yale University. Goldstein, M., De Janvry, A. and Sadoulet, E. (2005). Is a Friend in Need a Friend

Indeed? Inclusion and Exclusion in Mutual Insurance Networks in Southern Ghana. In S. Dercon (Ed.) Insurance against Poverty. New York: Oxford University Press.

Goldstein, M. and Udry, C. (1999). Agricultural Innovation and Resource

Management in Ghana, Washington D.C.: International Food Policy Research Institute.

Grantham-McGregor, S., Walker, C., Chang, S. and Powell, C. (1997). Effects of

Early Childhood Supplementation with and without Stimulation on Later Development in Stunted Jamaican Children, American Journal of Clinical Nutrition, 66, 247-53.

Haddad, L. and Kanbur, R. (1990). How Serious Is the Neglect of Intrahousehold

Inequality?, Economic Journal, 11, 866-81. Haddad, L., Kanbur, R. and Bouis, H. (1993). Intrahousehold Inequality at Different

Welfare Levels: Energy Intake and Energy Expenditure Data from the Philippines. Oxford Bulletin of Economic and Statistics, 57 (3), 389-409.

Hardle, W. (1990). Applied Nonparametric Regression. Cambridge: Cambridge

University Press.

136

Hoddinott, J. and Kinsey, B. (1999). Adult Health in Time of Drought, Washington

D.C.: International Food and Policy Research Institute. Hoddinott, J. and Kinsey, B. (2001). Child Growth in Time of Drought, Oxford

Bulletin of Economics and Statistics, 64, 409-36. Kanbur, R. and Haddad, L.(1992). Is There an Intrahousehold Kuznets Curve? Some

Evidence from the Philippines, Public Finance, 47, 77-93. Kuznets, S. (1955). Economic Growth and Income Inequality. American Economic

Review, 45, 1-28.

Loh, W. (2002). Regression Trees with Unbiased Variable Selection and Interaction Detection. Statistica Sinica, 12, 361-386

Sahn, D. E. and Younger, S. (2007). Measuring Intra-Household Health Inequality:

Explorations using the Body Mass Index. Cornell University. Santos, P. and Barrett, C. (2004). Interest and Identity in Network Formation: Who Do

Smallholders Seek out for Information in Rural Ghana?, Cornell University. Silverman, B. (1986). Density Estimation for Statistics and Data Analysis, London:

Chapman & Hall. Strauss, J. and Thomas, D. (1998). Health, Nutrition, and Economic Development,

Journal of Economic Literature, 36 (2) 766-817. Townsend, R. M. (1994). Risk and Insurance in Village India, Econometrica, 62, 539-

91. Udry, C. and Conley, T. (2005). Social Networks in Ghana, in C. Barrett (Ed.)The

Social Economics of Poverty. London: Routeledge. Vanderpuye-Orgle, J. and Barrett, C. (2007). Risk Management and Social Visibility

in Ghana. Cornell University.

139

APPENDIX A

Field Notes

A.1 Field Activities

A study of credit transactions in response to inter-temporal risk requires the

use of panel data. We build on the data set generated by Christopher Udry47 and

Markus Goldstein48 in the Eastern Region of Ghana, from November 1996 to October

1998, to create a three-period panel. The research was conducted in the Akuapim

South District (specifically the Nsawam - Aburi area) in the Eastern Region of Ghana.

This dynamic agricultural region lying in the forest savanna transition zone of Ghana

is particularly interesting because of the evolving farming systems. Since the early

1990s farmers in this area have been switching from the cultivation of maize-cassava

intercrop for domestic production to pineapple cultivation for export. This involves

not only a significant amount of technological innovation and investment but also risk-

taking.

The sample was generated using a two-stage sampling procedure. First, four

village clusters were purposively selected within this area based on their participation

in fruit and vegetable production as well as their array of agronomic, market access

and geographic conditions. Sixty married couples (or triples) were then randomly

selected in each village cluster, except for the smallest village cluster where all

resident couples were interviewed.

The field research was undertaken in two phases: (i) a preliminary trip to the

survey area to establish protocol for another round of surveys with the same sample of

47 Professor, Department of Economics, Yale University. 48 Lecturer, Department of Development Studies, London School of Economics.

140

respondents; assess the level of attrition and undertake a quasi-qualitative study using

the community studies approach to gather information towards formulating/revising

survey instruments, and (ii) an extended trip for data-collection.

A.2 Preliminary Visit to Survey Area

Duration: 12/16/2003 - 01/22/2004

Upon arrival in Ghana, I set up meetings with Professor Ernest Aryeetey, the

Director of the Institute of Statistical, Social and Economic Research (ISSER) and Mr.

Ernest Appiah, a research fellow at ISSER who led the data collection team in the first

two panels. My first meeting was with Mr. Appiah, during which we discussed the

requirements for the project: customary token gifts for the chiefs (bottles of imported

Schnapps), contact persons, transportation, lodging, the research schedule,

enumerators, etc. Two factors weighed on the commencement of the field trips: (i)

establishing contact with Bob, a seasoned enumerator from the previous rounds who

would introduce me to the chiefs and people in the respective villages and (ii) the

break in formal activities due to the Yuletide season. We thus set the date for

beginning the actual field trips for Monday, 01/05/04.

In the mean time, I met with Professor Aryeetey to discuss my appointment as

a Research Associate at ISSER. Upon my return to Ghana, I would be provided with

office space at ISSER as well as three national service personnel who I could employ

as part of

my data collection team. I would also be expected to participate in the weekly seminar

series.

Starting on 01/05/04, Bob and I visited the villages in the previous sample,

namely Pokrom, Konkonuru, Oboadaka and Ashweriase - Daman. As custom

demands, we presented drinks to the chiefs and elders (or those acting in their

141

absence) in the respective villages and explained our intention of undertaking survey

in 06/04. We received a hearty welcome in each of the villages as they fondly

recollected personal experiences from the previous rounds.

We also talked to some of the people in the villages. This gave me some

insights into some changes in the sources of credit, among other things. For instance,

most of the villages now had farming cooperatives. However, the loans disbursed were

meager and could not be used for meaningful investments. The farmers talked

extensively about the types of risk they faced and how they tried to deal with it. A

discussion with a staff of Malinas Ghana Limited, a pineapple export firm which uses

farmers in Konkoronu and Oboadaka as outgrowers, revealed that they paid 30% of

expected sales prior to harvest if the farmer "forced" the plants49. This changes the

extent of risk borne by the farmers. In addition, recent developments in infrastructure,

by way of feeder roads, coupled with the aforementioned cooperatives may

significantly change the dynamics of social networks and access.

Information gathered from these interactions served as a basis for the

development of a survey questionnaire, which I later revised - rephrasing some of the

questions to be consistent with the previous rounds. I set in motion plans to pretest the

survey instrument using 20 households in Konkonuru. However, I was infected with a

serious case of malaria at the beginning of my fourth week. This slowed down the

pace of events. In the end, the ailment, along with a last minute hitch in logistics and

the fact that most of the questions in the survey instrument had been adapted from the

previous surveys and have thus been amply tested, resulted in a decision to postpone

the survey pretest till 06/04 and/ or adopt the grounded theory approach used by Udry

and Goldstein

49 That is, if the plants were induced to flower prematurely; thus reducing the time from sowing to harvest.

142

in the previous rounds.

On the whole, this preliminary trip was invaluable. It afforded me a better

understanding of the issues relating to risk management and social insurance in the

respective villages. It also made it possible for me to put in place logistics required for

the actual data collection thus paving the way for successful doctoral dissertation

research.

A.3 Data Collection trip

Duration: 07/02/2004 - 01/29/2005

I arrived in Ghana on Saturday, 07/03/2004. In the first two weeks, I met with

Professor Ernest Aryeetey to go over the field research schedule, the personnel

requirements and equipment to be used. I was assigned an office at ISSER and met

with the three national service personnel designated to work with me on the project.

After various discussions about time availability, readiness to relocate and language

issues only one of them was selected. We also retained Bob on the project as the lead

enumerator. I started a series of meetings with Mr. Ernest Appiah to recruit and

interview additional research assistants. I also met with the director of the Center for

Remote Sensing and Geographic Information Services (CERSGIS), University of

Ghana, Legon, Dr Amamoo Otchere, to explore the possibility of hiring personnel

from that office to map the respondents' farming plots. The latter proved to be quite

expensive so it was aborted in favor of hiring and training a research assistant from the

Geography department to map the plots and to administer the questionnaires relating

to the acquisition and use of plots as well.

I met with Markus Goldstein and Ernest Appiah on 07/16/04 to discuss

questionnaires and logistics. I had another meeting with Markus Goldstein and Chris

Udry on 07/19/04 to finalize the questionnaires, research schedule and logistics. It was

143

decided that due to budgetary constraints and concerns for respondent fatigue, I would

increase the survey interval to 8 weeks: administering the entire set of modules at a

sitting instead of staggering the questionnaires over a 4-weeek interval schedule. We

also reduced the number of enumerators from eight to four, with two of them being

assigned to two villages, respectively, and moving back and forth from one village to

the other every four weeks. As with the previous rounds, male enumerators were

assigned to male respondents and female enumerators to female respondents,

respectively, to preserve gender sensitivity and cultural norms. On 07/20/04, I

presented a paper at the ISSER and Cornell University International Conference,

Ghana at Half Century (held from 07/18/04- 07/20/04).

We hired two additional enumerators. In the meantime, we went to the villages

to present another round of drinks to the chiefs, informing them of my safe return and

a tentative date for the start of the project as well as to ask for help in securing

accommodation in the area. We were asked to pay fees to the respective gong-gong

beaters (town announcers) to make announcements and create awareness for the

project so we won't be viewed as strangers in the town. The process of securing

accommodation required several trips during which we identified our local ‘resource

persons’, introduced our enumerators to the respondents and identified the logistics on

the ground vis-à-vis lodging needs (such as bedding, food and water), transportation

and coordination. We initially rented two places in each of the villages: one for the

male enumerator and one for the female enumerator -- I would share with the latter

when the need arose to stay for extended periods of time. I was to be based in Accra

and would commute to the villages daily, supplying water and stationary and

questionnaires when needed. We later rented a third place for the plot mapping

personnel.

144

We held an all-day training session on 08/02/04 for the research assistants to

go over the wording of the questionnaires in twi (the local dialect in the survey area);

complying with Cornell’s University Committee on Human Subjects (UCHS)

regulations such as getting verbal consent from the respondents; as well as

comportment and dealing with sensitive questions. We also had mock interviews with

some role playing to clarify concepts. We moved to the villages on 08/08/04 and

started administering the questionnaires the next day. As mentioned earlier, there was

a male and female enumerator in each of two villages, at any point in time. Each

enumerator was assigned to two villages (either Ashweriase-Daman and Pokrom or

Konkonuru and Oboadaka). We started out in Ashweriase-Daman and Konkonuru (i.e.

