University of Southern California Dissertations and Theses

Essays on the empirics of risk and time preferences in Indonesia

(USC Thesis Other)

Essays on the empirics of risk and time preferences in Indonesia

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content ESSAYSONTHEEMPIRICSOF
RISKANDTIMEPREFERENCESININDONESIA
by
James L. Ng
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
August 2013
Copyright 2013 James L. Ng
Acknowledgments
This dissertation is the culmination of six years of work. Calling the journey towards
the PhD a marathon would not be wrong, but it would be inaccurate. While a
marathon is a solitary endeavor, I made it through this journey only with help from
friends, colleagues, and family.
I owe a debt of gratitude to my adviser, John Strauss. I would not have come
close to nishing without the knowledge and feedback which he so generously oered,
and the sometimes brutal but always constructive criticism that he never ceased to
provide. From working closely with him through the years, I like to think that I have
adopted his uncompromising insistence on careful empirical work, a trait that I have
no doubt will serve me well as I continue to grow as an applied economist.
I am no less grateful to the immensely knowledgeable and gracious Jerey Nugent.
I beneted tremendously from the informal Friday meetings which he organized, as
they gave me a way to present various stages of my research to other students. John
and Je are not just eminent scholars, they are role models. I count myself lucky to
be able to call them friends.
I am very grateful to Gary Painter, Geert Ridder, and Simon Wilkie for serving
on my qualifying and/or dissertation committees. Simon in particular brought to my
attention many useful papers from behavioral economics that I otherwise would not
have known about. I am also extremely grateful to Yong Kim for his help and support
in my job search.
Young Miller and Morgan Ponder ensured that bureaucratic nightmares were prac-
tically nonexistent throughout my time in the program. Funding from the Department
i
of Economics and Dornsife College meant I never had to worry about nances as a
PhD student. I cannot thank them enough.
I could well have false started out of the program were it not for my rst-year
core exam study companions: Arya Gaduh, Saurabh Singhal, and Osman Abbasoglu.
Arya and Saurabh were my frequent partners in discussions, be they research-related
or random. I also gained from picking Arya's brains about his native land and Stata
coding.
Last but not least, I am indebted to my family: my parents, Ng Ah Ngee and
Lee Fay May, for their unconditional support in whatever I choose to do, and my
wife, Shareen Lee, for being simultaneously my best friend, condante, and strongest
supporter.
ii
Table of Contents
Acknowledgments i
List of Tables vii
List of Figures x
Abstract xi
1 Introduction 1
1.1 Demographic and Cognitive Eects on Risk and Time Preferences . . 1
1.2 Interviewer Eects on Risk and Time Preferences . . . . . . . . . . . 4
1.3 Risk and Time Preferences and the Transition into Adulthood . . . . 5
2 Demographic and Cognitive Eects on Risk and Time Preferences 7
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Previous Research Eliciting Risk Aversion: Experiments . . . 8
2.2.2 Previous Research Eliciting Risk Aversion: Surveys . . . . . . 11
2.2.3 Previous Research Eliciting Time Preference . . . . . . . . . . 13
2.2.4 Preference Heterogeneity and Correlates . . . . . . . . . . . . 14
2.2.5 Cognition and Preferences . . . . . . . . . . . . . . . . . . . . 16
2.3 Data and Measurements . . . . . . . . . . . . . . . . . . . . . . . . . 17
iii
2.3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.2 A Note on Nomenclature . . . . . . . . . . . . . . . . . . . . . 17
2.3.3 Measuring Risk Aversion in IFLS-4 . . . . . . . . . . . . . . . 18
2.3.4 Measuring Rate of Time Preference in IFLS-4 . . . . . . . . . 22
2.3.5 Descriptive Statistics . . . . . . . . . . . . . . . . . . . . . . . 26
2.3.6 Comparison with Other Large-Scale Surveys . . . . . . . . . . 31
2.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.4.1 Correlates of Risk Aversion . . . . . . . . . . . . . . . . . . . 35
2.4.2 Correlates of Time Preference . . . . . . . . . . . . . . . . . . 37
2.4.3 Correlates of Gamble Aversion/Loving . . . . . . . . . . . . . 39
2.4.4 Correlates of Negative Time Discounting . . . . . . . . . . . . 39
2.4.5 Sex Dierences in the Eect of Cognition . . . . . . . . . . . . 42
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Appendices
2.A Selected Explanatory Variables . . . . . . . . . . . . . . . . . . . . . 46
2.B Logit Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3 Interviewer Eects on Risk and Time Preferences 48
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.3 Data and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.4.1 Interviewer Dummies are Signicant . . . . . . . . . . . . . . 53
3.4.2 Eect on Standard Preferences . . . . . . . . . . . . . . . . . 54
iv
3.4.3 Eect on Nonstandard Preferences . . . . . . . . . . . . . . . 57
3.4.4 Switching from Nonstandard to Standard . . . . . . . . . . . . 59
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Appendices
3.A Interviewer Fixed Eects . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.B Logit Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4 Risk and Time Preferences and the Transition into Adulthood 69
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.2 Dependent Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.2.1 Schooling Attainment . . . . . . . . . . . . . . . . . . . . . . . 70
4.2.2 Marriage Age . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.2.3 Fertility Timing . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.2.4 Migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2.5 Full Time Employment . . . . . . . . . . . . . . . . . . . . . . 76
4.3 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4.1 Schooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4.2 Marriage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.4.3 Fertility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4.4 Migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.4.5 Employment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.4.6 Controlling for Nonstandard Preferences . . . . . . . . . . . . 92
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.5.1 Measurement Error . . . . . . . . . . . . . . . . . . . . . . . . 93
4.5.2 Seemingly Unrelated Regressions . . . . . . . . . . . . . . . . 94
v
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
Appendices
4.A Number of Moves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.B Controlling for Nonstandard Preferences . . . . . . . . . . . . . . . . 97
4.C SUR Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
5 Conclusion 114
5.1 Demographic, Cognitive, and Interviewer Eects on Preferences . . . 114
5.2 Eect of Preferences on the Transition into Adulthood . . . . . . . . 116
Comprehensive bibliography 118
vi
List of Tables
2.1 Payos and corresponding preference coecients . . . . . . . . . 23
(A) Partial risk aversion coecients . . . . . . . . . . . . . . . . . . 23
(B) Discount rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Summary statistics of IFLS-4 respondents . . . . . . . . . . . . . 27
2.3 Cross tabulations of preferences A and B . . . . . . . . . . . . . . 28
(A) Cross tabulation of Risk A and Risk B . . . . . . . . . . . . . . 28
(B) Cross tabulation of Time A and Time B . . . . . . . . . . . . . 28
(C) Cross tabulation of Gamble Averse A and Gamble Loving B . . 28
(D) Cross tabulation of NTD A and NTD B . . . . . . . . . . . . . 28
2.4 Preference Distribution in Other National Surveys . . . . . . . . . 32
(A) Risk aversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
(B) Discount rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.5(A) Correlates of Risk Preference . . . . . . . . . . . . . . . . . . . . 36
2.5(B) Correlates of Time Preference . . . . . . . . . . . . . . . . . . . . 38
2.5(C) Correlates of Gamble Aversion/Loving . . . . . . . . . . . . . . . 40
2.5(D) Correlates of Negative Time Discounting . . . . . . . . . . . . . . 41
2.6(A) Correlates of Risk and Time Preferences by Sex . . . . . . . . . . 43
2.6(B) Correlates of Nonstandard Preferences by Sex . . . . . . . . . . . 44
2A.1 Logit Estimates of Tables 2.5(A) { 2.5(D) . . . . . . . . . . . . . . 47
vii
3.1(A) Summary statistics of IFLS-4 respondents . . . . . . . . . . . . . 50
3.1(B) Summary statistics of IFLS-4 interviewers . . . . . . . . . . . . . . 51
3.2(A) Eect of Interviewer Characteristics on Risk Preference . . . . . . 55
3.2(B) Eect of Interviewer Characteristics on Time Preference . . . . . 56
3.2(C) Eect of Interviewer Characteristics on Gamble Aversion/Loving 58
3.2(D) Eect of Interviewer Characteristics on Negative Time Discounting 59
3.3 Do Interviewers In
uence Change in Answer to Risk Filter Ques-
tion? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3A.1(A) Correlates of Risk Preference: Interviewer Fixed Eects . . . . . . 62
3A.1(B) Correlates of Time Preference: Interviewer Fixed Eects . . . . . 63
3A.1(C) Correlates of GA/GL: Interviewer Fixed Eects . . . . . . . . . . 64
3A.1(D)Correlates of NTD: Interviewer Fixed Eects . . . . . . . . . . . 65
3A.2(A) Interviewer Fixed Eects Do Not Change Risk & Time Main Results 66
3A.2(B) Interviewer Fixed Eects Do Not Change GA, GL & NTD Main
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3A.3 Logit Estimates of Tables 3.2(A) { 3.2(D) . . . . . . . . . . . . . . 68
4.1 Summary statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.2 Years of Schooling . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.3 Positive Assortative Mating on Preferences . . . . . . . . . . . . . 82
4.4 Marriage Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.5 Likelihood of First Birth by Age 18 . . . . . . . . . . . . . . . . . 84
4.6 Likelihood of Using Birth Control . . . . . . . . . . . . . . . . . . 86
4.7 Ever Migrate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.8 Hazard Model of Time to First Migration . . . . . . . . . . . . . 89
4.9 Likelihood of Full Time Employment . . . . . . . . . . . . . . . . 90
viii
4.10 Likelihood of Self Employment . . . . . . . . . . . . . . . . . . . 92
4A.1 Number of Moves . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4B.1 Years of Schooling . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4B.2 Marriage Timing . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
4B.3 Likelihood of First Birth by Age 18 . . . . . . . . . . . . . . . . . 99
4B.4 Likelihood of Using Birth Control . . . . . . . . . . . . . . . . . . 100
4B.5 Ever Migrate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4B.6 Hazard Model of Time to First Migration . . . . . . . . . . . . . 102
4B.7 Number of Moves . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4B.8 Likelihood of Full Time Employment . . . . . . . . . . . . . . . . 104
4B.9 Likelihood of Self Employment . . . . . . . . . . . . . . . . . . . 105
4C.1 SUR estimates of Risk A and other correlates, males . . . . . . . 106
4C.2 SUR estimates of Risk B and other correlates, males . . . . . . . 107
4C.3 SUR estimates of Time A and other correlates, males . . . . . . . 108
4C.4 SUR estimates of Time B and other correlates, males . . . . . . . 109
4C.5 SUR estimates of Risk A and other correlates, females . . . . . . 110
4C.6 SUR estimates of Risk B and other correlates, females . . . . . . 111
4C.7 SUR estimates of Time A and other correlates, females . . . . . . 112
4C.8 SUR estimates of Time B and other correlates, females . . . . . . 113
ix
List of Figures
2.1 Flowcharts illustrating elicitation of risk aversion . . . . . . . . . . . 20
(A) Risk Submodule A . . . . . . . . . . . . . . . . . . . . . . . . . 20
(B) Risk Submodule B . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.2 Flowcharts illustrating elicitation of time preference . . . . . . . . . 24
(A) Time Submodule A . . . . . . . . . . . . . . . . . . . . . . . . . 24
(B) Time Submodule B . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3 Distribution of risk aversion . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Distribution of time preference . . . . . . . . . . . . . . . . . . . . . 31
x
Abstract
This dissertation consists of three empirical essays investigating the shaping and con-
sequences of individual risk and time preferences in Indonesia. Risk and time prefer-
ences were elicited using hypothetical questions in the 2007/08 wave of the Indonesian
Family Life Survey.
The rst essay nds risk and time preferences to be systematically related to re-
spondent demographics and cognition. The main ndings are as follows. Men are
less risk averse than women, older respondents are more impatient, wealthier respon-
dents are less risk averse and less impatient, better educated adults are less impatient,
and adults with better cognitive capacity proxied by episodic memory are less im-
patient. In addition, the preference elicitation procedure revealed individuals with
nonstandard preferences, in that they chose a dominated payo. Measures of cog-
nitive capacity are associated with the likelihood of having nonstandard preferences
in expected ways. Respondents with better episodic memory are less likely to be
nonstandard. Better educated respondents are less likely to be nonstandard, in some
cases.
The second essay reports evidence that interviewer characteristics can in
uence
survey-elicited preferences. The empirical models from the rst essay are augmented
with controls for interviewer language ability, interview language, measures of social
distance between interviewers and respondents, and interviewer human capital. The
main nding is that where interviewers and respondents do not share a daily spoken
language, respondents are more likely to show nonstandard preferences. This eect
is found only in interviews conducted in languages other than the national language.
xi
However, estimates of the original explanatory variables are not changed by the inclu-
sion of interviewer characteristics, suggesting that interviewers are not an important
source of omitted variable bias in estimations of the determinants of survey-elicited
preferences.
The third essay asks whether survey-elicited risk and time preferences can explain
the transition into adulthood of Indonesian adults. I look at the correlations of survey-
elicited risk and time preferences with economic decisions commonly thought of as
markers of adulthood: schooling attainment, marriage, fertility behavior, migration,
and full time employment. The results suggest that dierences in risk and time
preferences do explain some of the variation in economic decisions. Risk averse men
attain fewer years of schooling, while risk aversion is not signicantly associated with
women's schooling. Impatient men and women attain fewer years of schooling. Risk
averse men and women marry earlier, and so do impatient women. Married couples
positively sort on risk and time preferences. Women who are more risk averse and
more patient are less likely to use birth control. Highly risk averse men are less likely
to migrate. Risk and time preferences are not signicantly correlated with entry into
full time employment. However, risk aversion is negatively correlated with selection
into self-employment among employed men and women.
xii
Chapter 1
Introduction
This dissertation consists of three essays which empirically investigate the shaping and
consequences of survey-elicited risk and time preferences of Indonesian adults. The
rst essay examines how risk and time preferences vary with individual socioeconomic
and cognitive characteristics. The second essay builds on the rst, and explores the
eect of interviewer characteristics in elicited risk and time preferences. The third
essay looks at how risk and time preferences may in
uence the individual's transition
into adulthood by considering their correlations with the timing of major life events:
schooling, marriage, fertility, migration, and employment. All three essays employ
data from the fourth wave of the Indonesian Family Life Survey | henceforth referred
to as IFLS-4 | a longitudinal survey of households representative of approximately
80 per cent of the Indonesian population.
1.1 Demographic and Cognitive Eects on Risk and Time
Preferences
The rst essay (Chapter 2) examines how Indonesians' preferences toward risk and
time vary with sociodemographic characteristics. Preferences towards risk and in-
tertemporal choice are fundamental components of models of economic behavior. In
standard economic theory, these attitudes are traditionally taken as given (Stigler
and Becker, 1977). In applied work, even more restrictive assumptions are adopted
1
for empirical tractability. For example, in the standard test of ecient risk sharing
against idiosyncratic income shocks, all households are assumed to have identical risk
preferences. While such identifying assumptions are necessary given the absence or
inadequacies of data on underlying preferences, what are the implications of such
assumptions? In the case of the standard test of ecient risk sharing, two recent
papers have shown that homogeneous risk aversion biases the test against the null
(Schulhofer-Wohl, 2011; Mazzocco and Saini, 2012). More generally, one implica-
tion of assigning identical and stable underlying preferences to all individuals is that
observed dierences in economic behavior are wholly attributable to dierences in
inputs such as prices, incomes or endowments.
This is not to say that economists have not examined the plausibility of xed pref-
erences. In a highly-in
uential pioneering study, Binswanger (1980, p. 395) examines
whether \dierences in behavior between farmers of dierent wealth levels are the
consequence of dierent attitudes towards risk or of dierent constraint sets such as
limitations on credit or on access to modern inputs." There is an ever-growing body
of work investigating the empirical validity of identical (across agents) preferences,
in dierent populations.
1
This has been made possible by the increased use of direct
preference elicitation methods in household surveys and of laboratory experiments.
In all these studies, preferences have been found to vary across individuals at a given
point in time. Given that this is the case, what can be said about how preferences
vary? For instance, are men more or less risk averse than women? Are older people
more or less impatient than younger ones? Perhaps surprisingly, the evidence on the
1
Barsky et al. (1997), Hamoudi (2006), Dohmen et al. (2011), Hryshko et al. (2011).
2
role of such characteristics in explaining the distribution of preferences is mixed.
2
This essay contributes to the ever-growing body of evidence on heterogeneous
preferences. I use data from the fourth wave of the Indonesian Family Life Survey
(IFLS-4), elded in late 2007 and early 2008, in which risk and time preferences
were directly elicited using hypothetical questions.
3
I conrm that elicited risk and
time preferences are heterogeneous across Indonesian adults. Furthermore, they are
systematically related to sex, age, wealth, education, and cognition.
An unusual feature of the IFLS-4 elicitation method is it identies adults with
\nonstandard" risk and time preferences, in that they chose a dominated outcome over
a dominating one. In the context of expected utility, respondents with nonstandard
risk preference are either gamble averse or hyper-gamble loving. Respondents with
nonstandard time preference can be characterized as having a negative time discount
in the discounted utility framework. Another aim of this chapter is to determine the
correlates of these nonstandard preferences.
The main ndings are as follows. Men are less risk averse than women, older
respondents are more impatient, wealthier respondents are less risk averse and less
impatient, better educated adults are less impatient, and respondents with better cog-
nitive capacity proxied by episodic memory are less impatient. In addition, cognitive
capacity is associated with a lower likelihood of nonstandard preference. Respondents
with better episodic memory, a measure of cognition, are less likely to be nonstandard
2
According to Eckel and Grossman (2008), most studies that use indirect measures of risk aver-
sion from eld data conclude that women are more risk averse than men, but laboratory experimental
evidence is less conclusive. They note, however, that \both feld and lab studies typically fail to con-
trol for knowledge, wealth, marital status and other demographic factors that might bias measures of
male/female dierences in risky choices." Examples of contrasting ndings on the risk aversion{age
relationship are Barsky et al. (1997), who nd an inverse U bottoming at 60{65 years in an HRS
sample of U.S. adults aged 50 and older with a mean age of 56, and Guiso and Paiella (2008), who
nd a positive slope in a PSID sample of adults aged between 22 to 89 with a mean of 54.
3
This is the rst-ever large-scale attempt to elicit risk and time preferences in Indonesia. Miyata
(2003) and Cameron and Shah (2012) collect experimental data on, respectively, 400 and 1550
Indonesian villagers in 2000 and 2008, from which they estimate risk aversion parameters.
3
across all specications. The evidence for the role of education is somewhat weaker,
but the relationship between education and the likelihood of nonstandard preference
still has a negative sign.
1.2 Interviewer Eects on Risk and Time Preferences
The second essay (Chapter 3) is motivated by Binswanger's (1980) nding of inter-
viewer bias in survey-elicited risk preference. I investigate whether interviewers sig-
nicantly aect respondents' answers to the IFLS-4 elicitation questions, and whether
the inclusion of interviewer eects changes the estimated eects of demographics and
cognition from the rst essay. I nd evidence of interviewer eects on preferences,
but no evidence that their omission biases the coecients of the demographic and
cognitive variables. The most interesting result is language eects signicantly af-
fect the likelihood of nonstandard preference. In particular, where interviewers and
respondents do not share a daily spoken language, respondents are more likely to
show nonstandard preferences. This eect is found only in interviews conducted in
languages other than the national language.
To my knowledge, there is currently no paper in the economics literature examin-
ing language eects in survey enumeration, which is surprising given the multilingual
nature of many parts of the world in which populations are surveyed. The evidence
presented in this chapter suggests that in a linguistically-diverse environment, lan-
guage can aect survey responses in nonrandom ways.
4
1.3 Risk and Time Preferences and the Transition into
Adulthood
The third essay (Chapter 4) seeks to answer the following question: Do individual
risk and time preferences drive dierences in the transition into adulthood among
Indonesians? To investigate this question, I again draw on data from IFLS-4. I
consider ve markers of adulthood: schooling attainment, marriage, fertility behavior,
migration, and full time employment. If risk and time preferences are signicantly
correlated with the timing of these markers even after controlling for socioeconomic
characteristics, the implication is that socioeconomic characteristics only partially
explain observed dierences in adulthood transition behavior, with at least some role
played by preferences which are traditionally held to be intrinsic and unchanging
(Stigler and Becker, 1977). Policymakers may need to account for such preferences in
helping youth transition smoothly into adulthood, especially in developing societies
with incomplete or missing markets.
The results indicate that risk and time preferences do play a signicant role in some
markers of adulthood transition. Men who are more risk averse or impatient, and
women who are more impatient attain fewer years of schooling. Risk averse men and
women marry earlier, and so do impatient women. Married couples positively sort on
risk and time preferences. Risk and time preferences are not signicantly correlated
with female fertility timing. However, high risk aversion and low impatience are
signicantly negatively correlated with birth control use among women. Highly risk
averse men are signicantly less likely to migrate. Risk and time preferences are
not signicantly correlated with entry into full time employment, but conditional on
being employed, highly risk averse men and women are signicantly less likely to be
5
self employed.
6
Chapter 2
Demographic and Cognitive Eects on Risk and
Time Preferences
2.1 Introduction
This chapter contributes to the body of evidence on heterogeneous preferences by
examining how risk aversion and time preference of Indonesians vary with sociode-
mographic characteristics, drawing from data from the Indonesian Family Life Survey,
a longitudinal survey of households. I use data from the fourth wave, IFLS-4, elded
in late 2007 and early 2008, in which risk and time preferences were directly elicited
using hypothetical questions. I conrm that elicited risk and time preferences are het-
erogeneous across Indonesian adults. Furthermore, they are systematically related to
sex, age, wealth, education, and cognition.
An unusual feature of the IFLS-4 elicitation method is it identies adults with
\nonstandard" risk and time preferences, in that they chose a dominated outcome over
a dominating one. In the context of expected utility, respondents with nonstandard
risk preference are either gamble averse or hyper-gamble loving. Respondents with
nonstandard time preference can be characterized as having a negative time discount
in the discounted utility framework. Another aim of this paper is to determine what,
if any, are the correlates of these nonstandard preferences.
1
1
A recent working paper by Choi et al. (2013) reports results from a large scale experiment to
test for consistency with utility maximization. The researchers nd a strong positive correlation
between consistency and wealth.
7
The main ndings of this paper are as follows. Men are less risk averse than
women, older respondents are more impatient, wealthier respondents are less risk
averse and less impatient, better educated adults are less impatient, and respondents
with better cognitive capacity proxied by episodic memory are less impatient. In
addition, cognitive capacity is associated with a lower likelihood of nonstandard pref-
erence. Respondents with better episodic memory, a measure of cognition, are less
likely to be nonstandard across all specications. The evidence for the role of educa-
tion is somewhat weaker, but the relationship between education and the likelihood
of nonstandard preference still has a negative sign.
The chapter is organized as follows. The next section oers a literature review
of previous elicitations of risk and time preferences, and of empirical evidence on the
heterogeneity and correlates of these preferences. Section 2.3 details the IFLS-4 pref-
erence elicitation procedure and the data used in the empirical analyses. Section 2.4
describes the estimation strategy used and presents the results. Section 2.5 concludes.
2.2 Literature Review
2.2.1 Previous Research Eliciting Risk Aversion: Experi-
ments
Early attempts at eliciting risk attitudes were carried out primarily by experimental
psychologists (for a review, see Slovic, 1964). Psychologists have a number of dierent
conceptualizations of risk preference, but the one that is most closely related to the
economic concept of risk aversion is what Slovic termed \probability and variance
preference measure". These studies used both hypothetical and actual payos, but
actual payos and sample sizes were typically small.
8
In economics, Binswanger (1980) is a highly in
uential study that represents a
marked improvement over previous attempts to elicit risk attitudes.
2
First, where
heretofore economists had used hypothetical gambles to measure risk aversion, Bin-
swanger used actual gambles with real payos. Second, the psychology experiments
to that point had low payos and small sample sizes, whereas Binswanger ran his
experiment on 330 randomly-selected farmers in rural India using payos which were
high relative to their incomes.
Subjects were asked to choose between eight lotteries, with the riskiest lottery
oering equally-likely payos of Rs 0 in the low case and Rs 200 in the high case,
and the safest lottery oering Rs 50 in both the low and high case. Based on his
choice, a subject could be ranked on a risk aversion scale. Binswanger played this
basic lottery-choice game 17 times with each subject over a span of 5 to 6 weeks.
Payo levels were varied by multiplying the basic payos by 1/100, 1/10 or 10. Some
rounds consisted of hypothetical lotteries to test the validity of hypothetical-lottery
answers vis-a-vis actual gambles.
The main results of Binswanger's study are as follows. (a) Risk aversion increased
with payo levels: At low payo levels, individual risk aversion was widely distributed
from risk loving to intermediate risk aversion. But at high payo levels, risk aversion
was concentrated in the moderate risk aversion category, and risk neutrality virtu-
ally vanished. (b) At high payo levels, wealth does not signicantly in
uence risk
aversion. (c) When subjects had played only low-payo actual gambles beforehand,
hypothetical answers yielded a risk aversion distribution signicantly more dispersed
than that from the actual gambles. However, once a high-payo actual gamble was
played, there was no signicant dierence between real choices and hypothetical an-
2
A survey of risk preference elicitations in the laboratory is by Harrison and Rutstr om (2008).
Carpenter and Cardenas (2008) is a survey of laboratory-elicited risk and time preferences in devel-
oping countries.
9
swers.
Another early paper that studied risk attitudes in poor agricultural communi-
ties is that of Walker (1981). To elicit risk aversion, Binswanger-type lottery ex-
periments and interviews were held with 40 farmers, selected at random, from two
maize-farming villages in El Salvador. The villages were similar in terms of access to
markets and agroclimatic environment, but very dissimilar in levels of hybrid maize
adoption. Walker found that dierences in risk perceptions (of their potential expo-
sure to drought) explained regional dierences in adoption. However, he found no
signicant dierence in elicited risk aversion between adopters and nonadopters of
hybrid maize technology.
Kachelmeier and Shehata (1992) conducted a series of laboratory experiments
on 185 student volunteers in China to elicit risk aversion. They oered real-money
payos that were very large relative to the students' living costs. Their study diers
from Binswanger's in that the students were not asked to choose among dierent
lotteries but asked to provide their own certainty equivalents in individual lotteries.
Besides, various probabilities were used, not just the equal-probability coin toss used
by Binswanger.
3
Despite the methodological dierences, they too found that risk
aversion increased as payos increased.
Holt and Laury (2002) used a menu of paired lottery choices, similar to Bin-
swanger, but they varied the probabilities from 1/10 to 10/10. They ran four consec-
utive tasks in which subjects chose from the menu of paired lottery choices. The rst
task had low-payos but subjects were paid in real money. The second task involved
hypothetical payos scaled up by a factor of 20 compared to the rst task. The third
task was also high-payo, except the payos were real-money. The fourth and nal
3
They were working with a sample of undergraduates, whereas Binswanger's subjects were rural
farmers in India with very low levels of education, for which the concept of probabilities was not
well understood.
10
task was a repeat of the rst task, a "return to baseline" as the authors put it. To
control for potential wealth eects between the high and low real-payo treatments,
subjects were required to give up what they had earned in the rst, low-payo task
in order to participate in the third, high-payo task. Whereas Binswanger only in-
serted hypothetical lotteries sparingly, Holt and Laury conducted both actual and
hypothetical lotteries in almost all rounds. Among their dings was risk aversion in-
creased, once again, going from low-payo rounds to high-payo rounds. This eect
was not observed in the hypothetical lotteries, which the authors took to mean that
preferences elicited from hypothetical lotteries with high stakes may not be valid.
2.2.2 Previous Research Eliciting Risk Aversion: Surveys
While experimental studies have the advantage of oering real-money payos and thus
ensuring some level of incentive compatibility, they are usually too time consuming
and cost prohibitive to be administered on a large scale. Incorporating hypothetical
lotteries into household surveys allows the direct elicitation of preferences from large,
nationally-representative samples. In the U.S., the Health and Retirement Study
(HRS), Panel Study of Income Dynamics (PSID), and National Longitudinal Survey
of Youth 1979 (NLSY79) all administered survey questionnaires involving hypotheti-
cal lottery choice. The HRS also directly elicited time preference for a small subset of
respondents. Among large-scale household surveys in developing countries, the 2005
MxFLS is probably the rst to incorporate risk or time preference elicitation ques-
tionnaires. The disadvantage, of course, is responses to hypothetical lotteries may
not be behaviorally valid.
Fortunately, some recent studies have found that survey-elicited risk preference is
consistent with experimentally-derived measures, which contradicts Holt and Laury,
11
and to a lesser extent Binswanger. Dohmen et al. (2011) consider a \general risk
question" in the 2004 German Socioeconomic Panel (GSOEP), in which the 22,000
respondents were asked to rate their willingness to take risks on an 11-point scale.
They then run a laboratory experiment on a representative sample of 450 German
adults formed using the same sampling methodology as the GSOEP.
4
On this exper-
imental sample, they ask the GSOEP's general risk question and also play the usual
lottery-choice game with real-money payos. They nd that risk aversion measured
by the general risk question is signicantly positively correlated with risk aversion
measured by choices in the real-money game, and interpret this as validation of the
GSOEP's general risk question. While informative, this result is irrelevant to house-
hold surveys which elicit risk aversion using hypothetical lottery choice. Furthermore,
there is no guarantee that the particular validation established for the experimental
sample would carry over to the GSOEP sample.
5
Perhaps the most relevant evidence (to the present paper) on the validity of hy-
pothetical lotteries comes from the MxFLS, a nationally-representative survey of all
adult residents in 8200 households (Hamoudi, 2006). A subset of respondents was
chosen to play a Binswanger-style lottery-choice game involving real-money payos.
Participants were presented with six bisected circles and asked to choose one of the
six circles. The circles were arranged from a zero-risk option, with the same mone-
tary value written in both halves, to increasingly riskier ones. Respondents were paid
based on a coin
ip; with 50% probability they would receive the amount written
in the left half of their chosen circle, otherwise they received the amount written in
the right half. After a substantial amount of time had passed from the completion
4
The authors do not mention whether the experimental sample consisted of the same adults who
were in the GSOEP.
5
This study was essentially replicated for rural Thailand where the same conclusions were reached
(Hardeweg et al., 2011).
12
of the the real-money game, the same respondents were interviewed. The interviews
involved two instruments meant to measure respondents' willingness to take nan-
cial risk. One was similar to the real-money game but with hypothetical payos.
The other was a series of binary choices over hypothetical gambles similar to the one
employed in IFLS-4. Hamoudi nds that elicited levels of risk aversion in the two
hypothetical instruments were strongly predicted by elicited risk aversion in the real-
money game; respondents who were highly risk averse in the real-money game were
signicantly likely to be highly risk averse in the hypothetical instruments as well.
2.2.3 Previous Research Eliciting Time Preference
There is currently no gold standard for measuring time preference (Chabris et al.,
2007). The most widely-used method, which is also the method used in IFLS, asks a
series of questions, each of which asks the subject to choose between a sooner, smaller
payo and a later, larger payo. The competing payos are denominated in the same
goods, typically amounts of money. The subject's discount rate is backed out by
assuming a discount function. Most studies assume that the utility function is linear
in consumption, i.e. agents are risk neutral. Most studies also assume no savings,
i.e. the payo is consumed the moment it is received. This method is used both in
experiments with actual payos and in surveys with hypothetical payos.
To the best my knowledge, besides IFLS-4, discount rates have been directly
elicited using the above method for large national samples only in the 2005 MxFLS
and 2006 NLSY. The 1972 PSID asked context-specic subjective questions to score
respondents on a 9-point scale of impatience (Leigh, 1986). The NLSY prior to
2006 contained interviewer assessments of subjects' in-interview impatience and other
self-reported measures, which researches have used to proxy for time preference (for
13
example DellaVigna and Paserman, 2005).
2.2.4 Preference Heterogeneity and Correlates
Previous studies of nationally-representative surveys have found considerable indi-
vidual heterogeneity in directly-elicited risk preferences, although most respondents
appear to be at least moderately risk averse (see Table 2.4 for a summary). In their
survey of measurements of risk preference in the laboratory, Harrison and Rutstr om
(2008, p. 130) conclude that \At a substantive level, the most important conclu-
sion is that the average subject is moderately risk averse, but there is evidence of
considerable individual heterogeneity in risk attitudes in the laboratory."
Many studies have investigated the correlates of risk preferences in survey data
and in the lab. To cite just a few of the studies using survey data, Leigh (1986) nds
in the 1972 PSID that schooling, marriage, number of children, and wages are posi-
tively associated with risk aversion. The risk aversion measure used in the 1972 PSID
was elicited not from lottery choice questions, however, but from context-specic
\willingness to take risk"-type questions similar to the kinds used in the GSOEP.
In the GSOEP, Dohmen et al. (2011) show that age and being female are positively
associated with risk aversion, and that height and parental education are negatively
associated with risk aversion. In the 1996 PSID, Hryshko et al. (2011) nd that
parental education, instrumented by changes in compulsory schooling laws which af-
fected the parents, negatively aects hypothetical lottery-elicited risk aversion. They
also nd age, being a female head of household, and parental risk aversion to be
positively associated with risk aversion.
The survey paper by Frederick et al. (2002) provides an exhaustive summary of
studies that elicit discount rates from 1978 to 2002, the vast majority of which are
14
experimental. The HRS has survey-elicited discount rates for a very small subset of
respondents but demographic correlations were not explored due to the small sample
size (Barsky et al., 1997). Two studies that look at correlates of time preference in
national samples are by Leigh (1986) and Rubalcava et al. (2009). The former nds in
the 1972 PSID that schooling, marriage, number of children and wages are negatively
correlated with impatience. The latter study nds that women in the 2005 MxFLS
are more patient than men.
Economic theory has traditionally taken risk and time preferences to be invariant
over time, although there have been exceptions, e.g. the work of Becker and Mulligan
(1997) on endogenous time preference. An emerging literature examines the reaction
of risk aversion and time preference to large shocks. For example, Cassar et al.
(2011) and Cameron and Shah (2012) nd in repeated cross sections that Thai and
Indonesian villages which experienced a natural disaster are on average more risk
averse than those which did not. Guiso et al. (2011) elicit risk aversion from a large
sample of clients of an Italian bank before and after the 2008 nancial crisis and nd
that the same individuals are more risk averse post-crisis.
I am aware of only two working papers focusing on the stability of time prefer-
ence. In a study conducted over two years with 1,400 Boston-area residents, Meier
and Sprenger (2010) nd no change in the aggregate distribution of incentivized task-
elicited discount factors. While they do nd that some individuals had unstable dis-
count factors, they are uncorrelated with socioeconomic and demographic variables.
Krupka and Stephens (2010) show for a sample of Seattle and Denver residents that
discount rates elicited through hypothetical choice questions over two years changed
over time. With cross-sectional data, I cannot examine within-respondent stability of
preferences across time, but it makes for a natural extension to this paper after the
next wave of the IFLS is elded.
15
2.2.5 Cognition and Preferences
Does an individual's cognitive ability have anything to do with his level of risk aver-
sion or impatience? The notion that more-intelligent people are more patient is an
old one among economists, but there is no widely shared prior on the relationship
between cognition and risk aversion (Benjamin et al., 2006; Frederick, 2005). Using
GSOEP data, Dohmen et al. (2010) nd that nonverbal IQ and word
uency are pos-
itively correlated with patience and risk tolerance. Burks et al. (2009) report similar
ndings for a sample of U.S. truckers. Frederick (2005) nds that performance in
cognitive re
ection tests is positively correlated with patience and risk tolerance in
the domain of gains, but negatively correlated with risk tolerance in the domain of
losses. Interestingly, he nds strong evidence of gender dierences; \smarter" women
are more patient while \smarter" men take more risks.
It should be emphasized that cognition is not a monolithic trait. Psychologists
have long distinguished between two broad dimensions of intelligence:
uid intelli-
gence and crystallized intelligence (Cattell and Experiment, 1963). Fluid intelligence
involves abstract reasoning and executive function, whereas crystallized intelligence
comes from the acquisition of education and experience. Later researchers, in par-
ticular Horn (1985), have rened this distinction to include up to 10 dimensions,
including short-term, or episodic, memory (McArdle et al., 2002, 2009).
I look for a relationship between performance in a word recall test and risk and
time preferences. Word recall measures episodic memory, which is \a very general
measure of an important aspect of
uid intelligence since access to memory is basic
to any type of cognitive ability" (McArdle et al., 2009, p. 9), and has been used in
the HRS.
In the test, respondents were read a list of ten nouns and then asked to repeat
16
as many words as they can recall, in any order. 12 to 15 minutes later, after other,
unrelated questions were asked, respondents were again asked to repeat the words.
The average number of correctly recalled words over both attempts is the measure of
cognition used in this paper.
2.3 Data and Measurements
2.3.1 Data
IFLS is an ongoing, multi-purpose longitudinal household survey, based on a sample
of households representing about 83% of the Indonesian population living in 13 of the
nation's then 26 provinces in 1993. To date, four waves have been elded, in 1993/94,
1997, 2000, and 2007/08. This paper uses data from the fourth wave, IFLS-4.
6
IFLS-4 features risk and time elicitation modules which were administered to
29,054 respondents who were at least 15 years of age. This is the sample which
this paper studies. In the following subsections, I explain the modules in detail and
provide some important descriptive statistics of the sample.
2.3.2 A Note on Nomenclature
The rest of the paper follows the nomenclature described here. As will be explained
in the next two subsections, the risk elicitation and time elicitation modules each
contain two sets of questions, A and B. \Risk Submodule A(B)" refers to the risk
preference elicitation submodule which uses set A(B) questions. \Risk A(B)" refers
to the elicited risk aversion measure from Risk Set A(B). The same goes for time pref-
erence. Recall that respondents who chose a dominated option under expected utility
6
For more information about IFLS, see the user guide (Strauss et al., 2009) which can be found
on the ocial website: http://www.rand.org/labor/FLS/IFLS.html
17
maximization are labeled gamble aversion, gamble loving, or negative time discount-
ing. These labels are shortened to GA A, GL B and NTD, where necessary. Gamble
averse/loving and NTD respondents are collectively referred to as \nonstandard" re-
spondents. Respondents who chose the dominating option under expected utility
maximization and thus could be ranked on the risk or time preference categories are
referred to as \standard" respondents.
2.3.3 Measuring Risk Aversion in IFLS-4
The elicitation of risk aversion followed the commonly-used approach of a lottery
choice task. Two sets of questions were asked. Risk Submodule A was asked rst
followed by B. The two sets dier in the magnitude of the payos and the variance of
their expected payos. Submodule B's certain and expected payos are higher than
A's. The uncertain payos in Submodule B have higher coecients of variation than
those in A, re
ecting a higher risk-reward ratio.
The respondent was asked a series of questions. In each question, the respondent
was presented with two choices, a sure amount and a probability-based alternative,
and asked to choose which he would prefer. Importantly, the questions were not
framed as gambles over varying probabilities but rather presented as equal-probability
outcomes, since in environments where education levels are low, understanding of
probabilities is poor. The monetary values were economically meaningful, as monthly
GDP per capita was about Rp 1.4 million in 2007 (World Bank). Specically, the
respondent was asked:
\Suppose you are given two options of receiving income. In the rst
option you are guaranteedX rupiah per month. In the second option
you are guaranteed Y or Z rupiah, each with equal chance. Which
option would you choose?"
18
Monetary values X, Y and Z varied according to the
owchart in Figure 2.1.
The rst question in each submodule was a lter meant to identify respondents
who passed on the option which would have made them at least as well o as their
chosen option. In Risk Submodule A's lter question, the dominated option was
a sure amount, Rp 800 thousand, whereas the rst order stochastically dominant
alternative was an equal-probability option (equal chance of 800 thousand or 1.6
million). These respondents were given a chance to reverse their decision. They
stuck with their choice even after the interviewer had explained to them that the
probability-based alternative assured them of at least as much as the sure amount.
Following Hamoudi (2006), I refer to those respondents who chose the dominated
certain payo as \gamble averse". Gamble averse respondents exited Risk Submodule
A at this stage. As shown in Table 2.2, 42 percent of respondents exited this module
as gamble averse, even though no one should be expected to do so. In a similar
module conducted in the Mexican Family Life Survey, 5 percent of respondents were
classied as gamble averse (Hamoudi, 2006). We can only speculate as to the reasons
for the cross-country dierence, although the higher level of human development in
Mexico is one plausible explanation.
The lter question in Risk Submodule B was dierent. Here the dominated option
was the uncertain gamble (4 million or 2 million with equal chance), and the domi-
nating option was 4 million for sure. These respondents may have misunderstood the
question, misreported their choice, or they may in fact be hyper, even irrationally,
gamble loving. I refer to respondents who chose the dominated payo in the B lter
as \gamble loving". Gamble loving respondents exited Risk Submodule B at this
stage.
Respondents who passed the lter question faced a sequence of questions whereby
they had to choose between a certain payo and an uncertain payo. Where the
19
Figure 2.1: Flowcharts illustrating elicitation of risk aversion
Panel A: Risk Submodule A
or 800K | 1600K 800K
Exit
Gamble
Averse
or 400K | 1600K 800K
600K | 1600K or 800K 800K or 200K | 1600K
Exit
Risk A = 4
Exit
Risk A = 3
Exit
Risk A = 2
Exit
Risk A = 1
Panel B: Risk Submodule B
or 4M 4M | 2M
Exit
Gamble
Loving
or 12M | 0 4M
8M | 2M or 4M 4M or 16M | -2M
Exit
Risk B = 4
Exit
Risk B = 3
Exit
Risk B = 2
Exit
Risk B = 1
Questions posed to respondents were of the form: Suppose you are given two options of receiving income. In the rst
option you are guaranteed X rupiah per month. In the second option you are guaranteed Y or Z rupiah, each with
equal chance. Which option would you choose? where monetary valuesX,Y andZ vary according to the
owcharts
above.
20
respondent found himself in the sequence depended on his choice in the previous
question. If he picked the sure amount, the next question contained an uncertain
payo that was less risky; on the other hand if he picked the uncertain payo, his next
question contains an uncertain payo that is more risky.
7
In this way, respondents can
be grouped into four ordinal levels of risk aversion based on their certainty equivalents
at the termination of the interview.
The most risk averse respondents will exit the interview at the terminal node
represented in the lower left corner of Figure 2.1; the least averse exit in the lower
right node. The terminal nodes therefore represent an ordinal ranking of risk aversion
among the respondents. Respondents with risk aversion = 4 are the most risk averse,
and those with risk aversion = 1 are least risk averse.
An expected utility maximizer will choose the uncertain payo over the certain
payo if the expected utility from the uncertain payo is equal to or greater than
that from the certain payo, that is, if:
1
2
U(Y ) +
1
2
U(Z)U(X) (2.1)
and vice versa. The values for the payos X, Y , and Z are given in Figure 2.1.
Although respondents can be ranked on risk aversion free of functional form, for
the purpose of comparison with previous research, let us suppose, following much of
the literature, that respondents possess CRRA utility so that for a monetary value c,
U(c) =
c
(1r)
1r
(2.2)
The parameter r is the constant coecient of relative risk aversion. Substituting
7
The riskiness of an uncertain payo can be represented by its coecient of variation, which is
the ratio of the standard deviation of the two probabilistic payos to their mean.
21
c for the payos and solving for r in (2.1) for each question in the elicitation module
gives ranges for the coecient of relative risk aversion as shown in Table 2.1, Panel A.
8
2.3.4 Measuring Rate of Time Preference in IFLS-4
The process of eliciting time preference was similar to that of risk aversion. Respon-
dents were asked a series of questions of the form:
\Suppose you have won a prize. How will you choose to be paid?"
The respondent had to answer the question by choosing one of two options. The rst
option was an amount to be paid today and was held constant throughout the series.
The second option was a larger nominal amount to be paid at a future date (either one
year or ve years later). The payo for the second option was changed over the course
of the series to re
ect dierent subjective discount rates. The question a respondent
faced at a given point in the series depended on his choice in the previous question.
The elicitation process is best illustrated in a
owchart (Figure 2.2). Respondents
who exit at Category 1 have the lowest time preference, and so are the most patient.
At the other end, those who exit at Category 4 are the most impatient as they have
the highest time preference.
As with the risk questions, the time preference questions were asked in two sepa-
rate submodules. Time Submodule A questions oer a higher annualized return than
B questions and have a shorter time horizon (one year as opposed to ve for the
latter).
We can back out a range of discount rates for each of the four categories if we
assume that each response results from a discounted utility calculation (Samuelson,
8
I ignore background wealth in the calculation of r. Most expected utility-based measurements
of risk aversion in developing countries (Carpenter and Cardenas, 2008) and developed countries
(Harrison and Rutstr om, 2008) alike ignore income from outside the experiment.
22
Table 2.1: Payos and corresponding preference coecients
Panel A: Risk aversion
Category Terminal payo chosen
(Rupiah)
Terminal payo forgone
(Rupiah)
Constant relative risk
aversion coecient, r
% of standard
respondents
Risk A
4 800,000 for sure Equal chance of 600,000 (2.915,1) 50.3
or 1,600,000
3 Equal chance of 600,000 800,000 for sure (0.999, 2.915) 9.5
or 1,600,000
2 800,000 for sure Equal chance of 200,000 (0.306, 0.999) 13.7
or 1,600,000
1 Equal chance of 200,000 800,000 for sure (0, 0.306) 26.5
or 1,600,000
Risk B
4 4,000,000 for sure Equal chance of 8,000,000 (0.999,1) 84.1
or 2,000,000
3 Equal chance of 8,000,000 4,000,000 for sure (0.369, 0.999) 7.5
or 2,000,000
2 4,000,000 for sure Equal chance of 16,000,000 (0, 0.369) 2.9
or loss of 2,000,000
1 Equal chance of 16,000,000 4,000,000 for sure (0, 0.369) 5.6
or loss of 2,000,000
Panel B: Time preference
Category Terminal payo chosen
(Rupiah)
Terminal payo forgone
(Rupiah)
Annual discount rate,
100%
% of standard
respondents
Time A
4 1,000,000 now 6,000,000 a year from now (500,1) 71.5
3 6,000,000 a year from now 1,000,000 now (200, 500) 14.6
2 1,000,000 now 2,000,000 a year from now (100, 200) 5.9
1 2,000,000 a year from now 1,000,000 now (0, 100) 8.0
Time B
4 1,000,000 now 10,000,000 ve years from now (58.5,1) 81.0
3 10,000,000 ve years from now 1,000,000 now (32, 58.5) 13.1
2 1,000,000 now 2,000,000 ve years from now (14.9, 32) 3.8
1 2,000,000 ve years from now 1,000,000 now (0, 14.9) 2.1
1937). A discounted utility maximizer will choose the present payo X over a payo
Y at period from today if the utility from the present payo is at least as good
as the present value of the future payo. Following most of the literature (Chabris
et al., 2007), I assume that utility is linear in its argument,
9
so that the respondent
chooses X over Y if:
9
The linear utility function implies that agents are assumed to be risk neutral. Andersen et
al. (2008) have shown that discount rates are signicantly lower when the curvature of the utility
function is accounted for than when linear utility is assumed. Hence, they advocated joint elicitation
of risk and time preferences.
23
Figure 2.2: Flowcharts illustrating elicitation of time preference
Panel A: Time Submodule A
or 1M now
1M one year
from now
Exit
Negative
Time
Discount
or
3M one year
from now
1M now
6M one year
from now
or 1M now 1M now or
2M one year
from now
Exit
Time A = 4
Exit
Time A = 3
Exit
Time A = 2
Exit
Time A = 1
Panel B: Time Submodule B
or 1M now
0.5M five years
from now
Exit
Negative
Time
Discount
or
4M five years
from now
1M now
10M five years
from now
or 1M now 1M now or
2M five years
from now
Exit
Time B = 4
Exit
Time B = 3
Exit
Time B = 2
Exit
Time B = 1
Questions posed to respondents were of the form: Suppose you have won a prize. How will you choose to be paid?
The
owcharts above contain the two choices available for each question, one an immediate but smaller sum of money,
the other a later but larger sum.
24
X (
1
1 +
)

