 

 

 

7'3. 4'
("3:

‘1

52:3;

.4 u:
15.-
m

 

41%“ 1,.

 

This is to certify that the
dissertation entitled

TWO ESSAYS ON THE ECONOMICS OF THE HOUSEHOLD
OF THE DEVELOPING COUNTRIES

presented by

FIRMAN WITOELAR

has been accepted towards fulﬁllment
of the requirements for the

DOCTORAL degree in ECONOMICS

 

 

 

/ 1 Major P'rofe‘ssorTSVSignature

August 3, 2004

 

Date

MSU is an Afﬁnnative Action/Equal Opportunity Institution

--—-o-.-O-o—o-I-i—Q-I-u—I-U.< —.-‘-- ..-

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

| DATE DUE DATE DUE DATE DUE

 

JUIII l 0 2005

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRCIDatoDuopss-DJS

 

 

 

TWO ESSAYS ON THE ECONOMICS OF THE HOUSEHOLD
OF THE DEVELOPING COUNTRIES

By

Firman Witoelar

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Economics

2004

ABSTRACT

TWO ESSAYS ON THE ECONOMICS OF THE HOUSEHOLD
OF THE DEVELOPING COUNTRIES

By

F irman Witoelar

This dissertation consists of two essays on the economics of the household of the
developing countries. I use data from the Indonesia Family Life Survey to analyze
household allocation decisions and to investigate the underlying factor determining
household decision.

In Chapter 1, “Inter-household Allocations within Extended Families: Evidence
from the Indonesia Family Life Survey”, I investigate whether households that belong to
the same extended family pool their income to smooth their consumption. I use data from
two waves of the IFLS (IF LS 2-1997 and IFLS 3-2000) to test the income-pooling
hypothesis in both the static and the dynamic settings.

The ﬁndings suggest that in contradiction to the null hypothesis of income
pooling, the distribution of income between households in an extended family does affect
the distribution of their consumption. I also ﬁnd that the distribution of income changes
between them affects the distribution of consumption changes. The results stand even
after correcting for potential measurement error and endogeneity of the income variables.
However, the magnitudes of the coefﬁcient on income changes are small.

While complete risk-sharing is rejected, the results suggest some evidence of risk-
sharing within extended families. From a set of reduced form estimations, I also ﬁnd that

household consumption is affected by characteristics of other households in the same

extended families, highlighting the importance of inter-household ties in affecting
household behavior. This essay contributes to the body of literature on inter-household
risk sharing by shedding lights on the role that extended family play in household
allocation decisions.

In Chapter 2, “The Determinants of Household Division: A Case of a Developing
Country”, I investigate the underlying factors determining the probability of household
division in Indonesia. I use data from three waves of the IFLS (IFLS 1 - 1993, IFLSZ -
1997, and IFLS 3- 2000). The longitudinal nature of the IFLS data allows us to study
household division over the survey waves. Adopting the collective household model of
household division and the empirical framework developed by Foster and Rosenzweig
(2002), I estimate the probability of household division by the subsequent waves of the
survey, using household variables from an earlier wave as the explanatory variables. The
ﬁndings suggest that education variables play an important, although limited, role in
determining household division. There is evidence that higher education of household
head is associated with lower propensity of household to break up. On the other hand,
higher maximum years of schooling of other members in the household are associated
with higher probability of household division. These results, along with the ﬁnding
showing that rural households are more likely to divide, indicate that household division
in Indonesia may largely be associated with the mobility of young, more educated
members. While the empirical framework is based on a collective household model, the
results can be explained within the context of unitary household model. This essay
contributes to our understanding of household division, which in most previous studies is

treated as exogenous.

C0pyright by
FIRMAN WITOELAR
2004

This dissertation is dedicated to my parents,
my wife, Nina, and our daughter, Lakshmi.

ACKNOWLEDGEMENTS

I would like to deeply thank my main advisor and dissertation chair, Prof. John A.
Strauss, for his excellent guidance and constant support needed to complete this
dissertation. I am extremely grateful to him for his sound advice and counsel, and for
having his door always open for me throughout my graduate years at Michigan State
University. I am very fortunate to have him as a teacher and a mentor.

I would like to express my sincere gratitude to the other members of my
dissertation committee whose guidance has been indispensable: Prof. Jeff Biddle, for his
constructive and thorough suggestions, Prof. John Giles, for the discussions and
encouragement, and Prof. Richard Bernsten, for his insights and valuable comments.

I would also like to take this opportunity to thank those who helped shape my
interest to do graduate work in this ﬁeld: Prof. Mayling Oey—Gardiner of the University
of Indonesia and Prof. Mark Pitt of Brown University.

To my colleagues and dear friends at Michigan State: Mu Ren, Pungpond
Rukumnuaykit, Lebohang Lij ane, and Elan Satriawan, thank you for all your help and
ﬁ'iendship, and for making this journey much more enjoyable. I would also like to thank
my long-time friends Rezal Kusumaatmadja and Sjamsu Rahardj a for providing the
necessary distractions to keep me sane.

I am very fortunate to enjoy the tremendous support of an extended family. I wish
to thank my sister, Shanti Hasan, and her family, who always root for me in everything I
do. I am very grateful to my mother-in-law, Koesmartanti Soesetio, for her support and

encouragement. I would also like to thank my staunchest supporters: my uncle and my

vi

late aunt, Wimar and Suvatchara Witoelar, who always treat me almost like their own
son. A special thanks also goes to my late uncle, Toerki Witoelar, for his support and
interest in my work.

I wish to dedicate this dissertation to my two loving parents, Luki and Augusta
Witoelar, who always have faith in me and without whose support, none of this would
have been possible. I am also dedicating this dissertation to my wife and my very best
friend, Nina, for her continuous love, encouragement, and understanding, and to our

daughter, Lakshmi, who brings such joy and happiness to our life.

vii

TABLE OF CONTENTS

LIST OF TABLES .............................................................................................................. x
LIST OF FIGURES .......................................................................................................... xv
CHAPTER 1

INTER-HOUSEHOLD ALLOCATIONS WITHIN EXTENDED FAMILIES:
EVIDENCE FROM THE INDONESIA FAMILY LIFE SURVEY

l .1. Introduction .................................................................................................................. l
1.2. Background .................................................................................................................. 7
1.2.1 Evidence on Interhousehold Allocations in Indonesia ....................................... 7
1.2.2 Indonesia Family Life Survey ............................................................................ 9
1.2.3 Tracking Respondents in IFLS ........................................................................... 9
1.2.4 Household Structure of the IFLS Households .................................................. 13
1.3 Model and Empirical Speciﬁcation ............................................................................ 19
1.3.1 Model ................................................................................................................ 19
1.3.2 Empirical speciﬁcation of the static model ...................................................... 21
1.3.3 Dynamic Speciﬁcation ..................................................................................... 24
1.4 Data and Sample construction .................................................................................... 27
1.4.1 Sample Construction ........................................................................................ 27
1.4.2 Data .................................................................................................................. 30
1.4.3 Instrumenting Income ...................................................................................... 31
1.5. Empirical Results ....................................................................................................... 34
1.5.1 Static Speciﬁcations ......................................................................................... 34
1.5.2 Dynamic Speciﬁcation ..................................................................................... 38
1.5.3 Do Other Households’ Resources Affect Own Household’s Consumption?... 40
1.6. Conclusions ................................................................................................................ 41
TABLES ........................................................................................................................... 44
BIBLIOGRAPHY ............................................................................................................. 79
CHAPTER 2

DETERMINANTS OF HOUSEHOLD DIVISION:
A CASE FROM A DEVELOPING COUNTRY

2. 1 Introduction ............................................................................................................... 82

viii

2.2 Previous Literature on Household Division ................................................................ 87

2.3 Model ......................................................................................................................... 90
2.3.1 A Model of Household Division ..................................................................... 91
2.3.2 Parameterization of Preference and Some Comparative Static Results .......... 94
2.3.3 Empirical Strategy ........................................................................................... 97
2.3.4 Discussion ....................................................................................................... 98

2.4. The Indonesian Settings and Household Division in the IFLS ............................... 100
2.4.1 The Indonesian Settings ................................................................................ 100

Postnuptial Residence ............................................................................. 101

Inheritance ............................................................................................... 102

Marital Dissolution ................................................................................. 102

2.4.2 Descriptive Analysis of Household Division in [F LS ................................... 103

2.5 Empirical Speciﬁcation ............................................................................................ 107
Education Variables ................................................................................ 109

2.6 Results ...................................................................................................................... 111

2.7 Conclusions .............................................................................................................. 111

TABLES ........................................................................................................................ 118

BIBLIOGRAPHY ........................................................................................................... 183

ix

LIST OF TABLES

Table 1.2.1
Household Re-Contact Rates ............................................................................................ 44

Table 1.2.2
Number of Household Interviewed: Target vs. Split-off Households .............................. 45

Table 1.2.3
Relationship to the Head of the Target Households, 2000 ................................................ 45

Table 1.2.4
Descriptive Statistics: Target vs. Split-off Households, 2000 ........................................... 46

Table 1.2.5
Number of Households and Extended Families ................................................................. 47

Table 1.2.6
Current Marital Status of the Households Heads: 2000 ..................................................... 48

Table 1.4.1
Sample Sizes ...................................................................................................................... 49

Table 1.4.2
Descriptive Statistics: All Extended Families, 2000 .......................................................... 50

Table 1.4.3
Descriptive Statistics: Parent-Child Extended Families, 2000 .......................................... 51

Table 1.5.1
Estimates of the Effect of Household Own Income on Household Consumption ............. 52

Table 1.5.2
Estimates of the Effect of Changes in Household Own Income on Changes
in Household Consumption ................................................................................................ 53

Appendix Table 1.4.1
Descriptive Statistics: Parent-Son Extended Families, 2000 ............................................. 54

Appendix Table 1.4.2
Descriptive Statistics: Parent-Daughter Extended Families, 2000 .................................... 55

Appendix Table 1.5.1
Coefﬁcient on Income and Income Changes: No Instrumental Variables ........................ 56

Appendix Table 1.5.2
Static Tests: All Extended Families and Parent-Child Extended Families, 2000 .............. 57

Appendix Table 1.5.3
Static Tests, Two-Stage Least Squares, 2"d Stage: All Extended Families and Parent-
Child Extended Families, 2000 .......................................................................................... 58

Appendix Table 1.5.4
Static Tests, Two-Stage Least Squares, lst Stage: All Extended Families and Parent-
Child Extended Families, 2000 .......................................................................................... 59

Appendix Table 1.5.5
Static Tests: Parent-Son and Parent-Daughter Extended Families, 2000 .......................... 60

Appendix Table 1.5.6
Static Tests, Two-Stage Least Squares, 2nd Stage: Parent-Son and Parent-Daughter

Extended Families, 2000 .................................................................................................... 61
Appendix Table 1.5.7

Static Tests, Two-Stage Least Squares, lst Stage: Parent-Son and Parent-Daughter
Extended Families, 2000 .................................................................................................... 62

Appendix Table 1.5.8
Static Tests, Two-Stage Least Squares with Community Dummy Variables, 2nd Stage:
All Extended Families and Parent-Child Extended Families, 2000 .................................. 63

Appendix Table 1.5.9
Static Tests, Two-Stage Least Squares with Community Dummy Variables, lst Stage:
All Extended Families and Parent-Child Extended Families, 2000 .................................. 64

Appendix Table 1.5.10
Dynamic Tests: All Extended Families and Parent-Child Extended Families, 2000 ........ 65

Appendix Table 1.5.11
Dynamic Tests, Two—Stage Least Squares, 2nd Stage: All Extended Families ................. 66

Appendix Table 1.5.12
Dynamic Tests, Two-Stage Least Squares, 1St Stage: All Extended Families ................... 67

Appendix Table 1.5.13
Dynamic Tests, Two-Stage Least Squares, 2nd Stage: Parent-Child Extended Families ..68

Appendix Table 1.5.14
Dynamic Tests, Two-Stage Least Squares, 1st Stage: Parent-Child Extended Families ..69

xi

Appendix Table 1.5.15
Do Other Households’ Resources Affect Own Household’s Consumption ?

Two Stage Least Squares, 2nd Stage .................................................................................. 72
Appendix Table 1.5.16

Do Other Households’ Resources Affect Own Household’s Consumption ?

Two Stage Least Squares, 1St Stage ................................................................................... 74

Appendix Table 1.5.17
Do Other Households’ Resources Affect Own Household’s Consumption ?
Reduced Form Regression ................................................................................................. 76

Table 2.4.1
Number of Households Interviewed: 1993, 1997, and 2000 ........................................... 119

Table 2.4.2
Household Division: 1997 and 2000 ................................................................................ 119

Table 2.4.3
Percentage Household Interviewed by Types of Household Members:
1993, 1997, and 2000 ....................................................................................................... 120

Table 2.4.4a
Household Headship Status in 1997 of 1993 Household Member, by Sex ..................... 121

Table 2.4.4b
Household Headship Status in 2000 of 1993 Household Member, by Sex ..................... 121

Table 2.4.4c
Household Headship Status in 2000 of 1997 Household Member, by Sex ..................... 122

Table 2.4.5a
Household Headship Status in 1997 of 1993 Household Member,
by Sex and Age Group ..................................................................................................... 124

Table 2.4.5b
Household Headship Status in 1997 of 1993 Household Member,
by Sex and Age Group ..................................................................................................... 125

Table 2.4.5c
Household Headship Status in 1997 of 1993 Household Member,
by Sex and Age Group ..................................................................................................... 126

Table 2.4.6a

Reason for Leaving the Household of Household Members Not Found in the Target
Households in 1997, by Sex ............................................................................................ 127

xii

Table 2.4.6b
Reason for Leaving the Household of Household Members Not Found in the Target
Households in 2000, by Sex ............................................................................................ 127

Table 2.4.7
Moves by Location:
1993 Household Members Found in Any Split-off HH in 2000 ..................................... 128

Table 2.4.8
1993 Household Demographic Variables by Household Status in 2000 ......................... 129

Table 2.4.9
1993 Household Assets Variables by Household Status in 2000 .................................... 130

Table 2.5.1
Sample Sizes .................................................................................................................... 131

Table 2.5.2a
Summary Statistics: Base Year 1993, Division by 1997,
Claimant 1 and Claimant 2 .............................................................................................. 132

Table 2.5.2a
Summary Statistics: Base Year 1993, Division by 2000,
Claimant 1 and Claimant 2 .............................................................................................. 133

Table 2.5.2a
Summary Statistics: Base Year 1997, Division by 2000,
Claimant 1 and Claimant 2 .............................................................................................. 134

Table 2.5.3a
Means and Difference in Means of Key Variables of 1993 Households Between
Households that Have Divided and Households that Have Not by 1997 ........................ 135

Table 2.5.3a
Means and Difference in Means of Key Variables of 1993 Households Between
Households that Have Divided and Households that Have Not by 1997 ........................ 136

Table 2.5.3a
Means and Difference in Means of Key Variables of 1993 Households Between
Households that Have Divided and Households that Have Not by 1997 ........................ 137

Table 2.5.4

Household Education Variables, by the Number of Adults in the Household,
IF LS 1993 ........................................................................................................................ 138

xiii

Table 2.6.1

Determinant of Household Division between 1993 and 1997, Claimant l ..................... 139
Table 2.6.2

Determinant of Household Division between 1993 and 2000, Claimant 1 ..................... 141
Table 2.6.3

Determinant of Household Division between 1997 and 2000, Claimant 1 ..................... 143
Table 2.6.4

Determinant of Household Division between 1993 and 1997, Claimant 2 ..................... 145
Table 2.6.5

Determinant of Household Division between 1993 and 2000, Claimant 2 ..................... 147
Table 2.6.6

Determinant of Household Division between 1997 and 2000, Claimant 2 ......................... 149
Table 2.6.7

Probability of Household Division of Panel Households, LPM and LPM with HH Fixed
Effects, Claimant 1 ........................................................................................................... 151
Table 2.6.7

Probability of Household Division of Panel Households, LPM and LPM with HH Fixed
Effects, Claimant l ........................................................................................................... 152

Appendix Table 2.6.1
Determinant of Household Division between 1993 and 1997, Claimant 1: Urban ........... 153

Appendix Table 2.6.2

Determinant of Household Division between 1993 and 1997, Claimant 1: Rural .......... 155
Appendix Table 2.6.3

Determinant of Household Division between 1993 and 2000, Claimant 1: Urban ......... 157
Appendix Table 2.4

Determinant of Household Division between 1993 and 2000, Claimant 1: Rural .......... 159
Appendix Table 2.6.5

Determinant of Household Division between 1997 and 2000, Claimant 1: Urban ......... 161
Appendix Table 2.6.6

Determinant of Household Division between 1997 and 2000, Claimant 1: Rural .......... 163
Appendix Table 2.6.7

Determinant of Household Division between 1993 and 1997, Claimant 2: Urban ......... 165
Appendix Table 2.6.8

Determinant of Household Division between 1993 and 1997, Claimant 2: Rural .......... 167
Appendix Table 2.6.9

Determinant of Household Division between 1993 and 2000, Claimant 2: Urban ......... 169

xiv

Appendix Table 2.6.10
Determinant of Household Division between 1993 and 2000, Claimant 2: Rural .......... 171

Appendix Table 2.6.11

Determinant of Household Division between 1997 and 2000, Claimant 2: Urban ......... 173
Appendix Table 2.6.12
Determinant of Household Division between 1997 and 2000, Claimant 2: Rural .......... 175

Appendix Table 2.6.13
Determinant of Household Division between 1993 and 1997, Claimants are
Household Heads, Their Sons, and Brothers Aged 15 or Above in 1993 ....................... 17 7

Appendix Table 2.6.14

Determinant of Household Division between 1993 and 2000, Claimants are

Household Heads, Their Sons, and Brothers Aged 12 or Above in 1993 ...................... 179
Appendix Table 2.6.15

Determinant of Household Division between 1997 and 2000, Claimants are

Household Heads, Their Sons, and Brothers Aged 16 or Above in 1997 ...................... 181
LIST OF FIGURES

Figure 2.4.1
Relationship to Household Head by Age, Male and Female, 1993 and 2000 ................. 123

XV

CHAPTER 1
INTER-HOUSEHOLD ALLOCATION S WITHIN EXTENDED FAMILY:

EVIDENCE FROM THE INDONESIA FAMILY LIFE SURVEY

1.1. Introduction

Households break-up over time for several reasons such as members migrating to
other villages or cities to ﬁnd jobs, adult children leaving to form new households, or
marriage dissolutions. However, households with familial links may still have economic
ties with each other. For example, between these households there may be transfers of
income, exchanges of giﬁs, or informal loans provided by one household to another.
These inter-household transactions may be motivated by altruistic feelings of the
households in the extended families toward each other. Parents may transfer income to
their child’s household because they derive utility from their child’s consumption. But
the transfers may also be motivated by self-interest: parents may provide transfers to their
child in anticipation of receiving old age support from their child. In any case, one
household’s resource allocation decision may affect and be affected by allocation
decisions of other households within the extended family.

While there have been many studies on intra-household allocations in developing
countries, there are still few studies focusing on the role that an extended family plays in
a household’s allocation decisions. This essay focuses on this issue, and in particular asks
whether or not extended family provide a means for households to smooth their
consumption.

In the absence of complete ﬁnancial and insurance markets, households may be

involved in informal arrangements with each other in order to smooth their
consumptions.1 Previous studies on inter-household allocations as consumption
smoothing mechanism have focused on various links through which the mechanism
works. Many studies on consumption smoothing focus on how households in a
geographic location insure themselves against consumption risk face by the community.
In a study on villages in southern India, Townsend (1994) argues that households within
a village can make informal arrangements using local institutions to mitigate risks from
uncertainty faced by an agricultural economy.2 But geographical proximity may not be
the only grounds for informal arrangements. Households may also be involved in inter-
household allocation with relatives or members of the extended families living in
different villages or regions. Indeed, pooling resources with members of the extended
family living in a different village may protect the household from village-speciﬁc
economic shocks. Rosenzweig and Stark (1989) study the practice in rural India of
marrying daughters off to households living in different geographic locations. They ﬁnd
evidence that the marriage cum migration patterns plays a role in reducing household
consumption variability. A study by Grimard (1997) on households in Cote d’Ivoire
focuses on consumption smoothing between households with the same ethnicity,
allowing for members of extended families to reside in different regions. The study

shows some evidence of partial insurance performed by individual household with the

 

' There are of course mechanisms other than inter-household arrangement that households can use to limit
their consumption risks in the absence of complete ﬁnancial and insurance markets. For example,
households may adjust their labor supply, deplete their non-ﬁnancial assets, or withdraw their children
from school.

2 Using household data from the three villages sampled by the International Crops Research Institute for the
Semi-Arid Tropics (ICRISAT), Townsend (1994) found that controlling for village consumption,
household consumptions are not affected by contemporaneous changes in own income as well as other
idiosyncratic shock. See also Ravallion and Chauduri (1997) for a very closely related work using data
from the same ICRISAT villages.

 

members of the same ethnic group living across different geographic locations. However,
the study can only identify the ethnic group, not the particular lineage that the households
belong to.

In different social settings, using ethnic group as the “insurance group” may not
be appropriate. Closer relationship such as family ties between members of an extended
family rather than ethnicity or geographical proximity may be a more important factor on
which households base their informal arrangements. Altonji, Hayashi, and Kotlikoff
(1992) investigate whether or not households in an extended family in the United States
smooth their consumption. Using data from several waves of Panel Study of Income
Dynamics, they reject the null hypothesis of dynastic altruism among families in the
sample. They ﬁnd that at a point in time, the distribution of consumption between parents
and children is affected by the distribution of their income. They also ﬁnd that changes in
distribution of income within extended family affect changes in the distribution of
consumption.

It is important to note that Altonji et a1 (1992) are looking at extended families in
the United States. There are several reasons why focusing on households in developing
countries might produce different results. Households in these countries face very
different risk environments from their counterparts in developed countries. The majority
of the households depend on the agricultural sector, where variability in income is high.
As has already been mentioned above, the absence of complete ﬁnancial and insurance
markets may cause households in these countries to rely on inter-household informal
arrangements as a way to smooth their consumption. It is therefore reasonable to believe

that extended families in developing countries may play a larger role than they do in the

developed countries.

However, focusing on extended family imposes a data requirement that is hard to
meet with most household surveys. This is especially true for household surveys from
developing countries. Many of the surveys do not purposely collect information on
households that have familial links with each other. This essay takes advantage of a
somewhat unique feature of the Indonesia Family Life Survey, namely the fact that this
longitudinal household survey tracks a large fraction household members who have
moved out of their original households and re-interviews them in their new households
(the split-0]?r households) in the follow-up surveys (I will discuss the tracking rules in
more detail in the next section). By identifying the household from which the members
originated, I can identify the households that have family ties and deﬁne the extended
families.

Using the information on the linked households, I adopt the approach used by
Altonji et a1 (1992). Speciﬁcally, I test whether households within an extended family
pool their income to smooth their consumption, using data from two waves of the IFLS -
IFLSZ (1997) and [FLS3 (2000). These two waves include an important period:
Indonesia was hit by a ﬁnancial crisis that started in 1997 and reached its peak in mid-
1998. How the crisis has affected the welfare of Indonesian households has been and still
is an important and interesting subject. This essay contributes towards our understanding
of the dynamics of household behavior during a period of economic crisis.

This essay is also motivated by the question of how to take advantage of the
longitudinal household surveys that interview original as well as split-off households.

Collecting information from the split-off households in addition to the original

households helps to reduce sample attrition, a problem that is faced by all longitudinal
surveys. However, deﬁning what constitutes a household in a panel for the purpose of
economic analysis then become a question, since analysis using panel households that
consists of only the original households may be biased to the extent that households
break-up non-randomly. In addition, using a panel of original household is also
problematic because the rules used by surveys to deﬁne “original” and “split-off”
households are often designed for ease in the ﬁeldwork rather than being based on some
analytical underpinnings.3 This, coupled with the concern that dropping split-off
households may non-randomly exclude particular subgroups of the sample, make the
option of creating a panel of only the original households unappealing. On the other
hand, some economists choose to deﬁne the panel household by treating an original
household and its split-off households as a single extended family. This approach,
however, implies that the extended family acts as if it were a single household, or to put it
differently, this approach assumes that household decisions are made at the extended
family level.4

In this essay, I am particularly interested in studying household allocation
decisions. Speciﬁcally, I am interested in looking at changes in household consumption
over the period that was covered by the two waves of the survey. Analyzing. household
consumption and income while treating original and split-off households as extended
families amounts to assuming that households within the extended family pool their

income. A test for income pooling may then help us to judge whether household

 

3 I discuss brieﬂy the rule established in IFLSZ to assign “origin” and “split-oft” households in section two.
The same rule was used in IFL82+ and IFLS3. See Frankenberg and Thomas (2000) for full documentation
of IFLSZ.

4 In analyzing changes in household outcomes in Indonesia between 1997 and 1998, Frankenberg, Thomas,

allocation decisions are indeed made at the extended—family level and thus whether
analysis of household consumption at the level of extended family is justiﬁed.

The ﬁndings show some evidence against income pooling within extended
families among the IFLS households, both in 1997 and in 2000. To control for the
potential measurement error and endogeneity of income, I estimate the models using
instrument variables to instrument income. The distribution of income matters for the
distribution of consumption within an extended family even after controlling for
extended-family ﬁxed-effects. I then estimate the ﬁrst-difference version of the model to
control for the possibility that there are household-speciﬁc ﬁxed-effects that are
correlated with income. As in the static tests, I use instrumental variables estimation to
correct for the potential measurement error and endogeneity of income changes. The
dynamic tests return estimates of coefﬁcients on income changes that are statistically
different from zero, even after controlling for extended-family ﬁxed-effects. However,
the magnitudes of these coefﬁcients are small.

While the rejection of full risk sharing is in itself not a surprising ﬁnding,
especially in the light of similar results found in studies in other developing countries, the
small magnitude of the coefﬁcients suggest that risk-sharing within extended families
maybe important among Indonesian households, even though full risk-sharing do not
occur. The results from estimating a set of reduced form regressions, using variables from
other households in the same extended family to explain own household consumption,
suggest that variables from other households in the same extended family do indeed play

a role, although limited, in inﬂuencing household consumption.

 

and Beegle (1999) only look at the households that were interviewed in both waves.

The essay is organized as follows. The next section brieﬂy reviews some evidence
on inter-household transfers in Indonesia. The section also provides a brief background of
the IFLS. I also discuss the composition of households that constitute the sample in this
section. Section three discusses the model used in the estimation. The sample
construction and the data used in the estimation are discussed in section four. Section ﬁve

contains the estimation results, and I conclude the essay in section six.

1.2. Background
1.2.1 Evidence on Interhousehold Allocations in Indonesia

In the past years there have been numerous empirical studies that look at inter-
household transfers in both developed and developing countries. Altonji, et a1 (1997) use
data ﬁom the PSID to look at inter- generational transfers and test whether inter-vivos
transfers from parents to child are motivated by altruism. In another study, Hayashi,
Altonji, and Kotlikoff (1996) tests whether there is complete risk-sharing between and
within the PSII) families. They reject both inter- and intra-family full risk-sharing. Other
studies that examine distribution of resources within extended families look at data on
transfers explicitly. An example is the study by McGarry and Schoeni (1995) looking at
how transfers are distributed within extended families. Using data from the Health and
Retirement Survey they found that parents give more to their less well off children and
elderly parents.

Empirically, there is evidence that interhousehold transfers are an important

source of income for households in developing countries (Cox and Jimenez, 1990).5

 

5 The review by Cox and Jimenez (1990) of studies on transfers in developing countries reports the
percentage of households receiving transfers as well as the average transfer amount as percentage of

7

While motives for transfers could vary (e. g. altruism, self-interest motivated), evidence
have shown that transfers narrow inequality and serve as social insurance (Cox and
Jiminez, 1990). Lillard and Willis (1997) ﬁnd evidence that transfers from children to
parents are an important source of old age support among Malaysian families. More
recently Foster and Rosenzweig (2001) incorporate altruism into a model of risk sharing
under imperfect commitment to study the inter-household transfers in rural India and
Pakistan.

Interhousehold transfers are also important among Indonesian households. For
example, around 31 percent of rural households and 44 percent in rural Java give
transfers (Ravallion and Deardon, 1988), between 50 percent to 70 percent of elderly
receive transfers (Cameron and Cobb-Clark, 2001), around 44 percent of couples transfer
money to their noncoresident children and 55 percent of couples receive transfers from
adult children (Frankenberg, Lillard, and Willis, 2002). The study by Ravallion and
Dearden (1988) on interhousehold transfers in Java Indonesia suggests that transfers are
targeted to disadvantaged households. Levine and Kevane (2000) look speciﬁcally at
transfers from parents to daughters in Indonesia to see whether parents invest less in
daughters who move away after marriage. Looking at schooling and health outcomes,
they also ﬁnd that there is no evidence that parents invest less in daughters who move
away after marriage. There also transfers in the other direction. Studying old age
support, Cameron and Cobb-Clark (2001) ﬁnd that transfers from children to parents are
not strongly related to parental need or ability of children to give to their parents. Another

recent study by Frankenberg, et a1 (2002) ﬁnd that inter-household transfers in Indonesia

 

average income. For example, 93 percent of households in nrral South India receive transfers. Transfers
account for 46 percent of the average income of the Malaysian households in the lowest income quintile.

8

are consistent with all three motives: insurance motive, exchange of money over time, as
well as repayment of educational loans.

Regardless of the motives or the direction of transfers, it is evident that inter-
household transfers play a role in household allocations in Indonesia. In this essay I do
not attempt to look at transfers directly. Instead, drawing on the evidence, the study looks
at what happens to household consumption, taking into account the inter-household ties.
To the extent that inter-household transfers help households to smooth their consumption,

I can examine whether extended families pool their resources.

1.2.2 Indonesia Family Life Survey

IFLS is a longitudinal household and community survey that collects a large
amount of information from households, including information about their consumption,
income, and assets. It also collects data from each individual on fertility, education,
health, as well as migration, and labor market variables. In addition the survey collects
information about the community and school and health facilities. The ﬁrst wave of the
sample was collected in 1993 and is representative of about 83 percent of the Indonesian
population living in 13 of the 27 provinces in the country.6 Since then there have been
two other full sample follow-ups (IF LSZ in 1997, and IFLS3 in 2000) and a follow-up of
a 25 percent sub-sample in 1998 (IF LS2+). This essay focuses on consumption and

consumption changes between 1997 and 2000, using the data from IFLS2 and IF LS3.

1.2.3 Tracking Respondents in IFLS

One of the main concerns faced by all longitudinal surveys is sample attrition.

When respondents drop out from a longitudinal survey, the survey may lose its
population representativeness. Moreover, if the non-random attrition is related to the
factor being analyzed, the sample will suffer from selectivity bias. At the survey design
level, there are many ways to deal with the problems caused by sample attrition. One
procedure is to re-weight the sample after each wave of survey to maintain the
representativeness of the sample. One of the disadvantages of re-weighting the sample is
that it requires a speciﬁc model for attrition. Some surveys “refresh” their sample after
several waves by enrolling a new set of respondents. But perhaps one of the most
important procedures is to reduce attrition from happening in the ﬁrst place by following
individuals and households when they move. Although tracking the movers will not
prevent selection from attrition, it will help reduce it.

IFLS is one of the very few surveys conducted in developing countries that track
its target respondents when they move.7 IF LS interviewers track certain respondents
when they move from location where they were last found and even if they move to areas
outside their own village. Respondent moving from the original survey location is one of
the main causes of sample attrition in other household surveys. For the longitudinal
surveys that are conducted in developed countries such as the Panel Study of Income
Dynamics (PSID) and the Health and Retirement Survey (HRS) in the US, or the British
Household Panel Surveys (BI-IPS) in the United Kingdom, ﬁnding and re-interviewing

the movers does not pose such a big problem, since transportation and telecommunication

 

6 See Frankenberg and Karoly (1995) for full documentation of IFLSl.
7 See Frankenberg and Thomas (2000) for full documentation of IFLSZ. Thomas, Smith, and Frankenberg
(2001) study the sample attrition in IF LS.

10

infrastructures are already well-developed.8 Moreover, many of these surveys do not
require face-to-face interviews with the respondents.9 Still, many surveys do not track
their respondents when they move; the PSID, HRS, and BHPS are exceptions. In
developing countries, the cost for ﬁnding and re-interviewing the movers often deemed to
be prohibitive and the movers are dropped from the sample. IFLS is indeed one of the
very few exceptions.

To determine whether a respondent has to be tracked when he moves, IF LS
employs a set of tracking rules.10 In brief, the rules are as follow. Each individual listed in
the household is assigned a status determining whether the individual has to be tracked or
not. Any individual who has a tracking status will be tracked so long as he is in one of the
13 IFLS provinces and he can be found.1 1’12 These individuals are the “target”
respondents that get tracked if they had moved from the location where they were last
interviewed. In the split-off households where these respondents are found, the spouse
and biological children were interviewed, and general information about the households
was also collected.

The tracking rules were implemented by gathering much contact information in

the previous wave, which was used together with current contacts to locate individuals.

 

8 The main cause for attrition in longitudinal surveys in developed countries is respondents’ refusal. Some
studies in survey response literature have tried to model survey non-response explicitly. For example, Hill
and Willis (2001) developed a model of survey response decision process using data from the HRS and
conclude: (1) length of interview does not affect refusal for the next wave, (2) assigning same interviewer
reduce refusal rate.

9 Phone-based interviews account for the majority of interviews in PSID since 1973, and HRS since 1992.
The BHPS administer questionnaires and self-completion surveys.

‘0 See Frankenberg and Thomas (2000) for full documentation of IFLSZ.

” In IFLSZ and IF LSZ+, the tracking status is given to: (1) all of the individuals who were interviewed in
1993, (2) all members of 1993 households who were born before 1967 (including those who were not
interviewed in 1993). See Frankenberg and Thomas (2000) for details.

'2 In IFLS3, additional tracking rules were employed. In addition to rules (1) and (2) above, the following
individuals were also given tracking status: (3) all children born after 1993 to the parents who were 1993
household members, and (4) a random sample of other 1993 household members who were born after

11

These tracking procedures in IFLS explain why the survey has a very low attrition rate,
even compared to surveys conducted in developed countries.13 At the baseline in 1993,
7,224 households were interviewed. This number represents 93 percent of the total
original target sample of 7,330 households. IF L82, which was conducted in 1997 had a
recontact rate of 93.4 percent (see Table 1.2.1) as 6,752 original households as well as
877 split-off households were recontacted.

The IF LS3 that was conducted in 2000 managed to recontact 94.7 percent of the
target households that consisted of all of the original 1993 households plus split-off
households formed in 1997 and 1998. Some of the households that were not found in
1997 (IF L82) and 1998 (IFLS2+) were found and reinterviewed in 2000. In addition,
2,645 new split-off households were contacted in 2000. Compared to the surveys done in
developing countries, the recontact rates are among the highest, if not the highest. 14

In addition to reducing sample attrition, tracking the respondents proves to have
additional scientiﬁc value. Thomas et a1 (2001), investigate the attrition and the follow-up
in the IFLS, using IFLSl , IF L82, and IF L82+. Using household-level as well as
community-level variables of the households in 1997, they estimate a multinomial logit

model with the following outcomes: households that did not move from the baseline

 

1967. See Strauss, et al (2004) for full documentation of IFLS}.

'3 Thomas, et al (2001) provides a corrrparison of attrition rate between IFLS and other longitudinal
surveys. The PSID interviewed 78 percent of the target households at the baseline survey in 1968. The
recontact rate in the following year was 88.1 percent, and the year after 86 percent. The HRS, interviewed
81.6 percent of the target households at the baseline, and by the second follow-up survey, the cumulative
recontact rate was 83.7 percent.

" Thomas, et al (2001) also discusses the recontact rates of some of the longitudinal household surveys
conducted in developing countries. The Cebu Longitudinal Health Survey interviewed around 3,600
women between 1983-1984 in one province of the Philippines. In this two-year window, 17 percent
respondents were lost because of out-migration. By the second follow-up survey in 1991-92, only around
67 percent of the original respondents were interviewed. The World Bank’s Living Standard Measurement
Survey in Peru collected information from 1,280 dwellings in Lima in 1985-86. The follow-up survey in
1990 collect information from respondents living in the same dwelling; less than 55 percent of the
households in the ﬁrst round were interviewed.

12

survey, households that move locally, as well as households that move to areas outside
the original locality. They ﬁnd that local movers are very similar to the households that
stay at the baseline locations. Households that move considerably far from the original
location are very similar in many observable characteristics to those not interviewed in
the follow-up survey. This suggests that tracking these long-distance movers may provide
valuable information about households that are not found. They argue that the
information content of these movers is high and that tracking them is a very worthwhile

investment.

1.2.4 Household Structure of the IFLS Households

This section discusses the structure and the characteristics of the households that
constitute the sample. Who are the split-offs and how are their households different ﬁ'om
the old households? This information is essential because it may tell us whether or not the
test of income pooling within an extended family is plausible. In particular, it is important
to know how big the ﬁaction of all extended families actually represent inter-generational
(i.e., parent-child relationships) linkages. Since the model is derived from a household
model where a parent is altruistic towards his child, it is this inter-generational
relationship that I am mostly interested in. Another concern is that a lot of the
households that split might did so as a result of divorce, or marital separation.15 In this

case, the altruistic linkage between households may not be a plausible assumption.

 

'5 In this essay, family formation and dissolution are assumed to be exogenous. For a literature review on
models that treat family formation and dissolution as results of individuals’ decision making, see, for
example Weiss (1997) and Bergstorrn (1996). Foster and Rosenzweig (2002) formulate and estimate a
structural model of household division. They look at how household size and intra-household allocation
interact with exogenous income growth affect which households divide, division of assets among
households, amount of household public goods consumed and evolution of incomes of the new

13

I deﬁne an extended family as the set of households that originate from the same
1993 households. These households can be identiﬁed by looking at the household
identiﬁcation numbers in the data set. A target household is a household that was
interviewed in any prior wave of the survey. In 1997, the target households were the 1993
original households. Target households in 2000 include original 1993 households, 1997
split-off households, and 1998 split-off households. A split-oﬁ' household consists of an
individual or group of individuals who moved out from the one of the 1993 original
households to form a new household.

Table 1.2.2 shows the number of households and extended family interviewed in
IFLS], IFLSZ and IFLS3. The number of households that were interviewed in 1993 is
7,224. In the follow-up surveys in 1997, a total number of 7,619 households were re-
interviewed. This number includes both 1993 origin households as well as the households
that split-off by 1997. The nmnber of origin households interviewed in 1997 was actually
6,742 (93.3 percent of 1993 households). The other 877 households were split-off
households spawned from 791 original households. In 2000, the number of households
interviewed was 10,435, which came from 6,774 extended families (93.8 percent of 1993
households). Out of the 10,435 households interviewed, 7,790 were target households
and 2,645 were new 2000 split-off households.

The rule used in the survey to assign which households are original and which are
split-off households turns out to be somewhat arbitrary. At the ﬁrst point of ﬁeld contact
with any 1993 household member, the household where the individual was found was

assigned to be the original household.16 In practice, this will be the household living at

 

conﬁgurations of households.
'6 See Frankenberg and Thomas (2000).

14

the same address as the previous wave. This “ﬁrst-contact” rule has the advantage of
ensuring at least some information on all target household members was gathered, and it
also minimizes the risk of losing information of whereabouts of other 1993 household
members. However, the rule also results in a great deal of arbitrariness, in that the split-
off household may retain more of the household characteristics of the target household
from the previous survey than the household that is deﬁned in the current survey year as
the target household. It is not clear how one can deﬁne whether a household is still the
“same” household in different survey year. Table 1.2.3 may help illustrate this point. The
table shows the relationship of the respondents to the head of household in the 2000
target households and whether or not the respondents were new members of the
household. Note that there are 330 household heads that are new household members
(note that they may or may not be respondents from earlier surveys, “new” refers to
his/her membership in the household during the current survey). These cases may
include instances where a respondent joined the household by marrying the head of the
household and then became head of the household. They may also include instances
where a child returned to her parents’ household to assume the responsibility of the
household. Such examples increase the concern that the “target” household being
observed is not the same household as the original target after all. Then the question that
arises when one wants to restrict the analysis on only the panel of the original households
is: are the 'correct' households being chosen?

While analysis using only the panel of original households may suffer from the
fact that those households may not be the “same” households, a potentially more serious

problem comes from the fact that split-off households may have very different

15

characteristics than target households. Table 1.2.4 shows the descriptive statistics of
some of the economic and demographic variables of the 2000 target and 2000 new split-
off households. Total real income and real expenditure of the main households are higher
than those of split-off households. Household size of the split-off households tends to be
smaller. Per capita expenditure, which is a very common measure of welfare, is higher
for the split-off households than the target households. The same is true for per capita
income. The proportions of adult members aged 15-59 years are very similar in each
group, however the proportion of elderly is higher in the main households. On average,
the heads of the split-off households are younger, better educated. The maximum years of
education are also higher in the split-off households. The proportion of split-off
households residing in urban areas is higher. In short, the split-off households have
different characteristics from the target households, suggesting that household break-ups
may have occurred non-randomly. Analysis excluding the split-off households will suffer
from selection bias.

How many of the split-off households are really formed by children leaving their
parents’ household? Table 1.2.5 shows the number of extended families with multiple
sub-households, parent-child extended families, parent—son extended families, and parent-
daughter extended families. I deﬁne parent-child extended family as an extended family
in which there is at least one parent-child relationship between individuals in different
sub-households. By this deﬁnition it is possible that a sub-household can have someone
identiﬁed as “parent” and “child. In fact in some cases, one individual is both a parent
and a child. Using the similar approach I deﬁne the sample of parent-son household and

parent-daughter household to see whether there are differences between these two

16

samples. 17 A parent-son (daughter) extended family is an extended family in which there
is at least one parent-son (daughter) relationship between individuals in different sub-
households. By this deﬁnition a parent-son extended family can also be a parent-
daughter extended family. I describe in more detail how these samples are constructed in
section 4.

In 1997, there are 791 extended families with multiple sub-household of which
653 were parent-child extended families (82 percent). There are 287 extended families
with parent-son relationships and 388 extended families with parent-daughter
relationships. By 2000, there are 2,610 extended families with multiple sub-households.
Around 83 percent of them (2,176 extended families) are parent-child extended families,
48 percent have parent-son relationships, and around 52 percent have parent-daughter
relationships.

If a household split into two as a result of marriage dissolution, one may question
whether it is still plausible to think that the households have any altruistic linkage. For
cases of divorce where no children are present, altruistic behavior between the
households may indeed seem to be unrealistic. On the other hand, with the presence of
children, the divorced parents may still pool resources in order to improve their children's

welfare. If this is the case, some pooling of resources can still be observed although it

 

’7 Looking at parent-son and parent-daughter extended family separately may be of particular interest.
Parents may behave differently towards their son’s household and their daughter’s household due to factors
such as local norms, traditions, and other social institutions. For example, transfer behavior may depend on
where the adult children reside after marriage. Levine and Kevane (2003) investigate, in the context of
Indonesia the variations in residence after marriage based on the information on the local norms and
traditional law that applied in each community. Using the 1993 and 1997 waves of the IFLS (IFLSl and
IFLSZ), they ﬁnd that there is a lot of variations between communities: daughters tend to reside with or
near their parents in around 53 percent of the regions, with or near their husband’s parents in 23 percent of
the region, and in about 23 percent communities new couples tend to live with or near either set of parents.

17

might not necessarily be motivated by altruism.18

Table 1.2.6 shows the current marital status of the head of households in 2000 in
the target households and the new split-off households. Only about 2 percent of head of
the households in the split-off households were either divorced or separated. The
percentages among the target households were similar (2.9 percent).19 The low
percentage of the heads of split-off households that were divorced or separated help to
support the case that marital break-ups do not seem to play a major role in the spawning
of new split-off households in the data. However, it is important to note that the table
only shows the current marital status of the respondents at the time of the survey.
Therefore it does not tell us whether or not the household split because of a change in
marital status. Also, split-off households headed by divorced people may still be
economically related to origin households, for example if the origin household is the
parents' household.

The discussion about the household structure above can be concluded as follows.
Split-off households account for a large fraction of households in the sample and they are
indeed different from the original households. There is also some degree of arbitrariness
in deﬁning which households are "original" and which are "split-off". These facts
suggest that analyzing a panel of only the original households may not be appropriate and

looking at panel of extended families seem to be preferable. Moreover, the data shows

 

'8 In reality, parents with no custodial rights over the children often make inadequate transfers to the ones
with custody. To explain this, Weiss and Willis (1985) modeled children as collective consumption goods
within marriage, and they argued that altruism within marriage serves to overcome the “free—rider” problem
of the provision of public goods. They showed that, after a divorce, the non-custodial parent may lose
control over the allocation decisions of the custodial parent and therefore chooses to make inadequate
payment or even to stop making payment at all. This suggests that we need to pay attention to the
pervasiveness of divorces and marital separation among the households in the sample.

9 Indonesia used to have a very high rate of divorce: 13 per 1,000 population age 15 and above in 1960,
compared to 1.8 in developed countries in the same period (Jones [1994]; p. 180). However, the rate has

18

that inter-generational relationships account for most of the relationships between the
original and the split-off households. The data suggest about the fact that marital
dissolutions are not an important factor in the sample. All these facts seem to work in the
favor of treating the extended families as a unitary household. It seems plausible to

hypothesize about altruism linkages within extended families in the data.

1.3 Model and Empirical Specification
1.3.1 Model

In this essay I do not look at transfers directly. Rather, I follow the strategy used
by Altonji, et al (1992). Borrowing from the literatures on testing the dynastic nature of
households and the closely related intra-household allocation literature, they look at
parent-children dynasties in the PSID to test the hypothesis of extended family altruism.
They investigate whether or not the distribution of consumption between parent and
children households is affected by the distribution of their income. They argued that if
parents and children were altruistically linked, then the distribution of consumption
would be independent of the distribution of income.

The model is similar to that of intra-household allocations where parent’s utility
depends not only of his/her consumption but also on his/her child’s consumption.20
Parent and child behave as if their consumption is based on a unitary budget constraint. In
the context of extended family, we can think of the model in terms of household of the

head of the extended family (e.g., household of the parents) and other sub-households in

 

decline to 4.6 by 1975 and 1.1 by 1990

2° For review on the subject of intra-household allocations, see for example, the volume edited by Haddad,
Hoddinot, and Alderman (1997). Strauss, Mwabu, and Beegle (2000) review the theories and empirical
evidence on the subject. See also Thomas (1990, 1993, 1994).