1a in the Appendix table) and then moved to Pokrom and Oboadaka (i.e. 1b in the

Appendix table), the cycle was repeated two more times. There were a total of six

rounds; each respondent was interviewed three times.

In all, 436 individuals were surveyed fifteen times in the first and second

panels. Of which, 332 individuals were surveyed three times in the third panel. The

sample attrition rate over the 1998-2004 interval was 23.85%. Several modules were

administered. A brief description of each module and the survey schedule are given in

the appendix. I returned to the United States on 01/29/05. I went back to Ghana for a

workshop organized by the Consortium of Development Partnerships and ISSER on

04/14/05-04/15/05 (viz., CDP Workshop on Entrepreneurship and Poverty Alleviation

in Agribusiness) to give a presentation on Data Availability and Research Issues

relating to the study of pineapple as a crop system in Ghana.

Field work was quite an experience — a good one, on the whole. It was

challenging, though, for the most part. We had to deal with issues of respondent

fatigue; a peak tomato farming season in one of the villages and hence very busy

respondents; constant demands for 'big' gifts and occasionally disgruntled respondents;

145

as well as illnesses (we had on average two weeks of slowed activity for all four

enumerators at different points in time): one enumerator was eventually given a whole

week off by the doctor and the plot mapping personnel was also admitted in hospital

for a while. The list goes on. However, it was a excellent research experience and I'll

do it all over again, given the chance.

146

APPENDIX B

Survey Modules and Interview Schedule

Household Roster and Anthropometrics - This data was collected at the start of the

survey and again in the final round. A short update covering additions to and

departures from the households since the last survey in 1998 was administered in the

first round. Whether or not the household was still in the same community and

reasons for moving, if they had, was also documented. This information was updated

in the final round. Data on age, gender, education, occupation, migration and marital

status were collected on all children ever borne to each wife, as well as on all current

household members. Anthropometric data (i.e. weight and height) of the respondents

was also collected.

Birth History – This adopts questions from the Demographic and Health Surveys. It

asks detailed questions on all children ever born to the respondents in the last six years

including birth weights and heights, antenatal care, who assisted in delivery and where

delivery occurred. This questionnaire, in addition to the household roster, would allow

for the use of a concept of poverty which goes beyond the unidimensional

income/expenditure measure to capture a fuller multidimensional notion.

Assets and Trading Stocks - This gathered data on stocks of food, farm inputs,

livestock, financial assets and participation in ROSCAs. Information on trading stocks

and business assets was also collected.

Credit Practices - This refers respondents to lenders and borrowers identified in 1998

and asks if there are still transacting with them and if not why. It also asks if they

147

would lend/ borrow to individuals in three specified locations, each farther apart from

the present location. It is aimed at capturing penalties for default as well as the spatial

gradient in credit activities.

Lending, Lending Continuation, Borrowing, and Borrowing Continuation – These

questionnaires are based on tracking each loan from the point at which it is extended

until it is either repaid or written-off. Data is collected on both contractual terms and

actual payments, and on the use of any enforcement mechanisms.

Non-Farm Income - Non-farm income provides about half of the income of the sample

respondents. Costs and revenues of the respondents’ non-farm enterprises are

collected. In addition, this questionnaire records income from employment, from

farms outside of the study area (many own, or have some rights over, land in other

areas), and from miscellaneous sources such as inheritances or pensions. Respondents

are also asked to estimate their spouses’ income from non-farm sources which,

combined with a similar question on the sale of farm output, will allow us to see what

spouses know about each others income.

Gifts - Requests data on gifts received or given from individuals other than the spouse

as well as gifts received by or given to the spouse.

Expenditure - The purchased food, food from own farm, and family expenses

questionnaires constitute the expenditure survey. This requests standard data on

expenditures, but at the level of the individual rather than the household. In addition,

respondents are asked about the expenditures of their spouse(s), and about the transfer

of “chop money” - the main means of spouse to spouse transfer.

148

Family Background - This provides information on the wealth of the respondents’

families, the respondents’ parents’ education and occupation, migration histories, the

origins of the family, inheritances as well village offices held and the privileges

thereof.

Learning – This questionnaire is designed based on matching the respondent with a

random sample of six other respondents, and with two other village-specific “focal”

individuals identified from community-studies approach taken in the preliminary field

trip as individuals in the villages from whom advice is commonly sought. Each

respondent is asked about his/her relationship with each of these individuals, and

whether or not he/she could approach the individual to deal with one of a set of

specific issues.

Shocks - An account of significant unexpected events which occurred on the plots of

the respondent and his/her spouse. It also asked about illness and other unexpected

events, including death of livestock and theft.

Plot Mapping - All of the plots were mapped using GPS equipment and GIS software.

This procedure yields much more accurate measures of plot size than are available in

most surveys in developing countries. Moreover, by mapping the relative locations of

all plots (and associated roads, paths and villages) it may be possible to distinguish

between information transmission that is associated with social connections from that

associated with geographical proximity. The mapping procedure opens the opportunity

to investigate the importance of unobserved (and commonly unconsidered) spatial

autocorrelation in production shocks. Finally, by accurately determining the location

of each plot, it will be possible to match the agronomic and fertility data to medium-

149

term historical information available from satellite photography and also to collect

follow-up information indefinitely in the future. In association with the mapping, the

Plot questionnaire was administered. This questionnaire is concerned with the history

of each plot. The questionnaire also requests information on the current contractual

status of the land, focusing in particular on the details of sharecropping contracts,

which are very frequently used in the study area.

Plot Wrap-Up - This and other administrative questionnaires are used to keep track of

new plots as well as dropped and fallowed plots. This final questionnaire requests data

on the value of crops still standing as the survey ends and the property rights the

spouse and children of the respondent would have if the respondent were to die.

1

Schedule of Surveys

2004 - 2005

Round50

1a 1b 2a 2b 3a 3b

date started 9-Aug 6-Sept 4-Oct 1-Nov 29-Nov 27-Dec

Questionnaire

Household roster

Anthropometrics

Credit practices

Lending

Lending continuation

Borrowing

Borrowing continuation

Gifts and marital transfers

Events / Shocks

50 Grayed out cells indicate periods over which the questionnaires were administered.

150

2

Schedule of Surveys Continued

Assets and Trading stocks

Family Background

Learning

Expenditure

Other income

Plot questionnaire and Plot

mapping

Plot wrap up

Birth history 151

APPENDIX C

Questionnaires

152

Risk Management and Social Insurance Study

Section 1: Household Roster

VILLAGE HHN RESPONDENT RESPOND. # DATE ROUND

IS THIS HOUSEHOLD IN THE SAME COMMUNITY AS 1998? (Circle one) Yes……………………………………1 No…………………………………….2 IF NO, WHERE DID THEY MOVE TO? (See location codes on page 1)

Write any reasons given for relocation here:

Good (morning/ afternoon/ evening). I’m and we are conducting a survey of households in this area. From 1996-1998 we interviewed people living in this compound. This earlier study has been very helpful in learning about people’s living conditions and in planning government policies. As a follow- up to this earlier work we would again like to ask questions about the people who lived here in 1996-1998. The information you give us will be kept confidential. You and your household members will not be identified by name in any of the documents we plan to write.

HOUSEHOLD ROSTER PAGE 1

HOUSEHOLD HEAD, FIRST SPOUSE ALL OF HER CHILDREN EVER BORN (INLUDING CHILDREN LIVING ELSEWHERE) AS LISTED IN 1996-98 (Complete Questions for all Pre-printed Persons)

Schooling Name Sex Is […] alive? Yes…1 No…WRITE YEAR OF DEATH THEN GO TO NEXT PERSON

Approximate Current Age in Years in 2004

HH member? Yes…..1 No…...2

If no, where does s/he stay?

Highest level attended

Highest grade in that level

Occup-ation

Marital Status

Weight

Height

Page and ID of spouse

Months Away in Last 12

0 Head

1 Spouse

2 Child

3 Child

4 Child

Marital Status Location 8 Eastern region town 1 Primary Occupation 8 Non- agric labor 1 Married 1Daman/Ahweriase 9 Accra 2 Middle/ JSS 1 Farming 9 Trading 2 Consensual union 2 Pokrom/Nsabaa 10 Other Urban 3 SSS/ Tech/ Comm/ 2 Teaching 10 Student/pupil 3 Divorced 3 Oboadaka/ Kwamekrom 11 Other Rural training college 3 Artisan 11 Unemployed 4 Separated 4 Konkonuru 12 Not in Ghana 4 Post sec/ Nursing/ Poly 4 Office Worker 12 Not in the labor force 5 Widow 5 Aburi 5 Higher 5 Civil Servant 6 Never Married 6 Nsawam Level of School 6 Agric labor 7 NA (under 15 yrs) 7 Eastern region rural 0 None 7 Health worker

5 Child

6 Child

7 Child

8 Child

9 Child

10 Child

11 Child

HOUSEHOLD ROSTER PAGE 2 ADDITIONAL SPOUSE ALL OF HER CHILDREN EVER BORN (INLUDING CHILDREN LIVING ELSEWHERE) AS LISTED IN 1996-98 (Complete Questions for all Pre-printed Persons)

Schooling Wife #2

Name Sex Is […] alive? Yes…1 No…WRITE YEAR OF DEATH THEN GO TO NEXT PERSON

Approximate Current Age in Years in 2004

HH member?

If no, where does s/he stay?