Y (2.3)
where (
1
1+
)

is the exponential discount factor, and is the discount rate, assumed
to be constant.
10
Plugging in the X and Y values into (2.3) (see Figure 2.2) and solving for
gives annual discount rate ranges for each time preference category, as shown in
Table 2.1, Panel B.
11
The higher a person's discount rate, the more he discounts
future utility, implying a higher degree of impatience.
As in the risk module, respondents were rst given a lter question. The lter
question asked respondents to choose between receiving an amount today and the
same amount at a future date. Respondents who chose the latter option thus preferred
to defer receiving money, even without compensatory interest. As in the risk module,
these respondents were given a chance to reverse their decision; they were informed
by the interviewer that they could receive the same amount of money today and that
waiting will not increase the amount of money they could receive.
Respondents who chose to defer receiving money without compensation may have
done so for a number of reasons. They could be extremely patient. They could be
10
Constant exponential discounting implies preferences over a given bundle do not change for
a given time horizon no matter when they occur, a property known as time consistency. (To be
sure, the implication is a little more nuanced: constant discounting implies time consistency only
if the discount function is the same in all periods, i.e. stationary.) First, time consistency implies
if today a person decides he prefers $2000 now over $2200 in one month, he would also prefer
$2000 in 11 months over $2200 in 12 months. In reality, when faced with this choice today, most
people would choose to receive the $2200 in 12 months rather than $2000 in 11 months. Second,
time consistency implies when subjects are asked to compare a sooner but smaller payo to a later
but larger payo, their implicit discount rates should be the same under both time horizons. In
reality we frequently observe declining discount rate over longer time horizons. This goes against
time consistent preferences. The standard DU model does not permit this lowering of the discount
rates over longer time horizons. The quasi-hyperbolic discount factor of Laibson (1997) has been
found to better account for time inconsistency in empirical data. Courtemanche et al. (2012) report
evidence of time inconsistency in 85% of 2006 PSID respondents. IFLS is not suited to detecting
time inconsistency since respndents do not get to name their own future payos.
11
The time horizon is 1 year for Time A and 5 years for Time B. Since the per-period discount
rate is a constant and I take a period to be a year, is an annual rate.
25
exhibiting anticipatory saving or using deferment as a commitment savings device.
12
They could simply be irrational or have not understood the question. Whatever the
case, these respondents could dier in unobserved ways from other respondents, who
would accept a windfall now instead of waiting a year or ve years to receive the same
windfall without interest, and so should be considered separately. Since the only way
this is possible in the discounted utility framework of 2.3 is for to be negative, I call
these respondents \negative time discounters".
13
2.3.5 Descriptive Statistics
Table 2.2 presents summary statistics of the individual and household characteristics
used in the analyses. 48 percent of respondents are male, with the average age of
all respondents being 37 years. The average years of schooling attained is a little
more than 8. Respondents are split almost equally between rural and urban dwellers.
89 percent are Muslim. 98 percent of respondents profess to be at least somewhat
religious. Only about 37% of all respondents speak the national language, Indonesian,
at home in daily life. 16 percent speak more than one language.
Risk and time preferences are heterogeneous across individuals, although the vast
majority of respondents are categorized as most risk averse and most impatient.
Almost half of the respondents are GA A, but only 9% are GL B. Incidence of negative
time discounting is low (not more than 2%) in both Time submodules. Nonresponse
12
Bryan et al. (2010) have dened a commitment device as \an arrangement entered into by an
agent which restricts his or her future choice set by making certain choices more expensive, perhaps
innitely expensive, while also satisfying two conditions: (a) the agent would, on the margin, pay
something in the present to make those choices more expensive, even if he or she received no other
benet for the payment; and (b) the arrangement does not have a strategic purpose with respect to
others" (p. 3).
13
Experimental studies have shown that people prefer improving sequences of outcomes to de-
clining sequences, a nding that implies negative time preference (< 0) in the standard discounted
utility model (for a summary, see Frederick et al., 2002). However, Loewenstein (1987) has shown
that apparently negative time preference in the standard DU model can in fact be made positive if
the DU model is modied to incorporate the discounting of the anticipation of future consumption.
26
Table 2.2: Summary statistics of IFLS-4 respondents
Mean Std Dev Count
Preference parameters:
Risk A = 1 0.26 0.44 16817
Risk A = 2 0.14 0.34 16817
Risk A = 3 0.09 0.29 16817
Risk A = 4 0.50 0.50 16817
Gamble averse A 0.42 0.49 28876
Nonresponse in Risk Set A 0.01 0.08 29054
Risk B = 1 0.06 0.23 26496
Risk B = 2 0.03 0.17 26496
Risk B = 3 0.07 0.26 26496
Risk B = 4 0.84 0.37 26496
Gamble averse B 0.08 0.28 28879
Nonresponse in Risk Set B 0.01 0.08 29054
Time A = 1 0.08 0.27 28487
Time A = 2 0.06 0.24 28487
Time A = 3 0.15 0.35 28487
Time A = 4 0.72 0.45 28487
Negative time discounting A 0.02 0.14 29026
Nonresponse in Time Set A 0.00 0.03 29054
Time B = 1 0.02 0.14 28818
Time B = 2 0.04 0.19 28818
Time B = 3 0.13 0.34 28818
Time B = 4 0.81 0.39 28818
Negative time discounting B 0.01 0.08 29025
Nonresponse in Time Set B 0.00 0.03 29054
Nonstandard in both Risk A and B 0.03 0.17 28918
Nonstandard in both Time A and B 0.00 0.06 29027
Socio-demographics:
Male 0.48 0.50 29054
Age (years) 36.87 15.62 29051
Age 15-24 years 0.24 0.43 29050
Age 25-34 0.27 0.45 29050
Age 35-44 0.20 0.40 29050
Age 45-54 0.14 0.34 29050
Age 55-64 0.08 0.27 29050
Age 65 0.07 0.25 29050
Years of schooling 8.10 4.36 28755
Highest education level: No education 0.07 0.25 28756
Highest education level: Elementary 0.35 0.48 28756
Highest education level: Junior high 0.19 0.39 28756
Highest education level: Senior high/University 0.39 0.49 28756
Monthly household expenditure per capita (Rp million) 0.41 0.46 28915
Rural 0.47 0.50 28886
Muslim 0.89 0.31 29054
Religiosity: Not at all 0.03 0.16 28979
Religiosity: Somewhat 0.19 0.39 28979
Religiosity: Religious 0.73 0.44 28979
Religiosity: Very 0.06 0.24 28979
Cognition:
Word recall (average #correct out of 10) 4.37 1.89 28387
Parental education:
Father's education: None 0.26 0.44 28176
Father's education: Elementary 0.51 0.50 28176
Father's education: Junior High 0.09 0.29 28176
Father's education: Sr High/Univ 0.14 0.34 28176
Mother's education: None 0.36 0.48 28458
Mother's education: Elementary 0.48 0.50 28458
Mother's education: Junior High 0.08 0.27 28458
Mother's education: Sr High/Univ 0.08 0.26 28458
Total Number of Respondents 29054
27
Table 2.3: Cross tabulations of preferences A and B
Panel A: Risk A and Risk B
P
P
P
P
P
P
P P
Risk A
Risk B
1 2 3 4 Total
1 1048 281 590 1491 3410
2 114 137 389 1480 2120
3 74 61 441 923 1499
4 73 99 279 7780 8231
Total 1309 578 1699 11674 15260
Panel B: Time A and Time B
X
X
X
X
X
X
X
X
X
X
Time A
Time B
1 2 3 4 Total
1 352 515 683 707 2257
2 41 286 535 810 1672
3 80 144 1783 2133 4140
4 50 98 722 19460 20330
Total 523 1043 3723 23110 28399
Panel C: Gamble Averse A and Gamble Lov-
ing B
P
P
P
P
P
P
P P
GA A
GL B
0 1 Total
0 15260 1539 16799
1 11202 836 12038
Total 26462 2375 28837
Panel D: NTD A and NTD B
X
X
X
X
X
X
X
X
X
X
NTD A
NTD B
0 1 Total
0 28399 86 28485
1 418 121 539
Total 28817 207 29024
is negligible in all elicitation modules.
Table 2.3 presents cross tabulations of the preference measures across submodules,
for respondents with nonmissing responses in both Submodules A and B. Clearly, the
matrices are not symmetric; there are substantial numbers of respondents in the
o-diagonals.
28
Panel A shows that out of 15,260 standard respondents in both Risk Submodules,
7,780 fall under category 4 (most risk averse) in both A and B. Far fewer respondents
fall under category 1 (least risk averse) for Risk B compared to Risk A. A high per-
centage of respondents (between 44% to 70%) are observed to switch from the three
lower Risk A categories to the highest Risk B category. In other words, respondents
appear to be more risk averse in the higher-payo Risk B. Many studies have reported
increasing risk aversion with higher payos (Binswanger, 1980; Kachelmeier and She-
hata, 1992; Holt and Laury, 2002). Figure 2.3 illustrates the rightward skewing of
the risk aversion distribution going from submodule A to B.
Panel B cross tabulates the Time A and Time B categories. Out of 28,400 standard
respondents in both submodules, nearly 20,000 have the highest time preference in
both A and B. Of the 2,257 respondents who are most patient in Time A, only 352 are
also most patient in Time B. Far fewer adults are most patient in Time B compared
to Time A (523 vs. 2,257), and more adults were least patient in Time B compared
to Time A (23,111 vs. 20,331). Figure 2.4 illustrates the rightward skewing of the
time preference distribution going from submodule A to B. This pattern make sense.
B payos oer lower annualized returns than A payos, so respondents should be less
patient under Time B, which is exactly what we see.
The fact that the B questions were posed after A raises the question of the eect of
sequencing on respondent behavior. Respondents may have approached B questions
dierently from A, perhaps through learning. Panels C and D shed some light on this
question by cross tabulating the nonstandard counts across submodules. In Panel C,
we see that of the 12,236 adults who were GA A, 11,235 of them were not GL B. This
suggests that respondents \improved" on average, going from Risk Submodule A to B.
Panel D cross tabulates NTD A and NTD B. Although only 2 percent of respondents
were NTD in at least one Time submodule, we still see signs of such improvement
29
Figure 2.3: Distribution of risk aversion
among the nonstandard respondents going from A to B; out of 566 NTD A adults,
418 were not NTD B. However, the causal eect of sequencing cannot be assessed
because sequencing of A and B was not randomized.
Learning is not the only potential explanation for the stark decrease in nonstan-
dard responses observed going from Risk Submodule A to B. Recall that the \cor-
rect", dominating option in Submodule A is the uncertain payo, which justies the
characterization of nonstandard respondents as gamble averse. However, the \cor-
rect", dominating option in Risk B is the certain payo. If a preference for certainty
dominates all other considerations in some respondents' decision-making, standard
behavior in Submodule B may have increased simply because the certain payo hap-
pened to also be the \correct" option. Panel C is consistent with this explanation.
Of 12,038 Gamble Averse A respondents, 93% also preferred the certain option in
Submodule B.
30
Figure 2.4: Distribution of time preference
2.3.6 Comparison with Other Large-Scale Surveys
Prior to IFLS-4, nationally-representative data on risk and time preferences had been
limited to high-income, developed countries, an important exception being the 2005
Mexican Family Life Survey (MxFLS) (Rubalcava and Teruel, 2008). A question of
considerable interest is whether populations of less-developed countries have prefer-
ences that exhibit similar within-country patterns as developed-country populations.
This study helps answer this question.
Table 2.4, Panel A summarizes the distributions of elicited risk aversion in large-
scale national surveys. I stress that the distributions and means are not directly
comparable across countries because the elicitation methodologies are dierent. For
instance, we cannot say whether the average Indonesian is more (or less) risk averse
than the average American. The purpose of Table 2.4 is to show that despite diering
methodologies, within-country patterns of risk aversion are quite similar across all
the countries. Most people regardless of country fall in the moderate to highest risk
aversion categories. Few choose the most risk seeking outcome. IFLS-4 risk aversion
measures do not stand out as unusually skewed one way or the other when compared
31
Table 2.4: Preference Distribution in Other National Surveys
Panel A: Risk aversion
Source Survey Method # ordered
categories
Percent lowest
risk aversion
Percent moderate
risk aversion
Percent highest
risk aversion
Barsky et al. (1997) 1992 HRS Hypothetical lottery 4 12.8 22.5 64.6
Hamoudi (2006) 2005 Mexican FLS Hypothetical lottery 5 9.5 55.8 34.7
Spivey (2010) 1979 NLSY Hypothetical lottery 4 25.0 29.0 46.0
Hryshko et al. (2011) 1996 PSID Hypothetical lottery 6 6.6 62.1 31.3
Dohmen et al. (2011) 2004 German SOEP Self-assess on scale 10 1.0 92.0 7.0
This paper: Risk A IFLS-4 Hypothetical lottery 4 26.5 23.2 50.3
This paper: Risk B IFLS-4 Hypothetical lottery 4 5.6 10.3 84.1
Panel B: Discount rate
Source Survey Method # ordered
categories
Percent lowest
discount rate
Percent moderate
discount rate
Percent highest
discount rate
Rubalcava et al. (2009) 2005 MxFLS Sooner, smaller vs. later, larger 5 11.0 36.5 52.5
This paper: Time A IFLS-4 Sooner, smaller vs. later, larger 4 8.0 20.5 71.5
This paper: Time B IFLS-4 Sooner, smaller vs. later, larger 4 2.1 16.9 81.0
32
to the distribution of other surveys which use hypothetical lottery choice.
To my knowledge only the MxFLS and 2006 PSID measure time preference in the
time discounting sense. Courtemanche et al. (2012) report a mean discount factor of
0.59 (standard deviation of 0.25) when the time horizon is one year and 0.28 (standard
deviation of 0.34) when the time horizon is a month among PSID respondents. The
discount factors were calculated assuming linear utility as in (2.3).
In IFLS-4 Time Submodule A (where the time horizon is one year), 72% of re-
spondents have a discount factor falling between 0 and 0.17. In Time Submodule
B (time horizon of 5 years), 81% of respondents have a discount factor somewhere
between 0 and 0.10. IFLS respondents seem more impatient than PSID respondents.
Although IFLS respondents also seem more impatient than MxFLS respondents, in
both surveys, time preference is heavily skewed towards highest degree of impatience
(see Table 2.4, Panel B).
2.4 Results
The main purpose of this paper is to investigate whether individual preference param-
eters vary systematically with demographic characteristics. The empirical strategy is
to estimate linear probability models
14
of the form
y
i
= + X
i
0
+
i
(2.4)
where the dependent variable y
i
is one of Risk A, Risk B, Time A or Time B for
each individual i. Recall that each preference measure is an ordinal variable with
four categories. Because a high percentage of respondents fall under category 4 (most
14
The estimations are robust to nonlinear binary choice models. Appendix 2.B contains logit
estimates; they are similar to the LPM estimates.
33
risk averse) in the Risk module, Risk A and Risk B are converted into binary vari-
ables where category 4 takes the value 1 and other categories are 0 (nonstandard
respondents are dropped, coded as nonstandard, and analyzed separately). Likewise,
in the Time module, the vast majority of respondents fall under category 4 (most
impatient); therefore category 4 time preference is coded as 1 and the three other
categories of lower time preference grouped as 0.
As explained in Section 2.3, some respondents were categorized as nonstandard
based on their choices in the Risk or Time modules. Another objective of this paper
is to shed some light on the characteristics of these nonstandard respondents. To do
this, I estimate (2.4) using as dependent variables the four dummy variables indicating
nonstandard preference: gamble aversion A, gamble loving B, and negative time
discounting A and B.
The vector of explanatory variables, X
i
, consists of individual and household
characteristics. I include the following arguably exogenous variables in all regressions:
sex, age and whether the respondent is a Muslim. Following the literature, I also
include important variables that are potentially endogenous, i.e. highest level of
education attained, religiosity, a dummy for rural residence, and the logarithm of
monthly per capita household expenditure.
15
I also include respondent's word recall
test performance z-score to account for the in
uence of episodic memory which is not
captured by education.
Other researchers have found strong evidence of parental eects on preferences,
including direct intergenerational eects (Dohmen et al., 2012) and parental education
eects (Hryshko et al., 2011). Failure to control for parental characteristics could
15
I use expenditure as a measure of household welfare instead of current income because it
is a better proxy for permanent income. Furthermore, self-reported expenditure likely has less
measurement error than self-reported income, which is notoriously unreliable in developing countries
(Deaton, 1997).
34
therefore be a source of omitted variable bias in estimations of (2.4). I therefore also
control for parents' highest level of education attained.
In addition, I reestimate equation 2.4 rst with community dummies, then with
household dummies. This controls for unobserved heterogeneity common to all in-
dividuals within each IFLS community or household (for example, individuals living
in a community that had just suered a drought could be systematically more risk
averse than individuals in other communities). These xed eect OLS regressions are
of the form
y
ik
= + X
ik
0
+
k
+
i
(2.5)
where
k
is a dummy for each community, k (or household).
In all regressions, I cluster the standard errors at the community level to allow for
arbitrary correlations within communities. Finally, I stress that this is a descriptive
exercise and coecient estimates are not to be interpreted as causal. I do occasionally
use words such as \eect" and \impact" for expository convenience.
2.4.1 Correlates of Risk Aversion
Table 2.5(A) presents estimates of the correlates of risk aversion. Men are signicantly
less risk averse than women. This result agrees with a large literature examining
gender dierences in risk aversion in eld survey data. This pattern is robust to
submodules A and B and to the inclusion of community dummies (columns 2 and 5)
and household dummies (columns 3 and 6).
Age categories are jointly signicant in four of the six specications. In terms of
the individual age categories, the oldest respondents seem to be signicantly more
risk averse relative to the youngest respondents (columns 4 and 5). However, we
35
Table 2.5(A): Correlates of Risk Preference
(1) (2) (3) (4) (5) (6)
Risk A Risk A Risk A Risk B Risk B Risk B
Male -0.083 -0.087 -0.087 -0.061 -0.063 -0.066
(0.008)

(0.008)

(0.015)

(0.005)

(0.005)

(0.008)

Age categories (years):
25-34 -0.009 -0.016 -0.030 -0.001 -0.009 -0.002
(0.012) (0.012) (0.029) (0.008) (0.007) (0.013)
35-44 -0.019 -0.033 -0.024 0.002 -0.011 0.000
(0.014) (0.014)

(0.033) (0.009) (0.008) (0.015)
45-54 -0.003 -0.015 0.018 0.011 0.002 0.013
(0.018) (0.016) (0.037) (0.010) (0.010) (0.018)
55-64 0.015 0.004 0.044 0.031 0.019 0.042
(0.023) (0.020) (0.049) (0.012)

(0.010) (0.021)

65+ 0.009 0.001 0.069 0.040 0.033 0.048
(0.028) (0.024) (0.062) (0.015)

(0.013)

(0.027)
Education level:
Elementary -0.039 -0.039 -0.048 0.026 0.026 -0.002
(0.028) (0.022) (0.057) (0.013)

(0.010)

(0.021)
Junior high -0.031 -0.014 0.000 0.015 0.018 -0.002
(0.031) (0.025) (0.060) (0.014) (0.011) (0.025)
Senior high/university -0.024 -0.004 -0.006 -0.003 0.005 -0.021
(0.032) (0.025) (0.063) (0.015) (0.012) (0.026)
Word recall z-score -0.007 -0.007 0.005 -0.005 -0.004 -0.003
(0.006) (0.006) (0.012) (0.004) (0.004) (0.005)
Muslim 0.113 0.104 0.219 0.026 0.033 -0.030
(0.060) (0.061) (0.214) (0.031) (0.035) (0.075)
Religiosity (0-not, 3-very) 0.060 0.055 0.085 0.017 0.026 0.014
(0.025)

(0.027)

(0.072) (0.014) (0.016) (0.028)
Muslim Religiosity -0.058 -0.043 -0.071 -0.015 -0.017 -0.008
(0.027)

(0.028) (0.074) (0.015) (0.017) (0.030)
Father's education level -0.009 -0.008 0.007 0.001 0.001 0.005
(0.007) (0.006) (0.015) (0.004) (0.004) (0.007)
Mother's education level 0.001 -0.010 -0.010 -0.009 -0.014 -0.010
(0.008) (0.007) (0.016) (0.005) (0.004)

(0.008)
Log PCE -0.025 -0.036 -0.014 -0.023
(0.008)

(0.007)

(0.005)

(0.004)

Rural -0.055 0.002 0.026 -0.025 0.005 0.134
(0.020)

(0.022) (0.147) (0.011)

(0.013) (0.068)

Constant 0.811 0.927 0.323 1.017 1.099 0.829
(0.126)

(0.099)

(0.214) (0.072)

(0.065)

(0.082)

Observations 15644 15644 15644 24446 24446 24446
R
2
0.013 0.148 0.680 0.015 0.101 0.574
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.500 0.500 0.500 0.840 0.840 0.840
Joint signicance of (p-value):
Age categories 0.410 0.058 0.253 0.025 0.002 0.118
Education levels 0.429 0.013 0.296 0.003 0.002 0.424
Parents' education 0.358 0.027 0.804 0.135 0.001 0.440
Dependent variable Risk = 1 if highest risk aversion (category 4), = 0 if less risk averse (categories 1{
3). All regressions are OLS. Omitted category for age: 15-24, for education: no education. Standard
errors in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
cannot rule out cohort eects.
Household log per capita expenditure is signicantly negatively correlated with
risk aversion, even after controlling for community xed eects, suggesting that risk
36
aversion decreases with wealth. Education levels however are jointly insignicant in
four of the six columns. Word recall is not signicantly correlated with risk aversion.
2.4.2 Correlates of Time Preference
The correlates of time preference are shown in Table 2.5(B). The coecient on male
is negative but is signicant only in three of the six columns.
16
Age categories are
always jointly signicant; older respondents are signicantly more likely to be most
impatient than the youngest respondents with the eect appearing to be monotonic
over age categories. Log PCE is signicantly negatively related with time preference,
suggesting impatience decreases with wealth.
Education levels are also always jointly signicant. Relative to no formal school-
ing, education level eects are monotonically decreasing (although only the senior
high/university category is signicantly negative). This suggests that impatience de-
creases with more education. No claim is made on the direction of causality, although
in Becker and Mulligan's (1997) model of patience formation, schooling teaches chil-
dren to value the future more and thus become more patient.
Word recall is signicantly correlated with time preference. In Table 2.5(B),
columns 1 and 4, a one standard deviation increase in the word recall test score
is signicantly associated with a 1.6 percentage point decrease in the likelihood of be-
ing most impatient for Time A and a 1.2 percentage point decrease for Time B (this
eect is still signicant after controlling for community xed eects). People with
better episodic memory appear to be more patient, even controlling for education.
16
Compared to the literature on risk aversion, there is not much evidence on how time preference
varies by gender, nor a generally accepted prior. Two separate studies in Vietnam have found no
signicant gender dierences in the discount rate (Anderson et al., 2004; Tanaka et al., 2010). van
Praag and Booij (2003) have found that men are more patient than women, though their sample of
newspaper readers in the Netherlands was not representative of the population.
37
Table 2.5(B): Correlates of Time Preference
(1) (2) (3) (4) (5) (6)
Time A Time A Time A Time B Time B Time B
Male -0.015 -0.014 -0.012 -0.013 -0.012 -0.012
(0.006)

(0.006)

(0.009) (0.005)

(0.006)

(0.008)
Age categories (years):
25-34 0.047 0.048 0.045 0.025 0.024 0.013
(0.008)

(0.009)

(0.015)

(0.007)

(0.007)

(0.014)
35-44 0.043 0.045 0.070 0.036 0.032 0.039
(0.010)

(0.009)

(0.017)

(0.008)

(0.008)

(0.015)

45-54 0.046 0.051 0.072 0.045 0.044 0.053
(0.012)

(0.011)

(0.019)

(0.010)

(0.010)

(0.017)

55-64 0.069 0.077 0.100 0.074 0.073 0.088
(0.013)

(0.013)

(0.024)

(0.011)

(0.011)

(0.020)

65+ 0.092 0.107 0.112 0.081 0.084 0.087
(0.015)

(0.015)

(0.030)

(0.013)

(0.013)

(0.026)

Education level:
Elementary -0.001 -0.002 -0.028 0.003 -0.002 -0.014
(0.014) (0.015) (0.024) (0.011) (0.012) (0.019)
Junior high -0.002 -0.003 -0.041 0.007 0.004 -0.008
(0.016) (0.017) (0.029) (0.013) (0.014) (0.021)
Senior high/university -0.042 -0.037 -0.067 -0.025 -0.026 -0.035
(0.016)

(0.017)

(0.029)

(0.014) (0.015) (0.022)
Word recall z-score -0.016 -0.014 -0.007 -0.011 -0.011 -0.008
(0.004)

(0.004)

(0.007) (0.004)

(0.003)

(0.006)
Muslim -0.036 0.035 0.005 -0.049 -0.003 -0.022
(0.040) (0.035) (0.100) (0.033) (0.031) (0.086)
Religiosity (0-not, 3-very) -0.037 -0.002 0.003 -0.041 -0.003 -0.005
(0.021) (0.017) (0.031) (0.018)

(0.014) (0.025)
Muslim Religiosity 0.021 -0.007 -0.004 0.036 0.003 0.010
(0.022) (0.017) (0.033) (0.018)

(0.014) (0.027)
Father's education level -0.011 -0.010 -0.006 -0.004 -0.005 -0.004
(0.005)

(0.005)

(0.008) (0.004) (0.004) (0.007)
Mother's education level -0.008 -0.011 -0.007 -0.004 -0.007 -0.008
(0.006) (0.005)

(0.010) (0.005) (0.004) (0.008)
Log PCE -0.018 -0.020 -0.017 -0.017
(0.005)

(0.005)

(0.004)

(0.004)

Rural 0.012 0.013 -0.003 0.005 0.011 -0.000
(0.011) (0.016) (0.068) (0.010) (0.013) (0.054)
Constant 0.998 0.939 0.720 1.065 1.014 0.826
(0.074)