19

the extended family (e.g., households of their children) operating on a unitary, extended-
household budget constraint.
The parent’s utility maximization problem is given by:

(1) Uh =9hU(Ch)+9kU(Ck)

(2) S.t. Ch.ph +Ck.pk :Rh +Rk
where c is quantity of good consumed, p is price, R is resources, and h, k stands for

parent and child respectively. The parameter (9;, and 6;, is the weight attached by the

parent to his utility and on the utility of his child. Parent and child may face different
prices, for example, if they reside in different communities. The parent can transfer some
resources T to the child so that ck . pk = Rk + T and ch . p}. = Rh — T. If the child takes T
as given, then the parent will maximize his utility over his own consumption and transfer.
The ability to make transfers is the key in this model; it is what results in a uniﬁed budget
constraint. This is a typical model that can be found in intra-household allocation
literature (e. g., Thomas 1990).

The ﬁrst order condition of the maximization problem above is:

tamer) = 6mm) = ,1
Ph Pk

 

(3)

where A is the marginal utility of income. Suppose now that the utility function is of the
form U(c) = c ['7/ (1-7), then, from the ﬁrst order conditions, the following can be

obtained:

(4) log ch = {ljlog A + [i] log 6;, — [-1—] log ph for the parent, and
7 7 7

20

(5) log ck = {-1—} log it + (1] log 6k — (ljlog pk for the child.
7 7 7

I add an index i to denote an extended family and error terms to have the

statistical representations of these demand functions as:

(6) log Ci}, = —[—I—]log/l,- + [—1—]log 6i}, —(—1-]logph + “ih , and

l l l

I I 1
( 7) 108011: = —[f]log 4i + [—Jlog ark — [—JIOg Pik + uik

l I l

The parameter 2,- can be interpreted as an extended family ﬁxed-effect. Since It,- is the
marginal utility of income, this model assumes that in an altruistic extended family, the
marginal utility of income is common among the extended family members. Note that
members’ own resources, Rk and R}, do not enter either of the consumption functions.
Rather, it is the extended family uniﬁed resources R that enters the consumption
equations through 2,, the marginal utility of income.

It is clear how the test works: if extended family has altruistic linkages, the
marginal utilities of income of the members are the same. In the empirical speciﬁcation
the marginal utility of income is represented by extended-household ﬁxed effect.

Controlling for this ﬁxed effect, then the parent’s income should not affect his

consumption and child’s income should not affect her consumption.

1.3.2 Empirical specification of the static model

The demand ﬁrnction resulting from the ﬁrst order conditions can be written as:

(8) Cikt = C(A'It’pikt’. Xikt) , k = 0,1,...,n" i=1,2,...,N

21

where cik, is logarithm of consumption at time t of household k , which is a member of
extended family i, pm is the price vector, and xik, represents household observable
characteristics and other variables that might affect household weights (9,, and Q. The
empirical speciﬁcation of the demand ﬁrnction can be written as:

(9) Cikt = B'Xikr + 5 Pikt '1’ “it + “ikt
where am is the error term that uncorrelated with xikt. The altruistic linkage between
households in an extended family is the common marginal utility of income (A in the
model), and it is represented by the extended families ﬁxed effect, or” in equation (9) -
thus org, represents log 2,. If all members of an extended family reside in the same
community, it is likely that they will face the same price vector pg“. The extended family
ﬁxed effect or), will then also capture the price vector. However, when some members of

the extended family live in other community, this may no longer be true. In this case I
need to add community prices as additional explanatory variables.

The income pooling test is performed by estimating the following equation.
(10) cikt = B'Xikz + W Yikt + 5 Pa: + an + “a:
where Yik, is household k’s own income. The error term uik, contains unobserved
household characteristics that may or may not be correlated with income.

I ﬁrst assume that the error term um, which is uncorrelated with xi)“, is also
uncorrelated with Y ,1“. Under the null hypothesis of the extended family altruism, the

coefﬁcient on Y W should be zero. That is, after controlling for own household

characteristics and the extended family ﬁxed effect, household own income should not

22

affect its consumption.

However, the assumption that the error term am is uncorrelated with Yik, may not
hold. Observable household characteristics erz may not fully capture the factors that
belong to (9;, and 6k. These omitted variables will end up in the am and they maybe

correlated with Y1)“. Extended family ﬁxed effect estimation only sweeps away parts of

the unobservables that are common across all sub-households, while parts that are
household-speciﬁc and vary across the sub-households will remain. One way to deal with
the problem is to ﬁnd instrumental variables for income and use 2SLS estimation. In
addition, ZSLS estimation could also help us deal with problem of measurement error in

our income variable.
Even if one fails to accept the null that the coefﬁcient on Y ”a is zero, it is still

interesting to see whether household consumption is affected by income of other
household in the extended family. For example, one could directly estimate the following

equation:
(11) Cikt = B'xikt + l” Yikt + Zj Yrjt + 5 Pikt + “it + “ikt . 1371'C
where ZYU, is the sum of logarithm of income of other households in the extended

family. Under the null hypothesis that households within an extended family do not pool
their income at all, the coefﬁcient on the other households’ income variable, y, is zero.

Again, here one also needs to worry whether the error terms are correlated with Yik, or

2 Yijt.

23

1.3.3 Dynamic Specification

Consider an extended family i, with sub-households k=1,. . ., m, facing the state of
nature 5 =1, ..., S with known probability of occurrence as. The discount factor is given
by ,6, 0 < ,6 < 1. As before, & denotes the household weight of sub-household k in the
extended family i. When the sub-households pool their resources, the maximization
program that is faced by the extended family is that of maximizing the sum of weighted
utilities:

(12) max 2299,, 2:1 3’21, ”Sure,“ )
subject to, assuming no borrowing:

(13) ZZLIth-Pkt 52:1: IRkt

For each household k in state s at time t, the ﬁrst order conditions with respect to

ck, is:

(14) 6k.,8’.7rs.U'(ck,)=/1,.pk,, or

9k U701“) =
Pkt

(15) A.

where A, =,u,/,B' 7t, and ,u, is the Lagrange multiplier on the resource constraints at time

t. 2' Note that A, is common across all sub-households k e i. Note also that this ﬁrst-order

 

2' This dynamic maximization problem is similar to the problem studied by Townsend ( 1994). The
difference there is that the weighted sum of utilities is over all individuals and over all households in the
village, resulting in common Lagrange multipliers across all individuals in the village. The maximization
problem is also similar to that discussed by Mace (1991) and Cochrane (1991). Mace (1991) uses the data
ﬁ'om the Consumer Expenditure Survey to test the hypothesis of full consurrrption insurance by regressing
the change in consunrption on aggregate consumption, changes in household income, and other explanatory
variables. Sirrrilarly, using data from the PSID, Cochrane (1991) tests for consumption insurance by
regressing changes in consumption growth on idiosyncratic variables that are argued to be exogenous to
consumers.

24

condition should hold at any time t. Assuming that the utility function is U(c) = c ['7/ (1-

}), one can take logs and solve for consumption of sub-household k and add an index i
denoting the extended family to obtain:
(16) 10g 6.1.: = {4}!th 1:: + [iJlog 9]. - (4)101; pm + uikt
r r r
Focusing on the changes in consumption, I ﬁrst-difference consumption over the two
period t and H to obtain:
(1 7)

1 I
f](108 21,-, " 108 ’I'it-I)— [—JUog Pikt - [08 Pikr-l)+ (Viki _ “ikr—l)

Because I am assuming that the household weight 9;, is time-invariant, it

108 Cikt ‘108’ cikt—l = "[

l l

disappears when I do the ﬁrst-differencing. From equation (17), it is clear that household
k’s own income change does not enter into determination of the household’s consumption
change. The extended family income change does, however, affect change in
consumption through the change in marginal utility of income. The statistical
representation of equation (18) can be written as follows:

(18) Acikt = B'Axikt “I” If APikt + Aait + Auikt
Where A Cikr = Cikr - Cikt-I , A xikr = xikr - Kin-1, A épikt =Pikt‘Pikt-l. A (1n = an - Ola—1 and A
uik, = uik, - um-) The dynamic test is performed by including the change in sub-
household k’s income, A Yik, = Yik, - Yin-1-

(19) 46m = B'Axikz + W AYikt + 5 Apr/a + 405i: + Aura

Note that equation (19) is also the ﬁrst-differenced version of the empirical

speciﬁcation given by equation (10). In the static speciﬁcation, the extended-household

25

ﬁxed effect or,- represent the log of marginal utility of income that is common across all
sub-households in extended family i. Here, Aori, represents the difference in the log of

marginal utility of income across periods. Since Aori, is independent of k, then it will be

the same across all sub—households. Controlling for these ﬁxed-effects, changes in

household own income should not affect changes in household consumption.

Household-speciﬁc factors belonging to the household weights (9;, but that are not

fully captured by Axik, are swept away by the ﬁrst-differencing, provided they are time-

invariant. This means that the test allows for the possibility that the extended families

have different — but time-invariant - preferences over the sub-households. Consider the
case where the extended families consist of a parents’ household, the son’s household,
and the daughter’s household. Suppose that the parents prefer to invest more in human

capital of the sons’ household. Then the static version of this model at time t, would be:

(20) cik, = B'xl-k, + 91/ Yik, + g pik, + a“ + 6,), + um, k = O,1,...,n,- i=1,2,...,N

where by, represents the household-speciﬁc time-invariant constant. In other words, 8,1,
can be seen as sub-household ﬁxed-effects that differ between the son’s and the
daughter’s household and that may be correlated with the son’s and daughter’s income.
Preferences towards the son’s household imply that 8,), is larger for his household than

that for his sister’s. Everything else the same, the son’s household will have higher

consumption and eamings. If the sub-household ﬁxed effect is time-invariant, the ﬁrst-
differencing will sweep 511: away, and this is what is shown in equation (1 8). To correct
for potential measurement error in income, I also employ ZSLS for the dynamic tests.

The assumption that the household weights, 6),, are time-invariant may not be true.

26

In fact, it is in contrast with results in collective household models (e. g. Chiappori, 1988).
In the collective household models, the sharing rule that governs how much each member
can spend the remaining income after the household allocates its resources on household
public goods is endogenous. Prices, total household income, and other time-varying
factors such as assets of each individual may determine the sharing rule. On the contrary,

in the example above that the extended family has an unequal but time-invariant sharing

rule, which is in favor toward the son’s household. Under our null above, 6), are time-

invariant. If 6), include time-varying household speciﬁc factors, ﬁrst differencing will not
get rid of these factors, even with extended family ﬁxed effect, and there may still be
factors that determine 6;, that are correlated with A I’m. This is another reason why

instrumenting changes in income would be helpful even after adding the extended-

household ﬁxed-effects.

1.4 Data and Sample construction
1.4.1 Sample Construction

Since the tests involve estimating extended families ﬁxed effect I need to restrict
the sample to the extended families that have more than one member households. The
identiﬁcation comes from these extended families. There may be concerns that
constructing the sample in this way would induce a selection bias. For example, there
may be unobserved variables among households that affect the decision to split and that
also affect consumption. Adding the extended-household ﬁxed effects take care of this
problem.

In addition I also create a sample of extended families consisting only of parent-

27

child households, parent-son households, as well as parent-daughter households. Below I
deﬁne the samples and describe how I generate them.22

I start with household roster in IFLSl (1993). From this roster, I can identify
parent-child relationship within each household. In the following survey (1997), if the
household split, and provided that the 1993 members were found, I can observe these
individuals, the households they were in, the households their parents/children were in, as
well as the extended families they belong to. Therefore I can link the 1997 households of
the individuals with the 1997 households that their parents/children are in, and I can
identify the linkage between these households as a parent-child linkage. In 1997 (and
also in 1998), there were new entrants to the survey: new household members whose
household memberships and relationships with other members will also be followed and
identiﬁed. By 2000, there are more split-off households, spawned not only by individuals
who were members of the 1993 roster but also by individuals who were new members in
1997 (or 1998). Using similar procedure, I identify parent-child as well as parent-son and
parent-daughter linkages among the individuals in the different households within the
extended families.

Since I deﬁne a parent-child extended family as an extended family in which
there is at least one parent-child relationship between individuals in different households,
it is possible that in some cases, a household has members identiﬁed as a parent and
another member as a son/daughter. In some cases, an individual can be a parent and a

child (for example, the individual may have a parent residing in another household and a

 

22 At this point, it is appropriate to note the problem that we can never observe a “complete” extended
family in our sample. For exanrple, one may observe the extended family consisting of a parent’s
household and a rmrried son’s household in our sample, but it is unlikely that the household of son’s
parent-in-law is also interviewed in the survey. Pooling of resources may occur within the larger extended

28

son in yet another household in the extended families).

For the ﬁrst-difference version of the test, I need households that appeared in both
survey years, 1997 and 2000. The number of households that were interviewed in 1997
and re-interviewed in 2000 is 7,107. By deﬁnition, the new 2000 split-off households
did not exist in 1997. Since including the split-off household is essential to the test, I try
to match each of the new split-off household to the household where the split-off member
was in during the 1997 interview. I ﬁrst match the split-off to the 1997 household of
origin for the panel individual who was tracked. If I cannot make a match, I then try to
match it to the 1997 household of origin of the spouse of the tracked member. If a match
still cannot be found, I try to match the child of the tracked member, and so on. For the
cases that I am able to match, I use the 1997 household variables as the 1997 household
variables for the 2000 split-off households.

After restricting the sample to those households in multiple member extended
families and also parent-child, parent-son, parent-daughter extended families, and
restricting ﬁrrther by households that can be matched with 1997 households, I end up
with samples that are shown in Table 1.4. 1. The samples used to test the static version of
the model are shown in column (2) and (5) for 1997 and 2000, respectively.23 For the
dynamic test, I use the sample shown in column (5), and match the households with 1997

households as described above.

 

family that include the parent-in—law’s household. While I acknowledge this problem, I do not attempt to
solve it in this essay.

23 For the static test using 2000 households, I also estimate the model using a larger sample, namely by not
only restricting that extended families have multiple households. The results are similar. I report the results
of the sample of 2000 households that can be matched with 1997 households so that comparison with the
dynamic results can be made.

29

1.4.2 Data

Monthly household consumption is calculated using all consumption expenditures
including durable goods. For housing expenditures, I use rental value of housing (actual
if available, imputed otherwise). The household composition variables used as
explanatory variables are household size, proportion of children age 0-5, 6-14, adult 15-
59 (male, female), 60 or above (male, female), age of the head of household. I also use a
dummy variable whether a household is a male-headed household, and whether the
household is a farm household. For the education of the household I use the maximum
years of education of adults in the household. I also use dummy variables for province
and urban residence. Community median wages for males and females were calculated
from earnings and hours worked of those who earned labor income including those who
were self-employed. I use community median prices of sugar and oil since these were the
only two prices for which data were available for the majority of the households both in
1997 and 2000. The prices were prices that the household paid for the last purchases in
the past month.

Monthly household income was calculated using labor earnings of individuals in
the household, earnings from self-employment, net sales of farm and non-farm assets,
rental income from household assets, gross sales of individual assets, and other non-labor
income excluding transfers. 1 exclude transfer income since what I want to test is whether
the extended family's resources matter to household consumption after controlling for
household income, without explicitly accounting for transfers. All monetary values are in

December 2000 prices. The descriptive statistics for each of the sample are shown in

30

Table 1.4.2-1.4.3.

1.4.3 Instrumenting Income

Income variables are notoriously hard to measure without error. 24 In particular,
the income variables may be measured with errors in the sense of classical errors-in-
variables. If they are, then the estimates of income coefﬁcients may be biased towards
zero. Failure to take into account possible measurement error will result in
underestimating the coefﬁcient on income, which I expect to be positive.

The concern seems to be well-substantiated by looking at the value of monthly
household expenditure with the value of monthly household income in the data set shown
in Table 1.4.2. For the 2000 main households, the mean value of household income is
roughly 70 percent of household expenditure, and for the split-off households the
corresponding number is about 65 percent. The numbers seem to indicate there may be
under-reporting of income in the IFLS. Under-reporting of income is certainly not unique
to the IFLS; indeed, as noted by Deaton (1997), it is the case in many surveys, including
those in industrialized countries.25

To correct for this potential problem, I use instrumental variables that are
predictive of income but that can reasonably be excluded from the consumption
regressions. Note, however, that while the instrument variables may help correct for
random measurement errors, they may not help in solving the systematic measurement

errors. For example, it may be possible that household with higher income underreport

 

2’ As noted by Deaton (1997: p.29), “. . .All of the difﬁculties of measuring consumption — irnputations,
recall bias, seasonality, long questionaiers- apply with a greater force to the measurement of income...”
25 Another problem associated with data on household income is that collecting them often affects response
rate. This problem is not shared by the IFLS: collecting data on components of household income does not

31

more than poorer households.

The ﬁrst set of IV 5 used consists of the log of value of land, farm business assets,
and non-farm business assets (all in real values).26 Farm business assets include plants,
house or building used for farm business, livestock/poultry/ ﬁsh pond, vehicles, tractor,
heavy farm equipment, and other assets used in the farm business. Non-farm business
assets include building, vehicles, and other equipment used in the non-farm business. By
using these variables as identifying IVs, I claim that these assets are predictive of income
but they are not correlated with the error term in the consumption regressions.

In addition, in some of the static speciﬁcations I use community infrastructure
variables as additional instrumental variables. In order to obtain these variables, I use
data from the Village Potential Statistics (the PODES) collected by the Central Bureau of
Statistics. The PODES contains a rich set of information of village characteristics in
Indonesia. I can link most of the villages in the IFLS with the villages in PODES.
Several variables such percent of land that is irrigated, percent of households with
electricity, and percent of households with telephone seems to be suitable candidates for
instrumental variables. These variables may indicate the availability of infrastructure at
the village level that may be correlated with household income but not with consumption,
conditional on household income. In addition, I use population density, existence of
manufacturing industry, and existence of a bank in the village as measures of the level of

development of a market economy in the villages. However, none of the NS that I obtain

 

affect the response rate (Deaton, 1997: p.30).

26 In place of log of the value of land owned, I also estimate the models using log of the area of land owned
but the ﬁrst stage regression results show that land area is only marginally signiﬁcant in predicting
household income. Moreover, the speciﬁcation using land area along with other business assets as
instruments variables does not pass the overidentiﬁcation test. The results of adding the interaction terms
between land area owned and a dummy variable indicating a farm household are similar.

32

from the PODES contribute signiﬁcantly to explaining variation in household income in
the ﬁrst stage regressions.

I also employ instrumental variables in the dynamic tests. In addition to allowing
the possibility of household speciﬁc ﬁxed-effects, the dynamic test may help us solve the
systematic measurement error problem. For example, if richer households under-report
more than poorer households, and if the measurement errors are unchanged between
survey waves, then, these errors will be differenced out. However, there is still a potential
problem of random measurement error. In addition, changes in income may also be
endogenous. To correct for these I use lagged value of land owned as well as lagged
value of non-land business assets as instrument variables. Using changes in business
assets to instrument changes in income may potentially induce additional endogeneity
into the model. Changes in income may affect investment in business assets which in turn
may be correlated with consumption changes. The lagged value of assets used as
instruments are 1997 value of land owned and 1997 values of non-land business assets of
the households. I argue that these lagged values of assets are not correlated with the error
terms in the ﬁrst-differenced consumption equation.

In addition to the 1997 lagged value of land owned, I include the change in log
value of land. My claim is that the potential problem of endogeneity resulting from using
the change in land value is substantially less than if I were to use changes in value of
other business assets. The data shows that between 1997 and 2000 there were very few
incidence of land sales, only 1.5 percent out of all land ownerships. The total value of
those sales was only about 0.5 percent out of total values of land owned. The change in

land value might be the result of investment in land such as improvement in irrigation

33

system. However, during the three-year period there was no large irrigation project that
was being canied out, at least none that I am aware of. The variation of real land values
owned between the periods is likely driven by the change in prices that occur between
1997 and 2000.

In other speciﬁcations I add interaction of changes in the median wage of males
and females with the 1997 household maximum years of education. Changes in wages
between the periods may affect households differently depending on the level of
education in the households. I also tried to add interactions of price changes with 1997
household maximum years of education. The use of the latter set of NS turns out not to

be ﬁuitful since they do not pass the over-identiﬁcation tests.

1.5. Empirical Results
1.5.1 Static Specifications

I begin by estimating the static model with and without extended-household
ﬁxed-effects for all households in 1997 and 2000. For each sample, I estimate the
consumption regressions with and without the extended families ﬁxed effect. First, I
estimate the models without instrumental variables. Next, I estimate the models using
ZSLS.

Table 1.5.1 summarizes the result of our static tests. The table reports only the
coefﬁcient on log of household income from the various speciﬁcations. Regression
results showing coefﬁcients on the other covariates are reported in Appendix Table 1.5.2-
1.5.7.

The ﬁrst panel of Table 1.5.1 shows the result from estimating equation (5) using

34

the sample of 1997 households. The second panel shows the result for the 2000
households. While the results are similar qualitatively, I focus the attention to 2000
households since split-off households consists a much larger fraction of households in
2000 than in 1997.27

The ﬁrst thing to note is that estimations without using any instrumental variables
return coefﬁcients on income that are although small in magnitude, statistically
signiﬁcant. The coefﬁcient ranges from 0.015 (parent-daughter extended families, with
ﬁxed effect) to 0.026 (parent-son extended families, without ﬁxed effect); The small
magnitudes of the coefﬁcients translate into income elasticity of consumption of 0.01 5 to
0.026. This goes in line with the suspicion that the income variables suffer from
measurement error. 28

Looking closely at the regression results in Appendix Table 1.5.2 it is clear that
most of the explanatory variables appear to be statistically signiﬁcant when I estimate the
consumption equation without extended family ﬁxed effect. Column (1) shows that the

coefﬁcient on income for the sample of household in multi-member extended families is

0.022 and it is statistically signiﬁcant at 1 percent level. Having ﬁxed effect in the

 

27 The estimates of the coefﬁcients on household income under ZSLS using the sample of parent-daughter
extended farrrilies for 1997 stood out as much greater those of other samples; 1.509 (standard error 0.494)
and 1.205 (0.646) without and with ﬁxed effect, respectively. However, the identifying instrumental
variables fail the over-identiﬁcation tests miserably; the null hypothesis that the equation is properly
speciﬁed and the instruments are valid instruments (p-values is 0.000 in each case) is rejected. The same is
also true for the sample of parent-son extended families in 1997, although the estimates are not as high as
those for parent-daughter extended families

28 I also estimate a bivariate regression using only household income as the explanatory variables (see
Appendix Table 1.5.1 for the results). As shown in the tables, the coefﬁcient on household income or its
changes gets smaller as more explanatory variables are added. One explanation is that household income is
capturing effects of other, valid, explanatory variables such as household composition and education
variables. Another explanation is that the small coefﬁcients are the results of the measurement error
problem in income that gets worse as more explanatory variables are added. The intuition is that as more
explanatory variables are added, there are less variations available in the data to explain the variation in
household consumption. The contribution of measurement error to the remaining variation becomes
proportionally larger, and so the attenuation bias becomes worse.

35

estimation does not seem to change this coefﬁcient by any signiﬁcant magnitude. Note
however that some of the community-level variable became statistically insigniﬁcant
after using the ﬁxed effect. Large fraction of the households reside in the same
community as the other households in their extended families, so the extended
household ﬁxed-effects sweep away some of the community level variables. Similar
results are obtained using the sample of parent-child extended families (column 3 and 4
in Appendix Table 1.5.2), as well as parent-son, parent-daughter extended families
(Appendix Table 1.5.5).

The summary of the results of the ZSLS estimation is also presented in Table
1.5.1. The second stage regressions for the sample of extended families with multiple
sub-households and parent-child extended families are reported in Appendix Table 1.5.3.
For 1997, only the estimates of the coefﬁcients on household income are presented in
Table 1.5. 1. The regression results are not reported but are available upon request. The
corresponding ﬁrst stage regressions are reported in Appendix Table 1.5.4.

The ﬁrst thing to note is that the coefﬁcients on household income in the ZSLS
estimations are greater than in the OLS estimations by as much as ten-fold. For the
sample of 2000 extended families, the coefﬁcient on household income under 2SLS is
0.216 (with standard error 0.031).29 Controlling for ﬁxed effect, the coefﬁcient drops to
0.135 (with standard error 0.028) — lower than without controlling for ﬁxed effect- but
much higher than in the speciﬁcation without instrumental variables (0.021). Our IVs
pass the over-identiﬁcation tests for this sample. The p-values for our Hausman tests with

the null that the variable log of household real income is exogenous are 0.000 (column 1-

 

29 Altonj i et al (1992) reports income elasticities of food consumption of around 0240-0286 for US
households. However, this is arguably not comparable to the results above since one would expect that food

36

4, Appendix Table 5.3) suggesting that instrumental variables estimations are required.

Since the model is derived from a model of parental altruism, I am particularly
interested in whether using the sample of only parent-child extended families would
produce different results. It turns out that the results are very similar (column 3 and 4 in
Appendix Table 5.3). For the parent-child extended families the coefﬁcient on income
under ZSLS but without ﬁxed effect is 0.218 (standard error 0.035) and after ﬁxed effect
it drops to 0.128 (standard error 0.044). Coefﬁcient on income before ﬁxed effect is
highest using the sample of parent-son extended families; it is 0.284 (standard error
0.062), but after ﬁxed effect it became virtually the same as from the sample of parent-
daughter extended families. Except for the ZSLS estimate with ﬁxed effect for the
sample of parent-daughter extended families, the IV 5 pass the over-identiﬁcation tests.
As noted above, the Hausman tests suggest that income is indeed endogenous.

The results thus far seem to suggest that the household's own income matters even
after controlling for the extended-household ﬁxed-effects. The estimations without
instrumental variables show that the coefﬁcients on income are small in magnitude and
almost the same with and without the ﬁxed effect. ZSLS estimations provide us with
more reasonable estimates of the coefﬁcients on income. Under ZSLS, controlling for
extended family ﬁxed effect does decrease the coefﬁcients on income signiﬁcantly
(around 40 to 60 percent decrease for 2000 households) but the coefﬁcients on income

after accounting for ﬁxed effect are still statistically signiﬁcant.30

 

in the US would be far less sensitive than all consumption in a developing country.

3° I also estimate the static regressions using community dummy variables in place of prices and wages.
The results are shown in Appendix Table 5.8 and 5.9. Identiﬁcation comes from extended families with
members living in different communities. Around 60 percent of the extended families do span across
different communities. The coefﬁcient on own income is 20 percent lower (0.203 compared to 0.256) after
controlling for extended-family ﬁxed-effects and it is statistically signiﬁcant.

37

1.5.2 Dynamic Specification

Table 1.5.3 presents the summary of results from the dynamic tests. First I ﬁrst-
difference the variables and estimate the model by OLS. The coefﬁcient on changes in
log (household real income) is positive and statistically signiﬁcant, although the
magnitude, 0.017 is very small. When I add extended family ﬁxed effect, the income
coefﬁcient is slightly greater (0.020), and it is still statistically signiﬁcant. As in the static
version, 2SLS would provide a better estimate about the effects of the changes in
household income on changes in consumption, provided the instrumental variables are
valid. The second stage regression results for all extended families and for parent-child
extended families are reported in Appendix Table 1.5.11 and 1.5.13, respectively. The
ﬁrst stage regression results for the corresponding samples are reported in Appendix
Table 1.5.12 and 1.5.14, respectively.

The F-tests for the identifying instrumental variables reported in Appendix Table
1.5. 12 and 1.5.14 seem to suggest that the instrumental variables contribute considerably
well in predicting changes in income. The ﬁrst ZSLS speciﬁcation use lagged land value
1997 as well as changes in land value between the 1997 and 2000. Using the sample of
all extended families, the coefﬁcient on income changes is 0.132 (standard error 0.049)
before adding the ﬁxed effects. After accounting for extended family ﬁxed effect, the
coefﬁcient drops to 0.059 (standard error 0.033). Similar results were obtained using the
sample of parent-child extended families. Note that while the test can reject the null that
the change in household income is endogenous in the speciﬁcation without the extended-

household ﬁxed effects, the test fail to reject the null when the ﬁxed-effects are

38

controlled for. This seems to suggest that there is no need to treat the change in

income as an endogenous variable when I include the ﬁxed-effects. Nonetheless, the
estimations appear to be well identiﬁed and the coefﬁcient after controlling for ﬁxed-
effects under 2SLS (0.059), and under OLS (0.020) are both signiﬁcantly lower than the
estimates under ZSLS before controlling for ﬁxed effects. It is reasonable to conclude that
controlling for extended family ﬁxed-effects, the effects of a change in household own
income on the change in household own consumption are small.

When I add lagged value of business assets as an additional instruments, the
results do not change much, although now the coefﬁcient on income changes after
accounting for ﬁxed effect are slightly higher. However the instrumental variables did
not pass the over-identiﬁcation test, especially after accounting for extended family ﬁxed
effect. Using changes in median wage of male and female interacted with the household
education variable in 1997 to capture different effects of wage changes on household
income depending on education level of the household, the coefﬁcient on income before
accounting for ﬁxed effects are much lower than in previous speciﬁcations. But again,
the instrumental variables perform very poorly in over-identiﬁcation tests.

The results from the dynamic speciﬁcations show that changes in distribution of
resources does affect changes in distribution of consumption among households in
extended families, suggesting that households may not fully pool their resources to cope
with economic shock they were facing. However, the coefﬁcients on the change of own

income after controlling for extended family ﬁxed-effects become small.

39

1.5.3 Do Other Households’ Resources Affect Own Household’s Consumption?

Including income of other sub-households from the same extended family directly
in consumption regression may provide some insight about the role of other households'
income in household consumption. Note that this is not a formal test of consumption-
smootlring hypothesis. The estimation will tell us whether the coefﬁcient on other
households' income is statistically signiﬁcant. If it is, then it indicates that resources of
other households do play a role in determining household consumption.

As instruments, I use own household's as well as other households' log of land
value, farm and non-farm business assets. However, the instrumental variables failed the
over-identiﬁcation tests (p-values = 0.039 and 0.0524 for the sample of extended families
with multiple sub-households and parent-child extended families, respectively. The
Hausman tests seem to suggest that other households' income is not endogenous in
household consumption equation. Overall, the results seems to show, at least in the static
context, that income of other households does not play any role in determining own
household consumption (see Appendix Table 1.5.15 and 1.5.16).

In addition, I also estimate reduced form regression of household consumption by
regressing log of household consumption not on income variables but on all of the
exogenous explanatory variables included in other speciﬁcations belonging to own
household as well as other households, and also on log value of land, farm, and non-farm
business assets of own and other households (see Appendix Table 1.5.17). The F -test of
other household variables in these regressions are 2.09 (p-value=0.001) and 2.02 (p-
value=0.001) for the sample of extended families with multiple sub-households and

parent-child extended families, respectively. This suggests that other households'

40

variables have some effects, if only small, on households' own consumption.

This result is related to the body of literature that looks at the outcome of linked
households. For example, Foster (1993) looked at the effects of household partition in
rural Bangladesh on child’s schooling. In particular the study asks whether decision on
child’s schooling depend on resources available to a particular household or to resources
available to all the linked-households as a whole. The study rejects the null hypothesis of
no effects of linked-household, although controlling for linked households’ resources,

own household resources still affect child’s schooling.

1.6. Conclusions

In this essay I have shown that there is evidence against income pooling in
extended families, both in the static and dynamic settings. The ﬁndings show that
household own income and income changes affect consumption and consumption
changes even after adding the extended family ﬁxed-effects. In terms of the magnitudes
of the income (and income changes) coefﬁcients, the results are mixed. The static tests
return estimates that range from 0.127 to 0.135 after controlling for ﬁxed effect. These
magnitudes are economically signiﬁcant and suggest that I can strongly reject the
income-pooling hypothesis. This in itself is perhaps not surprising: even within
households, income-pooling hypothesis is almost always rejected in most empirical
studies.

The more interesting results come from the dynamic tests that show that
controlling for extended family ﬁxed effects, the magnitudes of the coefﬁcient on income

change seem to be small (0.067 to 0.09), although they are statistically signiﬁcant. This

41

suggests that at least to some degree, households within an extended family do pool their
resources. The ﬁndings also suggest that although households do not fully act as a unitary
(extended) household, allocation decisions may be made at the extended-household level.
This implies that, under some conditions, looking at a panel of extended families may be
preferable to using only panel of "original" household when one wants to analyze
household consumption or income changes.

It is also important to note that pooling resources is by far not the only mechanism
available to the households to cope with economic crisis. F rankenberg, Srrrith, and
Thomas (2003), using data from IFLSZ (1997) and IF LS2+ (1998) — a shorter period of
observation- shows how households in Indonesia use a type of asset that was least
affected by the crisis, namely gold, as a way to cope with the crisis. Yet another
mechanism that may have been used by the households is to change living arrangement
(i.e., moving out of households in some cases, or joining households in other), a
household decision that I assume to be exogenous in this essay.

Perhaps one of the more important lesson to be learned from this research is the
fact that inter-household ties may be inﬂuential in shaping household allocation
decisions. Rejection against extended family income pooling does not mean that extended
families do not behave as a single-household in other dimension of household behavior.
One possible extension of this study is then to look whether and how inter-household ties
affect other household behavior such as labor market supply, home production, and
investment in human capital. The study by Foster (1993) on the effects of household
partition in rural Bangladesh on child schooling is one of the few studies that looks at the

effects of linked household resources on household outcome.

42

The analysis has also raised some interesting questions. The results suggest that
extended-household does not fully act as a unitary household. The next obvious question
would be whether inter-household allocation decisions across households within an
extended family are consistent with the collective (extended-) household model. Another
important question concerns the determinants of household break-ups and formation. In
this essay I treat them as exogenous, although family formation and dissolution are
themselves results of economic decisions. One could further the study by incorporating
the endogeneity of household division into the analysis of interhousehold ties and

household outcomes.

43

$20525:

boéim +~mqn= use mEosomso: toéim mmqn: £22395: 5:“: 08 $3283: How—S mmqn: 6808.80 mEogomso: mo 33:5: 05 E
3335 03 $3525: eoﬁo 95 3:12:82 35 2:28:02 .w=_>= Eon—:08 Eonomao: 05% Bee. an no 3:25:50 2e v.38 BSEobM

.35: .quE .59.: u8.50m

 

 

:2: 4m $3 B ES Basso £238.. 33
- £3 - - - - - - “20:38; ESE; BE
is 83 em 23 - - - - ”228:2. ewes 8H,:
cg :m o 3 - - - - maﬁasoseoeam AS":
Ea E N E - E - - £288; eoée NS”:
a? $3 2 SE is 3.6 3 £3 £288: 5H:

0 :6 o— n O 850 U :6 —
..

 

n35— uoﬁneoem 22—85:: ﬂu; 935.

44

Table 1.2.2 Number of Households Interviewed: Target vs. Split-off Households

 

 

1993 1997 2000
Households interviewed 7,224 7,619 10,435
Target households interviewed 7,224 6,742 7,790
Split-off households interviewed - 877 2,645

 

Source: IFLS], IFLSZ, IFLS3

Target households are households that were interviewed in any prior wave of the survey. IF L82
target households are IFLS] original households. IFLS3 target households are IFLSl original
households, IF LS2 split-off households and IFLSZ+ split-off households.

Table 1.2.3 Relationship to the Head of the 2000 Target Households

 

HH

 

Relationship to household New HH

members re- Total
head members

interviewed
Head 7,460 330 7,790
Spouse 5,708 277 5,985
Child,S/D-in-law 13,075 2,675 15,750
Parent,F/M-in-law 812 174 986
Sibs.,B/S-in-law 378 140 518
Other relative 1,537 1,289 2,826
Non-relative 95 174 269
Total 29,065 5,059 34,124

 

Source: IFLS3

45

Table 1.2.4. Household Characteristics Target vs. Split-off Households, 2000

 

 

 

 

 

2000 Target households 2000 Split-off Households
Number of households 7,505 2,517
Mean Std. Dev Mean Std. Dev
HH real expenditure (000 Rp) 1,031 (1,169) 979 (1,101)
HH real income (000 Rp) 727 (1,169) 641 (1,120)
Per capita real expenditure (000 Rp) 261 (301) 329 (378)
Per capita real income (000 Rp) 179 (328) 202 (363)
Household size 4.39 (2.01) 3.62 (2.10)
Number of hh members:
0-5 years 0.47 (0.68) 0.63 (0.72)
6-14 years 0.85 (0.99) 0.36 (0.73)
15-59 years, male 1.26 (0.96) 1.22 (0.94)
15-59 years, female 1.37 (0.89) 1.23 (0.90)
60+ years, male 0.19 (0.40) 0.07 (0.27)
60+ years, female 0.24 (0.45) 0.10 (0.31)
Male household head (=1) 0.82 (0.39) 0.85 (0.36)
Age of hh head 49.41 (14.10) 34.72 (13.94)
Maximum years of education 9.04 (4.24) 10.19 (3.93)
Farm households (=1) 0.41 (0.49) 0.24 (0.43)
Urban 0.46 (0.50) 0.54 (0.50)
Source: IF LS3

* After dropping observations with missing values

46

Table 1.2.5 Number of Households and Extended Families

 

 

1993 1997 2000

All extended families - 6,742 6,774
Households 7,224 7,619 10,435
Extended families with multiple households - 791 2,610
Households - 1,668 6,271
Parent-child extended families - 653 2,176
Households - l ,343 5,075
Parent-son extended families - 287 1,256
Households - 578 2,730
Parent-daughter extended families - 388 1,361
Households - 787 2,952

 

Source: IFLSl, IFLSZ, and IFLS3
See text for discussion on different deﬁnitions of extended families.

47

Table 1.2.6 Current Marital Status of the Household Heads, 2000

 

 

 

 

 

2000 “Target” Households Split-off households

Male Female Total Male Female Total
% has never married 1.5 6.5 2.4 11.3 49.4 17.4
% married 95.0 15.7 80.7 87.6 18.2 76.5
% separated 0.2 3.3 0.7 0.0 3.5 0.6
% divorced 0.5 10.0 2.2 0.3 7.3 1.4
% widow/er 2.8 64.5 13.9 0.8 21.5 4.1
Total 100.0 100.0 100.0 100.0 100.0 100.0
Number of households 63 84 1406 7790 2222 423 2645
Source: IFLS3

48

Table 1.4.1. Sample Sizes

 

 

 

1997 2000
(1) (2) (3) (4) (5)

All extended families 6,742 6,382 6,774 6,698 6,175

Households 7,619 7,152 10,435 10,022 8,351
Extended families with multiple
households 791 703 2,610 2,450 1,723

Households 1,668 1,473 6,271 5,774 3,899
Parent.child extended families 653 562 2,176 2,070 1,510

Households 1,343 1,172 5,075 4,785 3,377
Parent-son extended families 287 240 1,256 1,172 834

Households 578 495 2,730 2,546 1,785
Parent-daughter extended families 388 339 1,361 1,275 907

Households 787 696 2,952 2,771 1,933

 

See text for discussion on different deﬁnitions of extended families.

1) All 1997 households

2) After dropping households with missing observations

3) All 2000 households

4) After dropping households with missing observations

5) After dropping households that cannot be matched with 1997 households

49

Table 1.4.2 Descriptive Statistics: All Extended Families with Multiple Households, 2000

 

 

 

Number of extended families 1,723
Number of households 3,339
Mean Std. Dev
HH real expenditure (000 Rp) 983 (1,000)
HH real income (000 Rp) 672 (1,086)
Per capita real expenditure (000 Rp) 295 (332)
Per capita real income (000 Rp) 190 (330)
Household size 3.93 (2.07)
Number of hh members:
0-5 years 0.49 (0.68)
6-14 years 0.56 (0.87)
15-59 years, male 1.24 (0.97)
15-59 years, female 1.30 (0.89)
60+ years, male 0.15 (0.36)
60+ years, female 0.18 (0.40)
Male household head (=1) 0.82 (0.39)
Age of hh head 43.54 (16.23)
Maximum years of education 9.56 (4.06)
Farm households (=1) 0.32 (0.47)
Urban (=1) 0.50 (0.50)
Median wage, male (Rp) 1,888 (2,221)
Median wage, female (Rp) 1,098 (2,608)
Median prices of sugar (Rp) 3,692 (451)
Median prices of oil (Rp) 3,510 (241)
Real land value (000 Rp) 8,412 (39,100)
Real value of farm bus. assets (000 Rp) 1,302 (8,585)
Real value of non-farm bus. assets (000 Rp) 4,149 (32,500)

 

50

Table 1.4.3 Descriptive Statistics: Parent-Child Extended Families, 2000

 

 

 

 

Number of extended families 1,510
Number of households 3,3 77
Parent households Child households Parent, child
(n=1,451) (n=1672) households
(n=254)
Mean Std. Dev Mean Std. Dev Mean Std. Dev

HH real expenditure (000 Rp) 982 (947) 983 (1,038) 883 (786)
HH real income (000 Rp) 727 (1000) 618 (1,109) 630 (1,352)
Per capita real expenditure (000 Rp) 265 (288) 329 (379) 242 (244)
Per capita real income (000 Rp) 194 (288) 190 (3 64) 167 (388)
Household size 4.22 (2.06) 3.64 (2.04) 4.13 (2.02)
Number of hh members:

0-5 years 0.34 (0.64) 0.60 (0.69) 0.54 (0.68)

6-14 years 0.66 (0.95) 0.44 (0.79) 0.84 (0.95)

15-59 years, male 1.28 (0.97) 1.23 (0.94) 1.07 (1.16)

15-59 years, female 1.40 (0.89) 1.24 (0.89) 1.30 (0.79)

60+ years, male 0.27 (0.44) 0.06 (0.24) 0.13 (0.34)

60+ years, female 0.28 (0.46) 0.07 (0.27) 0.24 (0.43)
Male household head (=1) 0.82 (0.39) 0.86 (0.35) 0.59 (0.49)
Age of hh head 54.41 (11.80) 33.29 (13.23) 46.87 (14.13)
Maximum years of education 8.95 (4.28) 10.30 (3.69) 8.62 (3.92)
Farm households (=1) 0.42 (0.49) 0.24 (0.42) 0.36 (0.48)
Urban (=1) 0.47 (0.50) 0.54 (0.50) 0.46 (0.50)
Median wage, male (Rp) 1,602 (1,090) 2,178 (2,966) 1,687 (1,934)
Median wage, female (Rp) 855 (1,383) 1,266 (2,401) 1,320 (6,835)
Median prices of sugar (Rp) 3,680 (466) 3,705 (450) 3,721 (402)
Median prices of oil (Rp) 3,499 (243) 3,521 (240) 3,502 (23 8)
Real land value (000 Rp) 13,100 (48,100) 4,844 (28,900) 7,188 (40,700)
Real value of farm bus. assets (000
Rp) 1,974 (11,900) 769 (3,918) 1,504 (9,487)
Real value of non-farm bus. assets
(000 Rp) 3,805 (22,100) 4,273 £10,300) 3,151 (20,40)

 

51

an. - an. 2.3 522...... a. 8:82

0.... mud—058.5... ooom no 2953 05 .8 831830 .050 .5 353508 2: 3365 3:52 :o_mmo..wom $82228 5 Echo 8.82.3.0. 6.53.

 

 

 

...,... .888 r... 83.8 r. 88.8 r... 88.8

B... So R... E... 2.... es... 85...... Buses 5.82.8225.
2.... .358 2.... £88 1... 88.8 r... 88.8

35 3... em... as... 3.... es... m3.8a. .5285 8..-...an
r... 28.8 r... 88.8 r... 88.8 3.... 9.8.8

:3 8%. as... ...N... 8.3 2.... 8....5. 82.35 220-22....
r... 88.8 r... 28.8 r... 88.8 ...r. 88.8

8.3 an. 2.... EN... .8... S; 8.2.8.5.. 2.5.2.. 3 sass .8853

8.2.82... 83 5.3 35...... 2. ...a 2.... 8.2.0.5.... 8.."
.. .888 t... 88.8 r. 85.8 r. 88.8

8.. a... SN. 82 as... S... a...E... 885.5 .aﬁsavaas.
88.8 83.8 ...,... 28.8 r... 88.8

3.. SN as... can... .8... N8... 8.85. 328.3 83.2....
r... :38 r... €38 r... :88 r... 88.8

E... new .2... EN... 88... m8... mazes 82.26 2235......
r... 83.8 r... 88.8 r... 88.8 r... 88.8

E... m: 5.... SN... 48... 88... 8.2.3.5.. 2.5.2.. .3 ease 323$

3.2.32... 8.:
8.2.2.5: 8.....3... are”. 385 $85. age...
a 8.585 -34.... .58.... 8...... ..z 62.... 2.8.... 8...... 2

a 828...... seesaw
mum" mac

 

:etnﬁsneou ...eaemsol .8 2:3:— EpO 22.8.5: no “ooh”.— 95 he 8:35am —.m.— 03:.

52

.32 5.80.60 .0 200% 2......er 5.3 0.020.. .225 Emma .00. 5.00:. e. 0028...“. ..o 82.028... .0.....> 2....
.00. e. 82.2.0 .32 80mm... 8058... ..o we. .32 0.....> .05. ..o we. ”0.8 2.0.8030. 0%.... 2.008 0... ... v0.5.0... .2. 830...?» .85....50... Am.

.02.... 2.2

.00. a. 32.2.0 .32 8080 805...... .0 we. .32 0......» .05.. ..o we. 0.... 25.3030. 02% 2.000.. 0... :. 80.5.0... .2. 0030...; 3:08.53. AN.
.0.....> 2.... :. 0.02.8.0 2... 32 0.....> 2.0. ..o we. .08 28.8030. 09.: 2.000.. 0... :. 00.5.0... .2. 330...? 3:08.58. C.
v. .m.. -o. .m. 0.2.... 022.02.... a. 80:80. 0... 3.0.850 .050 .8 8.8.0508 2.35% 8.3.0. 5.00030... 882.308.. E 220 0.8.0530. .253.

r... 88.8 ....8...... r... :88 £1898 ...,..88 r. .888
E... 83 .8... 8.8 88... 8.... 8.8 8....

_. 88.8 :LN88 ...,. :88 $158.8 .. £8.81... 8.88

0...... £598 .12.. 9598
as... as... $2.80..
80.2.8.8 220-208..