Highest level attended

Highest grade in that level

Occup-ation

Marital Status

Weight

Height

Page and ID of spouse

Months Away in Last 12

12 Spouse

Marital Status Location 8 Eastern region town 1 Primary Occupation 8 Non- agric labor 1 Married 1Daman/Ahweriase 9 Accra 2 Middle/ JSS 1 Farming 9 Trading

13 Child

14 Child

15 Child

16 Child

17 Child

18 Child

19 Child

20 Child

21 Child

2 Consensual union 2 Pokrom/Nsabaa 10 Other Urban 3 SSS/ Tech/ Comm/ 2 Teaching 10 Student/pupil 3 Divorced 3 Oboadaka/ Kwamekrom 11 Other Rural training college 3 Artisan 11 Unemployed 4 Separated 4 Konkonuru 12 Not in Ghana 4 Post sec/ Nursing/ Poly 4 Office Worker 12 Not in the labor force 5 Widow 5 Aburi 5 Higher 5 Civil Servant 6 Never Married 6 Nsawam Level of School 6 Agric labor 7 NA (under 15 yrs) 7 Eastern region rura l 0 None 7 Health worker

22 Child

23 Child

24 Child

25 Child

HOUSEHOLD ROSTER PAGE 3 ALL OTHER HOUSEHOLD MEMBERS AS LISTED IN 1996-98

(Complete Questions for all Pre-printed Persons)

Please specify Other family member / Other (id code):_________________________________________________________________________________________

Relationship 12. Wofa 4 Separated Occupation 10 Student/pupil 1 Head 13 Uncle 5 Widow 1 Farming 11 Unemployed 2 Wife 14 Aunt 6 Never Married 2 Teaching 12 Not in the labor force 3 Child 15 Other family member 7 NA (under 15 yrs) 3 Artisan 4 Grandson 16 Househelp Level of School 4 Office Worker 5 Grand daughter 17 Permanent worker 0 None 5 Civil Servant 6 Mother 18 Friend 1 Primary 6 Agric labor 7 Father 19 Other 2 Middle/ JSS 7 Health worker 8 Brother Marital Status 3 SSS/ Tech/ Comm/ 8 Non- agric labor 9 Sister 1 Married training college 9 Trading 10 Nephew 2 Consensual union 4 Post sec/ Nursing/ Poly 11 Niece 3 Divorced 5 Higher

Schooling ID code

Name Sex Is […] alive? Yes…1 No…WRITE YEAR OF DEATH THEN GO TO NEXT PERSON

ApproximateCurrent Age in Years in 2004

Relation -ship to head

Relation -ship to

head’s spouse

HH member?

If no, where does s/he stay?

Highest level attended

Highest grade in that level

Occup-ation

Marital Status

Weight

Height

Page and ID of spouse

Months Away in Last 12

26

27

28

29

30

31

32

33

34

35

HOUSEHOLD ROSTER PAGE 4 NEW HOUSEHOLD HEAD, FIRST SPOUSE AND ALL OF HER CHILDREN EVER BORN (INCLUDING CHILDREN LIVING ELSEWHERE)

Schooling

ID

Name

Sex

Age

Child of head?

HH member?

If no, where does s/he stay?

Highest level attended

Highest grade in that level

Occup ation

Marital Status

Weight

Height

Page and ID of spouse

Months Away in Last 12

Correction by head

02 Head

36 Spouse

37 Child

38 Child

39 Child

40 Child

41 Child

42 Child

43 Child

44 Child

45 Child

46 Child

47 Child

Marital Status1 Married 2 Consensual union 3 Divorced 4 Separated 5 Widow 6 Never Married 7 NA (under 15 yrs) Location 1Daman/Ahweriase

2 Pokrom/Nsabaa 3 Oboadaka/Kwamekrom 4 Konkonuru 5 Aburi 6 Nsawam 7 Eastern region rural 8 Eastern region town 9 Accra 10 Other Urban 11 Other Rural 12 Not in Ghana

Level of School0 None 1 Primary 2 Middle/JSS 3 SSS/Tech/Comm/ training college 4 Post sec/ Nursing/Poly 5 Higher Occupation1 Farming 2 Teaching

3 Artisan 4 Office worker 5 Civil Servant 6 Agric labor 7 Health worker 8 Non-agric labor 9 Trading 10 Student/pupil 11 Unemployed 12 Not in labor force

HOUSEHOLD ROSTER PAGE 5

NEW ADDITIONAL SPOUSE AND ALL OF HER CHILDREN EVER BORN (INCLUDING CHILDREN LIVING ELSEWHERE) Schooling

Name

Sex

Age

Child of head?

HH member?

If no, where does s/he stay?

Highest level attended

Highest grade in that level

Occu pation

Marital Status

Weight

Height

Page and ID of spouse

Months Away in Last 12

Correction by head

48 Spouse

49 Child

50 Child

51 Child

52 Child

53 Child

54 Child

55 Child

56 Child

57 Child

58 Child

59 Child

60 Child

61 Child

Marital Status1 Married 2 Consensual union 3 Divorced 4 Separated 5 Widow 6 Never Married 7 NA (under 15 yrs) Location 1Daman/Ahweriase

2 Pokrom/Nsabaa 3 Oboadaka/ Kwamekrom 4 Konkonuru 5 Aburi 6 Nsawam 7 Eastern region rural 8 Eastern region town 9 Accra 10 Other Urban 11 Other Rural

12 Not in Ghana Level of School0 None 1 Primary 2 Middle/JSS 3 SSS/Tech/Comm/ training college 4 Post sec/ Nursing/Poly 5 Higher Occupation

1 Farming 2 Teaching 3 Artisan 4 Office worker 5 Civil Servant 6 Agric labor 7 Health worker 8 Non-agric labor 9 Trading 10 Student/pupil 11 Unemployed

12 Not in Labor Force

HOUSEHOLD ROSTER PAGE 6 ALL OTHER NEW HOUSEHOLD MEMBERS

Schooling

ID code

Name

Sex

Age

Relation-ship to head

Relat-ionship to head’s spouse

Years in house-hold

Highest level attended

Highest grade in that level

Occu- pation

Marital Status

Weight

Height

Page and ID of spouse

Months Away in Last 12

Correction by head

62

63

64

65

66

67

68

69

70

71

72

Please specify Other family member / Other (id code):_________________________________________________________________________________________ Relationship 12. Wofa 4 Separated Occupation 10 Student/pupil 1 Head 13 Uncle 5 Widow 1 Farming 11 Unemployed 2 Wife 14 Aunt 6 Never Married 2 Teaching 12 Not in the labor force 3 Child 15 Other family member 7 NA (under 15 yrs) 3 Artisan 4 Grandson 16 Househelp Level of School 4 Office Worker 5 Grand daughter 17 Permanent worker 0 None 5 Civil Servant 6 Mother 18 Friend 1 Primary 6 Agric labor 7 Father 19 Other 2 Middle/ JSS 7 Health worker 8 Brother Marital Status 3 SSS/ Tech/ Comm/ 8 Non- agric labor 9 Sister 1 Married training college 9 Trading 10 Nephew 2 Consensual union 4 Post sec/ Nursing/ Poly 11 Niece 3 Divorced 5 Higher

PAGE 7

Section 2a: Credit Practices

1. Have you ever refused to give someone a loan? Y / N If yes, why___________________________________ 2. Would you lend to someone from the next village? Y / N If no, why ________________________________ 3. Would you lend to someone from Aburi/ Nsawam? Y / N If no, why ________________________________ 4. Would you lend to someone from Accra? Y / N If no, why ________________________________________ 5. Do you have a bank account? Y / N If no, why ____________________________________________ 7. Would you borrow from a bank? Y / N If no, why ____________________________________________

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE

ROUND

In the last survey you gave loans to [….] Please fill in from 1998

Do you still give [….] loans?

Y / N

If NO, WHY………

1.

2.

3.

4.

5.

6.

7.

In the last survey you got loans from [….] Please fill in from 1998

Do you still get loans from [….] ? Y / N

If NO, WHY………

1.

2.

3.

4.

5.

6.

PAGE 8

Section 2b: Lending

(One form to be filled for EACH loan)

Ask question 1 ONLY ONCE, for each respondent 1. In the last two months, have you given anyone any loans that have NOT YET BEEN REPAID? This could include

money, food, or goods, but not land. Y / N [If no, END] 2. What did you lend? Cash (enter amount) ____________________________________________________

Goods (enter items) ____________________________ Value __________________________________ 3. When did you make the loan (date/ how many days ago)? ___________________________________ 4. Who did you make the loan to? Name (the name is optional/use IRID)______________________________

residence_________________ relationship _________________________________________________ 5. Was there a witness to the loan? Y / N If yes, relationship to witness (record IRID)____________________ 6. What will he use the loan for? Purpose _____________________________________________________ 7. When do you expect the loan to be repaid? __________________________________________________ 8. How much will he repay? Cash____________ or Interest rate___________________________________ or Goods (enter items) ____________________________ Value _____________________________ 9. Did the borrower provide any collateral? Y/N If Yes, write details here:

Approximate value of collateral______________________________________________________________ Who is using the collateral now? Lender or Borrower ___________________________________

Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11Brother 12 Sister 13Wofa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin 18 Local Friend

19 Non-resident friend 20 Extension Agent 21 Market 22 Cooperative 23 NGO 24 Exporter 25 Village head 26 Lineage head 27 Priest 28 Trader Residence 1 Daman/Ahweriase 2 Pokrom/Nsabaa 3 Oboadaka/ Kwamekrom 4 Konkonuru 5 Aburi 6 Nsawam

7 Eastern region rural 8 Eastern region town 9 Accra 10 Other Urban 11 Other Rural 12 Not in Ghana Purpose 1 Farming 2 Medical 3 Consumption 4 Trading 5 Other business 6 Travel 7 Funeral 8 Other Ceremony 9 School fees

VILLAGE

HHN

RESPONDENT RESPOND.

# DATE

ROUND

PAGE 9

Section 2c: Lending Continuation

(One form to be filled for EACH outstanding loan) Bring out the loan form completed in earlier round. Remind the respondent of the loan in question

Name of Lender

Cash / Goods Enter amount/ items

Value

Ask questions 1-4 of all respondents 1. Since our last interview, did the […..] give you any repayments, or any gifts in thanks for this loan? Y / N

If the answer to 1 is No, ask questions 6 and 7 2. If Yes, what did he give you? Cash Goods (enter items) ____________________________ Value ______________________________ Date/ how many days ago_______________________________________ 3. Since our last interview, did you give him any further installments of money or goods that are part of this loan? Y / N If

yes, what did you give him? Cash Goods (enter items) ____________________________ Value

Date/ how many days ago_______________________________________

4. Do you expect to receive anything from the borrower in the future on account of this loan? Y / N If yes, what do you expect to receive? When do you expect to receive it?

5. Do you consider this loan to be fully repaid? Y / N If no, have you taken any action to attempt to be repaid? Describe the action:

6. Would you lend to this borrower again, if he requests a loan? Y / N 7. Has the borrower experienced any unexpected problems on his farm or in his family since you gave him this loan?Y/N

If yes, what was the problem and when did it occur?

Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11Brother 12 Sister 13Wofa

14 Uncle 15 Aunt 16 Co-wife 17 Cousin 18 Local Friend 19 Non-resident friend 20 Extension Agent 21 Market 22 Cooperative 23 NGO 24 Exporter 25 Village head 26 Lineage head 27 Priest

28 Trader Residence 1 Daman/Ahweriase 2 Pokrom/Nsabaa 3 Oboadaka/ Kwamekrom 4 Konkonuru 5 Aburi 6 Nsawam 7 Eastern region rural 8 Eastern region town 9 Accra 10 Other Urban 11 Other Rural

12 Not in Ghana Purpose 1 Farming 2 Medical 3 Consumption 4 Trading 5 Other business 6 Travel 7 Funeral 8 Other Ceremony 9 School fees

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE

ROUND

PAGE 10

Section 3a: Borrowing

(One form to be filled for each loan) Ask questions 1 only once, for each respondent 1. In the last two months, have you received any loans that you have NOT YET REPAID? This could include money,

food, or goods, but not land. Y/N [If no, END] 2. What did you borrow? Cash (enter amount)________________________________________________________

Goods (enter items) ____________________________ Value _______________________________________ 3. When did you get the loan (date/ how many days ago)?____________________________________________ 4. Who did you borrow from? Name(the name is optional/use IRID) _______________________________________

Residence_________________ Relationship ________________________________________________________ 5. Was there a witness to the loan? Y/N If yes, relationship to witness(record IRID)___________________________ 6. What will you use the loan for? Purpose ___________________________________________________________ 7. When do you expect to repay the loan(date/ how many days ago)?__________________________________ 8. How much will you repay? Cash_____________ or Interest rate _______________________________________ or Goods (enter items) ____________________________ Value_____________________________________ 9. Did you provide any collateral? Y / N If Yes, write details here:

Approximate value of collateral______________________________________________________________ Who is using the collateral now? Lender or Borrower

Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11Brother 12 Sister 13WBfa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin

18 Local Friend 19 Non-resident friend 20 Extension Agent 21 Market 22 Cooperative 23 NGO 24 Exporter 25 Village head 26 Lineage head 27 Priest 28 Trader Residence 1 Daman/Ahweriase 2 Pokrom/Nsabaa 3 Oboadaka/ Kwamekrom 4 Konkonuru 5 Aburi

6 Nsawam 7 Eastern region rural 8 Eastern region town 9 Accra 10 Other Urban 11 Other Rural 12 Not in Ghana Purpose 1 Farming 2 Medical 3 Consumption 4 Trading 5 Other business 6 Travel 7 Funeral 8 Other Ceremony 9 School fees

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

PAGE 11

Section 3b: Borrowing Continuation

(One form to be filled for each outstanding loan) Bring out the loan form completed in earlier round. Remind the respondent of the loan in question

Name of Borrower

Cash / Goods Enter amount/ items

Value

Ask questions 1-5 on all respondents 3. Since our last interview, did you give the lender any repayments, or any gifts in thanks for the loan? Y / N

If yes, what did you give him? Cash____________________________________________________ Goods (enter items) ____________________________ Value_________________________________

Date/ how many days ago_______________________________________

If the answer to 3 is no, ask question 6 4. Since our last interview did he give you any further installments of money/goods that are part of this loan? Y / N

If yes, what did he give you? Cash _________________________________________________________ Goods (enter items) ___________________________ Value_____________________________________

Date/ how many days ago_______________________________________ 5. Do you expect to give any further repayment of this loan to the lender? Y / N

If yes, what do you expect to give? _________________________________________________________ When do you expect to give it? ____________________________________________________________

6. Do you consider the loan fully repaid? Y/ N

If No: Has the lender taken any action to attempt to be repaid? Describe the action: 7. Has the lender asked for repayment? Y/N

Why (ask for both Y and N. If answer is needs the money, probe why) 8. Would you approach this lender to borrow again, if you needed a loan? Y / N Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11Brother 12 Sister 13Wbfa 14 Uncle 15 Aunt

16 Co-wife 17 Cousin 18 Local Friend 19 Non-resident friend 20 Extension Agent 21 Market 22 Cooperative 23 NGO 24 Exporter 25 Village head 26 Lineage head 27 Priest 28 Trader Residence 1 Daman/Ahweriase 2 Pokrom/Nsabaa

3 Oboadaka/ Kwamekrom 4 Konkonuru 5 Aburi 6 Nsawam 7 Eastern region rural 8 Eastern region town 9 Accra 10 Other Urban 11 Other Rural 12 Not in Ghana Action 1 Requested payment 2 Discussed the issue with your relatives 3 Discussed the issue with (relationship code) 4 Kept collateral 5 Taken collateral

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

PAGE 12 _________________________________________________________________________________________________

Section 4: Gifts and Transfers _________________________________________________________________________________________________

1. How much chop money did your spouse give directly to you for this past two months? ____________ 2. How much chop money did you give directly to your spouse for this past two months? ____________ 3. How much did your spouse get/keep from selling crops from your farm for this past two months? ___________________ 4. How much did you get/keep from selling output from your spouse’s farms for this past two months? _________________ 5. In the last two months have you receive a gift from anyone?

6. In the last two months did you give any gifts?

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

Name/ I. Roster #

Residence of giver

Item

Value

I. Roster #

Residence of giver

Item

Value

PAGE 13 _______________________________________________________________________________________________

Section 5: EVENTS QUESTIONNAIRE _______________________________________________________________________________________________ I would like to ask you about recent events that have happened to you and your household. In the last 6 months have you lost any of the crop you have to pests? Y / N Date noticed

Crop

Units

Quantity

Value

Have you lost a job or not gotten work you were expecting any time during the last 2 months (Note: this includes all work except that on the respondents own farms)? Y / N

If yes, date _________________________ Have you not been paid an amount you were expecting for work in the last 2 months? Y / N

Date

Amount

Date

Amount

Date

Amount

Have you been paid an amount you were not expecting for work in the last 2 months? Y/N

Date

Amount

Date

Amount

Date

Amount

In the last 2 months, have you received any loan repayment you were not expecting?

Date

Amount

Date

Amount

Date

Amount

In the last 2 months, have you not received a loan repayment you were expecting?

Date

Amount

Date

Amount

Date

Amount

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

PAGE 14

Now I would like to ask you about your health for the last year. Please tell me about significant events that have happened regarding your health in the last year (including illness, pregnancies, etc)

Event

Date started

Duration # of days

Days of work/farming missed

Resulting expenses

Who paid?

Has anything been stolen from you in the last 6 months? Y / N

If yes, value ____________ Has anything been stolen from the household since last 6 months? Y / N

If yes, value ____________ Have you had a sudden death in your family in the last 6 months? Y/N

If yes, did you make any contributions towards the organization of the funeral? Y/N If yes, how much?_______________________ date_______________________

In the last 2 months have you had any other unexpected expenses? Y / N

event

date

amount

event

date

amount

In the last 2 months has your spouse had any unexpected expenses? Y / N

event

date

amount

event

date

amount

PAGE 15 If you farm pineapple, did you sell some in the last 6 months? Y / N sale date

did the exporter reject some?

was this more than expected?

Amount received

Amount that you expected to receive

Y / N

Y / N

Y / N

Y / N

Y / N

Y / N

Y / N

Y / N

Y / N

Y / N

Did you receive any payments in advance for the output? If Yes, how much? _________________________________ Did you have to return any of that amount? Y / N Why?______________________________________ Crops: 1 Cassava 2 Maize 3 Plantain 4 Cocoyam 5 Yam 6 Pineapple 7 Tomato 8 Garden egg 9 Okro 10 Pepper 11 Oranges 12 Banana 13 Avocado 14 Oil Palm 15 Cocoa 16 Sugar Cane 17 Bean

18 Groundnut Unit: 1 Pounds 2 Kilos 3 Tons 4 Minibag 5 Maxibag 6 Liter 7 Gallon 8 Beer bottle 9 Hundred Fruits 10 Cobs 11 American tin (olonka) 12 Basket 13 Bowl

Units continued 14 Bunch 15 Sucker 16 Kerosine tin 17 Rope 18 Headload 19 Tomato tin

Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11 Brother 12 Sister 13 Wofa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin

18 Local Friend m/f 19 Non-resident m/f friend 20 Extension Agent m/f 21 Market 22 Cooperative 23 NGO 24 Exporter 25 Village head 26 Lineage head 27 Priest 28 Trader m/f

Events 1 termites 2 other insects 3 rodents 4 grasscutter 5 other animal 6 wilt 7 leaves fall off 8 mushroom/fungus 9 rotten roots 10 rotten crops 11 not enough water 12 flood 13 chemical damage 14 planting error 15 unidentified disease 16 no market for crop

Severity: 1 Minor 2 Noticable loss 3 Significant loss 4 Major loss 5 Total Crop Failure Response: 1 None 2 Replant 3 Specific chemicals 4 Hand picking 5 Stop cultivating 6 Uproot affected area 7 Uproot entire farm

8 Destroy affected crop 9 Make traps 10 Change seed or variety 11 Change method of planting 12 Treat seeds before planting 13 Treat suckers 14 Can watering 15 Pump irrigation

PAGE 16 Tell me about the problems on your farm in the last six months (such as wilt, termites, pilfered crops, crop loss for some other reason)?

Dates

Plot

Event (use codes, if possible)

Crops

% of plot

severity

response

discussed with

value of damage

Have you had any sudden livestock deaths in the last 6 months? Y / N If yes, what type of livestock? ________________________ How many died?__________________________________ What did you do with the dead livestock? _______________________________________________________________

How many do you have left? _________________________

PAGE 17 _________________________________________________________________________________________________

Section 6: Asset Holdings and Trading Stocks _________________________________________________________________________________________________

I. A. I would like to ask you about your own stores of food and farm output: Food/crop

Unit

Quantity

Value

When stored?

How Treated?

Is this yours, your spouses’, or joint?

B. I would like to ask you about your own stores of seeds and planting material: Seed

Unit

Quantity

Value

When stored?

How Treated?