(0.072)

(0.104)

(0.059)

(0.059)

(0.089)

Observations 26254 26254 26254 26553 26553 26553
R
2
0.021 0.080 0.554 0.016 0.073 0.548
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.710 0.710 0.710 0.810 0.810 0.810
Joint signicance of (p-value):
Age categories 0.000 0.000 0.000 0.000 0.000 0.001
Education levels 0.000 0.000 0.060 0.000 0.001 0.198
Parents' education 0.001 0.000 0.356 0.291 0.028 0.343
Dependent variable Time = 1 if highest impatience (category 4), = 0 if less impatient (categories 1{3).
All regressions are OLS. Omitted category for age: 15-24, for education: no education. Standard errors
in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
38
2.4.3 Correlates of Gamble Aversion/Loving
Table 2.5(C) reveals that gamble aversion A (GA A) is signicantly correlated with
gender, age, and education (columns 1{3). Men are less likely to be GA A compared
to women. Age categories are jointly signicant in the rst three columns. Older
respondents are more likely than the youngest respondents to be GA A, and this eect
is almost monotonically increasing in age. Education levels are jointly signicant in all
three GA A models. Respondents with more education are less likely than those with
no formal education to be GA A. Log PCE and rural residence are not signicantly
correlated with GA A.
As can be seen in the last three columns, GL B is signicantly correlated only with
gender and word recall. This is unsurprising because only 8 percent of respondents in
the GL B analytical samples were gamble averse, with the vast majority exhibiting
standard behavior.
Word recall is robustly signicantly negatively related with both GA A and GL B.
This suggests that people with better episodic memory are more likely to pick the
\correct", dominating payo in the Risk lter questions, consistent with standard
expected utility theory. Episodic memory also seems to matter more than educa-
tional attainment; although each education category has a negative sign, they are all
individually insignicant and jointly signicant only in the GA A regressions.
2.4.4 Correlates of Negative Time Discounting
Table 2.5(D) present the correlates of negative time discounting. The estimates are
broadly similar across columns. Males are signicantly more likely to be NTD A/B.
However, signicance disappears when the household xed eect is controlled for. Age
categories are jointly signicant in the NTD A regressions, with older respondents be-
39
Table 2.5(C): Correlates of Gamble Aversion/Loving
(1) (2) (3) (4) (5) (6)
GA A GA A GA A GL B GL B GL B
Male -0.084 -0.085 -0.091 0.011 0.013 0.012
(0.007)

(0.007)

(0.011)

(0.004)

(0.004)

(0.006)

Age categories (years):
25-34 -0.011 -0.019 -0.005 -0.002 0.002 -0.005
(0.009) (0.008)

(0.016) (0.005) (0.005) (0.010)
35-44 -0.012 -0.011 0.011 0.001 0.007 0.007
(0.009) (0.009) (0.018) (0.006) (0.006) (0.010)
45-54 0.007 0.005 0.029 -0.004 -0.001 0.002
(0.012) (0.012) (0.022) (0.007) (0.007) (0.012)
55-64 -0.002 0.002 0.033 -0.009 -0.006 -0.003
(0.015) (0.013) (0.027) (0.008) (0.008) (0.015)
65+ 0.049 0.068 0.076 -0.002 -0.003 -0.011
(0.019)

(0.018)

(0.034)

(0.010) (0.010) (0.019)
Education level:
Elementary 0.031 0.031 0.015 0.002 -0.005 -0.006
(0.017) (0.015)

(0.029) (0.010) (0.010) (0.018)
Junior high 0.015 0.010 -0.002 0.001 -0.009 -0.020
(0.019) (0.017) (0.031) (0.011) (0.010) (0.020)
Senior high/university -0.044 -0.043 -0.045 -0.002 -0.013 -0.022
(0.020)

(0.018)

(0.034) (0.012) (0.011) (0.021)
Word recall z-score -0.032 -0.020 -0.015 -0.010 -0.010 -0.003
(0.005)

(0.004)

(0.007)

(0.002)

(0.002)

(0.004)
Muslim 0.027 0.036 -0.027 -0.013 -0.004 -0.036
(0.041) (0.040) (0.107) (0.025) (0.024) (0.066)
Religiosity (0-not, 3-very) 0.005 0.010 0.007 -0.002 0.001 -0.005
(0.018) (0.019) (0.031) (0.012) (0.011) (0.021)
Muslim Religiosity -0.035 -0.013 0.001 -0.001 -0.001 0.009
(0.020) (0.020) (0.034) (0.012) (0.012) (0.022)
Father's education level -0.017 -0.016 -0.009 0.003 0.002 0.003
(0.005)

(0.005)

(0.009) (0.003) (0.003) (0.005)
Mother's education level -0.001 -0.007 -0.007 0.001 0.005 0.004
(0.006) (0.005) (0.010) (0.003) (0.003) (0.006)
Log PCE -0.005 -0.019 -0.001 0.003
(0.006) (0.005)

(0.003) (0.003)
Rural 0.016 -0.001 -0.021 0.009 0.004 -0.024
(0.014) (0.017) (0.080) (0.007) (0.008) (0.041)
Constant 0.557 0.693 0.494 0.094 0.045 0.122
(0.088)

(0.072)

(0.109)

(0.049) (0.043) (0.067)
Observations 26646 26646 26646 26651 26651 26651
R
2
0.032 0.112 0.563 0.002 0.058 0.520
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.410 0.410 0.410 0.080 0.080 0.080
Joint signicance of (p-value):
Age categories 0.005 0.000 0.206 0.870 0.570 0.887
Education levels 0.000 0.000 0.003 0.907 0.466 0.415
Parents' education 0.000 0.000 0.220 0.331 0.108 0.435
Dependent variable GA = 1 if gamble averse, = 0 if standard (risk aversion categories 1{4). All
regressions are OLS. Omitted category for age: 15-24, for education: no education. Standard errors
in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
ing signicantly less likely to pick the nonstandard answer than the youngest respon-
dents. Interestingly, religiosity plays a signicant role in the likelihood of NTD A/B
for non-Muslims but not for Muslims. Religious non-Muslims are more likely to be
NTD A/B than their less religious counterparts, but a similar eect is not observed
40
Table 2.5(D): Correlates of Negative Time Discounting
(1) (2) (3) (4) (5) (6)
NTD A NTD A NTD A NTD B NTD B NTD B
Male 0.004 0.004 0.005 0.003 0.003 0.001
(0.002)

(0.002)

(0.003) (0.001)

(0.001)

(0.002)
Age categories (years):
25-34 -0.011 -0.009 -0.009 -0.004 -0.003 -0.004
(0.002)

(0.002)

(0.005) (0.001)

(0.002) (0.003)
35-44 -0.010 -0.010 -0.015 -0.002 -0.001 -0.002
(0.003)

(0.003)

(0.005)

(0.002) (0.002) (0.003)
45-54 -0.013 -0.013 -0.018 -0.003 -0.003 -0.004
(0.003)

(0.003)

(0.006)

(0.002) (0.002) (0.004)
55-64 -0.019 -0.020 -0.023 -0.006 -0.006 -0.005
(0.004)

(0.004)

(0.007)

(0.003)

(0.003)

(0.005)
65+ -0.015 -0.017 -0.019 -0.004 -0.005 -0.002
(0.005)

(0.005)

(0.009)

(0.003) (0.004) (0.007)
Education level:
Elementary -0.000 -0.004 -0.008 -0.003 -0.006 -0.002
(0.005) (0.005) (0.010) (0.003) (0.003) (0.006)
Junior high -0.007 -0.010 -0.015 -0.006 -0.008 -0.005
(0.005) (0.006) (0.011) (0.003) (0.004)

(0.006)
Senior high/university -0.006 -0.009 -0.014 -0.005 -0.008 -0.004
(0.005) (0.005) (0.011) (0.003) (0.004)

(0.007)
Word recall z-score -0.005 -0.003 -0.001 -0.002 -0.001 -0.001
(0.001)

(0.001)

(0.002) (0.001)

(0.001)

(0.001)
Muslim 0.020 0.021 0.063 0.021 0.020 0.017
(0.010) (0.012) (0.035) (0.009)

(0.009)

(0.027)
Religiosity (0-not, 3-very) 0.015 0.010 0.021 0.014 0.012 0.010
(0.005)

(0.006) (0.015) (0.005)

(0.005)

(0.009)
Muslim Religiosity -0.016 -0.010 -0.019 -0.014 -0.011 -0.008
(0.006)

(0.006) (0.016) (0.005)

(0.005)

(0.009)
Father's education level 0.001 -0.000 -0.001 -0.001 -0.001 -0.002
(0.001) (0.001) (0.002) (0.001) (0.001) (0.002)
Mother's education level 0.003 0.003 0.003 0.002 0.001 0.002
(0.002)

(0.002)

(0.003) (0.001) (0.001) (0.002)
Log PCE 0.001 0.000 -0.000 -0.000
(0.001) (0.001) (0.001) (0.001)
Rural -0.003 0.002 -0.007 -0.001 -0.000 -0.000
(0.002) (0.004) (0.014) (0.001) (0.003) (0.001)
Constant -0.002 0.005 -0.024 -0.010 -0.005 -0.007
(0.019) (0.021) (0.035) (0.013) (0.013) (0.027)
Observations 26741 26741 26741 26740 26740 26740
R
2
0.004 0.031 0.474 0.003 0.030 0.480
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.020 0.020 0.020 0.010 0.010 0.010
Joint signicance of (p-value):
Age categories 0.000 0.000 0.021 0.149 0.181 0.760
Education levels 0.068 0.113 0.419 0.231 0.161 0.770
Parents' education 0.023 0.068 0.511 0.163 0.200 0.286
Dependent variable NTD = 1 if negative time discount, 0 if standard (time pref. categories 1{4).
All regressions are OLS. Omitted category for age: 15-24, for education: no education. Standard
errors in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
among Muslims.
Just as for gamble aversion/loving, the role of word recall in the NTD regressions
is strong; word recall is signicantly negatively related to NTD A and NTD B, even
41
after controlling for community xed eects.
2.4.5 Sex Dierences in the Eect of Cognition
In a sample of nearly 3,500 U.S. undergraduates, Frederick (2005) nds evidence of
sex dierences in the eect of cognition on preference. Cognition was measured by
what he calls cognitive re
ection tests.
17
Cognitive re
ection is the ability to resist
reporting the response that rst comes to mind. Better-performing men on this test
are more risk tolerant than other men, whereas better-performing women are more
patient than other women.
Tables 2.6(A) and 2.6(B) report estimates of the cognition eect by sex for the
IFLS sample. In Table 2.6(A), men with better word recall are less risk averse in both
A and B, although the result is signicant only in A (columns 1 and 3). Women's
word recall is not signicantly associated with their risk aversion (columns 2 and 4),
consistent with Frederick's nding of no signicant association between cognitive
capacity and risk aversion among women.
Women with better word recall are more patient than other women, but the eect
is signicant only for Time A (Table 2.6(A), column 6). I nd that men with better
word recall are also more patient, signicant at 1% across both submodules. Unlike
the men in Frederick's sample, Indonesian men with better cognitive capacity are
both more risk tolerant and patient than men with poorer cognition.
Table 2.6(B) presents estimated eects on nonstandard preferences by sex. The
signicant, negative eect of word recall on likelihood of nonstandard found in the
main results is still very evident when the sample is split across gender. Other ex-
17
One of the questions asked was: \A bat and a ball cost $1.10. The bat costs $1.00 more than
the ball. How much does the ball cost? cents" (Frederick, 2005, p. 26). An impulsive person
would give the incorrect answer that intuitively springs to mind, 10 cents. Those who re
ect a little
more carefully almost always give the correct answer, 5 cents.
42
Table 2.6(A): Correlates of Risk and Time Preferences by Sex
(1) (2) (3) (4) (5) (6) (7) (8)
Risk A
Men
Risk A
Women
Risk B
Men
Risk B
Women
Time A
Men
Time A
Women
Time B
Men
Time B
Women
Age categories (years):
25-34 -0.038 0.023 -0.023 0.020 0.056 0.039 0.027 0.024
(0.018)

(0.016) (0.012)

(0.009)

(0.012)

(0.011)

(0.011)

(0.009)

35-44 -0.033 -0.003 -0.020 0.024 0.054 0.035 0.039 0.036
(0.019) (0.019) (0.013) (0.010)

(0.013)

(0.013)

(0.012)

(0.011)

45-54 -0.010 0.005 -0.003 0.026 0.044 0.051 0.044 0.049
(0.023) (0.024) (0.015) (0.012)

(0.015)

(0.015)

(0.014)

(0.013)

55-64 0.002 0.029 0.017 0.044 0.084 0.056 0.070 0.082
(0.028) (0.032) (0.018) (0.015)

(0.017)

(0.018)

(0.016)

(0.015)

65+ -0.012 0.025 0.021 0.050 0.092 0.093 0.064 0.100
(0.033) (0.040) (0.018) (0.018)

(0.021)

(0.019)

(0.018)

(0.017)

Education level:
Elementary -0.072 -0.040 0.012 0.019 0.016 -0.010 -0.024 0.014
(0.039) (0.032) (0.022) (0.014) (0.025) (0.016) (0.017) (0.015)
Junior high -0.085 -0.009 -0.015 0.022 0.001 0.003 -0.025 0.025
(0.041)

(0.038) (0.024) (0.016) (0.026) (0.019) (0.018) (0.018)
Senior high/university -0.070 -0.011 -0.037 0.008 -0.041 -0.035 -0.059 -0.005
(0.041) (0.040) (0.024) (0.017) (0.026) (0.022) (0.019)

(0.019)
Word recall z-score -0.017 0.002 -0.010 -0.001 -0.021 -0.011 -0.019 -0.005
(0.008)

(0.008) (0.005) (0.004) (0.006)

(0.005)

(0.005)

(0.005)
Muslim 0.091 0.145 0.004 0.049 -0.033 -0.040 -0.035 -0.068
(0.073) (0.082) (0.045) (0.041) (0.043) (0.060) (0.042) (0.043)
Religiosity (0-not, 3-very) 0.063 0.059 0.018 0.014 -0.032 -0.042 -0.038 -0.045
(0.032) (0.035) (0.020) (0.019) (0.023) (0.029) (0.021) (0.021)

Muslim Religiosity -0.058 -0.063 -0.006 -0.024 0.019 0.023 0.031 0.044
(0.035) (0.037) (0.022) (0.020) (0.024) (0.030) (0.022) (0.022)

Father's education level -0.009 -0.010 0.003 -0.002 -0.012 -0.010 -0.001 -0.007
(0.009) (0.009) (0.006) (0.005) (0.007) (0.007) (0.006) (0.006)
Mother's education level -0.003 0.005 -0.017 -0.003 0.003 -0.018 -0.001 -0.008
(0.011) (0.011) (0.007)

(0.006) (0.008) (0.008)

(0.007) (0.006)
Log PCE -0.025 -0.025 -0.020 -0.009 -0.016 -0.019 -0.017 -0.016
(0.010)

(0.010)

(0.006)

(0.006) (0.007)

(0.006)

(0.006)

(0.005)

Rural -0.052 -0.059 -0.027 -0.024 0.009 0.014 0.008 0.003
(0.021)

(0.023)

(0.013)

(0.012)

(0.014) (0.012) (0.012) (0.011)
Constant 0.802 0.774 1.077 0.933 0.936 1.036 1.073 1.061
(0.145)

(0.155)

(0.092)

(0.084)

(0.097)

(0.093)

(0.081)

(0.075)

Observations 8056 7588 11604 12842 12504 13750 12654 13899
R
2
0.009 0.006 0.014 0.004 0.021 0.021 0.017 0.016
Community FE No No No No No No No No
Household FE No No No No No No No No
Mean of dependent variable 0.460 0.550 0.810 0.870 0.700 0.720 0.800 0.810
Joint signicance of (p-value):
Age categories 0.190 0.567 0.024 0.038 0.000 0.000 0.001 0.000
Education levels 0.213 0.188 0.002 0.174 0.000 0.036 0.002 0.015
Parents' education 0.467 0.569 0.024 0.717 0.141 0.000 0.973 0.061
Dependent variable Time = 1 if highest impatience (category 4), = 0 if less impatient (categories 1{3). All
regressions are OLS. Omitted category for age: 15-24, for education: no education. Standard errors in parentheses
clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
planatory variables exhibit largely similar patterns across gender. However, pooling
tests for inter-gender dierences in the estimated coecients reveal that the pooling
hypothesis is rejected for each specication. This suggests that men and women in
this sample should be studied separately in future work.
2.5 Conclusion
The 2007 Indonesian Family Life Survey (IFLS-4) has directly elicited measures of risk
aversion and time preference for a large, nationally-representative sample of adults.
43
Table 2.6(B): Correlates of Nonstandard Preferences by Sex
(1) (2) (3) (4) (5) (6) (7) (8)
GA A
Men
GA A
Women
GL B
Men
GL B
Women
NTD A
Men
NTD A
Women
NTD B
Men
NTD B
Women
Age categories (years):
25-34 -0.040 0.015 0.010 -0.012 -0.013 -0.008 -0.005 -0.002
(0.012)

(0.012) (0.008) (0.006) (0.004)

(0.003)

(0.002)

(0.002)
35-44 -0.033 0.009 0.016 -0.013 -0.009 -0.011 -0.002 -0.000
(0.014)

(0.013) (0.009) (0.007) (0.004)

(0.003)

(0.003) (0.002)
45-54 -0.022 0.035 0.005 -0.014 -0.013 -0.011 -0.001 -0.005
(0.017) (0.017)

(0.010) (0.009) (0.005)

(0.004)

(0.003) (0.002)

55-64 -0.018 0.014 0.008 -0.025 -0.023 -0.015 -0.006 -0.005
(0.020) (0.021) (0.011) (0.011)

(0.005)

(0.005)

(0.004) (0.003)
65+ 0.025 0.073 -0.005 0.002 -0.011 -0.020 -0.003 -0.005
(0.024) (0.027)

(0.014) (0.015) (0.007) (0.006)

(0.005) (0.004)
Education level:
Elementary 0.039 0.029 -0.005 0.009 -0.002 -0.001 -0.002 -0.003
(0.025) (0.020) (0.017) (0.012) (0.009) (0.006) (0.007) (0.003)
Junior high 0.028 0.007 0.003 0.001 -0.009 -0.006 -0.005 -0.007
(0.027) (0.025) (0.019) (0.012) (0.010) (0.006) (0.007) (0.004)
Senior high/university -0.039 -0.042 -0.005 0.001 -0.010 -0.004 -0.005 -0.005
(0.028) (0.026) (0.019) (0.013) (0.010) (0.006) (0.007) (0.004)
Word recall z-score -0.037 -0.026 -0.009 -0.012 -0.006 -0.003 -0.003 -0.001
(0.006)

(0.006)

(0.004)

(0.003)

(0.002)

(0.001)

(0.001)

(0.001)
Muslim 0.024 0.029 -0.037 0.020 0.006 0.040 0.008 0.039
(0.051) (0.061) (0.032) (0.039) (0.015) (0.015)

(0.010) (0.014)

Religiosity (0-not, 3-very) 0.026 -0.019 -0.014 0.013 0.010 0.022 0.011 0.019
(0.022) (0.028) (0.015) (0.018) (0.008) (0.007)

(0.006)

(0.007)

Muslim Religiosity -0.037 -0.033 0.016 -0.023 -0.008 -0.027 -0.008 -0.022
(0.024) (0.030) (0.016) (0.019) (0.008) (0.008)

(0.006) (0.007)

Father's education level -0.022 -0.014 0.003 0.003 0.000 0.001 -0.000 -0.001
(0.007)

(0.007)

(0.004) (0.004) (0.002) (0.002) (0.001) (0.001)
Mother's education level 0.011 -0.013 0.003 0.000 0.007 -0.000 0.003 0.000
(0.007) (0.008) (0.005) (0.004) (0.002)

(0.002) (0.001)

(0.001)
Log PCE -0.008 -0.001 0.001 -0.002 -0.002 0.003 -0.000 0.000
(0.007) (0.008) (0.004) (0.004) (0.002) (0.002) (0.001) (0.001)
Rural 0.018 0.015 0.011 0.007 -0.006 -0.001 -0.001 -0.001
(0.014) (0.017) (0.008) (0.008) (0.003)

(0.003) (0.002) (0.001)
Constant 0.492 0.537 0.095 0.096 0.043 -0.042 0.002 -0.021
(0.100)

(0.115)

(0.068) (0.061) (0.028) (0.028) (0.017) (0.019)
Observations 12734 13912 12727 13924 12757 13984 12756 13984
R
2
0.026 0.025 0.002 0.004 0.006 0.004 0.004 0.003
Community FE No No No No No No No No
Household FE No No No No No No No No
Mean of dependent variable 0.370 0.450 0.090 0.080 0.020 0.020 0.010 0.010
Joint signicance of (p-value):
Age categories 0.002 0.059 0.374 0.112 0.001 0.002 0.167 0.238
Education levels 0.000 0.000 0.758 0.618 0.143 0.539 0.561 0.316
Parents' education 0.006 0.001 0.343 0.592 0.003 0.916 0.079 0.766
GA = 1 if gamble averse, 0 otherwise. All regressions are OLS. Omitted category for age: 15-24, for education:
no education. Standard errors in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
Coupled with data on a rich set of characteristics of respondents and interviewers,
IFLS-4 permits the precise estimation of demographic, cognitive and interviewer cor-
relates of elicited risk and time preferences.
I have found that although most survey respondents are highly risk averse and
highly impatient, there is considerable heterogeneity in these preferences. Further-
more, these preferences are systematically related to respondent characteristics. The
main results are: (i) men are less risk averse than women; (ii) older adults are more
impatient; (iii) wealthier adults are less risk averse and less impatient; (iv) adults
who perform better in a word recall test, a measure of cognition proxied by episodic
44
memory, are less impatient; (v) better educated respondents are less impatient; and,
(vi) those with better word recall are more likely to exhibit \standard" preferences
(which are consistent with standard discounted expected utility theory).
45
Appendix 2.A Selected Explanatory Variables
Education Levels
There are ve education level dummies corresponding to no education, elementary,
junior high, and senior high/university. These categories refer to the highest level of
education attained, not necessarily completed.
Household Expenditures Per Capita
Household expenditures consist of spending on education, food, durable nonfood
items, and nondurable nonfood items, as reported by the spouse of the household head
or another person most knowledgeable about household aairs. To obtain monthly
measures of each component expenditure, I divide annual education and durable non-
food expenses by 12, and multiply weekly food expenses by 30/7. Monthly household
expenditures is the sum of the monthly component expenditures.
Word Recall Test Score
Respondents were read a list of ten nouns and then asked to repeat as many words as
they can recall, in any order. 12 to 15 minutes later, after other, unrelated questions
were asked, respondents were again asked to repeat the words. Four lists of nouns
were used, which were randomized across individuals within the household, so that
one person could not learn from another's experience. Word recall test score = the
average number of correctly recalled words over both attempts.
46
Appendix 2.B Logit Estimates
Table 2A.1: Logit Estimates of Tables 2.5(A) { 2.5(D)
Risk A Risk B Time A Time B GA A GL B NTD A NTD B
Male (d) -0.084 -0.061 -0.015 -0.013 -0.085 0.011 0.004 0.002
(0.008)

(0.005)

(0.006)

(0.005)

(0.007)

(0.004)

(0.002)

(0.001)

Age categories (years):
25-34 (d) -0.009 -0.001 0.043 0.021 -0.012 -0.002 -0.008 -0.003
(0.013) (0.007) (0.008)

(0.006)

(0.009) (0.005) (0.002)

(0.001)

35-44 (d) -0.019 0.002 0.039 0.031 -0.012 0.001 -0.007 -0.001
(0.014) (0.008) (0.009)

(0.007)

(0.010) (0.006) (0.002)

(0.001)
45-54 (d) -0.003 0.011 0.042 0.040 0.007 -0.004 -0.009 -0.002
(0.018) (0.010) (0.011)

(0.009)

(0.012) (0.006) (0.002)

(0.001)
55-64 (d) 0.016 0.032 0.065 0.069 -0.002 -0.008 -0.012 -0.004
(0.023) (0.012)

(0.012)

(0.010)

(0.015) (0.008) (0.002)

(0.001)

65+ (d) 0.009 0.042 0.092 0.078 0.048 -0.002 -0.010 -0.003
(0.028) (0.013)

(0.014)

(0.012)

(0.020)

(0.009) (0.002)

(0.002)
Education level:
Elementary (d) -0.040 0.026 -0.009 -0.004 0.034 0.002 -0.000 -0.002
(0.029) (0.014) (0.018) (0.015) (0.017)

(0.010) (0.004) (0.002)
Junior high (d) -0.031 0.014 -0.012 -0.002 0.018 0.001 -0.005 -0.004
(0.032) (0.015) (0.019) (0.017) (0.019) (0.011) (0.004) (0.002)

Senior high/university (d) -0.025 -0.003 -0.050 -0.032 -0.043 -0.002 -0.005 -0.004
(0.033) (0.017) (0.020)

(0.017) (0.020)

(0.011) (0.005) (0.002)
Word recall z-score -0.007 -0.005 -0.016 -0.012 -0.032 -0.010 -0.004 -0.002
(0.006) (0.004) (0.004)

(0.004)

(0.005)

(0.002)

(0.001)

(0.001)

Muslim (d) 0.115 0.024 -0.034 -0.042 0.027 -0.012 0.010 0.006
(0.060) (0.030) (0.037) (0.027) (0.042) (0.024) (0.004)

(0.002)

Religiosity (0-not, 3-very) 0.062 0.015 -0.036 -0.038 0.005 -0.002 0.010 0.007
(0.026)

(0.013) (0.021) (0.016)

(0.019) (0.010) (0.003)

(0.002)

Muslim Religiosity -0.060 -0.014 0.020 0.033 -0.036 -0.002 -0.011 -0.007
(0.028)

(0.014) (0.021) (0.016)

(0.020) (0.011) (0.004)

(0.002)

Father's education level -0.009 0.001 -0.010 -0.003 -0.018 0.003 0.000 -0.001
(0.007) (0.004) (0.005)

(0.004) (0.005)

(0.003) (0.001) (0.001)
Mother's education level 0.001 -0.008 -0.007 -0.003 -0.002 0.001 0.003 0.001
(0.008) (0.005) (0.006) (0.004) (0.006) (0.003) (0.001)

(0.001)
Log PCE -0.025 -0.014 -0.018 -0.016 -0.005 -0.000 0.001 -0.000
(0.009)

(0.005)

(0.005)

(0.004)

(0.006) (0.003) (0.001) (0.001)
Rural (d) -0.056 -0.025 0.012 0.005 0.016 0.009 -0.003 -0.001
(0.020)

(0.011)

(0.011) (0.010) (0.014) (0.007) (0.002) (0.001)
Observations 15644 24446 26254 26553 26646 26651 26741 26740
Pseudo R
2
0.010 0.017 0.017 0.017 0.024 0.004 0.019 0.028
Community FE No No No No No No No No
Household FE No No No No No No No No
Mean of dependent variable 0.500 0.840 0.710 0.810 0.410 0.080 0.020 0.010
Joint signicance of (p-value):
Age categories 0.406 0.018 0.000 0.000 0.005 0.873 0.000 0.134
Education levels 0.430 0.003 0.000 0.000 0.000 0.903 0.081 0.202
Parents' education 0.357 0.137 0.001 0.382 0.000 0.336 0.020 0.139
All regressions are logit. Marginal eects reported. (d) indicates dummy variable. Omitted category for age:
15-24, for education: no education. Constant term not reported. Standard errors in parentheses clustered by
community ID. *p<0.10 **p<0.05 ***p<0.01.
47
Chapter 3
Interviewer Eects on Risk and Time Preferences
3.1 Introduction
This essay is motivated by Binswanger's (1980) nding of interviewer bias in survey-
elicited risk preference. The main objective is to investigate the role of the interviewer
in responses to the elicitation modules. My approach is to reestimate the models
of Chapter 2 by introducing interviewer characteristics into the specications. My
purpose is twofold. First, I want to investigate whether not controlling for interviewer
eects introduces omitted variable bias. If the base-specication estimates change
substantially with the inclusion of interviewer variables, then interviewer variables
are correlated with X and estimates of are biased. We would expect to not nd
evidence of interviewer omitted variable bias, as a correlation between interviewer
and respondent characteristics should be exceedingly unlikely in a careful survey
design. Second, the direct eect of interviewer characteristics on respondent behavior
is interesting in and of itself.
3.2 Literature Review
Interviewer eects in surveys have been studied quite widely in other social sciences,
though recently it has gotten increased attention in economics. Why would inter-
viewers aect responses if they were trained to follow a common procedure? The
political science literature has oered one explanation: the \social distance" (that is,
48
dierences in social class, sex, race and age) between interviewer and respondent can
aect responses (Dohrenwend and Colombotos, 1968).
1
This would naturally be a
factor in subjective environments, such as when soliciting an opinion.
Whether interviewer eects would be a factor in objective procedures such as
lottery-choice questions is a priori less clear. Kemsley (1965) has shown that inter-
viewer eects are present in expenditure surveys in the British General Social Survey,
an ostensibly objective endeavor. Why would this be so? One possible explanation
is social distance, in that respondents may try to present a particular image of them-
selves depending on the perceived social distance between them and their interviewer.
Another possible explanation is interviewer idiosyncrasy: although they may follow
the same procedures, interviewers may still perform idiosyncratically during the in-
terview, for example in tone of voice, thus aecting responses in unknown ways. In
other words, some subjective element may be present in the interview process.
I am aware of one paper to have investigated interviewer eects in risk preference
elicitation. Binswanger (1980) nds evidence of severe interviewer bias in his classic
study on risk aversion among rural Indian farmers. He assigned interviewer A to the
village of Shirapur, and interviewer B to the neighboring village of Kalman. The
interviewers elicited certainty equivalents from their subjects based on hypothetical
income streams, upon which they classied respondents into ve risk aversion cate-
gories. The villages were then resurveyed, switching investigators. In each village,
investigator B classied respondents as more risk averse than investigator A, and
chi-square tests showed the dierences were signicant.
Other economists have looked at interviewer eects on longitudinal survey attri-
1
Social distance has not been ignored in economics. Akerlof (1997) has developed a generalized
model of social distance within the context of social decisions, such as the demand for education,
the decision to marry and the decision to discriminate. His model explains the existence of social
class and linguistic dialect with particular relevance to the United States.
49
tion (e.g. Thomas et al., 2012 for IFLS), item nonresponse (Riphahn and Ser
ing,
2005; Blom et al., 2011), measurement error (O'Muircheartaigh and Campanelli, 1998;
Davis et al., 2010), responses to trust questions (Cuxart, 2011), and contingent val-
uation (Loureiro and Lotade, 2005).
3.3 Data and Methods
3.3.1 Data
Table 3.1(A) presents summary statistics of the individual and household charac-
teristics used in the analyses. Only about 37% of all respondents speak the national
language, Indonesian, at home in daily life. 16 percent speak more than one language.
Table 3.1(B) presents summary statistics of the 239 interviewers. Men make up a
slight majority, 56 percent. The average age of interviewers is 27 years. 79 percent
of interviewers have prior experience working as survey interviewers. 91 percent
have a bachelor's degree, and 70% intend to continue his/her studies. Only half of
all interviewers speak Indonesian at home in daily life. 32 percent speak multiple
languages.
Table 3.1(A): Summary statistics of IFLS-4 respondents
Mean Std Dev Count
Language abilities:
Speaks Indonesian in daily life 0.37 0.48 29054
Speaks multiple languages in daily life 0.16 0.37 29054
Interviewer characteristics:
Male interviewer 0.54 0.50 27740
Interviewer age (years) 25.83 3.40 27571
Interviewer older than respondent 0.27 0.44 27568
Interviewer same sex as respondent 0.60 0.49 27740
Has past interviewer experience 0.76 0.43 27740
Plan to continue studies 0.68 0.47 27740
Has bachelor's degree 0.91 0.29 27740
Monthly income in last job (Rp1; 000; 000) 1.81 1.16 24667
Interview not conducted in Indonesian 0.35 0.48 29053
Interviewer speaks respondent's language 0.56 0.50 29054
Speaks Indonesian in daily life 0.51 0.50 29054
continued on next page
50
continued from previous page
Speaks multiple languages in daily life 0.34 0.47 29054
Total Number of Respondents 29054
Table 3.1(B): Summary statistics of IFLS-4 interviewers
Mean Std Dev Count
Male interviewer 0.56 0.50 218
Interviewer age (years) 26.58 4.32 217
Has past interviewer experience 0.79 0.41 218
Plan to continue studies 0.70 0.46 218
Has bachelor's degree 0.91 0.29 218
Monthly income in last job (Rp1; 000; 000) 1.87 1.16 198
Speaks Indonesian in daily life 0.50 0.50 239
Speaks multiple languages in daily life 0.32 0.47 239
Total Number of Interviewers 239
3.3.2 Methods
In all regressions, I cluster the standard errors at the community level to allow for
arbitrary correlations within communities. Finally, I stress that this is a descriptive
exercise and coecient estimates are not to be interpreted as causal. I do occasionally
use words such as \eect" and \impact" for expository convenience.
Interviewers were recruited from each of the 13 provinces surveyed in the IFLS.
They were selected from potential candidates who had undergone standardized train-
ing in Solo, Province of Jawa Tengah. They were then divided into 23 teams, each
containing between 5 to 10 household-questionnaire interviewers working under a su-
pervisor. Each team was assigned to a specic province (some provinces needed more
than one team).
To elucidate the examination of interviewer eects, recall that the original model
from the rst essay (Chapter 2) is:
y
i
= + X
i
0
+
i
(3.1)
If we suspect that interviewer characteristics H should be included in the model,
51
then the \true" model is:
y
i
= + X
i
0
+ H
i
0

+
i
(3.2)
For there to exist omitted variable bias from the omission of H in estimation of
(3.1), two conditions must hold. H must be correlated with the preference variable
y, and H must be correlated with the original regressors X. On the one hand, there
is no reason a priori to suppose that H should be uncorrelated with y. Indeed,
researchers in economics and other disciplines have found evidence of interviewer
eects on survey instruments in face-to-face interviews (see Section 3.2). On the other
hand, we would expect and desire that interviewer characteristics are uncorrelated
with X. Therefore, I expect interviewers to not be a source of omitted variable bias;
their inclusion in (3.2) should still produce 's which are similar to those estimated
under (3.1). This section's objectives are to conrm this, and to estimate
, the
direct eect of interviewer characteristics on preferences.
I start by introducing interviewer dummies into the main regressions. To disen-
tangle specic channels through which direct interviewer eects operate, I include
select measures of interviewer human capital, social distance, and language. Follow-
ing Thomas et al. (2012), the human capital measures are whether the interviewer
has previous enumeration experience, whether he plans to further his studies after
this stint, whether he has a bachelor's degree, and the income from his last job.
The social distance measures are whether the interviewer is older than the respon-
dent and whether the interviewer is of the same sex as the respondent. The dummy
for whether the interviewer is older than the respondent captures an \age dierence
eect". Similarly, there could be a \sex dierence eect" at play if the two parties
are of opposing sex.
52
The main language measure I use is a dummy for whether the interviewer speaks
the respondent's language. The country's national language is Indonesian, or Bahasa
Indonesia, but daily usage varies widely across space and ethnicity: 63% of all IFLS
respondents do not speak Indonesian at home in daily life (IFLS recognizes 21 dis-
tinct languages). Interviewer training was done in Indonesian. The questionnaires
administered by interviewers in the eld are written in Indonesian. However, if a
respondent could not communicate well in Indonesian, the interview was conducted
in a local language other than Indonesian. Interviewer selection did take the need
for diverse language abilities into account,
2
but 44% of respondents still had an in-
terviewer who did not speak their daily language. Translators were used in many of
the cases that no interviewer spoke the local language. Even then, the translator was
not trained in the questionnaire so issues of meaning can arise.
In any communication between people who do not speak the same language, some
degree of \lost in translation" is possible. To see if the possible instances of lost in
translation aected responses in the elicitation modules, I also include a dummy vari-
able for whether the interview was conducted in a language other than Indonesian,
and its interaction with the dummy for whether the interviewer speaks the respon-
dent's language.
3.4 Results
3.4.1 Interviewer Dummies are Signicant
Interviewer xed eects are jointly signicant in all specications, as shown in Ta-
bles 3A.1(A) { 3A.1(D) in Appendix 3.A. However, estimates of the other explanatory
2
For example, the team that was sent to the island of Madura contained some Maduranese-
speaking interviewers (Strauss et al., 2009, p. 21).
53
variables are roughly unchanged with the inclusion of interviewer xed eects. This
is shown in Tables 3A.2(A) and 3A.2(B) in Appendix 3.A, which compare the inter-
viewer xed eect regression results to the main regression results in Tables 2.5(A){
2.5(D) from the rst essay. This conrms that interviewers are not an important
source of omitted variable bias in regression equations 2.4 and 2.5. Tables 3.2(A) {
3.2(D) present the estimates replacing interviewer dummies with specic interviewer
characteristics. The next two subsections describe the results.
3.4.2 Eect on Standard Preferences
Table 3.2(A) shows the results for risk aversion. Interviewer human capital variables
are jointly signicant in the basic specication (columns 1 and 4), and are robust
to community xed eects. Social distance variables are also jointly signicant in
columns 1 and 4. In particular, if a respondent is interviewed by someone older,
he becomes less risk averse, contrary to what we would expect. There is no clear
evidence of signicant language eects.
Table 3.2(B) presents the results for time preference. Interviewer human capital
variables are jointly signicant only for Time A. Social distance variables are jointly
insignicant in all specications. There is some evidence of language eects. In
interviews held in a language other than Indonesian, having an interviewer who speaks
the respondent's language results in the respondent becoming less impatient, in both
Time A and B. However the eect disappears when community or household xed
eects are added.
54
Table 3.2(A): Eect of Interviewer Characteristics on Risk Preference
(1) (2) (3) (4) (5) (6)
Risk A Risk A Risk A Risk A Risk B Risk B
Word recall z-score -0.008 -0.010 0.004 -0.004 -0.001 -0.001
(0.007) (0.006) (0.013) (0.004) (0.004) (0.007)
Language eects:
(I): Interview not conducted in Indonesian -0.018 0.057 0.033 -0.019 -0.007 0.015
(0.029) (0.032) (0.077) (0.016) (0.015) (0.035)
(II): Interviewer speaks respondent's language -0.016 -0.050 -0.034 0.016 0.002 -0.011
(0.020) (0.017)