1.... .888 r... .888

 

 

3w.m Mme; 2A.... 35... m3... em . .c 33... mm. .c owe... 5 5... 0.0.2.0022.
0......2: 5.3
00...:8. 30:02.2
H.32. 00.....5... BREW 300.....— 300hm 500....— ..»ng 800....H 300....H magnum
.025: t 60.2.023— 62....— uon... e7. .6052 60.....— 07. 68.3 63.... e7. 60...... 2.8...— eoura e7.
3. .52....— haﬁam 2.8....— 4.0....0umm
Ive—503nm Ito—ESNH 4.0.2.0....—
m N — actuated
502.00....9 «0......—

 

wdmu 62......th 30:20:2— 3.....—

..e.........m..eO 3300...... ... Buaaao 0... ..o 0:50..— ..BO 32.028! 2 89:25 no «gum 0... we 8.0:...0H ..o.30..u0~— «.mA 2......

53

Appendix Table 1.4.1 Descriptive Statistics: Parent-Son Extended Families, 2000

 

 

 

 

Number of extended families 834
Number of households 1’785
Parent Son households Parent, son
households(n=780) (n=812) households (n=19ZQ
Mean Std. Dev Mean Std. Dev Mean Std. Dev
HH real expenditure (000 Rp) 973 (978) 1,003 (1,033) 914 (832)
HH real income (000 Rp) 706 (957) 629 (1,127) 611 (1,071)
Per capita real expenditure (000
Rp) 273 (315) 343 (418) 248 (266)
Per capita real income (000 Rp) 197 (289) 192 (3 80) 156 (232)
Household size 4.11 (2.10) 3.75 (2.22) 4.23 (2.15)
Number of hh members:
0-5 years 0.36 (0.64) 0.58 (0.70) 0.52 (0.69)
6-14 years 0.61 (0.91) 0.45 (0.79) 0.87 (0.98)
15-59 years, male 1.22 (0.94) 1.51 (1.04) 1.15 (1.25)
15-59 years, female 1.34 (0.91) 1.08 (0.90) 1.28 (0.82)
60+ years, male 0.28 (0.45) 0.05 (0.23) 0.15 (0.35)
60+ years, female 0.30 (0.48) 0.07 (0.26) 0.26 (0.44)
Male household head (=1) 0.80 (0.40) 0.94 (0.23) 0.59 (0.49)
Age ofhh head 55.05 (12.01) 33.12 (13.04) 48.08 (14.39)
Maximum years of education 8.93 (4.30) 10.29 (3.60) 8.84 (3.81)
Urban (=1) 0.48 (0.50) 0.54 (0.50) 0.48 (0.50)
Farm households (=1) 0.39 (0.49) 0.23 (0.42) 0.35 (0.48)
Median wage, male (Rp) 1,646 (1,078) 2,062 (2,356) 1,736 (2,158)
Median wage, female (Rp) 936 (1,718) 1,226 (1,93 8) 1,508 (7,829)
Median prices of sugar (Rp) 3,670 (413) 3,709 (426) 3,721 (349)
Median prices of oil (Rp) 3,500 (248) 3,526 (242) 3,504 (247)
Real land value (000 Rp) 12,600 (48,100) 3,908 (25,600) 8,774 (46,500)
Real value of farm bus. assets
(000 Rp) 2,370 (14,900) 812 (3,956) 1,861 (10,900)
Real value of non-farm bus. .
assets (000 Rp) 4,624 (27,000) 3,554 (26,200) 2,985 (17,800)

 

54

Appendix Table 1.4.2 Parent-Daughter Extended Families, 2000

 

 

 

 

Number of extended families 907
Number of households 1,933
Parent Daughter households Parent, daughter
households(n=85 3) (n=893) households (n=187)
Mean Std. Dev Mean Std. Dev Mean Std. Dev
HH real expenditure (000 Rp) 961 (867) 963 (1,023) 866 (784)
HH real income (000 Rp) 734 (1,014) 610 (1,072) 691 (1,522)
Per capita real expenditure (000
Rp) 256 (243) 3 10 (337) 232 (204)
Per capita real income (000 Rp) 194 (282) 187 (345) 180 (437)
Household size 4.27 (2.12) 3.60 (1.88) 4.13 (2.03)
Number of hh members:
0-5 years 0.33 (0.64) 0.62 (0.68) 0.55 (0.70)
6-14 years 0.69 (0.99) 0.46 (0.80) 0.86 (0.96)
15-59 years, male 1.31 (0.99) 0.98 (0.77) 1.06 (1.20)
15-59 years, female 1.42 (0.89) 1.40 (0.87) 1.30 (0.74)
60+ years, male 0.26 (0.44) 0.06 (0.25) 0.14 (0.35)
60+ years, female 0.26 (0.45) 0.08 (0.28) 0.22 (0.41)
Male household head (=1) 0.83 (0.37) 0.78 (0.41) 0.60 (0.49)
Age ofhh head 53.98 (11.49) 33.74 (13.58) 46.27 (13.37)
Maximum years of education 8.92 (4.23) 10.30 (3.76) 8.56 (4.08)
Farm households (=1) 0.45 (0.50) 0.24 (0.43) 0.35 (0.48)
Urban (=1) 0.47 (0.50) 0.54 (0.50) 0.45 (0.50)
Median wage, male (Rp) 1,584 (1,157) 2,268 (3,391) 1,730 (2,111)
Median wage, female (Rp) 883 (1,678) 1,280 (2,720) 1,462 (7,947)
Median prices of sugar (Rp) 3,690 (503) 3,706 (473) 3,732 (43 8)
Median prices of oil (Rp) 3,511 (243) 3,521 (242) 3,508 (241)
Real land value (000 Rp) 13,400 (47,200) 5,631 (31,200) 4,781 (14,000)
Real value of farm bus. assets
(000 Rp) 1,956 (11,600,) 722 (3,820) 1,157 (5,145)
Real value of non-farm bus.
assets 1000 Rp) 2,631 (13,300) 4,813 (49,200) 3,633 (23,600)

 

55

Appendix Table 1.5.1
Coefﬁcient on Income and Income Changes, No Instrumental Variables

 

 

Static Dynamic
Coefﬁcient on Coefﬁcient on changes in
Explanatory variables: log(household income) log(household income)
Household income variable ') 0.070 0.044
(0.004) "* (0.003) "*
+ Education variable 9 0.048 0.036
(0.004) *** (0.003) "*
+ Household composition variables " 0.032 0.020
(0.004) "* (0.003) "‘"
Full speciﬁcation " 0.022 0.017
(0.003) *** (0.003) ***
Number of households 3,899 3,899

 

The sample consists of households in extended household with multiple sub-households. The dependent
variable for the static speciﬁcation is log(household expenditure) and for the dynamic speciﬁcation the
change in log(household expenditure) between 1997 and 2000. Standard errors (in parentheses) are robust
to serial correlation and heteroskedasticity. "* indicates statistical signiﬁcance at 1%, ** at 5%, and "' at
10 %.

a) Household income (static) or changes in household income (dynamic) is the only explanatory variable
in the bivariate regression.

b) Maximum education of adult (for the static speciﬁcation) or its change (for the dynamic speciﬁcation is
added) as an additional explanatory variable.

c) Household composition variables (or their changes) are added as additional explanatory variables.
These variables include: log(household size), proportion of household members aged 6-14, 15-59 male
and female, 60+ male and female.

(1) The full speciﬁcations corresponds to the result shown in Table 1.5.1 (for the static speciﬁcation) and
Table 1.5.2 (for the dynamic speciﬁcation). See Appendix Table 1.5.2 (static) and Appendix Table 1.5. 10
(dynamic) for the complete sets of explanatory variables.

56

Appendix Table 1.5.2
Static Tests: All Extended Families and Parent-Child Extended Families, 2000

 

 

 

 

Dependent variable: log All Extended Families with . . .
(household real expenditure) Multiple Households Parent-child Extended Fanulies
Extended-Family No Fixed- Extended-Family
No Fixed-Effects Fixed-Effects Effects Fixed-Effects
log(hh real income) 0.022 0.021 0.020 0.019
, (0.004) *** (0.004)""‘ (0.004) *“ (0.004) ""
log(hh size) 0.408 0.521 0.410 0.556
(0.035) *** (0.042)*** (0.037) "" (0.045)"'"‘*
Proportion of hhmembers:
6-14 years 0.564 0.303 0.594 0.307
(0.077) *** (0.098)*"'* (0.082)*" (0.105)***
15-59 years, male 0.424 0.299 0.483 0.377
(0.078) *** (0.094)*** (0.085)""‘ (0.101)"*
15-59 years, female 0.498 0.320 0.513 0.355
(0.080) *** (0.102)“* (0.086)"* (0.110)"*
60+ years, male 0.548 0.276 0.536 0.257
(0.110)*** (0.133)" (0.122)"‘“ (0.148)*
60+ years, female 0.258 0.102 0.245 0.205
(0.140)* (0.158) (0.155) (0.172)
Male household head (=1) 0.147 0.138 0.146 0.123
(0.032) *** (0.042)*'"* (0.034) *" (0.045) *”
Age of hh head 0.019 0.014 0.021 0.017
(0.004) *** (0.004)""‘ (0.004)*" (0.005)*"""
Age of hh head (squared) -0.000 -0.000 -0.000 -0.000
(0.000) ”* (0.000)*** (0.000) "”” (0.000) ""'
Max. years of education 0.068 0.042 0.068 0.040
(0.003) "* (0.005)""‘ (0.003)""' (0.005) **"'
Farm households (=1) -0.036 -0.040 -0.036 -0.048
(0.024) (0.033) (0.026) (0.035)
log (median wage), male 0.092 0.043 0.089 0.040
(0.018)*** (0.022)* (0.019)*** (0.024)*
log (median wage), female 0.032 0.025 0.032 0.011
(0.012) "* (0.017) (0.013) " (0.018)
log (median prices of sugar) -0.414 0.313 -0.336 0.310
(0.170)" (0.321) (0.178)* (0.342)
log (median prices of oil) 0.268 0.017 0.281 -0.137
(0.109)" (0.196) (0.117)" (0.207)
Constant 11.623 8.505 10.871 9.840
(1.605)*** (3.034)*** (1.700)*** (3.248)***
Number of households 3899 3899 3377 3377
R-squared 0.42 0.33 0.42 0.33
Number of extended families 1723 1510

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. ‘** indicates

statistical signiﬁcance at 1%, " at 5%, and "' at 10 %. Omitted variables are proportion of hh members age
0-4, female hh head, and non-farm household. Variables included in the estimations but not reported on the
table are province and urban dummy variables and province-urban dummy interactions.

Appendix Table 1.5.3
Static Tests, ZSLS, 2"d Stage: All Extended Families and Parent-Child Extended families, 2000

 

 

 

Dependent variable: log All Extended Families with Parent-child Extended Families
(household real expenditure) Multiple Households
No Fixed Extended-Family N o Fixed Extended-Family
Effects Fixed-Effects Effects Fixed-Effects
Log(hh real income) 0.216 0.135 0.218 0.128
(0.031)*** (0.028) "* (0.035)"* (0.031)"‘
Log(hh size) 0.086 0.361 0.109 0.429
(0.069) (0.063) *** (0.073) (0.063) “*
Prgmrtion of hh members:
6-14 years 1.099 0.585 1.200 0.604
(0.141)**"‘ (0.135)*** (0.162)*** (0.149)*"
15-59 years, male 0.528 0.381 0.638 0.471
(0.109)*"‘* (0.113)"* (0.121)"'** (0.121)"*
15-59 years, female 0.569 0.353 0.643 0.427
(0.115)*" (0.120)*” (0.127)"* (0.130)*”
60+ years, male 0.775 0.336 0.695 0.257
(0.156)"* (0.157)" (0.173)"* (0.172)
60+ years, female 0.471 0.237 0.489 0.334
(0.186)" (0.189) (0.207) ** (0.204)
Male household head (=1) -0.202 -0.086 -0.211 -0.091
(0.072) *** (0.074) (0.080)*" (0.080)
Age of h head -0.038 -0.020 -0.042 -0.020
(0.010)""‘ (0.010)" (0.012)"* (0.012)*
Age of hh head (squared) 0.000 0.000 0.000 0.000
(0.000)*" (0.000)“ (0.000)*" (0.000)
Maximum years of education 0.052 0.029 0.053 0.027
(0.005)*** (0.006) "* (0.005)*" (0.007) "*
Farm households (=1) -0.092 -0. 105 -0.074 -0. 100
(0.035)"‘" (0.042) ** (0.037)" (0.044) "
log (median wage), male 0.053 0.017 0.055 0.026
(0.023)" (0.027) (0.025)" (0.028)
log (median wage), female 0.014 0.008 0.005 -0.009
(0.016) (0.020) (0.018) (0.022)
log (median prices of sugar) -0.383 0.289 -0.298 0.289
(0.232)* (0.379) (0.251) (0.399)
log (median prices of oil) 0.149 0.176 0.236 0.096
(0.160) (0.234) (0.173) (0.250)
Constant 12.411 7.488 11.021 8.111
(2.220)*""" (3.585)” (2.417)*"‘* (3.821)"
p-values for null hypothesis that;
- IV s are valid 0.89 0.21 0.93 0.22
-log (hh income) is exogenous 0.00 0.00 0.00 0.00
Number of households 3899 3899 3377 3377
Number of extended-families 1723 1510

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. *** indicates
statistical significance at 1%, ** at 5%, and " at 10 %. Omitted variables are proportion of hh members
age 0-4, female hh head, and non-farm household. Variables included in the estimations but not reported
on the table are province and urban dummy variables and province-urban dummy interactions. For the
2SLS estimations, instrumental variables not included in the ﬁrst stage regressions are: log of real value of
land owned, farm business assets, and non-farm business assets.

Appendix Table 1.5.4
Static Test, ZSLS, 1" Stﬂe: All Extended Families and Parent-Child Extended Families, 2000

Dependent variable: log All Extended Families with Parent-chﬂd Extended Families
(household real income) Multiple Households

 

 

 

 

No Fixed Extended-Family No Fixed Extended-Family
Effects Fixed-Effects Effects Fixed-Effects
log(household size) 1.487 1.196 1.418 1.011
(0.146)*** (0.177)*** (0.207)*** (0.242) "'"
Proportion of hh members:
6-14 years -2.717 -2.324 3.116 -2.863
(0.332)*** (0.403)*" (0.440)*** (0.568) "*
15-59 years, male -0.388 -0.441 -0.767 -0.982
(0.338) (0.386) (0455)" (0.548) "'
15-59 years, female -0.227 0.057 -0.738 -0.878
(0.368) (0.418) (0.493) (0.596)
60+ years, male -1.107 -0.458 -0.859 -0.165
(0.516)" (0.566) (0.689) (0.799)
60+ years, female -0.854 -1.347 -l.181 -l.368
(0.627) (0.685)" (0.802) (0.933)
Male household head (=1) 1.771 1.906 1.672 1.821
(0.155)*** (0.176)"* (0.196)*** (0.240) “"
Age of hh head 0.255 0.262 0.296 0.305
(0.022)"‘** (0.018)*** (0.028)*** (0.024) ***
Age of hh head (squared) -0.002 -0.003 -0.003 -0.003
(0.000)*** (0.000)*** (0.000)*"”" (0.000) ""'
Maximum years of education 0.056 0.081 0.065 0.106
(0.013)**"' (0.019)*” (0.017)"* (0.027) ""‘
Farm households (=1) -0. 160 -0.324 -0.213 -0.438
(0.151) (0.254) (0.187) (0.344)
log (median wage), male 0.300 0.288 0.185 0.140
(0.072)*** (0.091)*** (0.102)‘ (0.130)
log (median wage), female 0.131 0.132 0.157 0.203
(0.048)"* (0.070)‘ (0.063)" (0.097) *"'
log (median prices of sugar) -0.415 -1.002 -0.092 0.490
(0.724) (1.327) (0.975) (1.855)
log (median prices of oil) 0.460 0.121 0.271 -1.857
(0.466) (0.811) (0.640) (1.121) *
log (real land value) 0.011 0.039 0.009 0.048
(0.005)" (0.011)*** (0.007) (0.016) "*
log (real value of farm bus. assets) 0.037 0.047 0.036 0.047
(0.012)"'** (0.021)" (0.015)" (0.028) *
log (real value of non-farm bus. 0.065 0.070 0.062 0.060
assets)
(0.005)*" (0.008)*“ (0.006)*" (0.011) ""‘
Constant -0.685 7.120 -1.605 11.754
(6.775) (12.412) (9.255) (17.601)
(continued)

59

(continued)

 

F-test of exclusionary restrictions 42.28 19.43 32.41 15.82
(p~values) (0.000) (0.000) (0.000) (0.000)
Number of households 3899 3899 3377 3377
R-squared 0.30 0.33 0.29 0.34
Number of extended-families 1723 1510

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. *" indicates
statistical signiﬁcance at 1%, " at 5%, and * at 10 %. Omitted variables are proportion of hh members age
04, female hh head, and non-farm household. Instrumental variables not included in the ﬁrst stage
regressions are: log of real value of land owned, farm business assets, and non-farm business assets.
Variables included in the estimations but not reported on the table are province and urban dummy variables
and province-urban dummy interactions.

60

Appendix Table 1.5.5
Static Tests: 2000 Parent-Son and Parent-Daughter Extended Families

Parent-Daughter Extended
Families
Extended-Family Extended-Family
No Fixed Effects Fixed-Effects No Fixed Effects Fixed-Effects

Dependent variable: log Parent-Son Extended Families

(household real expenditure)

 

 

 

log(household real income) 0.026 0.025 0.016 0.015
(0.006)*** (0.006)*** (0.005) *** (0.006)“
log(household size) 0.389 0.541 0.420 0.577
(0.051)**"' (0.059)*** (0.047) *** (0.062)"*
Proportion 011m members:
6-14 years 0.515 0.210 0.674 0.425
(0.112)*** (0.149) (0.110) *** (0.141)"*
15-59 years, male 0.439 0.352 0.453 0.440
(0.115)*** (0.138)“ (0.122) *** (0.152)***
15-59 years, female 0.407 0.172 0.616 0.517
(0.131)*** (0.163) (0.107) *” (0.147)*"”"
60+ years, male 0.351 0.001 0.697 0.493
(0.165) ** (0.195) (0.163) *** (0.205)"
60+ years, female -0.004 -0.012 0.260 0.398
(O. 185) (0.224) (0.216) (0.236)*
Male household head (= 1) 0.089 0.021 0.231 0.188
(0.044) ** (0.058) (0.047) *** (0.063)***
Age of hh head 0.019 0.012 0.021 0.019
(0.005) *" (0.006)" (0.006) *** (0.007) "*
Age of hh head (squared) -0.000 -0.000 -0.000 -0.000
(0.000) *** (0.000)"' (0.000) *" (0.000) **"'
Maximum years of education 0.067 0.036 0.068 0.039
(0.004) *** (0.007)*** (0.004) **"' (0.007) ***
Farm households (=1) -0.016 0.021 -0.065 -0.099
(0.037) (0.051) (0.034) * (0.046)"
log (median wage), male 0.066 -0.034 0.116 0.111
(0.025)*** (0.034) (0.027) ”* (0.033)*"
log (median wage), female 0.040 0.049 0.021 -0.016
(0.016)" (0.023)” (0.017) (0.024)
log (median prices of sugar) -0.597 -0.230 0165 0.632
(0.245)" (0.432) (0.226) (0.505)
log (median prices of oil) 0.246 -0.622 0.325 -0.044
(0.172) (0.311)" (0.148) """ (0.263)
Constant 13.515 18.798 8.873 5.854
(2.363) *" (4.378)*** (2.167) *** (4.559)
Number of households 1785 1785 1933 1933
R-squared 0.42 0.36 0.43 0.34
Number of extended-households 834 907

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. *** indicates

statistical signiﬁcance at 1%, "”" at 5%, and "' at 10 %. Omitted variables are proportion of hh members age
04, female hh head, and non-farm household. Variables included in the estimations but not reported on the
table are province and urban dummy variables and province-urban dummy interactions.

Appendix Table 1.5.6
Static Tests, Two-State Least Squares 2"d Stage:
Parent-Son and Parent-Daughter Extended Families, 2000

Dependent variable: log (household Parent-Son Extended Families Parent-Daughter Extended Families

 

real ex nditure
pe ) Extended-Family

No Fixed Effects Fixed -Effects

Extended-Family
No Fixed Effects Fixed -Effects

log(household real income) 0.284 0.126 0.173 0.127
(0.062)“"' (0.044) “" (0.040)" (0.042) "”"
log(household size) -0.052 0.401 0.248 0.493
(0.130) (0.090) **"‘ (0.074) "”' (0.079) *”
Promrtion of hh members:
6-14 years 1.283 0.484 0.975 0.659
(0.272)‘" (0.208) " (0.159)" (0.185)""
15-59 years, male 0.635 0.433 0.237 0.383
(0.192)*" (0.163)"" (0.162) (0.179)”
15-59 years, female 0.245 0.172 0.800 0.673
(0.225) (0.188) (0.147)" (0.181)“*
60+ years, male 0.555 0.025 0.688 0.407
(0.276)" (0.226) (0.199)" (0241)"
60+ years, female -0.077 -0.004 0.373 0.532
(0.317) (0.259) (0.236) (0.280) “
Male household head (= 1) -0.299 -0.158 -0.043 -0.037
(0.121)" (0.102) (0.090) (0.111)
Age of hh head -0.050 -0.019 -0.023 -0.015
(0.018)”‘ (0.015) (0.013)‘ (0.015)
Age of hh head (squared) 0.000 0.000 0.000 0.000
(0.000) "“" (0.000) (0.000) " (0.000)
Maximum years of education 0.046 0.025 0.054 0.024
(0.009)”"' (0.009) *" (0.006)” (0.009) ""
Farm households (=1) -0.087 -0.045 -0.083 -0.1 18
(0.062) (0.065) (0.042) "”" (0.054) "
log (median wage), male 0.057 -0.023 0.075 0.081
(0.040) (0.039) (0.030) ” (0.040) "
log (median wage), female 0.039 0.049 -0.015 -0.053
(0.028) (0.027) " (0.022) (0.031)‘
log (median prices of sugar) -0.439 -0.320 -0.247 0.507
(0.399) (0.500) (0.292) (0.592)
log (median prices of oil) 0.104 -0.443 0.277 0.144
(0.298) (0.366) (0.192) (0.315)
Constant 12.840 17.724 10.102 5.518
(3.935)"* (5.068) *" (2.782)“ (5.327)
Q-values for null hypothesis that:
- le are valid (over-identiﬁcation 0.90 0.47 0.56 0.01
- log (hh income) is exogenous 0.00 0.01 0.00 0.01
Number of households 1785 1785 1933 1933
Number of extended- families 834 907

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. "" indicates statistical
signiﬁcance at 1%, ” at 5%, and " at 10 %. Omitted variables are proportion of hh members age 0-4, female hh head,
and non-farm household. Variables included in the estimations but not reported on the table are province and urban
dummy variables and province-urban dummy interactions. For the ZSLS estimations, instrumental variables not
included in the ﬁrst stage regressions are: log of real value of land owned, farm business assets, and non-farm business

assets.

Dependent variable: log (household
real income)

Appendix Table 1.5.7
Static Tests, Two-State Least Squares 1‘t Stage:
Parent-Son and Parent-Daughter Extended Families, 2000

Parent-Son Extended Families

Parent-Daughter Extended Families

 

Extended -Family

Extended oFamily

 

 

No Fixed Effects Fixed-Effects No Fixed Effects Fixed-Effects
log(household size) 1.612 1.200 0.971 0.642
(0.264)*** (0.329)*** (0.272)*" (0.327)"I
Proportion of hh members:
6-14 years -3.114 -2.840 -1.924 -2.243
(0.630)*** (0.831)**"' (0.578)**"' (0.741)*"
15-59 years, male -0.763 -0.890 1.307 0.340
(0.556) (0.774) (0.613)" (0.804)
15-59 years, female 0.521 -0.175 -1.230 -1.602
(0.651) (0.914) (0.632)‘I (0.776)"
60+ years, male -0.870 -0.374 0.053 0.720
(0.878) (1.094) (0.882) (1.084)
60+ years, female 0.302 -0.157 -0.791 -l.512
(0.867) (1.257) (1.087) (1.250)
Male household head (=1) 1.374 1.619 1.628 1.889
(0.256)"“""I (0.324)“* (0.248)*** (0.329)*”
Age of hh head 0.248 0.287 0.267 0.277
(0.037)*** (0.034)"'M (0.034)*"‘* (0.034)"*
Age of hh head (squared) -0.002 -0.003 -0.003 -0.003
(0.000)*** (0.000)*"”" (0.000)*** (0.000)"*
Maximum years of education 0.070 0.095 0.071 0.112
(0.022)*** (0.038)" (0.021)*" (0.036)*"
Farm households (=1) -0.043 0.075 -0.301 -0.824
(0.227) (0.475) (0.270) (0.458)*
log (median wage), male 0.061 -0.054 0.258 0.249
(0.143) (0.189) (0.135)* (0.174)
log (median wage), female 0.020 0.006 0.256 0.371
(0.087) (0.130) (0.078)*" (0.129)"*
log (median prices of sugar) -0.239 1.649 0.380 1.043
(1.386) (2.425) (1.228) (2.673)
log (median prices of oil) 0.436 -1.790 0.489 -1.195
(0.928) (1.739) (0.793) (1.398)
log (real land value) 0.014 0.039 0.012 0.062
(0.010) (0.023)* (0.009) (0.019)***
log (real value of farm bus. assets) 0.025 0.026 0.036 0.044
(0.017) (0.039) (0.021)‘I (0.037)
log (real value of non-farm bus. assets) 0.057 0.066 0.062 0.051
(0.009)"* (0.016)*" (0.008)*** (0.015)“*
Constant 0.751 5.050 -7.753 0.218
(12.841) (24.532) (11.549) (24.114)
F-test of exclusionary restrictions 14.26 7.45 21.3 9.23
(p-values) (0.000) (0.000) (0.000) (0.000)
Number of households 1785 1785 1933 1933
R-squared 0.27 0.35 0.33 0.37
Number of extended- families 834 907

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. ‘"" indicates statistical

signiﬁcance at 1%, " at 5%, and "' at 10 %. Omitted variables are proportion of hh members age 0-4, female hh head,
and non-farm household. instrumental variables not included in the ﬁrst stage regressions are: log of real value of land
owned, farm business assets, and non-farm business assets. Variables included in the estimations but not reported on
the table are province and urban dummy variables and province-urban dummy interactions.

63

Appendix Table 1.5.8

Static Tests, Two-Stage Least Squares with Community Dummy Variables, 2"I Stage:

All Extended Families and Parent-Child Extended Families, 2000

 

All Extended Families with Parent-Child Extended
Multiple Households Families
Dependent variable: log (household Extended Family Extended Family

real expenditure)

 

No Fixed Effects Fixed Effects No Fixed Effects Fixed Effects
log (household real income) 0.256 0.203 0.263 0.212
(0.040)*" (0.065)*** (0.048)""' (0.082)"“"'"
log(household size) -0.023 0.163 -0.009 0.206
(0.080) (0.135) (0.091) (0.145)
Proportion of hh members:
6-14 years 0.871 0.561 0.995 0.569
(0.131)**"‘ (0.184)*** (0.155)*** (0.215)“"'
15-59 years, male 0.210 0.187 0.333 0.223
(0.115)* (0.168) (0.125)**"' (0.185)
15-59 years, female 0.330 0.214 0.454 0.323
(0.117)*** (0.173) (0.128)*" (0.197)
60+ years, male 0.572 0.152 0.576 0.280
(0.156)*" (0.225) (0.173)“* (0.259)
60+ years, female 0.149 -0.181 0.161 -0.l84
(0.185) (0.283) (0.205) (0.319)
Male household head (=1) -0.187 -0.179 ~0.l70 -0.127
(0.073)" (0.1 17) (0.080)“ (0.128)
Age of hh head -0.023 -0.023 -0.028 -0.020
(0.010)" (0.016) (0.013)“ (0.019)
Maximum years of education 0.036 0.022 0.035 0.019
(0.007)"‘** (0.011)” (0.008)*"”" (0.013)
Farm households (=1) -0.125 -0. 136 -0.090 -0.146
(0.039)*** (0.067)" (0.041)“ (0.074)M
Constant 10.140 12.010 10.057 9.287
(0.769)*** (1.275)*** (0.787)*** (1.810)***
p-values for null hypothesis that:
- IVs are valid (over-identiﬁcation test) 0.515 0.577 0.867 0.471
- log (hh real income) is exogenous 0.000 0.000 0.000 0.000
Number of households 3899 3899 3377 3377
Number of extended families 1723 1510
Number of extended families
with households in different communities 1124 985

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. "" indicates statistical

signiﬁcance at 1%, " at 5%, and "' at 10 %. Omitted variables are proportion of hh members age 0-4, female hh head,
and non—farm household. Instrumental variables not included in the second stage regressions are: log of real value of
land owned, farm business assets, and non-farm business assets. Variables included in the estimations but not reported
on the table are province and urban dummy variables and province-urban dummy interaction, and community dummy
variables

All Extended Families and Parent-Child Extended Families, 2000
All Extended Families with

Dependent variable: log

Appendix Table 1.5.9
Static Tests, Two-Stage Least Squares with Community Dummy Variables, 1“ Stage:

Multiple Households

Households in Parent-child
Extended Families

 

Extended-Family

Extended-Family

 

 

(household re" income) No Fixed Effects Fixed-Effects No Fixed Effects Fixed-Effects
log(household size) 1.513 1.601 1.471 1.279
(0.198)*** (0.286)*** (0.215)*** (0.316)*"
Proportion of hh members:
6-14 years -l.613 -1.338 -1.956 -1.435
(0.483)*** (0.614)" (0.528)*” (0.660)“
15-59 years, male 0.330 0.336 0.115 0.139
(0.493) (0.611) (0.525) (0.655)
15-59 years, female -0.1 19 -0.095 -0.419 -0.473
(0.515) (0.639) (0.553) (0.694)
60+ years, male -0.976 -0.635 -0.739 -0.826
(0.714) (0.821) (0.771) (0.896)
60+ years, female 0.238 1.238 0.220 0.999
(0.876) (0.993) (0.869) (1.080)
Male household head (=1) 1.330 1.382 1.217 1.181
(0.206)*" (0.274)*" (0.216)*"”" (0.298)*"
Age of hh head 0.211 0.190 0.229 0.192
(0.028)*" (0.029)"* (0.030)*** (0.031)""‘
Maximum years of education 0.121 0.093 0.113 0.099
(0.018)*** (0.031)"* (0.021)*" (0.033)***
Farm households (=1) -0.017 0.002 -0. 177 -0.187
(0.206) (0.377) (0.194) (0.394)
log (real value of land) 0.006 0.030 0.008 0.020
(0.009) (0.017)* (0.009) (0.018)
log(real value of farm assets) 0.033 0.023 0.031 0.043
(0.017)* (0.031) (0.016)‘ (0.032)
log(real value of non-farm assets) 0.051 0.042 0.046 0.035
(0.008)*** (0.013)”* (0.008)*** (0.013)*"
F-test of exclusionary restrictions 17.36 5.74 12.28 3.88
(p-values) 0.000 0.0007 0.000 0.009
Number of households 3,899 3,899 3,377 3,377
Number of extended families 1,723 1,510
Number of extended families
with households in different communities 1,124 985
R-squared 0.57 0.71 0.58 0.74

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. ‘”" indicates statistical
signiﬁcance at 1%, ** at 5%, and * at 10 %. Omitted variables are proportion of hh members age 04, female hh head,
and non-farm household. Instrumental variables not included in the second stage regressions are: log of real
value of land owned, farm business assets, and non-farm business assets. Variables included in the
estimations but not reported on the table are province and urban dummy variables and province-urban

dummy interaction, and community dummy variables.

Appendix Table 1.5.10
Dynamic tests: First Difference Estimation without [Vs

All Extended Families with Multiple

Parent-child Extended Families

 

 

 

 

Dependent variable: A log Households
(h°“‘°h°'d "Mam” Extended—Family Extended Family
No Fixed Effects Fixed Effects No Fixed Effects Fixed Effects
1 2 3 4
A log(household income) 0.017 0.020 0.019 0.019
(0.003)"" (0.004)"" (0.004)""' (0.004)*"”"
A log(household size) 0.030 0.035 0.030 0.034
(0.005)""‘ (0.005)"‘" (0.005)""' (0.005)‘"
A proportion of hh members:
6-14 years 0.542 0.507 0.552 0.529
(0.036)"" (0.045)‘" (0.039)"" (0.049)""
15-59 years, male 0.345 0.316 0.330 0.352
(0.086)""‘ (0.104)”‘ (0.091)"* (0.1 12)"‘
15-59 years, female 0.440 0.348 0.480 0.405
(0.088)"" (0.105)*" (0.093)""' (0.112)”‘
60+ years, male 0.404 0.321 0.442 0.377
(0.093)"" (0.111)"‘" (0.099)"”" (0.120)‘“
60+ years, female -0.079 -0.127 -0.226 -0.129
(0.152) (0.163) (0.167) (0.178)
A Male household head (=1) 0.216 0.094 0.379 0.129
(0.130)‘ (0.145) (0.143)"" (0.160)
A Age of hh head 0.103 0.102 0.128 0.123
(0.035)"" (0.044)" (0.037)’” (0.049)“
A Maximum years of education 0.083 0.180 0.075 0.162
(0.041)" (0.045)"* (0.044)‘ (0.048)""
A Farm households (=1) -0.018 -0.077 -0.010 -0.059
(0.025) (0.033)" (0.026) (0.036)
A log (median wage), male 0.038 0.069 0.040 0.078
(0.018)" (0.023)"" (0.020)" (0.025)"“
A log (median wage), female 0.032 0.042 0.036 0.037
(0.012)“" (0.016)”" (0.013)"" (0.018)”
A log (median prices of sugar) 0.230 0.117 0.176 -0.026
(0.075)’” (0.153) (0.082)" (0.168)
A log (median prices of oil) 0.319 0.351 0.344 0.302
(0.1 15)""‘ (0.276) (0.123)"“ (0.302)
Constant -0. 107 0.257 -0.072 0. 140
(0.045)" (0.226) (0.049) (0.262)
Number of households 3899 3899 3377 3377
R-squared 0.22 0.26 0.22 0.26
Number of extended- families 1723 1510

 

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. "" indicates statistical
signiﬁcance at 1%, " at 5%, and " at 10 %. Variables included in the estimations but not reported on the table are
1997 province and 1997 urban dummy variables and province-urban dummy interactions.

66

Appendix Table 1.5.11

Dynamic tests ZSLS, 2"d Stage, Households in Extended Families with Multiple Households

Dependent variable: A log
(household expenditure)

No Fixed Extended— No Fixed Extended- No Fixed Extended-
Effects Family FE Effects Family FE Effects Family FE

1 2 3 4 5 6

 

A log(household income)

A log(household size)

A promrtion of hh members:
6-14 years

15-59 years, male

15-59 years, female

60+ years, male

60+ years, female
A Male household head (=1)
A Age of hh head
A Maximum years of education
A Farm households (=1)
A log (median wage), male
A log (median wage), female
A log (median prices of sugar)
A log (median prices of oil)
Constant

Q-values for null hypothesis th_a_t_:
- le are valid (overidentiﬁcation
test)

- log (hh real income) is exogenous
Number of households

Number of extended- families

0.132 0.059 0.136 0.072 0.084 0.049
(0.049)*** (0.033)* (0.047)*** (0.031)" (0032):" (0.028)*
0.023 0.032 0.023 0.031 0.026 0.033
(0.006)*** (0.006)*** (0.006)*** (0.006)*" (0.005)*** (0.006)**"'

0.316 0.443 0.309 0.421 0.410 0.459
(0.104)*** (0071):" (0101):” (0.070)*** (0.072)*** (0.064)""'
0.508 0.394 0.513 0.421 0.440 0.375
(0123):“ (0.125)*** (0.122)*** (0125):" (0.102)*** (0.119)m
0.410 0.366 0.409 0.372 0.423 0.362
(0.104)*** (0.109)*** (0.105)*** (0.111)*** (0.093)*** (0.108)***
0.441 0.342 0.442 0.349 0.426 0.337
(0.110)*** (0.115)*** (0.111)*** (0.117)*** (0.098)*** (0.114)***
0.215 0.031 0.223 0.086 0.092 -0.008
(0.208) (0.213) (0.206) (0.212) (0.170) (0.200)
0.463 0.196 0.471 0.231 0.360 0.171
(0.180)" (0.172) (0.179)*** (0.172) (0.148)" (0.164)
-0095 0.024 -0101 -0004 -0013 0.043
(0.092) (0.080) (0.089) (0.078) (0.064) (0.072)
-0094 0.100 -0099 0.073 -0020 0.120
(0.087) (0.081) (0.085) (0.079) (0.063) (0.072)*
-0.081 -0109 -0.083 -012 -0055 -0101
(0.039)" (0.043)" (0.039)" (0.043)*** (0.031)* (0.040)"
-0001 0.057 -0002 0.052 0.016 0.060
(0.026) (0.025)" (0.026) (0.026)" (0.021) (0.025)"
0.022 0.040 0.022 0.039 0.026 0.040
(0.014) (0.017)M (0.014) (0.017)" (0.012)" (0.016)"
0.319 0.116 0.321 0.116 0.282 0.116
(0.089)*** (0.156) (0.089)"* (0.159) (0.077)": (0.155)
0.144 0.240 0.138 0.201 0.217 0.267
(0.156) (0.298) (0.155) (0.302) (0.132) (0.291)
-0051 0.163 -005 0.13 -0074 0.186
(0.060) (0.245) (0.060) (0.248) (0.052) (0.239)

0.32 0.41 0.60 0.21 0.19 0.00
0.00 0.12 0.00 0.08 0.02 0.28
3899 3899 3899 3899 3899 3899

1723 1723 1723

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. *** indicates
statistical signiﬁcance at 1%, ** at 5%, and "‘ at 10 %. In columns (1) and (2), the instrumental variables
not included in the second regressions are: log of land value 1997, changes in land value. In columns (3)
and (4), the instrumental variables not included in the second regressions are: log of land value 1997, log of
business assets 1997, changes in real land value. In columns (5) and (6) Instrumental variables not included
in the second regressions are: log of land value 1997, log of business assets 1997, changes in real land
value, interactions of changes in median real wages (male, female) with maximum years of education 1997.

67

$63538»

 

 

 

 

ASN.8 $2.3 388 $2.8 38.8 $2 .8
an; e86 9.3. memo on; $2. :8 £288; as» <
3.888 1.598 ...,...888 :988 ...amo8 .588
35¢ «88 e85 685 83 63° 8:828 do as» 8:832 4
2.2438 ...EN8 2.52.8 :22.on :528 2.538
3:: as: $3 3: E: 5: Be; .2 .6 6w... 4
seaavmdv :aAw _ Ndv sesAomNdv 3.2..3— N.8 eessmmdv :33 _ Ndv
«SN 2: ”SN 2: «SN a: :8 82. 22636: 282 <
...828 $228 2:: $8 at: 3.8 2...: 3.8 3338
Sam. 82- $2- a: .N- 8:- :3. 22.8 .88» +8
ssaawmadv saeAWNadv sesAwNodv «ssAMN@dV seaﬁwmadv sseQVNodv
33. 3.3- ”m3. tea- PS. 83- 6.8: .88; +8
38.8 2: m8 988 2 $8 38.8 88.8
6%.? was. 82.. 82- Red. 22- ease 58» 3-2
38.8 :38 3.8.8 2 3.8 9.88 83.8
83. 33 one? 82 33. $2 one .88» $2
assAmamdv «sea—DES sausage essacvdv seismmdv assAwwvdv
53. $2- 2.3- 2 2- 83- 33- ea» 38
”$382: an: we aomtoqoa <
3:238 :588 ...838 :588 :538 ...G88
83 83 «a: 82 mm: 3.3 3% 228883 a
$8 5 A8 5 Q E
A an 88.5— 13: h :2 mm 33:”..— mm 88:”.— 3585
Each Bogota— »..Eah cot—.85— voxE oz 355— eoveoaxm cour— ez .3.— 23835 we. < 333...; 2.9.52.2—

 

ezegaae: 2.532 a? 83:5... 82.65 5 8.2.88: .1639: £qu .38. 252.3
2.3 2%... 532.3.

68

..x. o. 8 e can .6\6m .8 e... .6\6_ .8 356538 .3538... 8.36:. .3...

86:836—888.. new 5:22.86 3.8 2 .858 08 880558.. .5 ﬁche BEESm

 

 

m8. m8. m8. 8:88. 8.86.8 .6 88:52
.8... 2.6 8.6 2.6 8... .2 .6 8868-...
688 86.8. 668 668 66.:- 668 8.62.82. .6 .8882

88.8 88.8 88.8 88.8 88.8 88.8 828.8
..N... .88 6...: 3... 8.2 3.2 .2 858.828 .8.-.
..8. m. 8 .888 :8. m. .8 .888 .88.: ..888
$68 8.....- 83 8.....- 83 $2.- .6888
..iAVchv sans—OAS
awed god 82 onto 88:. x $88 awn.» 5:825 wo_ <
88.8 ...888
356. 2b.? 32 03.0 .88 x GEE awn? 5:625 wo_ <
..888 88.8 ..888 88.8
86.6. 288- 68.6- 88.6. 88. 8.88 8.. .6 28> .88 mo.
8883—98 8.5: 2:: 3.33—98 3.1.: —o.oV eaeAﬂodv :5: ~98
88.6 86.6 68.6 86... ms... 88... 82.; .8... .88 8. <
88.8 ...888 88.8 ...888 88.8 ...2 8.8
36.8. 68.6 86.6- 68... m8...- 868 86. 8...... .8... .88 86.
News... 2.8% 05 89c 38288 833%
.88. .8 ...888 .88.: ...888 .88. 8 ...888
868 88.. .88 86.. 66.2 E... 86 .6 86.... 8.85 8. <
:88 2.888 88.8 ..888 88.8 ..888
68.6- .8.6- 868- 6......- 888- 2.....- .8836 86.... 8.85 8. 4
38.8.8 .828 .868 .888 88.8 .888
86.....- 82.- 86... 8. .6 68... 8.... 8.856. .883 8.8.5 8. 6
.888 ...,..88 ..828 ...828 :82 .8 ...888
8:. m3. 8...... 8m... 3 no .8... 8.6... .883 8.868 8. 6

 

33:22.80»

69

Dynamic tests, ZSLS, 2"

Dependent variable: A log
(household expenditure)

Appendix Table 5.13
Stage, Parent-Child Extended Families

No Fixed Extended No Fixed Extended No Fixed Extended

Effects

FamilyFE Effects FamilyFE Effects FamilyFE

 

 

1 2 3 4 5 6
A log(household income) 0.133 0.067 0.157 0.090 0.085 0.073
(0.053)" (0.031)" (0.053)*** (0.031)“* (0.033)"'** (0.028)*"‘* (0.028)*
A log(household size) 0.021 0.030 0.020 0.028 0.025 0.029
(0.007)*** (0.006)*** (0.007)*** (0.007)*** (0.005)*** (0.006)***
A promrtion of hh members:
6-14 years 0.346 0.456 0.304 0.421 0.433 0.447
(0.105)*** (0.069)*** (0.106)“* (0.071)*** (0.070)*** (0.066)“*
15-59 years, male 0.510 0.444 0.547 0.489 0.434 0.456
(0.137)*** (0.131)*** (0.142)*** (0.135)*** (0.110)*** (0.129)*"
15-59 years, female 0.476 0.437 0.475 0.452 0.477 0.441
(0.110)**"' (0.118)*** (0.116)"'** (0.123)*“ (0.099)**"‘ (0.119)*“
60+ years, male 0.504 0.410 0.517 0.427 0.478 0.415
(0.120)*** (0.126)*" (0.127)*" (0.131)""""' (0.106)*"’* (0.127)*"'*
60+ years, female 0.061 0.079 0.119 0.179 -0.061 0.105
(0.224) (0.229) (0.233) (0.235) (0.182) (0.221)
A Male household head (=1) 0.636 0.250 0.689 0.308 0.527 0.265
(0.201)""' (0.184) (0.208)""‘ (0.189) (0.163)*** (0.181)
A Age of hh head -0.083 0.015 -0.127 -0.037 0.006 0.002
(0.107) (0.087) (0.107) (0.088) (0.071) (0.080)
A Maximum years of education -0.127 0.057 -0. 169 0.006 -0.042 0.044
(0.106) (0.085) (0.106) (0.086) (0.072) (0.078)
A Farm households (=1) -0.052 -0.086 -0.061 -0.098 -0.034 -0.089
(0.036) (0.041)” (0.038) (0.042)“ (0.030) (0.040)"
A log (median wage), male -0.004 0.067 -0.013 0.062 0.015 0.066
(0.029) (0.026)“ (0.030) (0.027)" (0.023) (0.026)“
A log (median wage), female 0.018 0.030 0.014 0.027 0.025 0.030
(0.016) (0.019) (0.017) (0.020) (0.014)* (0.019)
A log (median prices of sugar) 0.255 -0.037 0.272 -0.042 0.222 -0.038
(0.094)*" (0.175) (0.099)*** (0.182) (0.081)*" (0.176)
A log (median prices of oil) 0.180 0.179 0.146 0.120 0.249 0.164
(0.164) (0.323) (0.171) (0.335) (0.139)* (0.324)
Constant -0.01 1 -0.021 0.002 -0.099 -0.037 -0.041
(0.066) (0.291) (0.069) (0.301) (0.056) (0.289)
D-values for null hypothesis that;
- IVs are valid (overidentification 0.61 0.59 0.47 0.00 0.09 0.00
-test)
- log (hh real income) is exogenous 0.01 0.12 0.00 0.01 0.03 0.04
Number of households 3377 3377 3377 3377 3377 3377
Number of extended-families 1510 1510 1510

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. "* indicates
statistical signiﬁcance at 1%, *"‘ at 5%, and * at 10 %. In columns (1) and (2) , the instrumental variables
not included in the second regressions are: log of land value 1997, changes in land value. In columns (3)
and (4), the instrumental variables not included in the second regressions are: log of land value 1997, log of
business assets 1997, changes in real land value. In columns (5) and (6), the instrumental variables not
included in the second regressions are: log of land value 1997, log of business assets 1997, changes in real
land value, interactions of changes in median real wages (male, female) with maximum years of education

1997.