Crops/Food: 1 Cassava 2 Maize 3 Plantain 4 Cocoyam 5 Yam 6 Pineapple 7 Tomato 8 Garden egg 9 Okro 10 Pepper 11 Oranges 12 Banana

13 Avocado 14 Oil Palm 15 Cocoa 16 Sugar Cane 17 Bean 18 Groundnut Unit: 1 Pounds 2 Kilos 3 Metric Tonnes 4 Minibag 5 Maxibag

6 Liter 7 Gallon 8 Beer bottle 9 Hundred 10 Fruits 11Olonka 12 Basket 13 Bowl 14 Bunch 15 Sucker 16 Kerosine tin 17 Rope 18 Headload

19 Tomato tin 20 Marg. Tin 21 Cobs How Treated 1 None/Dried 2 Smoke 3 Dusban 4 Other Chemical (Specify)

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

PAGE 18 C. I would like to ask about your own stores of farm chemicals: Chemical

Unit

Quantity

Current Value

D. I would like to ask you about your own farm equipment, machines, motor vehicles or other equipment (e.g., chain saw, gun):

Equipment

Quantity

Age

Condition

Total current Value

E. I would like to ask you about any equipment you own for your businesses outside of farming (for example distilling equipment, a car, sewing machine, cooking utensils, grinding mill, carpentry tools)

Equipment

Quantity

Age

Condition

Total Current Value

PAGE 19

F. I would like to ask you about any goods you have to trade at this moment: Goods

Quantity

Units

Value

Chemicals: 1 Karate 2 Acellic 3 Dusban 4 Cymbush 5 Cymethoate 6 Sumithion 7 Furadan 8 Diazinon 11 Biobit 12 Dipel 2x 13 Of-nak 21 Atrazine 22 Hyvar X 23 Diuron 24 Gramoxone 25 Fusilad 26 Roundup 27 Basta

31 Cocide 32 Champion 33 Topsin M 34 Bavistin 35 Alielte 41 NPK 15-15-15 42 NPK 23-15-5 43 NPKMg 11-5-27-5 44 Ammonia 45 Urea 46 Lobi 44 47 Grofol (NPK) 48 K Fol (Potash) 49 Wuxal 51 Carbide

52 Biozyome 53 Almephon Unit: 1 Pounds 2 Kilos 3 Tons 4 Minibag 5 Maxibag 6 Liter 7 Gallon 8 Beer bottle 9 Hundred 10 Fruits 11 American tin 12 Basket 13 Bowl 14 Bunch 15 Sucker

16 Kerosine tin 17 Rope 18 Headload 19 Tomato tin Condition: 1 Like new 2 Good 3 Average 4 Poor 5 Not working

PAGE 20 II. Livestock: Please tell me about the animals that you own:

Number

Price of each if sold today

Shared? (If yes, enter Roster #)

How Shared?

Male

Female

Calves-Male

Goats

Calves-Female

Male

Female

Calves-Male

Sheep

Calves-Female

Male

Female

Pig

Piglets

Layers

Chickens

Broilers

Guinea Fowl

Other Poultry

Rabbits

Other:

Relationship: 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather

10Grandmother 11Brother 12 Sister 13 Wofa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin (m/f) 18 Local friend (m/f) 19 Non-residident

friend (m/f) 20 Extension Agent (m/f) 21 Market 22 Cooperative 23 NGO 24 Exporter 25 Village head 26 Lineage head 27 Priest 28 Trader (m/f)

PAGE 21 III. Financial Saving: A. Do you use a susu collector? Y/N [If no, go to B] Frequency

Deposit Amount

Accumulated Savings

Collection Date

Relationship to collector

B. Are you a member of an esusu group? Y/N [If no, go to C] Frequency

Deposit Amount

Pot size

Collection date

Relationship with group leader

C. Are you holding money for anyone? Y/N [If no, go to D] Relationship

Location

Amount

Since when

D. Is anyone holding any money for you? Y/N [If no, go to E] Relationship

Location

Amount

Since when

Relationship: 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter

9 Grandfather 10Grandmother 11Brother 12 Sister 13 Wofa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin (m/f)

18 Local friend (m/f) 19 Non-residident friend (m/f) 20 Extension Agent (m/f) 21 Market 22 Cooperative 23 NGO 24 Exporter

25 Village head 26 Lineage head 27 Priest 28 Trader (m/f)

PAGE 22 E. Do you have any deposits in a bank? Y/N [If no, go to F] Location

Type of Account

Balance

F. Do you have any bonds, stock or other financial asset? Y/N [If no, go to G] Use this space to record details (cost of bond, maturity date, face value, or company and number of shares):

G. Do you currently receive, or do you anticipate receiving, a pension? Y/N

Source:____________________________________________________________

Year started receiving/expect to start receiving:____________________________

Current monthly amount:___________ or anticipated monthly amount: _________

H. Could you tell us the approximate current value of your jewelry, if any? ___________

I. Could you tell us the approximate current value of your cloth, if any? ______________

At this point, reiterate the confidential nature of the interview!

J. Could you tell us how much cash you have now? ______________________________

In addition to these cedis, do you have any foreign currency? _________________

If yes, how much ___________________________________________________

PAGE 23 Now I would like to ask you about the stocks of goods that you own for trading. Could you please list for me the goods that you own that you are planning to sell? This includes raw materials to be used in any business (such as food for cooking for sale), but it does not include items harvested from your own plots.

Item Quantity

Total Value

PAGE 24

_________________________________________________________________________________________________________ Section 7: Family Background

_________________________________________________________________________________________________________

1. Family structure and wealth Family or village office you hold _________________________________________________________ Please describe any privilege this office gives you:

Family home town__________ and region__________ Your first language_______________________ Are you the first of your family to reside here? Y/N

If no, for approximately how many years has your family lived here?_______________________ If yes, for approximately how many years have you lived here?____________________________

Which abusua/clan are you a member of: (1) Oyoko (2) Asone (3) Agona Other __________________________________________________ Your income this year from family land rented out ____________ Size of family land sold or leased (long-term) in last 12 months ___________ Your income from sale _________

2. Parents’ Economic Background

If dead, Year of death

Years of school

First occupation

Second occupation

birthplace

residence

wife # or # of wives

# of children

mother

father

Family or village office your parent’s hold/held _________________________________________________________ Please describe any privilege this office gives you:

Land Size Unit 1 Ropes 2 Poles 3 Acres 4 Hectares

5 Square miles 6 Farms Period

1 Daily 2 Past week 3 Past two weeks 4 Previous full month 5 Since January 1

6 Annual 7 Last major season 8 Last minor season 9 School term

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE

ROUND

PAGE 25 Age at start of independent farming (or first marriage, if not a farmer)_______________ Please give us the following information about your parents. If they are already death, please state at year of death…

3. Fosterage Did you live with anyone away from your parents for at least a year while growing up?

4. Land Inheritances, received or anticipated

Do you have someone you might expect to inherit/ have inherited land from? Y/N If yes, what is your relationship to the person?______________________________

What is the size of the land?______________________________________ What is the current value of the land?_______________________________

Do you expect to inherit anything else? Y/N If yes, what is your relationship to the person?______________________________

What is the current value of the item?_______________________________ Residence/Place Roof Floor 7 Grandson 18 Local Friend Purpose 1 Daman/Ahweriase 12 Not in Ghana 1 Roofing Sheets 1 Concrete 8 Granddaughter 19 Non-resident friend 1 Househelp 2 Pokrom/Nsabaa 13 Family home town 2 Bamboo 2 Dirt 9 Grandfather 20 Extension Agent 2 Schooling 3 Oboadaka/ Kwamekrom 3 Grass/palm thatch 10 Grandmother 21 Market 3 Adopted 4 Konkonuru Land Area Unit Relationship 11Brother 22 Cooperative 4 Farming 5 Aburi 1 Ropes Walls 1 Self 12 Sister 23 NGO 5 Apprentice 6 Nsawam 2 Poles 1 Brick 2 Spouse 13Wbfa 24 Exporter Occupation 7 Eastern region rural 3 Acres 2 Plaster over brick 3 Son 14 Uncle 25 Village head 1 Farmer 8 Eastern region town 4 Hectares 3 Concrete block 4 Daughter 15 Aunt 26 Lineage head 2 Trader 9 Accra 5 Farms 4 Mud 5 Father 16 Co-wife 27 Priest 3 Artisan

10 Other Urban 5 Bamboo 6 Mother 17 Cousin 28 Trader 4 Civil Servant

5 Labourer

Size of farms cultivated by mother and father own

house? Y/N

material for

unit

total

cocoa

oil palm

food

other

roof

walls

floor

mother

father

ages

relationship

place purpose foster parent’s

occupation

number of farms

main farming activity

PAGE 26 Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11Brother 12 Sister 13WBfa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin 18 Local Friend 19 Non-resident friend 20 Extension Agent 21 Market 22 Cooperative 23 NGO 24 Exporter 25 Village head 26 Lineage head 27 Priest 28 Trader Residence/Location 1 Daman/Ahweriase 2 Pokrom/Nsabaa 3 Oboadaka/ Kwamekrom 4 Konkonuru 5 Aburi 6 Nsawam 7 Eastern region rural 8 Eastern region town 9 Accra 10 Other Urban

11 Other Rural 12 Not in Ghana Land Area Unit 1 Ropes 2 Poles 3 Acres 4 Hectares 5 Farms

PAGE 27 _______________________________________________________________________________________________________

Section 8: Learning (Village1) _______________________________________________________________________________________________________

Use y=yes, n=no, 99=yes but he wouldn’t know/have any Name --> xxxxx Traditional

Leader Extension officer

HHN/ID ---> 31 0 100 9 200 9 Do you know ---? Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N When did you last talk with ---?

In a normal month, how often do you talk with ----?

Have you ever gone to --- for advice about your farm?

Could you go to --- if you had a problem with unhealthy crops?

Could you go to --- if you wanted to find a buyer for any of your crops?

Could you ever go to---if you had financial problems?

Could you ever go to---to borrow money?

Could you ever go to – for advice on how to access loans?

y – yes Codes for Last spoke n – no 1 today 99 - yes, but he wouldn’t know/have any 2 yesterday

3 past week 4 past two weeks 5 past month 6 past six months 7 past year 8 more than a year now

HHN

RESPONDENT RESPOND. #

DATE

ROUND

VILLAGE 1

PAGE 28

Organizations Are you still a member of _________________________ Y /N ________________________ Y / N ________________________________ Y / N ______________________________________ Y / N (Fill in from 1996 groups identified) If not, why did you leave the organization? Which of the groups is no longer functioning?____________________________________________ Ask if respondent is currently a member of the following organizations 1. Onipa Hiamoa Kuw (community improvement group) (0) not a member (1) active member (2) inactive/past member (Circle one) 2. Sie wo ho (funeral organization) (0) not a member (1) active member (2) inactive/past member (Circle one) 3. Biakoy kuo (farming organization) (0) not a member (1) active member (2) inactive/past member (Circle one) 4. Youth group (0) not a member (1) active member (2) inactive/past member (Circle one) 5. Community works group (0) not a member (1) active member (2) inactive/past member (Circle one) 6. Any other organization? Name: _____________ (0) not a member (1) active member (2) inactive/past member (Circle one)

PAGE 27 ________________________________________________________________________________________________________

Section 8: Learning (Village2) ________________________________________________________________________________________________________

Use y=yes, n=no, 99=yes but he wouldn’t know/have any Name --> xxxxx Traditional

Leader Extension officer

HHN/ID ---> 61 0 100 9 200 9 Do you know ---? Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N When did you last talk with ---?

In a normal month, how often do you talk with ----?