(0.047) (0.012) (0.010) (0.020)
(III): (I) (II) 0.024 -0.024 0.016 -0.024 -0.009 -0.032
(0.034) (0.034) (0.084) (0.018) (0.018) (0.041)
Interviewer human capital:
Past interview experience -0.055 0.002 -0.004 -0.004 0.039 0.043
(0.021)

(0.020) (0.054) (0.011) (0.011)

(0.022)

Plan to continue studies -0.005 0.066 0.052 -0.028 0.016 0.038
(0.017) (0.016)

(0.043) (0.008)

(0.009) (0.019)

Has bachelor's degree 0.073 0.116 0.089 0.009 0.034 0.022
(0.026)

(0.032)

(0.083) (0.015) (0.016)

(0.031)
Log income in last job (monthly) 0.002 0.000 -0.002 0.002 0.001 -0.002
(0.002) (0.002) (0.004) (0.001) (0.001) (0.002)
Interviewer social distance:
Interviewer older than respondent -0.056 -0.025 -0.039 -0.033 -0.010 0.000
(0.022)

(0.018) (0.048) (0.015)

(0.012) (0.023)
Interviewer same sex as respondent -0.016 -0.002 0.016 -0.006 0.003 0.011
(0.009) (0.009) (0.021) (0.005) (0.005) (0.009)
Observations 13552 13552 13552 20840 20840 20840
R
2
0.018 0.167 0.704 0.019 0.114 0.610
(II) + (III) 0.008 -0.074 -0.018 -0.008 -0.006 -0.043
(0.029) (0.033)
**
(0.078) (0.015) (0.018) (0.041)
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.510 0.510 0.510 0.840 0.840 0.840
Joint signicance of (p-value):
Age categories 0.044 0.063 0.293 0.025 0.004 0.275
Education levels 0.717 0.038 0.338 0.003 0.011 0.606
Parents' education 0.112 0.015 0.840 0.190 0.001 0.391
Interviewer human capital 0.006 0.000 0.733 0.014 0.000 0.053
Interviewer social distance 0.012 0.385 0.541 0.080 0.530 0.532
Language eects: (I) and (III) 0.775 0.046 0.506 0.000 0.305 0.679
Language eects: (II) and (III) 0.689 0.003 0.762 0.293 0.883 0.565
Language eects: (I), (II) and (III) 0.862 0.002 0.649 0.001 0.496 0.711
Dependent variable Risk = 1 if highest risk aversion (category 4), = 0 if less risk averse (categories 1{3). All
regressions are OLS. Not shown: Male, age categories, education categories, Muslim, religiosity, Muslim Reli-
giosity, log PCE, rural, father's education, mother's education, constant. Standard errors in parentheses clustered
by community ID. *p<0.10 **p<0.05 ***p<0.01.
55
Table 3.2(B): Eect of Interviewer Characteristics on Time Preference
(1) (2) (3) (4) (5) (6)
Time A Time A Time A Time A Time B Time B
Word recall z-score -0.017 -0.014 -0.005 -0.012 -0.011 -0.004
(0.004)

(0.004)

(0.008) (0.004)

(0.004)

(0.007)
Language eects:
(I): Interview not conducted in Indonesian 0.016 0.026 0.036 -0.006 0.001 0.020
(0.016) (0.021) (0.043) (0.015) (0.019) (0.035)
(II): Interviewer speaks respondent's language -0.007 -0.016 -0.015 -0.006 0.010 0.005
(0.011) (0.011) (0.023) (0.009) (0.010) (0.021)
(III): (I) (II) -0.052 -0.011 -0.011 -0.015 0.019 0.006
(0.019)

(0.022) (0.045) (0.017) (0.021) (0.037)
Interviewer human capital:
Past interview experience -0.051 -0.018 -0.015 -0.025 0.008 0.004
(0.014)

(0.016) (0.026) (0.012)

(0.012) (0.022)
Plan to continue studies -0.022 -0.028 -0.018 -0.007 -0.004 -0.002
(0.011)

(0.011)

(0.020) (0.009) (0.009) (0.018)
Has bachelor's degree -0.038 -0.033 0.006 -0.016 0.002 0.027
(0.015)

(0.015)

(0.031) (0.014) (0.013) (0.028)
Log income in last job (monthly) 0.001 -0.001 -0.000 0.001 -0.000 0.001
(0.001) (0.001) (0.002) (0.001) (0.001) (0.002)
Interviewer social distance:
Interviewer older than respondent -0.006 -0.013 -0.018 0.026 0.019 0.020
(0.016) (0.014) (0.026) (0.013) (0.013) (0.020)
Interviewer same sex as respondent -0.008 -0.003 -0.008 -0.007 -0.001 -0.009
(0.006) (0.007) (0.012) (0.005) (0.005) (0.010)
Observations 22260 22260 22260 22495 22495 22495
R
2
0.028 0.088 0.594 0.019 0.077 0.580
(II) + (III) -0.059 -0.027 -0.026 -0.022 0.029 0.011
(0.017)
***
(0.022) (0.046) (0.014) (0.021) (0.038)
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.710 0.710 0.710 0.810 0.810 0.810
Joint signicance of (p-value):
Age categories 0.001 0.000 0.267 0.000 0.000 0.009
Education levels 0.000 0.000 0.043 0.000 0.000 0.040
Parents' education 0.000 0.000 0.571 0.499 0.123 0.718
Interviewer human capital 0.000 0.003 0.854 0.194 0.957 0.831
Interviewer social distance 0.383 0.603 0.640 0.057 0.312 0.370
Language eects: (I) and (III) 0.013 0.308 0.487 0.159 0.184 0.426
Language eects: (II) and (III) 0.002 0.203 0.743 0.242 0.312 0.947
Language eects: (I), (II) and (III) 0.001 0.187 0.634 0.054 0.143 0.623
Dependent variable Time = 1 if highest impatience (category 4), = 0 if less impatient (categories 1{3). All
regressions are OLS. Omitted category for age: 15-24, for education: no education. Not shown: Male, age
categories, education categories, Muslim, religiosity, Muslim Religiosity, log PCE, rural, father's education,
mother's education. Standard errors in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
56
3.4.3 Eect on Nonstandard Preferences
The results for nonstandard preferences are stronger and more interesting. Ta-
ble 3.2(C) presents the results for GA A, where language eects are strong and
interesting (column 1). In interviews conducted in the national language, Bahasa,
the respondent is not any more or less likely to be GA A whether or not the respon-
dent and his interviewer share the same daily language. However, in interviews held
in a non-Bahasa language, if they speak the same daily language, the respondent is
9.4 percentage points less likely to be GA A. The magnitude of the eect drops to
6 points with community xed eects but is still signicant at 5%. To underscore
the importance of interviewer language ability in non-Bahasa interviews, if the in-
terviewer does not regularly speak the respondent's language, the respondent is 8.8
percentage points more likely to be GA A.
Interviewer human capital variables are robustly jointly signicantly. Respon-
dents are less likely to be GA A if their interviewer has past enumeration experience
(column 1). However, GL B shows exactly the opposite pattern (column 4). Recall
however that the nonstandard behavior in Risk Submodule B is not gamble aversion,
but gamble loving (the dominated option that nonstandard respondents picked was
actually a gamble).
The signicant negative sign on whether the interviewer plans to continue stud-
ies (column 1) and positive sign on whether he already has a bachelor's degree
(columns 1 and 4) suggest that interviewers looking to increase their educational
capital are more likely to elicit standard responses. Perhaps these interviewers, who
are lower down the socioeconomic ladder but are looking to move up, are more moti-
vated to obtain the correct response. The positive sign on income in interviewer's last
job lends credence to this interpretation. Social distance variables are not signicant.
57
Table 3.2(C): Eect of Interviewer Characteristics on Gamble Aversion/Loving
(1) (2) (3) (4) (5) (6)
GA A GA A GA A GA A GL B GL B
Word recall z-score -0.031 -0.018 -0.010 -0.011 -0.010 -0.005
(0.005)

(0.005)

(0.008) (0.003)

(0.003)

(0.005)
Language eects:
(I): Interview not conducted in Indonesian 0.088 0.041 0.048 0.016 0.012 0.009
(0.020)

(0.024) (0.051) (0.012) (0.014) (0.028)
(II): Interviewer speaks respondent's language 0.006 -0.019 -0.032 0.002 0.001 0.013
(0.014) (0.013) (0.027) (0.006) (0.008) (0.014)
(III): (I) (II) -0.099 -0.030 -0.004 -0.011 0.007 -0.004
(0.024)

(0.027) (0.057) (0.014) (0.015) (0.030)
Interviewer human capital:
Past interview experience -0.050 -0.011 0.025 0.020 0.011 0.004
(0.019)

(0.019) (0.039) (0.006)

(0.007) (0.014)
Plan to continue studies -0.061 -0.055 -0.018 0.012 0.005 -0.002
(0.014)

(0.017)

(0.031) (0.007) (0.009) (0.018)
Has bachelor's degree 0.062 0.043 0.016 0.044 0.034 0.042
(0.022)

(0.026) (0.045) (0.007)

(0.008)

(0.015)

Log income in last job (monthly) 0.007 0.006 0.004 -0.000 0.001 0.002
(0.001)

(0.001)

(0.003) (0.000) (0.000) (0.001)

Interviewer social distance:
Interviewer older than respondent -0.028 -0.017 -0.037 0.014 0.004 0.008
(0.017) (0.016) (0.027) (0.008) (0.008) (0.015)
Interviewer same sex as respondent 0.001 0.007 0.010 0.003 -0.002 -0.003
(0.007) (0.007) (0.011) (0.004) (0.004) (0.007)
Observations 22559 22559 22559 22565 22565 22565
R
2
0.041 0.128 0.609 0.008 0.072 0.566
(II) + (III) -0.093 -0.049 -0.036 -0.009 0.008 0.009
(0.020)
***
(0.026)
*
(0.055) (0.012) (0.014) (0.028)
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.400 0.400 0.400 0.080 0.080 0.080
Joint signicance of (p-value):
Age categories 0.003 0.000 0.333 0.457 0.365 0.951
Education levels 0.000 0.000 0.012 0.840 0.431 0.328
Parents' education 0.006 0.000 0.687 0.659 0.268 0.475
Interviewer human capital 0.000 0.000 0.485 0.000 0.000 0.006
Interviewer social distance 0.219 0.244 0.232 0.186 0.804 0.817
Language eects: (I) and (III) 0.000 0.203 0.269 0.367 0.093 0.909
Language eects: (II) and (III) 0.000 0.080 0.447 0.732 0.859 0.652
Language eects: (I), (II) and (III) 0.000 0.086 0.279 0.567 0.110 0.729
Dependent variable GA = 1 if gamble averse, = 0 if standard (risk aversion categories 1{4). All regressions are OLS.
Omitted category for age: 15-24, for education: no education. Not shown: Male, age categories, education cate-
gories, Muslim, religiosity, Muslim Religiosity, log PCE, rural, father's education, mother's education. Standard
errors in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
Table 3.2(D) shows similar language eects for NTD, and moreover, they are
robust to submodules A and B. That they show up is quite remarkable given that
NTD A and B make up only 2% and 1% of the sample. In a language-heterogeneous
setting, interviewers must sometimes communicate with respondents in a language
they may not be comfortable speaking. My results indicate that some information
may have been lost in translation in these instances, leading to \incorrect" responses.
58
Table 3.2(D): Eect of Interviewer Characteristics on Negative Time Discounting
(1) (2) (3) (4) (5) (6)
NTD A NTD A NTD A NTD A NTD B NTD B
Word recall z-score -0.004 -0.004 -0.001 -0.002 -0.001 -0.001
(0.001)

(0.001)

(0.002) (0.001)

(0.001)

(0.001)
Language eects:
(I): Interview not conducted in Indonesian 0.013 -0.003 0.005 0.017 0.006 0.004
(0.005)

(0.005) (0.014) (0.004)

(0.003) (0.008)
(II): Interviewer speaks respondent's language 0.001 0.000 0.001 0.001 0.001 0.000
(0.002) (0.003) (0.007) (0.001) (0.001) (0.003)
(III): (I) (II) -0.012 -0.000 -0.011 -0.018 -0.006 -0.004
(0.005)

(0.005) (0.013) (0.005)

(0.003) (0.007)
Interviewer human capital:
Past interview experience 0.003 0.001 0.006 0.004 0.005 0.006
(0.003) (0.003) (0.008) (0.002)

(0.002)

(0.005)
Plan to continue studies -0.003 -0.002 0.000 0.001 -0.000 -0.002
(0.002) (0.003) (0.006) (0.001) (0.002) (0.003)
Has bachelor's degree 0.005 0.007 0.002 0.003 0.002 0.000
(0.003) (0.004) (0.008) (0.002) (0.003) (0.006)
Log income in last job (monthly) 0.000 0.000 0.000 -0.000 -0.000 -0.000
(0.000) (0.000) (0.001) (0.000) (0.000) (0.000)
Interviewer social distance:
Interviewer older than respondent 0.002 0.006 0.007 -0.002 0.000 -0.000
(0.003) (0.003) (0.006) (0.002) (0.002) (0.004)
Interviewer same sex as respondent -0.002 -0.001 -0.003 -0.002 -0.002 -0.002
(0.002) (0.002) (0.004) (0.001) (0.001) (0.002)
Observations 22639 22639 22639 22638 22638 22638
R
2
0.005 0.037 0.509 0.006 0.041 0.581
(II) + (III) -0.011 -0.000 -0.010 -0.017 -0.005 -0.004
(0.005)
**
(0.004) (0.012) (0.004)
***
(0.003) (0.006)
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.020 0.020 0.020 0.010 0.010 0.010
Joint signicance of (p-value):
Age categories 0.038 0.064 0.870 0.059 0.312 0.768
Education levels 0.297 0.329 0.631 0.863 0.870 0.962
Parents' education 0.042 0.110 0.365 0.072 0.133 0.563
Interviewer human capital 0.149 0.117 0.899 0.143 0.159 0.670
Interviewer social distance 0.427 0.168 0.428 0.172 0.326 0.633
Language eects: (I) and (III) 0.030 0.554 0.535 0.001 0.205 0.836
Language eects: (II) and (III) 0.063 1.000 0.678 0.001 0.244 0.827
Language eects: (I), (II) and (III) 0.070 0.738 0.673 0.002 0.354 0.942
Dependent variable NTD = 1 if negative time discount, 0 if standard (time pref. categories 1{4). All regressions
are OLS. Omitted category for age: 15-24, for education: no education. Not shown: Male, age categories,
education categories, Muslim, religiosity, Muslim Religiosity, log PCE, rural, father's education, mother's
education. Standard errors in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
3.4.4 Switching from Nonstandard to Standard
Recall for the Risk lter questions, respondents were given a chance to rethink their
answers if they picked the dominated payo. For example, if a respondent picked
\Rp 800,000 guaranteed" over \equal chance of Rp 800,000 or Rp 1,600,000", the
interviewer would ask whether he was sure about his choice, explain that the alterna-
tive option was at least as good, and give the respondent an opportunity to change
59
Table 3.3: Do Interviewers In
uence Change in Answer to Risk Filter Question?
(1) (2) (3) (4)
Change A Change A Change B Change B
Word recall z-score 0.008 0.003 0.006 0.005
(0.003)

(0.004) (0.010) (0.014)
Language eects:
(I): Interview not conducted in Indonesian -0.005 0.003 0.045 0.026
(0.012) (0.014) (0.040) (0.070)
(II): Interviewer speaks respondent's language -0.011 0.006 0.018 0.007
(0.010) (0.015) (0.023) (0.036)
(III): (I) (II) -0.002 -0.013 -0.091 -0.073
(0.015) (0.019) (0.048) (0.077)
Interviewer human capital:
Past interview experience -0.010 -0.009 -0.016 -0.043
(0.012) (0.016) (0.031) (0.049)
Plan to continue studies 0.031 0.030 -0.026 -0.014
(0.008)

(0.010)

(0.022) (0.036)
Has bachelor's degree -0.024 -0.042 -0.147 -0.132
(0.018) (0.023) (0.051)

(0.101)
Log income in last job (monthly) -0.003 -0.001 -0.004 0.001
(0.001) (0.002) (0.002) (0.004)
Interviewer social distance:
Interviewer older than respondent -0.003 0.002 -0.051 -0.044
(0.012) (0.013) (0.030) (0.042)
Interviewer same sex as respondent -0.002 0.001 -0.000 -0.005
(0.006) (0.006) (0.018) (0.025)
Observations 9787 9787 2069 2069
R
2
0.010 0.180 0.030 0.384
(II) + (III) -0.013 -0.006 -0.073 -0.066
(0.012) (0.019) (0.040)
*
(0.073)
Community FE No Yes No Yes
Household FE No No No No
Mean of dependent variable 0.080 0.080 0.170 0.170
Joint signicance of (p-value):
Age categories 0.684 0.036 0.070 0.492
Education levels 0.172 0.156 0.192 0.353
Parents' education 0.900 0.975 0.456 0.671
Interviewer human capital 0.000 0.011 0.009 0.698
Interviewer social distance 0.919 0.966 0.247 0.571
Language eects: (I) and (III) 0.757 0.759 0.121 0.482
Language eects: (II) and (III) 0.359 0.775 0.155 0.633
Language eects: (I), (II) and (III) 0.369 0.897 0.201 0.680
Dependent variable Change A(B) = 1 if respondent changed response from nonstandard to stan-
dard, = 0 if respondent stayed nonstandard, in Risk A(B) lter question. All regressions are
OLS. Not shown: Male, age categories, education categories, Muslim, religiosity, Muslim Re-
ligiosity, log PCE, rural, father's education, mother's education, constant. Standard errors in
parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
his answer. In other words, if a respondent initially gave a nonstandard response,
he could have corrected himself and chosen a standard response upon interviewer
prompt. In Risk Submodule A, 1,015 out of 12,055 initially nonstandard respondents
switched to the standard response. The percentage of switchers was even higher in
Submodule B, where 457 out of 2,380 switched.
I test whether interviewer characteristics aected the likelihood of an initially
60
nonstandard respondent changing his choice to become standard. The results can
be found in Table 3.3. I do not nd language or social distance eects. Interviewer
human capital variables are jointly signicant, but are individually insignicant with
two exceptions. Planning to continue studies is signicantly positive for A, and having
a bachelor's degree is signicantly negative for B. This is consistent with the inter-
pretation given for estimated eects on the likelihood of being nonstandard in the
full sample (Table 3.2(C)), i.e. interviewers planning to move up the socioeconomic
ladder may have been more motivated to induce a correct response.
I do not consider NTD switching because of the extremely small sample sizes
involved (98 out of 627 switched from NTD A to standard, 207 out of 231 switched
from NTD B to standard).
3.5 Conclusion
In addition, I have conrmed that interviewer eects are not an important source of
omitted variable bias in the estimation of preference correlates. There is, however,
strong evidence of language eects on elicited preferences in interviews conducted in
a language other than the national language. In these interviews, where interviewers
and respondents share the same language, respondents are more likely to exhibit
standard preferences, indicating more eective communication. This suggests that
interviewer language ability could be an especially important consideration for survey
designers where the surveyed population has great language diversity.
61
Appendix 3.A Interviewer Fixed Eects
Table 3A.1(A): Correlates of Risk Preference: Interviewer Fixed Eects
(1) (2) (3) (4) (5) (6)
Risk A Risk A Risk A Risk B Risk B Risk B
Male -0.089 -0.088 -0.086 -0.060 -0.060 -0.061
(0.008)

(0.008)

(0.015)

(0.004)

(0.005)

(0.008)

Age category (years):
25-34 -0.014 -0.014 -0.013 0.001 -0.001 0.003
(0.011) (0.011) (0.028) (0.007) (0.007) (0.013)
35-44 -0.035 -0.035 -0.009 -0.003 -0.006 0.006
(0.012)

(0.012)

(0.031) (0.008) (0.008) (0.015)
45-54 -0.014 -0.018 0.022 0.012 0.010 0.021
(0.014) (0.015) (0.037) (0.009) (0.009) (0.017)
55-64 -0.001 0.004 0.070 0.029 0.025 0.050
(0.018) (0.018) (0.046) (0.010)

(0.010)

(0.021)

65+ 0.001 -0.001 0.081 0.040 0.038 0.053
(0.022) (0.022) (0.061) (0.012)

(0.012)

(0.026)

Education level:
Elementary -0.040 -0.034 -0.044 0.014 0.018 -0.005
(0.021) (0.020) (0.055) (0.010) (0.010) (0.020)
Junior high -0.024 -0.021 -0.008 0.008 0.009 -0.006
(0.024) (0.022) (0.058) (0.012) (0.011) (0.024)
Senior high/university -0.008 -0.007 -0.004 -0.006 -0.003 -0.022
(0.024) (0.023) (0.061) (0.013) (0.012) (0.024)
Word recall z-score -0.001 -0.002 0.008 -0.001 -0.003 -0.002
(0.005) (0.005) (0.012) (0.003) (0.003) (0.005)
Muslim 0.131 0.134 0.185 0.040 0.043 -0.037
(0.053)

(0.055)

(0.207) (0.032) (0.034) (0.070)
Religiosity (0-not, 3-very) 0.053 0.065 0.084 0.017 0.026 0.017
(0.025)

(0.025)

(0.074) (0.015) (0.016) (0.029)
Muslim Religiosity -0.050 -0.060 -0.076 -0.017 -0.024 -0.014
(0.026) (0.026)

(0.076) (0.016) (0.016) (0.030)
Father's education level -0.012 -0.011 0.009 0.002 0.002 0.006
(0.006)

(0.006) (0.015) (0.003) (0.003) (0.007)
Mother's education level -0.008 -0.006 -0.013 -0.017 -0.015 -0.012
(0.006) (0.007) (0.016) (0.004)

(0.004)

(0.008)
Log PCE -0.026 -0.026 -0.018 -0.018
(0.006)

(0.006)

(0.004)

(0.004)

Rural 0.005 -0.015 -0.017 -0.012 -0.004 0.127
(0.010) (0.020) (0.117) (0.007) (0.010) (0.061)

Constant 0.842 1.093 0.332 1.268 1.088 1.594
(0.334)

(0.399)

(0.297) (0.067)

(0.183)

(0.174)

Observations 15643 15643 15643 24445 24445 24445
R
2
0.225 0.267 0.720 0.174 0.203 0.612
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.500 0.500 0.500 0.840 0.840 0.840
Joint signicance of (p-value):
Age categories 0.038 0.031 0.236 0.000 0.000 0.084
Education levels 0.011 0.051 0.422 0.018 0.006 0.554
Parents' education 0.001 0.012 0.694 0.000 0.000 0.286
Interviewer xed eects 0.000 0.000 0.000 0.000 0.000 0.000
Dependent variable Risk = 1 if highest risk aversion (category 4), = 0 if less risk averse (categories 1-3).
All regressions are OLS. Omitted category for age: 15-24, for education: no education. Standard errors
in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
62
Table 3A.1(B): Correlates of Time Preference: Interviewer Fixed Eects
(1) (2) (3) (4) (5) (6)
Time A Time A Time A Time B Time B Time B
Male -0.001 -0.002 0.002 -0.007 -0.007 -0.006
(0.005) (0.005) (0.008) (0.005) (0.005) (0.008)
Age category (years):
25-34 0.045 0.048 0.047 0.023 0.024 0.016
(0.008)

(0.008)

(0.015)

(0.007)

(0.007)

(0.014)
35-44 0.040 0.043 0.070 0.031 0.028 0.038
(0.009)

(0.009)

(0.016)

(0.007)

(0.007)

(0.015)

45-54 0.047 0.052 0.073 0.044 0.044 0.052
(0.011)

(0.011)

(0.019)

(0.009)

(0.010)

(0.017)

55-64 0.068 0.073 0.103 0.073 0.070 0.088
(0.013)

(0.013)

(0.025)

(0.011)

(0.011)

(0.020)

65+ 0.097 0.105 0.114 0.083 0.081 0.091
(0.014)

(0.015)

(0.030)

(0.012)

(0.012)

(0.026)

Education level:
Elementary -0.010 -0.000 -0.009 0.001 -0.001 0.000
(0.012) (0.013) (0.023) (0.010) (0.011) (0.019)
Junior high -0.013 -0.003 -0.026 0.005 0.004 0.005
(0.014) (0.015) (0.027) (0.012) (0.013) (0.021)
Senior high/university -0.052 -0.040 -0.052 -0.027 -0.027 -0.021
(0.013)

(0.015)

(0.027) (0.012)

(0.014) (0.022)
Word recall z-score -0.011 -0.011 -0.005 -0.009 -0.010 -0.007
(0.004)

(0.004)

(0.007) (0.003)

(0.003)

(0.006)
Muslim 0.008 0.036 -0.058 -0.029 -0.008 -0.066
(0.035) (0.036) (0.098) (0.029) (0.029) (0.082)
Religiosity (0-not, 3-very) -0.007 0.001 -0.005 -0.019 -0.004 -0.007
(0.017) (0.017) (0.030) (0.014) (0.013) (0.025)
Muslim Religiosity 0.002 -0.005 0.004 0.021 0.007 0.015
(0.017) (0.017) (0.032) (0.014) (0.013) (0.027)
Father's education level -0.008 -0.008 -0.006 -0.004 -0.005 -0.004
(0.004) (0.004) (0.008) (0.004) (0.004) (0.007)
Mother's education level -0.014 -0.012 -0.006 -0.008 -0.007 -0.007
(0.005)

(0.005)

(0.009) (0.004) (0.004) (0.008)
Log PCE -0.020 -0.019 -0.018 -0.016
(0.004)

(0.004)

(0.003)

(0.004)

Rural -0.003 0.017 -0.031 -0.001 0.012 -0.004
(0.008) (0.015) (0.061) (0.007) (0.013) (0.051)
Constant 0.869 0.597 0.180 1.318 1.048 0.319
(0.336)

(0.326) (0.172) (0.060)

(0.167)

(0.180)
Observations 26253 26253 26253 26552 26552 26552
R
2
0.135 0.161 0.584 0.109 0.136 0.570
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.710 0.710 0.710 0.810 0.810 0.810
Joint signicance of (p-value):
Age categories 0.000 0.000 0.000 0.000 0.000 0.001
Education levels 0.000 0.000 0.058 0.000 0.000 0.229
Parents' education 0.000 0.000 0.398 0.035 0.025 0.399
Interviewer xed eects 0.000 0.000 0.000 0.000 0.000 0.000
Dependent variable Time = 1 if highest impatience (category 4), = 0 if less impatient (categories 1{3).
All regressions are OLS. Omitted category for age: 15-24, for education: no education. Standard errors
in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
63
Table 3A.1(C): Correlates of GA/GL: Interviewer Fixed Eects
(1) (2) (3) (4) (5) (6)
GA A GA A GA A GL B GL B GL B
Male -0.062 -0.063 -0.067 0.017 0.018 0.015
(0.006)

(0.006)

(0.009)

(0.004)

(0.004)

(0.005)

Age category (years):
25-34 -0.011 -0.010 0.003 0.001 0.004 -0.006
(0.007) (0.007) (0.015) (0.005) (0.005) (0.009)
35-44 -0.012 -0.007 0.012 0.007 0.010 0.007
(0.008) (0.008) (0.016) (0.005) (0.005) (0.010)
45-54 0.009 0.015 0.042 -0.001 0.001 -0.002
(0.011) (0.011) (0.020)

(0.006) (0.006) (0.012)
55-64 0.008 0.016 0.051 -0.004 -0.003 -0.004
(0.012) (0.012) (0.025)

(0.007) (0.008) (0.015)
65+ 0.055 0.069 0.080 0.001 0.001 -0.009
(0.016)

(0.016)

(0.033)

(0.009) (0.009) (0.018)
Education level:
Elementary 0.002 0.019 0.015 -0.002 -0.006 -0.004
(0.013) (0.013) (0.026) (0.009) (0.009) (0.016)
Junior high -0.016 0.003 0.004 -0.004 -0.006 -0.015
(0.014) (0.014) (0.028) (0.009) (0.010) (0.018)
Senior high/university -0.069 -0.051 -0.041 -0.012 -0.014 -0.022
(0.015)

(0.015)

(0.030) (0.010) (0.010) (0.020)
Word recall z-score -0.022 -0.022 -0.017 -0.007 -0.007 -0.003
(0.003)

(0.003)

(0.006)

(0.002)

(0.002)

(0.004)
Muslim 0.029 0.020 -0.037 0.013 0.012 -0.015
(0.036) (0.037) (0.104) (0.024) (0.024) (0.058)
Religiosity (0-not, 3-very) 0.008 -0.000 -0.001 0.005 0.001 -0.010
(0.017) (0.017) (0.029) (0.011) (0.011) (0.019)
Muslim Religiosity -0.012 -0.000 0.010 -0.008 -0.005 0.010
(0.018) (0.018) (0.032) (0.012) (0.012) (0.020)
Father's education level -0.011 -0.011 -0.001 0.001 0.001 0.003
(0.004)

(0.004)

(0.008) (0.003) (0.003) (0.005)
Mother's education level -0.011 -0.008 -0.008 0.004 0.004 0.003
(0.005)

(0.005) (0.009) (0.003) (0.003) (0.006)
Log PCE -0.018 -0.021 -0.002 -0.000
(0.004)

(0.004)

(0.002) (0.002)
Rural -0.000 0.003 0.016 -0.004 0.009 -0.005
(0.008) (0.015) (0.072) (0.005) (0.008) (0.037)
Constant 0.409 0.607 1.115 0.030 0.029 -0.416
(0.070)

(0.166)

(0.314)

(0.045) (0.103) (0.113)

Observations 26645 26645 26645 26650 26650 26650
R
2
0.269 0.294 0.630 0.145 0.175 0.565
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.410 0.410 0.410 0.080 0.080 0.080
Joint signicance of (p-value):
Age categories 0.000 0.000 0.090 0.580 0.318 0.817
Education levels 0.000 0.000 0.004 0.280 0.318 0.369
Parents' education 0.000 0.000 0.580 0.171 0.184 0.553
Interviewer xed eects 0.000 0.000 0.000 0.000 0.000 0.000
Dependent variable GA = 1 if gamble averse, = 0 if standard (risk aversion categories 1{4). All re-
gressions are OLS. Omitted category for age: 15-24, for education: no education. Standard errors in
parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
64
Table 3A.1(D): Correlates of NTD: Interviewer Fixed Eects
(1) (2) (3) (4) (5) (6)
NTD A NTD A NTD A NTD B NTD B NTD B
Male 0.005 0.005 0.007 0.003 0.004 0.002
(0.002)

(0.002)

(0.003)

(0.001)

(0.001)

(0.002)
Age category (years):
25-34 -0.010 -0.009 -0.009 -0.002 -0.002 -0.003
(0.002)

(0.002)

(0.005)

(0.001) (0.001) (0.003)
35-44 -0.010 -0.010 -0.013 -0.000 -0.001 -0.001
(0.003)

(0.003)

(0.005)

(0.002) (0.002) (0.003)
45-54 -0.012 -0.012 -0.017 -0.002 -0.003 -0.003
(0.003)

(0.003)

(0.006)

(0.002) (0.002) (0.004)
55-64 -0.019 -0.018 -0.020 -0.004 -0.005 -0.003
(0.004)

(0.004)

(0.007)

(0.002) (0.003) (0.005)
65+ -0.016 -0.016 -0.018 -0.003 -0.004 -0.000
(0.005)

(0.005)

(0.009) (0.003) (0.003) (0.007)
Education level:
Elementary -0.003 -0.005 -0.009 -0.005 -0.006 -0.003
(0.005) (0.005) (0.009) (0.003) (0.003) (0.006)
Junior high -0.008 -0.010 -0.014 -0.008 -0.009 -0.006
(0.005) (0.005) (0.011) (0.004)

(0.004)

(0.006)
Senior high/university -0.009 -0.010 -0.014 -0.008 -0.009 -0.005
(0.005) (0.005) (0.011) (0.004)

(0.004)

(0.007)
Word recall z-score -0.003 -0.003 -0.001 -0.001 -0.001 -0.000
(0.001)

(0.001)

(0.002) (0.001) (0.001) (0.001)
Muslim 0.019 0.024 0.062 0.022 0.022 0.019
(0.011) (0.012)

(0.035) (0.008)

(0.008)

(0.027)
Religiosity (0-not, 3-very) 0.012 0.010 0.020 0.013 0.012 0.010
(0.005)

(0.006) (0.015) (0.004)

(0.004)

(0.009)
Muslim Religiosity -0.011 -0.009 -0.020 -0.012 -0.010 -0.009
(0.006) (0.007) (0.015) (0.004)

(0.004)

(0.009)
Father's education level 0.000 -0.000 0.000 -0.000 -0.001 -0.002
(0.001) (0.001) (0.002) (0.001) (0.001) (0.002)
Mother's education level 0.002 0.003 0.003 0.001 0.001 0.002
(0.001) (0.002) (0.003) (0.001) (0.001) (0.002)
Log PCE -0.001 -0.000 -0.000 -0.000
(0.001) (0.001) (0.001) (0.001)
Rural -0.001 0.003 -0.001 -0.000 0.000 0.001
(0.002) (0.003) (0.009) (0.001) (0.002) (0.002)
Constant 0.012 0.129 -0.415 -0.010 -0.020 -0.033
(0.020) (0.060)

(0.057)

(0.013) (0.017) (0.031)
Observations 26740 26740 26740 26739 26739 26739
R
2
0.047 0.068 0.498 0.076 0.093 0.509
Community FE No Yes No No Yes No
Household FE No No Yes No No Yes
Mean of dependent variable 0.020 0.020 0.020 0.010 0.010 0.010
Joint signicance of (p-value):
Age categories 0.000 0.000 0.061 0.356 0.367 0.769
Education levels 0.088 0.118 0.524 0.116 0.075 0.754
Parents' education 0.144 0.125 0.547 0.289 0.353 0.400
Interviewer xed eects 0.000 0.000 0.000 0.000 0.000 0.000
Dependent variable NTD = 1 if negative time discount, 0 if standard (time pref. categories 1{4).
All regressions are OLS. Omitted category for age: 15-24, for education: no education. Standard
errors in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
65
Table 3A.2(A): Interviewer Fixed Eects Do Not Change Risk & Time Main Results
(1) (2) (3) (4) (5) (6) (7) (8)
Risk A
Main
Risk A
IvwrFE
Risk B
Main
Risk B
IvwrFE
Time A
Main
Time A
IvwrFE
Time B
Main
Time B
IvwrFE
Male -0.083 -0.089 -0.061 -0.060 -0.015 -0.001 -0.013 -0.007
(0.008)

(0.008)

(0.005)

(0.004)

(0.006)

(0.005) (0.005)

(0.005)
Age category (years):
25-34 -0.009 -0.014 -0.001 0.001 0.047 0.045 0.025 0.023
(0.012) (0.011) (0.008) (0.007) (0.008)

(0.008)

(0.007)

(0.007)

35-44 -0.019 -0.035 0.002 -0.003 0.043 0.040 0.036 0.031
(0.014) (0.012)

(0.009) (0.008) (0.010)

(0.009)

(0.008)

(0.007)

45-54 -0.003 -0.014 0.011 0.012 0.046 0.047 0.045 0.044
(0.018) (0.014) (0.010) (0.009) (0.012)

(0.011)

(0.010)

(0.009)

55-64 0.015 -0.001 0.031 0.029 0.069 0.068 0.074 0.073
(0.023) (0.018) (0.012)