70

«33338»

 

 

 

 

ASN.8 GE .8 9.88 G»: .8 ASN.8 32.8
$3- ~35 82.- $56 2.3- 2.3 :8 86889. 58,.» 4
.13: 2:5 1.3qu 32% — modv ...komadv 3.. ...A — no.8 3.8228
855 $3 Sod 29° 38.0 ES 8:838 8 :8» 8:882 <
...838 £588 :538 :538 5:88 ...888
23 no? 23 $3 83 «8.. 98.. .2 8o om< q
iiacNdv «...Ahmmdv fixawmdv ...iawmmdv ﬁzacmdv .IiAmedv
83 mm: $3 8.3 £3 83: euv 82» 20882» on: <
...isoadv ...,...Amcwdv «..."Ccodv #:36de {*iA—Oadv iiaahdv
82. a :8- 22. 8 3. $3- $3. 22:8 .88» +8.
iiAwoadv 88*Amnadv ..ikamdv {EA—bodv ..ikaodv .1...kade
$3. :3- 831 8.3- $2. $3- 28. £8» +8
858 $8.8 $5.8 $8.8 83.8 88.8
88.8 33. ES- 83.. w _ 2.- 83. 28.8 £8» 3.:
$8.8 $8.8 33.8 $8.8 $8.8 $3.8
SS. ~85- and. 83.. Sad. 82. 22: £8» 8.:
2:: 8.8 ...acm8 ...388 ...38.8 :588 :588
me .N- E. _- £3- 23- N: .m- 83- £8» :8
”$3808 .2 mo 59235 <
2.358 ...GNN8 .3388 .3298 ...,...»E8 ...388
:2 mm: 82 SS 82 8: 8% 228858. <
8 6 E 5 S E
385. 85m 385. 88... 82 £85 8:... 385 £85 8:... £85 858...
2:53 31:85— .»:Eah Banana coma cz £55m Basia no.3...— c2 .3.- 223255 we. 4 "czar—.3 25.52.09

 

8:25... 820:”,— EEQEES 88% 8. .33 £8. 252.»;

3.3 2%... 58...?

71

..x. S 8 .. EB 58¢. 8 .8 ..x: 8 3523,86 :85sz «.2365 .3...

56:888—838.: 98 noun—9:8 .888 9 7.38 28 $805589 :6 Echo vaugm

 

 

 

 

8 2 8 2 8 2 88.5-8828 8 8882
8.8 8.8 8.8 w _ .o 8.8 8 .8 888...
E8 :8 :8 :8 2.8 :8 882.82. 8 .3832

88.8 88.8 88.8 88.8 88.8 88.8 8288
8.8 8.8 8.2 8.8 8.2 8.»: ..Z 82882 8 .8;
...88. 24 .858 588.. 2% 58.8 :98. a ..888
8:. :88- 88+ 2 no- :3 28.8. 8588
..888 ..388
88.». N88 82 88 .88 x 888 883 88.5 we 4
888 .3288
Sad: N066: h0¢~ onto .me x 372: .oww3 :ﬁﬁoev me Q
.888 88.8 .888 38.8
88.8. 888- 88.8- 88.8- 88 888 .88 8 8.8 .88 82
...888 ...c 8.8 ...888 ...2 8.8 ...888 ...c 8.8
88.8 88.8 88.8 «88 88.8 88.8 828 82 88 m2 4
:88 558.8 2 8.8 .5588 .388 .588
88.8. 28.8 8.8- 88.8 88.8- 88.8 88 828 8.2 88 8:
wwwum 988m 05 Bob veg—88 £31533
.2». z ..888 3:. 8 .588 8:. 8 ..888
8...... 82 8.8 82 8mm 83 :8 8 828 8828 3 <
:88 .888 28.8 .888 28.8 .2888
83 8.3- «8.8 88.8. «.88 8.3- 88. 8 838 888 we 4
.388 8:8 .88 .8 .2888 .»8_.8 .2888
8.8- 28. 8:. 82 .o 8. .o 2: .8 228 8.8.. 888 we 4
£888 ...288 82.8 ...»2 _.8 .82 .8 ...»2 _.8
88.8 $2 83 88.8 88 83 2.8 .883 888 we 4.
«83:22:88

72

Appendix Table 1.5.15
Do other households' resources affect own household's consumption?
2SLS: Second Sta e Re ressions

Dependent variable: log All extended families with

 

 

(household real expenditure) multiple sub-households Parent-child extended families
log (hh's income) 0.181 0.180
(0.027)"* (0.029)""'
log (other hh's income) 0.032 0.030
(0.031) (0.032)
Own hh's variables
log(household size) 0.167 0.188
(0.061)*"“" (0.065)""‘
Promnion of hh members:
6-14 years 0.929 0.992
(0.282)*" (0.309)“'"
1559 years, male 0.253 0.313
(0.219) (0.234)
15-59 years, female 0.765 0.781
(0.221)"'" (0.245)"""
60+ years, male 0.636 0.597
(0.348)* (0.407)
60+ years, female 0.076 -0.047
(0.405) (0.461)
Male household head (=1) -0.131 -0.139
(0.065)” (0.070)“
Age of hh head -0.028 -0.031
(0.009)‘" (0.010)"*
Maximum years of education 0.038 0.037
(0.008)”* (0.009)""
Farm households (=1) -0. 100 -0.081
(0.047)" (0.051)
log (median wage), male 0.046 0.049
(0.022)" (0.024)"
log (median wage), female 0.008 -0.002
(0.016) (0.017)
log (median prices of sugar) 0.012 0.172
(0.403) (0.434)
log (median prices of oil) 0.056 0.021
(0.307) (0.335)
Other hh's variables
log(household size) 0.010 0.015
(0.021) (0.023)
Prop_ortion of hh members:
6-14 years 0.061 0.069
(0.303) (0.325)
15-59 years, male 0.276 0.311
(0.226) (0.238)
15-59 years, female -0.275 -0.223
(0.230) (0.247)
60+ years, male 0.381 0.536
(0.430) (0.482)
60+ years, female 0.110 0.096
(0.372) (0.433)
(continued )

73

(continued!

 

Other hh's variables
Male household head (=1)

Age of hh head

Maximum years of education
Farm households (= 1)

log (median wage), male

log (median wage), female
log (median prices of sugar)
log (median prices of oil)

Constant

p-value : signiﬁcance of other hh’s variables

g-values of null hymthesis that:
- le are valid (overidentiﬁcation test)

- Own hh income exogenous
- Other hh income exogenous

Number of households
Number of extended families

-0.088
(0.069)
-0.010
(0.01 1)
0.013
(0.006)"
0.021
(0.033)
0.035
(0.024)
0.009
(0.017)
-0099
(0.171)
-0271
(0.235)
12.653
(2.344)***

61.61 (0.012)

0.039
0.000
0.221

3899
1723

-0.081
(0.071)
-0.01 1
(0.012)
0.014
(0.006)"
0.018
(0.035)
0.034
(0.025)
0.021
(0.018)
-0005
(0.184)
-0.241
(0.252)
10.639
(2.532)"*

59.18 (0.020)

0.052
0.000
0.258

3377
1510

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. *" indicates statistical
signiﬁcance at 1%, ** at 5%, and "' at 10 %. Omitted variables are proportion of hh members age 0-4, female hh head,
and non-farm household. Variables included in the estimations but not reported on the table are province and urban
dummy variables and province-urban dummy interactions. For the ZSLS estimations, instrumental variables not
included in the ﬁrst stage regressions are: own household's and other households' log of real value of land owned, farm

business assets, non-farm business assets.

74

Appendix Table 1.5.16

Do other households' resources affect own household's consumption?

2SLS: First Sta e Re ressions

All extended families with

multiple households Parent-child extended families

 

Dependent variable: log (other hh's real income)

 

Own hh's variables

 

 

 

75

log(household size) 1.504 0.194 1.388 0.301
(0.191)*** (0.162) (0.208)*** (0.176)’
Pronortion of 11h members:
6-14 years -2.599 4.263 -2.570 4.176
(0.984)*"'* (0.554)*"”" (1.092)" (0.635)""‘
15-59 years, male 0.520 4.718 0.615 4.782
(0.800) (0.452)**"' (0.858) (0.509)*"
15-59 years, female -1.485 4.250 -1.659 4.484
(0.907) (0.492)*" (1.013) (0.579)***
60+ years, male -1.021 3.214 -0.100 2.936
(1.888) (0.850)*** (2.058) (1.038)*"'*
60+ years, female -2.191 1.327 -2.153 1.437
(1.971) (1.232) (1.890) (1.330)
Male household head (=1) 1.694 -0.273 1.667 -0.275
(0.185)"”” (0.142)* (0.197)*"'* (0162)“
Age of hh head 0.263 -0.001 0.285 -0.003
(0.027)**"' (0.016) (0.029)"* (0.016)
Maximum years of education 0.081 -0.151 0.079 -0. 142
(0.025)*" (0.023)"'*"' (0.026)*" (0.025)"*
Farm households (=1) -0.001 0.290 -0.298 0.489
(0.268) (0.235) (0.292) (0.282)‘
log (median wage), male 0.218 -0.002 0.170 -0.004
(0.094)" (0.065) (0.106) (0.073)
log (median wage), female 0.128 0.007 0.167 0.008
(0.062)" (0.049) (0.067)" (0.053)
log (median prices of sugar) 1.177 -0.113 0.804 -0.920
(1.785) (1.464) (1.820) (1.709)
log (median prices of oil) -1.113 0.842 -l.504 1.447
(1.291) (1.095) (1.344) (1.214)
Other hh's variables
log(household size) -0.091 0.156 -0.115 0.185
(0.077) (0.058)"”'"" (0.085) (0.067)*"
Proportion of hh members:
6-14 years 0296 -5.035 -0.580 —4.691
(1.068) (0.705)**"‘ (1.183) (0.779)“*
15-59 years, male -1.347 -5.167 -1.556 -5.114
(0.827) (0.419)**"' (0.896)’ (0.459)”*
15-59 years, female 1.235 -4.912 1.189 -4.897
(0.917) (0.4433*** (1.015) (0.496)***
(continued)

(continued)

 

Other hh’s variables
60+ years, male

60+ years, female

Male household head (=1)
Age of hh head

Maximum years of education
Farm households (=1)

log (median wage), male

log (median wage), female
log (median prices of sugar)
log (median prices of oil)

IVs Excluded from 2nd Stage
log(own land)

log (own farm prod. assets)

log (own non-farm prod. assets)
log (other's land)

log (other's farm bus. assets)

log (other's non-farm bus. assets)
Constant

F_-t_est of identifying IVs (p-values)
Land and bus. assets (own hh)
Land and bus assets (other hh)
Land and bus assets (own, other hh)

Number of households

Number of extended families
R-squared

1.233
(2.129)
-0333
(1.999)
-0.087
(0.134)

0.009
(0.017)
-0009
(0.015)
-0095
(0.215)
-0029
(0.082)
-0.046
(0.062)

1.348

(0.708)*
-0.078
(1.045)

0.014
(0.007)"
0.032
(0.014)"
0.063
(0006)::38
-0040
(0.01 1)***
0.024
(0.020)
-0000
(0.008)
40.523
(9.133)

40.97 (0.000)
4.72 (0.003)
22.47(o.000)

3899
1723
0.31

-1291
(1.475)
4494

(0.944)***

1.842
(0.181)***
0.298
(0.027)***
0.131

(0.015)***
-0157
(0.141)

0.200
(0.096)"
0.067
(0.058)
0.018
(0.645)
0.145
(0.918)

-0.028
(0009):“:
0.007
(0.018)
-0000
(0.007)
0.002
(0.007)
0.041
(0.013)***
0.054
(0.005)***
-6.461
(8.383)

3.710 (0.011)
41.62 (0.000)
22.22 (0.000)

3899
1723
0.32

1.174
(2.055)
-0.854
(2.149)
-0.098
(0.143)

0.008
(0.017)
-001 1
(0.016)

0.093
(0.249)

0.016
(0.088)
0043
(0.065)

1.479

(0.750)M

0.081
(1.090)

0.013
(0.007)*
0.036
(0.015)"
0.061
(0.006)***
-0039
(0.012)***
0.006
(0.022)
0.003
(0.008)
-7.l46
(9.920)

32.35 (0.000)
4.15 (0.006)
18.07 (0.000)

3377
1510
0.31

-1105
(1.617)
.4296

(1.130)***

1.785
(0.193)***
0.311
(0.030)***
0.117

(0.017)***
-O.198
(0.157)

0.179
(0.107)*
0.099
(0.063)
-0.180
(0.705)
0.144
(1.082)

-0.028
(0.01 1)***
-0010
(0.022)
0.000
(0.008)
0.001
(0.008)
0.044
(0.014)***
0.056
(0.006)***
3.898
(9.224)

3.60 (0.013)
35.43 (0.006)
19.00 (0.000)

3377
1510
0.32

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. "" indicates statistical
signiﬁcance at 1%, " at 5%, and "' at 10 %. Omitted variables are proportion of hh members age 0-4, female hh head,
and non-farm household. Instrumental variables not included in the first stage regressions are: own hh's and other hh's
log of real value of land owned, farm business assets, and non-farm business assets. Variables included in the
estimations but not reported on the table are province and urban dummy variables and their interactions.

76

Appendix Table 1.5.17
Do other households' resources affect own household's consumption?
Reduced Form Regressions

Dependent variable: log OLS: All extended families OLS: Parent-child extended

 

 

 

 

(household real expenditure) with multiple sub-households families
Own hh's variables
log(household size) 0.447 0.448
(0.034)*** (0.037)*"""
Proportion of hh members:
6-14 years 0.563 0.615
(0.182)*** (0.196)"*
15-59 years, male 0.475 0.548
(0.134)*** (0.149)""'
15—59 years, female 0.612 0.598
(0.134)‘** (0.143)***
60+ years, male 0.542 0.657
(0.280)* (0341)“
60+ years, female 03 18 -0.443
(0.301) (0.318)
Male household head (=1) 0.162 0.149
(0.031)*** (0.034)"*
Age of hh head 0.020 0.021
(0.004)*""" (0.004)"*
Maximum years of education 0.048 0.047
(0.005)*** (0.005)***
Farm households (=1) -0.081 -0.092
(0.062) (0.066)
log (median wage), male 0.087 0.081
(0.018)"* (0.019)*"
log (median wage), female 0.032 0.030
(0.012)*** (0.013)“
log (median prices of sugar) 0.182 0.242
(0.312) (0.343)
log (median prices of oil) -0.084 -0. 168
(0.233) (0.251)
log(own land) x 10'2 0.318 0.274
(0.211) (0.002)
log (own farm prod. assets) x 10'2 0.611 0.669
(0.424) (0.004)
log(own non-farm prod. assets) x 1.328 1.296
(0.154)*** (0.166)"'"
Other hh's variables
log(household size) -0.003 -0.002
(0.016) (0.017)
Proportion of hh members:
6—14 years -0.118 -0.l34
(0.195) (0.210)
15-59 years, male -0.099 -0.092
(0.129) (0.144)
(continued)

77

(continued)

 

15-59 years, female
60+ years, male

60+ years, female

Male household head (=1)

Age of hh head

Maximum years of education
Farm households (=1)

log (median wage), male

log (median wage), female

log (median prices of sugar)

log (median prices of oil)

log (other's land) x 10‘2

log (other's farm bus. assets) x 10'2

log2(other's non-farm bus. assets) x
10'

F-test
Land and bus. assets (own)
Land and bus. assets (other)
All own hh's variables
All other hh's variables

Constant

Number of households

-0.181
(0.126)
0.632
(0.330)*
-0.085
(0.293)

-0043
(0.027)
0.001
(0.004)
0.015
(0.003)***
-0.026
(0.038)
0.039

(0.017)" -
0.006
(0.013)
0.135
(0.125)
-0253
(0.182)
0.099
(0.189)
0.160
(0.309)

0.190
(0.144)

28.36 (0.000)
0.96 (0.409)
27.01 (0.000)
2.09 (0.001)

10.417
(1.860)***
3899

-0.128
(0.133)
0.796
(0.352)"
-0.178
(0.359)

-0044
(0.029)
-0001
(0.004)
0.014
(0.003)***
-0021
(0.041)
0.044
(0.019)M
0.019
(0.014)
0.251
(0.131)*
-O.188
(0.200)

0.126
(0.207)
0.077
(0.339)

0.211
(0.157)

22.80 (0.000)
0.86 ( 0.462)
22.99 (0.000)
2.02 (0.0001)

9.040
(1.985)"“"*
3377

 

Standard errors (in parentheses) are robust to serial correlation and heteroskedasticity. "* indicates
statistical signiﬁcance at 1%, " at 5%, and * at 10 %. Omitted variables are proportion of hh members
age 04, female hh head, and non-farm household. Variables included in the estimations but not reported
on the table are province and urban dummy variables and province-urban dummy interactions.

78

BIBLIOGRAPHY

Altonji, Joseph G., Fumio Hayashi, and Laurence J. Kotlikoff (1997), “Parental Altruism
and Inter Vivos Transfers: Theory and Evidence”, Journal of Political Economy
105:6, p.1121-66.

Altonji, Joseph G., Fumio Hayashi, and Laurence J. Kotlikoff (1992), “Is the Extended
Family Altruistically Linked? Direct Tests Using Micro Data”, American
Economic Review, 82:5, p.1 177-98.

Bergstorm, Theodore C. (1996), “Economics in a Family Way”, Journal of Economic
Literature, 34:4, p.1903-34.

Cameron, Lisa and Deborah Cobb-Clark (2001), “Old-Age Support in Developing
Countries: Labor Supply, Intergenerational Transfers and Living Arrangements”
Working Paper No. 773, Department of Economics, University of Melbourne,
Australia.

Chiappori, Pierre-Andre (1988), “Nash-Bargained Household Decision”. International
Economic Review, 29:4, p. 791-96.

Cochrane, John (1991), “A Simple Test of Consumption Insurance”, Journal of Political
Economy, 99, p.857-76.

Cox, Donald and Emmanuel Jimenez (1990), “Achieving Social Objectives through
Private Transfers”, World Bank Research Observer, 5:2, p.205-18.

Foster, Andrew, and Mark R. Rosenzweig (2002), “Household Division and Rural
Economic Growth” Review of Economic Studies, 69, p.839-69.

Foster, Andrew, and Mark R. Rosenzweig (2001), “Imperfect Commitment, Altruism,
and the Family: Evidence from Transfer Behavior in Low-Income Rural Areas”,
The Review of Economics and Statistics, 83:3 p.389-407.

Foster, Andrew (1993), “Household Partition in Rural Bangladesh”, Population Studies,
47, p.97-114

Frankenberg, E., J .P. Smith, and D.Thomas (2003), “Economic Shocks, Wealth and
Welfare”, The Journal of Human Resources, 38:2, p. 280-321.

Frankenberg, Elizabeth, Lee Lillard, and Robert J. Willis (2002), “Patterns of
Intergenerational Transfers in Southeast Asia”, Journal of Marriage and Family,
64, p. 627-41.

Frankenberg, Elizabeth and Lynn Karoly (1995). The 1993 Indonesian Family Life

79

Survey: Overview and ﬁeld report, Publication No. DRU-1195/1-NICHD/AID,
RAND, Santa Monica, CA.

Frankenberg, Elizabeth and Duncan Thomas, (2000), "The Indonesia Family Life Survey
(IFLS): Study Design and Results from Waves 1 and 2." RAND, Santa Monica,
CA. DRU-223 8/ 1 -NIA/NICHD.

Frankenberg, Elizabeth, Duncan Thomas, and Kathleen Beegle (1999) “The Real Costs
of Indonesia’s Economic Crisis: Preliminary Findings from the Indonesia Family
Life Surveys” RAND DRU 2064. Santa Monica: RAND.

Grimard, Franque (1997), “Household Consumption Smoothing through Ethnic Ties:
Evidence from Cote d’Ivoire”, Journal of Development Economics, 53, p.391-
422.

Haddad, L., J. Hoddinott, and H. Alderman (1997) eds. Intrahousehold Resource
Allocation in Developing Countries: Models, Methods, and Policy, Baltimore:
John Hopkins University Press.

Hayashi, Fumio, Joseph Altonji, and Laurence Kotlikoff (1996), “Risk-Sharing between
and within Families” Econometrica, 62:2, p.261-94.

Hill, Daniel, and Robert J. Willis (2001), “Reducing Panel Attrition” The Journal of
Human Resources, 36:3 p.416-38.

Jones, Gavin (1994), Marriage and Divorce in Islamic Southeast Asia, Kuala Lumpur:
Oxford University Press.

Levine, David and Michael Kevane (2003), “Are Investments in Daughters Lower when
Daughters Move Away? Evidence from Indonesia”, World Development, 31 :6, p.
1065-1084.

Lillard, Lee and Robert J. Willis (1997), “Motives for Intergenerational Transfers:
Evidence from Malaysia”, Demography 34, p.115-34.

Mace, Barbara (1991), “Full Insurance in the Presence of Aggregate Uncertainty”, The
Journal of Political Economy, 99:5, p.928-956.

McGarry, Kathleen, Robert F. Schoeni, (1995), “Transfer Behavior in the Health and
Retirement Study: Measurement and Redistribution of Resources within the
Family”, The Journal of Human Resources, 30, p. 8184-8226.

Ravallion, Martin and Loraine Dearden (1988), “Social Security in a “Moral Economy”:

An Empirical Analysis of Java”, The Review of Economics and Statistics, 70: 1,
p.36-44.

80

Rosenzweig, Mark R., Oded Stark (1989), “Consumption Smoothing, Migration, and
Marriage: Evidence from Rural India”, The Journal of Political Economy, 97:4 p.
905-926.

Strauss, John, Kathleen Beegle, Bondan Sikoki, and Agus Dwiyanto, Yulia Herawati and
Firman Witoelar, The Third Wave of the Indonesia Family Life Survey (IFLS3):
Overveiw and Field Report. RAND, Santa Monica, California.

Strauss, John, Gerrnano Mwabu and Kathleen Beegle (2000), “Intrahousehold
Allocations: a Review of Theories and Empirical Evidence”, Journal of African
Economies, 9, p.83-143.

Thomas, Duncan, James P. Smith, and Elizabeth Frankenberg (2001), “Lost but Not
Forgotten”, Journal of Human Resources, 36:3, p. 556-92.

Thomas, Duncan (1994), “Like Father, Like Son; Like Mother Like Daughter: Parental
Resources and Child Height” The Journal of Human Resources, 29:4 p.951-88

Thomas, Duncan (1993), “The Distribution of Income and Expenditure within the
Household”, Annales d ’Economie et de Statistique 29, p. 110-35.

Thomas, Duncan (1990), “Intra-household Resource Allocations: An Inferential
Approach”, Journal of Human Resources, 25:4, p. 635-64.

Townsend, Robert (1994), “Risk and Insurance in Village India”, Econometrica, 62,
p.539-92.

Weiss, Yoram (1997) “The Formation and Dissolution of Families: Why Marry? Who
Marries Whom? And What Happens Upon Divorce”, in M.Rosenzweig and O.
Stark (eds). Handbook of Population and Family Economics, vol, Amsterdam:
North-Holland.

Weiss, Yoram and Robert J. Willis (1985), “Children as Collective Goods and Divorce
Settlements”, Journal of Labor Economics 3:3, p. 268-92.

81

CHAPTER 2
DETERMINANTS OF HOUSEHOLD DIVISION:

A CASE FROM A DEVELOPING COUNTRY

2. 1 Introduction

When a daughter leaves her parents to set up a new household with her husband,
or a son sets out to other villages to ﬁnd employment opportunity outside his village, the
parents as well as the child may lose some of the beneﬁts associated with living in a joint
household. For instance, the ability to pool resources and to share consumption may
diminish as the child ceases to contribute directly to the household economy. When the
departing child also has some claims on the parents’ assets, for instance through
inheritance, the household may lose not only a potential source of family labor, but also
some of the capital used in household production. On the other hand, departure of adult
children may ease parents of the responsibility to provide for their children (and perhaps
for their children’s young children), and relieve the children from the obligation to take
care of their aging parents. In many cases, there will still be economic ties between the
children and the parents, for instance through transfers. In this case, having family
members residing in different region may in fact increase insurance possibilities for the
household against local risks, such as weather risks for agricultural households.

This essay looks at factors underlying household division in Indonesia. As argued
by Foster (1993), if household division in a society occurs solely because of marriage,
one could study the age of ﬁrst marriage to be able to predict household division.l But

studies have shown that the process of household division does seem to vary between and

82

even within societies. For instance, it is common in some parts of Indonesia for a young
married couple to live with either sets of parents “. . .until they are considered to be able
to manage their own affairs...” (Koentjaraningrat, 1985: p.133). Anthropological studies
on Indonesia have long concluded that there is no strong cultural preference regarding
with whom a newly married couple resides (Geertz, 1961, p.76, Jay 1969: p, 40-41,
Koentjaraningrat, 1985: p.133. See also Jones, 1994: 113). Indeed, the choice of
postnuptial residence and thus the timing of household division may depend more on
economic reasoning:2

“Economic considerations appear most important in the choice of residence [of
the newlyweds]: the ﬁnancial advantages to be gained, the number of single
children remaining in the two families, the parents’ need for extra help, or the
availability of land rights. . ..” (Jay, 1969: p, 40-41)

Economic considerations also can explain the practice in rural India of parents
marrying daughters to families residing in other regions. Rosenzweig and Stark (1989)
ﬁnd evidence that these marriages also serve as a way to reduce consumption variability.
The economic motive is the primary motive of young male adults in Malaysia who leave
the household (Smith and Thomas [1998], Johnson and DaVanzo [1998]), while moves
of young women are more related to fertility and family considerations (Smith and
Thomas, 1998). 3 Household division may also be induced by other, non-economic
events. Death of the head of household may cause households to split. Foster and

Rosenzweig (2002), looking at household division in rural India, ﬁnd that most

 

' A similar point is also argued by Johnson and DaVanzo (1998) who study nest-leaving in peninsular
Malaysia.
2 There is a slight preference towards uxorilocality (residence with the bride’s family) as opposed to
virilocality (residence with the groom’s family) (See, for example Jones [1994], Levine and Kevane
2003]).
See also Lucas (1997) for a review on models and empirical ﬁndings on internal migration in developing

83

household splits occur after the death of the head. Negative economic shocks may also
cause household structure and composition to change. For instance, F rankenberg, Smith,
and Thomas (2003) ﬁnd that households in Indonesia reorganize their composition and
living arrangement to cope with economic crisis. In short, household division is likely to
be responsive to economic incentives. This implies that decisions underlying household
division are likely to be endogenous with a number of household economic outcomes.

There is a potentially high value in understanding the process of household
division with regards to the increasing availability of longitudinal household surveys in
developing countries. Some longitudinal surveys based their selection of respondents
conditional on the residence in the previous wave or the baseline year (Thomas et a1
[2001], Foster and Rosenzweig [2002], Rosenzweig [2003]). When a household
interviewed in the baseline survey has divided, inattention to this fact may cause
inference based only on the characteristics of the remaining household members to be
biased. 4

It is therefore unfortunate that the overwhelming majority of the analyses of
households in the development literature treat household composition as ﬁxed and
exogenous (see Strauss and Thomas [1995] for a review on empirical modeling of
household decisions). While the importance of understanding the underlying process of

how household composition changes seems obvious, there are still very few studies that

 

countries.

4 For example, Rosenzweig (2003) demonstrates how the estimates of economic mobility among adult
males in Bangladesh are affected by household division. He estimates the wage of adult male in 2000 based
on household income, assets, and the maximum years of schooling in 1982. He frnds that since less able
male are more likely to leave the joint households, the relationship between schooling in the household in
the initial year (1982) with the status of young males in the household in 2000 is signiﬁcantly understated.

84

try to explicitly model how households divide.5 One notable exception is the study by
Foster and Rosenzweig (2002) in the context of Indian rural households. They model
household division explicitly and investigate how exogenous income growth interacts
with household structure and intrahousehold inequality to inﬂuence decisions on joint
residence.

This essay is an attempt to apply the model developed by Foster and Rosenzweig
(2002) to the Indonesian context. It is very appealing to choose Indonesia as a case to
study household division. First, Indonesia is a very diverse society with different
traditions and norms inﬂuencing decisions on household structure and living
arrangement. It is interesting to see whether we can uncover the underlying process of
household division by focusing on some of the key economic and demographic variables.
The availability of a longitudinal household survey of a high quality, namely the
Indonesia Family Life Survey, also makes it very appealing to look at household division
in Indonesia.

Using three waves of the Indonesia Family Life Survey (1993, 1997, and 2000), I
ﬁrst present a descriptive analysis of household division. I then estimate the determinants
of household division by using origin-household variables as explanatory variables in a
probit regression framework. Using 1993 as the initial period, I estimate the probability
that a household divides by 1997 and by 2000. I also estimate the probability that a

household divide by 2000, using 1997 as the initial period.

 

5 More have been written on the closely related literature on coresidency in developed as well as in
developing countries. The subject of household division is also naturally related to the literature on
household formation.

85

The ﬁndings suggest that, consistent with the model, the probability of household
division increases with the number of claimants in the household (household size).
Higher education of the head seems to be associated with lower probability of household
break-up. On the other hand, higher maximum years of education of non-head members
seems to be associated with higher probability of household division. While one could
explain this ﬁnding in terms of the collective household model, one could also interpret
the ﬁnding to be largely related to the migration of young household members, which can
very well be explained within the context of unitary household model.

The result shows that standard deviation of the household members education,
used as a measure for the intra-household inequality, does not explain the probability of
household division. High correlation between different measures of education makes it
difﬁcult to estimate and interpret the effects of education variables effectively.

Allowing for the values of land and business assets to be interacted with dummy
variable for urban region, it seems that land and assets are not important in determining
household division. Controlling for household head’s education and age, as well as the
age composition of household members, households with female heads tend to have
higher propensity to divide, especially in rural areas.

In addition, I also estimate the probability of household division of the panel
households (i.e., households appearing in all three waves of the survey). Estimating the
model using a pooled Linear Probability Model (LPM), the results also show the positive
association between household division and maximum years of education of the non-head
members. The results are similar when I estimate the model using LPM with origin-

household ﬁxed effects.

86

The essay is organized as follows. The next section will brieﬂy discuss previous
literature related to household division. The third section discusses the model developed
by Foster and Rosenzweig (2002). The fourth section brieﬂy discusses the Indonesian
setting and provides some description analysis of household structure and composition in
Indonesia using the IFLS data. Section ﬁve will describe the empirical strategy for the
multivariate analysis and discuss how the samples are constructed. The ﬁndings are

discussed in section six and I conclude the essay in section seven.

2.2 Previous Literature on Household Division

With the increasing availability of longitudinal household surveys, it is now more
than ever, possible to study what happens to household structure over time. While the
advantages of having longitudinal data are obvious, the question of how one deﬁnes a
household - not unambiguous even in cross-sectional setting - becomes increasingly
important. First, there is the question of whether it is more appropriate to focus on
household as the decision-making unit or to treat household as a collection of individual
decision makers, a question that is at the center of the literature on intra-household
allocations.‘S As discussed by Strauss and Thomas (1995), the question of appropriateness
is an empirical one and the answer really depends on the issue at hand. Although in some
cases it may be more appropriate to put the locus of decision making on individuals, in

other cases household decisions may be made at the household level, or even at a higher

 

6 See Strauss and Thomas (1995) for an extensive discussion on this issue. For review on the subject of
intra-household allocations including collective household models, see for example, the volume edited by
Haddad, Hoddinot, and Alderman (1997). Strauss, Mwabu, and Beegle (2000) review the theories and
empirical evidence on the subject. For a review of testing among interhousehold models, see Doss (1996).
See also Thomas (1990, 1993, 1994).

87

level such as the extended family.7 Secondly, as discussed above, the household itself is
a dynamic concept: household structure and composition changes over time.

One of the few studies that focuses on household division is a study by Foster
(1993) in Bangladesh. The study focuses on the institution of bari, a collection of related
households living in large compound. Foster uses the data on the characteristics of
households in 1974 and match them with the data from 1982 to identify households that
are partitioned and to see whether there are effects of household partition on household
outcomes such as child’s schooling. While the study does provide some evidence of the
effects of household partition on household outcome, it does not model household
partition explicitly.

A more recent study by Foster and Rosenzweig (2002) does explicitly model
household division. In the context of rural Indian households at the beginning of the
“Green Revolution”, they develop a model of household division to study how exogenous
income grth interacts with household structure and intrahousehold inequality to affect
household division. They ﬁnd that estimates of the extent to which households are better
off due to technical change in agriculture are overestimated when household division is
treated as exogenous.8 More recently, Rosenzweig (2003) adopt a similar approach to
study economic mobility in Bangladesh, taking into account the selection problem caused
by nonrandom household division. He ﬁnds that the effects of maximum education of
adults in the origin-household when the individuals were young on individuals’

educational attainment and earnings in later years are understated, and the effects of

 

7 In the ﬁrst essay of this dissertation (Witoelar, 2004) I investigate whether resource allocation decisions
are made at the extended family level.

8 Foster and Roszenzweig (2002) ﬁnd that technical change in agriculture not only affect household income
growth, but also affect the probability of household division. In particular, technical change, interacting

88

origin-households income are overstated, when one only looks at adult males in
households that were undivided compared to using sample of all adult males. The
patterns of bias in the estimates of economic mobility are consistent with the facts that
household division occurred nonrandomly. 9

To date, at least to the author’s knowledge, there have been no economic studies
that attempt to look at household division in Indonesia. There have been, however, some
related studies on living arrangement and co-residence in Indonesia, focusing primarily
on the elderly. Cameron (2000) studies the residence decision of elderly Indonesian,
using the 1993 wave of the IFLS. She found that children’s and parent’s demographic
characteristics play an important role in the residency decision. Cameron and Cobb-
Clark (2001) look at determinants of coresidency, ﬁnancial transfers, and the labor-
supply of the elderly in Indonesia. They ﬁnd that the characteristics of the elderly
parents’ children play a more important role than the characteristics of the parents
themselves. A more recent study by Frankenberg, Chan, and Ofstedal (2002) focuses on
the stability and change in living arrangement using longitudinal household data from
Indonesia, Singapore, and Taiwan. The study shows that characteristics found to be
associated with co-residence at the initial period (baseline interview) exhibit an even
stronger association with continued co—residence over time, suggesting some stability of
living arrangements over time. Note however that this study, as in other studies looking
at co-residence mentioned above also focuses only on elderly.

Finally, another closely related study of Indonesian households is the study by

 

with intra-household inequality and household size, reduce the probability of household division of
households with more land resources per capita.

9 Rosenzweig (2003) ﬁnds that household income in 1982 is positively associated, and maximum years of
education in 1982 negatively associated, with household division between 1982 and 2000.

89

Frankenberg, et a1 (2003) looking at household coping strategies in the face of economic
crisis. Using two waves of the IFLS (IF L82 1997 and IF LSZ+ 1998) they investigate
various ways by which households in Indonesia cope with the economic crisis that hit
Asia in 1997-1998. They ﬁnd that household dependents tend to move into households
residing in locations with lower costs of consumption (i.e., places less severely affected
by the crisis), while working age family members tend to move into households that are
able to absorb more workers. While this study does not model household division
explicitly, the ﬁndings illuminate the need to look at household division as an

endogenous decision.

2.3 Model

This essay adopts the model of household division developed by Foster and
Rosenzweig (2002). The model is a collective household model in which conﬂicts
among the household members over the level of household public good provided in the
household may lead to household division. As with the collective household models
developed by Chiappori (1988, 1992), the model assumes that intrahousehold allocations
are efﬁcient. However, unlike most of the previous collective household models where
household composition is assumed to be ﬁxed and exogenous, here household division is
explicitly modeled as an outcome of household members’ decision making.

The model was developed in the context of rural Indian farm households.10 Foster
and Rosenzweig speciﬁcally look at the period of the Indian “Green Revolution” between

1960 and 1980s — the period during which new agricultural technologies were introduced

90

in India - to study the effects of technical change on household division. Focusing on this
particular period, they investigate how exogenous income grth interacts with
household size and intrahousehold inequality to affect household division.

In the following section I brieﬂy review the model, illuminating some key
assumptions and comparative static results that will be relevant to this essay. The readers

are referred to the original paper to see the complete exposition of the model.

2.3.1 A Model of Household Division

According to the model, household members may beneﬁt from residing in a joint
household by sharing the cost of household public goods, taking advantage of economies
of scale in production, and by sharing information on the best-practice farming
technique. ” On the other hand, gains from joint household may be offset by a direct
preference to live separately (i.e., desire for privacy)”, the possibility of diseconomies of
scale in production, and also by the increased insurance possibilities associated with
interhousehold transfers from departed members. 1 3

A joint household j consists of N individuals (i = l, ..., N). These individuals are

claimants, who have property rights over a divisible asset that produces income stream to

 

'0 This implies that the households allocate time and resources in both production and consumption
activities. See Singh, Squire, and Strauss (1986) for a review of the literature on agricultural household
models.

” A study by Foster and Rosenzweig (1995) models the learning spillover among rural Indian households
during the Green Revolution.

’2 Preference to live in private has long been acknowledged in the literature on co-residency and does not
seem to apply only to younger individuals seeking independence from the parents. A study by Costa (1997)
on retired Union Army veterans in the US. shows that rising incomes of the elderly were the most
important factor for the elderly to live alone. Frankenberg et a1 (2003) ﬁnd that in Indonesia privacy may be
substitutable with consumption from joint household, as shown by the ﬁnding that sub-households
recombine to minimize the impact of the economic crisis.

'3 This is an important consideration since there is evidence that households in rural India take advantage of
marital ties between family residing in different regions to deal with variability of income (Rosenzweig and
Stark, 1988).

91

the household.14

Each claimant i in a joint household j, has his own nuclear family, described by

the vector n,- and has the utility of:

“ij =u(x,j,zj,r',-j;n,-j) (I)
where xi!- denotes private good, 2] represents household public goods, and rij the
household structure. The presence of rij allows for direct preference for residing

separately from the joint household. The vector ny- represents the characteristics of

claimant i’s nuclear family in the joint household j.
The budget for the joint household is given by:

N N
2:11:3- + 2,. = y,- (2)

where yjN is the joint household income. If an individual i lives separately (possibly with
his own nuclear family), he will earn “autarchic income” y). The joint household

income yjN is not the same as the sum of each claimant’s autarchic income, Zy,.'5 Indeed,
the expected gains ﬁom the joint household depend on the expected value of the
difference between the two.

The time sequence of household decisions is as follows. Household members

make decisions about joint residence based on the expected utility maximization problem

 

‘4 Each claimant has a share Ky- of the household production asset A j.

'5 The autarchic income y,- is a function of claimant i wages Wi, his claim of asset try-Aj- times the individual
speciﬁc productivity factor 19,-, income shock e,, and transfer, 172 y,- =19,-K,-J-Aj + W,- + e,- +r. Joint household
income is yJ-N =6}NAj + N W,- + ejN+r,-N. Note that this implies that each claimant earns the same labor
income W), even though there is an individual-speciﬁc productivity factor 19,-. This productivity factor us

ssumed to affect joint household income through household assets A j.

92

above, before income shocks are revealed. After income shocks are realized, they make
decisions about consumption of private and public goods. It is also assumed that,
conditional on residence and income realizations, intrahousehold consumption is ex ante
efﬁcient, i.e., it is not possible for any claimant to have higher expected utility without
decreasing the expected utility of some other claimants. As noted by Foster and
Rosenzweig (2003), this timing sequence amounts to assuming ex post efﬁciency in the
sense that for each pair of residential choice and income realization, no claimant can
obtain higher utility without decreasing the utility of some other claimants. The last
restriction is that each claimant has to obtain an ex ante expected utility that is at least
equal to his reservation utility. The household will divide whenever this last condition
cannot be satisﬁed.

The gain from living in the joint household to each claimant is the difference
between expected utility achieved by the claimant in the joint household and his
reservation utility subject to the conditions that each other claimants receive his
respective reservation utility. In other words, for claimant 1, the expected gain from joint
household is:

E(u(x[,z,r1;nl)—Ev(y1,r1,n1) (3)
where 1102;, r), m) is the utility that claimant 1 would have received had he lived
separately. Claimant 1 maximizes (3) with respect to private good x1 and public good z
and subject to:

E(u(x,-,z,r,-;n,~) 2 Ev(y,~,r,-,'n,-) for i = 2,...,N, (4)

N

and the joint household budget: 2:; xi]- + 2,]- = y

As long as the gains ﬁom joint residence given by equation (3) is positive, the

93

joint household remains intact.

2.3.2 Parameterization of Preference and Some Comparative Static Results
Foster and Rosenzweig (2002) then assume a parametric utility function for each
claimant of the form:

u(x,z;n) = ln((x — ,B'n — az)(z + 7)) + 6r, (5)
where r,- is one for autharchic households and zero otherwise.16 This utility ﬁrnction

exhibits transferable utility, which, as discussed in Bergstrom (1996), implies that
distribution of assets within the household does not affect public good consumption.17
This is a key assumption since this implies that distribution of assets does not affect the ‘
decision to break up from the household, or at least not through the disagreement over
household public goods.

From the ﬁrst order conditions, Foster and Rosenzweig (2002) show that
claimants with differing characteristics or incomes will demand different levels for public
and private goods when they co-reside than the levels they demand when they live
separately.‘8 In particular the consumption of public goods in an autarchic household is
given by:

1

Z=m(y—ﬂ'n—7(1+a» (6)

while the consumption of public goods in a joint households is given by:

 

'6 Thus 6 represents direct preference for living separately.
'7 The necessary and sufﬁcient condition for a utility function with one private good and n public goods is

that the preferences for each household member can be written as Ui=f(z)x,- + gi(z) (Bergstrom, 1996).
‘8 The reader is referred to Foster and Rosenzweig (2002) for the complete derivation of the ﬁrst order
conditions and comparative statics.

94

_ 1 N_ N ._
Z_2(1+Na)(y EMF" WM”) (7)

It is then assumed that there only two states of the world; a good state, and a bad
state, with the corresponding income shock +A and -A. These shocks along with a
multiplicative factor Q affect income additively. In addition, in a bad state of the world,
claimant i will receive a net transfer 2'.- and in good state of the world, he will send a net
transfer of 2',- (see footnote 15). Thus, in a good state the autarchic income is

37i + 5A — 2'i , while in a bad state it is y, — 5A + ri . The joint household incomes in good

and bad state of the world are y” + {AN -— r" and y” — 4%” + r" , respectively.

Assuming ﬁrst that the claimants are homogenous, by solving the maximization

problem above, one can obtain an expression for the gains ﬁom the joint household as:

1 2
91* 0,N,A,W, 'n =————6A+NW—N 'n+ 1+Na
( ﬂ ) 4(1+Na)[ 13 r( )]

——1—N[oﬁ+W—Np'n+ (1+a):|2
4(1+a) N 7

(8)
Some comparative static results, as discussed in Foster and Rosenzweig (2002)

are as follows:

atllt

1. < 0. An increase in the number of claimants given the amount of

 

household asset will reduce the gains from joint household, and thus increase the
propensity of the household to break up.19 An increase in the number of claimants, given

the same amount of household assets decreases per capita income and thus decreases the

 

'9 This will be true as long as the demand for private good is high relative the wage, and public good
consumption is at an interior solution (see Foster and Rosenzweig 2002).

95

gains from the joint household. A larger number of claimants may also mean a lower
average cost of household public goods, but under the speciﬁed conditons, the increased

saving is not enough to offset the decrease in per capita income.

lil¢

a
2. si
g" on

 

= sign 56—2— . The effect of any element in n (characteristics of the
k ”k

nuclear family) on household division depends on whether the element increases or
decreases consumption of household public goods. The increase in the number of
children of a claimant, for example, will decrease the demand for public goods relative to
private goods (e.g., children’s clothing), and thus the effects on the joint household
surplus will be negative.

atpt

3. > 0. An increase in household assets will increase the joint household

 

surplus and thus decrease the propensity of household division.

all!!!

4. >0. An increase in productivity will increase the gains from joint

 

household and discourage household division.
When claimants are allowed to be heterogenous, the gains from the joint

household can be written as:

1 _ 2
=WHW —Zil:lﬂ'n+7(1+Na)] 4:21” —rN)2]

_ Ne”
4(1 + a)
(9)

(varbN —- ,B'n)+mean()7N - ,B'ny(1 + 05))? — mean(é’AN - I'D/)2)

Higher intrahousehold variance of income implies a greater difference among the

claimants over the demand for public goods. In particular equation (9) shows that for a
given mean of y,- -,B’n,- and yN -2ﬂ’n,- , and increase in the variance of y,- -,6”n,- will

96

reduce gains from joint households \p and increase the propensity of household division.
Higher intrahousehold variance of y,- -,B ’n, implies greater disagreement over the level of

household public goods demanded. Since the consumption of public good must be equal
among claimants in a joint household, the higher variance increases the probability of

household break up.

2.3.3 Empirical Strategy

In the empirical part of the study Foster and Rosenzweig (2002) estimate the
probability of household division using a probit model with dependent variable equal to 1
if a household split by 1982. They used characteristics of the same households from 1971
survey as explanatory variables.20 The claimant is deﬁned as the head of household,
sons, and brothers of the head; claimants are male. The base speciﬁcation includes the
number of claimants, the number of claimants’ wives, age of head, means of schooling of
claimants, the number of children by sex, and household landholdings, The results show
that the number of claimants (and claimants’ wives) is positively associated with
household division. The number of boys (future claimants) is also positively associated
with household division while the number of girls does not have a statistically signiﬁcant
effect.

In the next speciﬁcation, in addition to means of schooling of the claimants, they
add other education variables, namely the maximum years of schooling of the claimants

and the variance of the claimants’ schooling. They argue that these variables are

 

2° By using 1971 characteristics to summarize the initial conditions, they implicitly assumed that 1971
household characteristics are given. To some degree, they deal with the problem by adding as explanatory
variables the number of previously departed children in one of their speciﬁcations.

97

measures of intra-household inequality.21 According to the model, variance of schooling
is positively associated with the probability of household division. On the other hand,
higher maximum education is associated with higher household surplus and thus lower
probability of household division. Their empirical results show that indeed the variance
of schooling appears to be associated positively and maximum of schooling negatively

with household division.

2.3.4 Discussion

The model provides a framework in which household division is seen as a process
that is associated with conﬂicts over household public goods. As with Chiappori’s
model, household members are assumed to make Pareto efﬁcient decisions. 22 The model
is also in line with some other models of intra-household allocations. For example, public
goods also play a central role in cooperative bargaining models developed by McElroy
and Horney (1981), McElroy (1990), and Lundberg and Pollak (1993). 23 However,
neither Chiappori’s collective household model nor the cooperative bargaining models
mentioned above explicitly model household division. Household composition is

assumed to be ﬁxed and exogenous. In contrast, the model developed by Foster and

 

2' It is not clear, however, how one should interpret the coefﬁcients of the education variables when the
mean, the maximum, and the variance are included together as explanatory variables. Furthermore,
controlling for two of the education variables, there is not much variation in the third education variable in
the data, especially when the number of claimants is small. Consider a sample where all households have
only two claimants. For any given pair of mean and maximum years of education, there is no variation in
the variance of schooling.