Have you ever gone to --- for advice about your farm?

Could you go to --- if you had a problem with unhealthy crops?

Could you go to --- if you wanted to find a buyer for any of your crops?

Could you ever go to---if you had financial problems?

Could you ever go to---to borrow money?

Could you ever go to – for advice on how to access loans?

y – yes Codes for Last spoke n – no 1 today 99 - yes, but he wouldn’t know/have any 2 yesterday

3 past week 4 past two weeks 5 past month 6 past six months 7 past year 8 more than a year now

HHN

RESPONDENT RESPOND. #

DATE

ROUND

VILLAGE 2

PAGE 28

Organizations Are you still a member of ____________________________________________________________________ (Fill in from 1996 groups identified) If not, why you leave the organization? Which of the groups is no longer functioning?____________________________________________ Ask if respondent is currently a member of the following organizations 1. Town Development Committee (0) not a member (1) active member (2) inactive/past member (Circle one) 2. School Management Committee (0) not a member (1) active member (2) inactive/past member (Circle one) 3. Assembly Committee (0) not a member (1) active member (2) inactive/past member (Circle one) 4. Funeral Committee (0) not a member (1) active member (2) inactive/past member (Circle one) 5. Xxx Pineapple Farmers Association (0) not a member (1) active member (2) inactive/past member (Circle one) 6. New (unnamed) Pineapple Farmers Association (0) not a member (1) active member (2) inactive/past member (Circle one) 7. Clinic Expansion Project (0) not a member (1) active member (2) inactive/past member (Circle one) 8. Any other organization?_______________________________ (name) (0) not a member (1) active member (2) inactive/past member (Circle one)

PAGE 27 _______________________________________________________________________________________________________

Section 8: Learning (Village3) ______________________________________________________________________________________________________

Use y=yes, n=no, 99=yes but he wouldn’t know/have any Name --> xxxxx Traditional

Leader Extension officer

HHN/ID ---> 44 0 51 or 68 0 200 9 Do you know ---? Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N When did you last talk with ---?

In a normal month, how often do you talk with ----?

Have you ever gone to --- for advice about your farm?

Could you go to --- if you had a problem with unhealthy crops?

Could you go to --- if you wanted to find a buyer for any of your crops?

Could you ever go to---if you had financial problems?

Could you ever go to---to borrow money?

Could you ever go to – for advice on how to access loans?

y – yes Codes for Last spoke n – no 1 today 99 - yes, but he wouldn’t know/have any 2 yesterday

3 past week 4 past two weeks 5 past month 6 past six months 7 past year 8 more than a year now

HHN

RESPONDENT RESPOND. #

DATE

ROUND

VILLAGE 3

PAGE 28

Organizations Are you still a member of ____________________________________________________________________ (Fill in from 1996 groups identified) If not, why you leave the organization? Which of the groups is no longer functioning?____________________________________________ Ask if respondent is currently a member of the following organizations 1. Xxx Farmers Association (0) not a member (1) active member (2) inactive/past member (Circle one) 2. Xxx Pineapple Growers and Cooperative Society (0) not a member (1) active member (2) inactive/past member (Circle one) 3. Road Improvement Project (headed by Xxx) (0) not a member (1) active member (2) inactive/past member (Circle one) 4. Community Improvement Group (headed by Xxxx) (0) not a member (1) active member (2) inactive/past member (Circle one)

5. Town Development Committee (0) not a member (1) active member (2) inactive/past member (Circle one) 6. Any other organization?_______________________________ (name) (0) not a member (1) active member (2) inactive/past member (Circle one)

PAGE 27 _______________________________________________________________________________________________________

Section 8: Learning (Village4) _______________________________________________________________________________________________________

Use y=yes, n=no, 99=yes but he wouldn’t know/have any Name --> xxxxx Traditional

Leader Extension officer

HHN/ID ---> 81 0 100 0 200 9 Do you know ---? Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N Y / N When did you last talk with ---?

In a normal month, how often do you talk with ----?

Have you ever gone to --- for advice about your farm?

Could you go to --- if you had a problem with unhealthy crops?

Could you go to --- if you wanted to find a buyer for any of your crops?

Could you ever go to---if you had financial problems?

Could you ever go to---to borrow money?

Could you ever go to – for advice on how to access loans?

y – yes Codes for Last spoke n – no 1 today 99 - yes, but he wouldn’t know/have any 2 yesterday

3 past week 4 past two weeks 5 past month 6 past six months 7 past year 8 more than a year now

HHN

RESPONDENT RESPOND. #

DATE

ROUND

VILLAGE 4

PAGE 28

Organizations Are you still a member of ____________________________________________________________________ (Fill in from 1996 groups identified) If not, why you leave the organization? Which of the groups is no longer functioning?____________________________________________ Ask if respondent is currently a member of the following organizations 1. Wo Yε Na Yε Yε Ma Wo (funeral organization) (0) not a member (1) active member (2) inactive/past member (Circle one) 2. Any other organization? _______________________________ (name) 0) not a member (1) active member (2) inactive/past member (Circle one)

Page 29 ____________________________________________________________________________________________________________________

Section 9a: Family Expenses ________________________________________________________________________________

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

Yourself Spouse All others Has anyone in your household spent anything on ... Y/N

Period

Value Value Value Relation School fees, parent assoc. School term School uniform and shoes Annual Other school expenses including books and stationary

School term

Medicine Previous full month Other medical expenses Previous full month Adult shoes Annual Children shoes Annual Cloth Annual Material for adult clothes Annual Material for child clothes Annual Adult clothes Previous full month Child clothes Previous full month Soap Past two weeks Personal care products (pomade, combs, shampoo...)

Previous full month

Brooms, home maint. products Previous full month Home repairs, painting Annual Building of new house in this town Annual Building of new house in another town Annual Public transportation Previous full month Petrol, motor oil Previous full month Repairs, other vehicle exp. Previous full month Newspapers Previous full month Charcoal, firewood Past two weeks Matches, candles Previous full month Other fuel for cooking, light Previous full month Stove, coal pot Annual Kitchen equipment (pots, dishes) Annual Lanterns Previous full month Furniture Last 3 months Sheets, towels Previous full month Hairdressing, haircuts Past two weeks Domestic servants Previous full month Jewelry and watches Last 3 months Entertainment (cinema, sports, tapes, toys)

Previous full month

Taxes Previous full month

Page 30

Yourself Spouse All others Has anyone in your household spent anything on ... in the past 12 months? Y/N

Period

Value Value Value Relation

Weddings, dowries Annual Contributions for organizing a family funeral

Last 3 months

Attending funerals Previous full month Lottery tickets Past two weeks Cigarettes, tobacco, cola Daily Sewing machine Annual Radio, tape player Annual Other major appliance Annual Bicycle Annual Motorbike Annual Car, other vehicle Annual Camera Annual

Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11 Brother 12 Sister 13 W]fa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin 18 Local Friend m/f 19 Non-resident friend m/f 20 Extension Agent m/f 21 Market 22 Cooperative 23 NGO 24 Exporter m/f 25 Village head 26 Lineage head 27 Priest 28 Trader m/f 29 employed by resp (employee) 30 employer of resp. 31 landlord 32 tenant/renter 33 niece 34 nephew

Page 31 ____________________________________________________________________________________________________________________

Section 9b: Food From Family Farms ________________________________________________________________________________

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

From your farms

From spouse’s farm

From other farms

Have people in your household eaten … from your own farms in the past 12 months? Y/N

Period

Value Value Value RelationMaize Past week Fresh Cassava Past week Gari, kokonte, other Previous full

month

Yams Annual Cocoyam Past week Plantain Previous full

month

Potato, sweet potato Previous full month

Oil Palm Fruit Past two weeks Groundnuts Annual Fish, crabs, snails, shellfish Past week Chicken Annual Pigeon, dove, duck, turkey, other fowl

Annual

Goat, sheep Annual Bushmeat Previous full

month

Eggs Past two weeks Palm oil Previous full

month

Other oil Previous full month

Pineapples Past week Oranges, mangoes, pawpaws, bannanas, coconut, other fruit

Previous full month

Sugar cane, honey Previous full month

Palm wine, akpeteshie Last 3 months Other drinks Last 3 months Tomatoes Annual Onions Annual Garden egg, Okro Annual Beans, peas Annual Pepper Past week Kontomle Past week Other vegetables Past two weeks Other foods Past week Firewood Past week

Page 32 Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11 Brother 12 Sister 13 W]fa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin 18 Local Friend m/f 19 Non-resident friend m/f 20 Extension Agent m/f 21 Market 22 Cooperative 23 NGO 24 Exporter m/f 25 Village head 26 Lineage head 27 Priest 28 Trader m/f 29 employed by resp (employee) 30 employer of resp. 31 landlord 32 tenant/renter 33 niece 34 nephew

Page 33

____________________________________________________________________________________________________________________

Section 9c: Purchased Food ________________________________________________________________________________

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

Yourself Spouse All Others Has anyone in your household purchased… for consumption in the past 12 months? Y/N

Period Value Value Value Relation

Maize Past week Corn dough Past week Rice Previous full

month

Bread, macaroni or flour Past week Cassava Past week Gari, kokonte, other Previous full

month

Yams Previous full month

Cocoyam Previous full month

Plantain Previous full month

Potato, sweet potato, millet, guinea corn, sorghum

Previous full month

Oil Palm Fruit Past week Groundnuts Past week Fish, crabs, snails, shellfish Daily Chicken Annual Pigeon, dove, duck, turkey, other fowl

Annual

Goat, sheep Previous full month

Beef Previous full month

Eggs Previous full month

Palm oil Past week Other oil Previous full

month

Margarine, butter Previous full month

Pineapples Previous full month

Oranges, mangoes, pawpaws, bannanas, coconut, other fruit

Previous full month

Sugar cane, toffee, honey Previous full month

Milk, milk powder, baby food

Previous full month

Page 34

Yourself Spouse All Others

Has anyone in your household purchased… for consumption in the past 12 months? Y/N

Period

Value Value Value Relation Sugar, milo, tea, coffee Previous full

month

Non-alchoholic beverages (home or away from home)

Previous full month

Alchoholic beverages (home or away from home)

Previous full month

Salt Previous full month

Tomatoes Past week Onions Past week Garden egg, Okro Past week Beans, peas Previous full

month

Pepper Past week Kontomle Past two weeks Other vegetables Past two weeks Fufu, kenkey, banku, other prepared foods, chop bars

Daily

Other food Past week

Page 35 ____________________________________________________________________________________________________________________

Section 10: Other Income ________________________________________________________________________________ 1. Please tell me about any business other than farming you did in the last two months: 2.

Please tell me about the work for pay for the past two weeks: Period: 1 Daily 7 Last major season 2 Past week 8 Last minor season 3 Past two weeks 4 Previous full month 5 Since January 1 6 Annual

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

Type of business Value of initial stock of raw materials or goods for sale

A. Expenses Purchased raw materials or goods for sale Raw materials or goods for sale from farm Own labor days:

daily hours: daily value:

days: daily hours: daily value:

days: daily hours: daily value:

Spouse labor days: daily hours: daily value:

days: daily hours: daily value:

days: daily hours: daily value:

Other unpaid labor days: daily hours: daily value:

days: daily hours: daily value:

days: daily hours: daily value:

Wages and payments for services Transport/fuel Building costs or equipment purchase Equipment or building repairs, or rental Taxes or fees paid Other B. Revenue: Sales/Other Payments Value of own and household consumption Value of gifts given from business Value of final stock of raw materials or goods for sale

1 2 3 Occupation Casual or permanent work? Days worked over the past two weeks Cash earnings the past two weeks Value of kind earnings the past two weeks Average hours per day Have you been paid in full

Page 36 3. Please tell me about your income from farms for the past two months. 4. In the past two months have you had any other source of income, beyond those that we have

discussed so far? For example, there might be a pension, or you might have received money from bride payments or inheritances. Also includes income from sales of land.

5. Does your spouse get any income from a business other than farming? Y / N

Does your spouse work for pay? Y / N Does your spouse get income from farms outside of this town? Y / N Does your spouse get income from any other source other than farming? Y / N

If the answer to any of these questions is yes, please estimate for me the net income earned by your spouse from her business or employment:

Period: 1 Daily 2 Past week 3 Past two weeks 4 Previous full month 5 Since January 1 6 Annual 7 Last major season 8 Last minor season Weather: 1 Better than usual 2 Same as usual 3 Worse than usual

Location Expenses Income from sale of output

Weather in that location over that period

1 2 3 4 5 6 7 8 9

Type of Income Amount of income since our last interview

Spouse’s business, job or other income source

Spouse ID

Period Net Income

Page 37 ____________________________________________________________________________________________________________________

Section 11: Plot Questionnaire ________________________________________________________________________________ I. Current Contracts

A. Is this your own land. Or your family land? Y / N [If yes, go to B]

1. Are you renting this plot? Y / N [If no, skip to 2]

Contract type: Abusa Abunu Cash Rent Other:_____________________

Year contract began:_____________ Month:_______________________

Crop/Input Share of tenant Is the rent dependent on the type of crop you plant? Y / N

If yes, how so?___________________________________________________________

Any planned division of land? Y / N If yes, map.

Rent is for how long?________________ Outstanding rent:_____________________

Value of payments made for this rental period:__________________________________

Residence of owner:_________________ Relationship to owner:_________________

Name of owner:__________________________________________________________

Do you do other business with the owner? (Y / N) If Yes, explain_________________

_______________________________________________________________________

Have there been any disputes over the rent during this period (Y / N) If Yes, explain

_______________________________________________________________________

2. Other (i.e. neither own/rent: explain):_________________________________________

Month/ Year when you began farming this plot: Have you given anything to the owner on account of the land? Y/ N

Cash:________ Kind:_________________ Other:__________________

Do you plan to give anything to the owner on account of the land? Y/ N

Cash:________ Kind:_________________ Other:__________________

Value of payments made for this rental period:__________________________________

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

Page 38

Residence of owner:_________________ Relationship to owner:_________________

Name of owner:__________________________________________________________

Do you do other business with the owner? (Y / N) If Yes, explain________________

_______________________________________________________________________

Have there been any disputes over the rent during this period? (Y / N) If Yes, explain

_______________________________________________________________________

B. Is someone else using this land? Y/N [If no go to II]

1. Are you renting this plot out to someone else? Y / N [If no, skip to 2]

Contract type: Abusa Abunu Cash Rent Other:_____________________

Year contract began:_____________ Month:_______________________

Crop/Input Share of landlord Any planned division of land? Y / N If yes, map.

Amount of fixed rent:________________ Date paid:___________________________

Rent is for how long?________________ Outstanding rent:_____________________

Value of payments made for this rental period:__________________________________

Residence of tenant:_________________ Relationship to tenant:_________________

Name of tenant:__________________________________________________________

Do you do other business with the owner? (Y / N) If Yes, explain _________________

_______________________________________________________________________

[Go to II]

2. Other (explain):__________________________________________________________ Month/ Year when he/she began cultivating this plot: Has he/she given you anything on account of the land? Y / N

Cash:________ Kind:_________________ Other:__________________

Do you expect to receive anything on account of the land? Y/ N

Cash:________ Kind:_________________ Other:__________________

Residence of other person:_______ Relationship to other person :_________________

Name of other person:_____________________________________________________

Page 39

Do you do other business with the owner? (Y/N, explain)_________________________

_______________________________________________________________________

[Go to II]

II. Plot History

How did you obtain this land?_____________________________________________________________

From whom did you obtain the land?_______________________________________________________

When did you obtain this land?____________________________________________________________

Did they pay anything on account of this land? Y/N

Amount____________________ Period__________________ To whom_______________

If respondent purchased, full price:__________________________________________________________

If respondent mortgaged, terms:_____________________________________________________________

Have you paid anything this year on account of the land to the person from whom you obtained it? Y/ N

Cash:________ Kind:____________ Other:_______________ Abunu/Abusa____________

Crop/Input Share of landlord

Do you expect to pay anything this year on account of the land? Y/ N

Cash:________ Kind:____________ Other:_______________ Abunu/Abusa____________

Crop/Input Share of landlord

Tenancy Disputes: Have you had a dispute with your tenant/ borrower? Y/N

Land Disputes: Have there been any disputes over the control of this plot in the family?

If Yes, explain______________________________________________________

Has the village head intervened in any dispute? Y/N

Have there been any court cases regarding this plot? Y/N

Any written documentation regarding the plot? Y/N

Page 40 Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11 Brother 12 Sister 13 W]fa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin 18 Local Friend m/f 19 Non-resident friend m/f 20 Extension Agent m/f 21 Market 22 Cooperative 23 NGO 24 Exporter m/f 25 Village head 26 Lineage head 27 Priest 28 Trader m/f 29 employed by resp (employee) 30 employer of resp. 31 landlord 32 tenant/renter 33 niece 34 nephew Abusua 1 agona 2 asona 3 bretuo 4 asakyiri 5 abrade 6 oyoko 7 weme 8 twidan 9 anlo 10 nkonya 11 kona 12 aduama/aduana 13 atsegeme Divorce 1. Self 2. Spouse 3. Share Crops 1 Cassava 2 Maize 3 Plantain 4 Cocoyam 5 Yam 6 Pineapple 7 Tomato 8 Garden egg

Page 41 9 Okro 10 Pepper 11 Oranges 12 Banana 13 Avocado 14 Oil Palm 15 Cocoa 16 Sugar Cane 17 Bean 18 Groundnut 19 Cabbage 20 Cocoa 21 Pawpaw 22 Firewood 23 Watermelon 24 Coconut 25 Cashew 26 Ginger 27 Sweet pepper 28 Potato 29 Sun flower Events 1 termites 2 other insects 3 rodents 4 grasscutter 5 other animal 6 wilt 7 leaves fall off 8 mushroom/fungus 9 rotten roots 10 rotten crops 11 not enough water 12 flood 13 chemical damage 14 planting error 15 unidentified disease 16 no market forcrop 17 theft 18 bush fire (unintended) Severity 1 Minor 2 Noticable loss 3 Significant loss 4 Major loss 5 Total Crop Failure Response 1 None 2 Replant 3 Specific chemicals 4 Hand picking 5 Stop cultivating 6 Uproot affected area 7 Uproot entire farm 8 Destroy affected crop 9 Make traps 10 Change seed or variety 11 Change method of planting 12 Treat seeds before planting 13 Treat suckers 14 Can watering 15 Pump irrigation 16 weeding

Page 42 ____________________________________________________________________________________________________________________

Section 13: Plot Wrap up ________________________________________________________________________________ One form to be filled for each of the plots on the respondent’s plot list. 1. If someone wanted to buy all of the crops on this land now (even if they are not ready), what

would they be worth?_________________________________________________________

2. Please tell me about each crop you are growing, when you will harvest it, and the total value all of that crop will have when you harvest it.

if the respondent is growing the same crop at different stages enter them as different crops (for example, 2a, 2b, 2c) 3. Who did you obtain this plot from? _______ (add to I. roster if not there yet) 4. This plot belongs to which abusua?

(1) agona (2) asona (3) bretuo (4) asakyiri (5) abrade (6)oyoko (Other) ___________ 5. Would your spouse have access to farm this plot if you (god forbid) died? Y/N

If Y for how many years? ______ (99 = as long as they want) 6. Would your children have access to farm this plot if you (god forbid) died? Y/N

If Y for how many years ? _______ (99 = as long as they want) 7. Will you still be able to farm this plot if you and your spouse were to be divorced? Y/N

If Y for how many years? ______ (99= as long as they want) 8. Who would get the crops growing on this plot if you were divorced? (1) me (2) spouse (3) share _____% resp _______% spouse 9. Comments:

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

Crop Date to harvest Final Value of all the crop growing now

Page 43

Relationship 1 Self 2 Spouse 3 Son 4 Daughter 5 Father 6 Mother 7 Grandson 8 Granddaughter 9 Grandfather 10 Grandmother 11 Brother 12 Sister 13 W]fa 14 Uncle 15 Aunt 16 Co-wife 17 Cousin 18 Local Friend m/f 19 Non-resident friend m/f 20 Extension Agent m/f 21 Market 22 Cooperative 23 NGO 24 Exporter m/f 25 Village head 26 Lineage head 27 Priest 28 Trader m/f 29 employed by resp (employee) 30 employer of resp. 31 landlord 32 tenant/renter 33 niece 34 nephew Abusua 1 agona 2 asona 3 bretuo 4 asakyiri 5 abrade 6 oyoko 7 weme 8 twidan 9 anlo 10 nkonya 11 kona 12 aduama/aduana 13 atsegeme Di vorce 1. Self 2. Spouse 3. Share Crops 1 Cassava 2 Maize 3 Plantain 4 Cocoyam 5 Yam 6 Pineapple 7 Tomato 8 Garden egg 9 Okro

Page 44 10 Pepper 11 Oranges 12 Banana 13 Avocado 14 Oil Palm 15 Cocoa 16 Sugar Cane 17 Bean 18 Groundnut 19 Cabbage 20 Cocoa 21 Pawpaw 22 Firewood 23 Watermelon 24 Coconut 25 Cashew 26 Ginger 27 Sweet pepper 28 Potato 29 Sun flower Events 1 termites 2 other insects 3 rodents 4 grasscutter 5 other animal 6 wilt 7 leaves fall off 8 mushroom/fungus 9 rotten roots 10 rotten crops 11 not enough water 12 flood 13 chemical damage 14 planting error 15 unidentified disease 16 no market forcrop 17 theft 18 bush fire (unintended) Severity 1 Minor 2 Noticable loss 3 Significant loss 4 Major loss 5 Total Crop Failure Response 1 None 2 Replant 3 Specific chemicals 4 Hand picking 5 Stop cultivating 6 Uproot affected area 7 Uproot entire farm 8 Destroy affected crop 9 Make traps 10 Change seed or variety 11 Change method of planting 12 Treat seeds before planting 13 Treat suckers 14 Can watering 15 Pump irrigation 16 weeding

Page 45 _____________________________________________________________________________________________________________________________

Section 14: Birth History ______________________________________________________________________________________

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND

NO.

QUESTIONS AND FILTERS CODING CATEGORIES SKIP

201

Now I would like to ask about all the births you have had during your life. Have you ever given birth?

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

──<206

202

Do you have any sons or daughters to whom you have given birth who are now living with you?

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

──<204

203

How many sons live with you? And how many daughters live with you? IF NONE, RECORD ‘00’.

┌──┬──┐SONS AT HOME .................│░░│░░│ ├──┼──┤DAUGHTERS AT HOME .......│░░│░░│ └──┴──┘

204

Do you have any sons or daughters to whom you have given birth who are alive but do not live with you?

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

──<206

205

How many sons are alive but do not live with you? And how many daughters are alive but do not live with you? IF NONE, RECORD ‘00’.

┌──┬──┐SONS ELSEWHERE ............│░░│░░│ ├──┼──┤DAUGHTERS ELSEWHERE ..│░░│░░│ └──┴──┘

206

Have you ever given birth to a boy or girl who was born alive but later died? IF NO, PROBE: Any baby who cried or showed signs of life but did

not survive?

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

──<208

207

How many boys have died? And how many girls have died? IF NONE, RECORD ‘00’.

┌──┬──┐BOYS DEAD.......................│░░│░░│ ├──┼──┤GIRLS DEAD ......................│░░│░░│ └──┴──┘

208

SUM ANSWERS TO 203, 205, AND 207, AND ENTER TOTAL.

IF NONE, RECORD ‘00’.

┌──┬──┐TOTAL ..............................│░░│░░│ └──┴──┘

209

CHECK 208: Just to make sure that I have this right: you have had in TOTAL _____ births during your life. Is that correct? ┌──┐ ┌──┐ PROBE AND YES ├──┘ NO └──┴──< CORRECT │ 201-208 AS ? NECESSARY.

210

CHECK 208: ONE OR MORE ┌──┐ NO BIRTHS ┌──┐ BIRTHS ├──┘ └──┴────────────────────────────────────── ?

──<END

Page 46

211

Now I would like to record the names of all your births, whether still alive or not, starting with the first one you had. RECORD NAMES OF ALL THE BIRTHS IN 212. RECORD TWINS AND TRIPLETS ON SEPARATE LINES.

212

213

214

215

216

217 IF ALIVE:

218 IF ALIVE

219 IF ALIVE:

220 IF DEAD:

221

What name was given to your (first/next) baby? (NAME)

Were any of these births twins?

Is (NAME) a boy or a girl?

In what month and year was (NAME) born? PROBE: What is his/her birthday?

Is (NAME) still alive?

How old was (NAME) at his/her last birthday? RECORD AGE IN COM-PLETED YEARS.

Is (NAME) living with you?

RECORD HOUSEHOLD LINE NUMBER OF CHILD (RECORD ‘00’ IF CHILD NOT LISTED IN HOUSEHOLD)

How old was (NAME) when he/she died? IF ‘1 YR’, PROBE: How many months old was (NAME)? RECORD DAYS IF LESS THAN 1 MONTH; MONTHS IF LESS THAN TWO YEARS; OR YEARS.

Were there any other live births between (NAME OF PREVIOUS BIRTH) and (NAME)?

01

SING...1 MULT..2

BOY.. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS.......1 │░░│░░│ ├──┼──┤MONTHS .2 │░░│░░│ ├──┼──┤YEARS ....3 │░░│░░│ └──┴──┘

02

SING...1 MULT..2

BOY.. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS.......1 │░░│░░│ ├──┼──┤MONTHS .2 │░░│░░│ ├──┼──┤YEARS ....3 │░░│░░│ └──┴──┘

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

03

SING...1 MULT..2

BOY.. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS.......1 │░░│░░│ ├──┼──┤MONTHS .2 │░░│░░│ ├──┼──┤YEARS ....3 │░░│░░│ └──┴──┘

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

04

SING...1 MULT..2

BOY.. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS.......1 │░░│░░│ ├──┼──┤MONTHS .2 │░░│░░│ ├──┼──┤YEARS ....3 │░░│░░│ └──┴──┘

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

05

SING...1 MULT..2

BOY.. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS.......1 │░░│░░│ ├──┼──┤MONTHS .2 │░░│░░│ ├──┼──┤YEARS ....3 │░░│░░│ └──┴──┘

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

06

SING...1 MULT..2

BOY.. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS.......1 │░░│░░│ ├──┼──┤MONTHS .2 │░░│░░│ ├──┼──┤YEARS ....3 │░░│░░│ └──┴──┘

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

Page 47

212

213

214

215 216 217

IF ALIVE: 218 IF ALIVE

219 IF ALIVE:

220 IF DEAD:

221

What name was given to your next baby? (NAME)

Were any of these births twins?

Is (NAME) a boy or a girl?

In what month and year was (NAME) born? PROBE: What is his/her birthday?

Is (NAME) still alive?

How old was (NAME) at his/her last birthday? RECORD AGE IN COM-PLETED YEARS.

Is (NAME) living with you?

RECORD HOUSEHOLD LINE NUMBER OF CHILD (RECORD �00' IF CHILD NOT LISTED IN HOUSEHOLD)

How old was (NAME) when he/she died? IF ‘1 YR’, PROBE: How many months old was (NAME)? RECORD DAYS IF LESS THAN 1 MONTH; MONTHS IF LESS THAN TWO YEARS; OR YEARS.

Were there any other live births between (NAME OF PREVIOUS BIRTH) and (NAME)?

07

SING ...1 MULT ..2

BOY .. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO ...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS ...... 1 │░░│░░│ ├──┼──┤MONTHS 2 │░░│░░│ ├──┼──┤YEARS.... 3 │░░│░░│ └──┴──┘

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

08

SING ...1 MULT ..2

BOY .. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO ...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS ...... 1 │░░│░░│ ├──┼──┤MONTHS 2 │░░│░░│ ├──┼──┤YEARS.... 3 │░░│░░│ └──┴──┘

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

09

SING ...1 MULT ..2

BOY .. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO ...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS ...... 1 │░░│░░│ ├──┼──┤MONTHS 2 │░░│░░│ ├──┼──┤YEARS.... 3 │░░│░░│ └──┴──┘

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

10

SING ...1 MULT ..2

BOY .. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO ...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS ...... 1 │░░│░░│ ├──┼──┤MONTHS 2 │░░│░░│ ├──┼──┤YEARS.... 3 │░░│░░│ └──┴──┘

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

11

SING ...1 MULT ..2

BOY .. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO ...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS ...... 1 │░░│░░│ ├──┼──┤MONTHS 2 │░░│░░│ ├──┼──┤YEARS.... 3 │░░│░░│ └──┴──┘

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

12

SING ...1 MULT ..2

BOY .. 1 GIRL . 2

┌──┬──┐ MONTH│░░│░░│ └──┴──┘ YEAR ┌──┬──┬──┬──┐ │░░│░░│░░│░░│ └──┴──┴──┴──┘

YES .... 1 NO ...... 2 │ ? 220

AGE IN YEARS ┌──┬──┐ │░░│░░│ └──┴──┘

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

LINE NUMBER

┌──┬──┐ │░░│░░│ └──┴──┘

│ ?

(NEXT BIRTH)

┌──┬──┐DAYS ...... 1 │░░│░░│ ├──┼──┤MONTHS 2 │░░│░░│ ├──┼──┤YEARS.... 3 │░░│░░│ └──┴──┘

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

222

Have you had any live births since the birth of (NAME OF LAST BIRTH)? YES

NO

223

COMPARE 208 WITH NUMBER OF BIRTHS IN HISTORY ABOVE AND MARK:

NUMBERS ┌──┐ NUMBERS ARE ┌──┐ ARE SAME ├──┘ DIFFERENT └──┴──< (PROBE AND RECONCILE) │ ?

CHECK: FOR EACH BIRTH: YEAR OF BIRTH IS RECORDED.

FOR EACH LIVING CHILD: CURRENT AGE IS RECORDED.

FOR EACH DEAD CHILD: AGE AT DEATH IS RECORDED.

FOR AGE AT DEATH 12 MONTHS OR 1 YR.: PROBE TO DETERMINE EXACT NUMBER OF MONTHS.

┌──┐ │░░│ ├──┤ │░░│ ├──┤ │░░│ ├──┤ │░░│ └──┘

Page 48 Now I would like to ask you about all your children born in the last 6 years (/ since 1998)

[Use names from 212-216, please fill out form for each respective child]

Name of child, explanation for absence of health card

Type of Antenatal Care / Delivery Place of Delivery Size of baby 1. Doctor 1. Your home 1. Very large 2. Nurse 2. Other home 2. Larger than average 3. Midwife 3. Government Hospital/ Clinic 3. Average 4. Traditional birth attendant 4. Private Hospital/ Clinic 4. Smaller than average 5. Relative/friend 5. Other 5. Very small 6. No one 6. Don’t know/remember 7. Other

Antenatal Care Delivery Name of Child (Please fill in from 212)

Antenatal care for preg-nancy? Y/N

Type Months into preg-nancy

Number of times

Who assisted with the delivery

Where did you give birth to […..]

Size of baby

Was baby measured at birth? Y / N

Were you given a health card? Y / N

If yes, did you lose it? Y / N If yes, skip next ques.

Do you still have it? Y / N If no, explain below

Birth weight in grams from health card

Height at birth from health card

Page 49 ________________________________________________________________________________________________________________________

Section 11b: Plot Mapping __________________________________________________________________________________

One form to be filled for each of the plots on the respondent’s plot list

North West

Comments:

VILLAGE

HHN

RESPONDENT RESPOND. #

DATE ROUND