(0.010)

(0.013)

(0.013)

(0.011)

(0.011)

65+ 0.009 0.001 0.040 0.040 0.092 0.097 0.081 0.083
(0.028) (0.022) (0.015)

(0.012)

(0.015)

(0.014)

(0.013)

(0.012)

Education level:
Elementary -0.039 -0.040 0.026 0.014 -0.001 -0.010 0.003 0.001
(0.028) (0.021) (0.013)

(0.010) (0.014) (0.012) (0.011) (0.010)
Junior high -0.031 -0.024 0.015 0.008 -0.002 -0.013 0.007 0.005
(0.031) (0.024) (0.014) (0.012) (0.016) (0.014) (0.013) (0.012)
Senior high/university -0.024 -0.008 -0.003 -0.006 -0.042 -0.052 -0.025 -0.027
(0.032) (0.024) (0.015) (0.013) (0.016)

(0.013)

(0.014) (0.012)

Word recall z-score -0.007 -0.001 -0.005 -0.001 -0.016 -0.011 -0.011 -0.009
(0.006) (0.005) (0.004) (0.003) (0.004)

(0.004)

(0.004)

(0.003)

Muslim 0.113 0.131 0.026 0.040 -0.036 0.008 -0.049 -0.029
(0.060) (0.053)

(0.031) (0.032) (0.040) (0.035) (0.033) (0.029)
Religiosity (0-not, 3-very) 0.060 0.053 0.017 0.017 -0.037 -0.007 -0.041 -0.019
(0.025)

(0.025)

(0.014) (0.015) (0.021) (0.017) (0.018)

(0.014)
Muslim Religiosity -0.058 -0.050 -0.015 -0.017 0.021 0.002 0.036 0.021
(0.027)

(0.026) (0.015) (0.016) (0.022) (0.017) (0.018)

(0.014)
Father's education level -0.009 -0.012 0.001 0.002 -0.011 -0.008 -0.004 -0.004
(0.007) (0.006)

(0.004) (0.003) (0.005)

(0.004) (0.004) (0.004)
Mother's education level 0.001 -0.008 -0.009 -0.017 -0.008 -0.014 -0.004 -0.008
(0.008) (0.006) (0.005) (0.004)

(0.006) (0.005)

(0.005) (0.004)
Log PCE -0.025 -0.026 -0.014 -0.018 -0.018 -0.020 -0.017 -0.018
(0.008)

(0.006)

(0.005)

(0.004)

(0.005)

(0.004)

(0.004)

(0.003)

Rural -0.055 0.005 -0.025 -0.012 0.012 -0.003 0.005 -0.001
(0.020)

(0.010) (0.011)

(0.007) (0.011) (0.008) (0.010) (0.007)
Constant 0.811 0.842 1.017 1.268 0.998 0.869 1.065 1.318
(0.126)

(0.334)

(0.072)

(0.067)

(0.074)

(0.336)

(0.059)

(0.060)

Observations 15644 15643 24446 24445 26254 26253 26553 26552
R
2
0.013 0.225 0.015 0.174 0.021 0.135 0.016 0.109
Community FE No No No No No No No No
Household FE No No No No No No No No
Mean of dependent variable 0.500 0.500 0.840 0.840 0.710 0.710 0.810 0.810
Joint signicance of (p-value):
Age categories 0.410 0.038 0.025 0.000 0.000 0.000 0.000 0.000
Education levels 0.429 0.011 0.003 0.018 0.000 0.000 0.000 0.000
Parents' education 0.358 0.001 0.135 0.000 0.001 0.000 0.291 0.035
Interviewer xed eects 0.000 0.000 0.000 0.000
See previous tables for denition of dependent variables. All regressions are OLS. Omitted category for age: 15-
24, for education: no education. Standard errors in parentheses clustered by community ID. *p<0.10 **p<0.05
***p<0.01.
66
Table 3A.2(B): Interviewer Fixed Eects Do Not Change GA, GL & NTD Main
Results
(1) (2) (3) (4) (5) (6) (7) (8)
GA A
Main
GA A
IvwrFE
GL B
Main
GL B
IvwrFE
NTD A
Main
NTD A
IvwrFE
NTD B
Main
NTD B
IvwrFE
Male -0.084 -0.062 0.011 0.017 0.004 0.005 0.003 0.003
(0.007)

(0.006)

(0.004)

(0.004)

(0.002)

(0.002)

(0.001)

(0.001)

Age category (years):
25-34 -0.011 -0.011 -0.002 0.001 -0.011 -0.010 -0.004 -0.002
(0.009) (0.007) (0.005) (0.005) (0.002)

(0.002)

(0.001)

(0.001)
35-44 -0.012 -0.012 0.001 0.007 -0.010 -0.010 -0.002 -0.000
(0.009) (0.008) (0.006) (0.005) (0.003)

(0.003)

(0.002) (0.002)
45-54 0.007 0.009 -0.004 -0.001 -0.013 -0.012 -0.003 -0.002
(0.012) (0.011) (0.007) (0.006) (0.003)

(0.003)

(0.002) (0.002)
55-64 -0.002 0.008 -0.009 -0.004 -0.019 -0.019 -0.006 -0.004
(0.015) (0.012) (0.008) (0.007) (0.004)

(0.004)

(0.003)

(0.002)
65+ 0.049 0.055 -0.002 0.001 -0.015 -0.016 -0.004 -0.003
(0.019)

(0.016)

(0.010) (0.009) (0.005)

(0.005)

(0.003) (0.003)
Education level:
Elementary 0.031 0.002 0.002 -0.002 -0.000 -0.003 -0.003 -0.005
(0.017) (0.013) (0.010) (0.009) (0.005) (0.005) (0.003) (0.003)
Junior high 0.015 -0.016 0.001 -0.004 -0.007 -0.008 -0.006 -0.008
(0.019) (0.014) (0.011) (0.009) (0.005) (0.005) (0.003) (0.004)

Senior high/university -0.044 -0.069 -0.002 -0.012 -0.006 -0.009 -0.005 -0.008
(0.020)

(0.015)

(0.012) (0.010) (0.005) (0.005) (0.003) (0.004)

Word recall z-score -0.032 -0.022 -0.010 -0.007 -0.005 -0.003 -0.002 -0.001
(0.005)

(0.003)

(0.002)

(0.002)

(0.001)

(0.001)

(0.001)

(0.001)
Muslim 0.027 0.029 -0.013 0.013 0.020 0.019 0.021 0.022
(0.041) (0.036) (0.025) (0.024) (0.010) (0.011) (0.009)

(0.008)

Religiosity (0-not, 3-very) 0.005 0.008 -0.002 0.005 0.015 0.012 0.014 0.013
(0.018) (0.017) (0.012) (0.011) (0.005)

(0.005)

(0.005)

(0.004)

Muslim Religiosity -0.035 -0.012 -0.001 -0.008 -0.016 -0.011 -0.014 -0.012
(0.020) (0.018) (0.012) (0.012) (0.006)

(0.006) (0.005)

(0.004)

Father's education level -0.017 -0.011 0.003 0.001 0.001 0.000 -0.001 -0.000
(0.005)

(0.004)

(0.003) (0.003) (0.001) (0.001) (0.001) (0.001)
Mother's education level -0.001 -0.011 0.001 0.004 0.003 0.002 0.002 0.001
(0.006) (0.005)

(0.003) (0.003) (0.002)

(0.001) (0.001) (0.001)
Log PCE -0.005 -0.018 -0.001 -0.002 0.001 -0.001 -0.000 -0.000
(0.006) (0.004)

(0.003) (0.002) (0.001) (0.001) (0.001) (0.001)
Rural 0.016 -0.000 0.009 -0.004 -0.003 -0.001 -0.001 -0.000
(0.014) (0.008) (0.007) (0.005) (0.002) (0.002) (0.001) (0.001)
Constant 0.557 0.409 0.094 0.030 -0.002 0.012 -0.010 -0.010
(0.088)

(0.070)

(0.049) (0.045) (0.019) (0.020) (0.013) (0.013)
Observations 26646 26645 26651 26650 26741 26740 26740 26739
R
2
0.032 0.269 0.002 0.145 0.004 0.047 0.003 0.076
Community FE No No No No No No No No
Household FE No No No No No No No No
Mean of dependent variable 0.410 0.410 0.080 0.080 0.020 0.020 0.010 0.010
Joint signicance of (p-value):
Age categories 0.005 0.000 0.870 0.580 0.000 0.000 0.149 0.356
Education levels 0.000 0.000 0.907 0.280 0.068 0.088 0.231 0.116
Parents' education 0.000 0.000 0.331 0.171 0.023 0.144 0.163 0.289
Interviewer xed eects 0.000 0.000 0.000 0.000
See previous tables for denition of dependent variables. All regressions are OLS. Omitted category for age: 15-
24, for education: no education. Standard errors in parentheses clustered by community ID. *p<0.10 **p<0.05
***p<0.01.
67
Appendix 3.B Logit Estimates
Table 3A.3: Logit Estimates of Tables 3.2(A) { 3.2(D)
(1) (2) (3) (4) (5) (6) (7) (8)
Risk A Risk B Time A Time B GA A GL B NTD A NTD B
Word recall z-score -0.008 -0.005 -0.017 -0.012 -0.032 -0.010 -0.004 -0.001
(0.007) (0.004) (0.005)

(0.004)

(0.005)

(0.002)

(0.001)

(0.001)

Language eects:
(I): Interview not conducted in Indonesian (d) -0.018 -0.018 0.022 -0.004 0.088 0.015 0.011 0.011
(0.030) (0.016) (0.019) (0.017) (0.020)

(0.011) (0.004)

(0.003)

(II): Interviewer speaks respondent's language (d) -0.016 0.015 -0.008 -0.007 0.005 0.002 0.001 0.001
(0.020) (0.012) (0.011) (0.009) (0.015) (0.006) (0.002) (0.001)
(III): (I) (II) (d) 0.025 -0.024 -0.059 -0.017 -0.096 -0.010 -0.008 -0.006
(0.035) (0.018) (0.022)

(0.018) (0.023)

(0.011) (0.003)

(0.001)

Interviewer human capital:
Past interview experience (d) -0.056 -0.004 -0.053 -0.024 -0.052 0.020 0.002 0.002
(0.021)

(0.011) (0.015)

(0.012)

(0.020)

(0.006)

(0.002) (0.001)

Plan to continue studies (d) -0.005 -0.027 -0.023 -0.007 -0.063 0.011 -0.003 0.001
(0.017) (0.008)

(0.011)

(0.009) (0.015)

(0.006) (0.002) (0.001)
Has bachelor's degree (d) 0.074 0.007 -0.040 -0.016 0.066 0.043 0.005 0.001
(0.027)

(0.014) (0.015)

(0.014) (0.023)

(0.006)

(0.003) (0.001)
Log income in last job (monthly) 0.002 0.002 0.001 0.001 0.008 0.000 0.000 -0.000
(0.002) (0.001)

(0.001) (0.001) (0.001)

(0.001) (0.000) (0.000)

Interviewer social distance:
Interviewer older than respondent (d) -0.057 -0.031 -0.005 0.024 -0.029 0.014 0.002 -0.001
(0.023)

(0.015)

(0.016) (0.012)

(0.017) (0.008) (0.003) (0.001)
Interviewer same sex as respondent (d) -0.016 -0.005 -0.009 -0.007 -0.001 0.003 -0.002 -0.001
(0.009) (0.005) (0.006) (0.005) (0.007) (0.004) (0.002) (0.001)
Observations 13552 20840 22260 22495 22559 22565 22639 22638
Pseudo R
2
0.013 0.022 0.024 0.020 0.031 0.016 0.027 0.063
(II) + (III) 0.033 -0.063 -0.324 -0.158 -0.392 -0.119 -0.549 -1.603
(0.118) (0.111) (0.092)
***
(0.103) (0.086)
***
(0.153) (0.211)
***
(0.304)
***
Community FE No No No No No No No No
Household FE No No No No No No No No
Mean of dependent variable 0.510 0.840 0.710 0.810 0.400 0.080 0.020 0.010
Joint signicance of (p-value):
Age categories 0.046 0.020 0.001 0.000 0.004 0.476 0.059 0.111
Education levels 0.726 0.002 0.000 0.000 0.000 0.843 0.353 0.895
Parents' education 0.109 0.162 0.000 0.569 0.004 0.656 0.036 0.047
Interviewer human capital 0.006 0.015 0.000 0.211 0.000 0.000 0.223 0.105
Interviewer social distance 0.012 0.068 0.368 0.055 0.234 0.166 0.480 0.183
Language eects: (I) and (III) 0.773 0.001 0.008 0.142 0.000 0.340 0.006 0.000
Language eects: (II) and (III) 0.681 0.322 0.002 0.229 0.000 0.705 0.028 0.000
Language eects: (I), (II) and (III) 0.857 0.001 0.001 0.047 0.000 0.534 0.014 0.000
All regressions are logit. Marginal eects reported. (d) indicates dummy variable. Not shown: Male, age categories, education
categories, Muslim, religiosity, Muslim Religiosity, log PCE, rural, father's education, mother's education, constant. Standard
errors in parentheses clustered by community ID. *p<0.10 **p<0.05 ***p<0.01.
68
Chapter 4
Risk and Time Preferences and the Transition into
Adulthood
4.1 Introduction
This essay seeks to answer the following question: Do individual risk and time pref-
erences drive dierences in the transition into adulthood among Indonesians? After
controlling for family background and other socioeconomic characteristics, what role
might the risk aversion and time preference of individuals play in determining pat-
terns of economic behavior? To investigate this question, I draw on data from the
fourth wave of the Indonesian Family Life Survey (IFLS-4).
Adulthood can be thought of as economic and social maturity. The set of deci-
sions made in the transition into adulthood are critical for an individual's economic
well-being over the life course. I consider ve markers of adulthood: schooling attain-
ment, marriage, fertility behavior, migration, and full time employment. If risk and
time preferences are signicantly correlated with the timing of these markers even
after controlling for socioeconomic characteristics and other background variables,
the implication is that socioeconomic characteristics only explain part of observed
dierences in adulthood transition behavior, with at least some role played by prefer-
ences which are traditionally held to be intrinsic and unchanging (Stigler and Becker,
1977). Policymakers may need to factor such preferences in policy instruments aimed
at helping youth transition smoothly into adulthood.
69
The results indicate that risk and time preferences play a signicant role in some
markers of adulthood transition. Men who are more risk averse or impatient, and
women who are more impatient attain fewer years of schooling. Risk averse men and
women marry earlier, and so do impatient women. Married couples positively sort on
risk and time preferences. Risk and time preferences are not signicantly correlated
with female fertility timing. However, high risk aversion and low impatience are
signicantly negatively correlated with birth control use among women. Highly risk
averse men are signicantly less likely to migrate. Risk and time preferences are
not signicantly correlated with entry into full time employment, but conditional on
being employed, highly risk averse men and women are signicantly less likely to be
self employed.
The rest of the essay is organized as follows. Section 4.2 describes the ve key
markers in the transition into adulthood. Attention will be drawn to existing empirical
evidence on the role of risk and time preferences in their determination. Section 4.3
describes the methods used to estimate the eects of risk and time preferences on the
ve markers. Section 4.4 presents the results, and Section 4.6 concludes.
4.2 Dependent Variables
4.2.1 Schooling Attainment
School-leaving is an important life event in the transition into adulthood. The eect
of time preference on schooling has not been studied empirically. However, there
are a number of papers that look at the in
uence of risk aversion on higher edu-
cation attainment (Belzil and Leonardi, 2009; Chen, 2003). In these papers, higher
education is seen as an investment and the decision to invest is partly determined
by attitude towards risk. They nd that risk aversion is negatively associated with
70
higher education attainment.
Care should be taken in thinking about the direction of causality in the preference{
schooling relationship. Although having preference measures as an explanatory vari-
able implies that preferences help in
uence how long a person stays in school, the
direction of causality is by no means clear even on a priori grounds. In our sam-
ple, risk and time preferences were measured after schooling has been completed by
most of the respondents, so a person's preferences could have been shaped by his
amount of schooling. Becker and Mulligan (1997, p. 736) point out that because
schooling \focuses students' attention on the future" and \through repeated practice
at problem-solving,[...] helps children learn the art of scenario simulation", it can
make students become more patient. Their prediction is borne out in an empirical
study on Mexican college students (Perez-Arce, 2011). Schooling may also in
uence
respondents' elicited risk preferences. A number of empirical studies have found that
education, particularly higher education, decreases risk aversion (Knight et al., 2003;
Hartog et al., 2002). In the rst essay using the same IFLS data, I nd that relative to
the uneducated, university-educated respondents have lower rate of time preference
(that is, they are more patient) relative to the uneducated, and weaker evidence that
they have lower risk aversion. This means that estimates of preference eects in the
present study would be overstated.
The average number of years of schooling in the IFLS sample is 8.1 years, with
men having almost one year more schooling than women (Table 4.1). In the regression
analysis, to avoid right censoring from including respondents who were still in school
at survey time, I drop respondents who were younger than 22. Of 2,255 respondents
who were still in school, 93 percent were younger than 22. Of 24,348 respondents who
were at least 22 years of age, only 153 were still in school.
71
4.2.2 Marriage Age
The economics of marriage has been widely studied by economists ever since the
pioneering work of Becker (1973). In Becker's theory of marriage, a utility maximizing
individual matches with a partner in the marriage market if the output as a couple
is at least as large as the sum of their single outputs. Becker shows that gains from
marriage for a man and woman are aected by their incomes, relative dierences in
their wage rates, and human capital. Keeley (1977) builds upon Becker's work by
modeling marriage timing as a sequential search problem. In the model, a single
person enters the marriage market only if the expected benets of search is at least
equal to the costs. Once the person enters, he/she expends resources in search of a
mate; search duration depends on individual characteristics such as education and
wages. In these models, individuals in the marriage market are assumed to be risk
neutral. Schmidt (2008) explicitly considers risk aversion in marriage search, pointing
out that mate seekers who are more risk tolerant are willing to hold out longer for a
mate who meets or exceeds their (higher) 'reservation wage' (i.e. the income stream
from being single). Her paper, and that of Spivey (2010) both nd that highly risk
tolerant women are more likely to delay marriage, using cross-sectional data from the
PSID and NLSY respectively. Spivey nds stronger results for men than women.
The eect of time preference on marriage timing has so far not been studied, either
theoretically or empirically. The only paper I have found to specically look at the
eect of time preference in marital behavior is a working paper by Compton (2009),
who nds in the NLSY sample that individuals with patient characteristics are less
likely to divorce than those with impatient characteristics.
I examine the eect of risk and time preferences on the likelihood of an individual
marrying for the rst time by age x, if the individual is at leastx years old. In other
72
words, the dependent variable is a binary variable taking the value 1 if the individual
is married by agex, and 0 otherwise, for individuals agedx and older. Setting up the
dependent variable this way saves us from having to deal with right censored data,
and permits the estimation of linear probability models. I regress the marriage age
variable on preferences and socioeconomic variables (explained later), separately for
men and women.
The data show that the average at rst marriage is 25 for men and 20 for women
(Table 4.1). Therefore the dependent variable for the male analytical sample is a
dummy for whether a man is married by 25; the cuto age for the female sample is
20.
4.2.3 Fertility Timing
Women's preferences can aect fertility timing through two competing mechanisms.
In essence, women have to weigh the net benet of having a child (rst mechanism)
against the net benet of postponing childbirth (second mechanism). The net eect
of preferences therefore depends on the relative importance of the two mechanisms.
In the rst mechanism, women who are highly risk averse may be less willing to accept
the cost of an unplanned pregnancy and therefore actively limit fertility, thus delaying
childbirth. As for time preference, impatient women may be less willing to incur the
immediate costs of raising a child and more steeply discount the future benets of
having grown children, so they delay births.
However, risk aversion may hasten childbirth as well, especially in less developed
countries where children may serve as a form of insurance. Portner (2001) proposes a
dynamic model of fertility decisions where households with more expected variation
in their income would have more children for a given income, with the children seen as
73
an incomplete insurance good. Although not dealing with the role of risk attitudes per
se in fertility decisions, Portner (2008) looks at the relationship between municipality-
specic hurricane risk in Guatemala and fertility, and nds support for the children-
as-insurance hypothesis. He nds that higher hurricane risk is associated with a
higher number of children for households with land, whereas he nds the opposite
for households without land. Households with land can benet from having more
children to help on the land after a hurricane hits, while households without land do
not benet from this mechanism.
As for the second mechanism, postponing childbirth may allow women to pursue
human capital-enhancing activities more aggressively, such as higher education, but
it carries with it the risk of decreased fecundity with age. Women who are more risk
averse might be more worried about lowered fecundity and therefore might hasten
childbearing. It is less clear what the prior for the time preference eect should be in
the second mechanism.
The data contain 6,347 women with self-reported fertility histories, out of a total
of 15,213 women. The mean age at rst birth is 25.5, with a standard deviation of 6
(Table 4.1). The dependent variable that is a measure of fertility timing is whether
or not a woman gave birth for the rst time by the age of 18. To avoid censoring, I
truncate the sample to include only women who are 18 years old or older.
Birth control use indicates, at least partially, whether a woman is actively limiting
fertility. IFLS asks all ever married women between the ages of 15 and 49 whether
they or their spouse are currently using birth control. On the one hand, risk averse
women would be more likely to use birth control if they want to limit fertility. On the
other hand, if a woman does not want to limit fertility, risk aversion could decrease
birth control use. The implication is that risk aversion eects likely vary by stages in
the lifecycle, of which age is a reasonable proxy.
74
4.2.4 Migration
Economics oers many reasons for why an individual might migrate. An in
uential
early work is the Harris and Todaro (1970) model of rural-to-urban migration. They
posit that the lower the urban{rural wage dierential, the lower is the rate of migra-
tion, and the higher the perceived probability of nding a job in the urban sector,
the greater is the rate of rural-to-urban migration. Stark (1991) shifts the focus of
migration research from migration as an individual decision (as in the Harris-Todaro
line of research) to migration as a collective, household decision. The decision to
migrate therefore is not simply a response to the urban{rural wage dierential. In
these models, a household may place members in dierent localities as part of a wel-
fare maximizing strategy for household income growth, risk sharing or easing credit
constraints in the presence of incomplete or missing markets. Household members
may migrate for work (Hoddinott, 1994), but they may also migrate for (exogamous)
marriage (Rosenzweig and Stark, 1989).
Whatever the motivation, migration is a risky activity, the rewards of which may
take a considerable amount of time to realize. Risk aversion conceivably is negatively
correlated with migration propensity. To date, Jaeger et al. (2010) is the only paper
to directly test this supposition, and they nd support for it in the German Socioeco-
nomic Panel. I test whether the propensity to migrate varies with risk aversion and
time preference. I look at two indicators of migration: whether a respondent had ever
migrated since age 18, and number of migrations since age 18. I omit those whose
only instances of migration happened when they were younger than 18, because they
were likely just following their parents, not migrating for their own purposes.
The correlates of the migration variables are estimated using OLS. A problem
with this approach is the data are right censored, inevitable since migration histories
75
are incomplete for living persons. A more suitable approach is to t a semiparametric
Cox hazard model because it allows right censored data. The time to `failure' is the
number of months elapsed from each person's 18th birthday to the rst instance of
migration. A hazard function of covariates is formed for each individual and entered
into a partial likelihood. The log of the partial likelihood is then maximized to give
estimates of each covariate's eect on the hazard ratio (details in Section 4.3).
4.2.5 Full Time Employment
Closely tied to the completion of schooling attainment is full time employment. The
transition of young adults from dependence on parents to nancial independence
can be measured by the age at which individuals enter into their rst full time job.
Risk aversion has gured prominently in job search models (Pissarides, 1974). Time
preference is a more recent consideration in theoretical and empirical studies of job
search (DellaVigna and Paserman, 2005). The gist of these studies is that risk averse
or impatient job seekers have a lower reservation wage and would therefore more
easily accept a job than less risk averse or more patient individuals.
In the IFLS sample, men on average start their rst full time job at 19.9 years of
age, whereas women start at 20.4.
1
Therefore, whether an individual is working by
the age of 20 can serve as a measure of the timing of the reservation wage decision.
I expect risk aversion and impatience to be positively associated with the dependent
variable. To avoid right censoring on the dependent variable, I restrict the sample to
respondents aged 20 and older.
1
IFLS asks each respondent \When did you start working full time for the rst time?" and \
What was your age when starting to work full time for the rst time?"
76
Table 4.1: Summary statistics
All Men Women
Mean Std Dev Count Mean Std Dev Count Mean Std Dev Count
Years of schooling 8.10 4.36 28755 8.56 4.16 13680 7.68 4.49 15075
Age at marriage (years) 22.36 6.49 22551 24.86 6.20 10085 20.33 5.98 12466
Ever migrated 0.31 0.46 29054 0.33 0.47 13836 0.30 0.46 15218
Age at rst migration (years) 24.41 10.57 9062 24.72 10.36 4555 24.10 10.77 4507
Times migrated 1.88 1.26 9135 2.02 1.37 4587 1.75 1.12 4548
Age at rst full-time employment (years) 20.10 6.75 19062 19.86 6.00 10701 20.41 7.59 8361
Self employed 0.30 0.46 25907 0.40 0.49 12005 0.21 0.41 13902
Age at rst birth (years) 25.50 6.29 6347
Uses birth control 0.56 0.50 9318
Total Number of Respondents 29043 13830 15213
4.3 Empirical Strategy
The results in this paper are mainly obtained via OLS regressions. In all regressions,
I cluster standard errors by IFLS community to allow intra-community correlations.
The baseline regressions are of the form:
y
i
= +p
i
+ X
i
0

+
j
+
i
(4.1)
where y is a variable dened in the previous section. The explanatory variable p is a
measure of risk aversion or time preference of individual i. IFLS provides two sets of
risk aversion measures, Risk A and Risk B, and two sets of time preference measures,
Time A and Time B.
2
. For tractability, I create binary measures of risk aversion
called Most Risk Averse A and Most Risk Averse B. They take the value of 1 if the
corresponding Risk measure is equal to 4, and 0 if it is less than 4 (missing values are
dropped). The binary measures of time preference are Most Impatient A and Most
Impatient B, taking the value of 1 when the corresponding Time measure is equal to
4, and 0 otherwise (missing values are dropped). X is a vector of control variables
which includes age, rural residence, Muslim, ethnicity, and province of residence. In
many cases, we would expect preferences to vary with age. In these cases, I include
age{preference interaction eects.
2
Details are in Strauss et al. (2009).
77
Previous studies have found strong intergenerational eects of parental education
on human capital accumulation (e.g. Black et al., 2005). Parents' education can be
thought of as a proxy for parental resources, which serve as exogenous inputs into
the general development of the child, including the child's transition into adulthood.
I therefore control for parents' educational attainment in all regressions.
I control for community xed eects, in all regressions to account for unob-
served characteristics that may aect each IFLS community. I run each regression
separately for men and women because they face dierent constraints in decision
making. Furthermore, IFLS women are on average more risk averse than men, as
shown in Chapter 2.
I am able to circumvent the problem of right censored data for most of the depen-
dent variables, with one exception: whether an individual migrated since turning 18.
Because migration data is right censored, OLS regression would result in coecient
estimates that are biased towards zero (Greene, 2008). The Cox proportional hazards
model is well suited to modeling censored duration data. The hazard function at time
t, h(t) assesses the instantaneous probability of migration at time t, conditional on
not having migrated to that point. In the Cox model, the hazard function at baseline
(18th birthday), h
0
(t) is left unspecied and the relationship between hazard and
regressors is:
h(t) =h
0
(t)exp(X) (4.2)
The linear combination of regressors enter exponentially to ensure thath(t) is always
non-negative. h(t) is shifted up or down relative toh
0
(t) by an order of proportionality
with changes in the regressors. This is the proportional hazards assumption.
The Cox model is estimated via maximum (partial) likelihood. Intuitively, the
78
partial likelihood is a product over the set of observed migration times of the condi-
tional probabilities of the migration events. At each migration time t
i
, given that t
i
is one of the observed migration times from the set R(t
i
), the conditional probability
that individual i migrates is:
Pr(individual i migrates at t
i
j has not migrated up until t
i
)
Pr(one migration at t
i
j has not migrated up until t
i
)
(4.3)
Because individuals are assumed to respond independently, (4.3) can be rewritten
as:
Pr(individual i migrates at t
i
)
P
k2R(t
i
)
Pr(individual k migrates at t
i
)
(4.4)
which is equivalent to:
h
i
(t
i
)
P
k2R(t
i
)
h
k
(t
i
)
=
h
0
(t
i
)exp(
i
X
i
)
P
k2R(t
i
)
h
0
(t
i
)exp(
k
X
k
)
(4.5)
The baseline hazards cancel out, so each observation's contribution to the likelihood
is:
exp(
i
X
i
)
P
k2R(t
i
)
exp(
k
X
k
)
(4.6)
The joint partial likelihood is then:
L() =
n
Y
i=1
"
exp(
i
X
i
)
P
k2R(t
i
)
exp(
k
X
k
)
#

i
(4.7)
where
i
= 0 ift
i
is a censoring time and 1 otherwise. The maximum partial likelihood
estimators can be found by solving
@
@
logL() = 0.
79
If we are to interpret risk and time preferences as causal in these regressions, we
will have to assume that preferences are time invariant. Recent research has found
that risk aversion is not immutable but can be changed by a shock such as a natural
disaster (Cassar et al., 2011; Cameron and Shah, 2012) or nancial crisis (Guiso et
al., 2011). Even in the absence of such shocks, preferences could be aected by the
dependent variables (e.g. years of schooling in Subsection 4.2.1) in ways not captured
by the other covariates in the models. Therefore, the preference estimates should be
seen as correlates of the dependent variables, holding the most obvious demographic
variables constant.
4.4 Results
4.4.1 Schooling
Table 4.2 presents the results for years of schooling completed for men and women
22 years old and older. Risk aversion has a signicantly negative eect on years of
schooling for men but is insignicant for women. Time preference has a signicantly
negative eect for both men and women. Overall, the evidence suggests that less
schooling is attained by risk averse men, impatient men, and impatient women, al-
though the magnitude of the eects is small (less than a year). Risk aversion does
not seem to play a signicant role in women's schooling attainment.
The regressions control for age, rural residence, whether the respondent is a Mus-
lim, religiosity, and the education of level of the father and mother. Across all speci-
cations, age is signicantly negative related to years of schooling, consistent with the
observation that younger generations had better access to education (Du
o, 2001). As
one would expect, rural residents have signicantly less schooling than urban counter-
parts. Parents' education is signicantly positively associated with years of schooling.
80
Table 4.2: Years of Schooling
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Highly Risk Averse A -0.239 0.002
(0.079)

(0.077)
Highly Risk Averse B -0.571 -0.104
(0.082)

(0.082)
Most Impatient A -0.356 -0.175
(0.063)

(0.061)

Most Impatient B -0.418 -0.155
(0.080)

(0.072)

Age (years) -0.045 -0.050 -0.050 -0.051 -0.087 -0.093 -0.094 -0.094
(0.004)

(0.003)

(0.003)

(0.003)

(0.004)

(0.003)

(0.003)

(0.003)

Rural -0.935 -0.877 -0.787 -0.793 -1.012 -0.772 -0.712 -0.722
(0.224)

(0.188)

(0.172)

(0.172)

(0.286)

(0.202)

(0.201)

(0.199)

Muslim -0.612 -0.542 -0.778 -0.710 -0.558 -0.460 -0.349 -0.344
(0.384) (0.353) (0.329)

(0.329)

(0.501) (0.434) (0.391) (0.400)
Religiosity 0.545 0.584 0.471 0.493 0.404 0.471 0.510 0.516
(0.181)

(0.168)

(0.159)

(0.157)

(0.211) (0.200)

(0.174)

(0.178)

Muslim Religiosity -0.214 -0.153 -0.033 -0.068 -0.148 -0.198 -0.244 -0.255
(0.193) (0.175) (0.166) (0.164) (0.229) (0.213) (0.187) (0.190)
Father's education:
Elementary 1.843 1.665 1.672 1.686 1.645 1.573 1.538 1.549
(0.146)

(0.117)

(0.114)

(0.114)

(0.137)

(0.098)

(0.096)

(0.096)

Junior High 3.005 2.745 2.782 2.802 2.766 2.734 2.732 2.718
(0.184)

(0.156)

(0.149)

(0.151)

(0.195)

(0.147)

(0.145)

(0.144)

Senior High/University 3.267 3.079 3.104 3.093 3.408 3.415 3.406 3.394
(0.194)

(0.166)

(0.162)

(0.163)

(0.188)

(0.146)

(0.141)

(0.141)

Mother's education:
Elementary 0.930 0.926 0.927 0.921 1.534 1.411 1.408 1.413
(0.135)

(0.110)

(0.107)

(0.107)

(0.114)

(0.086)

(0.083)

(0.083)

Junior High 1.222 1.203 1.242 1.258 2.248 2.121 2.170 2.192
(0.187)

(0.152)

(0.150)

(0.150)

(0.171)

(0.141)

(0.134)

(0.132)

Senior High/University 1.352 1.517 1.493 1.538 2.176 2.297 2.288 2.272
(0.207)

(0.179)

(0.177)

(0.175)

(0.201)

(0.160)

(0.153)

(0.152)

Constant 8.614 8.802 8.768 8.801 8.987 8.976 8.907 8.902
(0.417)

(0.391)

(0.356)

(0.362)

(0.537)

(0.470)

(0.418)

(0.423)

Observations 8185 11844 12764 12915 7766 13217 14179 14332
R
2
0.466 0.463 0.460 0.459 0.594 0.595 0.598 0.596
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 9.020 8.610 8.600 8.600 8.290 7.790 7.720 7.730
Joint signicance of (p-value):
Education categories
Father's education 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Mother's education 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Each column is for a separate regression. All regressions are OLS. Dependent variable is years of schooling.
Omitted category for education is no education. Religiosity is an ordinal variable from 0 (not religious) to 3 (very
religious). Standard errors in parentheses clustered by community. *p<0.10 **p<0.05 ***p<0.01.
For both father and mother, the positive eect is monotonically increasing across ed-
ucational levels, and the educational levels are jointly signicant. This speaks to the
importance of parental resources in the child's human capital accumulation.
4.4.2 Marriage
Assortative Mating
Ever since Becker's theory of marriage (Becker, 1973), economists have studied match-
ing and sorting patterns in marriage markets. The literature has documented sorting
along attributes such as age, education levels and physical traits (see Browning et al.,
81
Table 4.3: Positive Assortative Mating on Preferences
(1) (2) (3) (4) (5) (6) (7) (8)
Wife's
Risk A
Wife's
Risk B
Wife's
Time A
Wife's
Time B
Wife's
GA A
Wife's
GL B
Wife's
NTD A
Wife's
NTD B
Husband's:
Risk A 0.112
(0.035)

Risk B 0.130
(0.021)

Time A 0.143
(0.020)

Time B 0.157
(0.021)

GA A 0.105
(0.019)

GL B 0.103
(0.023)

NTD A 0.024
(0.028)
NTD B 0.038
(0.035)
Observations 1672 3865 4455 4524 4557 4551 4593 4593
R
2
0.357 0.223 0.165 0.177 0.196 0.153 0.112 0.094
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.510 0.880 0.740 0.830 0.470 0.080 0.020 0.010
Joint signicance of (p-value):
Wife's age categories 0.990 0.404 0.366 0.530 0.157 0.162 0.353 0.375
Wife's father's edu 0.532 0.933 0.795 0.454 0.592 0.063 0.954 0.918
Wife's mother's edu
Wife's education 0.481 0.048 0.004 0.043 0.052 0.463 0.542 0.638
Husband's age categories 0.893 0.985 0.189 0.222 0.068 0.113 0.423 0.409
Husband's father's edu 0.193 0.686 0.229 0.593 0.400 0.755 0.180 0.077
Regression controlled for the following husband and wife variables: age categories, highest education attained,
dummy for Muslim, dummy for rural residence, highest education attained by a parent. Risk = 1 for category
4 (most risk averse), and 0 for categories 1-3 (less risk averse). GA = 1 if gamble averse, 0 otherwise. GL =
1 if gamble loving, 0 otherwise. NTD = 1 if negative time discounting, 0 otherwise. All regressions are OLS.
Omitted category for age: 15-24, for education: no education. Standard errors in parentheses clustered by
community ID. *p<0.10 **p<0.05 ***p<0.01.
2011 for a survey). However, evidence on sorting along intrinsic preferences such as
risk aversion and impatience is comparatively uncharted territory.
Table 4.3 presents evidence of positive assortative mating on the preference mea-
sures within married couples. The signicant, positive correlation between husband's
preference and wife's preference stands even after controlling for a host of husband
and wife characteristics. These results are in agreement with the ndings of Dohmen
et al. (2012), the only other paper to examine assortative mating on preferences
(specically, on risk and trust attitudes).
Marriage Timing
Table 4.4, columns 1{4 present the results for the propensity of men to be married
for the rst time by age 25. There is some evidence that risk averse men are more
likely to marry by age 25. Columns 5{8 show the corresponding results for women.
82
Table 4.4: Marriage Timing
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A 0.016 -0.009
(0.015) (0.016)
Most Risk Averse B 0.037 0.033
(0.016)