2 In a typical Chiappori’s collective household model, household allocation decisions are ﬁrst made to
decide the level of household public goods. Decisions on the expenditure of private goods are then made
according to a sharing rule, using what is left of the household income.

23 In McElroy and Homey’s cooperative bargaining model, individuals in marriage receive utility from a
household public good in addition to consumption and leisure. They solve a Nash-bargained game in which
the threat point is the utility outside the marriage. McElroy (1990) introduces the concept of EEP (extra-
household environmental parameters), parameters that shift the threat points. Lundberg and Pollak (1993)

98

Rosenzweig above enables us to derive and sign the effects of changes in household
characteristics and income, as well as intrahousehold inequality, on household division,
providing the basis for the empirical work.

Unfortunately, some of the assumptions made in the Foster and Rosenzweig
model above are unquestionably strong. While the use of the utility function with
transferable utility helps to make the model tractable, it is inconsistent with some
assumptions of previous models of collective household. The assumption of transferable
utility assumes away the importance of distribution of assets in intrahousehold
allocations. The assumption is also not in agreement with some recent empirical ﬁndings.
As an example to the contrary, a study by Thomas, Contreras and Frankenberg (2002)
ﬁnd some evidence that assets brought into marriage by husbands and wives affects child
health differentially.24

The model also assumes that the magnitude and the sign of the gains from living
in a joint household are independent of the particular identity of the maximizer and they
are also independent of the decision of how the gains are distributed. This is somewhat in
contrast with ﬁndings showing that intrahousehold allocations are affected by
individuals’ exogenous income, such as non-labor income (Schultz [1990], Thomas
[1994]), or transfers (Lundberg, Pollak, and Wales [1993])

While the model above allows for a special role of the head of the household (by
allowing some elements of n to be characteristics of household head), the empirical

speciﬁcation does not differentiate the education of the household head from that of the

 

introduce the notion of “separate sphere” in which the threat point is not divorce, but the traditional gender
roles insidde marriage.

24 A review by Quisumbing (2003) discusses the recent development in the literature with regards to the
role of resources and power in the household.

99

other claimants. It is true that the role of household head’s education in intrahousehold
allocations is somewhat controversial. 25 There have been some discussions on whose
education matters the most in the household and some authors suggest it is the maximum
years of education that matters the most (Jolliffe [1997], Foster and Rosenzweig
[1996]).26

One the other hand, with regards to decisions of the joint household on public
goods allocation, one could argue for a case where the head is altruistic towards
household members (an assumption dating back to Becker’s “rotten kid theorem”
[1974]). As acknowledged by Foster and Rosenzweig (2002) the head of household may
identify more with the joint household, and thus will tend to allocate more resources
towards household public goods, discouraging household dissolution. In addition,
younger, non-head adults in the households with higher education may have employment

opportunities with higher earning outside the joint households.

2.4. The Indonesian Settings and Household Division in the IFLS
2.4.1 The Indonesian Settings

In adopting the model to the case of Indonesia, it is important to acknowledge
some important differences from the rural Indian context. In particular, it is not clear how
one should deﬁne a “claimant” in the Indonesian context. The law and norms inﬂuencing

household behavior in Indonesia and in particular with regards to claims on inheritance,

 

25 It is also true that the deﬁnition of the head in household surveys is sometimes arbitrary

26 Jolliffe (1997) uses data on households in Ghana to tests which of the following education variables
matter most in determining household income: household head’s schooling, maximum schooling of adults,
or average schooling of adult. The results show that either maximum or average schooling of adults is a
better measure than schooling of the household head. Foster and Rosenzweig ( 1996) ﬁnd that maximum
years of education is a better predictor of the adoption of the new agricultural technology during the Indian
“Green Revolution”.

100

distribution of resources within households, as well as co-residency are vastly different
from that in rural India, as will be discussed below. Furthermore, since the IFLS includes
households from 13 provinces in Indonesia, it is not clear what local norms and
traditional law operate in each community.

To better deﬁne the appropriate set of claimants, I will discuss brieﬂy some of the
institutional settings that are most relevant, namely the norms on postnuptial residence,

inheritance, and the incidence of marital dissolution.

Postnuptial Residence

If one practice dominates the pattern of postnuptial residence, for example, if
most marriages are virilocal, then for the purpose of the empirical work, one may need to
treat households with adult sons differently from households with no adult sons. As
discussed above, postnuptial residence varies among Indonesians and sometimes depends
more on economic considerations than some strict traditional law or local norms.
Anthropologists usually classify the Javanese (which constitute the majority of
Indonesian population) as ambilocal (reside with either set of parents) with some
preference toward uxorilocality (Geertz [1961], Jay [1969], Koentjaraningrat [1985],
Jones [1994]). Levine and Kevane (2003) study the variations in residence aﬁer marriage
based on the information on the local norms and traditional law. Using a special module
on adat (traditions) from 1997 IFLS they ﬁnd that there is a lot of variation between
communities and ethnic groups: daughters tend to reside with or near their parents in
around 53 percent of the regions, with or near their husband’s parents in 23 percent of the

region, and in about 23 percent communities new couples tend to live with or near either

101

set of parents. 27

Inheritance

In the context of rural India, inheritance customs determine adult males as heirs
(Foster and Rosenzweig 2002). It is less clear how the inheritance system operates in
Indonesia, although generally sons and daughters have some claims on their parents’
estates.28 The kinship is bilateral, ’ in the sense that one is equally related to father’s side
of the family as to the mother’s side (Jones, 1994). For the majority of the Indonesian
moslem population, inheritance is governed by Islamic law, which states that sons should
inherit two-thirds of their parents’ estates. But at the same time, Javanese traditional law
states that sons and daughters should have equal shares. In practice, some follow each
while others pay no attention to either (Geertz [1961], Jones [1994]). While these
ﬁndings do not offer any guide on how the inheritance is shared, what is unambiguous is

that both men and women have some claims on the parents’ assets.

Marital Dissolution

How assets are distributed when a household is divided may differ between the
case where a child leaves the household and the case where the division is due to marital
dissolution. Among some ethnic groups in Indonesia such as the Bataks and the
Makassarese, the husband will lose the payment made to the wife at the time of marriage,

while among the Javanese and the Minangs, the husband and the wife keep what they

 

27 Balinese and Sasaks reported 94 percent and 83 percent of the marriages as virilocal, respectively, while
Javanese, Bugis, and Minangs reported 53 percent, 71 percent, and 77 percent of the marriages as
uxorilocal.

102

brought into marriage (Jones [1994]: p. 220).

If the incidence of divorce is high, it may be necessary incorporate marital
dissolution into the model or the empirical speciﬁcation. As a matter of fact, Indonesia
used to have a very high rate of divorce: 13 per 1,000 population age 15 and above in
1960, compared to 1.8 in developed countries in the same period (Jones [1994]; p. 180).
However, the rate has decline to 4.6 by 1975 and 1.1 by 1990. A study by Heaton,
Carnmack, and Young (2001) uses data from IFLS 1993 and the Indonesian Fertility
Study (IFS) to study the dramatic decline. They ﬁnd that education variables are
becoming important in predicting marital dissolution while age at marriage and marital

duration are becoming less so. 29

The discussion above provides some insights on how the set of claimants should
be deﬁned. In contrast to the rural Indian context, in the Indonesian context one should
consider female household members as potential claimants. With regards to the
widespread practice of virilocality and uxorilocality, one possibility is to include not only

sons and daughter as potential claimants, but also sons- and daughter-in-laws.

2.4.2 Descriptive Analysis of Household Division in IFLS

The IF LS provides an excellent opportunity to look at the incidence of household

 

28 According to Jones (1994; p. 114), it is typical for the parents to inherit their home to the youngest
daughter, with the expectation that the daughter will care for the parents.

29 Previous authors (for example Jones [1994]) have identiﬁed some key factors explaining the decline in
the past decades: the increase in the age of marriage, the decrease in arranged marriage, and the passing of
the 1974 Marriage Law that makes it more difﬁcult for a husband to divorce a wife (this law also sets the
minimum age of marriage of 19 for male, and 16 for female).

103

division among Indonesian households.30 Since the survey tracks and interviews some
respondents when they leave the households and set up a new household in the
subsequent wave, I can identify households that have divided from the baseline interview
by the subsequent wave. In the ﬁrst wave, which was conducted in 1993, 7,224
households were interviewed. In 1997 the number of households interviewed was 7,619,
and more than 11 percent of those households are split-off households. When the IF LS3
was conducted in 2000, the number of splitoff households (including those that split in
1997 and 1998) accounts for around 35 percent of all households interviewed (see Table
2.4.1). In 1997, around 11 percent of the 1993 original households have divided. By
2000, the percentage is around 36 percent (Table 2.4.2). Using 1997 as the base year,
the table shows that of 7,619 households that were interviewed in 1997, around 21
percent have divided by 2000.

Does the household composition differ between the original and the split-off
households? The fraction of nuclear households, households that consists of at most the
head, spouse, and their children, is similar between the original households and the split-
offs, with around 62-66 percent households in either category. From Table 2.4.3, it is
clear that one would more likely ﬁnd intergenerational coresidency in original
households than in splitoff households.

Table 2.4.4.a-4.b shows the status of 1993 non-head household members in their
respective 1997, 2000 households. While most of them were still non-household

members in 1997, some of the individuals have moved on to form new households, and

 

3° See the ﬁrst chapter of this dissertation (Witoelar, 2004) for a brief description on IFLS and the tracking
rules used by the IFLS. For a full documentation of IFLSl see Frankenberg and Karoly (1995). See
Frankenberg and Thomas (2000) for a full documentation of IFLSZ, and Strauss, et al (2004) for a full
documentation of IFLS3.

104

some others become heads of the original households. Of the 1993 female non-heads
who became heads of the original household by 1997, most were spouses of the head.
This is consistent with the case where the woman (possibly the spouse of the head)
became the household head when the head died. On the other hand, of the 1993 male
non-heads who became head of the splitoff households, most were sons of the heads of
the original households. Table 2.4.4c similarly shows the number for the 1997 non-
household members in their respective 2000 households.

Figure 2.4.1 shows the relationship to the household head by age and sex in 1993
and 2000. Even though the ﬁgure only provides a snapshot of a cross section of the
population in a given year, one can infer some life cycle pattern with regards to
household living arrangement. Until they are 17 years old, most children are sons and
daughter of the household head (the other signiﬁcant portion is ‘other’, which also
includes grandchildren, nieces, and nephews). A small fraction of men age 18-23 are the
head of households, while a bigger fraction of women at that age are married to the head
of household. By the age of 23-29 these fractions increase for both sexes. The fraction of
men residing as son-in-law of the head and women as daughter-in-law are also the
highest in this age group, suggesting that in this age group , many of the individuals are
residing with their parents-in—law. The fraction of men and women living as sons- and
daughters-in-law is smaller among the 30-35 years old, and virtually disappear for the
older age groups, suggesting that the married children who reside with either set of
parents eventually leave the household or take over the headship themselves (or became
spouses of the heads). This seems to conform the previous anthropological ﬁndings of the

practice of coresidency after marriage.

105

Table 2.4.5a and 2.4.5b show the household headship status of 1993 individuals in
1997 and 2000 by age (and similarly, Table 2.4.5c for 1997 individuals in 2000). The top
panel of Table 2.4.5b shows that most of the males who became heads of the 2000 split-
off households were 25-29 in 1993. It is also shown that some of those who were 12-13
in 1993 have already become head of the splitoff households by 2000, or when they are
19-20 years old.

As noted before, if marriage is the only reason why young adults decide to leave
their household, then the study on household division is just a study of age of ﬁrst
marriage. The IFLS asks a question of what was the reason why a household member left
the household. The question is answered by a household member who remains (usually
the head). Table 2.4.6a tabulates the answers from the 1997 interview of why 1993
household members left their original household. Among men, ﬁnding work seems to be
the most important reason, while among women, to follow spouse or parent and marriage
seems to be the primary reason. Table 2.4.6b tabulates the answers given in the 2000
interview. In both years note also the numbers of individuals leaving due to marital
dissolution are relatively small.

How far did the individuals move? Table 2.4.7 shows the number and percentage
of 1993 respondents who are found in split-off households in 2000 by the location of the
split-off households relative to their 1993 households. Around 60 percent of 1993 male
respondents who were found in split-off households have moved out from their original
village. The number is somewhat similar with females.

Are there differences in the 1993 demographic and education characteristics

between households that remain intact by and the households that eventually split by?

106

Table 2.4.8 summarizes some of the key household-level demographic and education
variables of the 1993 households by the status of the household in 2000. One thing to
note is that the mean of household head’s education is slightly higher in the households
that remain intact by 1997. The maximum education of both male and female are higher
in households that eventually split. The standard deviations of age and of education are
also higher in households that eventually split.

Differences in household assets are shown in Table 2.4.9. It seems that
households that eventually split have higher initial value of landholdings as well as other
household business assets. In per capita terms, households that remain intact seem to
have higher per capita landholdings. Households with more non-business household

assets in 1993 seem to be more likely to eventually split by 2000.

2.5 Empirical Speciﬁcation

Following Foster and Rosenzweig (2002), I estimate the probability of household
division by using a probit model with characteristics of the households at the base year as
explanatory variables. Speciﬁcally I estimate the probability of household division
between 1993 and 1997 using 1993 household variables, between 1993 and 2000 using
1993 household variables, and the division between 1997 and 2000 using 1997 household
variables.3 1

Based on the discussions above, I deﬁne two sets of samples according to
different deﬁnitions of claimants. The ﬁrst set of claimants, ‘Claimant 1 ’ , include the

head, child of the head or sibling of the head who would have been 19 years old by the

 

3' Speciﬁcally, the dependent variable equals one if household have split by 1997 (or 2000), and zero
otherwise. I am ignoring household that were not found in the subsequent surveys, and I also do not take

107

time the household was next observed. The second deﬁnition of claimants, ‘Claimant
2’, includes not only the head, head’s child or sibling, but also any household members
who would be 19 by the year the household is next observed. The latter deﬁnition is
broader and may include in-laws as well as other members of the household.32

The choice of 19 as the cutoff age is somewhat arbitrary, although the descriptive
analysis above suggests that it is at the age of 19-20 that the fraction of males who
become heads of the splitoff households starts to increase.33 The restriction of age 19 or
more implies that when I estimate the probability of household division by 1997 for the
1993 households using Claimant 1, the claimants are the head’s children or siblings who
were at least 15 years old in 1993. Similarly they have to be at least 12 years old in 1993
to be deﬁned as a potential claimant when I estimate the probability of household
division between 1993 and 2000 , and 16 years old when I do the estimation for
household division between 1997 and 2000. Note that if I use the same deﬁnition of
claimants as Foster and Rosenzweig (i.e., sons or brother of the head), I would miss a
signiﬁcant fraction of households that divided due to other members leaving. 34

The sample sizes, alter dropping a small fraction of households who have missing
values in the key variables, are shown in Table 2.5.1. Note that there are 6 sets of

samples: three for each deﬁnition of claimants.

 

into account the possibility of households that have divided to join back together.

32 Around 45 percent of males who were sons-in-law of the household heads in 1993 became the heads of
the same households by 2000.

33 Age 19 is also the legal minimum age for marriage for males, although the limit seems to be non-binding.
Using data from the Indonesian Population Census, Jones (1994) estimate the mean age at ﬁrst marriage
for Indonesian rrrales to be 24.1 in 1980 and 25.4 in 1990.

3‘ I estimate the model using the set of claimants deﬁned similarly to that in Foster and Rosenzweig
(2002). Namely, the claimants are deﬁned as the household head, the head’s sons and brothers. Among
those, similar to the other deﬁnitions of claimants in this essay, I choose those who would have been 19 by
the time the household is observed as potential claimants. The results are presented in Appendix Table
2.6.14 and 2.6.15.

108

Table 2.5.2a —— 2.5.20 shows the summary statistics of the key variables used in
the estimation. Table 2.5.3a -— 2.5.3c shows means and the difference in the means of the
key variables for the households used in the estimation. It is worth noting that the
difference between the means of the claimant’s standard deviation of schooling is
signiﬁcant for Claimant 2 in the sample used to estimate the probability of household
division between 1993 and 2000 (Table 2.5.3b).

In the base regresion speciﬁcation I include the number of claimants, and the
proportion of claimants by age, the number of claimants, the number of young children
by sex. These account for the household demographic composition and household size.
Furthermore, young children represent ﬁrture claimants. I also include the age and age
squared of the head, and a dummy variable identifying male-headed household. I include
the log of the value of the land owned as well as the log of the value of the business

assets for farm and non-farm businesses.35

Education Variables

In addition to household size and composition variables, age and age squared of
the head, value of land and other business assets owned by the households, and dummy
variables indicating urban location and male headed households, the base speciﬁcation
includes two education variables: education of the head, and maximum education of non-
head claimants.36

In the other speciﬁcations I add the standard deviation of education among the

 

3’ Farm and non-farm business assets include house or buildings, vehicles, other equipment, and other. For
farm businesses, hard stem plans, livestock/poultry/ﬁshpond are also included, and for non-farm
businesses the assets also include supplies/merchandise, ofﬁce equipment, and other.

109

claimants (including the head) as an additional explanatory variable. As previously
discussed, this variable provides some measure of inequality among the claimants. In
another speciﬁcation I also include the means of education of the non-head claimants.

There is a problem associated with using all of these measures of household
education at the same time: these values are highly correlated. In particular maximum
years of education of non-head claimants is highly correlated with the mean years of
education especially when the number of non-head claimants is small. Table 2.5.4 shows
the comparison of the various measure of household education using all households in
IF LSl, by the number of adult in the household.37 Note that the number of households
with only two adults is 2,905, or 43 percent of all households with multiple adults.
Column (6) shows the correlation between the maximum years of education of non-head
adults with the mean years of education of non-head adults. Even among households with
more than 6 adults, the correlation is 0.832. The correlations between other measures of
education are also high. While the table reports the comparison for all 1993 households,
the same is true for the households in the estimation sample. 38

Because of this problem, in most speciﬁcations I only include at most three
measures of education: household head’s education, maximum education of non-head
claimants, and the standard deviation of education among the claimants, including the
head.

In addition to the education variables I also estimate the model with various sets

 

’6 Information in schooling is collected on the highest education level attended and the highest grade
completed at the level. The information is then converted into a variable on completed years of schooling.
The value ranges from 0 (no schooling or not completed ﬁrst grade) to 17 (university graduate).

37 The total number of observations 1s 7, 223 instead of 7 ,224 since one household does not have any adult.
(The head of household 18 14 years old).

38 In fact, the correlation between the maximum education of non-head claimants and the mean is above
0. 90 in each sample.

110

of interaction variables. Most importantly, I include the interaction of the dummy
variable of urban/rural residence with the log value of land and log value of business
assets, to control for region effects.39 The nature of the business that the households are
involved in may be different depending on where the households reside (e. g, rural
households are more likely to be engaged in farm business), and consequently land and
other business assets can affect household division differentially. In addition I also
estimate the full model using dummy variables for province, urban, and their interactions.
Finally, I estimate the model using dummy variables for community in place of the

province-urban interaction variables.40

2.6 Results

Table 2.6.1-2.6.3 show the results of the probit regression of household division
between 1993 and 1997, between 1993 and 2000, and between 1997 and 2000,
respectively, using the Claimant 1 sample. Table 2.6.4-2.6.6 report the result using the
sample of Claimant 2. The results from separating the urban and rural households for
each sample are reported in Appendix Table 2.6.1-2.6. 12. The marginal effects of a
change in explanatory variables are reported in these tables. The standard errors are
corrected for clustering at the base year community level and heteroskedasticity.

I am primarily interested in looking at the probability of household division
between 1993 and 2000 (Table 2.6.2 and 2.6.5) since the longer period allows for more

variation in the sample. As expected, the probability of household division is positively

 

’9 For the interaction terms, I use the demeaned value as opposed to the original value of the variable.
4° 1 also estimate separating the sample between urban and rural. For the rural sample I add a dummy
variable equals to one for farm household, zero otherwise. I also interact this dummy variable with the
education variables.

lll

associated with household size (the number of claimants and the number of non-
claimants) across all speciﬁcations, using either Claimant 1 or Claimant 2 as the sample.
For the sample of Claimant 1, assets variables seem to be associated positively with
household division (Table 2.6.2) , in contrast to the model’s prediction. However, as
shown in the third column of Table 2.6.2, when the business assets variable is interacted
with the dummy variable indicating urban residence, the coefﬁcient on business assets
become not statistically signiﬁcant with z-statistics = 0.514 (the p-value of the F-test of
joint signiﬁcance of the land variables is 0.07). This suggests that most of the assets
impacts were actually capturing urban — rural differences. Business assets does not
signiﬁcantly affect household division when I use Claimant 2 as the sample (Table 2.6.5).
The value of land owned does not appear to be signiﬁcant in determining household
divison in either sample.

The dummy for urban is negative and statistically signiﬁcant throughout the
results (see Table 2.6.2 and Table 2.6.5), suggesting that households in rural areas are
more likely to split than their urban counterpart.

Presence of children seems to be positively associated with the possibility of
household division. One explanation could be that these children are future claimants,
and as is with the number of claimants, the number of future claimants increases the
propensity of household to break up. Another explanation that is consistent with the
model is that the presence of children in the household implies that there is an increase in
demand for private goods (e. g. children’s food and clothing) relative to public goods, thus
encouraging household division.

Moving on to the education variables, the results show that the maximum years of

112

education of the non-head claimant to be positive and statistically signiﬁcant in most of
the speciﬁcations. Table 2.6.5 shows that the coefﬁcients on the maximum years of
education are close to 0.010 and statistically signiﬁcant, while the coefﬁcients on
education of the head are for the most part negative at around -0.004. The negative sign
on the coefﬁcient on head’s education is somewhat consistent with the story that heads
with higher education have more control over household resources and may be able to
increase the gains from joint household. It is interesting that for the most part, standard
deviation of education does not appear to be statistically signiﬁcant. In fact in some of the
speciﬁcations the coefﬁcients are negative, contrary to the prediction of the model.

The positive sign on the coefﬁcent on maximum years of education of the non-
head claimants implies that the higher is the maximum years of education of claimants,
controlling for household heads’ education, the more likely it is for the household to
break up. The results somewhat differ from previous ﬁndings in India (Foster and
Rosenzweig, 2002) or Bangladesh (Rosenzweig 2003) where the maximum years of
education is associated negatively with household division.41 Note however that in each
of these previous studies, head’s education is not separately controlled for, and the
maximum years of education include education of all claimants including the head.

One possible interpretation of the positive relationship between maximum years
of education and the probability of household division in Indonesia is that household
division may be primarily driven by the migration of the young, more educated, adult

males from the household.42 It is most likely that household members with the maximum

 

4' In Foster and Rosenzweig (2002) and Rosenzweig (2003) maximum years of education is positively
associated with surplus of joint household.

‘2 This is in contrast to the study by Rosenzweig (2003). He argues that among the households in rural
Bangladesh, education lower than the maximum is redundant in the households, so members with lower

113

years of education in the household are the young adult males. This, coupled with the
ﬁndings that rural househlds are more likely to divide, is consistent with the migration
interpretation. As has been shown in Table 2.4.7, around 60 percent of the 1993 male
respondents who are found in split-off households in 2000, reside outside their original
village.

In the context of the collective model used in this essay, the results indicate that
the higher is the education of non-claimant members, the higher is their autarchic
incomes, and the more likely that this members will leave the joint household. On the
other hand, and perhaps more interestingly, the results are not necessarily inconsistent
with the unitary household model. Indeed, the departure of highly educated members is
can be well explained within the context of unitary household. Households may send
their most educated sons to ﬁnd better opportunities elsewhere, and have them transfer
remittances back to the household, for example.

To interpret the results with respect to what happens with the female members is
more complicated. In particular it is not very clear how marriage of daughters would
affect household division in Indonesia. On one hand, as with males, higher educated
females may leave the household to earn higher earnings elsewhere and thus increase the
probability of household division. On the other hand, higher educated females marry later
in life than their lower educated counterparts, and thus lower the propensity for
household division.

Another interesting ﬁnding is that, in some speciﬁcations, male-headed

households tend to have lower probability of splitting up. This is interesting since this is

 

education are the ones more likely to leave the household.

114

saying that controlling for age and education of the head of households —thus controlling
for human capital of the head — and controlling for other household variables, households
headed by women have higher probability of breaking up. This is something that needs to

be investigated much ﬁrrther.

Panel Households

I also investigate what happens to the panel households -households appearing in
the three waves of the IFLS- by using a pooled Linear Probability Model (LPM) and
LPM with origin household ﬁxed effects.

The setup for the pooled LPM estimation is as follows. I ﬁrst restrict the sample
only to include households appearing in all three waves of the IFLS. For each household,
there are two observations: one representing the household in 1993, the other
representing the household in 1997. The dependent variable for the observations with
1993 values is 1 if the household split by 1997, 0 otherwise. The explanatory variables
for these households are their 1993 characteristics. Likewise, the dependent variable for
the observations with 1997 values is 1 if the household split by 2000, 0 otherwise; and
the explanatory variables for these households are their 1997 characteristics.43 The idea is
to estimate the linear probability model of household division conditional on the
household variables in the base year. Included as explanatory variables are household
composition, education variables, dummy for urban region, and a year dummy equals 1 if

the base year is 1997. The results are presented in the ﬁrst four columns of Table 2.6.7

 

’3 This implies that for the observations with 1997 as the base year, I did not take into account whether the
households have divided before 1997.

115

(for Claimant 1) and Table 2.6.8 (for Claimant 2).

The results seem to show that as before, maximum years education is positively
associated with the probability of household division. For the Claimant 2 speciﬁcation,
the coefﬁcient on the standard deviation of education is positive and signiﬁcant (Table
2.6.8, column (2)). When head’s education is added along with the maximum years and
the standard deviation, the result in (column (4)) shows that head’s education and
maximum education are negative, and positive, respectively. For Claimant 1, maximum
education is positive, and signiﬁcant, while standard deviation is not signiﬁcant when
head’s education is excluded (Table 2.6.7, column (2)). When head’s education is
included, the coefﬁCient on standard deviation becomes negative and statistically
signiﬁcant. Standard deviation becomes not signiﬁcant, while coefﬁcients on maximum
education are positive and statistically signiﬁcant.

The last three columns in Table 2.6.7 and Table 2.6.8 report the results ﬁ'om
estimating the model using OLS, controlling for 1993 household ﬁxed effects. The rest of
the setup is similar to the LPM estimation above. For both Claimant 1 and Claimant 2,
the maximum years of education is positive and signiﬁcant while standard deviation,

when included, does not seem to help explain the probability of household division.

2.7 Conclusions

In this essay, I have investigated the factors underlying household division. The
descriptive analysis suggests that household division is indeed a non-random process.
The data shows that age and sex of household members is an important factors

determining individuals’ decision to leave the household. Men in Indonesia leaves

116

household for different reasons than women, and on average they leave at different age in
life. Households with higher assets seem to be more likely to break up. This is somewhat
in contrast with the prediction of the theoretical model.

The multivariate analysis provides a better picture about this process. The probit
estimates show that assets, after controlling for urban/rural dunnny variable, do not seem
to affect the probability of household division. On the other hand, education variables
seem to play an important role in determining which households are more likely to break
up. Positive association between maximum years of schooling of non-claimant members
with the probability of household division is consistent with the idea that household
members may have different preferences, and that this may lead to conﬂicts with regards
to household allocation decisions. The ﬁnding does not necessarily imply that the
collective household model holds — the model is not a test of collective household model.

Indeed, one interpretation of the ﬁndings, namely that household division in
Indoensia is very much related to migration of household members can be very well
explained in the context of unitary household model.

This essay is but a ﬁrst attempt to uncover a very important aspect in household
and individual behavior. Household division is a much more complex issue than that can
be explained with the current model. One of the problems related to the empirical
strategy that has not been dealt with satisfactorily in the previous or in this essay is the
question of initial condition assumption.

A potential improvement that can be made is to make use of the information
available in the split-off households. Longitudinal household surveys such as the IFLS

collect information not only from the original households but also from the split-off

117

households. By using the information on the “destination household” as well as the
origin-household, one could perhaps better understand this process, although the existing

models are not appropriate for this.

118

Table 2.4.1
Number of Households Interviewed: 1993, 1997, and 2000

 

 

1993 1997 2000
Households interviewed 7,224 7,619 10,435
Target households interviewed 7,224 6,742 7,789
Split-off households interviewed - 877 2,646

 

Source: IFLS], IFLSZ, IFLS3

Target households are households that were interviewed in any prior wave of the
survey. IFLSZ target households are IFLS] original households. IF LS3 target
households are IF LSl original households, IFLSZ split-off households and IF LSZ+
split-off households

 

 

 

Table 2.4.2
Household Division: 1997 and 2000
1997 2000
# HH % # HH %

1993 Households (N =7, 224)

Household not found/all members died 482 6.7 . 450 6.2

Household undivided 5,951 82.4 4,165 57.7

Household has divided 791 10.9 2,609 36.1
1997 Households (N=7,619) ‘

Household not found/all members died - - 305 4.0

Household undivided - - 5,744 75.4

Household has divided - - 1,570 20.6

 

Source: IFLS], IFLSZ, IFLS3
' 1997 Households includes 1997 splitoff households as well as the 1993 original
households

119

Table 2.4.3
Percentage of Household Interviewed by Types of Household Members,

 

 

 

1993, 1997, and 2000
1993 1997 2000
HH 1237 Target Split-off 23310 Target Split-off
interviewed Household HH Household HH
HH HH
Nuclear households " 65 63 63 63 63 63 62
Vertical households 9 21 24 25 15 23 26 15
Other °’ 14 12 11 23 14 ll 23
Total 100 100 100 100 100 100 100

 

Source: IFLSl, IFLSZ, IFLS3

"Nuclear households include at most head, spouse, and their children

I” Vertical households include also grandchildren, fathers, mothers, sons-in-law or daughters-in-law
c) Other households are households with other types of members

120

Table 2.4.4a
Household Headship Status in 1997 of 1993 Household Member, by Sex

Relationship to HH Head in 1997

 

Found in 1993 Origin HH Found in 1997 Split-off HH

 

 

 

 

 

 

 

 

 

 

 

 

 

Relationship to Total
£192,“th Head in Head Spouse Other Head Spouse Other
Male
Head 5,229 11 43 59 0 11 5,353
Spouse 1 6 l 0 0 0 8
Sons 70 0 5,891 197 0 101 6,259
Brothers 4 0 107 5 0 4 120
Other 52 1 1,034 153 0 166 1,406
Total 5,356 18 7,076 414 0 282 13,146
Female
Head 838 62 70 3 2 6 981
Spouse 281 4,890 69 15 3 8 6 5,299
Daughters 29 43 5,455 56 172 124 5,879
Sisters 0 4 102 0 7 5 1 18
Other 29 51 1,427 18 99 189 1,813
Total 1,177 5,050 7,123 92 318 330 14,090
Source: IFLSI, IFLSZ
Table 2.4.4b
Household Headship Status in 2000 of 1993 Household Member, by Sex
Relationship to HH Head in 2000
Relationship to Found in 1993 Origin HH Found in Any Split-off HH' Total
3:92“th Head in Head Spouse Other Head Spouse Other
Male
Head 4,855 7 65 166 0 32 5,125
Spouse 6 2 0 0 0 0 8
Sons 131 0 4,758 824 3 407 6,123
Brothers 2 0 83 29 0 6 120
Other 101 0 646 376 0 348 1,471
Total 5,095 9 5,552 1,395 3 793 12,847
Female
Head 648 108 129 12 12 1 l 920
Spouse 438 4,463 133 43 101 27 5,205
Daughters 44 102 4,338 241 732 357 5,814
Sisters 1 6 74 3 22 9 1 15
Other 47 88 943 84 263 346 1,771
Total 1,178 4,767 5,617 383 1,130 750 13,825

 

Source: IFLS], IFLS2, IFLS3

‘ Includes 1997, 1998,2000 split-off households.

121

Table 2.4.4c
Household Headshipjtatus in 2000 of 1997 Household Member, by Sex

Relationship to HH Head in 2000

 

 

 

 

 

 

Relationship to Found in Origin HH' Found in 2000 Split-off HH Total

giggsehold Head in Head Spouse Other Head Spouse Other

Male
Head 5,512 7 108 110 0 25 5,762
Spouse 26 3 0 1 0 0 30
Sons 130 0 6,072 441 0 293 6,936
Brothers 0 0 87 12 0 6 105
Other 139 0 1,270 190 0 225 1,824
Total 5,807 10 7,537 754 0 549 14,657

Female
Head 918 112 143 9 10 5 1,197
Spouse 275 5,105 113 17 64 20 5,594
Daughters 44 l 12 5,601 163 367 225 6,512
Sisters 2 5 88 0 7 7 109
Other 73 120 1,699 44 150 255 2,341
Total 1,312 5,454 7,644 233 598 512 15,753

 

Source: IFLSl, IFLSZ, IFLS3

. Includes 1993 original households as well as 1997 split-off households

122

Figure 1. Relationship to Household Head by Age, Male and Female, 1993 and 2000

       

 

 

  

 

 

 

Male. 1993 Female. 1993
16 T - i W . . 7 16 i , 7 7
‘ 71 14: , A .7 i ”72*, 1
--1 £212 1 - W -l
(U
7. E10 - - i
1 “a 1 l
at» > , - i .9
S
c ...,.
2a" .. 0 ‘ ‘3
: ' [L 4 - - . . 7 ,
a ... . 1 a.
a 5 . ‘7- “z 01 .4. . E

 

0-5 611 12-17 1523 213—3 3045 3641 42.47 4&5:
Age group

0-5 6-11 12-17 18—23 24—29 1045 $41 42-47 4353 $59 60-

Age group
I Head I Hlsband I Head I ere
[1 Son [:1 in law [:1 Daughter |:| mughter-i'l-Iaw
I Brother/brother-in-Iaw I Fatherlfather-in-law I Sister/Sister-ln—Iaw I Mother/nother—in law
I Other Other
16 7 Malo.2£00 Female, 2000

’1

..A ..a
N #-

   

.a
o

_A

o

O)

 

Percentage of male
m
Percentage of female
at

a.

     

el . 7-11.
4~ m -‘
o.- . . .. .M.

0-5 6—11 12—17 18-23 2429 31-3 1341 42—47 4553 54-59 60-

Age group Age group
3 Head I Hlsband I Head I Wlfe
|:| Son E1 Son irl law [J Daughter |:| Daughter-ln-hw
I Brotherlbrother-in-Iaw I Father/father-in—Iaw I S' ter/Slster-ln-Iaw I lVother/rmther-In law
I Other Other

123

Table 2.4.5a

1997 Headship Status of 1993 HI] Members, by Sex and Age Group

 

 

 

 

 

 

Found in 1993 Origin HH, Found in 1997 Split-off HH,
Headship Status: Headship Status:
Age in 1993 Head Spouse Other Head Spouse Other Total
Male
0-5 0 0 1,895 0 O 96 1,991
6-11 1 0 2,132 11 0 75 2,219
12-13 1 0 921 32 0 53 1,007
15-17 11 0 659 12 0 14 696
18-20 19 0 439 16 0 8 482
21-24 152 O 420 36 0 6 614
25-29 543 2 272 139 O 14 970
30-39 1,605 2 183 107 0 9 1,906
40-49 1,248 3 40 32 0 3 1,326
50+ 1,776 11 115 29 O 4 1,935
Total 5,356 18 7,076 414 0 282 13,146
Female
0-5 0 0 1,784 0 0 104 1,888
6-11 1 2 2,065 11 2 88 2,169
12-13 2 2 826 31 36 72 969
15-17 11 28 566 8 23 9 645
18-20 17 1 14 376 4 46 6 563
21-24 24 383 322 3 61 13 806
25-29 52 769 259 9 68 13 1,170
30-39 175 1,742 207 13 55 5 2,197
40-49 219 1,020 103 6 12 4 1,364
50+ 676 990 615 7 15 16 2,319
Total 1,177 5,050 7,123 92 318 330 14,090

 

Source: IFLSl, IFLSZ

124

Table 2.4.5b
2000 Headship Status of 1993 HH Members, by Sex and Age Group

Found in 1993 Origin HH, Headship Found in Any Split-off HH',

 

 

 

 

 

Status: Headship Status:
Age in 1993 Head Spouse Other Head Spouse Other Total
0-5 0 0 1,753 4 0 213 1,970
6-11 2 0 1,820 91 0 207 2,120
12-13 8 0 643 136 0 114 901
15-17 9 0 454 126 0 82 671
18-20 29 0 256 144 1 62 492
21-24 150 0 246 276 2 56 730
25-29 542 l 154 308 0 24 1,029
30-39 1,593 1 109 182 0 12 1,897
40-49 1,208 1 25 57 0 6 1,297
50+ 1,554 6 92 71 0 17 1,740
Total 5,095 9 5,552 1,395 3 793 12,847
Female

0—5 0 0 1,650 3 0 213 1,866
6-11 3 5 1,716 77 60 226 2,087
12-13 0 8 529 105 158 114 914
15-17 8 31 356 31 176 55 657
18-20 1 1 126 214 29 190 30 600
21-24 24 390 187 27 239 26 893
25-29 57 762 165 36 152 18 1,190
30-39 226 1,700 133 26 104 17 2,206
40-49 248 935 94 18 25 7 1,327
50+ 601 810 573 31 26 44 2,085
Total 1,178 4,767 5,617 383 1,130 750 13,825

 

Source: IF LSl, IFLSZ, IF LS3
° Includes 1997, 1998, 2000 split-off households.

125

Table 2.4.5c
2000 Headship Status of 1997 HH Members, by Sex and Age Group

 

 

 

 

 

Found in 1997 Origin 1111', Found in 2000 Split-off HH,
Headship Status: Headship Status:
Age in 1997 Head Spouse Other Head Spouse Other Total
Male
0-5 0 0 1,706 0 0 138 1,844
6-11 2 0 2,025 1 1 0 89 2,127
12-13 6 0 1,052 40 0 80 1,178
15-17 22 0 815 89 0 86 1,012
18-20 33 0 590 95 0 42 760
21-24 73 O 493 134 0 54 754
25-29 401 l 391 170 0 29 992
30-39 1,699 1 276 142 0 10 2,128
40-49 1,532 2 46 25 0 5 1,610
50+ 2,039 6 143 48 O 16 2,252
Total 5,807 10 7,537 754 0 549 14,657
Female
0-5 0 0 1,638 0 0 153 1,791
6-1 1 0 0 1,936 8 3 91 2,038
12-13 10 4 1,020 35 28 62 1,159
15-17 12 45 691 70 88 75 981
18-20 16 72 509 37 1 15 51 800
21-24 29 254 409 17 152 25 886
25-29 55 723 299 1 1 106 10 1,204
30-39 199 1,906 280 24 71 10 2,490
40-49 262 1,315 120 9 19 5 1,730
50+ 729 1,135 742 22 16 30 2,674
Total 1,312 5,454 7,644 233 598 512 15,753

 

Source: IFLSl, IFLSZ, IFLS3
. Includes 1993 original households as well as 1997 split-off households

126

Table 2.4.6a
Reason for Leaving the Household of Household Members Not Found in the
Target Households in 1997, by Sex "

 

 

Male Female
Reason not in I-IH Freq. Percent Freq. Percent
Find work 896 36.5 484 19.7
School 275 1 1.2 263 10.7
Follow spouse/parent 377 15.4 655 26.7
Marriage 446 18.2 561 22.8
Divorce 44 1.8 39 1 .6
Help family 24 1.0 51 2.1
Need place to stay 151 6.1 122 5.0
Other 243 9.9 256 10.5
Total 2,456 100.0 2,431 100.0

 

Source: IFLSl and IFLSZ
*)Target households are households that were interviewed in any prior wave of the
survey. IFLSZ target households are IFLSl original households.

Table 2.4.6b
Reason for Leaving the Household of Household Members Not Found in the
TarEt Households in 2000, by Sex )

 

 

Male Female
Reason not in HH Freq. Percent Freq. Percent
Find work 1426 29.7 870 17.3
School 417 8.7 351 7.0
Follow spouse/parent 918 19.1 1,833 36.4
Marriage 798 16.6 880 17.5
Divorce 108 2.3 90 1.8
Help family 42 0.9 94 1.9
Need place to stay 553 11.5 390 7.8
Other 534 11.1 524 10.4
Total 4,796 100.0 5,032 100.0

 

Source: IFLSl, IF L82, and IFLS3

*)Target households are households that were interviewed in any prior wave of the
survey. IF LS3 target households are IF LSl original households, IFLSZ split-off
households and IF LSZ+ split-off households.

127

Table 2.4.7
Moves by Location: 1993 Household Members
Found in Agy Split-off BB in 2000

 

 

 

Male Female
# °/o # %
Not moved/moved within village 916 41.8 959 42.4
Moved within sub-district 225 10.3 232 10.3
Moved within district 307 14.0 267 11.8
Moved within province 388 17.7 462 20.4
Moved outside province 355 16.2 343 15.2
Total 2,191 100.0 2,263 100.0

 

Source: IFLSl, IFLSZ, and IFLS3

128

Table 2.4.8
1993 Household Demographic and Education Variables by Household Status in 2000

HE not found in 200 HH found undivided HH have divided by

 

 

(N=450) in 2000 (N=4,l65) 2000N=(2,609)
1993 BB Variables Mean Sd. Dev Mean Sd. Dev Mean Sd. Dev
Mean
Age 35.62 20.07 29.76 15.1 27.2 8.73
Age male 32.35 19.34 29.7 16.14 27.84 11.85
Age female 36.07 19.65 30.73 16.16 28.9 11.78
Educ (yrs) 6.86 4.77 5.28 3.77 5.93 3.36
Educ of male (yrs) 6.06 5.55 5.48 4.48 6.36 4.15
Educ of female (yrs) 4.98 4.85 4.44 3.95 5.01 3.71
Educ ofhead (yrs) 7.18 5.4 5.33 4.52 5.36 4.41
Standard deviation
Age 9.4 9.27 15.74 7.91 18.27 5.77
Age male 6.58 9.87 11.99 11.31 15.93 10.38
Age female 7.14 9.81 12.04 10.63 16.17 9.15
Educ (yrs) 1.76 2.10 2.18 1.98 2.87 1.73
Educ of male (yrs) 0.42 1.19 0.59 1.56 1.41 1.98
Educ of female (yrs) 0.86 2.04 0.92 1.97 1.92 2.35
Maximum
Educ of male (yrs) 6.4 5.75 5.93 4.76 7.48 4.61
Educ of female (yrs) 5.66 5.42 5.14 4.44 6.53 4.53
Educ (yrs) 8.37 5.30 7.20 4.49 8.76 4.06
HH size 3.1 2.23 4.04 1.85 5.7 2.19
0-5 yrs 0.35 0.69 0.61 0.78 0.60 0.80
6 -14 yrs (boys) 0.22 0.59 0.43 0.71 0.66 0.81
6 -14 yrs (girls) 0.27 0.65 0.42 0.68 0.67 0.83
15- 24 yrs male 0.32 0.66 0.28 0.57 0.68 0.88
15- 24 yrs female 0.37 0.79 0.34 0.66 0.66 0.82
25 -49 yrs male 0.48 0.6 0.63 0.52 0.76 0.63
25 -49 yrs female 0.54 0.62 0.68 0.53 0.83 0.59
50 - 64 yrs male 0.09 0.28 0.18 0.38 0.3 0.46
50 - 64 yrs female 0.16 0.37 0.23 0.43 0.32 0.48
65+ male 0.12 0.33 0.10 0.31 0.11 0.31
65+ female 0.19 0.4 0.12 0.33 0.11 0.32
Male headed hh 0.70 0.46 0.84 0.36 0.86 0.35
Urban 0.70 0.46 0.46 0.50 0.47 0.50

 

129

Table 2.4.9
1993 Household Assets Variables by Household Status in 2000

 

 

 

HH not found in HH found undivided HH have divided by
2000 in 2000 2000

1993 H Variables Mean Sd. Dev Mean Sd. Dev Mean Sd. Dev
H has farm business 0.19 0.39 0.39 0.49 0.42 0.49
H has nonfarm business 0.25 0.44 0.32 0.47 0.38 0.48
Household business assets
Landed 0.18 0.38 0.34 0.47 0.38 0.49
Farm land value ( 1000 Rp) 565 4,945 4 1,817 2,469 11,200
Non-farm land value (1000 Rp) 528 8,676 153 2,847 407 7,902
Land value (1000 Rp) 1,072 10,100 1,933 10,700 2,837 13,700
Other business assets (1000 Rp) 954 6,876 852 5,656 1,982 23,800
Household business assets per capita
Landed 373 3,172 513 2,438 462 1,893
Non-farm land value (1000 Rp) 167 2,877 35 562 77 1,523
Land value (1000 Rp) 530 4,381 537 2,523 531 2,425
Other business assets (1000 Rp) 325 2,717 233 1,650 306 2,267
Household assets
House occupied 10,600 45,300 5,617 21,100 8,279 30,300
Other house 2,141 13,700 744 5,145 1,643 19,000
Other land 1,567 10,900 1,451 9,813 2,084 12,400
Livestock 30 263 60 445 89 556
Vehicle 2,734 23,100 444 2,643 847 8,380
House appliances 763 2,105 401 1,995 473 1,785
Savings 638 3,849 193 1,283 197 1,047
Stock 23 475 6 188 9 336
Receivables l 13 674 134 2,481 219 2,844
Jewelry 240 560 191 2,244 197 803
Other 103 732 53 404 141 2,257
Total 17,700 60,900 8,928 28,500 13,600 52,000
Household assets per capita
House occupied 2,985 12,700 1,503 5,728 1,521 7,016
Other house 488 2,523 220 1,733 300 4,006
Other land 569 2,574 437 3,440 373 2,276
Livestock 11 76 17 93 17 108
Vehicle 588 3,897 l 18 804 147 1,594
House appliances 230 735 104 509 85 315
Savings 164 802 53 319 38 253
Stock 4 79 l 36 1 56
Receivables 36 185 34 483 39 480
Jewelry 89 206 51 464 36 1 19
Other 41 306 15 127 24 341
Total 4,855 14,300 2,454 8,170 2,473 11,300

 

Source: IFLSI, IFLSZ, IF LS3

130

.88 e Be as...» a .82 a

o: 2:03 0:3 EBEUE 20:92.0: 8:8 23 use: 20:08.0: 02:2: £09220 20:3 852.520 032:8 23 m20:om=0: :0 $22.00 N 55220

0:3 $536 223: 28 08220 223: .23: 20:33: 02:2: 8:522... 20:3 95850 0.022.: :23 220585: :0 $2300 _ “SEE—U

 

 

 

 

 

 

 

 

 

£88 m8 :3. 08$ w K 3.3. 3.8: 82 .86 S3 8 .5520
088V 8% 85 3.3 88 E..— 083 _8_ 5% 23 _ .5286
88
:3 e Bites 8.: 82288:
as.» a 8.22.88; Se .2
:88 3 82.5.0 28a :88 E .5220 use... :8" 3 8320 use... 8.... __ also.» 8.8..
8...; mean. a? E. “582.. a... 88 a. "cam

£83 $2 t} 85 a: ma." 08$ 88 83 83 N “5550
3.3 8: 35.8 3.3 :2 :2 03.3 Ra ma... 8:. _ .5586
coon
:3 e Bites 8% 2228.8:
88.8 829.88; 82 =<
n3. .30» Sam

:88 E 8220 use..— 888 3 82:0 case... :88 3 8220 use...