(0.016)

Most Impatient A 0.019 0.028
(0.013) (0.012)

Most Impatient B 0.013 0.025
(0.014) (0.013)
Age (years) -0.006 -0.006 -0.006 -0.006 -0.001 -0.002 -0.002 -0.002
(0.001)

(0.001)

(0.000)

(0.000)

(0.001) (0.000)

(0.000)

(0.000)

Rural 0.028 0.026 0.021 0.019 0.048 0.037 0.042 0.043
(0.036) (0.026) (0.024) (0.024) (0.043) (0.033) (0.032) (0.032)
Muslim 0.126 0.066 0.037 0.036 0.021 0.025 -0.033 -0.031
(0.090) (0.077) (0.069) (0.069) (0.096) (0.072) (0.070) (0.071)
Religiosity 0.053 0.017 0.006 0.010 -0.015 -0.030 -0.058 -0.055
(0.042) (0.036) (0.032) (0.032) (0.044) (0.031) (0.029)

(0.030)
Muslim Religiosity -0.037 -0.002 0.009 0.005 0.028 0.028 0.057 0.056
(0.044) (0.037) (0.033) (0.033) (0.046) (0.033) (0.032) (0.033)
Father's education:
Elementary -0.063 -0.068 -0.072 -0.073 -0.068 -0.051 -0.043 -0.042
(0.020)

(0.016)

(0.016)

(0.016)

(0.023)

(0.017)

(0.016)

(0.016)

Junior High -0.133 -0.135 -0.133 -0.135 -0.161 -0.122 -0.124 -0.122
(0.036)

(0.028)

(0.027)

(0.026)

(0.035)

(0.024)

(0.024)

(0.023)

Senior High/University -0.142 -0.153 -0.164 -0.161 -0.207 -0.175 -0.169 -0.171
(0.034)

(0.027)

(0.026)

(0.026)

(0.035)

(0.024)

(0.024)

(0.024)

Mother's education:
Elementary -0.049 -0.057 -0.052 -0.051 0.007 -0.015 -0.018 -0.017
(0.019)

(0.016)

(0.015)

(0.015)

(0.020) (0.015) (0.014) (0.014)
Junior High -0.125 -0.146 -0.138 -0.137 -0.067 -0.098 -0.104 -0.105
(0.039)

(0.032)

(0.030)

(0.030)

(0.034)

(0.026)

(0.025)

(0.025)

Senior High/University -0.095 -0.141 -0.135 -0.138 -0.086 -0.145 -0.147 -0.144
(0.041)

(0.036)

(0.033)

(0.033)

(0.038)

(0.027)

(0.027)

(0.026)

Constant 0.751 0.816 0.862 0.863 0.559 0.583 0.645 0.636
(0.096)

(0.082)

(0.072)

(0.073)

(0.101)

(0.073)

(0.070)

(0.071)

Observations 5540 8083 8762 8841 5504 9612 10316 10393
R
2
0.163 0.139 0.134 0.132 0.158 0.133 0.129 0.128
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.570 0.590 0.590 0.580 0.480 0.500 0.500 0.500
Joint signicance of (p-value):
Father's education 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Mother's education 0.006 0.000 0.000 0.000 0.017 0.000 0.000 0.000
Each column is for a separate regression. All regressions are OLS. Dependent variable for male (female) sample is
a dummy for being married by 25 (20). Omitted category for education is no education. Religiosity is an ordinal
variable from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered by community.
*p<0.10 **p<0.05 ***p<0.01.
Because women on average are married by 20, the dependent variable is married by
20. Women who are most risk averse (as measured by Submodule B) or impatient
(in Submodule A) are signicantly more likely to be married. Overall, there is some
support for the hypothesis that marriage market participants who are more risk averse
have a lower reservation wage.
The education level of the father and mother have jointly signicant eects on
the likelihood of marriage for both men and women. The higher the education of
either parent, the lower the likelihood of the respondent marrying by 25. Perhaps
surprisingly, age has a signicant negative relationship with the marriage variables for
both men and women, indicating that younger cohorts are more likely to be married
83
at the cuto age.
4.4.3 Fertility
First Birth
Table 4.5 shows the estimates for the correlates of having a rst birth by age 18, for
ever married women between the ages of 18{49. Risk aversion is negatively correlated
with the dependent variable, suggesting that women who are more risk averse start a
family earlier. Time preference has a positive sign, suggesting that impatient women
delay childbirth. However, none of the preference eects are statistically signicant.
Table 4.5: Likelihood of First Birth by Age 18
(1) (2) (3) (4)
Most Risk Averse A -0.004
(0.011)
Most Risk Averse B -0.006
(0.013)
Most Impatient A 0.005
(0.009)
Most Impatient B 0.015
(0.010)
Age (years) -0.004 -0.004 -0.004 -0.004
(0.001)

(0.001)

(0.001)

(0.001)

Rural 0.026 0.025 0.018 0.020
(0.022) (0.015) (0.014) (0.014)
Muslim 0.006 -0.024 -0.045 -0.031
(0.036) (0.033) (0.031) (0.028)
Religiosity -0.012 -0.012 -0.025 -0.017
(0.015) (0.015) (0.014) (0.012)
Muslim Religiosity 0.032 0.028 0.040 0.034
(0.018) (0.017) (0.017)

(0.015)

Father's education:
Elementary -0.005 -0.017 -0.006 -0.006
(0.023) (0.017) (0.017) (0.017)
Junior High -0.020 -0.027 -0.014 -0.014
(0.029) (0.022) (0.021) (0.021)
Senior High/University -0.061 -0.058 -0.047 -0.048
(0.029)

(0.023)

(0.022)

(0.022)

Mother's education:
Elementary -0.020 -0.023 -0.030 -0.029
(0.019) (0.015) (0.015)

(0.015)

Junior High -0.030 -0.047 -0.051 -0.050
(0.025) (0.022)

(0.021)

(0.021)

Senior High/University -0.010 -0.050 -0.049 -0.048
(0.028) (0.022)

(0.021)

(0.021)

Constant 0.195 0.254 0.268 0.244
(0.056)

(0.046)

(0.043)

(0.040)

Observations 3289 5563 5926 5962
R
2
0.164 0.118 0.117 0.116
Community FE Yes Yes Yes Yes
Mean of dependent variable 0.090 0.100 0.100 0.100
Joint signicance of (p-value):
Father's education 0.015 0.024 0.019 0.015
Mother's education 0.470 0.114 0.092 0.102
Each column is for a separate regression. All regressions are OLS. Dependent variable is a dummy for having a rst
birth by age 18. Sample consists of ever married women ages 18{49. Omitted category for education is no education.
Religiosity is an ordinal variable from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered by
community. *p<0.10 **p<0.05 ***p<0.01.
84
Interestingly, there seems to be a signicant interaction eect between whether
a woman is a Muslim and her self-assessed religiosity. In particular, Muslim women
who are more religious are more likely to have a child by 18. Also interesting is
the signicance of father's and mother's education level. The evidence indicates that
having parents who are more highly educated is signicantly negatively associated
with the likelihood of having a rst birth by 18.
Birth Control
Variation in age at rst birth does not shed light on whether a woman is actively
controlling her fertility. Birth control use should also be looked at for a more complete
picture of fertility timing. Table 4.6 presents the results for whether a woman (or her
husband) is currently using birth control. Because fertility behavior is likely to depend
on where a woman's position in her life cycle, the regressions include interactions of
preferences with age.
There is some evidence that preferences aect the use of birth control. Looking
at the second-to-last row in Table 4.6, the total eect of Most Risk Averse B on the
dependent variable is signicantly negative: highly risk averse women are less likely
to use birth control. This result suggests that on average, risk averse women seem to
act in a manner which hastens childbirth. As for time preference, the total eect of
Most Impatient B is signicantly positive, suggesting that highly impatient women
are more likely to contracept. One way to interpret this result is that in Indonesian
families, childbirth is valued highly, and patient women are willing to incur the costs
of childbearing now, while impatient women prefer to delay incurring these costs in
favor of gratication in the present.
There is weaker evidence that preference eects vary signicantly with age. The
interaction between Most Risk Averse A and age, and the interaction between Most
85
Table 4.6: Likelihood of Using Birth Control
(1) (2) (3) (4)
Most Risk Averse A -0.032
(0.062)
Most Risk Averse B -0.134
(0.068)
Most Impatient A 0.086
(0.051)
Most Impatient B 0.131
(0.056)

Age (years) -0.005 -0.010 -0.005 -0.003
(0.001)

(0.002)

(0.001)

(0.002)

Most Risk Averse A Age 0.001
(0.002)
Most Risk Averse B Age 0.004
(0.002)
Most Impatient A Age -0.002
(0.001)
Most Impatient B Age -0.004
(0.002)

Rural -0.023 -0.033 -0.032 -0.032
(0.041) (0.029) (0.028) (0.028)
Muslim 0.205 0.110 0.081 0.068
(0.115) (0.087) (0.084) (0.083)
Religiosity 0.091 0.030 0.014 0.004
(0.052) (0.042) (0.038) (0.038)
Muslim Religiosity -0.079 -0.027 -0.010 -0.003
(0.055) (0.042) (0.039) (0.039)
Father's education:
Elementary 0.025 0.027 0.016 0.017
(0.025) (0.019) (0.017) (0.017)
Junior High -0.041 -0.017 -0.033 -0.034
(0.040) (0.028) (0.026) (0.026)
Senior High/University -0.038 -0.002 -0.017 -0.019
(0.038) (0.028) (0.026) (0.026)
Mother's education:
Elementary 0.014 0.009 0.020 0.017
(0.022) (0.016) (0.015) (0.015)
Junior High 0.015 0.003 0.017 0.015
(0.035) (0.027) (0.026) (0.026)
Senior High/University 0.030 0.012 0.022 0.021
(0.044) (0.033) (0.032) (0.032)
Constant 0.505 0.776 0.625 0.602
(0.129)

(0.106)

(0.094)

(0.099)

Observations 4810 8222 8779 8846
R
2
0.137 0.091 0.093 0.093
Most Risk Averse + Most Risk AverseAge -0.031 -0.130
(0.060) (0.066)
*
Most Impatient + Most ImpatientAge 0.084 0.127
(0.049)
*
(0.054)
**
Community FE Yes Yes Yes Yes
Mean of dependent variable 0.560 0.560 0.560 0.560
Joint signicance of (p-value):
Father's education 0.058 0.088 0.064 0.055
Mother's education 0.890 0.941 0.633 0.731
Each column is for a separate regression. All regressions are OLS. Dependent
variable is a dummy for using birth control. Sample consists of women ages 18{49.
Omitted category for education is no education. Religiosity is an ordinal variable
from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered
by community. *p<0.10 **p<0.05 ***p<0.01.
Risk Averse B and age, are positive but not signicant (although the latter interaction
term is almost signicant at 10 percent). At older ages, highly risk averse women are
more likely to use birth control, which makes sense since late motherhood is risky
for biological, nancial and other pragmatic reasons. The interplay between time
preference and age is dicult to interpret: the interaction between Most Impatient
B and age is signicantly negative.
86
Women's education categories are jointly signicant across all specications. Women
belonging to all three education attainment categories are signicantly more likely to
use birth control relative to women with no education, although the eects are not
quite monotonic across the education levels. Interestingly, father's education levels
are jointly signicant in all regressions, although the individual coecients are all
insignicant. Mother's education categories however are not jointly signicant.
4.4.4 Migration
Propensity to Migrate
Let us rst look at the correlates of the propensity to migrate. The dependent variable
is a binary variable coded 1 if an individual had ever migrated since turning 18, and 0
otherwise. The rst four columns of Table 4.7 presents the results for men. High risk
aversion is negatively associated with propensity to migrate, but only B is signicant.
This is consistent with people viewing migration as a risky venture. The impatience
measures are also positively signed, but insignicant. Interestingly, rural men (at the
time of interview) are signicantly less likely than urban men to migrate. Table 4.7,
columns 5{8 present the results for women. Risk aversion and time preference are
insignicant predictors of the propensity to migrate. Rural residence is also negatively
associated with migration, although it is mostly insignicant.
Interestingly, older respondents are less likely to have ever migrated. Father's
education is also a strong predictor of migration propensity for both men and women:
the coecients for father's education levels are mostly signicantly positive and are
all jointly signicant.
87
Table 4.7: Ever Migrate
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.011 0.008
(0.011) (0.011)
Most Risk Averse B -0.032 -0.001
(0.010)

(0.010)
Most Impatient A 0.011 0.002
(0.008) (0.008)
Most Impatient B 0.007 -0.003
(0.009) (0.009)
Age (years) -0.002 -0.002 -0.002 -0.001 -0.002 -0.002 -0.002 -0.002
(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

Rural -0.052 -0.079 -0.070 -0.067 -0.087 -0.040 -0.041 -0.046
(0.030) (0.022)

(0.023)

(0.022)

(0.034)

(0.023) (0.022) (0.022)

Muslim -0.073 -0.101 -0.100 -0.088 -0.044 0.001 -0.006 -0.003
(0.057) (0.045)

(0.043)

(0.041)

(0.074) (0.050) (0.047) (0.048)
Religiosity -0.066 -0.066 -0.069 -0.065 -0.025 -0.006 -0.009 -0.005
(0.026)

(0.020)

(0.019)

(0.018)

(0.033) (0.023) (0.022) (0.022)
Muslim Religiosity 0.060 0.062 0.062 0.059 0.028 0.004 0.010 0.007
(0.027)

(0.021)

(0.020)

(0.019)

(0.035) (0.024) (0.023) (0.023)
Father's education:
Elementary 0.026 0.022 0.025 0.027 0.023 0.021 0.019 0.018
(0.014) (0.011)

(0.010)

(0.010)

(0.013) (0.010)

(0.010)

(0.010)
Junior High 0.069 0.063 0.065 0.067 0.068 0.051 0.055 0.051
(0.023)

(0.018)

(0.018)

(0.018)

(0.022)

(0.017)

(0.017)

(0.017)

Senior High/University 0.034 0.039 0.037 0.037 0.052 0.056 0.055 0.051
(0.024) (0.020) (0.020) (0.020) (0.023)

(0.017)

(0.017)

(0.017)

Mother's education:
Elementary 0.016 0.013 0.015 0.016 0.017 0.020 0.019 0.019
(0.015) (0.011) (0.011) (0.011) (0.014) (0.010)

(0.009)

(0.009)

Junior High 0.018 0.012 0.020 0.021 0.002 0.006 0.008 0.009
(0.026) (0.020) (0.020) (0.020) (0.023) (0.019) (0.018) (0.018)
Senior High/University -0.034 -0.035 -0.022 -0.015 0.002 0.035 0.034 0.034
(0.028) (0.022) (0.022) (0.021) (0.028) (0.021) (0.021) (0.021)
Constant 0.420 0.453 0.417 0.401 0.343 0.281 0.285 0.286
(0.060)

(0.049)

(0.046)

(0.045)

(0.076)

(0.052)

(0.049)

(0.049)

Observations 8277 11973 12907 13062 7834 13346 14311 14465
R
2
0.211 0.207 0.203 0.202 0.230 0.219 0.215 0.216
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.270 0.250 0.250 0.250 0.230 0.220 0.220 0.220
Joint signicance of (p-value):
Father's education 0.026 0.008 0.003 0.002 0.020 0.004 0.003 0.005
Mother's education 0.103 0.036 0.085 0.122 0.569 0.142 0.169 0.161
Each column is for a separate regression. All regressions are OLS. Dependent variable is a dummy for ever migrating
since age 18. Omitted category for education is no education. Religiosity is an ordinal variable from 0 (not religious)
to 3 (very religious). Standard errors in parentheses clustered by community. *p<0.10 **p<0.05 ***p<0.01.
Migration Frequency
Next, I use a ner measure of migration propensity | the number of migrations since
age 18 as reported by the individual. Broadly speaking, the results are qualitatively
similar to Table 4.7, suggesting that the ever migrate binary variable adequately
captures migration propensity. The results are included in the appendix to this
chapter.
Hazard Model of Migration
Data for the migration dependent variable are right censored, that is, migration in-
formation is available only up until the time of interview even though respondents
88
Table 4.8: Hazard Model of Time to First Migration
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.052 0.081
(0.058) (0.066)
Most Risk Averse B -0.151 0.001
(0.056)

(0.070)
Most Impatient A 0.031 -0.007
(0.045) (0.051)
Most Impatient B 0.018 0.003
(0.051) (0.059)
Age (years) -0.032 -0.032 -0.032 -0.031 -0.041 -0.043 -0.043 -0.043
(0.002)

(0.002)

(0.002)

(0.002)

(0.003)

(0.002)

(0.002)

(0.002)

Rural -0.107 -0.081 -0.021 0.029 -0.093 0.048 0.043 0.029
(0.329) (0.249) (0.219) (0.224) (0.474) (0.273) (0.273) (0.270)
Muslim -0.410 -0.743 -0.762 -0.675 -0.315 0.018 -0.023 0.017
(0.348) (0.269)

(0.265)

(0.252)

(0.449) (0.301) (0.279) (0.285)
Religiosity -0.345 -0.407 -0.443 -0.414 -0.050 0.037 0.011 0.046
(0.171)

(0.126)

(0.125)

(0.120)

(0.195) (0.136) (0.127) (0.130)
Muslim Religiosity 0.328 0.404 0.420 0.396 0.088 -0.036 -0.007 -0.030
(0.177) (0.131)

(0.129)

(0.124)

(0.207) (0.142) (0.134) (0.136)
Father's education:
Elementary 0.132 0.203 0.184 0.192 0.251 0.218 0.205 0.196
(0.101) (0.080)

(0.076)

(0.076)

(0.117)

(0.084)

(0.082)

(0.082)

Junior High 0.265 0.406 0.363 0.369 0.466 0.378 0.406 0.388
(0.129)

(0.103)

(0.099)

(0.099)

(0.152)

(0.116)

(0.112)

(0.111)

Senior High/University 0.176 0.272 0.232 0.233 0.359 0.336 0.298 0.283
(0.136) (0.112)

(0.107)

(0.106)

(0.152)

(0.115)

(0.112)

(0.110)

Mother's education:
Elementary 0.059 0.027 0.029 0.035 -0.090 -0.008 -0.012 -0.006
(0.086) (0.070) (0.064) (0.063) (0.100) (0.076) (0.071) (0.071)
Junior High 0.143 0.005 0.035 0.044 -0.165 -0.127 -0.127 -0.111
(0.138) (0.108) (0.104) (0.103) (0.130) (0.112) (0.104) (0.104)
Senior High/University -0.084 -0.195 -0.116 -0.083 -0.119 0.002 -0.011 0.001
(0.138) (0.115) (0.107) (0.105) (0.155) (0.117) (0.114) (0.114)
Observations 8261 11955 12888 13043 7823 13320 14282 14436
Pseudo R
2
0.039 0.039 0.037 0.036 0.053 0.056 0.056 0.055
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Joint signicance of (p-value):
Father's education 0.236 0.001 0.004 0.003 0.023 0.009 0.004 0.006
Mother's education 0.321 0.117 0.382 0.511 0.649 0.521 0.521 0.581
Each column is for a separate regression. All regressions are Cox proportional hazard models. Time to rst
migration is measured in months since 18th birthday. Omitted category for education is no education. Religiosity
is an ordinal variable from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered by
community. *p<0.10 **p<0.05 ***p<0.01.
could have moved after that. I therefore reestimate the ever migrate models of Table
4.7 in a Cox proportional hazards model framework.
Table 4.8 presents the estimated eects of preferences and other covariates on
the hazard rate of migration. Elapsed time to the rst instance of migration is
measured in months since turning 18. Each column in this table is comparable to
its counterpart in Table 4.7, the OLS estimates. The results are broadly similar to
the OLS estimates. Highly Risk Averse B has a negative and signicant eect on
the hazard for men (column 2). Father's education categories are jointly signicant
and the individual eects are mostly signicantly positive. The corresponding results
for women (columns 5{8) also do not show much deviation from the OLS pattern of
89
results. Women's preferences are not signicantly correlated with the hazard rate.
Overall, there is some evidence that risk aversion is negatively related to migration
and no evidence that impatience is signicantly correlated with migration.
4.4.5 Employment
First Full Time Job
Table 4.9: Likelihood of Full Time Employment
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A 0.018 -0.024
(0.015) (0.019)
Most Risk Averse B 0.034 0.019
(0.015)

(0.020)
Most Impatient A -0.006 -0.029
(0.012) (0.016)
Most Impatient B 0.007 -0.011
(0.015) (0.017)
Age (years) -0.009 -0.009 -0.009 -0.009 -0.009 -0.009 -0.009 -0.009
(0.001)

(0.000)

(0.000)

(0.000)

(0.001)

(0.001)

(0.001)

(0.001)

Rural -0.126 -0.039 0.014 0.026 -0.161 0.093 0.020 0.040
(0.160) (0.118) (0.106) (0.109) (0.239) (0.119) (0.127) (0.124)
Muslim 0.067 0.053 0.056 0.048 0.017 0.127 0.078 0.094
(0.087) (0.061) (0.060) (0.060) (0.136) (0.107) (0.104) (0.101)
Religiosity -0.026 -0.017 -0.015 -0.015 -0.049 0.006 -0.011 -0.003
(0.040) (0.028) (0.026) (0.026) (0.056) (0.047) (0.044) (0.043)
Muslim Religiosity 0.007 -0.001 -0.005 -0.004 0.027 -0.040 -0.020 -0.029
(0.043) (0.030) (0.028) (0.028) (0.059) (0.050) (0.047) (0.045)
Father's education:
Elementary -0.083 -0.073 -0.077 -0.076 -0.059 -0.021 -0.021 -0.017
(0.021)

(0.018)

(0.017)

(0.017)

(0.027)

(0.020) (0.019) (0.019)
Junior High -0.172 -0.163 -0.166 -0.170 -0.079 -0.031 -0.041 -0.033
(0.034)

(0.028)

(0.027)

(0.026)

(0.046) (0.033) (0.031) (0.031)
Senior High/University -0.156 -0.183 -0.183 -0.181 -0.084 -0.069 -0.076 -0.066
(0.038)

(0.032)

(0.030)

(0.030)

(0.049) (0.038) (0.037)

(0.037)
Mother's education:
Elementary -0.036 -0.028 -0.021 -0.019 -0.014 -0.026 -0.030 -0.032
(0.021) (0.017) (0.016) (0.016) (0.029) (0.020) (0.019) (0.019)
Junior High -0.068 -0.043 -0.031 -0.026 -0.046 -0.079 -0.086 -0.088
(0.040) (0.031) (0.030) (0.030) (0.049) (0.038)

(0.036)

(0.036)

Senior High/University -0.088 -0.073 -0.066 -0.065 -0.164 -0.165 -0.172 -0.173
(0.049) (0.041) (0.039) (0.039) (0.059)

(0.043)

(0.041)

(0.041)

Constant 1.081 1.026 1.023 1.011 1.151 0.920 1.032 0.997
(0.113)

(0.085)

(0.078)

(0.079)

(0.169)

(0.120)

(0.119)

(0.115)

Observations 5986 8728 9442 9537 3973 6896 7417 7484
R
2
0.297 0.250 0.243 0.242 0.329 0.276 0.267 0.267
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.580 0.580 0.580 0.580 0.610 0.610 0.600 0.600
Joint signicance of (p-value):
Father's education 0.000 0.000 0.000 0.000 0.134 0.344 0.216 0.347
Mother's education 0.144 0.192 0.317 0.341 0.044 0.002 0.000 0.000
Each column is for a separate regression. All regressions are OLS. Dependent variable is a dummy for full time
employment by age 20. Omitted category for education is no education. Religiosity is an ordinal variable from 0
(not religious) to 3 (very religious). Standard errors in parentheses clustered by community. *p<0.10 **p<0.05
***p<0.01.
Table 4.9, columns 1{4 display the estimates of the correlates of the propensity
to be in full time employment by age 20, for all men aged 20 or older. Risk aversion
and time preference are not signicant predictors of working age, with the exception
90
of Most Risk Averse B. The last four columns present the corresponding results for
women. None of the preference measures are signicant for women.
Interestingly, the eect of parental education seems to vary along gender lines.
In the male sample, father's education levels are jointly signicant, but mother's
education levels are not. Men with more highly educated fathers are signicantly less
likely to be in full time employment by age 20, relative to men with fathers who had
no education. In contrast, mother's education levels are jointly signicant for women,
but not fathers.
Self Employment
Although risk and time preferences do not appear to matter much in the transition to
full time employment, they could play a role in occupational choice. Here, I look at
the relationship between self employment and preferences. A substantial proportion
of Indonesians are self employed. Of employed men, 40 percent are self employed. The
proportion self employed among employed women is much lower at 20 percent, which
according to Croson and Gneezy (2009) is an empirical regularity. The negative eect
of risk aversion for men is well established in the literature (see e.g. van Praag and
Cramer, 2001; Hartog et al., 2002; Ahn, 2010). However, there is very little empirical
evidence on the relationship between risk aversion and occupational choice among
women.
Table 4.10 presents estimates of risk aversion eects on the likelihood of self em-
ployment for employed men (columns 1 and 2) and employed women (columns 5 and
6). Highly risk averse men are signicantly negatively associated with the likelihood
of being self employed, consistent with the usual nding in the literature. The same
eect is observed for women.
The relationship between time preference and selection into self employment has
91
Table 4.10: Likelihood of Self Employment
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.041 -0.037
(0.012)

(0.011)

Most Risk Averse B -0.041 -0.039
(0.012)

(0.012)

Most Impatient A 0.005 -0.008
(0.010) (0.008)
Most Impatient B 0.013 -0.007
(0.011) (0.009)
Age (years) 0.010 0.009 0.009 0.009 0.006 0.007 0.007 0.007
(0.001)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

Rural 0.085 0.062 0.060 0.056 0.021 0.022 0.022 0.021
(0.029)

(0.023)

(0.022)

(0.022)

(0.025) (0.016) (0.016) (0.016)
Muslim -0.025 0.001 -0.001 0.003 0.042 0.018 0.006 0.024
(0.069) (0.053) (0.054) (0.053) (0.070) (0.052) (0.049) (0.052)
Religiosity -0.008 0.021 0.010 0.014 0.018 0.015 0.005 0.016
(0.033) (0.025) (0.025) (0.024) (0.034) (0.025) (0.023) (0.025)
Muslim Religiosity 0.026 -0.008 0.001 -0.002 -0.009 -0.006 0.002 -0.008
(0.034) (0.026) (0.026) (0.025) (0.035) (0.025) (0.024) (0.026)
Father's education:
Elementary -0.037 -0.026 -0.027 -0.027 -0.010 0.003 0.001 0.001
(0.018)

(0.014) (0.014)

(0.013)

(0.018) (0.012) (0.011) (0.011)
Junior High -0.050 -0.048 -0.045 -0.047 -0.033 0.004 -0.007 -0.004
(0.025)

(0.019)

(0.019)

(0.019)

(0.024) (0.017) (0.017) (0.017)
Senior High/University -0.050 -0.055 -0.056 -0.053 -0.022 -0.005 -0.011 -0.010
(0.025)

(0.021)

(0.020)

(0.020)

(0.024) (0.017) (0.016) (0.016)
Mother's education:
Elementary -0.031 -0.039 -0.040 -0.040 -0.007 -0.005 0.001 0.001
(0.017) (0.014)

(0.013)

(0.013)

(0.016) (0.012) (0.011) (0.011)
Junior High 0.000 -0.025 -0.023 -0.022 -0.052 -0.042 -0.038 -0.037
(0.027) (0.024) (0.022) (0.022) (0.022)

(0.016)

(0.016)

(0.016)

Senior High/University -0.008 -0.028 -0.030 -0.032 -0.023 -0.017 -0.011 -0.010
(0.032) (0.028) (0.027) (0.026) (0.025) (0.019) (0.019) (0.018)
Constant 0.043 0.065 0.038 0.028 -0.060 -0.048 -0.056 -0.076
(0.072) (0.055) (0.057) (0.056) (0.072) (0.052) (0.051) (0.054)
Observations 7144 10336 11174 11298 7058 12194 13079 13207
R
2
0.255 0.245 0.244 0.243 0.149 0.132 0.128 0.126
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.390 0.390 0.400 0.400 0.200 0.210 0.210 0.210
Joint signicance of (p-value):
Father's education 0.132 0.034 0.031 0.032 0.525 0.899 0.816 0.872
Mother's education 0.190 0.054 0.026 0.024 0.071 0.040 0.029 0.042
Each column is for a separate regression. All regressions are OLS. Dependent variable is a dummy where 1 denotes
self employment, 0 denotes employment. Omitted category for education is no education. Religiosity is an ordinal
variable from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered by community.
*p<0.10 **p<0.05 ***p<0.01.
not been examined widely in the empirical literature. Barr and Packard (2002) nd
no signicant dierences in risk and time preferences between the self employed and
employees. However, they did not study men and women separately. I likewise do not
nd a signicant relationship between time preference and self employment in either
the male sample (columns 3 and 4) or female sample (columns 7 and 8).
4.4.6 Controlling for Nonstandard Preferences
The preceding analyses omit respondents with nonstandard preferences as determined
in the elicitation modules (see Chapter 2, Section 2.3). Nonetheless, how nonstan-
dard preferences vary with economic behavior is interesting in its own right. Some
92
researchers are studying whether heterogeneity in economic outcomes is driven by
\decision-making ability" (see e.g. Choi et al., 2007 and Choi et al., 2013). Choi et
al. (2013) measure an individual's decision-making ability by the quality of his/her de-
cisions, which in turn is measured by the consistency of his/her choices with economic
rationality.
3
They nd that this measure of decision-making quality is signicantly
positively correlated with household wealth, even after controlling for other socioeco-
nomic variables.
IFLS nonstandard preferences t within this strand of research. Appendix 4.B
reports results of regressions identical to the preceding regressions with the excep-
tion that they now include a dummy variable for nonstandard preference.
4
With the
exception of years of schooling, nonstandard preferences are mostly insignicantly
correlated with our measures of economic decisions Nonstandard preferences are sig-
nicantly negatively correlated with years of schooling(Table 4B.1). The caveat, of
course, is the direction of causality cannot be assigned in the present analysis.
4.5 Discussion
4.5.1 Measurement Error
The reliability of this paper's ndings is hampered by the likely presence of mea-
surement error. The rst potential problem is measurement error in the dependent
variables. For example, in the analysis of the timing into full time employment, the
dependent variable (a dummy for being employed at 18) says nothing about whether
3
This is done through an online experiment administered to 5,000 Dutch individuals. Partici-
pants choose to allocate money from a given budget constraint to a safe or risky portfolio. Consis-
tency in choices is measured by the extent to which a participant satises the Generalized Axiom of
Revealed Preference (due to Afriat, 1967). GARP requires that if a portfolio x is revealed preferred
to x
0
then x
0
is not strictly revealed preferred to x.
4
Nonstandard preferences are coded 0 under the standard risk/time preference binary variables
in these regressions.
93
an unemployed respondent was actually in the labor force. Since some as-yet un-
employed respondents could have been searching for employment, the ideal measure
would be whether the respondent was looking for work at age 18. Our dependent
variable therefore under-measures the intention to work. If the regressors are mea-
sured correctly, our results are inecient but still unbiased and consistent. However,
if theX variables are in fact correlated with the latent intention to work, the results
are inconsistent.
The more severe problem therefore is measurement error in the explanatory vari-
ables. As is well known, the estimated coecient of anX variable measured with error
is biased towards zero. What's more, even if only one explanatory variable is mea-
sured with error, all other coecients are biased as well, and in unknown directions
(Greene, 2008).
In practice, the best we can do is to deal with the attenuation bias resulting from
mismeasurement of a particular X variable and assume classical measurement error
in the Y variable. The X variable of utmost interest is the risk, or time, preference
measure. Instrumental variable estimation using a variable that is correlated with
the preference variable but uncorrelated with the main error term will identify the
true coecient of the preference variable. In future work, I intend to instrument risk
aversion or time preference with exogenous shocks such as weather shocks or natural
disasters.
4.5.2 Seemingly Unrelated Regressions
The interrelatedness of the adulthood transition dependent variables point to SUR
estimation of a combined model for improved eciency. I report SUR estimates in
Appendix 4.C. I refrain from interpretation of the results, however, for two reasons.
94
First, any eciency gains could have been negated by incorrect standard errors due to
heteroskedasticity. Second, the SUR specications do not control for community xed
eects and so the estimates are not comparable to the OLS estimates. Future work
requires programming a SUR estimator to properly account for heteroskedasticity
and community xed eects.
4.6 Conclusion
The paper nds mixed evidence of the signicance of risk and time preferences in de-
cisions pertaining to the transition into adulthood among Indonesians. The following
are the main ndings:
Men who are more risk averse attain fewer years of schooling. Men and women
who are more impatient attain fewer years of schooling.
Risk averse men and women marry earlier, and so do impatient women.
Married couples positively sort on risk and time preferences.
Risk and time preferences are not signicantly correlated with female fertility
timing.
High risk aversion and high patience are signicantly negatively correlated with
birth control use among women.
Highly risk averse men are signicantly less likely to migrate.
Risk and time preferences are not signicantly correlated with entry into full
time employment.
Conditional on being employed, highly risk averse men and women are signi-
cantly less likely to be self employed.
95
Appendix 4.A Number of Moves
Table 4A.1: Number of Moves
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.041 0.001
(0.025) (0.020)
Most Risk Averse B -0.102 -0.012
(0.025)

(0.022)
Most Impatient A 0.027 0.007
(0.020) (0.015)
Most Impatient B 0.012 -0.007
(0.022) (0.018)
Age (years) -0.002 -0.001 -0.001 -0.001 -0.002 -0.002 -0.002 -0.002
(0.001) (0.001) (0.001) (0.001) (0.001)

(0.001)

(0.001)

(0.001)

Education level:
Elementary 0.060 0.096 0.077 0.085 0.030 0.041 0.029 0.029
(0.057) (0.032)

(0.034)

(0.034)

(0.026) (0.019)

(0.019) (0.019)
Junior High 0.132 0.170 0.147 0.156 0.077 0.105 0.086 0.085
(0.069) (0.044)

(0.045)

(0.045)

(0.039) (0.028)

(0.028)

(0.028)

Senior High/University 0.293 0.315 0.298 0.310 0.251 0.236 0.224 0.220
(0.067)

(0.041)

(0.042)

(0.043)

(0.042)

(0.031)

(0.030)

(0.031)

Rural -0.071 -0.103 -0.083 -0.076 -0.141 -0.058 -0.057 -0.064
(0.073) (0.047)

(0.046) (0.045) (0.058)

(0.042) (0.041) (0.040)
Muslim -0.312 -0.314 -0.320 -0.285 0.043 0.174 0.155 0.164
(0.196) (0.143)

(0.135)

(0.135)

(0.143) (0.101) (0.089) (0.094)
Religiosity -0.232 -0.196 -0.207 -0.201 0.016 0.063 0.049 0.059
(0.099)

(0.070)

(0.068)

(0.067)

(0.062) (0.046) (0.040) (0.043)
Muslim Religiosity 0.216 0.183 0.188 0.177 -0.008 -0.070 -0.053 -0.060
(0.101)

(0.071)

(0.068)

(0.068)

(0.066) (0.049) (0.043) (0.046)
Father's education:
Elementary 0.039 0.027 0.033 0.038 0.021 0.013 0.011 0.010
(0.033) (0.028) (0.025) (0.026) (0.025) (0.019) (0.018) (0.018)
Junior High 0.103 0.100 0.095 0.095 0.069 0.032 0.038 0.036
(0.057) (0.047)

(0.044)

(0.044)

(0.045) (0.033) (0.032) (0.032)
Senior High/University 0.018 0.002 0.001 0.004 0.064 0.076 0.070 0.068
(0.056) (0.048) (0.046) (0.045) (0.052) (0.038)

(0.036) (0.036)
Mother's education:
Elementary 0.064 0.057 0.068 0.068 -0.006 0.026 0.023 0.023
(0.035) (0.027)

(0.025)

(0.025)

(0.028) (0.019) (0.018) (0.018)
Junior High 0.054 0.053 0.070 0.071 -0.051 -0.021 -0.025 -0.019
(0.060) (0.045) (0.044) (0.043) (0.049) (0.040) (0.037) (0.037)
Senior High/University -0.065 -0.029 -0.011 0.007 -0.010 0.050 0.070 0.064
(0.065) (0.054) (0.049) (0.050) (0.064) (0.049) (0.048) (0.047)
Constant 0.716 0.714 0.635 0.597 0.299 0.142 0.159 0.159
(0.186)

(0.137)

(0.137)

(0.137)

(0.153) (0.108) (0.094) (0.099)
Observations 8185 11845 12765 12916 7766 13217 14179 14332
R
2
0.146 0.133 0.130 0.129 0.172 0.154 0.150 0.149
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.500 0.460 0.460 0.470 0.380 0.350 0.350 0.350
Joint signicance of (p-value):
Education categories 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Father's education 0.252 0.093 0.074 0.074 0.466 0.242 0.267 0.288
Mother's education 0.045 0.046 0.017 0.028 0.758 0.328 0.189 0.267
Each column is for a separate regression. All regressions are OLS. Dependent variable is number of moves since
age 18. Omitted category for education is no education. Religiosity is an ordinal variable from 0 (not religious)
to 3 (very religious). Standard errors in parentheses clustered by community. *p<0.10 **p<0.05 ***p<0.01.
96
Appendix 4.B Controlling for Nonstandard Pref-
erences
Table 4B.1: Years of Schooling
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.227 -0.003
(0.076)

(0.073)
Most Risk Averse B -0.589 -0.098
(0.081)

(0.082)
Most Impatient A -0.354 -0.176
(0.063)

(0.061)

Most Impatient B -0.422 -0.155
(0.080)

(0.072)

Nonstandard: GA A -0.615 -0.301
(0.073)

(0.071)

Nonstandard: GL B -0.616 -0.307
(0.129)

(0.118)

Nonstandard: NTD A -1.034 -0.574
(0.227)

(0.219)

Nonstandard: NTD B -1.298 -0.730
(0.353)

(0.322)

Age (years) -0.050 -0.051 -0.051 -0.051 -0.093 -0.094 -0.094 -0.094
(0.003)

(0.003)

(0.003)

(0.003)

(0.003)

(0.003)

(0.003)

(0.003)

Rural -0.804 -0.804 -0.809 -0.802 -0.729 -0.727 -0.716 -0.722
(0.169)

(0.171)

(0.169)

(0.168)

(0.198)

(0.198)

(0.200)

(0.199)

Muslim -0.725 -0.753 -0.708 -0.715 -0.357 -0.332 -0.347 -0.353
(0.329)

(0.331)

(0.328)

(0.328)

(0.392) (0.389) (0.388) (0.386)
Religiosity 0.486 0.480 0.490 0.494 0.512 0.523 0.519 0.521
(0.157)

(0.158)

(0.156)

(0.156)

(0.176)

(0.175)

(0.173)

(0.172)

Muslim Religiosity -0.067 -0.053 -0.066 -0.069 -0.252 -0.262 -0.256 -0.256
(0.164) (0.165) (0.163) (0.163) (0.189) (0.187) (0.185) (0.184)
Father's education:
Elementary 1.662 1.685 1.687 1.693 1.561 1.567 1.556 1.560
(0.114)

(0.114)

(0.114)

(0.114)

(0.097)

(0.097)

(0.096)

(0.096)

Junior High 2.744 2.772 2.779 2.795 2.729 2.742 2.736 2.736
(0.150)

(0.150)

(0.150)

(0.151)

(0.145)

(0.145)

(0.144)

(0.144)

Senior High/University 3.060 3.111 3.103 3.111 3.400 3.419 3.409 3.408
(0.162)

(0.164)

(0.162)

(0.163)

(0.141)

(0.141)

(0.141)

(0.141)

Mother's education:
Elementary 0.919 0.908 0.919 0.918 1.401 1.402 1.409 1.407
(0.106)

(0.106)

(0.106)

(0.106)

(0.082)

(0.083)

(0.082)

(0.083)

Junior High 1.264 1.224 1.252 1.252 2.154 2.152 2.164 2.175
(0.149)

(0.149)

(0.149)

(0.150)

(0.132)

(0.134)

(0.133)

(0.132)

Senior High/University 1.537 1.504 1.548 1.541 2.257 2.281 2.262 2.267
(0.173)

(0.173)

(0.174)

(0.173)

(0.152)

(0.152)

(0.152)

(0.152)

Constant 8.791 9.001 8.736 8.805 8.914 8.862 8.900 8.897
(0.356)

(0.362)

(0.353)

(0.359)

(0.421)

(0.423)

(0.417)

(0.409)

Observations 12989 12984 13022 13021 14325 14333 14421 14421
R
2
0.459 0.458 0.459 0.459 0.595 0.595 0.597 0.597
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 8.600 8.600 8.590 8.590 7.750 7.750 7.720 7.720
Joint signicance of (p-value):
Father's education 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Mother's education 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Each column is for a separate regression. All regressions are OLS. Dependent variable is years of schooling.
Omitted category for education is no education. Religiosity is an ordinal variable from 0 (not religious) to 3 (very
religious). Standard errors in parentheses clustered by community. *p<0.10 **p<0.05 ***p<0.01.
97
Table 4B.2: Marriage Timing
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A 0.013 -0.003
(0.014) (0.015)
Most Risk Averse B 0.038 0.035
(0.016)

(0.016)

Most Impatient A 0.019 0.027
(0.013) (0.012)

Most Impatient B 0.014 0.026
(0.014) (0.013)

Nonstandard: GA A 0.036 0.020
(0.013)

(0.013)
Nonstandard: GL B 0.015 0.084
(0.023) (0.023)

Nonstandard: NTD A 0.042 0.006
(0.048) (0.039)
Nonstandard: NTD B 0.182 -0.013
(0.061)

(0.059)
Age (years) -0.006 -0.006 -0.006 -0.006 -0.001 -0.002 -0.002 -0.002
(0.001)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

Rural 0.020 0.021 0.020 0.020 0.042 0.042 0.043 0.043
(0.025) (0.025) (0.024) (0.024) (0.033) (0.032) (0.032) (0.032)
Muslim 0.027 0.028 0.043 0.042 -0.029 -0.032 -0.025 -0.022
(0.068) (0.068) (0.068) (0.068) (0.070) (0.070) (0.070) (0.069)
Religiosity 0.001 0.001 0.010 0.009 -0.056 -0.057 -0.054 -0.053
(0.032) (0.031) (0.031) (0.031) (0.030) (0.030) (0.029) (0.029)
Muslim Religiosity 0.014 0.014 0.004 0.005 0.056 0.057 0.053 0.052
(0.033) (0.033) (0.032) (0.033) (0.032) (0.032) (0.032) (0.032)
Father's education:
Elementary -0.071 -0.072 -0.072 -0.072 -0.042 -0.042 -0.042 -0.043
(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

(0.016)

Junior High -0.130 -0.133 -0.132 -0.133 -0.123 -0.124 -0.122 -0.123
(0.027)

(0.027)

(0.026)

(0.027)

(0.024)

(0.024)

(0.023)

(0.023)

Senior High/University -0.152 -0.156 -0.155 -0.156 -0.171 -0.171 -0.171 -0.171
(0.026)

(0.026)

(0.026)

(0.026)

(0.024)

(0.024)

(0.024)

(0.024)

Mother's education:
Elementary -0.053 -0.053 -0.053 -0.053 -0.014 -0.014 -0.016 -0.016
(0.015)

(0.015)

(0.015)

(0.015)

(0.014) (0.014) (0.014) (0.014)
Junior High -0.140 -0.134 -0.138 -0.138 -0.098 -0.099 -0.101 -0.102
(0.030)

(0.031)

(0.030)

(0.030)

(0.025)

(0.025)

(0.025)

(0.025)

Senior High/University -0.140 -0.136 -0.141 -0.141 -0.141 -0.144 -0.143 -0.144
(0.033)

(0.033)

(0.033)

(0.033)

(0.026)

(0.026)

(0.026)

(0.026)

Constant 0.867 0.851 0.854 0.858 0.641 0.620 0.631 0.628
(0.072)

(0.072)

(0.072)

(0.073)

(0.071)

(0.070)

(0.069)

(0.070)

Observations 8874 8874 8904 8903 10392 10400 10462 10462
R
2
0.132 0.131 0.131 0.131 0.128 0.129 0.128 0.128
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.590 0.590 0.590 0.590 0.500 0.500 0.500 0.500
Joint signicance of (p-value):
Father's education 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Mother's education 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Each column is for a separate regression. All regressions are OLS. Dependent variable for male (female) sample is
a dummy for being married by 25 (20). Omitted category for education is no education. Religiosity is an ordinal
variable from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered by community.
*p<0.10 **p<0.05 ***p<0.01.
98
Table 4B.3: Likelihood of First Birth by Age 18
(1) (2) (3) (4)
Most Risk Averse A -0.008
(0.010)
Most Risk Averse B -0.005
(0.013)
Most Impatient A 0.006
(0.009)
Most Impatient B 0.015
(0.010)
Nonstandard: GA A 0.014
(0.010)
Nonstandard: GL B -0.000
(0.020)
Nonstandard: NTD A -0.018
(0.038)
Nonstandard: NTD B -0.009
(0.043)
Age (years) -0.004 -0.004 -0.004 -0.004
(0.001)

(0.001)

(0.001)

(0.001)

Rural 0.021 0.021 0.021 0.020
(0.014) (0.014) (0.014) (0.014)
Muslim -0.047 -0.044 -0.041 -0.041
(0.032) (0.032) (0.030) (0.030)
Religiosity -0.023 -0.024 -0.022 -0.022
(0.015) (0.015) (0.014) (0.014)
Muslim Religiosity 0.040 0.040 0.039 0.039
(0.017)

(0.017)

(0.016)

(0.016)

Father's education:
Elementary -0.007 -0.007 -0.006 -0.006
(0.017) (0.017) (0.017) (0.017)
Junior High -0.013 -0.015 -0.014 -0.013
(0.021) (0.021) (0.021) (0.021)
Senior High/University -0.047 -0.049 -0.048 -0.047
(0.022)

(0.022)

(0.022)

(0.022)

Mother's education:
Elementary -0.028 -0.028 -0.028 -0.028
(0.015) (0.015) (0.015) (0.015)
Junior High -0.048 -0.048 -0.049 -0.049
(0.021)

(0.021)

(0.021)

(0.021)

Senior High/University -0.046 -0.048 -0.048 -0.048
(0.021)

(0.021)

(0.021)

(0.021)

Constant 0.265 0.273 0.261 0.254
(0.043)

(0.045)

(0.043)

(0.042)

Observations 5992 5997 6006 6005
R
2
0.116 0.115 0.115 0.116
Community FE Yes Yes Yes Yes
Mean of dependent variable 0.100 0.100 0.100 0.100
Joint signicance of (p-value):
Father's education 0.018 0.015 0.017 0.017
Mother's education 0.123 0.113 0.104 0.099
Each column is for a separate regression. All regressions are OLS. Dependent variable is a dummy for having a rst
birth by age 18. Sample consists of ever married women ages 18{49. Omitted category for education is no education.
Religiosity is an ordinal variable from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered by
community. *p<0.10 **p<0.05 ***p<0.01.
99
Table 4B.4: Likelihood of Using Birth Control
(1) (2) (3) (4)
Most Risk Averse A -0.108
(0.048)

Most Risk Averse B -0.116
(0.058)

Most Impatient A 0.079
(0.050)
Most Impatient B 0.131
(0.055)

Age (years) -0.007 -0.009 -0.005 -0.003
(0.001)

(0.001)

(0.001)

(0.001)

Most Risk Averse A Age 0.003
(0.001)

Most Risk Averse B Age 0.003
(0.002)
Most Impatient A Age -0.002
(0.001)
Most Impatient B Age -0.004
(0.002)

Nonstandard: GA A -0.004
(0.015)
Nonstandard: GL B -0.034
(0.024)
Nonstandard: NTD A -0.048
(0.049)
Nonstandard: NTD B -0.050
(0.073)
Rural -0.030 -0.031 -0.031 -0.032
(0.028) (0.028) (0.028) (0.028)
Muslim 0.063 0.066 0.065 0.067
(0.081) (0.082) (0.081) (0.080)
Religiosity 0.003 0.007 0.006 0.007
(0.037) (0.038) (0.037) (0.037)
Muslim Religiosity 0.000 -0.004 -0.002 -0.003
(0.038) (0.038) (0.038) (0.038)
Father's education:
Elementary 0.017 0.018 0.019 0.019
(0.017) (0.017) (0.017) (0.017)
Junior High -0.034 -0.033 -0.031 -0.032
(0.026) (0.026) (0.026) (0.026)
Senior High/University -0.018 -0.016 -0.015 -0.015
(0.026) (0.026) (0.026) (0.026)
Mother's education:
Elementary 0.018 0.017 0.016 0.016
(0.015) (0.015) (0.015) (0.015)
Junior High 0.018 0.013 0.014 0.013
(0.026) (0.026) (0.026) (0.026)
Senior High/University 0.019 0.018 0.019 0.018
(0.032) (0.032) (0.032) (0.032)
Constant 0.738 0.802 0.649 0.598
(0.087)

(0.093)

(0.092)

(0.095)

Observations 8882 8887 8906 8906
R
2
0.092 0.092 0.092 0.093
Most Risk Averse + Most Risk AverseAge -0.105 -0.113
(0.047)
**
(0.056)
**
Most Impatient + Most ImpatientAge 0.077 0.127
(0.048) (0.054)
**
Community FE Yes Yes Yes Yes
Mean of dependent variable 0.560 0.560 0.560 0.560
Joint signicance of (p-value):
Father's education 0.050 0.055 0.053 0.053
Mother's education 0.692 0.723 0.757 0.776
Each column is for a separate regression. All regressions are OLS. Dependent
variable is a dummy for using birth control. Sample consists of women ages 18{49.
Omitted category for education is no education. Religiosity is an ordinal variable
from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered
by community. *p<0.10 **p<0.05 ***p<0.01.
100
Table 4B.5: Ever Migrate
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.012 0.006
(0.010) (0.010)
Most Risk Averse B -0.030 -0.002
(0.010)

(0.010)
Most Impatient A 0.011 0.002
(0.008) (0.008)
Most Impatient B 0.007 -0.003
(0.009) (0.009)
Nonstandard: GA A -0.025 -0.001
(0.009)

(0.009)
Nonstandard: GL B -0.011 -0.004
(0.015) (0.015)
Nonstandard: NTD A 0.020 -0.030
(0.028) (0.024)
Nonstandard: NTD B -0.071 0.036
(0.036)

(0.045)
Age (years) -0.001 -0.001 -0.001 -0.001 -0.002 -0.002 -0.002 -0.002
(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

Rural -0.071 -0.069 -0.069 -0.069 -0.045 -0.044 -0.043 -0.044
(0.022)

(0.022)

(0.022)

(0.022)

(0.022)

(0.022)

(0.022) (0.022)
Muslim -0.085 -0.082 -0.089 -0.089 0.010 0.004 0.003 -0.000
(0.042)

(0.042)

(0.041)

(0.041)

(0.049) (0.048) (0.048) (0.048)
Religiosity -0.064 -0.062 -0.067 -0.066 -0.000 -0.003 -0.004 -0.005
(0.019)

(0.019)

(0.018)

(0.018)

(0.023) (0.023) (0.023) (0.022)
Muslim Religiosity 0.058 0.056 0.060 0.060 0.001 0.003 0.004 0.006
(0.020)

(0.020)

(0.019)

(0.019)

(0.024) (0.024) (0.024) (0.023)
Father's education:
Elementary 0.026 0.026 0.027 0.027 0.017 0.018 0.018 0.018
(0.010)

(0.010)

(0.010)

(0.010)

(0.009) (0.009) (0.009) (0.009)
Junior High 0.064 0.064 0.066 0.065 0.051 0.052 0.053 0.053
(0.018)

(0.018)

(0.018)

(0.018)

(0.016)

(0.017)

(0.016)

(0.016)

Senior High/University 0.035 0.036 0.038 0.037 0.051 0.051 0.052 0.052
(0.020) (0.020) (0.020) (0.020) (0.017)

(0.017)

(0.017)

(0.017)

Mother's education:
Elementary 0.016 0.015 0.015 0.015 0.019 0.019 0.019 0.019
(0.011) (0.011) (0.011) (0.011) (0.009)

(0.009)

(0.009)

(0.009)

Junior High 0.019 0.018 0.019 0.019 0.008 0.008 0.007 0.007
(0.020) (0.020) (0.020) (0.020) (0.018) (0.018) (0.018) (0.018)
Senior High/University -0.020 -0.022 -0.019 -0.019 0.030 0.033 0.031 0.031
(0.021) (0.021) (0.021) (0.021) (0.020) (0.020) (0.020) (0.021)
Constant 0.418 0.426 0.404 0.405 0.272 0.280 0.279 0.284
(0.045)

(0.045)

(0.044)

(0.045)

(0.051)

(0.051)

(0.050)

(0.050)

Observations 13136 13131 13169 13168 14461 14468 14557 14557
R
2
0.202 0.202 0.202 0.202 0.214 0.214 0.214 0.214
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.250 0.250 0.250 0.250 0.220 0.220 0.220 0.220
Joint signicance of (p-value):
Father's education 0.004 0.004 0.002 0.003 0.005 0.005 0.004 0.003
Mother's education 0.082 0.081 0.106 0.105 0.162 0.153 0.160 0.160
Each column is for a separate regression. All regressions are OLS. Dependent variable is a dummy for ever migrating
since age 18. Omitted category for education is no education. Religiosity is an ordinal variable from 0 (not religious)
to 3 (very religious). Standard errors in parentheses clustered by community. *p<0.10 **p<0.05 ***p<0.01.
101
Table 4B.6: Hazard Model of Time to First Migration
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.024 0.071
(0.056) (0.063)
Most Risk Averse B -0.140 0.002
(0.055)

(0.069)
Most Impatient A 0.035 -0.022
(0.045) (0.050)
Most Impatient B 0.023 -0.009
(0.051) (0.058)
Nonstandard: GA A -0.147 0.000
(0.050)

(0.058)
Nonstandard: GL B -0.019 -0.014
(0.086) (0.099)
Nonstandard: NTD A 0.040 -0.353
(0.155) (0.190)
Nonstandard: NTD B -0.433 0.233
(0.259) (0.266)
Age (years) -0.031 -0.031 -0.031 -0.031 -0.043 -0.043 -0.043 -0.043
(0.002)

(0.002)

(0.002)

(0.002)

(0.002)

(0.002)

(0.002)

(0.002)

Rural -0.008 -0.007 -0.007 -0.025 0.042 0.051 0.054 0.051
(0.213) (0.217) (0.214) (0.217) (0.274) (0.271) (0.272) (0.271)
Muslim -0.611 -0.592 -0.634 -0.631 0.063 0.027 0.012 -0.012
(0.250)

(0.248)

(0.244)

(0.244)

(0.298) (0.292) (0.288) (0.289)
Religiosity -0.382 -0.367 -0.395 -0.390 0.061 0.046 0.034 0.024
(0.116)

(0.115)

(0.113)

(0.113)

(0.139) (0.136) (0.134) (0.134)
Muslim Religiosity 0.363 0.350 0.374 0.372 -0.054 -0.035 -0.024 -0.011
(0.121)

(0.120)

(0.117)

(0.117)

(0.144) (0.141) (0.140) (0.140)
Father's education:
Elementary 0.176 0.185 0.193 0.192 0.196 0.200 0.202 0.203
(0.076)

(0.076)

(0.076)

(0.076)

(0.081)

(0.081)

(0.081)

(0.080)

Junior High 0.348 0.352 0.365 0.364 0.400 0.404 0.409 0.410
(0.100)

(0.099)

(0.099)

(0.099)

(0.110)

(0.111)

(0.110)

(0.110)

Senior High/University 0.216 0.226 0.237 0.235 0.296 0.293 0.298 0.300
(0.108)

(0.107)

(0.107)

(0.107)

(0.109)

(0.109)

(0.109)

(0.109)

Mother's education:
Elementary 0.043 0.037 0.033 0.033 -0.011 -0.012 -0.010 -0.009
(0.063) (0.063) (0.063) (0.063) (0.071) (0.071) (0.071) (0.070)
Junior High 0.042 0.034 0.030 0.031 -0.125 -0.127 -0.132 -0.130
(0.103) (0.103) (0.103) (0.103) (0.104) (0.104) (0.104) (0.104)
Senior High/University -0.122 -0.138 -0.124 -0.121 -0.012 -0.002 -0.007 -0.004
(0.106) (0.106) (0.106) (0.106) (0.113) (0.113) (0.113) (0.113)
Observations 13114 13109 13147 13146 14431 14438 14527 14527
Pseudo R
2
0.036 0.036 0.036 0.036 0.054 0.054 0.055 0.055
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Joint signicance of (p-value):
Father's education 0.006 0.006 0.003 0.003 0.004 0.004 0.003 0.003
Mother's education 0.262 0.228 0.319 0.337 0.538 0.505 0.473 0.479
Each column is for a separate regression. All regressions are Cox proportional hazard models. Time to rst
migration is measured in months since 18th birthday. Omitted category for education is no education. Religiosity
is an ordinal variable from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered by
community. *p<0.10 **p<0.05 ***p<0.01.
102
Table 4B.7: Number of Moves
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.040 0.001
(0.023) (0.019)
Most Risk Averse B -0.104 -0.019
(0.025)

(0.022)
Most Impatient A 0.020 -0.002
(0.020) (0.015)
Most Impatient B 0.004 -0.013
(0.021) (0.018)
Nonstandard: GA A -0.067 -0.015
(0.021)

(0.018)
Nonstandard: GL B -0.058 -0.009
(0.038) (0.031)
Nonstandard: NTD A 0.093 -0.091
(0.079) (0.047)
Nonstandard: NTD B -0.180 0.040
(0.070)

(0.084)
Age (years) -0.002 -0.002 -0.002 -0.002 -0.003 -0.003 -0.003 -0.003
(0.001)

(0.001)

(0.001)

(0.001)

(0.000)

(0.000)

(0.000)

(0.000)

Rural -0.089 -0.086 -0.086 -0.086 -0.080 -0.080 -0.078 -0.079
(0.044)

(0.044)

(0.044) (0.044) (0.040)

(0.040)

(0.040)

(0.040)

Muslim -0.299 -0.294 -0.309 -0.307 0.188 0.177 0.173 0.167
(0.133)

(0.134)

(0.131)

(0.131)

(0.096)

(0.095) (0.095) (0.094)
Religiosity -0.192 -0.189 -0.199 -0.197 0.074 0.069 0.067 0.065
(0.065)

(0.066)

(0.064)

(0.064)

(0.044) (0.044) (0.044) (0.043)
Muslim Religiosity 0.178 0.175 0.183 0.182 -0.078 -0.072 -0.070 -0.067
(0.066)

(0.067)

(0.065)

(0.065)

(0.046) (0.046) (0.046) (0.046)
Father's education:
Elementary 0.074 0.076 0.078 0.077 0.030 0.031 0.031 0.031
(0.025)

(0.025)

(0.025)

(0.025)

(0.017) (0.017) (0.017) (0.017)
Junior High 0.169 0.169 0.174 0.172 0.094 0.094 0.095 0.096
(0.043)

(0.043)

(0.043)

(0.043)

(0.031)

(0.031)

(0.031)

(0.031)

Senior High/University 0.079 0.083 0.087 0.084 0.141 0.139 0.143 0.143
(0.044) (0.044) (0.044)

(0.044) (0.035)

(0.035)

(0.035)

(0.035)

Mother's education:
Elementary 0.088 0.086 0.086 0.086 0.050 0.050 0.050 0.050
(0.024)

(0.025)

(0.024)

(0.024)

(0.018)

(0.018)

(0.018)

(0.018)

Junior High 0.107 0.102 0.107 0.107 0.030 0.031 0.030 0.030
(0.043)

(0.043)

(0.043)

(0.043)

(0.037) (0.037) (0.037) (0.037)
Senior High/University 0.033 0.027 0.035 0.037 0.102 0.107 0.105 0.105
(0.050) (0.050) (0.050) (0.050) (0.047)

(0.047)

(0.047)

(0.047)

Constant 0.832 0.870 0.794 0.804 0.264 0.283 0.274 0.284
(0.133)

(0.133)

(0.133)

(0.135)

(0.099)

(0.099)

(0.097)

(0.097)

Observations 13136 13131 13169 13168 14461 14468 14557 14557
R
2
0.121 0.122 0.121 0.121 0.142 0.143 0.143 0.142
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.460 0.460 0.460 0.460 0.350 0.350 0.350 0.350
Joint signicance of (p-value):
Father's education 0.001 0.001 0.001 0.001 0.000 0.001 0.000 0.000
Mother's education 0.002 0.002 0.002 0.002 0.024 0.020 0.020 0.021
Each column is for a separate regression. All regressions are OLS. Dependent variable is number of moves since
age 18. Omitted category for education is no education. Religiosity is an ordinal variable from 0 (not religious)
to 3 (very religious). Standard errors in parentheses clustered by community. *p<0.10 **p<0.05 ***p<0.01.
103
Table 4B.8: Likelihood of Full Time Employment
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A 0.023 -0.019
(0.014) (0.017)
Most Risk Averse B 0.033 0.018
(0.015)

(0.020)
Most Impatient A -0.007 -0.029
(0.012) (0.015)
Most Impatient B 0.008 -0.011
(0.015) (0.017)
Nonstandard: GA A 0.025 -0.010
(0.014) (0.016)
Nonstandard: GL B 0.067 -0.000
(0.023)

(0.029)
Nonstandard: NTD A 0.034 0.052
(0.044) (0.049)
Nonstandard: NTD B 0.100 -0.045
(0.071) (0.083)
Age (years) -0.009 -0.009 -0.009 -0.009 -0.009 -0.009 -0.009 -0.009
(0.000)

(0.000)

(0.000)

(0.000)

(0.001)

(0.001)

(0.001)

(0.001)

Rural 0.032 0.019 0.012 0.015 0.037 0.039 0.039 0.039
(0.105) (0.106) (0.106) (0.106) (0.124) (0.124) (0.123) (0.124)
Muslim 0.059 0.055 0.055 0.054 0.086 0.085 0.080 0.085
(0.060) (0.060) (0.060) (0.060) (0.101) (0.101) (0.101) (0.101)
Religiosity -0.011 -0.013 -0.013 -0.013 -0.012 -0.011 -0.014 -0.011
(0.026) (0.026) (0.025) (0.025) (0.043) (0.043) (0.043) (0.043)
Muslim Religiosity -0.008 -0.007 -0.008 -0.007 -0.023 -0.023 -0.022 -0.025
(0.028) (0.028) (0.027) (0.028) (0.046) (0.045) (0.045) (0.045)
Father's education:
Elementary -0.074 -0.074 -0.076 -0.076 -0.016 -0.017 -0.018 -0.018
(0.017)

(0.017)

(0.017)

(0.017)

(0.019) (0.019) (0.018) (0.018)
Junior High -0.167 -0.165 -0.171 -0.171 -0.029 -0.030 -0.032 -0.031
(0.027)

(0.027)

(0.026)

(0.026)

(0.031) (0.031) (0.031) (0.031)
Senior High/University -0.180 -0.180 -0.185 -0.184 -0.064 -0.065 -0.068 -0.067
(0.030)

(0.030)

(0.030)

(0.030)

(0.036) (0.036) (0.036) (0.036)
Mother's education:
Elementary -0.019 -0.020 -0.018 -0.018 -0.034 -0.034 -0.032 -0.032
(0.015) (0.016) (0.016) (0.016) (0.019) (0.019) (0.019) (0.019)
Junior High -0.028 -0.028 -0.026 -0.026 -0.089 -0.086 -0.087 -0.087
(0.030) (0.030) (0.030) (0.030) (0.036)

(0.036)

(0.036)

(0.036)

Senior High/University -0.062 -0.062 -0.064 -0.063 -0.171 -0.169 -0.170 -0.170
(0.039) (0.039) (0.039) (0.039) (0.041)

(0.041)

(0.041)

(0.041)

Constant 0.990 0.985 1.022 1.012 1.012 0.989 1.026 1.012
(0.078)

(0.080)

(0.078)

(0.078)

(0.115)

(0.115)

(0.115)

(0.115)

Observations 9580 9575 9609 9608 7473 7477 7529 7530
R
2
0.241 0.241 0.240 0.240 0.265 0.266 0.266 0.266
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.580 0.580 0.580 0.580 0.610 0.600 0.600 0.600
Joint signicance of (p-value):
Father's education 0.000 0.000 0.000 0.000 0.367 0.357 0.307 0.336
Mother's education 0.366 0.353 0.368 0.371 0.000 0.001 0.000 0.000
Each column is for a separate regression. All regressions are OLS. Dependent variable is a dummy for full time
employment by age 20. Omitted category for education is no education. Religiosity is an ordinal variable from 0
(not religious) to 3 (very religious). Standard errors in parentheses clustered by community. *p<0.10 **p<0.05
***p<0.01.
104
Table 4B.9: Likelihood of Self Employment
Men Women
(1) (2) (3) (4) (5) (6) (7) (8)
Most Risk Averse A -0.040 -0.035
(0.011)

(0.010)

Most Risk Averse B -0.040 -0.039
(0.012)

(0.012)

Most Impatient A 0.006 -0.007
(0.010) (0.008)
Most Impatient B 0.013 -0.007
(0.011) (0.009)
Nonstandard: GA A -0.043 -0.025
(0.011)

(0.009)

Nonstandard: GL B 0.007 -0.013
(0.017) (0.017)
Nonstandard: NTD A 0.012 -0.046
(0.035) (0.030)
Nonstandard: NTD B -0.005 -0.034
(0.051) (0.039)
Age (years) 0.009 0.009 0.009 0.009 0.007 0.007 0.007 0.007
(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

Rural 0.059 0.059 0.059 0.059 0.025 0.024 0.023 0.023
(0.022)

(0.022)

(0.022)

(0.022)

(0.016) (0.016) (0.015) (0.016)
Muslim 0.004 -0.001 0.004 0.004 0.022 0.018 0.019 0.018
(0.054) (0.054) (0.053) (0.053) (0.051) (0.051) (0.052) (0.051)
Religiosity 0.015 0.012 0.015 0.015 0.013 0.013 0.012 0.012
(0.024) (0.025) (0.024) (0.024) (0.024) (0.024) (0.024) (0.024)
Muslim Religiosity -0.001 0.002 -0.003 -0.003 -0.008 -0.006 -0.006 -0.006
(0.026) (0.026) (0.025) (0.025) (0.025) (0.025) (0.025) (0.025)
Father's education:
Elementary -0.028 -0.027 -0.026 -0.026 -0.001 0.000 0.002 0.002
(0.014)

(0.014)

(0.014) (0.014) (0.011) (0.011) (0.011) (0.011)
Junior High -0.048 -0.045 -0.045 -0.045 -0.004 -0.005 -0.004 -0.004
(0.019)

(0.019)

(0.019)

(0.019)

(0.017) (0.017) (0.017) (0.017)
Senior High/University -0.055 -0.053 -0.052 -0.052 -0.013 -0.012 -0.011 -0.011
(0.020)

(0.020)

(0.020)

(0.020)

(0.016) (0.016) (0.016) (0.016)
Mother's education:
Elementary -0.038 -0.040 -0.040 -0.040 0.001 0.001 0.000 -0.000
(0.013)

(0.013)

(0.013)

(0.013)

(0.011) (0.011) (0.011) (0.011)
Junior High -0.022 -0.023 -0.022 -0.021 -0.035 -0.036 -0.038 -0.037
(0.022) (0.022) (0.022) (0.022) (0.016)

(0.016)

(0.016)

(0.016)

Senior High/University -0.032 -0.032 -0.032 -0.031 -0.010 -0.009 -0.011 -0.010
(0.026) (0.026) (0.026) (0.026) (0.018) (0.018) (0.018) (0.018)
Constant 0.056 0.064 0.028 0.022 -0.060 -0.048 -0.069 -0.069
(0.057) (0.057) (0.056) (0.056) (0.052) (0.052) (0.053) (0.053)
Observations 11360 11355 11384 11383 13210 13216 13290 13290
R
2
0.245 0.245 0.243 0.243 0.128 0.128 0.127 0.127
Community FE Yes Yes Yes Yes Yes Yes Yes Yes
Mean of dependent variable 0.400 0.400 0.400 0.400 0.210 0.210 0.210 0.210
Joint signicance of (p-value):
Father's education 0.028 0.038 0.046 0.044 0.826 0.819 0.804 0.794
Mother's education 0.037 0.027 0.026 0.025 0.050 0.045 0.036 0.040
Each column is for a separate regression. All regressions are OLS. Dependent variable is a dummy where 1 denotes
self employment, 0 denotes employment. Omitted category for education is no education. Religiosity is an ordinal
variable from 0 (not religious) to 3 (very religious). Standard errors in parentheses clustered by community.
*p<0.10 **p<0.05 ***p<0.01.
105
Appendix 4.C SUR Estimates
Table 4C.1: SUR estimates of Risk A and other correlates, males
(1) (2) (3) (4) (5)
Years of schooling Married by 25 Migrated since 18 Full time work by 20 Self employed
Most Risk Averse A -0.312 0.019 -0.023 0.023 -0.048
(0.101)

(0.014) (0.012) (0.014) (0.014)

Age (years) -0.060 -0.006 -0.009 -0.008 0.006
(0.004)

(0.001)

(0.001)

(0.001)

(0.001)

Rural -1.641 0.040 -0.052 0.070 0.222
(0.105)

(0.014)

(0.013)

(0.014)

(0.014)

Muslim -1.650 0.183 -0.027 0.137 -0.082
(0.636)

(0.087)

(0.077) (0.087) (0.085)
Religiosity 0.360 0.083 -0.051 0.004 -0.000
(0.297) (0.041)

(0.036) (0.041) (0.040)
Muslim Religiosity 0.043 -0.082 0.028 -0.014 0.027
(0.312) (0.043) (0.038) (0.043) (0.042)
Father's education:
Elementary 2.142 -0.080 0.014 -0.095 -0.054
(0.146)

(0.020)

(0.018) (0.020)

(0.020)

Junior High 3.738 -0.168 0.103 -0.210 -0.105
(0.237)

(0.033)

(0.029)

(0.032)

(0.032)

Senior High/University 4.712 -0.229 0.029 -0.250 -0.135
(0.255)

(0.035)

(0.031) (0.035)

(0.034)

Mother's education:
Elementary 0.890 -0.055 0.051 -0.047 -0.037
(0.138)

(0.019)

(0.017)

(0.019)

(0.018)

Junior High 2.080 -0.153 0.151 -0.082 0.019
(0.266)

(0.036)

(0.032)

(0.036)

(0.036)
Senior High/University 2.018 -0.103 0.074 -0.127 -0.019
(0.314)

(0.043)

(0.038) (0.043)

(0.042)
Constant 10.128 0.756 0.707 0.884 0.214
(0.656)

(0.090)

(0.079)

(0.090)

(0.088)

Standard errors in parentheses. *p<0.10 **p<0.05 ***p<0.01.
106
Table 4C.2: SUR estimates of Risk B and other correlates, males
(1) (2) (3) (4) (5)
Years of schooling Married by 25 Migrated since 18 Full time work by 20 Self employed
Most Risk Averse B -0.764 0.048 -0.039 0.026 -0.051
(0.108)

(0.015)

(0.013)

(0.015) (0.015)

Age (years) -0.062 -0.006 -0.008 -0.008 0.006
(0.003)

(0.000)

(0.000)

(0.000)

(0.000)

Rural -1.706 0.059 -0.064 0.081 0.221
(0.086)

(0.012)

(0.010)

(0.012)

(0.012)

Muslim -1.302 0.096 -0.045 0.070 -0.019
(0.517)

(0.071) (0.061) (0.071) (0.070)
Religiosity 0.458 0.036 -0.060 -0.006 0.031
(0.236) (0.032) (0.028)

(0.032) (0.032)
Muslim Religiosity 0.010 -0.034 0.037 -0.003 -0.008
(0.248) (0.034) (0.029) (0.034) (0.034)
Father's education:
Elementary 1.976 -0.080 0.020 -0.081 -0.041
(0.117)

(0.016)

(0.014) (0.016)

(0.016)

Junior High 3.703 -0.172 0.114 -0.192 -0.101
(0.198)

(0.027)

(0.023)

(0.027)

(0.027)

Senior High/University 4.476 -0.228 0.041 -0.245 -0.144
(0.217)

(0.030)

(0.026) (0.030)

(0.029)

Mother's education:
Elementary 0.947 -0.056 0.031 -0.035 -0.041
(0.113)

(0.015)

(0.013)

(0.016)

(0.015)

Junior High 2.080 -0.156 0.113 -0.083 -0.008
(0.231)

(0.032)

(0.027)

(0.032)

(0.031)
Senior High/University 2.310 -0.131 0.090 -0.108 -0.032
(0.271)

(0.037)

(0.032)

(0.037)

(0.037)
Constant 10.157 0.801 0.726 0.898 0.187
(0.533)

(0.073)

(0.063)

(0.073)

(0.072)

Standard errors in parentheses. *p<0.10 **p<0.05 ***p<0.01.
107
Table 4C.3: SUR estimates of Time A and other correlates, males
(1) (2) (3) (4) (5)
Years of schooling Married by 25 Migrated since 18 Full time work by 20 Self employed
Most Impatient A -0.451 0.023 0.014 -0.016 0.005
(0.090)

(0.012) (0.011) (0.012) (0.012)
Age (years) -0.062 -0.006 -0.008 -0.008 0.006
(0.003)