8...: use? a? ==J582e use. .88 e. deem
£8: 88 83 £8: 88 88.8 08: a: 886 :3 8 $586
38: e8 23 08: 88 ”8.— 8m: 8m 88..” 8} 2553.0
3.:
. . . ~86 E 83258 82 8.2.88:
83 8.238; 82 =<
1J1
82 E 8220 use... 88. E 8220 use... 83 E .8220 use... 82 .3 sum
.23. + one: :25 as: + 555 n: a

 

8...!» ”28.8 :33 I: “5.50.... ..oca .33 a. 253m
82m 0.98am
—.m.u 03am.

131

Table 2.5.2a
Summary Statistics : Base Year 1993, Division by 1997, Claimant 1 and Claimant 2

Claimant 1 (N=g,27o) "Claimant 21N=6,627) ”’
1993 BB variables Means Std. Dev. Means Std. Dev.
# Claimants 2.97 1.17 3.18 1.43

 

 

Proportion of claimants:

 

15-19 years, male 0.16 0.21 0.07 0.14
20-29 years, male 0.15 0.21 0.1 1 0.18
30-49 years, male 0.18 0.22 0.20 0.21
50 years or older, male 0.15 0.19 0.11 0.17
15-19 years, female 0.14 0.20 0.07 0.14
20-29 years, female 0.11 0.18 0.14 0.20
30-49 years, female 0.06 0.15 0.20 0.20
50 years or older, female 0.05 0.14 0.12 0.18
# Nonclaimant age 15 and above 1.07 0.78 0.00 0.00
# Boys 0-11 0.59 0.81 0.68 0.85
# Girls 0-11 0.57 0.80 0.66 0.84
# Boys 12-14 0.23 0.46 0.18 0.42
# Girls 12-14 0.22 0.45 0.18 0.42
Age of household head 50.78 11.88 45.64 13.71
Male head (=1) 0.82 0.38 0.89 0.31
Head of hh's schooling 5.14 4.40 5.41 4.43
Max. of non-head claimants' schooling 8.77 3.79 7.13 4.44
Sd. dev. of claimants' schooling 2.78 1.86 2.62 1.85
Mean of non-head claimants' schooling 7.89 3.50 5.62 3.70
Land owned (000 Rp) 2,616 10,600 2,280 9,885
Farm/Non-farm business assets (000 Rp) 1,097 4,929 1,070 5,339
Urban (=1) 0.51 0.50 0.46 0.50
HH farm (=1) 0.41 0.49 0.42 0.49

 

a) Claimant 1 consists of households with multiple claimants where claimants include household
head, head’s children and head’s siblings who would be at least 19 years old in 1997.

b) Claimant 2 consists of households with multiple claimants where claimants include household
head and other household members who would be at least 19 years old in 1997.

132

Table 2.5.2b
Summall Statistics: Base Year 1993, Division by 2000, Claimant 1 and Claimant 2

Claimant 1 (N=3,951) "Claimant 2 QV=6,330L"’
Means Std. Dev. Means Std. Dev.
# Claimants 3.15 1.26 3.536 1.62

 

 

Proportion of claimants:

12-14 years, male 0.12 0.14 0.05 0.12
15-19 years, male 0.11 0.18 0.06 0.12
20-29 years, male 0.12 0.19 0.10 0.17
30-49 years, male 0.19 0.21 0.18 0.19
50 years or older, male 0.13 0.18 0.10 0.15
12-14 years, female 0.06 0.14 0.04 0.12
15-19 years, female 0.10 0.17 0.06 0.12
20-29 years, female 0.09 0.16 0.13 0.19
30-49 years, female 0.05 0.13 0.17 0.18
50 years or older, female 0.04 0.13 0.11 0.17
# Nonclaimant age 12 and above 0.57 0.95 -0.37 0.59
# Boys 0-5 0.27 0.53 0.33 0.57
# Girls 0-5 0.25 0.52 0.31 0.55
# Boys 6-11 0.37 0.61 0.36 0.60
# Girls 6-11 0.37 0.61 0.36 0.60
Age of household head 49.17 11.90 45.54 13.71
Male head (=1) 0.84 0.36 0.88 0.32
Head of hh's schooling 5.22 4.35 5.42 4.43
Max. of non-head claimants' schooling 8.28 3.64 7.39 4.18
Sd. dev. of claimants' schooling 2.68 1.75 2.64 1.74
Mean of non-head claimants' schooling 7.22 3.21 5.65 3.45
Land owned (000 Rp) 2,480 9,916 2,247 9,577
Farm/Non-farm business assets (000 Rp) 1,129 5,141 1,056 5,290
Urban 0.48 0.50 0.46 0.50
HH farm 0.42 0.49 0.41 0.49

 

a) Claimant 1 consists of households with multiple claimants where claimants include household
head, head’s children and head’s siblings who would be at least 19 years old in 2000.

b) Claimant 2 consists of households with multiple claimants where claimants include household
head and other household members who would be at least 19 years old in 2000.

133

Table 2.5.2c

Descriptive Statistics: Base Year 1997, Division by 2000 Claimant l and Claimant 2

Claimant 1 (N=3,417) "Claimant 2 (N=6,664) '”
Means Std. Dev. Means Std. Dev.

 

 

# Claimants

Proportion of c1aima_l;ts_:
16-19 years, male
20-29 years, male
30-49 years, male
50 years or older, male
16-19 years, female
20-29 years, female
30-49 years, female
50 years or older, female
# Nonclaimant age 16 and above
# Boys 0-11
# Girls 0-11
# Boys 12-15
# Girls 12-15
Male head (=1)
Age of household head
Head of hh's schooling

Max. of non-head claimants' schooling

Sd. dev. of claimants' schooling

Mean of non-head claimants' schooling

Land owned (000 Rp)

Farm/Non—farm business assets (000 Rp)

Urban
HH farm

2.91 1.09 3.12 1.36
0.13 0.04 0.05 0.12
0.16 0.21 0.09 0.16
0.19 0.22 0.21 0.21
0.16 0.20 0.12 0.17
0.12 0.19 0.06 0.13
0.12 0.18 0.12 0.19
0.07 0.16 0.22 0.20
0.05 0.14 0.13 0.18
1.03 0.77
0.49 0.75 0.59 0.80
0.48 0.73 0.57 0.77
0.26 0.49 0.23 0.48
0.27 0.51 0.23 0.48
0.81 0.40 0.88 0.33
52.40 11.72 47.49 13.71
5.45 4.50 5.87 4.59
9.36 3.80 7.73 4.53
2.91 1.94 2.70 1.96
8.50 3.61 6.21 3.89
4,21 1 21,000 3,639 17,700
2,372 14,700 2,195 13,800
0.50 0.50 0.46 0.50
0.36 0.48 0.36 0.48

 

a) Claimant 1 consists of households with multiple claimants where claimants include household
head, head’s children and head’s siblings who would be at least 19 years old in 2000.

b) Claimant 2 consists of households with multiple claimants where claimants include household
head and other household members who would be at least 19 years old in 2000.

134

Table 2.5.3a
Means and Difference in Means of Key Variables of 1993 Households between Households that Have
Divided and Households that Have Not Divided by 1997

Households Household

 

have not divided have divided Diff. p-value
1993 H Characteristics by 1997 by 1997
Claimant 1 '1 N= 3,270
# Claimants 2.9 3.4 0.5 0.000
Household size 5.5 6.5 1.0 0.000
Age, mean 29.2 28.8 -0.4 0.296
Age, sd. dev. 18.6 19.0 0.4 0.081
Head's age 50.6 51.7 1.1 0.043
Head's educ. (yrs.) 5.2 4.9 -0.4 0.086
Non-head claimants' mean educ (yrs). 7.9 7.9 0.0 0.833
Non-head claimants' max. educ. (yrs) 8.7 9.1 0.4 0.032
Claimants' sd. dev. of educ. (yrs) 2.8 2.8 0.0 0.622
Household sd. dev. of education (yrs) 3.4 3.6 0.1 0.021
Value of land owned (1000 Rp) 2,530 2,939 409 0.465
Value of other bus. assets (1000 Rp) 1,027 1,389 362 0.157
Claimant 2 ”’ N= 6,227
# Claimants 2.4 2.7 0.3 0.000
Household size 5.7 6.6 0.9 0.000
Age, mean 29.4 29.0 -0.4 0.302
Age, sd. dev. 18.9 19.1 0.2 0.434
Head's age 50.3 51.9 1.6 0.011
Head's educ. (yrs.) 5.3 4.8 -0.4 0.051
Non-head claimants' mean educ (yrs). 7.9 7.9 0.1 0.785
Non-head claimants’ max. educ. (yrs) 8.4 8.8 0.4 0.047
Claimants’ sd. dev. of educ. (yrs) 3.0 3.0 0.0 0.713
Household sd. dev. of education (yrs) 3.5 3.6 0.1 0.122
Value of land owned (1000 Rp) 2,724 3,065 341 0.608
Value of other bus. assets (1000 Rp) . 1,074 1,228 154 0.535

 

a) Claimant 1 consists of households with multiple claimants where claimants include household head,
head’s children and head’s siblings who would be at least 19 years old in 1997.

b) Claimant 2 consists of households with multiple claimants where claimants include household head
and other household members who would be at least 19 years old in 1997.

135

Table 2.5.3b
Means and Difference in Means of Key Variables of 1993 Households between Households that Have
Divided and Households that Have Not Divided by 2000

Households Household

 

have not divided have divided Diff. p-value
1993 HH Characteristics by 2000 by 2000
Claimant 1 " N= 3,951
# Claimants 2.8 3.5 0.7 0.000
Household size 5.0 6.0 1.1 0.000
Age, mean 29.0 27.3 -1.7 0.000
Age, sd. dev. 18.9 18.3 -0.6 0.000
Head's age 49.1 49.3 0.2 0.590
Head's educ. (yrs.) 5.1 5.3 0.2 0.275
Non-head claimants' mean educ (yrs). 6.9 7.5 0.5 0.000
Non-head claimants' max. educ. (yrs) 7.8 8.7 0.9 0.000
Claimants' sd. dev. of educ. (yrs) 2.7 2.7 0.0 0.638
Household sd. dev. of education (yrs) 3.3 3.4 0.1 0.011
Value of land owned (1000 Rp) 2,040 2,826 786 0.009
Value of other bus. assets (1000 Rp) 725 1,460 734 0.000
Claimant 2 '” N=6,330
# Claimants 3.0 4.3 1.3 0.000
Household size 4.3 5.7 1.4 0.000
Age, mean 28.3 27.2 -1.1 0.000
Age, sd. dev. 16.8 18.3 1.5 0.000
Head's age 43.9 48.0 4.1 0.000
Head's educ. (yrs.) 5.5 5.4 -0.1 0.392
Non-head claimants' mean educ (yrs). 5.3 6.1 0.8 0.000
Non-head claimants' max. educ. (yrs) 6.6 8.5 1.9 0.000
Claimants' sd. dev. of educ. (yrs) 2.5 2.9 0.5 0.000
Household sd. dev. of education (yrs) 3.2 3.4 0.1 0.004
Value of land owned (1000 Rp) 1,947 2,661 714 0.004
Value of other bus. assets (1000 Rp) 824 1,384 560 0.000

 

a) Claimant 1 consists of households with multiple claimants where claimants include household head,
head’s children and head’s siblings who would be at least 19 years old in 2000.

b) Claimant 2 consists of households with multiple claimants where claimants include household head and
other household members who would be at least 19 years old in 2000.

136

Table 2.5.3c
Means and Difference in Means of Key Variables of 1997 Households between Households that
Have Divided and Households that Have Not Divided by 2000

Households

 

have not Household .
divided by have div1ded lef. p-value
1997 HH Characteristics 2000 by 2000
Claimant 1 " N= 3,417
# Claimants 2.7 3.2 0.45 0.000
Household size 5.1 6 0.91 0.000
Age, mean 31.6 29.3 -2.27 0.000
Age, sd. dev. 19.3 18.9 -0.45 0.012
Head's age 52.5 52.2 -0.28 0.494
Head's educ. (yrs.) 5.4 5.5 0.2 0.372
Non-head claimants' mean educ (yrs). 8.4 8.8 0.37 0.005
Non-head claimants' max. educ. (yrs) 9.1 9.8 0.67 0.000
Claimants' sd. dev. of educ. (yrs) 2.9 2.9 0.00 0.951
Household sd. dev. of education (yrs) 3.5 3.7 0.12 0.030
Value of land owned (1000 Rp) 3,704 5,119 1,414 0.051
Value of other bus. assets (1000 Rp) 2,202 2,669 467 0.361
Claimant 2 '” N=6,664

# Claimants 2.9 3.9 1.07 0.000
Household size 4.4 5.8 1.34 0.000
Age, mean 30.3 29 -1.38 0.000
Age, sd. dev. 17.4 18.9 1.53 0.000
Head's age 46.6 50.5 3.93 0.000
Head's educ. (yrs.) 5.9 5.6 -0.29 0.030
Non-head claimants' mean educ (yrs). 6 6.8 0.81 0.000
Non-head claimants' max. educ. (yrs) 7.3 9.2 1.84 0.000
Claimants' sd. dev. of educ. (yrs) 2.6 3.1 0.56 0.000
Household sd. dev. of education (yrs) 3.4 3.6 0.15 0.004
Value of land owned (1000 Rp) 3,247 4,931 1,684 0.003
Value of other bus. assets (1000 Rp) 2,083 2,556 473 0.185

 

a) Claimant 1 consists of households with multiple claimants where claimants include household head,
head’s children and head’s siblings who would be at least 19 years old in 2000.

b) Claimant 2 consists of households with multiple claimants where claimants include household head
and other household members who would be at least 19 years old in 2000.

137

A20 800» E .2 20:02.0: 0:. ..0 :00:
0:5 00028.0 0. 0:00.: :02. 0.. ::3 20:05.0: 0: 005m 2.8.128 mmmn 2 2:0. 0:. 5 03.000. 20:08.0: ..0 .095... .80. 0:... . .020 ..0 20 0.80.: m _ ma 00.....0: m. :03.

—m1—n= ”ooh—om

 

 

 

:00 owed .506 00.4.0 9m 06 NS Wm NA

omm Sod mmwd mwvd tn ..w a... 0.: 0A

5mm mend 03.0 02.0 Wm 0.5 2 .0 0

new 000.0 2.»: and NM 50 wd 9m m

02: «mod Sad m . m6 NM 0.0 m6 Em v

Sm. 25.0 806 000.0 o.m 9m ms ﬁn m

mcam ~36 80.. 50.0 0.. a... 0.0 Wm N

NE - - v.0. _

3.. E G. a. A... 5 5 A:

:0 =0 :0 00:00:00 0.. :0. 00.. 0 v... :0. 00..
0.0—0:030: Amwgsmv AS Amman. .525 AC Amman“ .0005 0.0—=00 A a”... .w“_=:” A 0“... 3.0.3.0000 GEO 00:00:00 20:02.0: 0:.
u 00320.30 00320.30 00:20:00 ..0 0032.30: 000:-..02 000:-..07. .0000: 2 =3... ..0 .3502
2.00.35

 

«.8. 3.... 6.2.38: 2.. a. 3.32 .c a2.53.. 2.. E 8.8.2; 803...... 0.2.3.8:
4.3 28...

138

Determinant of Household Division between 1993 and 1997, Claimant 1

Table 2.6.1

 

 

 

1993 Variables: (1) (2) (3) (4) (5) (6)
# Claimants 0.060 0.058 0.058 0.058 0.059 0.023
[8247]." [7.862] *n [7.866] H. [7.870] t“ [8.187] "t [8.245] ”*
Progortion of claimants:
20—29 years, male 0.291 0.292 0.291 0.291 0.304 0.102
[6163]." [6.196] ”t [6.161] *" [6.161] "a [6.452] *” [6213]”.
30-49 years, male 0.386 0.393 0.392 0.390 0.4 0.144
[5989]." [6.068] "W [6.058] *" [6.040] "m [6.283] "w [6112]."
50 years or older, male 0.386 0.394 0.392 0.392 0.389 0.14
[4.456] an [4553]." [4.538] "a [4.528] t" [4.544] a" [4.425] "*
15-19 years, female 0.032 0.033 0.033 0.032 0.04 0.022
[0.673] [0.685] [0.696] [0.673] [0.874] [1.391]
20-29 years, female 0.231 0.232 0.232 0.233 0.248 0.101
[4.640] *" [4.675] "t [4.675] "a [4.683] "t [4.994] a" [5.737] "t
30—49 years, female 0.335 0.339 0.339 0.339 0.35 0.127
[5.017] *" [5.083] ..., [5.072] *" [5.081] W" [5.347] t" [5.445] W"
50 years or older, female 0.218 0.228 0.229 0.225 0.234 0.083
[2.593] "t [2.717] -" [2.713] "‘ [2.688] t" [2.844] "t [2.817] "t
# Non-claimant members 15+ 0.019 0.020 0.020 0.020 0.019 0.007
[2.256] at [2.354] " [2.333] " [2.294] " [2.200] " [2.000] "
Age of head -0.011 -0.011 -0.011 -0.011 -0.011 -0.003
[2.533] " [2.554] " [2.539] "I [2.555] “ [2.617] "I [2.140] "
Age of head squared x 10'3 0.086 0.087 0.086 0.087 0.087 0.024
[2.315] .. [2.328] “ [2.318] "I [2.342] H [2.347] at [1.781] *
Head of hh's educ. -0.003 -0.005 -0.005 -0.004 -0.004 -0.002
[1.364] [1.926] t [1.924] '- [1.664] t [1.760] a [1.526]
Max. schooling, non-head claimants 0.003 0.006 0.006 0.006 0.006 -0.002
[1.632] [2.182] ” [2.168] H [2.150] " [2.207] at [1.297]
Sd. dev. of claimants' schooling -0.007 -0.007 -0.007 -0.007 0.045
[1.397] [1.396] [1.366] [1.396] [3.733] *N
# Boys 0-11 0.018 0.018 0.018 0.018 0.018 0.006
[2.249] H [2.238] " [2.254] H [2.251] H [2.314] H [2.159] n
# Girls 0-11 0.016 0.016 0.016 0.016 0.017 0.005
[1.957] ' [1.932] a [1.931] a [1.951] a [2.057] '- [1.667] t
# Boys 12-14 0.004 0.004 0.004 0.004 0.003 0.004
[0.285] [0.279] [0.273] [0.233] [0.229] [0.727]
# Girls 12-14 0.091 0.090 0.090 0.089 0.088 0.032
[6.497] t" [6.462] a" [6.447] “a [6.449] "a [6.399] '" [6.192] "a
Male head (=1) -0.053 -0.054 -0.053 -0.051 -0.055 -0.019
[1.430] [1.443] [1.424] [1.382] [1.521] [1.371]
log(land owned) (Rp x 10") 1.736 1.764 2.338 2.573 2.414 0.002
[1.518] [1.541] [1.517] [1.700]. [1.655]. [1.955]‘
log(bus. assets) (Rp x 103) -1.449 -l.467 -1.756 -1.414 -1.796 0.675
[1.211] [1.225] [0.839] [0.678] [0.842] [1.123]
Urban (=1) -0.055 -0.054 -0.055 -0.055 -0.078 -0.186
[3.090] t" [3.081] ml [3.081]". [3.065] "W [1.386] [0.222]
(continued)

139

(continued)

 

 

 

 

Interactions
Urban x log(land owned) x 10'3 -1.589 -1.989 -1.881 -0.627
[0.660] [0.830] [0.802] [0.626]
Urban x log(bus. assets) x10'3 0.537 .0297 -0.366 -0313
[0.214] [0.1 13] [0.139] [0.297]
log(land owned) (Rp x 10'3)
x household head's schooling 0.150 0.033 0.057
[0.479] [0.104] [0.491]
x max. of non-head claimants' schooling -0.011 0.025 -0.029
[0.029] [0.070] [0.218]
log(bus. assets) (Rp x 10'3)
x household head's schooling 0.138 0.175 0
[0.454] [0.576] [0.003]
x max. of non-head claimants' schooling 0.139 0.119 0.025
[0.364] [0.315] [0.173]
Observations 3,270 3,270 3,270 3,270 3,270 3,270
Province, urban, interaction
dummies No No No No Yes No
Community dummies No No No No No Yes
F-test -va1ues
Head's education 0.167 0.200 0.317
Non-head claimants' max. educ. 0.193 0.176 0.265
Education variables 0.177 0.138 0.054 0.349 0.407 0.641
Land variables 0.294 0.531 0.564 0.858
Other assets variables 0.494 0.726 0.535 0.945
Land and assets variables 0.270 0.260 0.554 0.711 0.535 0.968

 

Claimants include head of household and children or siblings who were 12 or over in 1993. The table
shows the marginal effects of a change in explanatory variables. Standard errors are corrected for

clustering at the 1993 community level and heteroskedasticity. Absolute value of 2 statistics are in brackets

with statistical significance at 10% (*); 5% (*"‘); and 1% (***) indicated.

140

Table 2.6.2

Determinant of Household Division between 1993 and 2000, Claimant 1

 

1993 Variables:
# Claimants

Proportion of claimants:

(1) (2) (3) (4) (5) (6)
0.113 0.111 0.111 0.111 0.117 0.129
[10.374] m [10.074] m [10.1031-m [10.048] m [10.726] m [10.278] m

 

15-19 years, male -0.157 -0.163 -0. 161 -0.160 -0.l69 -0.243
[2.097] " [2.169] H [2.137] " [2.121] H [2.267] H [2.908] a"
20-29 years, male 0.104 0.101 0.101 0.102 0.117 0.065
[1.369] [1.318] [1.324] [1.337] [1.512] [0.749]
30-49 years, male -0.215 -0.214 -0.211 -0.214 -0.192 -0.31
[2.270] H [2.256] H [2.220] H [2.247] n [2.014] .. [2.808] "a
50 years or older, male -0.171 -0. 172 -0. 172 -0.170 -0.142 -0.323
[1.501] [1.519] [1.507] [1.480] [1.227] [2.391]
12-14 years, female 0.036 0.033 0.032 0.031 0.041 0.051
[0.518] [0.475] [0.471] [0.456] [0.595] [0.644]
15-19 years, female -O.133 -0.141 -0.141 -0.138 -0.130 -0.2
[1.837] a [1.945] * [1.941] * [1.899] t [1.790] t [2.460] "
20-29 years, female -0.058 -0.065 -0.063 -0.061 -0.047 -0.083
[0.750] [0.831] [0.811] [0.785] [0.592] [0.973]
3049 years, female -0.151 -0.154 -0.153 -0.153 -0.132 -0.176
[1.480] [1.498] [1.492] [1.495] [1.267] [1.487]
50 years or older, female -0.170 -0.162 -0.163 -0.164 -0.135 -0.181
[1.315] [1.246] [1.251] [1.260] [1.025] [1.290]
# Non-claimant members 12+ 0.069 0.069 0.069 0.069 0.069 0.086
[5.304] "'- [5.301] a" [5.307] t" [5.273] "9 [5.332] "9 [5.998] N”
Age of head -0.006 -0.006 -0.006 -0.006 -0.007 -0.005
[1.136] [1.101] [1.106] [1.215] [1.349] [0.899]
Age of head squared x 10'3 0.047 0.046 0.046 0.051 0.056 0.053
[0.984] [0.966] [0.966] [1.064] [1.151] [0.973]
Head of hh‘s educ. 0.000 -0.002 -0.002 -0.001 -0.001 0.003
[0.048] [0.667] [0.634] [0.423] [0.213] [0.912]
Max. schooling, non-head claimants 0.007 0.011 0.011 0.011 0.011 0.014
[2.404] " [3.040] "a [3.071] t" [3.173] t" [3.259] W" [3.678] "t
Sd. dev. of claimants' schooling -0.011 -0.011 -0.009 -0.009 -0.009
[1.831] t [1.840] t [1.609] [1.585] [1.326]
# Boys 0-5 0.027 0.027 0.027 0.028 0.029 0.034
[1.715] 9 [1.737] 9 [1.752] a [1.773] t [1.930] * [1.909] r
# Girls 0-5 0.033 0.032 0.032 0.033 0.031 0.051
[2.112] H [2.117] " [2.109] H [2.151] " [1.978] " [2.738] "s
# Boys 6-11 0.056 0.056 0.056 0.055 0.055 0.062
[4.127] "a [4.093] 9'" [4.122] a" [4.079] "a [3.974] ”t [3.850] 0‘
# Girls 6-11 0.039 0.038 0.037 0.037 0.039 0.044
[2.746] a" [2.675] a" [2.676] t" [2.656] "* [2.864] 0' [2.842] "t
Male head (=1) -0.010 -0.011 -0.009 -0.005 -0.012 0.001
[0.220] [0.230] [0.201] [0.107] [0.259] [0.025]
(continued)

141

(continued)

 

log(land owned) (Rp x 103) 0.477 0.559 1.040 1.021 1.025 1.522

[0.311] [0.365] [0.535] [0.500] [0.497] [0.634]

log(bus. assets) (Rp x 10") 3.654 3.589 1.428 2.197 1.799 2.832

[2.172] " [2.138] H [0.514] [0.773] [0.606] [0.898]

Urban (=1) -0.070 -0.069 -0.071 -0.070 -0.097 0.953
[3.138] "t [3.096] 9'" [3.122] "9' [3.099] a" [1.568] [23.364] "a

Interactions

Urban x log(land owned) x 10'3 -0100 0.990 0.470 -159

[0.031] [0.284] [0.135] [0.391]

Urban x log(bus. assets) x 10’3 3.337 1.758 1.429 2.46

[0.954] [0.460] [0.368] [0.567]
log(land owned) (Rp x 10'3)

x household head's schooling 0.231 0.073 0.174
[0.511] [0.162] [0.350]
x max. of non-head claimants' schooling -0.613 -0.527 -0.63

[1.182] [1.032] [1.093]
log(bus. assets) (Rp x 10'3)

 

 

 

x household head's schooling 0.703 0.728 0.83
[1.817] t [1.867] t [1.888] t
x max. of non-head claimants' schooling -0.193 -0.135 -0.203
[0.378] [0.266] [0.362]
Observations 3,951 3,951 3,951 3,951 3,951 3,951
Province, urban, interaction
dummies No No No No Yes No
Community dummies No No No No No Yes
F-test [Q-valuesz
Head's education 0.077 0.138 0.073
Non-head claimants' max. educ. 0.005 0.005 0.001
Education variables 0.045 0.020 0.018 0.019 0.020 0.001
Land variables 0.814 0.755 0.790 0.750
Other assets variables 0.070 0.076 0.114 0.042
Land and assets variables 0.028 0.028 0.069 0.029 0.107 0.035

 

Claimants include head of household and children or siblings who were 12 or over in 1993. The table
shows the marginal effects of a change in explanatory variables. Standard errors are corrected for
clustering at the individual level and heteroskedasticity. Absolute value of 2 statistics are in brackets
with statistical significance at 10% (*); 5% (**); and 1% (“*) indicated.

142

Table 2.6.3

Determinant of Household Division between 1997 and 2000, Claimant 1

1997 Variables:
# Claimants

Proportion of claimants:

(1)
0.069

(2)
0.068

(5)
0.07

(6)
0.07

(3)
0.068

(4)
0.067

[6.656] t" [6.471] u. [6.473] in [6.366] an [6.741] um [6.440] 1......

 

20-29 years, male 0.101 0.101 0.101 0.103 0.107 0.097
[1.838] * [1.846] t [1.861] "‘ [1.884] * [1.964] ** [1.803] "'

30-49 years, male 0.001 0.004 0.005 0.002 0.012 -0.014

[0.016] [0.052] [0.066] [0.029] [0. 150] [0.178]

50 years or older, male -0.083 -0.080 -0.080 -0.088 -0.071 -0.081

[0.812] [0.777] [0.782] [0.863] [0.675] [0.762]

16-19 years, female 0.118 0.117 0.117 0.118 0.125 0.098

[2.106] .. [2.096] H [2.097] ** [2.110] ** [2.218] ** [1.641]

20-29 years, female 0.134 0.135 0.136 0.135 0.137 0.127
[2.119] " [2.126] H [2.144] ** [2.125] ** [2.181] ** [2.020] **

30—49 years, female -0.187 -0.185 -0.186 -0.192 -0.196 -0.129

[2.400] .. [2.383] " [2.387] " [2.470] " [2.441] " [1.530]

50 years or older, female -0.227 -0.222 -0.221 -0.232 -0.224 -0. 149

[1.879] a [1.849] * [1.845] * [1.944] * [1.875] * [1.248]

# Non-claimant members 16+ 0.031 0.031 0.031 0.031 0.03 0.055
[2.452] *9 [2.463] " [2.400] "”" [2.444] ** [2.311] " [3.813] ”*

Age of head 0.006 0.006 0.006 0.006 0.005 0.006

[1.065] [1.028] [1.030] [1.034] [0.960] [0.933]

Age of head squared x 10-3 -0.040 -0.038 -0.038 ~0.039 -0.035 -0.041

[0.802] [0.769] [0.774] [0.770] [0.701] [0.728]

Head of hh's educ. 0.000 -0.001 -0.001 -0.002 -0.001 -0.001

[0.1 15] [0.426] [0.418] [0.488] [0.306] [0.182]

Max. schooling, non-head claimants 0.005 0.006 0.006 0.006 0.007 0.009
[1.645] [1.623] [1.641] [1.660] * [1.815] "' [2.291] "

Sd. dev. of claimants' schooling -0.003 -0.003 -0.003 -0.004 -0.007

[0.504] [0.509] [0.424] [0.516] [0.985]

# Boys 0-11 0.023 0.023 0.023 0.023 0.023 0.027
[1.973] "- [1.969] as [1.977] ** [1.946] "' [1.988] " [2.224] "

# Girls 0—1 1 0.021 0.021 0.020 0.021 0.021 0.023

[1.724] a [1.712] ' [1.707] * [1.757] "‘ [1.713] * [1.699] "‘

# Boys 12-15 0.057 0.057 0.057 0.058 0.062 0.062
[3.105] Wt [3.109] "a [3.104] "* [3.122] *** [3.373] *" [3.279] "*

# Girls 12-15 0.106 0.106 0.106 0.106 0.111 0.125
[6.438] "m [6.438] a" [6.478] *" [6.448] "* [6.692] ”'" [7.257] "*

Male head (=1) -0.078 -0.078 -0.076 -0.079 -0.085 -0.07

[1.775] 9 [1.770] t [1.720] "' [1.775] * [1.887] * [1.480]

log(land owned) (Rp x 103) 2.456 2.466 2.553 2.370 2.153 1.502

[1.748] ‘ [1.756] t [1.488] [1.318] [1.163] [0.812]

log(bus. assets) (Rp x 103) 4.190 4.188 2.995 2.423 2.379 2.652

[2.758] t” [2.755] a" [1.167] [0.915] [0.872] [0.962]

Urban (=1) -0.013 -0.013 -0.014 -0.014 -0.032 0.145

[0.622L [0.615] [0.641] [0.658] [0.422] [0.479]
(continued)

143

(continued)

 

 

 

 

Interactions
Urban x log(land owned) x 10'3 0.485 1.063 0.222 0.472
[0.160] [0.338] [0.070] [0.148]
Urban x log(bus. assets) x 10'3 1.841 3.493 3.306 1.076
[0.570] [1.008] [0.930] [0.304]
log(land owned) (Rp x 10'3)
x household head's schooling -0.115 -0.068 -0.126
[0.289] [0.170] [0.326]
x max. of non-head claimants' schooling -0.001 -0.089 -0.099
[0.003] [0.200] [0.227]
log(bus. assets) (Rp x 10'3)
x household head's schooling 0.020 0.013 0.122
[0.050] [0.032] [0.299]
x max. of non-head claimants' schooling -0.65 -0.607 -0.231
L351] [1.259] [0.479]
Observations 3,417 3,417 3,417 3,417 3,417 3,417
Province, urban, interaction dummies No No No No No Yes
Community dummies No No No No No No
F -test [Q-valuesz
Head's education 0.962 0.991 0.987
Non-head claimants' max. educ. 0.197 0.159 0.139
Education variables 0.217 0.340 0.329 0.493 0.411 0.336
Land variables 0.165 0.392 0.696 0.748
Other assets variables 0.020 0.032 0.054 0.351
Land and assets variables 0.000 0.000 0.001 0.005 0.022 0.349

 

Claimants include head of household and children or siblings who were 16 or over in 1997. The table shows
the marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at the
individual level and heteroskedasticity. Absolute value of 2 statistics are in brackets with statistical
signiﬁcance at 10% (*); 5% (**); and 1% (***) indicated.

144

Table 2.6.4
Determinant of Household Division between 1993 and 1997, Claimant 2
1993 Variables: (1) (2) (3) (4) (5) (6)
# Claimants 0.043 0.043 0.043 0.043 0.044 0.018
[12.524] m [12.531] m [12.530] W [12.556] W [12.559] m [12.004] m

 

Proportion of claimants:

20-29 years, male 0.204 0.203 0.202 0.200 0.204 0.078
[4.656]H# [4.618] *H [4.592] *H [4.563] ‘H [4.697] 'H [4.617] *H
30-49 years, male 0.244 0.243 0.242 0.240 0.242 0.097
[5.348] 9H [5.285]*H [5.280] H- [5.274] *H [5.326] *H [5.231]H*
50 years or older, male 0.209 0.208 0.207 0.204 0.207 0.088
[3.627] H‘ [3.599] *H [3.590] 9H [3.545] H. [3.584] H. [3.697]-H
15-19 years, female 0.019 0.018 0.018 0.018 0.021 0.02
[0.424] [0.406] [0.404] [0.395] [0.480] [1.170]
20-29 years, female 0.078 0.077 0.076 0.076 0.074 0.038
[1.683]* [1.647]* [1.636] [1.641] [1.616] [2.171]H
30-49 years, female 0.127 0.122 0.121 0.122 0.115 0.045
[2.551]H [2.464]H [2.448] H [2.468] [2.390] H [2.451]H
50 years or older, female 0.167 0.154 0.154 0.155 0.151 0.066
[3.705]*H [3.423]Hs [3.403] *H [3.433] 'H [3.327] H‘ [3.721]H*
Age of head -0.001 -0.001 -0.001 -0.001 0 0
[0.248] [0.236] [0.235] [0.235] [0.207] [0.108]
Age of head squared x 10‘3 0.008 0.008 0.008 0.008 0.007 -0001
[0.361] [0.349] [0.347] [0.359] [0.314] [0.087]
Head of hh's schooling -0.002 -0.002 -0.002 -0.002 -0.002 -0.001

[1.753] a [1.568] [1.560] [1.361] [1.468] [0.972]
Max. schooling, non-head claimants 0.004 0.003 0.003 0.003 0.003 0.001
[2.910] H" [1.980] H [1.979] [2.040] [2.201] H [1.393]

 

Sd. dev. of claimants' schooling 0.004 0.004 0.004 0.004 0.033
[1.681] a [1.693] [1.421] [1.498] [4.849] *H

# Boys 0-11 0.003 0.003 0.003 0.003 0.004 0.002

[0.682] [0.660] [0.674] [0.639] [0.820] [1.125]

# Girls 0-11 0.012 0.012 0.012 0.012 0.013 0.006
[2.442] H [2.363] H [2.362] H [2.338] H [2.562] H [2.911]H*

# Boys 12-14 0.026 0.026 0.026 0.025 0.025 0.012
[2.892].H [2.877] H- [2.873] *H [2786]." [2901]". [3.271]*H

# Girls 12-14 0.071 0.072 0.072 0.072 0.071 0.03
[7.886]*H [7.918]... [7.917] *H [7.965]-H [8.020] H' [8.122]H°

Male head (=1) -0.089 -0.089 -0.089 -0.087 -0.095 -0.042
[4.444]H* [4.467]m [4.462] H" [4.390] *H [4.814] H' [4.411]*H

log(land owned) (Rp x 103) 0.779 0.750 1.008 1.147 1.063 0.001

[1.122] [1.083] [1.144] [1.251] [1.199] [2.379]

log(bus. assets) (Rp x 103) -0.192 -0.233 -0.604 -0.840 -0.871 0.441

[0.277] [0.333] [0.539] [0.736] [0.737] [1.146]

Urban (=1) -0.035 -0.035 -0.035 -0.035 -0.032 -0.43

[3.183] 9H [3.185] H2 [3.201]H' [3.205] H- [1.039] [0.851]

(continued)

145

(continued)

 

 

 

Interactions
Urban x log(land owned) x 10'3 -0.606 -1.900 -l.901 -0.966
[0.397] [1.200] [1 .215] [1.366]
Urban x log(bus. assets) x 10'3 0.626 1.465 1.427 1.06
[0.442] [0.927] [0.886] [1.540]
log(land owned) (Rp x 10'3)
x household head's schooling 0.101 0.062 0.06
[0.521] [0.318] [0.730]
x max. of non-head claimants' schooling 0.326 0.356 0.156
[1.574] [1.731]. [1.786]*
log(bus. assets) (Rp x 10'3)
x household head's schooling 0.050 0.055 -0.049
[0.283] [0.317] [0.651]
x max. of non-head claimants' schooling -0.302 -0.311 -0.164
[1.485] [1.523] [1.962]
Observations 6,227 6,227 6,227 6,227 6,227 6,227
Province, urban, interaction
dummies No No No No Yes No
Community dummies No No No No No Yes
F-test [lg-values)
Head's education 0.336 0.349 0.498
Non-head claimants' max. educ. 0.077 0.049 0.023
Education variables 0.013 0.008 0.007 0.008 0.005 0.002
Land variables 0.498 0.089 0.087 0.050
Other assets variables 0.865 0.612 0.563 0.152
Land and assets variables 0.515 0.549 0.838 0.302 0.306 0.196

 

Claimants include head of household and any member who were 15 or over in 1993. The table shows
the marginal effects of a change in explanatory variables. Standard errors are corrected for clustering
at the 1993 community level and heteroskedasticity. Absolute value of 2 statistics are in brackets

with statistical significance at 10% (*); 5% ("); and 1% (“*) indicated.

146

Table 2.6.5

Determinant of Household Division between 1993 and 2000, Claimant 2

1993 Variables:
# Claimants

Proportion of claimants:
15-19 years, male

20-29 years, male
' 30—49 years, male
50 years or older, male
12-14 years, female
15-19 years, female
20-29 years, female
30-49 years, female
50 years or older, female
# Non-claimant members 12+
Age of head
Age of head squared x 10'3

Head of hh's schooling

(1) (2) (3) (4) (5) (6)
0.118 0.118 0.118 0.118 0.120 0.131
[9.633] [9.643] [9.660] [9.656] [9.823] [9.282]

-0.601 -0.596 -0597 -0599 -0.596 -0.691
[3.383] [3.353] [3.361] [3.369] [3.347] [3.544]
-0504 -0500 -0501 -0505 -0.489 -0591
[2.907] [2.882] [2.890] [2.909] [2.808] [3.056] m
-0742 -0739 -0739 -0743 -0722 -0.816
[4.536] [4.510] [4.517] [4.527] [4.375] [4.491]
-0.861 -0.857 -0.857 -0.858 -0.828 -0972
[5.096] [5.063] [5.072] [5.066] [4.853] [5.202]
-0001 0.000 0.000 -0002 0 0.022
[0.012] [0.004] [0.002] [0.036] [0.006] [0.310]
-0.650 -0.646 -0.647 -0.648 -0.636 -071
[3.869] [3.839] [3.852] [3.850] [3.782] [3.843]
-0700 -0.696 -0.697 -0.697 -0.679 -0791
[4.122] m [4.099] [4.111] [4.107] [4.004] ... [4.237]
-0740 -0739 -0740 -0740 -0.716 -0.856
[4.422] [4.422] [4.436] [4.432] [4.273] [4.590]
-0.782 -0.789 -0791 -0792 -0.765 -0.869
[4.560] [4.637] [4.652] [4.661] [4.491] [4.613]
0.135 0.134 0.134 0.135 0.131 0.15
[2.872] m [2.846] m [2.855] m [2.869] m [2.791] m [2.875]
0.014 0.014 0.014 0.014 0.014 0.015
[3.098] [3.108] [3.113] [3.104] [3.071] [3.108]
-0103 -0103 -0103 -0103 -0.103 -0107
[2.460] .. [2.470]" [2.475] .. [2.460] .. [2.440] .. [2.322] ..
-0004 -0004 -0004 -0003 -0003 0
[1.916] r [1.812] . [1.792] . [1.339] [1.477] [0.205]

Max. schooling, non-head claimants 0.011 0.010 0.010 0.010 0.011 0.013

Sd. dev. of claimants' schooling

# Boys 0-5
# Girls 0-5
# Boys 6-11
# Girls 6-11

Male head (=1)

[4.440] [3.851] [3.883] [3.948] [4.309] [4.763]
0.005 0.005 0.004 0.004 0.005
[0.949] [0.938] [0.846] [0.885] [0.971]
0.000 0.000 0.000 0.000 0.001 0.007
[0.001] [0.002] [0.008] [0.010] [0.059] [0.522]
0.007 0.006 0.006 0.007 0.006 0.016
[0.552] [0.523] [0.533] [0.546] [0.507] [1.175]
0.065 0.065 0.065 0.065 0.065 0.072
[5.930] [5.919] [5.938] [5.889] [5.785] [5.840]
0.059 0.059 0.059 0.060 0.061 0.071
[4.847] [4.837] [4.846] [4.870] [5.024] [5.407]
-0059 -0059 -0059 -0.056 -0.067 -0.083
[1.638] [1.639] [1.629] [1.538] [1.859] [2.038]

 

(continued)

147

(continued)

 

 

 

 

log(land owned) (Rp x 103) -0.209 -0.237 -0.344 0.056 -0.178 -0.21
[0.168] [0.189] [0.232] [0.037] [0.116] [0.121]
log(bus. assets) (Rp x 103) 3.695 3.678 3.087 3.325 3.379 3.013
[2.692] [2.677] [1.433] [1.511] [1.469] [1.216]
Urban (=1) -0.064 -0.064 -0.063 -0.063 -0.018 -0.994
[3.562] H. [3.558] H* [3.411] Ht [3.381] *H [0.397] [36.958] *H
Interactions
Urban x log(land owned) x 10'3 0.885 0.247 0.873 -0371
[0.327] [0.085] [0.301] [0.1 17]
Urban x log(bus. assets) x10'3 0.920 0.124 -012 2.048
[0.328] [0.040] [0.038] [0.583]
log(land owned) (Rp x 10'3)
x household head's schooling 0.318 0.197 0.329
[0.899] [0.551] [0.854]
x max. of non-head claimants' schooling -0.053 0.06 -0.013
[0.129] [0.144] [0.029]
log(bus. assets) (Rp x 10'3)
x household head's schooling 0.239 0.241 0.13
[0.755] [0.764] [0.381]
x max. of non-head claimants' schooling 0 0.054 0.108
[0.001] [0.145] [0.273]
Observations 6,330 6,330 6,330 6,330 6,330 6,330
Province, urban, interaction dummies No No No No Yes No
Community dummies No No No No No Yes
F—test [Q-valuesz
Head's education 0.109 0.178 0.578
Non-head claimants' max. educ. 0.001 0.000 0.000
Education variables 0.000 0.000 0.000 0.001 0.000 0.000
Land variables 0.946 0.883 0.914 0.880
Other assets variables 0.029 0.106 0.118 0.057
Land and assets variables 0.012 0.014 0.055 0.087 0.095 0.081

 

Claimants include head of household and any member who were 12 or over in 1993. The table shows
the marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at
the individual level and heteroskedasticity. Absolute value of z statistics are in brackets with statistical
signiﬁcance at 10% (*); 5% (**); and 1% (***) indicated.

148

1997 Variables:
# Claimants

Proportion of claimants:

( 1)
0.063

Table 2.6.6
Determinant of Household Division between 1997 and 2000, Claimant 2

(2)
0.062

(3)
0.062

(4)
0.062

(5)
0.063

(6)
0.051

[11.900] W [11.908] W [11.866] W [11.881] W [11.936] W [12.758] W

 

20-29 years, male 0.038 0.038 0.040 0.041 0.042 0.006
[0.739] [0.750] [0.773] [0.791] [0.832] [0.171]
30-49 years, male -0. 193 -0.192 -0. 191 -0. 190 -O. 19 -0. 141
[3.347] '“" [3.343] *” [3.329] *" [3.309] *" [3.321] "" [3.446] "'"
50 years or older, male -0.213 -0.213 -0.214 -0.214 -0.212 -0.152
[3.124] ”" [3.122] ”" [3.135] ”" [3.131] ”" [3.118] ”" [3.199] "*
15-19 years, female 0.121 0.121 0.121 0.122 0.13 0.070
[2.193] " [2.189] " [2.183] ” [2.207] " [2.340] " [1.728] "
20-29 years, female -0.006 -0.006 -0.006 -0.005 -0.006 -0.037
[0.109] [0.1 16] [0.106] [0.098] [0.1 1 1] [0.944]
30-49 years, female -0.054 -0.056 -0.056 -0.055 -0.056 -0.042
[0.953] [0.985] [0.991] [0.973] [1.004] [1.015]
50 years or older, female -0.154 -0.162 -0.164 -0. 163 -0.163 -0.106
[2.618] "'" [2.719] ”'" [2.735] *" [2.730] "" [2.695] "“ [2.489] "
Age of head 0.012 0.012 0.012 0.012 0.012 0.007
[4.019] "" [4.037] "‘ [4.074] ""‘ [4.111] ”‘ [4.145] ”" [3.208] "'"
Age of head squared x 10'3 -0.084 -0.084 -0.085 -0.086 -0.087 -0.048
[3.020] “" [3.034] "" [3.064] "* [3.095] "" [3.107] ‘" [2.334] "‘
Head of hh's schooling -0.002 -0.002 -0.002 -0.002 -0.002 0.000
[1.754] "' [1.551] [1 .523] [1.493] [1.475] [0.116]
Max. schooling, non-head claimants 0.005 0.005 0.005 0.004 0.005 0.004
[3.150] ”* [2.690] "W [2.736] "”' [2.530] " [2.722] "* [3.425]
Sd. dev. of claimants' schooling 0.003 0.003 0.003 0.003 0.000
[0.954] [0.923] [0.947] [0.872] [0.009]
# Boys 0-11 0.019 0.019 0.019 0.019 0.02 0.012
[2.617] "“ [2.602] '" [2.620] "" [2.602] "”” [2.853] ‘”" [2.432]
# Girls 0-1 1 0.004 0.004 0.004 0.004 0.005 0.002
[0.586] [0.582] [0.580] [0.583] [0.649] [0.309]
# Boys 12-15 0.074 0.074 0.074 0.074 0.077 0.055
[6.589] "‘ [6.569] ‘” [6.551] ”" [6.536] "“ [6.812] "“ [6.720] "‘*
# Girls 12-15 0.105 0.105 0.105 0.105 0.108 0.080
[10.465] "“" [10.438] "‘ [10.509] "'" [10.523] "* [10.728] "" [11.121] *”
Male head (= 1) -0.041 -0.042 -0.041 -0.042 -0.046 -0.034
[1.801] "' [1.822] " [1781]" [1.813] " [1.978] " [1958]"
log(land owned) (Rp x 10") 1.997 1.968 2.585 2.396 2.502 1.179
[2.263] " [2.225] ” [2.359] [2.123] [2.142] “ [1.487]
log(bus. assets) (Rp x 10") 1.575 1.550 0.095 -0.045 -0.031 -0.112
[1.592] [1.564] [0.062] [0.030] [0.020] [0.098]
Urban (=1) -0.027 -0.027 -0.028 -0.028 -0.017 -0.026
[2066]" [2.058] " [2.137] ” [2.168] " [0.411] [0.323
(continued)

149

(continued)

 

 

 

 

Interactions
Urban x log(land owned) x 10‘3 -1023 -1.185 -1.841 -0.640
[0.537] [0.613] [0.950] [0.460]
Urban x log(bus. assets) x 10'3 2.455 2.884 2.716 1.900
[1.231] [1.362] [1.273] [1.233]
log(land owned) (Rp x 10")
x household head's schooling -0.126 -0.129 -0.078
[0.533] [0.549] [0.483]
x max. of non-head claimants' schooling 0.145 0.131 0.089
[0.557] [0.503] [0.529]
log(bus. assets) (Rp x 10‘3)
x household head's schooling -0.184 -0.189 -0.188
[0.784] [0.800] [1 . 120]
x max. of non—head claimants' schooling 0018 0.054 0.032
[0073] [0.218] 40.183]
Observations 6,664 5,664 6,664 6.664 6,664 6,664
Province, urban, interaction dummies No No No No Yes No
Community dummies No No No No No Yes
F-test [lg-values)
Head's education 0.194 0.189 0.345
Non-head claimants' max. educ. 0.007 0.034 0.004
Education variables 0.007 0.016 0.002 0.065 0.042 0.005
Land variables 0.004 0.154 0.187 0.493
Other assets variables 0.143 0.343 0.382 0.381
Land and assets variables 0.001 0.001 0.004 0.030 0.062 0.167

 

Claimants include head of household and any member who were 16 or over in 1997. The table shows the marginal
effects of a change in explanatory variables. Standard errors are corrected for clustering at the individual level and
heteroskedasticity. Absolute value of 2 statistics are in brackets with statistical signiﬁcance at 10% (‘); 5% (""); and

1% ("*) indicated.

150

Table 2.6.6

Probability of Household Division of Panel Households, LPM and

LPM with H Fixed Effects, Claimant 1

 

 

LPM with 1993 HR
LPM Fixed Effects
Base year characteristics (1) (2) (3) (4) (5)
# Claimants 0.072 0.072 0.07 0.135 0.135

[11.255] ""‘ [11.104]"* [10.656] """‘
Proportion of claimants

[10.228] W

[10.216] W

 

20-29 years, male 0.137 0.139 0.137 0.306 0.306
[3.578] "* [3.665] "”” [3.594] "" [5.254] "* [5.254] *“
30-49 years, male 0.115 0.12 0.122 0.711 0.71
[2.204] *" [2.324] “ [2.372] ” [7.421] ”" [7.394] "*
50 years or older, male 0.034 0.041 0.038 0.86 0.859
[0.699] [0.853] [0.772] [7.629] "”" [7.613] ""
15-19 years, female 0.03 0.03 0.029 0.045 0.045
[0.888] [0.867] [0.840] [0.646] [0.650]
20-29 years, female 0.121 0.122 0.122 0.463 0.463
[3.123] "" [3.152] [3.132] [5.703] *" [5.698]
3049 years, female 0.059 0.062 0.061 0.594 0.593
[1.288] [1.380] [1.350] [5.449] "" [5.419] ""
50 years or older, female -0.009 0.003 -0.003 0.517 0.514
[0.189] [0.052] [0.069] [3.841] “" [3.785] “"
# Non-claimant members 0.022 0.021 0.024 0.013 0.014
[2.649] "* [2.577] [2.871] "" [0.706] [0.708]
Head's schooling —0.003
[1.681] "
Max. schooling, non-head claimants 0.004 0.004 0.007 0.018 0.017
[1.844] * [2.161] ” [2.855] "" [3.880] *" [3.487] *"
Sd. dev. of claimants' schooling -0.004 -0.009 0.001
[1.163] [1.933] " [0.194]
# Boys 0-11 0.01 0.01 0.01 -0.051 -0.051
[0.910] [0.930] [0.887] [2.835] "* [2.830] "'"
# Girls 0-1 1 0.023 0.023 0.023 0.174 0.174
[1.853] " [1.834] ‘ [1.791] "‘ [6.770] "* [6.771] *"
# Boys 12-14 0.028 0.028 0.028 0.107 0.107
[2.231] “ [2.236] “ [2.233] " [4.676] "" [4.679] ”"'
# Girls 12-14 0.043 0.043 0.043 0.213 0.213
[4.002] "‘" [4.005] "“ [3.975] "”" [7.789] *" [7.788] ”"
Year(1997=l) 0.217 0.217 0.217
[16.063] "'" [16.063] "'" [16.081] "”"
Urban -0.106 -0.107 -0.103 -0.041 -0.042
[2.290] " [2.317] " [2.200] " [0.265] [0.274]
Constant -0.219 -0.217 -0.204 -1.021 -0.9
3.090] ‘" [4.073]*" [3.707] "" [6.143] "‘" [4.889 *"
F-test (pvalues) for education variables - 0.092 0.071 - 0.001
Number of observations 5076 5076 5076 5076 5076
R-squared 0.1425 0.1427 0.1242 0.1624 0. 1624
Number of unique households 2538 2538 2538 2538 2538

 

Panel households are households appearing in 1993, 1997, and 2000. The estimation uses household variables in 1993
and 1997 to estimate the probability of household division by 1997, and 2000, respectively (i.e., for each household
there are two observations). The dependent variable is 1 for 1993 household if the household split by 1997, and 0
otherwise. The dependent variable is l for households that have split by 2000, and 0 otherwise. The explanatory
variables are the base year household variables, namely 1993 household variables for 1993 households, and 1997
household variables for households observed in 1997. Claimant 2 consists of households with multiple claimants where
claimants include household head, and other household members who would be at least 19 years old in 1997 for 1993

households, or 19 years old in 2000 for 1997 households.

151

Table 2.6.7
Probability of Household Division of Panel Households, LPM and
LPM with H Fixed Effects, Claimant 1

LPM

LPM with 1993 HH

 

 

 

 

Fixed Effects
Base year characteristics (1) (2) (3) (4) (5)
# Claimants 0.074 0.074 0.072 0.096 0.096
[19.690] *" [19.434] *“ [18.984] "”" [13.563] *“ [13.520] *"
Proportion of claiman_ts
20-29 years, male 0.121 0.121 0.126 0.339 0.34
[3.435] “'" [3.448] "‘ [3.563] "“ [6.492] *""“ [6.499] "*
30-49 years, male -0.013 -0.012 0.004 0.278 0.279
[0.352] [0.318] [0.116] [4.556] "" [4.577] "'"
50 years or older, male 0.023 0.024 0.033 0.412 0.414
[0.621] [0.656] [0.896] [5.603] ""* [5.627] ""'
15-19 years, female 0.052 0.052 0.056 0.076 0.075
[1.291] [1.289] [1 .380] [1.266] [1.247]
20-29 years, female 0.019 0.019 0.029 0.342 0.34
[0.518] [0.529] [0.784] [5.340] "" [5.306] ”*
30-49 years, female 0.05 0.048 0.057 0.31 0.307
[1.294] [1.258] [1 .475] [4.473] "" [4.432] ”"'
50 years or older, female 0.043 0.034 0.034 0.3 0.294
[1.165] [0.925] [0.923] [3.956] ”" [3.843] "‘
Head's schooling -0.004
[4.085] ""
Max. schooling, non-head claimants 0.004 0.003 0.005 0.016 0.016
[3.613] *** [3.048] *" [4.843] "“ [6.831] *" [6.102] *"
Sd. dev. of claimants' schooling 0.004 0.003 0.004
[1 .908] "‘ [1.486] [0.844]
# Boys 0-1 1 -0.012 -0.012 -0.011 -0.04 -0.04
[1.921] * [1.934] " [1.879] " [4.047] "" [4.059] "”""'
# Girls 0-1 1 0.039 0.039 0.039 0.108 0.108
[5.368] ”* [5.366] *" [5.274] "'" [8.441] "”” [8.449] ”"'
# Boys 12-14 0.007 0.007 0.008 0.073 0.073
[1.166] [1.158] [1.232] [5.424] *" [5.436] ”"
# Girls 12-14 0.036 0.036 0.036 0.155 0.155
[5.938] "" [5.888] "" [5.943] "" [10.299] "" [10.311] ”'"
Year(1997=1) 0.109 0.11 0.109
[13.844] "“ [13.864] "* [13.876] ""
Urban -0.047 -0.047 -0.044 0.084 0.081
[1.787] " [1.816] " [1.655] " [0.904] [0.871]
Constant -0.22 -0.222 -0.2 1 5 -0. 704 -0. 705
[5.341] "'" [5.382 *“ [5.149 '" [3.644 ""' [3.653 “"
F-test (pvalues) for education variables 0.124 0.124 0.126 0.11 1 0.1 1 1
Number of observations 1 1,774 1 1,774 1 1,774 11,774 1 1,774
R-squared 0.124 0.124 0.125 0.107 0.108
Number of ﬂue households 5887 5887 5887 5887 5887

 

Panel households are households appearing in 1993, 1997, and 2000. The estimation uses household variables in 1993 and
1997 to estimate the probability of household division by 1997, and 2000, respectively (i.e., for each household there are
two observations). The dependent variable is l for 1993 household if the household split by 1997, and 0 otherwise. The
dependent variable is 1 for households that have split by 2000, and 0 otherwise. The explanatory variables are the base
year household variables, namely 1993 household variables for 1993 households, and 1997 household variables for
households observed in 1997. Claimant 2 consists of households with multiple claimants Where claimants include
household head, and other household members who would be at least 19 years old in 1997 for 1993 households, or 19

years old in 2000 for 1997 households.

152

Appendix Table 2.6.1

Determinant of Household Division between 1993 and 1997, Claimant 1: Urban

 

 

 

 

1993 Variables: (1) (2) (3) (4) (5)
# Claimants 0.056 0.053 0.052 0.054 0.011
[6.642] "* [6.255] *" [6.163] *** [6.456] ”* [6.341] ***
Proportion of claimants;
20-29 years, male 0.280 0.282 0.284 0.297 0.058
[3.953] **"' [4.028] *" [4.054] “* [4.153] *“ [4.351] "*
30-49 years, male 0.368 0.381 0.380 0.387 0.078
[4.323] **"' [4.520] *" [4.519] *** [4.680] “* [4.594] ***
50 years or older, male 0.389 0.395 0.395 0.391 0.073
[3.234] *" [3.321] *** [3.320] *** [3.299] *" [2.961] *"
15-19 years, female 0.016 0.022 0.024 0.038 0.011
[0.249] [0.346] [0.374] [0.601] [0.891]
20—29 years, female 0.177 0.182 0.187 0.186 0.045
[2.589] *" [2.677] *" [2.754] *** [2.672] “”" [3.289] *"
30-49 years, female 0.342 0.355 0.358 0.348 0.064
[3.723] "'" [3.908] *** [3.972] *" [3.941] **"' [3.688] “*
50 years or older, female 0.214 0.238 0.232 0.233 0.041
[1.790] "' [1.992] ** [1.955] “ [2.024] " [1.779] *
# Non-claimant members 15 + 0.021 0.022 0.021 0.02 0.004
[2.218] "* [2.338] ** [2.259] "”' [2.080] ** [2.121] ”
Age of head -0.011 -0.010 -0.011 -0.011 -0.001
[1.899] * [1.845] * [1.879] * [1.890] "' [1.324]
Age of head squared x 10'3 0.081 0.078 0.080 0.08 0.009
[1.508] [1.458] [1.506] [1.508] [0.875]
Head of hh's schooling -0.003 -0.007 -0.005 -0.005 -0.001
[1.464] [2.370] [1.446] [1.336] [1.035]
Max. schooling, non-head claimants 0.006 0.010 0.010 0.011 0.002
[2.338] " [3.146] *" [3.072] **" [3.274] *" [2.976] "*
Sd. dev. of claimants' schooling -0.013 -0.013 -0.013 -0.002
[2.031] ** [1.967] ** [1.991] " [1.910] "'
# Boys 0-11 0.028 0.027 0.026 0.026 0.003
[2.388] ** [2.290] ** [2.255] ** [2.228] ” [1.458]
# Girls 0-11 0.019 0.019 0.019 0.021 0.002
[1.618] [1.612] [1.614] [1.854] "' [0.926]
# Boys 12-14 0.001 0.001 0.001 0.002 0.002
[0.049] [0.033] [0.031] [0.092] [0.546]
# Girls 12-14 0.054 0.053 0.052 0.051 0.01
[2.900] "* [2.854] "* [2.834] *” [2.825] "‘" [2.595] *"
Male head (=1) -0.092 -0.090 -0.087 -0.095 -0.025
[1.807] * [1.793] "' [1.745] " [2.004] *2 [2.175] *2
log(land owned) (Rp x 103) 0.809 0.891 0.535 0.588 -0.026
[0.455] [0.506] [0.275] [0.302] [0.064]
log(bus. assets) (Rp x 10'3) -l.077 -l.041 -1.998 -2.468 -0.344
[0.801] [0.776] [1.402] [1.701] "‘ [1.155]
(continued)

153

(continued)

 

 

 

 

Interactions
log(land owned) (Rp x 10")
x household head's schooling 0.310 0.253 0.073
[0.639] [0.521] [0.769]
x max. of non-head claimants' schooling -0.027 -0.066 -0.019
[0.048] [0.1 19] [0.178]
log(bus. assets) (Rp x 10")
x household head's schooling 0.146 0.16 0.012
[0.420] [0.463] [0.179]
x max. of non-head claimants' schooling 0.357 0.364 0.063
[0.809] [0.827] [0.701]
Observations 1,652 1,652 1,652 1,652 1,652
Province dummies No No No Yes No
Community dummies No No No No Yes
F-test [Q-valuesz
Head's education 0.095 0.130 0.230
Non-head claimants' max. educ. 0.022 0.011 0.029
Education variables 0.045 0.014 0.063 0.052 0.129
Land variables 0.851 0.920 0.884
Other assets variables 0.482 0.353 0.675
Land and assets variables 0.714 0.717 0.640 0.628 0.797

 

Claimants include head of household and children or siblings who were 12 or over in 1993. The table
shows the marginal effects of a change in explanatory variables. Standard errors are corrected for
clustering at the 1993 community level and heteroskedasticity. Absolute value of 2 statistics are in

brackets with statistical signiﬁcance at 10% (*); 5% (”); and 1% (***) indicated.

154

Appendix Table 2.6.2

Determinant of Household Division between 1993 and 1997, Claimant 1: Rural

1993 Variables:

(1) (2) (3) (4) (5) (6)

 

# Claimants

Proportion of claimants:
20-29 years, male

30.49 years, male
50 years or older, male
15-19 years, female
20-29 years, female
30—49 years, female
50 years or older, female
# Non-claimant members 15+
Age of head
Age of head squared x 10‘3

Head of hh's schooling

Sd. dev. of claimants' schooling

# Boys 0-11
# Girls 0-11

# Boys 12-14

0.062 0.063 0.063 0.063 0.063 0.041
[4.506] *** [4.447] *" [4.462] *" [4.420] *** [4.572] 2" [4.960] ***

0.281 0.281 0.282 0.281 0.29 0.149
[4.600] m [4.577] m [4.584] m [4.570] m [4.830] m [4.186] m
0.358 0.356 0.357 0.357 0.364 0.209
[3.670] m [3.618] m [3.615] m [3.617] m [3.742] m [3.765] m
0.341 0.337 0.338 0.337 0.329 0.2
[2.629] m [2.606] m [2.609] m [2.588] m [2.580] m [2.778] m
0.059 0.060 0.057 0.057 0.059 0.054
[0.843] [0.852] [0.821] [0.830] [0.908] [1.511]
0.283 0.282 0.280 0.279 0.309 0.201
[3.915] m [3.906] m [3.859] m [3.853] m [4.370] m [4.971] m
0.319 0.319 0.321 0.321 0.357 0.23
[3.124] m [3.105] m [3.119] m [3.141] m [3.519] m [3.993] m
0.229 0.226 0.227 0.228 0.246 0.162
[1.848] * [1.824] * [1.815] * [1.855] * [2.020] H [2.262] H
0.011 0.011 0.010 0.010 0.009 0.007

[0.678] [0.649] [0.568] [0.565] [0.530] [0.690]
-0010 -0010 -0010 -0010 -0011 -0.006
[1.497] [1.487] [1.539] [1.507] [1.649] [1.508]

0.085 0.084 0.087 0.086 0.091 0.047

[1.534] [1.525] [1.582] [1.550] [1.645] * [1.438]
-0.002 -0.001 -0.001 -0.001 -0.002 -0.001
[0.501] [0.190] [0.175] [0.158] [0.512] [0.306]
Max. schooling, non-head claimants 0.000 -0.001 -0.001 -0.001 -0.001 0.002
[0.105] [0.174] [0.198] [0.185] [0.131] [0.387]
0.003 0.003 0.002 0.001 -0.981

[0.358] [0.359] [0.305] [0.157] [0.515]

0.011 0.011 0.011 0.011 0.012 0.01
[1.020] [1.010] [1.001] [0.994] [1.098] [1.589]
0.012 0.013 0.012 0.012 0.011 0.009
[1 .030] [1.046] [1.007] [1.000] [0.940] [1.281]
0.012 0.012 0.011 0.011 0.011 0.011
[0.509] [0.511] [0.490] [0.465] [0.462] [0.807]

# Girls 12-14
Male head (=1)

log(land owned) (Rp x 10")

0.129 0.129 0.129 0.130 0.128 0.079
[6.173] m [6.188] m [6.216] m [6.319] m [6.212] m [6.267] m

0.012 0.013 0.012 0.012 0.016 0.013
[0.226] [0.234] [0.226] [0.225] [0.297] [0.429]

3.496 3.498 10.123 10.499 12.223 -0001

[1.773] * [1.778] * [1.359] [1.419] [1.731] * [0.202]
log(bus. assets) (Rp x10'3) -0.851 -0.827 -2.762 -3144 -357 6.34
[0.338] [0.329] [0.938] [1.049] [1.202] [1.578]
Farm household (=1) -0035 -0035 -0.067 -0.069 -0.088 -0043
[0.890] [0.892] [1.284] [1.303] [1.633] [1.338]

 

155

(continued)

 

 

 

 

Interactions

Farm hh (=1) x log(land owned) (Rp x 103) -7.386 -7.300 -8.899 -4.323
[0.968] [0.953] [1.203] [1.004]
Farm hh (=1) x log(bus. assets) (Rp x 103) 4.755 4.688 5.76 1.07
[0.782] [0.761] [0.903] [0.290]
log(land owned) (Rp x 10'3) x head's schooling 0.156 0.017 0.183
[0.291] [0.032] [0.635]
x max. of non-head claimants' schooling 0.084 0.199 0.018
[0.147] [0.349] [0.052]
log(bus. assets) (Rp x 10'3) x head's schooling -0.233 -0.125 -0.352
[0.383] [0.206] [1.014]
x max. of non-head claimants' schooling -0.05 -0.16 -0.112
[0.064] 40.204] [0.238]
Observations 1,618 1,618 1,618 1,618 1,618 1,618
Province dummies No No No No Yes No
Community dummies No No No No No Yes

F-test [Q-valuesz
Head's education 0.976 0.931 0.663
Non-head claimants' max. educ. 0.996 0.984 0.976
Education variables 0.879 0.941 0.942 0.998 0.990 0.804
Land variables 0.169 0.333 0.182 0.215
Other assets variables 0.604 0.867 0.811 0.753
Land and assets variables 0.208 0.206 0.366 0.702 0.487 0.606

 

Claimants include head of household and children or siblings who were 12 or over in 1993. The table shows
the marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at the
1993 community level and heteroskedasticity. Absolute value of 2 statistics are in brackets with statistical
signiﬁcance at 10% (*); 5% (**); and 1% (***) indicated.

156

Appendix Table 2.6.3
Determinant of Household Division between 1993 and 2000, Claimant 1: Urban
1993 Variables: (1) (2) (3) (4) (5)
# Claimants 0.102 0.101 0.101 0.108 0.101
[7.324] m [7.149] m [7.093] m [7.609] m [7.093] 1'"

 

Proportion of claimants:

15-19 years, male -0.210 -0.215 -0.212 -0.227 -0.212
[1.845] "‘ [1.877] * [1.845] "' [1.995] ** [1.845] "'
20-29 years, male 0.099 0.096 0.101 0.111 0.101
[0.903] [0.880] [0.918] [0.996] [0.918]
30-49 years, male -0.280 -0.280 -0.274 -0.253 -0.274
[2.071] " [2.072] ** [2.020] ** [1.884] * [2.020] **
50 years or older, male -0.215 -0.221 -0.213 —0. 178 -0.213
[1.370] [1.404] [1.351] [1.152] [1.351]
12-14 years, female -0.038 -0.039 -0.037 -0.016 -0.037
[0.354] [0.362] [0.338] [0.146] [0.338]
15-19 years, female -0.110 -0.115 -0.109 -0.093 -0.109
[1.005] [1.046] [0.978] [0.838] [0.978]
20-29 years, female 003? -0.042 -0.034 -0.034 -0.034
[0.324] [0.366] [0.296] [0.295] [0.296]
30-49 years, female -0.073 -0.074 -0.067 -0.059 -0.067
[0.487] [0.496] [0.451] [0.388] [0.451]
50 years or older, female -0.044 -0.036 -0.027 0.008 -0.027
[0.238] [0.197] [0.145] [0.042] [0.145]
# Non-claimant members 12+ 0.078 0.078 0.077 0.076 0.077
[4.337] *" [4.319] **"' [4.292] *" [4.271] *" [4.292] “*
Age of head -0.011 -0.011 -0.011 -0.012 -0.011
[1 .468] [1.426] [1 .476] [1.660] * [1.476]
Age of head squared x 10'3 0.081 0.080 0.083 0.092 0.083
[1.102] [1.084] [1.129] [1.259] [1.129]
Head of hh's schooling -0.003 -0.003 0.001 0.002 0.001
[0.732] [0.905] [0.206] [0.327] [0.206]
Max. schooling, non-head claimants 0.011 0.013 0.011 0.012 0.011
[2.671] **"' [2.737] "* [2.423] *2 [2.704] *"”" [2.423] **
Sd. dev. of claimants' schooling -0.006 -0.005 -0.007 -0.005
[0.729] [0.589] [0.797] [0.589]
# Boys 0-5 0.049 0.050 0.049 0.056 0.049
[2.189] ** [2.210] ** [2.198] ** [2.466] " [2.198] *"'
# Girls 0-5 0.003 0.003 0.004 -0.001 0.004
[0.1 10] [0. 130] [0.156] [0.057] [0.156]
# Boys 6-11 0.048 0.047 0.047 0.046 0.047
[2.347] *"‘ [2.315] ** [2.297] ” [2.126] ** [2.297] "
# Girls 6-11 0.020 0.019 0.019 0.02 0.019
[0.993] [0.969] [0.969] [1.017] [0.969]
Male head (=1) 0.048 0.049 0.053 0.051 0.053
[0.790] [0.797] [0.852] [0.817] [0.852]

 

157

 

(continued)

log(land owned) (Rp x 10") 0.611 0.719 1.043 0.608 1.043
[0.229] [0.271] [0.343] [0.201] [0.343]
log(bus. assets) (Rp x 103) 4.854 4.817 4.575 3.837 4.575
[2.247] ** [2.241] ** [1.907] * [1.611] [1.907] "‘
Interactions
log(land owned) (Rp x 10")
x household head's schooling 0.675 0.552 0.675
[0.959] [0.777] [0.959]
x max. of non-head claimants' schooling -0.355 -0.307 -0.355

[0.432] [0.383] [0.432]
log(bus. assets) (Rp x 10'3)

 

 

 

x household head's schooling 0.523 0.492 0.523
[1.221] [1.129] [1.221]
x max. of non-head claimants' schooling '0-47 -0.376 -0.47
[0764] [0.626L [0.764]
Observations 1,911 1,911 1,911 1,911 1,911
Province dummies No No No Yes No
Community dummies No No No No Yes
F-test (Q-valuesz
Head's education 0.050 0.374 0.228
Non-head claimants' max. educ. 0.228 0.037 0.050
Education variables 0.028 0.053 0.082 0.076 0.082
Land variables 0.774 0.884 0.774
Other assets variables 0.097 0.179 0.097
Land and assets variables 0.0291 0.028 0.040 0.162 0.040

 

Claimants include head of household and children or siblings who were 12 or over in 1993. The table
shows the marginal effects of a change in explanatory variables. Standard errors are corrected for
clustering at the individual level and heteroskedasticity. Absolute value of z statistics are in brackets
with statistical significance at 10% (*); 5% (**); and 1% (***) indicated.

158

Appendix Table 2.6.4
Determinant of Household Division between 1993 and 2000, Claimant 1: Rural

1993 Variables: (1) (2) (3) (4) (5) (6)

 

# Claimants 0.133 0.129 0.128 0.128 0.134 0.144
[7.872] m [7.577] m [7.475] m [7.435] m [7.983] m [7.440] m
Proportion of claimants: ‘

15-19 years, male -0.090 -0.096 -0.093 -0.093 -0.104 -0.155
[0.899] [0.960] [0.928] [0.927] [1.044] [1.415]
20-29 years, male 0.132 0.130 0.136 0.138 0.154 0.134
[1.207] [1.184] [1.240] [1.263] [1.381] [1.093]
30—49 years, male -0.103 -0.100 -0.094 -0.095 -0.073 -0.156
[0.734] [0.712] [0.670] [0.673] [0.507] [0.948]
50 years or older, male -0.082 -0.075 -0.068 -0.065 -0.045 -0.187
[0.478] [0.435] [0.399] [0.378] [0.255] [0.928]
12-14 years, female 0.081 0.077 0.079 0.074 0.075 0.082
[0.918] [0.869] [0.894] [0.843] [0.847] [0.841]
15-19 years, female -0.131 -0. 140 -0.137 -0.137 -0. 144 -0.19
[1.295] [1.403] [1.375] [1.375] [1.451] [1.754] *
20-29 years, female -0.078 -0.079 -0.080 -0.079 -0.057 -0.074
[0.702] [0.718] [0.723] [0.713] [0.500] [0.618]
30-49 years, female -0.227 -0.231 -0.229 -0.232 -0.l97 -0.186
[1.595] [1.613] [1.602] [1.626] [1.336] [1.134]
50 years or older, female -0.276 -0.271 -0.264 -0.280 -0.265 -0.263
[1.523] [1.487] [1.436] [1.541] [1.425] [1.402]
# Non-claimant members 12+ 0.055 0.056 0.055 0.055 0.058 0.062
[2.824] "* [2.880] *" [2.762] **"' [2.752] "" [2.918] "" [2.754] ""'
Age of head -0.001 -0.002 -0.002 —0.003 -0.004 -0.004
[0.217] [0.227] [0.281] [0.443] [0.508] [0.578]
Age of head squared x 10'3 0.018 0.018 0.022 0.033 0.037 0.056
[0.284] [0.290] [0.352] [0.517] [0.566] [0.810]
Head of hh's schooling 0.002 -0.001 -0.001 -0.003 -0.003 -0.001

[0.445] [0.242] [0.249] [0.606] [0.542] [0.260]
Max. schooling, non-head claimants 0.002 0.007 0.007 0.009 0.009 0.013
[0.519] [1.387] [1.377] [1.766] "' [1.712] * [2.130] **

Sd. dev. of claimants' schooling -0.013 -0.013 -0.012 -0.009 -0.012

[1.617] [1.631] [1.386] [1.119] [1.228]

# Boys 0-5 0.011 0.010 0.010 0.012 0.01 0.026

[0.491] [0.459] [0.481] [0.554] [0.472] [1.055]

# Girls 0-5 0.047 0.046 0.045 0.046 0.046 0.076
[2.277] "“" [2.244] *"' [2.223] " [2.243] " [2.223] "“" [3.099] "*

# Boys 6-11 0.068 0.068 0.069 0.068 0.069 0.075
[3.740] *** [3.742] *" [3.789] "'" [3.739] *" [3.767] *** [3.506] ***

# Girls 6-11 0.053 0.052 0.053 0.052 0.055 0.063
[2.616] *** [2.575] [2.620] *" [2.604] **"' [2.856] *** [2.846] "*

Male head (=1) -0.074 -0.077 -0.078 -0.077 -0.091 -0.055

[1.075] [1.119] [1.131] 11.127] [1.332] [0.757]

 

159

(continued)

 

 

 

 

log(land owned) (Rp x 10") -0.053 -0.055 -0.066 -0.061 -0.067 0.945
[1.348] [1.407] [1.380] [1.271] [1.320] [0.096]
log(bus. assets) (Rp x 103) 2.758 2.849 0.219 -0.257 1.161 3.811
[1.234] [1.274] [0.024] [0.029] [0.132] [0.917]
Farm household (=1) 3.337 3.337 1.019 2.359 1.88 -0.086
[1.053] [1 .046] [0.260] [0.596] [0.464] [1.566]
Interactions

Farm hh (=1) x log(land owned) (Rp x 103) 2.329 1.920 0.339 3.275
[0.244] [0.203] [0.036] [0.308]
Farm hh (=1) x log(bus. assets) (Rp x 103) 8.612 9.155 9.822 5.847
[1.389] [1.454] [1.560] [0.779]
log(land owned) (Rp x 10'3) x head's schooling -0.040 -0.207 0.101
[0.056] [0.293] [0.133]
x max. of non-head claimants' schooling -0.753 -0.721 -0.603
[0.960] [0.925] [0.714]
log(bus. assets) (Rp x 103) x head's schooling 1.119 1.308 1.008
[1.374] [1.574] [1.141]
x max. of non-head claimants' schooling 0109 0.14 -0.471
[0-1 171 [0.149] [0.489]
Observations 2,040 2,040 2,040 2,040 2,040 2,040
Province dummies No No No No Yes No
Community dummies No No No No No Yes

F-test (Q-valuesz
Head's education 0.329 0.302 0.394
Non-head claimants' max. educ. 0.251 0.268 0.1 14
Education variables 0.744 0.332 0.329 0.473 0.487 0.303
Land variables 0.550 0.659 0.595 0.501
Other assets variables 0.183 0.229 0.172 0.480
Land and assets variables 0.1821 0.172 0.262 0.349 0.321 0.284

 

Claimants include head of household and children or siblings who were 12 or over in 1993. The table
shows the marginal effects of a change in explanatory variables. Standard errors are corrected for
clustering at the individual level and heteroskedasticity. Absolute value of 2 statistics are in brackets

with statistical significance at 10% (‘); 5% (**); and 1% (***) indicated.

160

 

Appendix Table 2.6.5

Determinant of Household Division between 1997 and 2000, Claimant 1: Urban

1997 Variables:

(1)

(2)

(3)

(4)

(5)

 

# Claimants

Proportion of claimants:

0.068

[5.076] *H [5.031]*** [4.887] m [5.265] m [5.078] m

0.068

0.066

0.070

0.071

E

 

20-29 years, male 0.152 0.152 0.157 0.160 0.192
[2.007] ** [2.001] ** [2.066] ** [2.093] ** [2.630] "*
30-49 years, male 0.147 0.148 0.142 0.154 0.155
[1.373] [1.386] [1.319] [1.431] [1.423]
50 years or older, male 0.033 0.034 0.010 0.042 0.008
[0.225] [0.232] [0.070] [0.282] [0.053]
16-19 years, female 0.098 0.098 0.097 0.102 0.105
[1.231] [1.227] [1.222] [1.285] [1.360]
20-29 years, female 0.282 0.283 0.290 0.276 0.284
[3.120] *** [3.119] "* [3.188] "" [3.012] 2" [3.372] "*
30-49 years, female -0.140 -0.139 -0.155 ~0.159 -0.056
[1.271] [1.272] [1.429] [1.444] [0.514]
50 years or older, female -0.229 -0.228 -0.263 -0.257 -0.164
[1.355] [1.354] [1.565] [1.530] [0.966]
# Non-claimant members 16+ 0.019 0.019 0.021 0.019 0.047
[1.200] [1.203] [1.320] [1.170] [2.544] "
Age of head 0.013 0.013 0.014 0.013 0.014
[1.624] [1.619] [1.679] * [1.541] [1.582]
Age of head squared x 10'3 -0.100 -0.100 -0.106 -0.098 -0.117
[1.392] [1.386] [1.431] [1.322] [1.407]
Head of hh's schooling 0.002 0.002 0.001 0.001 0.000
[0.661] [0.438] [0.148] [0.147] [0.019]
Max. schooling, non-head claimants 0.002 0.003 0.002 0.003 0.010
[0.516] [0.516] [0.356] [0.577] [1.830] *
Sd. dev. of claimants' schooling -0.001 0.001 0.000 -0.006
[0.099] [0.071] [0.001] [0.612]
# Boys 0-11 0.040 0.040 0.041 0.041 0.047
[2.332] ** [2.339] ** [2.383] ** [2.354] " [2.715] ***
# Girls 0-11 0.001 0.001 0.002 0.000 -0.009
[0.079] [0.074] [0.1 14] [0.003] [0.481]
# Boys 12-15 0.019 0.019 0.018 0.022 0.024
[0.697] [0.695] [0.668] [0.794] [0.895]
# Girls 12-15 0.087 0.087 0.086 0.092 0.113
[3.822] *** [3.820] *** [3.779] **"' [3.955] **"‘ [4.825] ***
Male head (=1) -0.104 -0.104 -0.111 -0.114 -0.079
[1.820] * [1.819] * [1.938] * [1.960] ** [1.277]
log(land owned) (Rp x 103) 2.661 2.654 2.306 1.280 -0.603
[1.022] [1.020] [0.814] [0.452] [0.225]
log(bus. assets) (Rp x 103) 5.176 5.181 6.927 6.694 5.090
[2.696] *** [2.701] *" [3.339] *** [3.152] *" [2.477] **
(continued)

161

(continued)

 

 

 

Interactions
log(land owned) (Rp x 10'3)
x household head's schooling -0.430 -0.405 -0.340
[0.656] [0.621] [0.548]
x max. of non-head claimants' schooling 0.653 0.627 1.118
[0.917] [0.877] [1.610]
log(bus. assets) (Rp x 10'3)
x household head's schooling 0.386 0.369 0.429
[0.797] [0.767] [0.920]
x max. of non-head claimants' schooling -1 .456 -1.389 -1.314
[2.338] [2.234] [2.160]
Observations 1,721 1,721 1,721 1,721 1,721
Province dummies No No No Yes No
Community dummies No No No No Yes
F-test -values
Head's education 0.784 0.81 1 0.817
Non-head claimants' max. educ. 0.131 0.145 0.037
Education variables 0.546 0.743 0.460 0.477 0.208
Land variables 0.515 0.694 0.442
Other assets variables 0.004 0.007 0.039
Land and assets variables 0.003 0.003 0.005 0.016 0.126

 

Claimants include head of household and children or siblings who were 16 or over in 1997. The table
shows the marginal effects of a change in explanatory variables. Standard errors are corrected for

clustering at the individual level and heteroskedasticity. Absolute value of z statistics are in brackets
with statistical signiﬁcance at 10% C“); 5% (**); and 1% (***) indicated.

162

 

 

1997 Variables:

Appendix Table 2.6.6
Determinant of Household Division between 1997 and 2000, Claimant 1: Rural

(1)

(2)

(3)

(4)

(5)

(6)

 

# Claimants

Proportion of claimants:

0.068

0.065

0.065

0.065

0.066

0.064

[4.030] "* [3.771] "* [3.724] 9'" [3.754] Ma- [3951] stun [3.831] a"

 

20-29 years, male 0.058 0.062 0.061 0.06 0.059 0.011

[0.750] [0.812] [0.790] [0.781] [0.767] [0.142]

3049 years, male -0.156 -0.149 -0.153 -0.154 -0.155 -0.209

[1.411] [1.349] [1.383] [1.375] [1.340] [1.714] "'

50 years or older, male -0.246 -0.236 -0.238 -0.236 -0.239 -0.239

[1.659] * [1.595] [1.597] [1.580] [1.534] [1.485]

16-19 years, female 0.15 0.151 0.147 0.147 0.154 0.083

[1.909] * [1.912] * [1.861] * [1.846] * [1.916] "' [0.935]

20-29 years, female 0.014 0.018 0.017 0.015 0.031 -0.002

[0.163] [0.201] [0.198] [0.174] [0.361] [0.021]

3049 years, female -0.232 -0.228 -0.232 -0.234 -0.243 -0.204

[2.103] ** [2.063] ** [2.089] " [2.111] " [2.067] ** [1.584]

50 years or older, female -0.232 -0.219 -0.221 -0.218 -0.209 -0.142

[1.345] [1.284] [1.276] [1.271] [1.219] [0.814]

# Non-claimant members 16+ 0.058 0.058 0.057 0.058 0.059 0.071
[2.803] *** [2.792] *" [2.770] "* [2.797] "* [2.822] *** [3.154] **"‘

Age of head 0.003 0.002 0.002 0.002 0.003 0.001

[0.378] [0.289] [0.254] [0.293] [0.321] [0.083]

Age of head squared x 10'3 -0.01 -0.004 -0.002 -0.005 -0.006 0.008

[0.141] [0.060] [0.030] [0.068] [0.082] [0.104]

Head of hh's schooling -0.003 -0.006 -0.006 -0.004 -0.003 0

[0.818] [1.242] [1.227] [0.891] [0.637] [0.023]

Max. schooling, non-head claimants 0.005 0.009 0.009 0.009 0.01 0.007

[1.348] [1.605] [1.612] [1.643] [1.739] " [1.117]

Sd. dev. of claimants' schooling -0.008 -0.009 -0.008 -0.008 -0.007

[0.888] [0.898] [0.808] [0.820] [0.670]

# Boys 0-11 0.009 0.009 0.008 0.009 0.01 0.012

[0.527] [0.532] [0.513] [0.545] [0.609] [0.734]

# Girls 0-11 0.036 0.036 0.035 0.035 0.037 0.057
[2.247] ** [2.257] ** [2.216] ** [2.236] ** [2.260] ** [3.055] ***

# Boys 12-15 0.099 0.099 0.099 0.099 0.105 0.097
[4.111] 2" [4.120] *** [4.153] "* [4.126] **"' [4.547] *** [3.738] "'**

# Girls 12-15 0.129 0.129 0.128 0.128 0.133 0.137
[5.534] *** [5.538] "* [5.456] *" [5.451] “" [5.568] “* [5.374] ***

Male head (=1) -0.048 -0.048 -0.047 -0.05 -0.058 -0.034

[0.692] [0.685] [0.668] [0.708] [0.809] [0.468]

log(land owned) (Rp x 103) 0.001 0.001 -0.019 -0.023 -0.011 8.445

[0.022] [0.030] [0.392] [0.474] [0.237] [1.518]

log(bus. assets) (Rp x 103) 2.598 2.685 10.005 10.059 9.423 3.663

[1.069] [1.108] [1.988] ** [1.988] ** [1.802] "‘ [0.971]

Farm household (=1) 2.916 2.809 1.821 1.471 1.92 -0.008

[1.069] [1.030] [0.571] [0.444] [0.552] [0.160]
(continued)

163

(continued)

 

 

 

 

Interactions

Farm hh (=1) x log(land owned) (Rp x 103) -8.709 -8.802 -8.644 -8.311
[1 .454] [1.451] [1.398] [1.264]
Farmhh (=1) x log(bus. assets) (Rp x10'3) 1.854 2.43 0.444 -1915
[0.273] [0.352] [0.065] [0.263]
log(land owned) (Rp x 10'3) x head's schooling 0.283 0.277 -0.03
[0.468] [0.460] [0.046]
x max. of non-head claimants' schooling -0.633 —0.706 -0.749
[1.004] [1.071] [1.131]
log(bus. assets) (Rp x 10'3) x head's schooling -0.546 -0.537 -0.288
[0.743] [0.727] [0.351]
x max. of non-head claimants' schooling 0.45 0.459 1.011
[0.605] [0.606] [1.262]
Observations l ,696 1 ,696 1,696 1,696 1,696 1 ,696
Province dummies No No No No Yes No
Community dummies No No No No No Yes

F-test [Q-valuesz
Head's education 0.632 0.754 0.970
Non-head claimants' max. educ. 0.314 0.275 0.337
Education variables 0.376 0.417 0.417 0.749 0.719 0.789
Land variables 0.115 0.251 0.358 0.374
Other assets variables 0.691 0.805 0.831 0.599
Land and assets variables 0.283 0.285 0.225 0.469 0.522 0.435

 

Claimants include head of household and children or siblings who were 16 or over in 1997. The table
shows the marginal eﬂ‘ects of a change in explanatory variables. Standard errors are corrected for
clustering at the individual level and heteroskedasticity. Absolute value of z statistics are in brackets

with statistical signiﬁcance at 10% (*); 5% (**); and 1% (***) indicated.

164

Appendix Table 2.6.7

Determinant of Household Division between 1993 and 1997, Claimant 2: Urban

 

 

 

1993 Variables: (1) (2) (3) (4) (5)
# Claimants 0.039 0.039 0.039 0.039 0.009
[9.138] *“ [9.142] "* [9.167] *" [9.328] 2" [8.901] ***
Proportion of claimants:
20-29 years, male 0.205 0.204 0.204 0.215 0.048
[2.955] *" [2.936] *** [2.940] **"‘ [3.128] *** [3.217] ***
30-49 years, male 0.249 0.249 0.249 0.254 0.06
[3.683] **“' [3.664] *** [3.678] *" [3.790] *" [4.065] *"
50 years or older, male 0.290 0.290 0.283 0.291 0.071
[3.139] *** [3.137] *** [3.089] *** [3.132] *** [3.376] **“'
15-19 years, female 0.005 0.005 0.005 0.014 0.01
[0.082] [0.074] [0.081] [0.224] [0.736]
20—29 years, female 0.034 0.033 0.031 0.026 0.014
[0.506] [0.479] [0.452] [0.391] [0.970]
30-49 years, female 0.080 0.076 0.072 0.055 0.013
[1.046] [0.997] [0.958] [0.762] [0.835]
50 years or older, female 0.170 0.161 0.160 0.137 0.035
[2.376] "”" [2.227] ** [2.226] "”" [1.915] "' [2.211] *2
Age of head -0.001 -0.001 -0.001 -0.001 0
[0.361] [0.366] [0.392] [0.306] [0.482]
Age of head squared x 10'3 0.003 0.004 0.005 0.002 -0.006
[0.098] [0.101] [0.139] [0.055] [0.797]
Head of hh's schooling -0.003 -0.003 -0.001 -0.001 0
[1.876] "' [1.732] * [0.404] [0.424] [0.039]
Max. schooling, non-head claimants 0.005 0.005 0.004 0.004 0.001
[2.625] “* [2.105] ** [1.882] " [1.979] " [2.031] **
Sd. dev. of claimants' schooling 0.002 0.003 0.003 0
[0.584] [0.678] [0.778] [0.263]
# Boys 0-11 0.012 0.012 0.012 0.012 0.003
[1.666] * [1.659] "' [1.676] * [1.656] "' [1.588]
# Girls 0-11 0.013 0.012 0.012 0.013 0.003
[1.695] * [1.646] * [1.612] [1.769] "‘ [1.470]
# Boys 12-14 0.010 0.010 0.010 0.01 0.003
[0.788] [0.809] [0.759] [0.830] [1 . 184]
# Girls 12-14 0.047 0.047 0.047 0.047 0.011
[3.515] **"' [3.515] *" [3.565] *** [3.642] *** [3.677] ***
Male head (=1) -0.124 -0.125 -0.123 -0.136 -0.042
[4.363] "'** [4.427] "* [4.356] "'*"' [4.774] *** [4.624] *"
log(land owned) (Rp x 103) 0.395 0.362 -0.433 -0.437 -0.198
[0.318] [0.292] [0.257] [0.264] [0.513]
log(bus. assets) (Rp x 103) 0.166 0.155 0.921 0.887 0.38
[0.197] [0.183] [0.876] [0.818] [1 .595]
(continued)

165

(continued)

 

 

 

 

Interactions
log(land owned) (Rp x 103) 0.339 0.297 0.097
x household head's schooling
[0.988] [0.866] [1.194]
x max. of non-head claimants' schooling 0.115 0.131 0.03
[0.290] [0.326] [0.335]
log(bus. assets) (Rp x10'3) 0.104 0.106 -0014
x household head's schooling
[0.487] [0.501] [0.288]
x max. of non-head claimants' schooling -0-397 -0.406 -0.103
[1.4921 [1.545] [1.766]
Observations 2,862 2,862 2,862 2,862 2,862
Province dummies No No No Yes No
Community dummies No No No No Yes
F-test (p-values)
Head's education 0.262 0.310 0.054
Non-head claimants' max. educ. 0.116 0.108 0.476
Education variables 0.025 0.042 0092 0.090 0.084
Land variables 0.470 0.530 0.361
Other assets variables 0.517 0.486 0.205
Land and assets variables 0.908 0922 0505 0.547 0.331

 

Claimants include head of household and any member who were 15 or over in 1993. The table shows
the marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at
the 1993 community level and heteroskedasticity. Absolute value of z statistics are in brackets with
statistical signiﬁcance at 10% (*); 5% (**); and 1% (***) indicated.

166

 

 

Appendix Table 2.6.8

Determinant of Household Division between 1993 and 1997, Claimant 2: Rural

1993 Variables:

(1) (2) (3) (4) (5) (6)

 

# Claimants

Proportion of claimants:
20-29 years, male

30-49 years, male
50 years or older, male
15-19 years, female
20-29 years, female
30-49 years, female
50 years or older, female
Age of head
Age of head squared x 10'3

Head of hh's schooling

0.049 0.049 0.049 0.049 0.05 0.031
[8.332] *** [8.295] *** [8.195] “* [8.168] "* [8.177] *" [8.203] ‘**

0.182 0.180 0.181 0.179 0.176 0.098
[3.343] m [3.301] m [3.326] m [3.283] m [3.271] m [3.052] m
0.214 0.210 0.210 0.209 0.208 0.12
[3.522] m [3.430] m [3.426] m [3.406] m [3.394] m [3.127] m
0.130 0.126 0.126 0.125 0.124 0.079
[1.791] * [1.726] * [1.725] 3 [1.698] * [1.690] r [1.707] r
0.048 0.046 0.042 0.041 0.041 0.046
[0.760] [0.730] [0.669] [0.655] [0.669] [1.344]
0.134 0.132 0.129 0.127 0.127 0.085
[2111]" [2.090] H [2.038] H [1.988]“ [2.049] H [2.387]"
0.177 0.173 0.171 0.169 0.173 0.104
[2.716] m [2.651] m [2.603] m [2.579] m [2.689] m [2.837] m
0.170 0.157 0.154 0.154 0.164 0.109
[3.041] m [2.800] m [2.735] m [2.737] m [2.913] m [3.278] m
0.000 0.000 0.000 0.000 0.000 -0001
[0.045] [0.023] [0.035] [0.034] [0.111] [0.317]
0.014 0.013 0.014 0.014 0.015 0.012
[0.449] [0.429] [0.458] [0.454] [0.504] [0.628]
-0001 -0001 -0001 -0001 -0001 0
[0.517] [0.472] [0.445] [0.549] [0.727] [0.318]

Max. schooling, non-head claimants 0.003 0.001 0.001 0.001 0.001 0.004

Sd. dev. of claimants' schooling

# Boys 0-11

# Girls 0-11

# Boys 12-14

# Girls 12-14

Male head (=1)

log(land owned) (Rp x 10'3)
log(bus. assets) (Rp x 10'3)

Farm household (=1)

[1.376] [0.629] [0.565] [0.420] [0.531] [1.869] r
0.005 0.005 0.004 0.005 -0.896
[1.712] * [1.726] * [1.322] [1.378] [0.901]
-0003 -0004 -0004 -0004 -0002 0.001
[0.543] [0.554] [0.590] [0.622] [0.367] [0.149]
0.010 0.010 0.009 0.010 0.01 0.009
[1.534] [1.496] [1.427] [1.437] [1.572] [2.321] H
0.041 0.041 0.041 0.041 0.042 0.027
[3.312] m [3.279] m [3.304] m [3.292] m [3.326] m [3.436] m
0.094 0.095 0.095 0.094 0.094 0.059
[7.627] m [7.649] m [7.655] m [7.681] m [7.658] m [7.788] m
-0049 -0049 -0051 -0051 -0.056 -0.03
[1896]" [1.874]* [1966]" [1965]" [2.183]" [1.775]*
1.656 1.601 8.070 7.744 8.487 0.001
[1.463] [1.415] [2.214] H [2.123] H [2.418] H [0.855]
-0.128 -0243 -1.846 -1910 -1.728 4.611
[0.098] [0.187] [1.214] [1.227] [1.132] [2.183] H
-0020 -0019 -0.046 -0045 -005 -0027
[1.007] [0.938] [1.807] r [1.807] * [1.961] [1.693] *

 

(continued)

167

(continued)

 

 

 

 

Interactions
Farm hh (=1) ong(1and owned) (Rp x 103) -7.375 -6.986 ~7.744 -4.052
[1.996] ** [1.888] * [2.140] ** [1.838] *

Farmhh(=1) x log(bus. assets) (Rp x10'3) 3.687 3.408 2.983 0.706
[1.145] [1 .069] [0.924] [0.358]
log(land owned) (Rp x 10‘3) x head's schooling -0.032 -0.063 0.045
[0.115] [0.231] [0.270]
x max. of non-head claimants' schooling 0.358 0.396 0.218
[1.215] [1.382] [1.176]
log(bus. assets) (Rp x 10'3) x head's schooling -0.048 -0.035 -0. 171
[0.160] [0.121] [0.975]
x max. of non-head claimants' schooling -0. 192 -0.193 -0.175
[0.624] [0.627] [0.888]
Observations 3,365 3,365 3,365 3,365 3,365 3,365
Province dummies No No No No Yes No
Community dummies No No No No No Yes

F -test [Q-valuesz
Head's education 0.909 0.822 0.629
Non-head claimants' max. educ. 0.592 0.473 0.582
Education variables 0.382 0.199 0.212 0.507 0.387 0.088
Land variables 0.137 0.061 0.073
Other assets variables 0.724 0.768 0.459
Land and assets variables 0.340 0.367 0.167 0.416 0.263 0.336

 

Claimants include head of household and any member who were 15 or over in 1993. The table shows
the marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at

the community level and heteroskedasticity. Absolute value of z statistics are in brackets with
statistical signiﬁcance at 10% (*); 5% (**); and 1% (***) indicated.

168

Appendix Table 2.6.9
Determinant of Household Division between 1993 and 2000, Claimant 2: Urban

 

 

 

 

1993 Variables: (1) (2) (3) (4) (5)
# Claimants 0.103 0.103 0.103 0.104 0.113
[6.727] *** [6.730] *" [6.701] "* [6.831] *" [6.343] ***
Proportion of claima__n_t§;
15-19 years, male -0.526 -0.521 -0.530 -0.514 -0.617
[2.109] " [2.089] *"' [2.128] “ [2.071] "”" [2.234] "
20-29 years, male -0.393 -0.390 -0.399 -0.366 -0.469
[1.626] [1.618] [1.660] [1.526] [1.744]
30-49 years, male -0.668 -0.664 -0.671 -0.63 -0.729
[2.955] **"' [2.945] *" [2.984] *** [2.787] "* [2.920] "*
50 years or older, male -0.753 -0.747 -0.762 -0.699 -0.851
[3.216] *** [3.196] *** [3.256] “* [2.976] *” [3.291] "*
12-14 years, female -0.074 -0.074 -0.079 -0.071 -0.059
[0.725] [0.727] [0.777] [0.700] [0.518]
15-19 years, female -0.517 -0.512 -0.518 -0.484 -0.551
[2.221] ** [2.199] " [2.223] '“' [2.092] " [2.150] *"‘
20-29 years, female -0.541 -0.539 -0.545 -0.513 -0.61
[2.224] "”" [2.217] ** [2.250] " [2.128] *"‘ [2.287] **
30-49 years, female -0.702 -0.704 -0.713 -0.686 -0.846
[2.820] "W [2.837] *** [2.884] *” [2.772] 2“ [3.091] *"
50 years or older, female -0.540 -0.553 -0.558 -0.552 -0.649
[2.232] *"' [2.299] " [2.328] ** [2.294] " [2.428] **
# Non-claimant members 12+ 0.104 0.103 0.104 0.097 0.111
[1.689] " [1.678] "‘ [1.697] * [1.597] [1.621]
Age of head 0.017 0.017 0.017 0.017 0.02
[2.466] " [2.462] " [2.451] *"' [2.420] " [2.448] *"‘
Age of head squared x 10'3 -0.146 -0145 -0145 -0142 -0.158
[2.216] ** [2.214] ** [2.201] " [2.144] *"' [2.071] "
Head of hh's schooling -0.004 -0.004 -0.001 -0.001 0.003
[1 .402] [1.324] [0.203] [0.178] [0.720]
Max. schooling, non-head claimants 0.012 0.011 0.010 0.01 0.014
[3.417] **"‘ [3.071] "* [2.792] *""" [2.926] *“ [3.538] ""'
Sd. dev. of claimants' schooling 0.005 0.005 0.005 0.004
[0.781] [0.770] [0.773] [0.530]
# Boys 0-5 0.024 0.024 0.023 0.025 0.025
[1.330] [1.326] [1.295] [1.369] [1.285]
# Girls 0-5 0012 -0.013 -0.013 -0.016 -0.015
[0.659] [0.690] [0.719] [0.823] [0.739]
# Boys 6-11 0.055 0.055 0.055 0.055 0.06
[3.611] **"' [3.622] *“ [3.628] *" [3.482] *” [3.430] "’"
# Girls 6-11 0.031 0.031 0.031 0.033 0.039
[1.869] * [1.832] "' [1.843] * [1.935] "' [2.153] **
Male head (=1) -0.019 -0.019 -0.016 -0.028 -0.043
[0.363] [0.364] [0.305] [0.536] [0.717]
(continued)

169

(continued)

 

log(land owned) (Rp x 101‘) 0.437 0.351 0.444 0.957 -0.657
[0.185] [0.149] [0.157] [0.343] [0.211]
log(bus. assets) (Rp x 103) 4.157 4.148 4.597 4.47 5.734
[2.349] ** [2.338] "”" [2.248] ** [2.165] " [2.423] **
Interactions
log(land owned) (Rp x 10'3)
x household head's schooling 0.609 0.54 0.805
[1.071] [0.925] [1.305]
x max. of non-head claimants' schooling -0.252 -0.2 -0.226

[0.375] [0.300] [0.308]
log(bus. assets) (Rp x 10'3) -

 

 

 

x household head's schooling 0.140 0.107 0.015
[0.381] [0.290] [0.037]
x max. of non-head claimants' schooling -0.401 -0.352 -0.07
[0.847] [0.761] [0.145]
Observations 2,913 2,913 2,913 2,913 2,913
Province dummies No No No Yes No
Community dummies No No No No Yes
F-test [Q-valuesz
Head's education 0.353 0.476 0.558
Non-head claimants' max. educ. 0.010 0.012 0.004
Education variables 0.003 0.007 0.031 0.038 0.007
Land variables 0.723 0.754 0.599
Other assets variables 0.112 0.142 0.048
Land and assets variables 0.027 0.029 0.107 0.128 0.058

 

Claimants include head of household and any member who were 12 or over in 1993. The table shows the
marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at the
individual level and heteroskedasticity. Absolute value of z statistics are in brackets with statistical
signiﬁcance at 10% (*); 5% (**); and 1% (*”) indicated.

170

1993 Variables:

Appendix Table 2.6.10
Determinant of Household Division between 1993 and 2000, Claimant 2: Rural

(1)

(2)

(3)

(4)

(5)

(6)

 

# Claimants

Proportion of claimants:

0.145

0.145

0.144

0.144

0.150

0.161

[10.703] [10.705] [10.512] [10.551] [1 1.248] [10.653]

 

15-19 years, male -0.684 -0.678 -0.679 -0.671 -0.692 -0.781
[3.022] 'H [2.983]... [2.967] 'H [2.904] H'- [2.961]H* [3.128]"H
20-29 years, male -0.601 -0.594 -0.594 -0.583 -0.587 -0.687
[2.631]H* [2.591]... [2.576] Ht [2.515] H [2.494] H [2.707] H"
30-49 years, male -0.785 -0.780 -0.780 -0.774 -0.771 -0.861
[3.404] ‘H [3.372]." [3.351]:H [3.301]". [3.248] H' [3.348] *H
50 years or older, male -0.918 -0.913 -0.919 -0.906 -0.900 -1.033
[3.763] H. [3.732] 3H [3.723]H' [3.657] ‘H [3.593] H' [3.809] 3H
12-14 years, female 0.034 0.035 0.036 0.029 0.027 0.056
[0.407] [0.424] [0.430] [0.346] [0.322] [0.632]
15-19 years, female -0.776 -0.770 -0.774 -O.771 -0.787 -0.866
[3.291]*H [3.266] *H [3.256]H* [3.218] 9H [3.262] H. [3.332] *H
20-29 years, female —0.847 -0.841 -0.844 -0.842 -0.839 -0.973
[3.807] H‘ [3.781] ‘H [3.759] ‘H [3.726] ‘H [3.673] ‘H [3.940] H"
30-49 years, female -0.795 -0.790 -0.794 -0.793 -0.772 -0.909
[3.709]H* [3.684] H‘ [3.669] ‘H [3.634] H* [3.480] *H [3.747]*H
50 years or older, female -0.986 -0.989 -0.993 -0.993 -0.960 -1.067
[4.418] *H [4.437] 3H [4.415] 9H [4.395] Hr [4.190]*H [4.311]‘H
# Non-claimant members 12+ 0.171 0.169 0.169 0.167 0.171 0.191
[2.702] 'H [2.663] *H [2.632] ‘H [2.594] Ht [2.618]-Ht [2.750] H'-
Age of head 0.009 0.009 0.009 0.009 0.008 0.009
[1.748] t [1.762] a [1.750] a [1.665] * [1.611] [1.727] 8
Age of head squaredx 10'3 -0.048 -0.048 -0.047 -0.043 -0.043 -0.042
[1.002] [1.013] [0.978] [0.891] [0.883] [0.853]
Head of hh's schooling -0.003 -0.003 -0.003 -0.004 -0.005 -0.002
[1.070] [1.005] [1.004] [1 .424] [1.648] 3 [0.765]
Max. schooling, non-head claimants 0.008 0.007 0.007 0.007 0.009 0.010
[2.580]H' [1.997] H [1.982] H [1.735]. [2.170] H [2.285] H
Sd. dev. of claimants' schooling 0.004 0.004 0.004 0.004 0.008
[0.623] [0.591] [0.549] [0.607] [1.078]
# Boys 0-5 0019 -0.019 -0.018 -0.018 -0.018 -0.005
[1.141] [1.136] [1.122] [1.069] [1.116] [0.253]
# Girls 0-5 0.012 0.012 0.012 0.012 0.013 0.031
[0.785] [0.768] [0.764] [0.776] [0.833] [1.744] a
# Boys 6-11 0.075 0.075 0.075 0.074 0.076 0.084
[4.806] H— [4.791]'H [4.790] *H [4.737] 9H [4.675]*H [4.853]*H
# Girls 6-11 0.080 0.080 0.080 0.081 0.082 0.095
[4.458] 9H [4.467] ‘H [4.465] *H [4.545] Ht [4.664] 'H [4.944] *H
Male head (=1) -0.119 -0.119 -0.121 -0.119 -0.132 -0.148
[2.636] *H [2.629] H' [2.676] *H [2.641] *H [3.039] Ht [3.137]*H
(continued

171

 

 

(continued

 

 

 

 

log(land owned) (Rp x 103) -0.296 -0.335 2.270 2.188 3.869 2.675
[0.163] [0.184] [0.360] [0.343] [0.625] [0.408]
log(bus. assets) (Rp x 103) 2.745 2.710 0.193 1.558 1.547 2.837
[1.1 1 1] [1.099] [0.063] [0.497] [0.474] [0.829]
Farm household (=1) 0.000 0.001 -0.027 -0.019 -0.024 -0.051
[0.008] [0.023] [0.727] [0.522] [0.626] [1.223]
Interactions

Farm hh (=1) x log(land owned) (Rp x 103) -3.369 -3.098 -5.251 -2.372
[0.499] [0.461] [0.797] [0.340]
Farm hh (=1) x log(bus. assets) (Rp x 103) 8.523 8.088 7.846 4.491
[1.714] a [1.595] [1.557] [0.806]
log(land owned) (Rp x 10'3) x head's schooling 0.281 0.134 0.286
[0.526] [0.254] [0.506]
x max. of non-head claimants' schooling -0.227 -0.178 -0.055
[0.372] [0.294] [0.082]
log(bus. assets) (Rp x 10'3) x head's schooling 0.647 0.736 0.340
[1.070] [1.225] [0.531]
x max. of non-head claimants' schooling 0506 0.554 0.333
[0.771] [0.840] [0.471]
Observations 3,417 3,417 3,417 3,417 3,417 3,417
Province dummies No No No No Yes No
Community dummies No No No No No Yes

F-test [Q-valuesz
Head's education 0.133 0.157 0.580
Non-head claimants' max. educ. 0.153 0.058 0.049
Education variables 0.036 0.076 0.085 0.063 0.026 0.027
Land variables 0.813 0.960 0.919 0.974
Other assets variables 0.100 0.119 0.092 0.521
Land and assets variables 0.538 0.546 0.317 0.107 0.108 0.476

 

Claimants include head of household and any member who were 12 or over in 1993. The table shows the
marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at the
individual level and heteroskedasticity. Absolute value of z statistics are in brackets with statistical

signiﬁcance at 10% (*); 5% (**); and 1% (***) indicated.

172

Appendix Table 2.6.11

Determinant of Household Division between 1997 and 2000, Claimant 2: Urban

1997 Variables:

(1)

(2)

(3)

(4)

(5)

 

# Claimants

Proportion of claimants:

0.053

0.053

0.053

0.054

0.042

[7.412] m [7.404] m [7.459] m [7.477] m [8.690] m

 

20-29 years, male 0.056 0.057 0.057 0.06 0.038
[0.772] [0.787] [0.772] [0.832] [0.794]
30-49 years, male -0.187 -0.187 -0.l91 -0.184 40.113
[2.312] ** [2.316] ** [2.358] ** [2.292] " [2.054] "
50 years or older, male -0.256 -0.257 -0.264 -0.253 -0. 165
[2.518] ** [2.530] "'* [2.590] ** [2.493] ** [2.404] **
15-19 years, female 0.091 0.091 0.091 0.101 0.062
[1.169] [1.168] [1.165] [1.288] [1.196]
20-29 years, female 0.064 0.063 0.063 0.062 0.025
[0.810] [0.802] [0.806] [0.792] [0.483]
30-49 years, female -0.089 -0.092 -0.098 -0.102 -0.051
[1.089] [1.125] [1.196] [1.230] [0.891]
50 years or older, female -0.094 -0.103 -0.109 -0.115 -0.057
[1.143] [1.233] [1.307] [1.356] [1.035]
Age of head 0.016 0.016 0.016 0.016 0.01
[3.615] *” [3.622] *** [3.704] *" [3.750] "* [3.462] ***
Age of head squared x 10'3 -01 16 -0117 -0119 -012 -0.078
[2.846] 2" [2.851] "* [2.910] *** [2.954] "* [2.846] "*
Head of hh's schooling -0.002 -0.002 -0.003 -0.003 -0.001
[1.187] [1.030] [1.189] [1.147] [0.454]
Max. schooling, non-head claimants 0.006 0.005 0.005 0.005 0.004
[2.119] " [1.957] * [1.875] * [1.946] " [2.370] "
Sd. dev. of claimants' schooling 0.003 0.004 0.004 0.002
[0.594] [0.907] [0.914] [0.552]
# Boys 0-11 0.023 0.023 0.022 0.024 0.015
[2.361] ** [2.353] *"' [2.319] " [2.416] ** [2.202] *"
# Girls 0-11 0009 -0.009 -0.009 -0.009 -0.009
[0.804] [0.808] [0.783] [0.791] [1.111]
# Boys 12-15 0.047 0.047 0.048 0.051 0.032
[2.549] ** [2.550] ** [2.605] " [2.779] "* [2.630] ***
# Girls 12-15 0.088 0.088 0.088 0.091 0.063
[5.698] "* [5.683] *" [5.684] *** [5.870] *** [6.139] “*
Male head (=1) 0021 -0.021 -0.021 -0.024 -0.01
[0.695] [0.715] [0.706] [0.807] [0.496]
log(land owned) (Rp x 10") 1.273 1.253 -0009 -0312 -0303
[0.781] [0.770] [0.005] [0.161] [0.225]
log(bus. assets) (Rp x 103) 2.974 2.943 4.290 4.152 2.519
[2.328] ** [2.296] ** [2.905] *" [2.811] **"' [2.578] "'"
(continued)

173

(continued)

 

Interactions
log(land owned) (Rp x 10'3)

 

 

 

x household head's schooling -0.302 -0.274 -0.124
[0.727] [0.658] [0.464]
x max. of non-head claimants' schooling 0.515 0.45 0.316
[1.1 11] [0.968] [0.991]

log(bus. assets) (Rp x 10'3)
x household head's schooling -0.235 -0.229 -0.182
[0.772] [0.747] [0.916]
x max. of non-head claimants' schooling '0-3 83 -0.349 -0.267
[1.124] [1.027] [1.191]
Observations 3,047 3,047 3,047 3,047 3,047
Province dummies No No No Yes No
Community dummies No No No No Yes

F-test -values

Head's education 0.403 0.454 0.585
Non-head claimants' max. educ. 0.228 0.228 0.092
Education variables 0.105 0.210 0.252 0.277 0.106
Land variables 0.639 0.771 0.765
Other assets variables 0.035 0.046 0.053
Land and assets variables 0.013 0.015 0.060 0.108 0.130

 

Claimants include head of household and any member who were 16 or over in 1997. The table shows the
marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at the

individual level and heteroskedasticity. Absolute value of z statistics are in brackets with statistical
signiﬁcance at 10% (*); 5% (**); and 1% (***) indicated.

174

 

1997 Variables:

Appendix Table 2.6.12
Determinant of Household Division between 1997 and 2000, Claimant 2: Urban

(1)

(2)

(3)

(4)

(5)

(6)

 

# Claimants

Proportion of claimants:
20-29 years, male

30-49 years, male
50 years or older, male
15-19 years, female
20-29 years, female
30-49 years, female
50 years or older, female
Age of head
Age of head squared x 10'3

Head of hh's schooling

0.075

0.075

0.075

0.075

0.075

0.06

[9.619] *** [9.641] *" [9.621] "* [9.703] *** [9.911] **"' [9.044] ***

0.028
[0.382]
-0.185
[2.243] H
-0175
[1.881] *
0.163
[2.037] H
-0.06
[0.780]
-0037
[0.468]
-0199
[2.400] H
0.009
[2.127] H
-0.058
[1.520]
-0002
[1.142]

Max. schooling, non-head claimants 0.004

Sd. dev. of claimants' schooling

# Boys 0-11
# Girls 0-11
# Boys 12-15

# Girls 12-15

[1.843] *

0.016
[1.557]
0.011
[1.158]
0.093

0.028
[0.380]
-0.185
[2.240] H
-0175
[1 .879] *
0.162
[2.030] H
-0.06
[0.784]
-0.038
[0.480]
-0205
[2.442] H
0.009
[2.133] H
-0.058
[1.523]
-0002
[1.072]
0.003
[1.450]
0.002
[0.541]
0.016
[1.549]
0.01 1
[1.158]
0.093

0.028 0.025 0.024 -0.035
[0.378] [0.344] [0.340] [0.668]
.311: .. .318... .31.: .. 3.1.7.:
1.1317851" [181301] * 1.39101] * [2:3 910:
2.1:: .. .112: .. .128... .3167:
[0772] [0?72311 1:830] [3819381 *
1.233.? [$.21] [3.13.8] 1198:":
13.2.21" [39.1%] .. 3212?" [2.1.6.1 ..
.289. .. [£1229] .. [23139146331
[1'32 “2’23? [(3336] [02232]
[190122] [097522] [£6221] 1:63:11
[3.3%] [£323] [33?] [372%].
[£322] [£333] [3.332] [3.3.3.1]
[3521? [35313 [333‘s]. [1.31%]
[1.18] [1.3133 [3522'] [3.323
0.093 0.093 0.097 0.074

[6.371] *** [6.350] "* [6.352] **"‘ [6.380] "* [6.569] “* [6.931] *"

0.122

0.122

0.122

0.122

0.124

0.096

[9.746] "* [9.729] "* [9.700] *"”" [9.669] *** [9.670] "* [9.936] ***

 

Male head (=1) -0059 -0059 -0059 -0.062 -0.068 -0.058
[1.676] * [1.680] * [1.674] * [1.784] * [1.905] 3 [2.055] H
log(land owned) (Rp x 10") 2.836 2.811 3.531 3.477 3.2 3.012
[1.912] r [1.889] r [1.115] [1.092] [0.984] [1.272]
log(bus. assets) (Rp x10'3) -0057 -005 0.128 0.262 0.379 0.394
[0.036] [0.032] [0.074] [0.151] [0.214] [0.285]
Farm household (=1) -0005 -0005 -0002 0 0.003 0.005
[0.197] [0.205] [0.086] [0.005] [0.117] [0.233]
(continued)

175

(continued)

 

 

 

 

Interactions

Farm hh (=1) x log(land owned) (Rp x 103) -0.817 -1.249 -0.853 -2.092
[0.224] [0.344] [0.230] [0.771]
Farmhh (=1) x log(bus. assets) (Rp x10'3) -1.149 -0797 -1.452 -2.77
[0.322] [0.226] [0.414] [0.993]
log(land owned) (Rp x 10‘3) x head's schooling -0.215 -0.236 -0.209
[0.660] [0.734] [0.854]
x max. of non-head claimants' schooling -0.098 -0.088 0.001
[0.304] [0.270] [0.004]
log(bus. assets) (Rp x 10") x head's schooling 0.008 -0027 -0.189
[0.021] [0.072] [0.657]
x max. of non-head claimants' schooling 0531 0.562 0.398
[1.560] [1.638] [1.501]
Observations 3,617 3,617 3,617 3,617 3,617 3,617
Province dummies No No No No Yes No
Community dummies No No No No No Yes

F -test [Q—valuesz
Head's education 0.346 0.598 0.030
Non-head claimants' max. educ. 0.674 0.116 0.393
Education variables 0.164 0.289 0.276 0.191 0.249 0.088
Land variables 0.148 0.361 0.326 0.420
Other assets variables 0.948 0.577 0.542 0.553
Land and assets variables 0.149 0.156 0.408 0.470 0.395 0.247

 

Claimants include head of household and any member who were 16 or over in 1997. The table shows

the marginal effects of a change in explanatory variables. Standard errors are corrected for clustering at
the individual level and heteroskedasticity. Absolute value of 2 statistics are in brackets with statistical
signiﬁcance at 10% C“); 5% ("); and 1% ("*) indicated.

176

Appendix Table 2.6.13
Determinant of Household Division between 1993 and 1997, Claimants are
Household Heads, Their Sons, and Brothers Age 15 or Above in 1993
1993 Variables: (l) (2) (3) (4) (5)
# Claimants 0.082 0.076 0.076 0.076 0.076
[6.134] *H [5.488] H* [5.474] *H [5.474] *H [5.379] *H

 

Proportion of claimants:

20-29 years 0.258 0.255 0.254 0.254 0.266
[5.969] *** [5.930] **"' [5.914] **"‘ [5.915] *** [6.190] **"‘
30-49 years 0.375 0.382 0.382 0.379 0.393
[6.236] "* [6.411] *" [6.409] *** [6.366] ‘** [6.617] "*
50 years or older 0.392 0.4 0.401 0.401 0.387

[4.611] *" [4.754] *" [4.757] "* [4.740] "" [4.598] "*
# Non-claimant members 15 and over 0.027 0.027 0.027 0.027 0.028
[39781 13.6761 13.6701 [3.614] [3.661]

Age of head -0.013 -0.014 -0.014 -0.014 -0.013
[2.641] *" [2.731] *""" [2.716] "* [2.767] *" [2.689] **"'
Age ofhead squaredx 10'3 0.113 0.117 0.116 0.118 0.11
[2.530] *"' [2.626] *" [2.618] """* [2.664] [2.558] *"'
Head of hh's educ. -0.002 -0.008 -0.008 -0.008 -0.008
[0.837] [2.428] " [2.428] *"' [2.308] *" [2.526] *"'
Max. schooling, non-head claimants 0.003 0.010 0.010 0.011 0.011
[1.189] [2.832] "'" [2.822] "'** [2.909] [3.065]
Sd. dev. of claimants' schooling -0.016 -0.016 -0.016 -0.015
[2.614] *** [2.613] *** [2.548] [2.516] **
# Boys O-ll 0.011 0.011 0.011 0.011 0.012
[1.090] [1.057] [1.052] [1.087] [1.141]
# Girls 0-11 0.024 0.023 0.023 0.023 0.024
[2.369] *" [2.326] ** [2.332] *"' [2.358] "”" [2.487] '"'
# Boys 12-14 -0.02 -0.021 -0.021 -0.021 -0.023
[1.134] [1.212] [1.205] [1.175] [1.382]
# Girls 12-14 0.087 0.086 0.086 0.085 0.081
[5.056] *** [4.989] *** [4.994] *""" [4.947] *" [4.774] "*
Male head (=1) -0.022 -0.025 -0.025 -0.024 -0.03
[0.838] [0.955] [0.979] [0.915] [1.159]
log(land owned) (Rp x 10") 2.122 2.201 2.69 2.55 2.677
[1.484] [1.542] [1.380] [1.360] [1.467]
log(bus. assets) (Rp x 103) -2.47 -2.542 -2.428 -1.969 -2.595
[1.650] [1.703] [0.926] [0.746] [1.001]
Urban (=1) -0.041 -0.041 -0.041 -0.042 -0.026

[1.989] [1.982] [1.968] [2.017] [0.374]

Interactions

 

Urban x log (land owned) x 10'3 -l.533 -0.573 -0.832
[0.514] [0.198] [0.293]
Urban x log(bus. assets) x 10'3 -0.092 -l.l 16 -l.087
[0.029] [0.336] [0.336]

(continued)

177

(continued)

 

log(land owned) (Rp x 10")

 

 

 

x household head's schooling 0.112 -0.009
[0.280] [0.021]
x max. of non-head claimants' schooling -0.412 -0.347
[0.922] [0.821]

log(bus. assets) (Rp x 10")
x household head's schooling 0.144 0.143
[0.357] [0.356]
x max. of non-head claimants' schooling 0.191 0.196
[0.386] [0.412]
Observations 2395 2395 2395 2395 2395
Province, urban, interaction dummies No No No No Yes

F -test [Q-valuesz

Head's education 0.095 0.068
Non-head claimants' max. educ. 0.027 0.018
Education variables 0.4474 0.029 0.029 0.175 0.139
Land variables 0.341 0.554 0.508
Other assets variables 0.247 0.481 0.279
Land and assets variables 0.1889 0.167 0.422 0.747 0.586

 

The table shows the marginal effects of a change in explanatory variables. Standard errors are
corrected for clustering at the 1993 conununity level and heteroskedasticity. Absolute value of 2

statistics are in brackets with statistical signiﬁcance at 10% (*); 5% (**); and 1% (**"') indicated.

178

 

Appendix Table 2.6.14

Determinant of Household Division between 1993 and 2000, Claimants are
Household Heads, Their Sons, and Brothers Aged 12 or Above in 1993

1993 Variables:

(1) (2) (3) (4) (5)

 

# Claimants

Proportion of claimants:
15-19 years

20-29 years

30-49 years

50 years or older

# Non-claimant members 12 and over 0.084

Age of head

Age of head squared x 10‘3

Head of hh's educ.

Max. schooling, non-head claimants
Sd. dev. of claimants' schooling

# Boys 0-5

# Girls 0-5

# Boys 6-11

# Girls 6-11

0.118 0.116 0.116 0.117 0.122
[7.085] *H [6.899] *H [6.900] *H [6.910] *H [7.109] *H

-0073 -0.078 -0.078 -0.078 .0093
[1.146] [1.216] [1.217] [1.220] [1.438]
0.155 0.15 0.15 0.148 0.156
[2.268] H [2.186] H [2.188] H [2.145] H [2.190] H
-007 -0.072 -0072 -0079 -0.06
[0.798] [0.825] [0.829] [0.899] [0.670]
-0.037 -004 -004 -0039 -0034
[0.350] [0.381] [0.386] [0.372] [0.309]
0.084 0.084 0.084 0.087
[7.824] *H [7.810] *H [7.791] *H [7.771] *H [8.043] *H
-0003 -0003 -0003 -0004 0004
[0.469] [0.468] [0.471] [0.603] [0.656]
0.015 0.016 0.016 0.022 0.026
[0.277] [0.290] [0.293] [0.398] [0.456]
0.001 0.0001 0.0001 0.0001 0.0001
[0.391] [0.003] [0.001] [0.049] [0.084]
0.004 0.006 0.006 0.007 0.007
[1.220] [1.549] [1.557] [1.805] r [1.700] *
-0005 -0005 -0004 -0002
[0.872] [0.875] [0.714] [0.331]
0.014 0.015 0.015 0.016 0.017
[0.799] [0.817] [0.815] [0.894] [0.940]
0.028 0.029 0.029 0.031 0.026
[1.511] [1.527] [1.525] [1.638] [1.374]
0.055 0.054 0.054 0.056 0.05

[3.413] **"' [3.381] ""'"" [3.390] "" [3.486] *" [3.017] “*
0.046 0.045 0.046 0.045 0.041
[2.737] *** [2.715] *" [2.717] *"'* [2.685] "* [2.475] *"

 

Male head (=1) -0.034 -0.036 -0.035 -0.03 -0.035
[1.103] [1.148] [1.134] [0.965] [1.094]
log(land owned) (Rp x 10‘3) —l.026 -0.983 -1.349 -l.9 -l.889
[0.563] [0.540] [0.592] [0.806] [0.794]
log(bus. assets) (Rp x 10") 3.401 3.377 3.217 4.973 4.09
[l.719]"I [1.707]"' [1.015] [1.523] [1.181]
Urban (=1) -0.075 -0.074 -0.074 -0.074 -0. 12
[2.882] *** [2.853] ""[2807] "* [2.822] "”'”" [1.683]***
(continued)

179

 

(continued)

 

 

 

 

Interactions
Urban x log (land owned) x 10'3 1.265 4.135 3.488
[0.324] [1.007] [0.849]
Urban x log(bus. assets) x10'3 0.173 -3335 -3231
[0.043] [0.764] [0.723]
log(land owned) (Rp x 10")
x household head's schooling 0.039 -0.17
[0.074] [0.320]
x max. of non-head claimants' schooling -1.169 -0.957
[1.911]" [1.592]
log(bus. assets) (Rp x 10'3)
x household head's schooling 1.079 1.077
[2.596] """ [2.523] ""
x max. of non-head claimants' schooling 0.315 0.397
[0.536] [0.674]
Observations 2,904 2,904 2,904 2,904 2,904
Province, urban, interaction dummies No No No No Yes
F-test (Q-valuesz
Head's education 0.021 0.048
Non-head claimants' max. educ. 0.077 0.163
Education variables 0.299 0.366 0.361 0.041 0.064
Land variables 0.839 0.247 0.301
Other assets variables 0.245 0.077 0.033
Land and assets variables 0.220 0.223 0.531 0.036 0.119

 

The table shows the marginal effects of a change in explanatory variables. Standard errors are
corrected for clustering at the individual level and heteroskedasticity. Absolute value of z
statistics are in brackets with statistical signiﬁcance at 10% C"); 5% (""); and 1% (""")

indicated.

180

 

Appendix Table 2.6.15
Determinant of Household Division between 1997 and 2000, Claimants are Household
Heads, Their Sons, and Brothers Aged 16 or Above in 1997

1997 Variables:
# Claimants

Proportion of claimants:

(1)
0.070

(2)
0.069

(3)
0.068

(4)
0.067

(5)
0.068

l3-6921*** 13.6021 [3.5791H* [3.5121H* [3.oos1*H

 

181

20-29 years 0.150 0.151 0.149 0.149 0.161
[3.210] """ [3.216] """ [3.187] """ [3.165] """ [3.398] """
30-49 years 0.023 0.024 0.021 0.014 0.036
[0.317] [0.333] [0.296] [0.192] [0.496]
50 years or older -0.1 19 -0.118 -0.121 -0.133 -0.107
[1.180] [1.171] [1.193] [1.317] [1.036]
# Non-claimant members 16 and over 0.067 0.067 0.068 0.068 0.069
[5.863] """ [5.855] """ [5.946] """ [5.906] """ [5.906] """
Age of head 0.014 0.014 0.014 0.014 0.013
[2.054] "" [2.048] "" [2.040] "" [1.987] "" [1.878] "
Age of head squared x 10'3 -0110 -0110 -0109 -0.106 -0100
[1.774] " [1.764] " [1.751] " [1.686] " [1.578] "
Head of hh's educ. 0.002 0.001 0.001 0.001 0.002
[0.594] [0.202] [0.215] [0.354] [0.589]
Max. schooling, non-head claimants -0.0010 0.0000 -0.0010 0.0000 0.0000
[0.446] [0.103] [0.143] [0.008] [0.106]
Sd. dev. of claimants' schooling -0.003 -0.003 -0.002 -0.002
[0.424] [0.406] [0.241] [0.235]
# Boys 0-11 0.016 0.016 0.016 0.016 0.015
[1.207] [1.217] [1.181] [1.155] [1.140]
# Girls 0-11 0.021 0.021 0.022 0.022 0.021
[1.464] [1.457] [1.481] [1.532] [1.439]
# Boys 12-15 0.065 0.065 0.065 0.066 0.068
[2.973] """ [2.971] """ [2.974] """ [3.016] [3.101] """
# Girls 12-15 0.109 0.109 0.109 0.108 0.111
[5.137] """ [5.141] """ [5.123] """ [5.110] [5.144] """
Male head (=1) -0.07 -0.07 -0.073 -0.073 -0.073
[2.277] "" [2.292] "" [2.401] "" [2.396] "" [2.352] ""
log(land owned) (Rp x 103) 3.81 3.833 2.908 2.958 2.279
[2.314] "" [2.330] "" [1.512] [1.445] [1.051]
log(bus. assets) (Rp x10‘3) 2.624 2.614 5.318 4.371 4.618
[1.462] [1.455] [1.827] " [1.446] [1.491]
Urban (=1) -0.002 -0.002 0.001 0.001 -0.042
[0.080] [0.064] [0.027] [0.040] [0.496]
Interactions
Urban x log (land owned) x 10'3 1.176 2.129 1.814
[0.337] [0.570] [0.477]
Urban x log(bus. assets) x10'3 4.224 -1.908 -2419
[1.128] [0.473] [0.592]
(continued)

 

(continued)

 

log(land owned) (Rp x 10'3)
x household head's schooling

x max. of non-head claimants' schooling

log(bus. assets) (Rp x 10'3)
x household head's schooling

x max. of non-head claimants' schooling

Observations
Province, urban, interaction dummies
Community dummies
F-test [Q-valuesz
Head's education
Non-head claimants' max. educ.
Education variables
Land variables
Other assets variables
Land and assets variables

The table shows the marginal effects of a change in explanatory variables. Standard errors are

2,489 2,489
No No

0.582 0.472

0.164 0.155

2,489
No

0.421

0.160

0.259
[0.535]

-0441
[0.812]

-0.353
[0.770]
-0.53
[0.971]
2,489
No

No

0.584
0.464
0.276
0.244

0.339
[0.701]

-0515
[0.960]

-0.366
[0.796]
-0.436
[0.797]
2,489
No
Yes

0.821
0.402
0.549
0.486
0.196
0.092

corrected for clustering at the individual level and heteroskedasticity. Absolute value of 2 statistics
are in brackets with statistical signiﬁcance at 10% ("); 5% (""); and 1% (""") indicated.

 

182

 

Bibliography

Becker, Gary S. (1974), “A Theory of Social Interaction”, The Journal of Political
Economy 82:6, p. 1177-98.

Bergstorm, Theodore C. (1996), “A Survey of Theories of the Family”, in M.Rosenzweig
and O. Stark (eds). Handbook of Population and Family Economics, vol,
Amsterdam: North-Holland.

Cameron, Lisa and Deborah Cobb-Clark (2001 ), “Old-Age Support in Developing
Countries: Labor Supply, Intergenerational Transfers and Living Arrangements”
Working Paper No. 773, Department of Economics, University of Melbourne,
Australia

Cameron, Lisa (2000), “The Residency Decision of Elderly Indonesians: A Nested Legit
Analysis” Demography 37:1, p.17-27.

Chiappori, Pierre A. (1988), “Rational Household Labor Supply” Econometrica 56:1, p.
63-89.

Costa, D.L. (1997), “Displacing the Family: Union Army Pensions and Elderly Living
Arrangements”, The Journal of Political Economy 106:6, p.1269-92.

Doss, Cheryl. (1996 ), “Testing among Models of Intrahousehold Resource Allocation”
World Development 24:10, p. 1597-1609.

Deaton, Angus. (1997), Analysis of Household Surveys: A Microeconometric Approach
to Development Policy, Washington: The World Bank.

Foster, Andrew (1993), “Household Partition in Rural Bangladesh”, Population Studies,
47 , p.97-114

Foster, Andrew and Mark R. Rosenzweig (2002),”I—Iousehold Division and Rural
Economic Growth”, Review of Economic Studies, p.839-69.

Foster, Andrew and Mark R. Rosenzweig (1996). “Technical Change and Human Capital
Returns and Investment : Evidence from the Green Revolution”, American
Economic Review, 86:4, p.931-53

Foster, Andrew and Mark R. Rosenzweig (1995) “Learning by Doing and Learning from
Others: Human Capital and Technical Change in Agriculture”, The Journal of
Political Economy, 103: 6, p. 1176-1209.

Frankenberg, Elizabeth, James P. Smith, and DuncanThomas (2003), “Economic Shocks,
Wealth and Welfare”, The Journal of Human Resources, 38:2, p. 280-321.

183

 

F rankenberg, E1izabeth., Angelique Chan, M.B. Ofstedal (2002), “Stability and Change
in Living Arrangements in Indonesia, Singapore, and Taiwan, 1993-1999”
Population Studies, 56, p. 201-13.

F rankenberg, Elizabeth and L. Karoly (1995). The 1993 Indonesian Family Life Survey:
Overview and Field Report, Publication No. DRU-l 195/1-NICHD/AID, RAND,
Santa Monica, CA.

Frankenberg, Elizabeth. and DuncanThomas, (2000), The Indonesia Family Life Survey
(IFLS): Study Design and Results from Waves 1 and 2. RAND, Santa Monica,
CA. DRU-223 8/1-NIA/NICI-ID.

Geertz, Hildred (1961), The Javanese Family. New York: Free Press of Glencoe.

Haddad, L., J. Hoddinott, and H. Alderman (1997) eds. Intrahousehold Resource
Allocation in Developing Countries: Models, Methods, and Policy, Baltimore:
John Hopkins University Press.

Heaton, T.B., M. Carnmack, and L. Young (2001), “Why is the Divorce Rate Declining
in Indonesia?” Journal of Marriage and the Family 63:2. p,480—90.

Jay, Robert (1969), Javanese Villagers. Cambridge: MIT Press.

Johnson, R.W, and J. DaVanzo (1998),”Economic and Cultural Inﬂuences on the
Decision to Leave Home in Peninsular Malaysia”, Demography, 35:1 p. 97-114.

J olliffe, D. (1997), “Whose Education Matters in the Determination of Household
Income: Evidence from a Developing Country” IFPRI Food and Consumption
Nutrition Division Discussion Paper No. 39.

Jones, Gavin (1994), Marriage and Divorce in Islamic Southeast Asia, Kuala Lumpur:
Oxford University Press.

Koentjaraningrat, RM. (1985) Javanese Culture, Singapore: Oxford University Press.

Levine, David and Michael Kevane (2003), “Are Investments in Daughters Lower when
Daughters Move Away? Evidence from Indonesia”, World Development, 31:6, p.
1065-1084.

Lucas, R.E.B. (1997), “Internal Migration in Developing Countries” in Handbook of
Population and Family Economics, M. Rosenzweig and O.Stark, Vol. 1B,
Amsterdam: North Holland.

Lundberg, S., R. Pollak (1993), “Separate Spheres Bargaining and the Marriage Market”
The Journal of Political Economy, 101:6, p.988-1010.

184

 

Lundberg, S., R, Pollak, and T.J. Walles (1997), “Do Husbands and Wives Pool Their
Resources? Evidence from the United Kingdom Child Beneﬁt”, Journal of
Human Resources 32:3, p.463-80.

McElroy, MB. (1990), “The Empirical Content of Nash-Bargained Household
Behavior”, Journal of Human Resources, 25:4, p.559-83.

McElroy, MB. and M.J. Homey (1981), “Nash-Bargained Household Decisions: Toward
a Generalization of the Theory of Demand”, International Economic Review 22:2,
p. 333-49.

Quisumbing, Agnes eds. (2003), Household Decisions, Gender and Development: A
Synthesis of Recent Research, New York: IFPRI.

Rosenzweig, MR. (2003), “Payoff from Panels in Low-Income Countries: Economic
Development and Economic Mobility” American Economic Review, 93:2, p. 112-
17.

Rosenzweig, M.R., Oded Stark (1989), “Consumption Smoothing, Migration, and
Marriage: Evidence from Rural India”, The Journal of Political Economy, 97:4 p.
905-926.

Schultz, T.Paul (1990), “Testing the Neoclassical Model of Family Labor Supply and
Fertility”, The Journal of Human Resources, 24:4, p. 599-634.

Singh, 1., L. Squire, and J. Strauss (1986) Agricultural Household Models: Extensions,
Applications and Policy. Baltimore: John Hopkins University Press.

Smith, James P., and DuncanThomas (1998), “On the Road: Marriage and Mobility in
Malaysia”, The Journal of Human Resources, 33:4, p.805-32

Strauss, John, Kathleen Beegle, Bondan Sikoki, Agus Dwiyanto, Yulia Herawati and
F irman Witoelar (2003), The Third Wave of the Indonesia Family Life Survey
(IFLS3): Overview and Field Report. RAND, Santa Monica, California.

Strauss, John, and DuncanThomas (1995), “Human Resources: Empirical Modeling of
Household and Family Decisions”, in Handbook of Development Economics, J.

Behrman and T. N. Srinivasan, Vol. 3A, Amsterdam: North Holland.

Strauss, John, G. Mwabu and Kathleen Beegle (2000), “Intrahousehold Allocations: a
Review of Theories and Empirical Evidence”, Journal of African Economies, 9,
p.83-143.

Thomas, Duncan, D. Contreras, Elizabeth Frankenberg (2002),“Distribution of Power
within Household and Child Healt ”, RAND.

18S

 

Thomas, Duncan, James P. Smith, and Elizabeth Frankenberg (2001), “Lost but Not
Forgotten: Attrition and Follow-up in the Indonesia Family Life”, The Journal of
Human Resources, 36:3, p. 556-92.

Witoelar, F irrnan (2004), “Inter-household Allocations within Extended Households:
Evidence from the Indonesia Family Life Survey”, Two Essays on the Economics
of the Household of the Developing Countries, Dissertation, Michigan State
University.

186

 

       

S

l

 

11111111111111111