(0.000)

(0.000)

(0.000)

(0.000)

Rural -1.685 0.053 -0.063 0.083 0.219
(0.083)

(0.011)

(0.010)

(0.011)

(0.011)

Muslim -1.462 0.074 -0.039 0.072 -0.021
(0.490)

(0.067) (0.058) (0.067) (0.066)
Religiosity 0.396 0.033 -0.061 -0.006 0.022
(0.225) (0.031) (0.027)

(0.031) (0.030)
Muslim Religiosity 0.081 -0.029 0.038 -0.005 -0.002
(0.237) (0.032) (0.028) (0.033) (0.032)
Father's education:
Elementary 2.009 -0.085 0.017 -0.087 -0.043
(0.113)

(0.015)

(0.013) (0.015)

(0.015)

Junior High 3.751 -0.172 0.111 -0.196 -0.110
(0.191)

(0.026)

(0.023)

(0.026)

(0.026)

Senior High/University 4.532 -0.239 0.041 -0.249 -0.136
(0.210)

(0.029)

(0.025) (0.029)

(0.028)

Mother's education:
Elementary 0.909 -0.050 0.037 -0.029 -0.045
(0.109)

(0.015)

(0.013)

(0.015)

(0.015)

Junior High 2.085 -0.152 0.125 -0.073 -0.010
(0.220)

(0.030)

(0.026)

(0.030)

(0.030)
Senior High/University 2.225 -0.137 0.092 -0.106 -0.048
(0.262)

(0.036)

(0.031)

(0.036)

(0.035)
Constant 10.008 0.843 0.677 0.930 0.168
(0.507)

(0.069)

(0.060)

(0.070)

(0.068)

Standard errors in parentheses. *p<0.10 **p<0.05 ***p<0.01.
108
Table 4C.4: SUR estimates of Time B and other correlates, males
(1) (2) (3) (4) (5)
Years of schooling Married by 25 Migrated since 18 Full time work by 20 Self employed
Most Impatient B -0.531 0.024 0.014 0.001 0.009
(0.103)

(0.014) (0.012) (0.014) (0.014)
Age (years) -0.063 -0.006 -0.008 -0.008 0.006
(0.003)

(0.000)

(0.000)

(0.000)

(0.000)

Rural -1.658 0.052 -0.061 0.080 0.218
(0.083)

(0.011)

(0.010)

(0.011)

(0.011)

Muslim -1.450 0.088 -0.042 0.080 -0.016
(0.490)

(0.067) (0.058) (0.067) (0.066)
Religiosity 0.403 0.038 -0.061 -0.004 0.025
(0.225) (0.031) (0.027)

(0.031) (0.030)
Muslim Religiosity 0.075 -0.035 0.037 -0.007 -0.005
(0.237) (0.032) (0.028) (0.033) (0.032)
Father's education:
Elementary 2.016 -0.087 0.019 -0.087 -0.044
(0.113)

(0.015)

(0.013) (0.015)

(0.015)

Junior High 3.749 -0.174 0.110 -0.199 -0.112
(0.190)

(0.026)

(0.022)

(0.026)

(0.026)

Senior High/University 4.543 -0.235 0.043 -0.248 -0.134
(0.208)

(0.028)

(0.025) (0.029)

(0.028)

Mother's education:
Elementary 0.906 -0.049 0.038 -0.028 -0.044
(0.109)

(0.015)

(0.013)

(0.015) (0.015)

Junior High 2.095 -0.151 0.126 -0.071 -0.011
(0.219)

(0.030)

(0.026)

(0.030)

(0.029)
Senior High/University 2.216 -0.133 0.099 -0.106 -0.047
(0.260)

(0.036)

(0.031)

(0.036)

(0.035)
Constant 10.097 0.825 0.677 0.907 0.160
(0.509)

(0.070)

(0.060)

(0.070)

(0.069)

Standard errors in parentheses. *p<0.10 **p<0.05 ***p<0.01.
109
Table 4C.5: SUR estimates of Risk A and other correlates, females
(1) (2) (3) (4) (5) (6) (7)
Years of schooling Married by 25 First birth by 18 Use birth control Migrated since 18 Full time work by 20 Self employed
Most Risk Averse A 0.156 -0.030 -0.001 -0.006 0.005 -0.019 -0.056
(0.172) (0.021) (0.009) (0.026) (0.025) (0.024) (0.021)

Age (years) -0.029 -0.017 0.002 -0.008 -0.013 -0.021 0.013
(0.014)

(0.002)

(0.001)

(0.002)

(0.002)

(0.002)

(0.002)

Rural -1.265 0.054 0.011 0.015 -0.057 0.055 -0.014
(0.183)

(0.022)

(0.010) (0.027) (0.026)

(0.026)

(0.023)
Muslim -2.593 0.045 0.026 0.337 -0.245 -0.079 0.174
(1.042)

(0.126) (0.056) (0.155)

(0.150) (0.148) (0.130)
Religiosity -0.966 0.022 0.001 0.124 -0.102 -0.083 0.062
(0.466)

(0.056) (0.025) (0.069) (0.067) (0.066) (0.058)
Muslim Religiosity 1.074 -0.027 0.006 -0.169 0.101 0.051 -0.113
(0.505)

(0.061) (0.027) (0.075)

(0.073) (0.072) (0.063)
Father's education:
Elementary 1.806 -0.044 0.007 -0.017 0.084 -0.097 -0.075
(0.287)

(0.035) (0.015) (0.043) (0.041)

(0.041)

(0.036)

Junior High 3.297 -0.056 0.016 -0.022 0.096 -0.155 -0.028
(0.390)

(0.047) (0.021) (0.058) (0.056) (0.056)

(0.049)
Senior High/University 4.073 -0.161 0.004 -0.059 0.131 -0.215 -0.068
(0.398)

(0.048)

(0.021) (0.059) (0.057)

(0.057)

(0.050)
Mother's education:
Elementary 1.631 -0.056 -0.018 0.085 -0.023 -0.056 0.029
(0.250)

(0.030) (0.013) (0.037)

(0.036) (0.036) (0.031)
Junior High 3.166 -0.142 -0.038 0.013 0.143 -0.137 -0.062
(0.382)

(0.046)

(0.020) (0.057) (0.055)

(0.054)

(0.047)
Senior High/University 4.212 -0.167 -0.035 0.117 0.086 -0.181 -0.072
(0.431)

(0.052)

(0.023) (0.064) (0.062) (0.061)

(0.054)
Constant 9.456 1.432 -0.079 0.629 0.972 1.645 -0.212
(1.138)

(0.138)

(0.061) (0.169)

(0.164)

(0.162)

(0.142)
Standard errors in parentheses. *p<0.10 **p<0.05 ***p<0.01.
110
Table 4C.6: SUR estimates of Risk B and other correlates, females
(1) (2) (3) (4) (5) (6) (7)
Years of schooling Married by 25 First birth by 18 Use birth control Migrated since 18 Full time work by 20 Self employed
Most Risk Averse B -0.171 0.010 -0.007 -0.004 0.010 -0.006 -0.067
(0.186) (0.023) (0.012) (0.028) (0.027) (0.027) (0.024)

Age (years) -0.056 -0.014 0.004 -0.009 -0.013 -0.023 0.013
(0.011)

(0.001)

(0.001)

(0.002)

(0.002)

(0.002)

(0.001)

Rural -1.304 0.057 0.012 0.018 -0.068 0.055 0.001
(0.141)

(0.018)

(0.009) (0.021) (0.021)

(0.020)

(0.018)
Muslim -2.863 0.083 0.035 0.368 -0.192 0.003 0.049
(0.886)

(0.110) (0.055) (0.134)

(0.131) (0.128) (0.112)
Religiosity -0.848 0.019 -0.002 0.139 -0.091 -0.035 0.034
(0.410)

(0.051) (0.026) (0.062)

(0.060) (0.059) (0.052)
Muslim Religiosity 1.039 -0.034 0.001 -0.176 0.080 0.009 -0.056
(0.433)

(0.054) (0.027) (0.066)

(0.064) (0.063) (0.055)
Father's education:
Elementary 1.712 -0.067 0.001 0.001 0.050 -0.045 -0.028
(0.216)

(0.027)

(0.013) (0.033) (0.032) (0.031) (0.027)
Junior High 3.418 -0.102 -0.004 -0.018 0.060 -0.127 -0.006
(0.299)

(0.037)

(0.019) (0.045) (0.044) (0.043)

(0.038)
Senior High/University 4.411 -0.190 -0.015 -0.003 0.129 -0.168 -0.042
(0.307)

(0.038)

(0.019) (0.047) (0.045)

(0.045)

(0.039)
Mother's education:
Elementary 1.477 -0.033 -0.004 0.048 -0.001 -0.059 0.009
(0.190)

(0.024) (0.012) (0.029) (0.028) (0.028)

(0.024)
Junior High 2.879 -0.107 -0.033 0.000 0.119 -0.140 -0.064
(0.298)

(0.037)

(0.019) (0.045) (0.044)

(0.043)

(0.038)
Senior High/University 3.704 -0.148 -0.027 0.066 0.093 -0.159 -0.052
(0.339)

(0.042)

(0.021) (0.051) (0.050) (0.049)

(0.043)
Constant 10.650 1.290 -0.099 0.621 0.943 1.553 -0.171
(0.958)

(0.119)

(0.060) (0.145)

(0.141)

(0.139)

(0.121)
Standard errors in parentheses. *p<0.10 **p<0.05 ***p<0.01.
111
Table 4C.7: SUR estimates of Time A and other correlates, females
(1) (2) (3) (4) (5) (6) (7)
Years of schooling Married by 25 First birth by 18 Use birth control Migrated since 18 Full time work by 20 Self employed
Most Impatient A -0.152 -0.020 -0.004 -0.007 0.002 -0.006 -0.018
(0.138) (0.017) (0.009) (0.021) (0.020) (0.020) (0.018)
Age (years) -0.065 -0.014 0.004 -0.010 -0.012 -0.022 0.013
(0.011)

(0.001)

(0.001)

(0.002)

(0.002)

(0.002)

(0.001)

Rural -1.299 0.060 0.012 0.021 -0.067 0.058 0.002
(0.136)

(0.017)

(0.009) (0.021) (0.020)

(0.020)

(0.017)
Muslim -2.747 0.042 0.032 0.319 -0.152 -0.047 0.003
(0.849)

(0.105) (0.053) (0.129)

(0.125) (0.123) (0.109)
Religiosity -0.775 0.006 -0.006 0.127 -0.070 -0.057 0.015
(0.388)

(0.048) (0.024) (0.059)

(0.057) (0.056) (0.050)
Muslim Religiosity 1.001 -0.018 0.001 -0.156 0.060 0.036 -0.040
(0.412)

(0.051) (0.026) (0.062)

(0.061) (0.060) (0.053)
Father's education:
Elementary 1.727 -0.066 -0.000 0.005 0.036 -0.045 -0.036
(0.209)

(0.026)

(0.013) (0.032) (0.031) (0.030) (0.027)
Junior High 3.374 -0.100 -0.006 -0.006 0.052 -0.127 -0.019
(0.291)

(0.036)

(0.018) (0.044) (0.043) (0.042)

(0.037)
Senior High/University 4.329 -0.193 -0.017 0.010 0.111 -0.161 -0.048
(0.298)

(0.037)

(0.019) (0.045) (0.044)

(0.043)

(0.038)
Mother's education:
Elementary 1.451 -0.032 -0.006 0.047 0.015 -0.065 0.018
(0.185)

(0.023) (0.012) (0.028) (0.027) (0.027)

(0.024)
Junior High 2.980 -0.105 -0.034 -0.004 0.137 -0.146 -0.063
(0.289)

(0.036)

(0.018) (0.044) (0.043)

(0.042)

(0.037)
Senior High/University 3.843 -0.140 -0.027 0.051 0.105 -0.187 -0.052
(0.330)

(0.041)

(0.021) (0.050) (0.049)

(0.048)

(0.042)
Constant 10.713 1.353 -0.099 0.676 0.882 1.577 -0.145
(0.916)

(0.113)

(0.057) (0.139)

(0.135)

(0.133)

(0.117)
Standard errors in parentheses. *p<0.10 **p<0.05 ***p<0.01.
112
Table 4C.8: SUR estimates of Time B and other correlates, females
(1) (2) (3) (4) (5) (6) (7)
Years of schooling Married by 25 First birth by 18 Use birth control Migrated since 18 Full time work by 20 Self employed
Most Impatient B -0.144 -0.025 0.003 -0.020 -0.024 0.016 -0.002
(0.157) (0.019) (0.010) (0.024) (0.023) (0.023) (0.020)
Age (years) -0.061 -0.014 0.004 -0.010 -0.013 -0.022 0.013
(0.011)

(0.001)

(0.001)

(0.002)

(0.002)

(0.002)

(0.001)

Rural -1.282 0.058 0.011 0.018 -0.067 0.058 0.000
(0.137)

(0.017)

(0.009) (0.021) (0.020)

(0.020)

(0.017)
Muslim -2.526 0.048 0.027 0.323 -0.169 -0.056 0.021
(0.854)

(0.105) (0.053) (0.129)

(0.125) (0.123) (0.109)
Religiosity -0.677 0.008 -0.006 0.114 -0.069 -0.065 0.028
(0.391) (0.048) (0.024) (0.059) (0.057) (0.056) (0.050)
Muslim Religiosity 0.904 -0.019 0.004 -0.151 0.064 0.042 -0.047
(0.414)

(0.051) (0.026) (0.062)

(0.061) (0.060) (0.053)
Father's education:
Elementary 1.701 -0.061 0.000 -0.001 0.037 -0.042 -0.039
(0.210)

(0.026)

(0.013) (0.032) (0.031) (0.030) (0.027)
Junior High 3.341 -0.095 -0.007 -0.012 0.052 -0.124 -0.018
(0.292)

(0.036)

(0.018) (0.044) (0.043) (0.042)

(0.037)
Senior High/University 4.333 -0.185 -0.017 -0.005 0.112 -0.163 -0.052
(0.299)

(0.037)

(0.019) (0.045) (0.044)

(0.043)

(0.038)
Mother's education:
Elementary 1.495 -0.033 -0.005 0.050 0.016 -0.067 0.017
(0.186)

(0.023) (0.012) (0.028) (0.027) (0.027)

(0.024)
Junior High 3.013 -0.107 -0.033 0.004 0.137 -0.144 -0.062
(0.291)

(0.036)

(0.018) (0.044) (0.042)

(0.042)

(0.037)
Senior High/University 3.832 -0.151 -0.027 0.066 0.105 -0.179 -0.051
(0.332)

(0.041)

(0.021) (0.050) (0.048)

(0.048)

(0.042)
Constant 10.343 1.345 -0.100 0.707 0.920 1.577 -0.178
(0.924)

(0.113)

(0.058) (0.139)

(0.135)

(0.133)

(0.117)
Standard errors in parentheses. *p<0.10 **p<0.05 ***p<0.01.
113
Chapter 5
Conclusion
This dissertation has oered three essays on the shaping and consequences of risk and
time preferences among Indonesian adults. Chapter 2 examines how risk and time
preferences vary with individual socioeconomic and cognitive characteristics. Chapter
3 builds on the rst, and explores the eect of the interviewer in elicited risk and time
preferences. Finally, Chapter 4 looks at how risk and time preferences may in
uence
the transition into adulthood by considering their correlations with the timing of
major life events: schooling, marriage, fertility, migration, and employment. All
three essays employ data from the fourth wave of the Indonesian Family Life Survey.
5.1 Demographic, Cognitive, and Interviewer Eects on
Preferences
Chapter 2 examines how Indonesians' preferences toward risk and time vary with
sociodemographic characteristics. It contributes to the body of evidence on hetero-
geneous preferences. I use data from the fourth wave of the Indonesian Family Life
Survey (IFLS-4), in which risk and time preferences were directly elicited using hypo-
thetical questions. I conrm that elicited risk and time preferences are heterogeneous
across Indonesian adults. Furthermore, they are systematically related to sex, age,
wealth, education, and cognition.
An unusual feature of the IFLS-4 elicitation method is it identies adults with
114
\nonstandard" risk and time preferences, in that they chose a dominated outcome
over a dominating one. In the context of expected utility, respondents with nonstan-
dard risk preference are either gamble averse or hyper-gamble loving. Respondents
with nonstandard time preference can be characterized as having a negative time
discount in the discounted utility framework. Another contribution of this paper is
to determine the correlates of these nonstandard preferences.
The main ndings of Chapter 2 are as follows. Men are less risk averse than
women, older respondents are more impatient, wealthier respondents are less risk
averse and less impatient, better educated adults are less impatient, and respondents
with better cognitive capacity proxied by episodic memory are less impatient. In
addition, cognitive capacity is associated with a lower likelihood of nonstandard pref-
erence. Respondents with better episodic memory, a measure of cognition, are less
likely to be nonstandard across all specications. The evidence for the role of educa-
tion is somewhat weaker, but the relationship between education and the likelihood
of nonstandard preference still has a negative sign.
Chapter 3 is motivated by Binswanger's (1980) nding of interviewer bias in
survey-elicited risk preference. The objective is twofold. First, I investigate whether
interviewers signicantly aect respondents' answers to the elicitation questions. Sec-
ond, I seek to ascertain whether the inclusion of interviewer eects changes the es-
timated eects of demographic and cognitive variables from the previous chapter (if
they do, then interviewer characteristics are a source of omitted variable bias).
With respect to the second objective, I nd no evidence that their omission biases
the coecients of the demographic and cognitive variables. With respect to the rst
objective, I nd evidence of interviewer eects on preferences. In particular, signi-
cant language eects are evident in the likelihood of nonstandard preference. Where
interviewers and respondents do not share a daily spoken language, respondents are
115
more likely to show nonstandard preferences. This eect is found only in interviews
conducted in languages other than the national language. This nding suggests that
interviewer language ability could be an especially important consideration for sur-
vey designers where the surveyed population has great language diversity. To my
knowledge, there is currently no paper in the economics literature examining lan-
guage eects in survey enumeration, which is surprising given the multilingual nature
of many parts of the world in which populations are surveyed.
5.2 Eect of Preferences on the Transition into Adulthood
Chapter 4 asks whether individual risk and time preferences drive dierences in the
transition into adulthood. To investigate this question, I again draw on data from
IFLS-4. I consider ve markers of adulthood: schooling attainment, marriage, fertility
behavior, migration, and full time employment. The results indicate that risk and
time preferences do play a signicant role in some markers of adulthood transition.
Men who are more risk averse or impatient, and women who are more impatient
attain fewer years of schooling. Risk averse men and women marry earlier, and so
do impatient women. Married couples positively sort on risk and time preferences.
Risk and time preferences are not signicantly correlated with female fertility timing.
However, high risk aversion and low impatience are signicantly negatively correlated
with birth control use among women. Highly risk averse men are signicantly less
likely to migrate. Risk and time preferences are not signicantly correlated with entry
into full time employment, but conditional on being employed, highly risk averse men
and women are signicantly less likely to be self employed.
The jury is still out as to the precise nature of the role of risk and time preferences
in the transition into adulthood, due to endogeneity concerns with the explanatory
116
variable of interest. The next step in this study is to instrument the relevant prefer-
ence measure with a suitable instrument such as rainfall shocks or natural disasters.
117
Comprehensive bibliography
Afriat, Sydney N, \The Construction of Utility Functions from Expenditure Data,"
International Economic Review, 1967, 8 (1), 67{77.
Ahn, Taehyun, \Attitudes toward Risk and Self-Employment of Young Workers,"
Labour Economics, 2010, 17, 434{442.
Akerlof, George A., \Social Distance and Social Decisions," Econometrica, 1997,
65 (5), 1005{1027.
Andersen, Steen, Glenn W. Harrison, Morten I. Lau, and E. Elisabet
Rutstr om, \Eliciting Risk and Time Preferences," Econometrica, 2008, 76 (3),
583{618.
Anderson, C. Leigh, Maya Dietz, Andrew Gordon, and Marieka Klawitter,
\Discount Rates in Vietnam," Economic Development and Cultural Change, 2004,
52 (4), 873{887.
Barr, Abigail and Truman Packard, \Revealed Preference and Self-Insurance:
Can We Learn from the Self-Employed in Chile?," World Bank Policy Research
Working Paper 2754, 2002.
Barsky, Robert, F. Thomas Juster, Miles Kimball, and Matthew Shapiro,
\Preference Parameters and Behavioral Heterogeneity: An Experimental Approach
in the Health and Retirement Study," Quarterly Journal of Economics, 1997, 112
(2), 537{79.
118
Becker, Gary, \A Theory of Marriage: Part I," Journal of Political Economy, 1973,
81 (4), 813{846.
Becker, Gary S. and Casey B. Mulligan, \The Endogenous Determination of
Time Preference," Quarterly Journal of Economics, 1997, 112 (2), 729{758.
Belzil, Christian and Marco Leonardi, \Risk Aversion and Schooling Decisions,"
HAL Working Paper 00411099, 2009.
Benjamin, Daniel J., Sebaastian A. Brown, and Jesse M. Shapiro, \Who
is `Behavioral'? Cognitive Ability and Anomalous Preferences," Working Paper,
May 2006.
Binswanger, Hans, \Attitudes Toward Risk: Experimental Measurement in Rural
India," American Journal of Agricultural Economics, 1980, 62 (3), 395{407.
Black, Sandra E, Paul J Devereux, and Kjell G Salvanes, \Why the Apple
Doesn't Fall Far: Understanding Intergenerational Transmission of Human Capi-
tal," American Economic Review, 2005, 95 (1), 437{449.
Blom, Annelies G., Edith D. de Leeuw, and Joop J. Hox, \Interviewer Eects
on Nonresponse in the European Social Survey," Journal of Ocial Statistics, 2011,
27 (2), 359{377.
Browning, Martin, Pierre-Andr e Chiappori, and Yoram Weiss, \Family Eco-
nomics," 2011.
Bryan, Gharad, Dean Karlan, and Scott Nelson, \Commitment Devices,"
Annual Review of Economics, 2010, 2, 671{698.
Burks, Stephen V., Jerey P. Carpenter, Lorenz Goette, and Aldo Rusti-
chini, \Cognitive Skills Aect Economic Preferences, Strategic Behavior, and Job
119
Attachment," Proceedings of the National Academy of Sciences, 2009, 106 (19),
7745{50.
Cameron, Lisa and Manisha Shah, \Risk-Taking Behavior in the Wake of Natural
Disasters," IZA Discussion Paper 6756, March 2012.
Carpenter, Jerey and Juan Camilo Cardenas, \Behavioural Development
Economics: Lessons from Field Labs in the Developing World," Journal of Devel-
opment Studies, 2008, 44 (3), 311{338.
Cassar, Alessandra, Andrew Healy, and Carl Von Kessler, \Trust, Risk, and
Time Preferences after a Natural Disaster: Experimental Evidence from Thailand,"
Working Paper, June 2011.
Cattell, Raymond B. and A Critical Experiment, \Theory of Fluid and Crys-
tallized Intelligence," Journal of Educational Psychology, 1963, 54 (1), 1{22.
Chabris, Christopher F., David I. Laibson, and Jonathan P. Schuldt, \In-
tertemporal Choice," in Steven Durlauf and Lawrence Blume, eds., The New Pal-
grave Dictionary of Economics, 2nd ed., London: Palgrave Macmillan, 2007.
Chen, Stacey H, \Risk Attitude and College Attendance," University at Albany
Discussion Paper 03-03, 2003.
Choi, Syngjoo, Raymond Fisman, and Douglas Gale, \Consistency and Het-
erogeneity of Individual Behavior under Uncertainty," American Economic Review,
2007, 97 (5), 1921{1938.
, Shachar Kariv, Wieland M uller, and Dan Silverman, \Who Is (More)
Rational?," 2013.
120
Compton, Janice, \Why do Smokers Divorce? Time Preference and Marital Sta-
bility," Working Paper, 2009.
Courtemanche, Charles J., Garth Heute, and Patrick McAlvanah, \Impa-
tience , Incentives , and Obesity," Georgia State University Working Paper 2012-
7-4, July 2012.
Croson, Rachel and Uri Gneezy, \Gender Dierences in Preferences," Journal
of Economic Literature, 2009, 47 (2), 448{474.
Cuxart, Anna, \Design Eect and Interviewer Eects: Evidence from the Spanish
ESS3-2006 Data," Universitat Pompeu Fabra Working Paper, 2011.
Davis, Rachel E., Mick P. Couper, Nancy K. Janz, Cleo H. Caldwell, and
Ken Resnicow, \Interviewer Eects in Public Health Surveys," Health Education
Research, 2010, 25 (1), 14{26.
Deaton, Angus, The Analysis of Household Surveys: A Microeconometric Approach
to Development Policy, Baltimore, MD: Johns Hopkins University Press, 1997.
DellaVigna, Stefano and M. Daniele Paserman, \Job Search and Impatience,"
Journal of Labor Economics, 2005, 23 (3), 527{588.
Dohmen, Thomas, Armin Falk, David Human, and Uwe Sunde, \Are
Risk Aversion and Impatience Related to Cognitive Ability?," American Economic
Review, 2010, 100 (3), 1238{1260.
, , , and , \The Intergenerational Transmission of Risk and Trust Atti-
tudes," Review of Economic Studies, 2012, 79 (2), 645{677.
121
, , , , Jurgen Schupp, and Gert Wagner, \Individual Risk Attitudes:
Measurement , Determinants and Behavioral Consequences," Journal of the Euro-
pean Economic Association, 2011, 9 (3), 522{550.
Dohrenwend, Barbara Snell Bruce P. and John Colombotos, \Social Distance
and Interviewer Eects," Public Opinion Quarterly, 1968, 32 (3), 410.
Du
o, Esther, \Schooling and Labor Market Consequences of School Construction
in Indonesia: Evidence from an Unusual Policy Experiment," American Economic
Review, 2001, 91 (4), 795{813.
Eckel, Catherine C. and Philip J. Grossman, \Men, Women and Risk Aversion:
Experimental Evidence," in Charles R Plott and Vernon L Smith, eds., Handbook
of Experimental Economics Results, Vol. 1, Elsevier, 2008, chapter 113, pp. 1061{
1073.
Frederick, Shane, \Cognitive Re
ection and Decision Making," Journal of Eco-
nomic Perspectives, 2005, 19 (4), 25{42.
, George Loewenstein, Ted O'Donoghue, and Ted Odonoghue, \Time Dis-
counting and Time Preference: A Critical Review," Journal of Economic Literature,
June 2002, 40 (2), 351{401.
Greene, William H., Econometric Analysis, 6th ed., New Jersey: Pearson Educa-
tion, 2008.
Guiso, Luigi and Monica Paiella, \Risk Aversion, Wealth and Background Risk,"
Journal of the European Economic Association, 2008, 6 (6), 1109{1150.
, Paola Sapienza, and Luigi Zingales, \Time Varying Risk Aversion," 2011.
122
Hamoudi, Amar, \Risk Preferences in Households and Families," UCLA Working
Paper, December 2006.
Hardeweg, Bernd, Lukas Menkho, and Hermann Waibel, \Experimentally-
Validated Survey Evidence on Individual Risk Attitudes in Rural Thailand," Uni-
versit at Hannover Discussion Paper 464, 2011.
Harris, John R and Michael P Todaro, \Migration, Unemployment and De-
velopment: A Two-Sector Analysis," American Economic Review, 1970, 60 (1),
126{142.
Harrison, Glenn W. and E. Elisabet Rutstr om, \Risk Aversion in the Labora-
tory," in James C. Cox and Glenn W. Harrison, eds., Risk Aversion in Experiments
(Research in Experimental Economics, Volume 12), Vol. 12, Emerald Group Pub-
lishing Limited, 2008, pp. 41{196.
Hartog, Joop, Ada Ferrer i carbonell, and Nicole Jonker, \Linking Measured
Risk Aversion to Individual Characteristics," Kyklos, 2002, 55 (1), 3{26.
Hoddinott, John, \A Model of Migration and Remittances Applied to Western
Kenya," Oxford Economic Papers, 1994, 46 (3), 459{476.
Holt, Charles A and Susan K Laury, \Risk Aversion and Incentive Eects,"
American Economic Review, 2002, 92 (5), 1644{1655.
Horn, John L., \Remodeling Old Models of Intelligence: Gf-Gc Theory," in Ben-
jamin B. Wolman, ed., Handbook of Intelligence, New York: Wiley, 1985, pp. 267{
300.
123
Hryshko, Dmytro, Bent E. Sorensen, and Maria Jose Luengo-Prado, \Child-
hood Determinants of Risk Aversion : The Long Shadow of Compulsory Educa-
tion," Quantitative Economics, 2011, 2 (1), 37{72.
Jaeger, David A, Thomas Dohmen, Armin Falk, David Human, Uwe
Sunde, and Holger Bonin, \Direct Evidence on Risk Attitudes and Migration,"
Review of Economics and Statistics, 2010, 92 (3), 684{689.
Kachelmeier, Steven J. and Mohamed Shehata, \Examining Risk Preferences
Under High Monetary Incentives: Experimental Evidence from the People's Re-
public of China," American Economic Review, 1992, 82 (5), 1120{1141.
Keeley, Michael C., \The Economics of Family Formation," Economic Inquiry,
1977, 15 (2), 238{250.
Kemsley, William F. F., \Interviewer Variability in Expenditure Surveys," Journal
of the American Statistical Association, 1965, 128 (1), 118{139.
Knight, John, Sharada Weir, and Tassew Woldehanna, \The Role of Educa-
tion in Facilitating Risk-Taking and Innovation in Agriculture," Journal of Devel-
opment Studies, 2003, 39 (6), 1{22.
Krupka, Erin L. and Melvin Stephens, \The Stability of Elicited Discount Rates
Over Time," University of Michigan, May 2010.
Laibson, David I., \Golden Eggs and Hyperbolic Discounting," Quarterly Journal
of Economics, 1997, 112 (2), 443{477.
Leigh, J. Paul, \Accounting for Tastes: Correlates of Risk and Time Preferences,"
Journal of Post Keynesian Economics, 1986, 9 (1), 17{31.
124
Loewenstein, George, \Anticipation and the Valuation of Delayed Consumption,"
The Economic Journal, 1987, 97 (387), 666{684.
Loureiro, Maria L and Justus Lotade, \Interviewer Eects on the Valuation of
Goods with Ethical and Environmental Attributes," Environmental and Resource
Economics, 2005, 30 (1), 49{72.
Mazzocco, Maurizio and Shiv Saini, \Testing Ecient Risk Sharing with Het-
erogeneous Risk Preferences," American Economic Review, 2012, 102 (1), 428{68.
McArdle, John J., Emilio Ferrer-Caja, Fumiaki Hamagami, and
Richard W. Woodcock, \Comparative Longitudinal Structural Analyses of the
Growth and Decline of Multiple Intellectual Abilities over the Life Span," Devel-
opmental Psychology, 2002, 38 (1), 115{142.
, James P. Smith, and Robert Willis, \Cognition and Economic Outcomes in
the Health and Retirement Survey," NBER Working Paper Series 15266, August
2009.
Meier, Stephan and Charles D. Sprenger, \Stability of Time Preferences," IZA
Discussion Paper 4756, February 2010.
Miyata, Sachiko, \Household's Risk Attitudes in Indonesian Villages," Applied Eco-
nomics, 2003, 35 (5), 573{583.
O'Muircheartaigh, Colm and Pamela Campanelli, \The Relative Impact of
Interviewer Eects and Sample Design Eects on Survey Precision," Journal of the
Royal Statistical Society. Series A (Statistics in Society), 1998, 161 (1), 63{77.
Perez-Arce, Francisco, \The Eect of Education on Time Preferences," RAND
Working Paper Series WR- 844, 2011.
125
Pissarides, Christopher A, \Risk , Job Search , and Income Distribution," Journal
of Political Economy, 1974, 82 (6), 1255{1267.
Portner, Claus, \Children as Insurance," Journal of Population Economics, May
2001, 14, 119{136.
, \Gone with the Wind? Hurricane Risk, Fertility and Education," 2008.
Riphahn, Regina T. and Oliver Ser
ing, \Item Non-Response on Income and
Wealth Questions," Empirical Economics, September 2005, 30 (2), 521{538.
Rosenzweig, Mark R. and Oded Stark, \Consumption Smoothing, Migration,
and Marriage: Evidence from Rural India," Journal of Political Economy, January
1989, 97 (4), 905.
Rubalcava, Luis and Graciela Teruel, \User's Guide for the Mexican Family Life
Survey: Second Wave," Technical Report 2008.
, , and Duncan Thomas, \Investments , Time Preferences and Public Transfers
Paid to Women," Economic Development and Cultural Change, 2009, 57 (3), 507{
538.
Samuelson, Paul, \A Note on Measurement of Utility," Review of Economic Stud-
ies, 1937, 4 (2), 155{161.
Schmidt, Lucie, \Risk Preferences and the Timing of Marriage and Childbearing,"
Demography, 2008, 45 (2), 439{460.
Schulhofer-Wohl, Sam, \Heterogeneity and Tests of Risk Sharing," Journal of
Political Economy, 2011, 119 (5), 925{958.
126
Slovic, Paul, \Assessment of Risk Taking Behavior," Psychological Bulletin, 1964,
61 (3), 220{233.
Spivey, Christy, \Desperation or Desire? The Role of Risk Aversion in Marriage,"
Economic Inquiry, 2010, 48 (2), 499{516.
Stark, Oded, The Migration of Labor, Oxford: Blackwell, 1991.
Stigler, George J. and Gary S. Becker, \De Gustibus Non Est Disputandum,"
American Economic Review, 1977, 67 (2), 76{90.
Strauss, John, Firman Witoelar, Bondan Sikoki, and Anna Marie Wattie,
\The Fourth Wave of the Indonesia Family Life Survey: Overview and Field Report
Volume 1," RAND Labor and Population Working Paper WR-675/1-NIA/NICHD,
April 2009.
Tanaka, Tomomi, Colin F. Camerer, and Quang Nguyen, \Risk and Time
Preferences: Linking Experimental and Household Survey Data from Vietnam,"
American Economic Review, 2010, 100 (1), 557{571.
Thomas, Duncan, Firman Witoelar, Elizabeth Frankenberg, Bondan
Sikoki, John Strauss, Cecep Sumantri, and Wayan Suriastini, \Cutting
the Costs of Attrition: Results from the Indonesia Family Life Survey," Journal of
Development Economics, 2012, 98 (1), 108{123.
van Praag, Bernard and Adam Booij, \Risk Aversion and the Subjective Time
Discount Rate: A Joint Approach," CESifo Working Paper 923, 2003.
van Praag, CM and JS Cramer, \The Roots of Entrepreneurship and Labour
Demand: Individual Ability and Low Risk Aversion," Economica, 2001, 68 (269),
45{62.
127
Walker, Thomas Steven, \Risk and Adoption of Hybrid Maize in El Salvador,"
Food Research Institute Studies, 1981, 18 (1), 59{88.
128

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays on development and health economics

PDF

Three essays on cooperation, social interactions, and religion

PDF

Individual vs. group behavior: an empirical assessment of time preferences using experimental data

PDF

Essays on health economics

PDF

Essays on economics of education and private tutoring

PDF

Essays on health and aging with focus on the spillover of human capital

PDF

Essays on education: from Indonesia to Los Angeles, CA

PDF

Three essays on health & aging

PDF

Beyond revealed preferences: how gender, relative socioeconomic status, and social norms drive happiness and behavior

PDF

The role of the timing of school changes and school quality in the impact of student mobility: evidence from Clark County, Nevada

PDF

Essays on family and labor economics

PDF

A "second chance" program for vulnerable young adults: a program evaluation and a randomized controlled trial of adjunctive motivational interviewing to improve retention

PDF

Three essays on economics of early life health in developing countries

PDF

Essays in development economics

PDF

Essays in the economics of education and conflict

PDF

US biosimilar market entry and uptake: US commercial payer preferences in access decision making

PDF

Marital matching in West Africa: an examination of interethnic and interreligious first marriages in Benin

PDF

The determinants and measurement of human capital

PDF

Essays in labor economics: demographic determinants of labor supply

PDF

Three essays on strategic commuters with late arrival penalties and toll lanes

Asset Metadata

Creator Ng, James L. (author)

Core Title Essays on the empirics of risk and time preferences in Indonesia

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 06/27/2013

Defense Date 04/29/2013

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag adulthood transition,Indonesia,interviewer effect,OAI-PMH Harvest,risk aversion,time preference

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Strauss, John A. (committee chair), Nugent, Jeffrey B. (committee member), Painter, Gary Dean (committee member), Wilkie, Simon J. (committee member)

Creator Email jamesltng@gmail.com,jamesng@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-280742

Unique identifier UC11293579

Identifier etd-NgJamesL-1712.pdf (filename),usctheses-c3-280742 (legacy record id)

Legacy Identifier etd-NgJamesL-1712.pdf

Dmrecord 280742

Document Type Dissertation

Format application/pdf (imt)

Rights Ng, James L.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA