Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Improvements in women's status, decision-making and child nutrition: evidence from Bangladesh and Indonesia
(USC Thesis Other)
Improvements in women's status, decision-making and child nutrition: evidence from Bangladesh and Indonesia
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
IMPROVEMENTS IN WOMEN’S STATUS, DECISION-MAKING
AND CHILD NUTRITION:
EVIDENCE FROM BANGLADESH AND INDONESIA
by
Radheeka Ranmali Jayasundera
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(SOCIOLOGY)
August 2012
Copyright 2012 Radheeka Ranmali Jayasundera
ii
Dedication
To
My Parents
Sunil and Mangalee Jayasundera
iii
Acknowledgements
I would like to thank my wonderful advisor Prof. Lynne Casper for guiding me in my
research throughout these years, for instilling in me a love for demography, and for
pushing me to fully realize my true potential as a researcher. Special thanks go to Prof.
John Strauss for introducing me to the fascinating world of development economics, and
for inspiring me to pay great attention to detail in statistical modeling. A gracious thank
you goes to Prof. Tim Biblarz for instilling a positive attitude in me, and for always been
there to support me despite his busy schedule as department chair. I would also like to
thank Prof. Merril Silverstein who enlightened me with his wisdom, insight and expertise
in family research. Special thanks also go to Prof. Amon Emeka for his thoughtful advice
on my research and for his encouragement and support throughout my years at USC.
I gratefully acknowledge financial support from the University of Southern
California, the Wallis Annenberg Foundation and the Frederick and Dorothy Quimby
Memorial Scholarship.
I have enjoyed the friendship and support of Sandra Florian, Sum-yan Ng,
Hyeyoung Kwon, Emir Estrada, Edson Rodriguez, Nimna Ranatunga, and Joy Lam who
have been great friends and colleagues at USC.
This dissertation would not have been possible without my family’s support. I
thank my parents Sunil and Mangalee for their encouragement, unwavering support and
love that guided me through the long road of graduate school. I owe special thanks to my
siblings Nadeeka, Mandu, Devin, Heminda and my little niece Nadine for always being
there for me, for lifting my spirits when I was feeling down and for keeping me abreast of
iv
the news at home. Last, but not least I thank my best friend and fiancé – Harsha for his
continuous emotional support, love and phenomenal patience. My family is what makes
me whole; I thank them for giving me strength when I’ve needed it most.
v
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables vii
List of Figures ix
Abstract x
Chapter One: Women’s Status in Developing Countries: An Introduction 1
1. Introduction 1
2. The Genesis of Gender Analysis in Development 3
3. The Definition of Women’s Status 7
4. Micro and Macro-Level Processes 11
5. Modernization Theory 12
6. Summary 15
Chapter Two: A Summary of the Socioeconomic and Demographic Contexts 18
of Bangladesh and Indonesia
1. Introduction 18
2. Bangladesh 18
3. Indonesia 24
4. Summary 31
Chapter Three: Methods 33
1. Introduction 33
2. Maximum Likelihood Estimation 33
3. Oaxaca-Blinder Decomposition 34
Chapter Four: Improvements in Women’s Participation in Household 42
Decision-Making in Bangladesh from 1999 to 2007
1. Introduction 42
2. Background 44
3. Data 54
4. Methods and Analyses 63
5. Results 66
6. Conclusion 86
vi
Chapter Five: The Role of Social Context in Household Decision-Making in 88
Bangladesh and Indonesia
1. Introduction 88
2. Background 89
3. Data 96
4. Empirical Specification 100
5. Results 102
6. Conclusion 122
Chapter Six: Women’s Status and Child Nutrition in Bangladesh 125
1. Introduction 125
2. Background 125
3. Data 140
4. Empirical Strategy 144
5. Results 146
6. Discussion 152
7. Conclusion 154
Chapter Seven: Discussion and Conclusion 157
1. Background 157
2. Methods Used 159
3. Results 161
4. Conclusion 174
Bibliography 176
Appendices 191
Appendix – A: Additional Tables for Chapter Four 191
Appendix – B: Additional Tables for Chapter Five 192
vii
List of Tables
Table 4.1 Summary Statistics of the Explanatory Variables, Bangladesh in
1999 and 2007
67
Table 4.2 Pooled Sample Bivariate Regression: Women’s Participation in
Decision-Making Predicted by Women’s Characteristics and
Control Variables
69
Table 4.3 Woman’s Participation in Decision Making: Combined Scale (0-5) 76
Table 4.4 Woman’s Participation in Decision Making Regarding Her Own
Health
77
Table 4.5 Woman’s Participation in Decision Making Regarding Large
Purchases
78
Table 4.6 Woman’s Participation in Decision Making Regarding Daily
Purchases
79
Table 4.7 Woman’s Participation in Decision Making Regarding Visits to
Family and Friends
80
Table 4.8 Woman’s Participation in Decision Making Regarding Children’s
Health Care
81
Table 4.9A Oaxaca-Blinder Decomposition: Predicted Changes in Decision-
Making
85
Table 4.9B Oaxaca-Blinder Decomposition: Percentage Differences in
Decision-Making if Characteristics in 1999 were at 2007 Levels
85
Table 5.1 Summary Statistics of the Explanatory Variables in Indonesia 104
Table 5.2 Bivariate Regression Results for 1997 and 2007 in Indonesia 106
Table 5.3 Woman’s Participation in Decision Making Regarding Children’s
Health in Indonesia
109
Table 5.4 Woman’s Participation in Decision Making Regarding Large
Expensive Purchases in Indonesia
111
Table 5.5 Oaxaca-Blinder Decomposition: Indonesia in 1997 and 2007 113
viii
Table 5.6 Summary Statistics of the Explanatory Variables: Bangladesh and
Indonesia in 2007
115
Table 5.7 Bivariate Regression Results: Bangladesh and Indonesia, 2007 116
Table 5.8 Woman’s Participation in Decision Making Regarding Children’s
Health: Comparison Between Indonesia and Bangladesh in 2007
118
Table 5.9 Woman’s Participation in Decision Making Regarding Large
Purchases: Comparison Between Indonesia and Bangladesh in
2007
119
Table 5.10 Oaxaca-Blinder Decomposition: Bangladesh and Indonesia in
2007
121
Table 6.1 Summary Statistics of the Explanatory Variables: Child Stunting 147
Table 6.2 Multivariate Logistic Regressions: Predicting Child Stunting in
Bangladesh, 1999 and 2007
149
Table 6.3 Oaxaca-Blinder Decomposition: Child Stunting – All Children 151
Table 6.4A Oaxaca-Blinder Decomposition: Child Stunting – Boys 153
Table 6.4B Oaxaca-Blinder Decomposition: Child Stunting – Girls 153
Table A1 Pooled Sample F test / Chi-square statistics (p values in
parentheses)
191
Table A2 Joint Tests of Significance for Categorical Variables for Table 4.3
to 4.8
191
Table B1 Pooled Sample Chi-Square test for Table 5.3 and 5.4 192
Table B2 Joint Tests of Significance for Categorical Variables for Table 5.3
and 5.4
192
Table B3 Pooled Sample Chi-Square test statistics for Table 5.8 and 5.9 193
Table B4 Joint Tests of Significance for Categorical Variables for Table 5.8
and 5.9
193
ix
List of Figures
Figure 1.1 Males per 100 Females Among 0-4 Year Olds 5
Figure 3.1 Oaxaca-Blinder Decomposition 38
Figure 4.1 Excerpts from the 1999 BDHS Women’s Questionnaire 56
Figure 4.2 Excerpts from the 2007 BDHS Women’s Questionnaire 56
Figure 4.3
Weighted Percentage of Women who Participate in Decision-
Making in Bangladesh
59
Figure 5.1 Weighted Percentages of Women who Participate in Decision
Making in Indonesia from 1997 to 2007
99
Figure 5.2
Weighted Percentages of Women who Participate in Decision
Making - Indonesia and Bangladesh in 2007
99
Figure 6.1A Conceptual Framework 136
Figure 6.1B Operational Framework 137
Figure 6.2 Height-for-Age Z Score Distribution of Children Under 5 Years –
Normal Curves
143
Figure 6.3 Cumulative Distribution Functions of Height-for-Age Z-scores in
1999 and 2007
144
x
Abstract
The goal of this dissertation is (1) to analyze how the determinants of women’s decision-
making power and child stunting have changed over time and (2) to identify the sources
of improvement in women’s decision-making and child health in two less-developed
Muslim countries. According to past literature, the agents of change in women’s
decision-making at the household-level should derive from changes in women’s
education and wage employment. Sources of change in child health should derive from
parental education, household wealth and community infrastructure.
This dissertation contributes to the current literature by looking at change in
women’s status and its determinants over time, and its implications for child health in one
of the most gender biased and malnourished societies in the world—Bangladesh. It also
examines the role of social context on women’s status by comparing two predominantly
Muslim societies—Bangladesh and Indonesia that had similar origins in terms of
economic and social development.
In the first paper, I ask two questions: (1) what factors predict women’s decision-
making in Bangladesh in 1999 and 2007? and (2) do increases in women’s
socioeconomic characteristics contribute to improvements in women’s decision-making
in Bangladesh from 1999 to 2007? I use logistic regression and Oaxaca-Blinder
decomposition to answer these questions. The results indicate that, the strength of the
relationships between predictors and women’s decision-making has changed over time.
And the mean increase in women’s socioeconomic status only has a modest effect on
decision-making. Much of the difference between 1999 and 2007 is accounted for by the
xi
intercept, which may have subsumed unobserved variables that represent change in
norms and attitudes towards women and economic development.
In the second paper, I attempt to answer the same questions in Indonesia. The
results indicate that improvements in women’s decision-making from 1997 to 2007 are
partly due to improvements in women’s socioeconomic status, but similar to Bangladesh,
the majority of the improvement derives from the differences in intercepts. In the second
part of this paper, I harmonize data from Bangladesh and Indonesia to examine the role of
social context on women’s decision-making. The results suggest that the underlying
cultural context in Indonesia and Bangladesh plays an important role in determining
women’s status regardless of their socioeconomic characteristics. Ethnographic evidence
from Indonesia and Bangladesh on family systems and traditions support these results.
Therefore, it can be concluded that social context is an important factor (net of
socioeconomic characteristics) in women’s decision-making.
In the third paper, I look at the impact of women’s socioeconomic status on child
stunting in Bangladesh, where child malnutrition is widespread and detrimental to
development. In recent years, the number of children who are stunted has been
significantly reduced. The results from Oaxaca-Blinder decompositions reveal two
things: 1) improvements in women’s education (net of partner’s education), delayed first
marriage, reduction in fertility (i.e. the lower birth order of the child) and access to
information through television viewership account for about half of the decline in child
stunting. The other half is accounted for by improvements in household wealth,
availability of electricity, drinking water and advanced sewerage systems. Because
xii
parents favor sons over daughters in Bangladesh, the effect of parent’s education on
stunting is weaker for girls than for boys. In fact, the majority of the reduction in stunting
for girls comes from improvements in household assets and community infrastructure
that are usually available for all children in the household regardless of their gender.
These results partly support the notion that increases in women’s socioeconomic
characteristics have an impact on improvements in decision-making and child nutrition in
Bangladesh and Indonesia. But other community level changes such as norms and
attitudes towards women, improvements in infrastructure and wealth, which usually
accompany modernization and economic development, also play an important role in the
improvements of women’s decision-making and child nutrition. From a welfare policy
point of view, these results are encouraging; as empowering women in developing
countries have had a positive effect on their wellbeing as well as their children’s
wellbeing.
1
Chapter One
Women’s Status in Developing Countries: An Introduction
1. Introduction
Since the 1970s ‘gender’ has been a central focus of development studies. From being a
critical and political irritant, the concern for gender analysis has now become de rigeur
for development analysts and practitioners (Pearson 2007). Development programs in the
past were directed at improving the wellbeing of households and communities as a whole,
and development interventions were usually aimed at the householders (i.e. usually men).
It was assumed that improvements made at the householder’s level would trickle down to
other family members including women and children. Thus, in the early development
literature, reflections on the particular problems of women were few and far between.
Ester Boserup’s Women’s Role in Economic Development published in 1970 launched
the first thorough feminist attack on the conventional view of development and
modernization, which resulted in the new development agenda “Women in
Development.”
Since then, a myriad of studies have focused on the economic and social benefit
of empowering women as a form of human capital investment. Access to education,
employment, credit, contraceptive use, etc. has been identified as important avenues in
which women are empowered. Therefore, international organizations such as the World
Bank and United Nations, and local governments have taken significant steps to improve
women’s position in developing societies as a pathway for economic development. As a
2
result, there have been massive improvements in the women’s socioeconomic status in
developing countries.
The goal of this dissertation is (1) to analyze how the determinants of women’s
decision-making power and child stunting has changed over time and (2) to identify the
sources of improvement in women’s decision-making and child health in developing
countries. According to past literature, the agents of change in women’s decision-making
at the household-level should derive from changes in women’s education and wage
employment. Sources of change in child health should derive from parent’s education,
household wealth and community infrastructure.
This study contributes to the current literature by looking at change in women’s
status and its determinants over time, and its implications for child health in one of the
most gender biased and malnourished societies in the world—Bangladesh. It also
examines the role of social context on women’s status by comparing two predominantly
Muslim societies—Bangladesh and Indonesia that had similar origins in terms of
economic and social development.
In this chapter, I will first discuss the origin of gender analysis in the development
literature to explain why it is instrumental to involve women in economic development.
Then I will define what women’s status means in the development context, and how this
term is used in this dissertation. Next, I will explain how micro-level and macro-level
analyses help us to understand population level changes in women’s status. Finally, I will
review modernization theory that provides us some insights as to how economic
development and women’s status are linked.
3
2. The Genesis of Gender Analysis in Development
Although gender analysis has been an important subfield in sociology, history and
anthropology for more than a century, in the development literature, gender analysis was
acknowledged only after the publication of Ester Boserup’s Women’s Role in Economic
Development in 1970. Her work led to the ‘Women in Development’ thesis, which
focuses on how economic development affects women and how women can become
agents of change in less-developed societies. As a result a new development agenda
highlighted in the UN International Year for Women 1975, the UN Decade on Women
1976-1985 and the four world conferences on women from 1975-1995 (Corner 2008).
Boserup explains that traditional forms of agricultural preparation of land were
performed by men (e.g. felling trees, slash and burn technology). But, more time
consuming tasks such as planting, weeding and harvesting were done by women. In fact,
women spent more hours per week on farming activities than men. Therefore, women’s
labor was an important economic commodity for the agricultural communities in addition
to their reproductive and domestic utility. For this reason, at the time of marriage, a man
pays a bride price (transfer of assets from groom’s family to bride’s family) to the
woman’s family in exchange for her labor and fertility. However, with increasing
population density, land became scarce, and the fallow periods were shortened which
made the soil less fertile. New technologies were imported to clear and till the land (e.g.
plough). These technologies were mainly used by men who then devoted longer hours to
farm work than women; thus changing the gender dynamics of agricultural societies.
Agricultural changes also affected land-owning patterns, which evolved from communal
4
ownership during slash and burn period to individual ownership. This led to an income
disparity between poorer landless families and wealthy landowners who hired labor to
work in their fields. Women’s labor in these wealthy families was then replaced by hired
labor. As a result of these changes, a woman’s worth was reduced to her fertility, and the
bride price was replaced by the dowry (transfer of assets from the bride’s family to the
groom’s family). In Asia when plough cultivation and landownership became
predominant, women were exempted from farm work, only performed domestic duties
and lived in seclusion in their homes. This arrangement greatly reduced women’s
mobility and autonomy (Boserup 1970).
Although gender relations in developing countries are much more complex than
Boserup originally described, her work legitimized the scholarship of women activists
around the world who were demanding that women’s economic activities be recognized
and rewarded, not ignored and undercut, as most international assistance programs were
doing in the 1960s (Tinker 2007). Her book drew attention to women’s contributions to
productive work and critiqued the gender-biased allocation of resources, which led to a
growing literature on gender differentials in the allocation of food and health.
Later, Jennie Dey (1981) brought attention to unfair gender relations in land
tenure and farming in Gambia. In her study, she explains that men traditionally controlled
upland groundnuts and millet crops and women controlled low-value rice fields in
lowlands, which constantly flooded. But when the Taiwanese Agricultural Mission
introduced irrigated rice projects in Gambia, men saw the opportunity and took over the
rice fields from women. This project was not very successful because men were not as
5
experienced in rice cultivation as women, and were not able to fully benefit from the
program. Dye explains that if investments were focused on a few simple improvements in
women’s rain-fed and swamp rice production they could have obtained better results.
Further, by excluding women, the project increased women’s economic dependence on
men; who now controlled an additional food and cash crop. This was one of the first
studies that showed how ignorance of gender issues could lead to unsuccessful
development programs and increase the vulnerability of women.
In 1990, Nobel Laureate—Amartya Sen published an essay titled More Than 100
Million Women Are Missing which highlighted the consequences of differential resource
allocation for women. This publication had a profound impact on the way development
scholars viewed gender inequality in the developing world. Sen claimed that while the
ratios of males to females in the Western countries are nearly equal, in countries like
China, South Korea and India, there are far fewer women than men (see Figure 1.1).
Figure 1.1: Males per 100 Females Among 0-4 year Olds
6
Sen (1990) explored two possible explanations for this anomaly. One, which emphasizes
the cultural contrast between East and West, which implies that western civilization, is
less sexist than eastern civilization. And another that attributes it to different trajectories
of economic development, where unequal nutrition and health care for women is seen as
characteristic of underdevelopment. But, Sen points out that the East-West explanation is
flawed because countries like Japan have a ratio of males to females that is similar to
Europe and North America. At the same time he argues that the development trajectory
explanation is also dubious, because many poor countries in sub-Saharan Africa do not
have a deficit of women. He further explains that within the Indian states of Punjab and
Haryana (the richest and most economically advanced Indian states) the ratio of males to
females is very high—1.16 in contrast in a much poorer state—Kerala, the males to
females ratio is only 0.97. Sen explains that neither the alleged cultural contrast between
East and West, nor the simple hypothesis of female deprivation as a characteristic of
economic underdevelopment gives us an adequate understanding of the missing women
problem. He states, instead that we should consider the complex economic, social and
cultural factors that affect the status and power of women in the family. Sen concludes
his essay by providing some important variables that may explain the phenomenon of
missing women, such as gainful employment outside the home, education and ownership
of assets which are all supported by economic theory. But he argues that the more
important question is why such outside employment is more prevalent in certain societies
and not in others. Sen speculates that social and cultural variables, including family
7
organization, religion, and historic backgrounds might be the driving factors that
contribute to the relative deprivation of women in certain regions of the world.
3. The Definition of Women’s Status
According to the sociologist - Max Weber, the notion of status explains a form of
stratification in society. He states that status refers to differences between social groups in
the social honor or prestige they are accorded by others (Gerth and Mills 1946). Weber
explains that markers and symbols of status—such as housing, dress, manner of speech
and occupation help to shape an individual’s social standing in the eyes of others.
Goffman (1959) further explained that people with certain social statuses are expected to
perform certain roles that are “socially and culturally defined.”
Furthermore, sociologists distinguish between ascribed status and achieved status.
An ascribed status is one ‘assigned’ to you on the basis of biological factors such as sex,
race, skin-complexion or age. An achieved status is one that is earned through an
individual’s own effort. While we may like to believe that it is our achieved statuses that
are the most important, our ascribed statuses frequently override everything else we have
achieved as an individual (Hughes 1945; Becker 1963). And these statuses have priority
over all other statuses. The most prominent ascribed statuses are those based on Gender
and Race (Omi and Winant 1994). In the next few paragraphs, I will explore the meaning
of women’s status as a social status in the development context.
“Women’s Status” is a commonly used term in the development literature. It is
associated with women’s autonomy, power, empowerment, authority, valuation,
8
wellbeing and position in society (Smith, Ramakrishnan, Ndiaye, Haddad and Martorell
2003). Scholars classify the concept as being “non-unitary,” “multidimensional,” and
“multilevel,” as it can be viewed on more than one dimension (Safilios-Rothschild 1982).
In the past, it has been described as “elusive” (Curtin 1982; Dixon 1976) and “ambiguous
and poorly defined” (Powers 1985). This ambiguity makes it difficult to develop a
consensus on its definition. The failure to define it is however not an obstacle to
understanding the changes in women’s status, but it is very important to clearly specify
what “women’s status” mean, each time it is used (Mason 1986, 1993; Sen and Batliwala
2000; Pasternak, Ember, and Ember 1997).
There are certain commonalities in the varying array of definitions in the
women’s status literature. Most definitions address aspects of gender inequality and
concentrate on power, prestige, resource control and access (Mason 1985). Some authors
focus either on the position of women relative to others in the community or society as a
whole, or on their position at the interpersonal level (Safilios-Rothschild 1982). Williams
(1989) used power within the household, which includes a woman’s ability to influence
decisions about household financial matters, such as major purchases, wages, or income,
as well as other central decisions such as whether or not, and if so when to migrate or to
have an additional child. This “intra-household status” (Williams and Guest 1985) is of
particular importance, since variations in the distribution of power between household
members can have an effect on decision outcomes that are independent of more
commonly considered variables (Hull 1981).
9
In this dissertation I use the following definition: “Women’s status is women’s
power to exercise their preferences and choices in the households, communities and
nations in which they live.” I use the terms women’s power and women’s status
interchangeably. The above definition incorporates two important aspects.
First, it is founded on the concept of power, which is the ability to pursue one’s
goals, even in the face of adversity. Power is exercised through decision-making. And
decisions can be made alone or jointly with another person through a process of
bargaining and negotiation (Smith et al. 2003). But sometimes power is taken away by
way of deception, manipulation, violence, coercion or threat, which may lead to “non
decision-making,” in which a person accepts the status quo and allows others to make
decision for them (Kabeer 1999; Riley 1997; Safilios-Rothschild 1982; Sen 1990).
Economic, human, and social resources enhance power to exercise one’s preferences and
choices. Such economic resources may include work for pay (cash or in kind), assets
(jewelry, land, furniture), time, access to credit, etc. Human resources may include
education, skills and knowledge. Finally, social resources may include membership in
groups and access to social networks (Quisumbing and Maluccio 2003; Kabeer 1999, Sen
and Batliwala 2000). Thus, it can be said that, unequal women’s statuses reflect the
inequality in control over these resources.
The second part of the definition highlights the differences women experience not
only in their households but also in their communities and nations. Social norms, beliefs,
traditions, family systems (patriarchy vs. matriarchy), values and attitudes often dictate
differential roles, behaviors, rights and opportunities for women (Safilios-Rothschild
10
1982; World Bank 2001; Agarwal 1997; Kishor 1999; Kevane 2000; Sen and Batliwala
2000; England 2000). This is consistent with Weber’s and Goffman’s definition of status,
which is an individual’s social standing in the eyes of others and the expectation that
individuals with certain social statuses will perform certain roles that are socially and
culturally defined (Gerth and Mills 1946; Goffman 1959).
In societies where the purdah
1
is practiced, women cannot travel without been
accompanied by a male relative; this prevents them from going to school or obtaining
outside employment. Further, women may face greater risk of assault, or be treated as
intellectually inferior to men when they come into contact with institutions outside of
their homes (Smith et al. 2003). They may find certain female-specific services
unavailable for them such as maternity care. Thus, certain social norms, values, and
beliefs that govern social behaviors do not conceive some alternatives to be within the
realm of possibility for women (Kabeer 1999; Riley 1997; Sen and Batliwala 2000;
Barosso and Jacobson 2000; Smith et al. 2003). Thus, even if women and men have
egalitarian relationships at household level, societal norms and values may constrain
women’s access to resources and their decision-making power. In most societies, such
gender-biased social norms and beliefs influence household-level power relations
between couples. For example, norms and customs may require women to be silent when
men display anger, which prevents women from bargaining or negotiating with their
spouses (Kevane 2000). Further, patriarchal family systems such as found in Bangladesh
put young brides at the bottom of the pecking order in the household of her husband’s
1
A curtain or screen, used to keep women separate from men.
11
parents. This means according to the patriarchal social norms the woman will be the last
to receive food and other resources and has little influence on household decisions;
including decisions regarding herself and her children (Cain 1991). Additionally,
discriminatory labor policies, divorce laws and the lack of enforcement of laws against
domestic violence further reduces women’s status by forcing them to stay in abusive
relationships, as there are no viable livelihood outside of marriage (McElroy and Horney
1981; Haddad, Hoddinott, and Alderman 1997; Katz 1997; Hoddinot and Adam 1998;
England 2000). For these reasons, it is important to take into account the broader
institutional context in which women and men operate to understand women’s status.
4. Micro and Macro-Level Processes
According to Goldscheider (1995) demographic analysis operates at two extreme levels
of human behavior—the individual and the national. Before the advent of computers,
scholars who were interested in human population processes had to use printed tables
published by the census bureaus and vital statistics offices of governments for their
analyses, and it was impossible to theorize about the behavior of individuals with these
types of data. However, this limitation was overcome in the 1960s when computers were
utilized to analyze sample surveys. Demographers were able to develop causal models of
individual behavior and to ask why people migrate, marry, and take care of their health.
Family demographers in particular has focused on marriage, divorce, child-bearing and
living arrangements in order to understand both why individuals behave as they do
12
towards each other (micro-level), and why when those individual behaviors are
aggregated into nations (macro-level), societies looks similar or dissimilar.
This dissertation focuses on the micro-level interactions (e.g., negotiation,
decision-making) that take place between household members, especially between
husbands and wives, and how on aggregate (macro-level) these interpersonal interactions
affect certain outcomes. For example, it has been shown that at micro-level, a woman’s
higher level of education positively affects her decision-making regarding child nutrition
(because she is able to negotiate with her partner, read and process information, which
prevent childhood diseases and malnourishment), then we should also expect a percent
increase in female education at population level (i.e., macro-level) to positively affect
child nutrition in a country or a larger society. Hence, it can be said that this dissertation
is a macro-level analyses of a micro-level process i.e., household decision-making.
5. Modernization Theory
Modernization theory is perhaps the most influential sociological theory regarding
changes in women’s status over time. Although gender differences are usually not
discussed explicitly in this theory, the underlying assumption is that traditional societies
serve to fulfill larger family interests such as the preservation and enhancement of
family’s social status, and the reproductive and economic control of the younger
generation (Shorter 1977; United Nations 1988). Further, because traditional marriage
systems usually are organized to protect a woman’s sexuality and maximize her
reproductive value, parental and social interests are best served by marrying daughters
13
close to puberty (Malhotra 1997). In the absence of alternative opportunities such as
schooling and employment, marriage may be the only socially legitimate option for an
adult woman (Caldwell 1982; Goode 1963; McDonald 1985).
Modernization theory stresses that with urbanization, economic development and
the accompanying changes in the market-oriented society, the younger generation has
fewer reasons to be obligated to or be dependent on parents. Education and employment
serve as important means for this independence, especially in terms of better, more equal
options for women. Schooling and work not only offer socially legitimate alternatives to
marriage and childbearing, they also are instrumental in motivating young men and
women to emulate a modern conceptualization of marriage in terms of self-selection of
spouses and more nuclear, conjugal and egalitarian marital relationships (Domingo and
King 1992; Goode 1963; Hirschman 1985; Smith 1980; Thornton et al. 1984; United
Nations 1988). Therefore in addition to the direct impact of schooling and work on
women’s status, we should expect the byproducts of economic development such as
exposure to modern concepts on women’s rights and freedom and egalitarian conjugal
relationships to have a positive influence on decision-making power in developing
societies.
Similar changes in social norms and attitudes towards women were witnessed in
western societies during the late nineteenth century. Søland (2000) in her book titled
Becoming Modern documents changes in women’s role during the industrialization
period. The legal and political gains women have made since the late nineteenth century,
not only gave new educational opportunities to women, it also led to feminist and
14
suffrage movement that had brought women out of the home and into the public eye. By
the end of the 1910s, feminists had succeeded in securing women’s suffrage and in
pushing through other kinds of reform legislation that improved their legal status. This
liberated women from conventional ties and obligations. They rejected Victorian
concepts of domesticity and instead got involved in broad range of public activities
previously deemed incompatible with proper womanhood. During the 1920s an enormous
preoccupation with issues of female identity and women’s proper role emerged, and the
foundation of a new social and sexual order was established. According to modernization
theory, similar changes in women’s role in society may emerge in less-developed nations
as a result of economic development, market-orientation and urbanization.
Like any sociological theory, modernization theory has come under attack for its
inapplicability in certain contexts. For example, the strict dichotomy between the terms
“traditional” and “modern” may not apply to non-Western societies like Asia where
complex and contradictory practices exist within their marriage and family systems.
Evidence shows that parents’ involvement in their children’s marriages continues despite
a shift in self-selection of spouses in countries as diverse as Indonesia, China and Sri
Lanka, or the continued importance of extended family residence in economically
advanced countries like Taiwan and Japan (Morgan and Hiroshima 1983; Riley 1994;
Thornton and Lin 1994). Another criticism of this theory is the tendency for modern
institutions to also be gender biased. For instance, the two agents of independence—
education and wage employment—are usually not equally accessible to men and women.
15
A good example is the labor markets in the U.S. where women are discriminated against
with regard to wages and job opportunities (Blau and Kahn 1994).
Nevertheless, the modernization theory helps us to understand how social norms
and attitudes towards women change with economic development in an entire nation. For
instance, the increase in women’s education and employment not only affects decision-
making power and mobility among women who possess those characteristics; it also
affects the community that they live in through spillover effects (positive externalities).
What this means is if educated women in a certain community are freely commuting to
their workplace, it becomes the norm for other women in that community (who are not
educated or employed) to freely move and visit health centers when they are sick.
Therefore, such changes in attitudes towards women affect all women in a community
regardless of their socioeconomic characteristics.
6. Summary
The above discussion helps us to understand several concepts: (1) why gender analysis is
an important aspect of economic development, (2) what does “women’s status” mean in a
development context, (3) the impact of micro-level processes on macro-level outcomes
and (4) mechanisms in which urbanization and market-orientation affects attitudes
towards women in a society.
Esther Boserup’s Women in Development thesis started a new era of development
research that focused on the role of women in economic development. It shed light on the
negative outcomes of certain development programs on women and emphasized the
16
importance of using women as agents of change by empowering them with education,
employment and credit. These studies help us to understand why women’s power or
status is important in improving the wellbeing of women and children in less-developed
societies. Women’s status is associated with women’s autonomy, power, empowerment,
authority, valuation, wellbeing and position in society, but its ambiguity makes it difficult
to develop a consensus on its definition. Therefore it is very important to clearly specify
what it means each time it is used. Throughout this dissertation I use the following
definition of women’s status: “Women’s power to exercise their preferences and choices
in the households, communities and nations in which they live.”
In this chapter, I also discussed the relationship between micro-level interactions
such as negotiations between spouses within the household and macro-level outcomes
such as child nutrition levels in a population. Such aggregate impact of micro-level
processes is the foundation of this dissertation. Finally, I summarize the modernization
theory, which is the most influential sociological theory regarding changes in women’s
status over time. It helps us to understand why social norms of a society changes with
urbanization and market-orientation.
I apply these theories and concepts to two predominantly Muslim countries in
Asia – Bangladesh and Indonesia to analyze how and why the determinants of women’s
decision-making power and child stunting has changed over time and to identify the
sources of improvements in women’s decision-making and child health in these less-
developed countries. Before I dive into the empirical analysis, it is imperative to
17
understand the current socioeconomic and demographic contexts of the Bangladesh and
Indonesia, which I will describe in the next chapter.
18
Chapter Two
A Summary of the Socioeconomic and Demographic Contexts of
Bangladesh and Indonesia
1. Introduction
The question of how women’s status changes as countries develop has been given little
attention in the current literature. But, it is important to investigate whether
improvements in social, economic and demographic characteristics of women are
actually making an impact on women’s statuses as suggested in the previous chapter. In
this dissertation, I take advantage of the massive socioeconomic improvements
demonstrated by Bangladeshi women and Indonesian women from the late 1990s to the
late 2000’s to investigate whether these improvements have increased women’s status.
These two countries are remarkably similar in a number of contextual features such as
religion and the organization of the agricultural sector, making cross-country comparison
possible. But the countries differ in two aspects: (1) systems of family organization where
relationships are far more patriarchal in Bangladesh than in Indonesia, and (2) economic
development where Indonesia has proceeded more rapidly than Bangladesh.
2. Bangladesh
With over 140 million people squeezed into an area of 145,000 km
2
(i.e., about the size of
Iowa state in the US), Bangladesh is one of the most densely populated countries in the
world (UNFPA 2011). It is located in the northeastern part of South Asia and is almost
entirely surrounded by India, except for a short southeastern frontier with Myanmar and
19
southern coastline on the Bay of Bengal. The country is divided into 6 divisions and 64
districts. Muslims make up almost 90 percent of the population of Bangladesh and
Hindus account for 9 percent. The national language of Bangladesh is Bangla, which is
spoken and understood by all (National Institute of Population Research and Training
(NIPORT), Mitra and Associates and Macro International 2007).
Until 1971, Bangladesh was part of Pakistan (1947-1971) and before 1947, part of
India. Like its neighbors, Bangladesh has deeply rooted patriarchic ideals and has
relatively low valuation of women when compared with other less-developed countries.
High levels of female infanticide, domestic violence against women and low resource
allocation for girls are indicators of gender discrimination (Chen, Huq and D’Souza
1981). Despite continuous international and domestic efforts to improve Bangladesh’s
economic and demographic prospects, it is still among the world’s least developed
nations.
2.1 The Economy and Demography of Bangladesh
Agriculture is the single largest producing sector of the Bangladeshi economy; it
contributes about 22 percent to the GDP and employs 48 percent of the labor force
(NIPORT et al. 2007; BBS 2008). Rice, wheat, jute, sugar cane, tobacco, oil seeds and
potatoes are the principal crops. The manufacturing sector contributes about 17 percent of
GDP and is becoming increasingly important as a result of foreign investments. In fact,
the ready-made garment sector is the industry, which employs the most women.
20
According to the United Nations Development Program’s Human Poverty Index
(HPI), in 2007, Bangladesh ranked 93
rd
poorest of 108 less-developed countries (the
Human Poverty Index is a multidimensional measure of poverty for less-developed
countries; it takes into account social exclusion, lack of economic opportunities and
deprivations in survival, livelihood, and knowledge - UNDP 2007). However, current
trends show that the poverty level in Bangladesh has fallen over the past decade from 58
percent to 49 percent (UNFPA 2011). And the country has graduated from the status of
“low economic potential” to an “emerging market economy,” defying the gloomy
predictions made by many in the mid-1970s.
According to the Human Development Index (HDI) Bangladesh ranks 146 out of
187 countries and territories (the Human Development Index is a summary measure for
assessing long-term progress in three basic dimensions of human development: a long
and healthy life, access to knowledge and a decent standard of living - UNDP 2011). The
HDI value for 2011 is 0.500—in the low human development category. The Gender
Inequality Index (GII) which reflects gender-based inequalities in three dimensions –
reproductive health, empowerment and economic activity is estimated at 0.550 for
Bangladesh, ranking it 112 out of 146 countries in the 2011 index (UNDP 2011).
Currently, Bangladesh is the 8
th
most populous country in the world. During the
second half of the 20
th
century the population growth of Bangladesh tripled. To slow this
rapid growth, the Bangladeshi government and non-governmental organizations
implemented several family planning programs, and social and health interventions
(NIPORT et al. 2007). These programs were first introduced in Bangladesh (then East
21
Pakistan) in the early 1950s through the voluntary efforts of social and medical workers.
After recognizing the urgency of curtailing population growth, the Bangladeshi
government adopted family planning as a government-sector program in 1965. All
subsequent governments have identified population control as the top priority for
government action. As a result of this strong political commitment, increased
involvement of nongovernmental organizations and strong support of the international aid
community, contraceptive use has steadily become widespread in Bangladesh over the
past 20 years (NIPORT et al. 2007). Between 1994 and 2004, the contraceptive
prevalence rate increased from 45 percent to 58 percent. The total fertility rate (TFR) fell
from 6.3 in 1975 to 3.3 in 1994; the current TFR in Bangladesh is 2.36 births per woman
(UNFPA 2011).
In addition to the declining TFR, other human development indicators such as
infant mortality have decreased from 102 deaths per 1,000 live births in 1990 to 41 deaths
per 1,000 live births in 2009. Similarly, child mortality (under five years) decreased from
148 deaths per 1,000 live births in 1990 to 52 deaths per 1,000 live births in 2009
(UNICEF 2012). Further, Bangladesh has achieved gender parity (i.e., relative equal
access to education of males and females) in primary education and nearly removed the
gender gap in secondary education. Although female labor force participation is still low,
it increased from 23.9 percent in 2000 to 29.2 percent in 2006 (UNFPA 2011). Further,
among women aged 20-49 years who were surveyed, median age at first marriage
increased from 14.2 years in 1996-97 to 15.3 years in 2007 (the median age at marriage is
defined as the age by which 50 percent of all women in particular age group were
22
married). All these improvements were achieved in spite of low per capita yearly income
(i.e., US$599 during the fiscal year July 2007–June 2008). Due to these demographic and
socioeconomic changes, Bangladesh makes an interesting research site to track
improvements in women’s status and child malnutrition.
2.2 Patterns of Family Organization and the Status of Women in Bangladesh
According to Islamic tenets, the responsibility for one’s parents is in the hands of adult
children. In Bangladesh this responsibility largely falls onto sons (Kabir, Szebehely and
Tishelman 2000; Mahmood 1992; Rahman 1999). In fact, sons have been described as
the best risk insurance available to women (Arthur and McNicoll 1978; Cain, Khanam
and Nahar 1979). This difference in the relative importance of sons and daughters is
deeply rooted in the patrilineal family organization, where the joint family (i.e. extended
family) is the predominant system of organization. Marriage of sons and daughters is a
highly planned part of family strategy; it is an arrangement between the couple’s parents
(Arthur and McNicoll 1978). Although daughters do useful household work, they do not
bring in outside wages like sons. Under such circumstances, families tend to marry off
daughters at an early age but hold sons as long as possible. This contributes to the large
age difference between brides and grooms, where brides are much younger than grooms
(Arthur and McNicoll 1978).
After marriage, the woman moves in with the husband, his parents, his brothers
and their wives. Her own parents no longer have any claim on her labor. The practice of
village exogamy usually takes a woman some distance from her own parent’s home when
23
she marries (Frankenberg and Khun 2004). According to Cain (1991), Bangladesh has
both a patriarchal and gerontocratic (i.e. group governed by old people) family system,
where young women who move in with their husband’s family are usually at the bottom
of the family hierarchy with little or no control over household resources. These family
systems provide the greatest scope of structural bias based on age and sex (Skinner
1997). In addition, early marriage of women enables extended families to further
dominate the young bride (Arthur and McNicoll 1978).
The low status of women in Bangladesh is a result of this rigid family system
coupled with strong traditions of purdah. According to the tradition of purdah,
respectable women do not engage in trade or fieldwork or leave the family for other than
traditionally specified visits to relatives (Arthur and McNicoll 1978). This means they
cannot find employment outside their homes. Women work mostly on tasks that can be
done at home, such as processing harvested rice and producing handicrafts (Balk 1997;
Cain, Khanam and Nahar 1979; Zaman 1995). However, men are responsible for selling
these products, thus women have little or no control over the profits gained by these
products (Frankenberg and Kuhn 2004). For these reasons, separation and divorce are
real threats to women, especially if they don’t have sons. Therefore, fulfilling duties to
one’s husband and in-laws take on a special importance, as does producing sons as
security for later years. As a result, women find themselves carrying out roles dictated by
the interests of their husbands, sons, or in-laws (Arthur and McNicoll 1978).
24
3. Indonesia
The Republic of Indonesia is an archipelago consisting of 17,000 islands, which lies
between Asia and Australia. It is bounded by the South China Sea in the north, the Pacific
Ocean in the north and east, and the Indian Ocean in the south and west. There are five
major islands: Sumatera, Java, Kalimantan, Sulawesi and Papua. Due to the large number
of islands and their dispersion over a wide area, the country has a diverse culture and
hundreds of ethnic groups, each with its own language (Badan Pusat Statistick (BPS) and
Macro International 2008). Because of this diversity, there are a variety of traditions
known as adat with respect to the organization of family and community life. These adats
define the customary laws regarding marriage, women’s rights, divorce, inheritance and
residence. About 87 percent of Indonesia’s population is Muslim and about 10 percent is
Christian (Protestant or Catholic). However, unlike Bangladesh the kinship systems and
social structure generally support a relatively high social and economic status for women
(Heaton, Cammack and Young 2001; Jones 1994).
During the colonial period, the Dutch dominated Indonesia. In 1602, they
established the Dutch East India Company and became the dominant European power in
the islands. Since proclaiming independence from the Dutch in 1945, Indonesia
experienced several political shifts and periods of instability. In 1966, President Soeharto
began a new era with the establishment of the New Order Government, which was
oriented toward overall development. Soeharto’s government made substantial progress,
particularly in stabilizing political and economic conditions (BPS and Macro
International 2008). But, in 1998, after the Asian economy collapsed, Indonesia went
25
through its worst economic crisis; the economic growth rate dropped to negative 13
percent (BPS 2003). After the year 2000, however the economy recovered with a growth
rate of 5 percent in 2000 and 6 percent in 2007.
3.1 The Economy and Demography of Indonesia
In 2010, the manufacturing sector of Indonesia contributed 24.8 percent to the country’s
GDP. Garments, footwear, electronic goods, furniture, paper products and automobiles
are the principal goods produced by this sector. The agriculture sector contributes the
second largest share of GDP –15.3 percent; the principal crops include rice, timber,
rubber, palm oil and coffee. More than half of Indonesia’s population lives in rural areas,
and proceeds from agriculture average half of rural household incomes, with the rest
coming from rural non-agricultural activities, petty trading and seasonal construction
work. Natural resources (e.g., oil and gas, bauxite, silver, tin, copper, gold, coal) also play
an important role in the Indonesian economy contributing 11.2 percent to the GDP. In
1996, the per capita income of Indonesia was approximately US$ 1,124; by 2007 this has
increased to US$ 3,900 (U.S. Department of State 2012).
By the mid-1990s, Indonesia had enjoyed over three decades of remarkable
social, economic, and demographic change and was on the cusp of joining the middle-
income countries. Increases in educational attainment and decreases in fertility and infant
mortality over the same period reflected impressive investments in infrastructure.
However, during the Asian economic crisis in the 1990s the Indonesian rupiah collapsed
and the gross domestic product contracted. This along with political instability slowed the
26
progress of the country. However, since the mid-2000s the Indonesian economy has
improved. It is now the largest economy of Southeast Asia, and is one of the emerging
market economies of the world. In 2007, economic growth accelerated to 6.3%, and
unlike its export-dependent neighbors, Indonesia has managed to avoid the 2007-08
global recession due to strong domestic demand.
Indonesia’s Human Development Index (HDI) value for 2011 is 0.617. It ranks
124 out of 187 countries and territories. Indonesia’s HDI value is below the average for
countries in East Asia and the Pacific (such as China and Philippines), but it’s above
Bangladesh, which has a HDI value of 0.500. The Gender Inequality Index of Indonesia
is estimated at 0.505, ranking it 100 out of 146 countries in the 2011 index (UNDP 2011).
In comparison Philippines and China are ranked at 75 and 35 respectively on this index,
and Bangladesh is ranked at 112.
With more than 237 million inhabitants, Indonesia is the 4
th
most populous
country in the world, i.e., after China, India and the United States (BPS 2012). In the last
two decades Indonesia’s population growth has slowed down as a result of several family
planning programs implemented by the government. The total fertility rate (TFR) has
decreased by more than 50% in just 35 years— from 5.6 in 1968 to 2.4 births per woman
by 2003. Unmet need for family planning has decreased from 13% in 1991 to 8.6% in
2003. Currently 60.3% of married women use contraception (for most women in
Indonesia first sexual intercourse occurs at the time of marriage).
Since the establishment of the new order government, the country has shown
great improvements in other areas of human development by ensuring the availability of
27
adequate food, clothing and housing, as well as providing adequate education and health
services. Improvements can be seen especially in women’s education. In 1971, school
attendance among children age 7-12 years was 62 percent for boys and 58 percent for
girls; in 2007, the corresponding rates were 93 percent and 98 percent respectively. At all
levels of education (including junior high school and higher education), the increase in
schooling among females has been greater than that of males (CBS 1972; BPS 2008).
This high rate of school enrollment among girls has a direct impact on the age of first
marriage and labor force participation in Indonesia (BPS 2002a; 2008). The median age
at first marriage increased from 18.6 years in 1997 to 19.8 years in 2007 (BPS and Macro
International 2008). Labor force participation among women aged 10 and older increased
from 33 percent in 1971 to 50 percent in 2007. Other human development indicators such
as the infant (under 1) mortality rate declined from 56 deaths per 1,000 live births in 1990
to 30 deaths per 1,000 live births in 2009. Similarly, the child (under 5 years) mortality
rate declined from 86 deaths per 1,000 live births to 39 deaths per 1,000 live births over
the same period (UNICEF 2012).
3.2 Patterns of Family Organization and the Status of Women in Indonesia
According to Islamic tenets and traditional law (adat) in Indonesia, children are obligated
to care for their older parents (Mahmood 1992; Frankenberg, Lillard and Willis 2002).
But unlike in Bangladesh, the obligation is assigned to children of both genders
(Keasberry 2001). Therefore, the difference in the relative importance of sons and
28
daughters does not exist in the Indonesian culture. Further, purdah is not at all practiced
in Indonesia; women have complete freedom of movement.
Indonesian women have traditionally played a prominent role in both the public
and domestic realms. It is not uncommon for couples to work together, not only as part of
an economic survival strategy (Koentjaraningrat 1967), but also as an interactive
decision-making unit (Bangun 1981). Wives are given an equal say in the determination
of important household matters, particularly those involving control over financial
resources (Geertz 1961; Mangkuprawira 1981). Women work in agriculture, sometimes
with their husbands and sometimes on their own, and they engage in trade. Women also
own property separate from their husbands and participate in household management and
decision-making—sometimes to the point of domination. Because women are not
dependent on their husbands economically, divorce is not a major threat to their survival.
If divorced, women could usually rely on their parents to provide them and their children
with a place to live (Heaton, Cammack and Young 2001; Jones 1994).
As explained earlier, Indonesia is extremely diverse in terms of ethnicity. There
are a variety of traditions known as adat with respect to the organization of family and
community life. These adats define the customary laws regarding marriage, divorce,
women’s rights, inheritance, residence, land and property rights. The predominant
kinship system in the region is bilateral, although there are large matrilineal and
patrilineal groups. Through ethnographic literature, researchers have identified some of
the principal features that distinguish the major ethnic groups in the islands of Indonesia,
29
which include Java, Bali, Bugis, Batak and Minangkabau (Frankenberg and Thomas
2003).
In Java, social organization is loosely structured (Schweizer 1988). Descent is
bilateral and nuclear families are the primary unit of social organization. Household
members share resources and work together. Women play a central role in the
organization of the household economy. Javanese society traditionally has incorporated
some major bases of power and independence for women, including economic
participation, property rights and a matrifocal bias in relationships and residence
(Malhotra 1997). In fact, women are considered to be clever, good financial managers
and equal economic partners in marriage (Malhotra 1997; Williams 1989). Although it is
common for some couples, especially those who marry early, to live with either set of
parents (most often with those of the bride) for some time after marriage, most couples
eventually establish a separate residence (Jay 1969; Nag, White and Reet 1980). As a
result, women are able to avoid the constraints imposed by frequent contact with
extended kin that arise under the strictest patrilocal scenario. And the couple’s ability to
make decisions without the direct interference of either set of parents is strengthened.
This results in within-household autonomy for women, and relationship between a
woman and her husband becomes more egalitarian. More than 43 percent of the study
sample belongs to this Javanese ethnic group.
On the islands of Bali, descent is patrilineal and men are viewed as superior to
women. The household is only committed economically to the maintenance of the man’s
ancestral temple; but the woman can maintain strong ties to her natal family. Only sons
30
can inherit their fathers’ estates. About 5 percent of the sample belongs to this ethnic
group.
The Bugis of South Sulawesi maintain bilateral kinship ties and neither bride nor
groom loses ties to their natal family. Bride wealth is an integral part of a Bugis wedding,
as it measures the social status of the bride (and the groom’s status as well, since
marriage tends to be between equals). The children are not considered part of the father’s
kinship line until the bride price has been paid in full. About 3.4 percent of the sample is
Bugis.
The Bataks of North Sumatra are patrilineal. Among the “Toba Batak” the bride
moves in with the husband’s parents after marriage, but she remains part of her birth clan.
She is merely under the “jural control” of the lineage of her husband. Both Toba and
Karo Batak prefer that their sons marry the daughters of their maternal uncles. Kinship
ties are extremely hierarchical and all social interactions occur within this framework
(Kipp 1984). Four percent of the sample belongs to this ethnic group.
The Minangkabaus of West Sumatra are matrilineal, and there is no bride price in
this culture. The husband remains part of his own clan, and children belong to the clan of
the mother. Property accrues to the nuclear family only if both the husband and the wife
participated in acquiring the property. More than 4 percent of the sample is
Minangkabaus (Frankenberg and Thomas 2003).
31
4. Summary
Bangladesh and Indonesia are remarkably similar on a number of contextual features
such as religion and the organization of the agricultural sector. However the countries
differ in two aspects (1) the system of family organization in Bangladesh is strictly
patriarchal and gender biased compared with Indonesia, (2) after 1960s economic
development in Indonesia has proceeded more rapidly than Bangladesh.
Although Bangladesh is till among the world’s least developed nations, it has
shown massive improvements in social indicators such as infant and child mortality,
female school enrollment, contraceptive use, total fertility rate and women’s age at first
marriage. All these improvements were achieved in spite of low per capita income (i.e.,
US$ 599 in 2007). Bangladesh’s family organization is patriarchal. Responsibility for
one’s elderly parents is primarily in the hands of sons, therefore sons are relatively more
important than daughters. The relative low status of women in Bangladesh is a result of
this rigid family system coupled with strong traditions of purdah, which limits woman’s
mobility and autonomy.
In contrast, in Indonesia women have a relatively higher status compared with
Bangladesh. In fact, Indonesian women have traditionally played a prominent role in both
the public and domestic realms. Purdah is not at all practiced in Indonesia, women are
free to move, engage in gainful employment and own property separate from their
husbands. Responsibility for one’s elderly parents falls on to both sons and daughters,
thus daughters are as important as sons. There are a variety of traditions known as adat
with respect to the organization of family and community life, which defines the
32
customary laws regarding marriage, divorce, women’s rights, inheritance, residence, land
and property rights. This makes Indonesia an interesting research site for social scientists.
Before 1960s Indonesia had comparable levels of socioeconomic indicators with
Bangladesh. However, after the 1960s Indonesia progressed rapidly both economically
and socially, this was mainly due to President Soeharto’s new order government, which
ensured the availability of adequate food, clothing, housing, education and health services
for its citizens. In 2007, the per capita income of Indonesia rose to US$ 3,900 and it is
now one of the largest economies in South East Asia.
These two countries have shown rapid improvements in number of socioeconomic
indicators in the past two decades. I take advantage of these changes to analyze how
determinants of women’s decision-making and child nutrition have changed overtime,
and to identify the sources of improvement in decision-making and child nutrition
between 1990s and 2007. I also investigate the role of social context by directly
comparing Bangladesh to Indonesia. There are several methods to accomplish these
goals; in the next chapter I explain the methodologies that I used for my analyses.
33
Chapter Three
Methods
1. Introduction
It is evident from the previous chapter that both Bangladesh and Indonesia has shown
significant improvements in women’s socioeconomic characteristics from late 1990s to
2007. Past research studies have shown that women’s status as measured by
socioeconomic characteristics is strongly associated with women’s decision-making
power in the household and child nutrition and health. Therefore, the first goal of this
dissertation is to identify the determinants of women’s participation in household
decision-making and child stunting, and to investigate whether the relationship between
these determinants and outcomes have changed over time. The second goal is to analyze
the improvements in women’s participation in decision-making and child stunting from
the late 1990s to 2007, and explore to what extent the changes in means (i.e., women’s
socioeconomic characteristics) and changes in coefficients (i.e., the relationship between
determinants and outcomes) account for the improvements in decision-making and child
nutrition. I use multivariate logistic regression to accomplish the first goal and Oaxaca-
Blinder decomposition to accomplish the second goal. In this chapter I explain why these
techniques are appropriate to answer these research questions.
2. Maximum Likelihood Estimation
All outcome variables in this dissertation are binary (0, 1). The method of maximum
likelihood estimation (MLE) selects values of the model parameters that produce a
34
distribution (logistic or probit) that gives the observed data the greatest probability. MLE
methods are preferred for binary response models over OLS because the predicted values
fall within (0, 1), estimates of the marginal effects are consistent, and variance is constant
(homoscedasticity). Further, the logistic coefficients can be converted into odds-ratios,
which have an institutive interpretation that most sociologists like to use. Therefore, I
will use multivariate logistic regression to analyze the change in the relationship between
women’s characteristics and outcomes (i.e., change in coefficients). I will conduct pooled
sample chi-square tests after interacting survey-year dummy variable with all the
variables in the model to investigate whether the logistic coefficients have changed over
time.
3. Oaxaca-Blinder Decomposition
The counterfactual decomposition technique popularized independently by both Ronald
Oaxaca (1973) and Alan Blinder (1973) is widely used to study mean outcome
differences between groups. This method was first employed to study labor market
outcomes by gender and race (e.g. male vs. female or white vs. black). But the technique
can be useful in other fields as well, for example O’Donnell et al. (2008) used it to
analyze health inequalities by poverty status.
In this dissertation, I use this methodology to accomplish my second research
goal, which is to analyze the improvements in women’s participation in decision-making
and child stunting from the late 1990s to 2007, and explore to what extent the changes in
means (i.e., women’s socioeconomic characteristics) and changes in coefficients (i.e., the
35
relationship between determinants and outcomes) account for the improvements in
decision-making and child nutrition. Oaxaca-Blinder decomposition is useful for these
analyses because it decomposes the difference in outcomes of two groups into two
additive elements: one attributed to the existence of differences in observable
characteristics between the two groups and the other attributed to differences in the
association between those characteristics and the outcomes.
3.1 Background: Oaxaca-Blinder Decomposition
Regression analysis helps us to identify the factors (explanatory variables) that explain
the variation in the outcome variable. Such analyses are purely descriptive, revealing the
associations that characterize the inequalities of the outcome. When comparing outcomes
of two groups, decomposition techniques are helpful in explaining the “gap” in the means
of an outcome variable (e.g. men’s wages vs. women’s wages). The gap is decomposed
into three main parts: (1) a part that is due to group differences in the magnitude of the
explanatory variables (i.e., in the example of wage differentials, education or years of
experience), (2) another part due to group difference in the effects of these explanatory
variables (i.e., the association between education and wages), and (3) the last part is the
interaction of the first two parts (1)*(2), which is known as the residual. It is important to
recognize that part (2) captures all potential effects of difference in unobserved variables,
because it subsumes the differences in intercepts.
In the past, sociologists and demographers have used similar decomposition
methods to analyze outcomes of different groups. Demographers have used
36
decomposition methods as early as the 1950s. Kitagawa (1955) proposed a technique to
explain the difference between the total rates of two groups in terms of differences in
their specific rates and differences in their compositions. According to Kitagawa (1955)
similar methods were been used in research under various headings even before the
1950s. For example, in the 1920’s Ogburn determined the part of the difference between
the percent of the U.S. population married in 1890 and the percent married in 1920 that
was due to change in the age composition of the population between these two years. In a
more recent analysis, Preston, Heuveline and Guillot (2001) demonstrated the
decomposition of the difference between the crude death rates in France, 1991, and
Japan, 1992. France’s crude death rate is higher than Japan’s by 0.003116. Their analysis
showed that differences in age composition accounted for 75 percent of the differences in
crude death rates and differences in the rate schedule accounted for the remaining 25
percent. Moreover, in their analysis of the gender poverty gap in eight industrialized
countries, Casper, McLanahan and Garfinkel (1994) employed standardization techniques
to assess whether differences in demographic composition explain differences in the
gender poverty gap. For within-country differences, they substituted the mean values of
demographic characteristics for men into the logistic regression models for women. They
use the same standardization technique to examine how demographic composition
differences explain cross-national differences in the gender poverty gap.
The Oaxaca-Blinder method applies a similar decomposition technique to
regression results, where group differences are accounted for by not only differences in
population characteristics, but also by the association between those characteristics and
37
outcomes. In the next section I describe in detail the formulas used to decompose
regression results. I rely on Owen O’Donnel et al. (2008) chapter on “Explaining
Differences between Groups: Oaxaca Decomposition” in Analyzing Health Equity Using
Household Survey Data, The World Bank Learning Resources Series and Ben Jann’s
(2008) Stata Journal article “A Stata Implementation of the Blinder-Oaxaca
Decomposition.”
3.2 Methods and Formulas: Oaxaca-Blinder Decomposition
Suppose we have a variable, y, which is our outcome variable of interest. We have two
groups, men and women. We assume y is explained by a vector of determinants, x,
according to the following regression model:
y
i
= !
men
x
i
+ "
i
men
…………………….....(Equation 3.1)
y
i
=!
women
x
i
+ "
i
women
where the vectors of ! parameters include intercepts.
In the case of a single regressor, drawn in figure 3.1 the men are assumed to have
a more advantageous regression line than the women. At each value of x
i
, the outcome y
is better for men. In addition, the men are assumed to have a higher mean of x. The result
is that the women have a lower mean value of y than do the men.
38
Figure 3.1: Oaxaca-Blinder Decomposition
In figure 3.1 x
men
and x
women
are vectors of explanatory variables evaluated at the means
for men and women respectively (assuming exogeneity, the conditional expectation of the
error terms in 3.1 are zero). The gap between the mean outcomes, y
men
and y
women
can be
expressed in either of two ways (refer to curly brackets in figure 3.1):
y
men
– y
women
= #x!
women
+ #!x
men
.......................(Equation 3.2)
where #x = x
men
– x
women
and #! = !
men
—!
women
, or as
y
men
– y
women
= #x!
men
+ #!x
women
……................(Equation 3.3)
39
As figure 3.1 makes clear, these two decompositions are equally valid. In the first
equation (Equation 3.2), the differences in the x’s are weighted by the coefficients of the
women and the differences in the coefficients are weighted by the x’s of the men,
whereas in the second equation (Equation 3.3), the differences in the x’s are weighted by
the coefficients of the men and the differences in the coefficients are weighted by the x’s
of the women. Either way, we have a way of partitioning the gap in outcomes between
men and women into a part attributable to the fact that women have worse x’s than men,
and a part attributable to the fact that women have worse!’s than the men.
However, the decomposition shown in figure 3.1 and equations 3.2 and 3.3 are
special cases of the more general decomposition:
y
men
– y
women
= #x!
women
+ #!x
women
+ #x#!……........(Equation 3.4)
= M + C + CM
where the gap in mean outcomes can be thought of as deriving from a gap in mean(M), a
gap in coefficients(C), and a gap arising from the interaction of means and coefficients
(CM). Equations 3.2 and 3.3 are special cases in which;
y
men
– y
women
= #x!
women
+ #!x
men
= M + (CM+C)....................(Equation 3.5)
y
men
– y
women
= #x!
men
+ #!x
women
= (M+CM) + C…..…..........(Equation 3.6)
40
In effect, the first decomposition places the interaction in the coefficient portion, whereas
the second places it in the mean portion. The rationale for this is that, in the past,
decompositions were devised to look at discrimination in the labor market. Thus, in the
first decomposition the presumption was that it’s women who are paid according to their
characteristics, whereas the men receive unduly generous remuneration. In the second
decomposition, the presumption was that men are paid according to their characteristics,
and it’s the women who are discriminated in the work force. Although there were reasons
to assume that discrimination is directed towards one of the groups only, in other
applications there is no specific reason to assume that the coefficients of one or the other
group are non-discriminating.
Several methods have been proposed to overcome this problem. Reimers (1983)
proposes to use the average coefficients over both groups. Cotton (1988) suggested
weighting the differences in the x’s by the mean of the coefficient vectors. Neumark
(1988) made use of the coefficients obtained from the pooled data regression. However,
these methods and assumptions are no longer used in the literature; the interaction term is
simply the residual that is not accounted for by the change in means or the change in
coefficients. It does not have a specific meaning, and therefore most scholars ignore it
completely, this is what I intend to do in my analyses.
Although I use MLE for the multivariate analysis to investigate the change in
coefficients, for the purpose of decomposition, I specify a Linear Probability Model
(LPM). I choose to use LPM even if the dependent variable is binary because a linear
model makes the decomposition results easier to interpret. Moreover, the slopes
41
estimated with linear probability models are usually very close to the marginal effects
routinely estimated for binary dependent variable models such as logit and probit (see
Nielsen 1998; Tarrozi and Mahajan 2007). This approach is acceptable, as long as the
predicted values fall comfortably between 0 and 1, which in my case they are. Further,
heteroskedasticity, which is an inherent problem with the LPM, is resolved by specifying
robust standard errors, which I will do in all regression models. In the next three chapters,
I present the regression and decomposition results that answers my research questions.
42
Chapter Four
Improvements in Women’s Participation in Household Decision-Making
in Bangladesh from 1999 to 2007
1. Introduction
The purpose of this chapter is to examine whether (1) the relationship between women’s
characteristics and participation in household decision-making has changed over time and
(2) to identify the sources of improvement in women’s participation in decision-making
in Bangladesh. I analyze five decision-making spheres: woman’s own health, purchasing
large items, purchasing items for daily needs, visits to friends/family and child
healthcare.
Previous research studies using cross-sectional data and experiments have
examined the association between women’s characteristics and their ability to influence
various household decisions, including decisions regarding contraceptive use, family size,
migration, food eaten at home, child health and education, prenatal and delivery care, etc
(Mason 1987; Williams 1989; Lundberg and Poallk 1993; Shultz 1993, 1999; Balk 1994;
Lundberg, Pollak and Wales 1997; Thomas, Contreras and Frankenberg 1997; Beegle,
Frankenberg and Thomas 2001; Frankenberg and Thomas 2003). This study adds another
dimension to this growing literature by comparing women’s decision making in two time
periods 1999 and 2007.
The above-mentioned studies have shown that, women’s demographic and
socioeconomic characteristics such as having more education, skills, knowledge, and
income tend to increase their decision-making power. Therefore I expect that if womens’
43
demographic and socioeconomic characteristics have improved in Bangladesh from 1999
to 2007, so will their ability to influence household decision-making. In addition, I expect
that the strength of the relationship between determinants and outcomes will weaken
overtime, because when certain attributes are newly introduced to a population (e.g., free
contraception, primary schools for girls, etc.) there will be a strong association between
these new attributes and outcomes. But as time passes, certain characteristics become
ubiquitous (e.g., universal primary education for girls) then the strength of relationship
between these attributes and outcomes weakens, as there will be little variation (or no
variation) in the population with regard to the availability of these attributes. Therefore, it
is important to investigate whether the variables that we used to think that are predictive
of women’s decision-making power are as strong as they used to be. This will ensure that
governments and multilateral organizations will direct their efforts to the development of
relevant attributes of a population, which will optimize their development goals.
As explained in chapter 1 of this dissertation, the focus of this analysis is the
macro-level outcome of the aggregate micro-level interactions between household
members (especially spouses). Therefore, the results should be understood as population-
level changes that have occurred due to improvements in women’s characteristics (e.g.,
percent employed in the female population or percent completed primary school in the
female population).
The data for this study derives from the 1999 and 2007 Bangladesh Demographic
& Health Surveys (BDHS), which are nationally representative cross-sectional surveys.
The BDHSs include detailed information about women aged 15-49 years, their partners
44
and children. Starting in 1999, a decision-making module was included in the BDHS; it
collects information on women’s participation in five household decision-making
spheres. The empirical strategy is to first analyze the determinants of women’s
participation in decision-making in each of these five spheres. The next step is to see
whether the regression coefficient and the means differ systematically between 1999 and
2007. If there are significant differences in the coefficients and means, then the Oaxaca-
Blinder decomposition will indicate whether the change in the outcome variable is due to
changes in means or changes in coefficients.
The rest of this chapter is organized as follows. Section 2 examines theories and
the relevant literature. Section 3 describes the data set, and the five decision-making areas
and their predictors. Section 4 discusses the empirical approach used in this chapter. The
results are presented in Section 5 and conclusion in Section 6.
2. Background
2.1 Gendered Models of Household Decision-Making
In the past, the most common model of the household decision-making assumed that all
household members have identical preferences or that the preferences of one member
determine resource allocation, this is known as the unitary model. However, such models
came under scrutiny because in reality each individual has their own preferences and
resources. As a result, researchers started considering more general models (collective
models), which take individuals as the basic element and treat household decisions as the
outcomes of interactions and bargaining among the members (McElroy and Horney
45
1981; Manser and Brown 1980). Therefore, the collective model has been used to analyze
the differences in preferences and distribution of resources between household members,
especially between men and women.
Scholars who study development have had a long-standing interest in how a
woman's preference relative to that of her spouse affects behaviors and outcomes related
to household welfare. Since differential preferences do not necessarily mean that a
woman will be able to exercise her preferences, the relative power or status of the women
plays a central role in household decisions. In fact, most of the literature has focused
especially on male-female equity in intra-household decision-making power and
allocation of resources, and on the economic and social benefit of educating girls and
women as a form of human capital investment. Women may derive power from multiple
sources, especially through education, employment and assets. Contexts of power such as
customs and norms regarding marriage and family life are also sources of power.
Since both preferences and decision-making are gendered, and are socially
constructed through day-to-day interactions with individuals and social structures, this
dissertation will focus on the gendered nature of household decision-making. Lundberg
and Pollak (1993) have also emphasized the different roles that men and women play in
the household and the implications of these “spheres of interest” for models of household
decision-making. These spheres may include expenditure on daily purchases; expenditure
on children; large purchases/expenses and the decision to use contraceptive. Who
influences each sphere has consequences for the overall welfare of the household. For
instance, as women have different preferences from men, when they control food
46
expenditure, they may buy nutritional food for the children, which improves their growth,
and make them less susceptible to illness (Schultz 1999). Further, when power within a
household changes, the decision-making spheres of each member and their ability to
exercise their preferences may also change. Therefore, it is important to investigate the
changes in household decision-making power and their consequences over time.
2.2 Determinants of Women’s Status
Education – A large literature has documented that increased female education is
associated with increased participation in household decision-making (Frankenberg and
Thomas 2003). Schooling affects decision-making in several ways and one mechanism is
through wage employment. However, at micro-level, sometimes women from higher
social classes have more opportunities to get an education than women from lower social
classes, but due to certain traditions like purdah, women from higher social classes are
not allowed to work outside the home in order maintain the family’s social status. By
contrast, women form poorer backgrounds who don’t have an education do not have a
choice but to find employment for the survival of the household; therefore, women with
no education may be more likely to work than women who have some education. Thus,
in certain contexts education and employment may not be positively correlated.
Nevertheless, it is expected that as education levels are enhanced, women will have
increased agency as well as negotiating powers both at home and in the community.
Further, women who are educated marry at a later age as they spend most of their
adolescence in school, this means they are more mature and have more life experiences
47
when they enter married life, giving them even more power within the household. So if
the female population becomes more educated, women will have more influence on
household decisions.
Age—In a gerontocratic society such as Bangladesh, age confers authority and
status; therefore older women tend to influence decision-making more than younger
women (Kamal and Zunaid 2011). So as the female population ages women will have
more influence on household decisions.
Employment—Wage employment and individual labor income have been used as
indicators of women’s control over resources in several studies (Blumberg 1988;
Bourguignon et al. 1994). In general, when women contribute to the household budget,
they are able to influence household decision-making. Therefore woman’s individual
earning potential is an important source of power for women, which influences household
decision-making. Thus, as women’s participation in the labor force increases in a
country, women’s decision-making power in the household also increases.
Access to Microcredit—Microcredit programs such as Grameen Banks were one
of the earliest development concepts that allowed rural communities to obtain group-
based credit to invest in small-scale, self-employment activities. Because the bank
recognizes women’s centrality to the development process—as beneficiaries of it and as
active agents in promoting it—it has increasingly focused on providing credit to women.
Therefore, a significant majority of its borrowers are women. The Bank primarily focuses
on improving women’s economic status, which it views as the foundation on which better
social and political status can be built. In addition to providing credit, the Bank has
48
promoted specific social developmental goals known as the Sixteen Decisions. It is a list
of goals for borrowers and their families to aspire to and work towards. The goals include
commitment to better sanitation, education, family planning practices, housing, and better
nutrition. By participating in microcredit programs, women also have access to a larger
social network and are able to exchange ideas and information. In a recent evaluation of
the microcredit program, researchers found that there are significant improvements in
maternal care, nutritional status, sanitation and consumption of clean drinking water in
microcredit communities (Bernasek 2003). Thus, there are two mechanisms in which
Grameen Banks increase women’s decision-making power: (1) by giving them credit for
self-employment activities and (2) by providing them with a social network (i.e., social
capital) where information and ideas are exchanged. Therefore, I expect that if more
women in the population have membership in these programs, then more women will
participate in decision-making.
Household Headship—In a setting such as Bangladesh, it is rare for a woman to
assume household headship unless her partner is absent from the household. Using data
from Kenya and Malawi, Kennedy & Peters (2002) show how gender of the household
head affects child wellbeing. They disaggregated the households by male- and female-
headed households. The female-headed households were further disaggregated into de
jure (legal head of household is a woman) and de facto (male head of the household is
absent more than 50% of the time). In both Kenya and Malawi, the de facto household
had the lowest income. Despite this low income, preschoolers’ nutritional status was
significantly better than in the higher income male-headed. The authors state that the
49
ability to improve nutritional status in a low-income environment in de facto households
is related to a combination of child feeding practices and other nurturing behavior. Their
findings substantiate the claim that women allocate more resources to their children when
they have the decision making power at home. Therefore, if there are more female
household-heads in a population, then I expect to see more women participating in
household decision-making.
2.3 Women’s Decision-Making Literature
The women’s status literature can be divided into two areas; one area, that looks at the
association between women’s characteristics and welfare outcomes such as fertility and
infant mortality, and a more recent area that examines the association between women’s
characteristics and household decision-making. Although it was useful to see how
varying characteristics of women can influence welfare outcomes, it was not clear who in
the household made decisions that influence these outcomes. It was assumed that, if
women have higher power in the household, she may be able to influence decision-
making in certain spheres (Lundberg and Pollak 1993), but without data on decision-
making, it was difficult to make this claim. For instance, say it has been argued that
women who have more control over resources (power), allocate a higher share of the
household budget to buy nutritional food and healthcare; as a result children are less
stunted (low height-for-age). Then, it should be the case that women have more say in
decisions regarding budget allocation (to food and healthcare) in those households in
which they have more power.
50
In an effort to pry open this “black-box” of household decision-making,
researchers began to directly ask survey respondents to describe who they perceived to be
the primary decision-maker for a series of household decision-making spheres. Most
household surveys including the Demographic and Health Surveys now use explicit
questions about decision-making within the household and use them as outcomes of
relative power within the household; these indicators can shed light on how power
manifests itself in everyday life.
Deborah Balk (1994) analyzed one of the first small scale surveys on women’s
status in two sub-districts of rural Bangladesh – Abhoynagar and Sirajgong. The survey
was conducted in 1988 and was designed explicitly to measure women’s status (the data
was collected by the MCH-FP Extension Project of the International Center for Diarrheal
Disease Research). Out of 25 survey questions on women’s status, Balk constructed four
indices
2
(scales), they are: mobility, authority, attitudes, and the leniency of her
household towards her. For the most part, these indices attempted to measure the
propensity of women to do things that are typically reserved for men (e.g. work outside
the home). Men were not surveyed, nor were mothers-in-law or other representatives of
the patriarchal authority (Balk 1994: 23).
Mobility questionnaire included questions like “When you go outside the village,
who usually accompanies you?” authority questionnaire included questions like “Who
makes the decision on how long a child should attend school?” attitudes questionnaire
2
Balk (1994) used Cronbach’s !-coefficient to measure the “reliability”, or internal cohesion, of these
indices. While each index varies by the number of questions it includes in its construction, the index scales
were intended to equal one another intrinsically. For this reason, the range was standardized to span a low
of zero and a high around three. Balk also showed that these indexes are conceptually and statistically
distinct.
51
included questions like “How do you feel about women in this society traveling on their
own?” leniency questionnaire included questions like “Are you permitted to earn
money?”
Ordinary Least Square (OLS) regressions were used to identify the determinants
of these four women’s status indices. The multivariate analyses reveal that older women
are more likely than younger women to have higher authority. The authority of women
who live in the homes of their in-laws is lower than that of women who live in an
independent household or in their natal household. Education, the variable most
commonly used as a proxy for women’s status is positively and very significantly
associated with the authority index. The status of Muslim women is lower than that of
Hindu women in all indices; and the status of residents in the more traditional and
conservative Sirajgong region in central Bangladesh is lower than that of women in the
more religiously heterogeneous, liberal southwestern region, Abhoynagar.
The size of the family dwelling unit produces negative effects on authority, while
landlessness produces a positive effect. This is an important finding, as it implies that the
status of poorer women (i.e. women from poorer households) is considerably higher than
that of wealthier women. The author argues that poorer women are more autonomous,
come from more lenient homes (where purdah is not maintained), and are more likely to
take part in household decisions because their relative worth (i.e. relative to the men in
the household) is considerably greater, as they have to work for pay for the survival of
the household.
52
This study was one of the first to analyze decision-making as a function of
demographic, socioeconomic and socio-cultural variables. However, there are some
limitations in this study. For instance, the bundling of decision-making variables to create
a single decision-making disregards Lundberg and Pollak (1993) notion of separate
spheres. Further, responses to the decision-making questions were converted into a 0-3
linear scale to perform OLS, but responses such as: “primarily the wife”, “primarily the
husband” or “other family members” does not have an additive meaning. Overall,
however, this paper is a good starting point for scholars who are interested in revealing
the “black-box” of household decision-making.
2.4 Evidence from Other Countries
Evidence from other less-developed countries including Pakistan shows that women’s
age and family structure are the strongest determinants of women’s authority in decision-
making (Sathar and Kazi 2000). Jejeebhoy (2000) shows that the socio-cultural context in
rural India conditions the relationship of women’s individual-level characteristics to
decision-making, and that autonomy is the key-mediating factor between women’s status
and welfare outcomes such as reproduction. In a study of rural China, Jin (1995) found
that women who have significantly more influence on reproductive matters tend to be
more educated, spend more time on household economic activities and marry later. In a
study in Nigeria, Kritz and Adebusoye (1999) show that ethnicity plays a very important
role in shaping a wife’s decision-making authority and is even more important than other
individual-level characteristics. Becker, Fonseca-Becker and Schenck-Yglesias (2006)
53
compared husbands’ and wives’ reports of women’s decision-making power in Western
Guatemala and found that, wives tend to under-report their household decision-making
power. But in couples with both partners educated and in couples in which women work
for pay, both husband and wife were significantly more likely to report that they jointly
make final decisions about household activities. Frankenberg and Thomas (2003)
analyzed the response of both male and female reports on five
3
decision-making spheres
in Indonesia, and found that 25 percent of couples did not report the same decision-
maker. The authors also found that ethnicity
4
is a powerful predictor of decision-making.
Additionally, they found that higher the woman’s education; less likely is for her husband
to make decisions about food expenditures, relative to decision being joint. Higher levels
of education among women also increase the probability that she makes the decisions
alone relative to the husband deciding them alone.
Previous literature has thoroughly analyzed the determinants of decision-making
in less-developed countries. However, comparative studies between different contexts or
time periods are rare. Such comparative studies will allow us to understand how nations
are developing, and whether attributes that used to predict women’s decision-making
power has the same effect they used to have. Such analyses also help us to investigate
whether empowering women through education and employment is actually affecting
decision-making in the household. Therefore, this study extends the current literature by
looking at the changes in decision-making in a rapidly changing society.
3
The five decision-making questions ask the respondents about who makes decisions regarding: (1)
Expenditure on food eaten at home; (2) Child education; (3) Child health; (4) Purchasing durables and (5)
Contraceptive use.
4
See chapter 2 for a detailed discussion on Indonesia’s ethnicities and family systems.
54
3. Data
This study is based on the 1999 and 2007 Bangladesh Demographic and Health Surveys
(BDHS), which are nationally representative cross-sectional data sets. Since 1984, the
MEASURE DHS project has provided technical assistance to more than 260 DHSs in
over 90 countries to advance the global understanding of health and population trends in
developing nations. DHS has earned a worldwide reputation for collecting and
disseminating accurate, nationally representative data on fertility, family planning,
maternal and child health, gender inequality, HIV/AIDS, malaria, nutrition and domestic
violence. The 1999 and 2007 BDHSs were implemented through a collaborative effort of
the National Institute of Population Research and Training (NIPORT), Macro
International and Mitra & Associates. The financial supports for the surveys were
provided by the United States Agency for International Development (USAID).
The BDHSs
5
are based on a two-stage sample design. In the first stage,
enumeration areas or “clusters” were selected from the larger regional units within the
country. Next, households were randomly selected within clusters with probability
proportional to PSU size. Some enumeration areas were oversampled to accommodate
small divisions (Barisal and Sylhet) and some urban areas due to the small proportion of
the population living in them. Fieldwork for 1999 BDHS was carried out from November
!
""The BDHS covers the entire population residing in private dwelling units in Bangladesh. The country is
divided into six administrative divisions: Barisal, Chittagong, Dhaka, Khulna, Rajashani and Sylthet. Each
division is divided into zilas and each zila into upazilas. Rural areas in an upazila are divided into union
parishads (UPs), and UPs are further divided into mouzas. Urban areas in an upazila are divided into
wards, and wards are subdivided into mahallas. These subdivisions are used as the Primary Sampling Units
(PSUs) for the surveys and each enumeration area consists of about 100 households, on average, and is
equivalent to a mauza in rural areas and to a mohallah in urban areas.
55
10
th
, 1999 to March 15
th
, 2000; and fieldwork for 2007 BDHS was carried out from
March 24
th
to August 11
th
, 2007. The 1999 BDHS interviewed 9,854 households and
2007 BDHS interviewed 10,400 households. Detailed information on ever-married
women, their male partners (if they had one), and their children under age five was
gathered. My analytical sample comprises ever-married women between ages 15-49
years, resulting in sample sizes of 10,373 in 1999 and 10,996 in 2007. Using the
women’s detailed questionnaire and sample weights I can make inferences about macro
level relationships between women’s evolving socioeconomic status and their decision-
making ability in the household.
3.1 Decision Making
This section discusses the dependent variables of interest that are used to measure
women’s influence on household decision-making. The BDHSs collect information about
women’s participation in household decision-making in five areas: woman’s own health
care; child health care; large household purchases; daily household purchases and visits
to family and relatives (see Figure 4.1 and Figure 4.2).
56
Figure 4.1: Excerpts from the 1999 BDHS Women’s Questionnaire
Figure 4.2: Excerpts from the 2007 BDHS Women’s Questionnaire
If the woman indicates that she makes these decisions alone, jointly with her husband or
jointly with someone else; then the dependent variable is coded as one. And if the woman
indicates that her husband makes these decisions alone, or someone else in the household
is making decisions without consulting her, then the dependent variable is coded as zero.
There are two reasons for collapsing the five categories into two binary categories. First,
the interpretation of a five-category multinomial logistic regression coefficient is
cumbersome and provides little insights as to how women’s decision-making has
changed overtime. The second reason is due to the distribution of the responses in the two
RESPONDENT=1, HUSBAND=2,
Who in your family usually has the final say on the following decisions. RESPONDENT & HUSBAND JOINTLY=3,
SOMEONE ELSE=4, RESPONDENT &
SOMEONE ELSE JOINTLY=5
Your own health care? 1 2 3 4 5
Child health care? 1 2 3 4 5
Making large household purchases? 1 2 3 4 5
Making household purchases for daily needs? 1 2 3 4 5
Visits to family, friend, or relatives? 1 2 3 4 5
What food should be cooked each day? 1 2 3 4 5
Who usually makes decisions about health care for yourself: RESPONDENT = 1
you, your husband, you and your husband jointly, or someone HUSBAND = 2
else? RESPONDENT & HUSBAND JOINTLY = 3
SOMEONE ELSE = 4
RESPODENTLY AND SOMEONE ELSE JOINTLY = 5
1 2 3 4 5
Who usually makes decisions about making major
household purchases? 1 2 3 4 5
Who usually makes decisions about making purchases
for daily household needs? 1 2 3 4 5
Who usually makes decisions about visits to your family
or relatives? 1 2 3 4 5
Who usually makes decisions about your child health care? 1 2 3 4 5
57
data sets. In 1999 more women said they make decisions alone than in 2007, while more
women said they jointly make decisions with their spouses or other household members
in 2007 than in 1999. Therefore, if we only consider the percentage of women who made
decisions alone, then we would see a reduction from 1999 to 2007. By collapsing the
three categories into one, i.e., decisions made (1) alone, (2) jointly with her husband or
(3) jointly with someone else, we can get a more accurate measure woman’s participation
in decision-making.
Figure 4.3 illustrates the weighted percentages of women in 1999 and 2007 who
have claimed that they participate in decision making about their own health, large
purchases, daily purchases, visits to family/friends and child care. In addition, the chart
depicts the combined decision-making scale,
6
which adds each of the five variables into
one, yielding a scale of 0-5. Among the five decision-making domains, child health
shows the highest percentage increase (13 percentage points) between 1999 and 2007,
this is encouraging because women tend to invest more resources on children’s health by
giving them nutritional food and take them to a doctor when they are sick. Women’s
participation in decisions regarding her health and daily household purchases has
increased by 8 percentage-points. Again this is a good indication, because women are
taking more control of their own health and day-to-day purchases for the household such
as food. In a society like Bangladesh where purdah system doesn’t allow women to
freely move outside their homes, women have limited freedom to visit their families and
6
Factor analysis was conducted to measure the internal cohesion of the five decision-making variables. The
results indicated that all five indicators load into one factor with an Eigen value of 2.55. Factor loadings
were greater than 0.65 for all five indicators. Therefore, adding the variables to create a combined decision-
making measure is warranted.
58
friends, their husbands and the community they live in usually implement this rule. But
from 1999 to 2007, women have shown significantly more involvement in decision
regarding her visits to friends and family, which has increased from 62 to 68 percent.
When women are able to freely move in their village or town, they can pursue an
education, engage in gainful employment and take their children for vaccination and
doctor’s visits. Therefore, this is an encouraging finding. The final decision-making
sphere relates to large expensive household purchases such as television, refrigerator,
radio, bicycle, etc. Since these items require a substantial financial investment and since
women usually don’t contribute a significant amount (if any) to the household income,
they have little influence on decisions regarding large purchases. However, even this
decision-making sphere has shown significant improvements from 61 to 66 percent. This
maybe an indication of women’s increased participation in gainful employment, which
allows them to the household income. The combined decision-making scale, which is a
summary measure of the five binary variables show a 0.46 increase (i.e., about 15%
increase) from 1999 to 2007. All outcome variables including the combined scale show
statistically significant increases between the two survey years, indicating that there have
been considerable changes in women’s status over time.
59
78%
68%
71%
66%
64%
65%
62%
63%
61%
56%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90%
Child Health
Visits
Daily Purchases
Large Purchases
Own Health
Figure 4.3: Weighted Percentage of Women who Participate in
Decision Making in Bangladesh
1999 2007
#$!#"
#$%&"
0 1 2 3 4 5
2007
1999
Notes: The weighted percentage and mean differences between 1999
and 2007 are all statistically signficant at 0.05 level.
Source: Author's calculations.
Combined Decision Making Scale (0-5)
60
3.2 Explanatory Variables
The main objective of this study is (1) to explore whether the determinants of women’s
decision-making power and the strength of their relationships has changed over time, and
(2) to investigate the sources of improvement in women’s decision-making from 1999 to
2007 in Bangladesh. I use five characteristics of women that are expected to influence
decision-making, they are: age, education, employment, household headship and
membership in microcredit programs. I include woman’s age in three categories 15-24
years; 25-34 years; and 35-49 years to analyze how women’s status changes over the life
course. I use age categories instead of a continuous age variable because there are only
subtle differences between one-year age gaps (e.g., 23 years vs. 24 years), thus the size of
the coefficient is small, which means the strength of the relationship is weak. Since one
of my goals is to measure the change in coefficient from 1999 to 2007, I decided to use
three age categories, which provides large coefficients for further analyses. Woman’s
education is divided into four categories according to Bangladesh’s schooling system: no
education, primary education, secondary education and higher education. Employment
status and household headship are indicated by dummy variables. Membership in a
Grameen Bank, Bangladesh Rural Advancement Committee (BRAC), Bangladesh Rural
Development Board (BRDB), Mother’s Club or any other organization is indicated with a
dummy variable.
Since women’s status is considered to be relative to men’s rather than absolute or
relative to other women’s, it is imperative to use partner characteristics as controls. I
considered using partner’s age, age difference between spouses, education and presence
61
in the household. However, since woman’s age and partner’s age,
7
and woman’s
household headship and partner’s presence in the household are strongly correlated ($ >
±0.5), I only use partner’s education in the multivariate models. Controlling for partner’s
education, the woman’s education can be interpreted as relative education (or a measure
of relative power). Since most men are better educated than their wives, holding his
education constant, higher-levels of education among women implies a reduction in the
gap (Frankenberg and Thomas 2001).
Finally, four variables capture the background characteristics of the household
and societal context, which include a household asset variable, rural/urban residence, the
four most common religions in Bangladesh (Islam, Hinduism, Buddhism and
Christianity) and the six administrative divisions (Barisal, Chittagong, Dhaka, Khulna,
Rajashani and Sylthet). Women who live in rural areas have fewer opportunities to
pursue an education due to lack of schools and other facilities. Further, certain gender-
biased attitudes are more prevalent in rural areas than in urban areas, as rural residence
have little exposure (e.g., through newspapers, radio and television) to modern views
compared with urban residents. Therefore, it is important to control for residence when
predicting women’s decision-making power in the household. Similarly, certain religions
promote gender-biased views through their teachings, for example Buddhism and
Christianity are more tolerant towards women’s autonomy, while Islam and Hinduism
advocate restrictive practices such as purdah. The main religion practiced in a woman’s
7
I also considered using the difference between partners’ age, but this was not significant in the
multivariate
62
household therefore, may dictate to what extend she can influence household decision-
making.
Bangladesh’s six divisions capture variation in industrial and infrastructural
growth, as well as social development. Most of these developments are centered around
Dhaka (Kamal and Zunaid 2011), which means according to modernization theory, social
norms and attitudes towards women in Dhaka should be more progressive than the rest of
the country. Therefore we should see stronger relationship between women who live in
Dhaka and their participation in household decision-making. In contrast, Barisal division,
which is closest to the Bay of Bengal frequently suffers the consequences of environment
disasters; therefore economic development is much slower in this division compared with
Dhaka, therefore we should see fewer opportunities for women’s education and
employment, and stronger gender-biased attitudes towards women in this region. The
Sylhet division has the lowest Human Development Index (HDI) in the nation. In terms
of health indicators it has the highest child mortality (under 5 years) and fertility rates,
and lowest rates of immunization. Further, school enrollment rates at primary and
secondary levels are much lower than the national averages (Nath et al. 2012). These
factors affect women’s status in the community and their decision-making power in the
household.
As part of the 1999 BDHS, a Service Provision Assessment (SPA) survey
collected information on the infrastructure and social development indicators of all
BDHS communities, as well as information on the accessibility and availability of health
and family planning services. However, the SPA survey was not conducted in 2007. In
63
the absence of such community-level data, division variable provide us useful
information about the community infrastructure and social development indicators that
may affect women’s decision-making outcomes. Therefore it is important to control for
division in the multivariate analyses.
Household wealth and assets are important determinants of women’s decision-
making power. But unfortunately BDHSs do not include household income or
consumption data.
8
However, the surveys include detailed information on household
assets such as chair/table, almirah (i.e. cupboard), radio, television, telephone, bicycle
and motor bicycle. I created a household asset variable using household items as a proxy
for wealth (Rutstein 1999). Principle component factor analysis indicated that out of
seven common items in Bangladeshi households, the following items: almirah, television
and chair/table had a loading of 0.6 or greater on one factor with an eigen value of 2.33.
Therefore I used these variables to create a continuous scale indicating the wealth of the
household.
4. Methods and Analyses
Analysis is conducted using STATA version 10.0. Sample weights are used in order to
adjust for the sample design; this ensures that the results are representative at a national
level. The association between explanatory variables and the five outcomes of women’s
decision-making are explored using bivariate regression. Predictors found to be
8
The recoded wealth index provided in the BDHS data sets are not comparable across surveys (Rutstein
2008)
64
significantly associated (p<0.05) with the outcomes are then used for the multivariate
logistic regression (see Equation 4.1) for the two survey years separately.
u C WC y + + + =
2 1 0
! ! !
…….. Equation 4.1
where y denotes a binary variable which equals to one if the respondent participates in
decision-making, else zero, WC—are women’s characteristics including age, education,
work status, household headship and membership in microcredit programs; and C—are
control variables including partner’s education, household asset scale, rural/urban
residence, division and religion.
Since this study compares two groups—one group that lived in an era where
women had less education, fewer employment opportunities and lower access to credit
and social capital (i.e. 1999 survey sample); and another group that has enjoyed the fruits
of economic development including gender parity in education, establishment of garment
factories with female employees, more self-employment opportunities through
microfinance, and fewer children to care for, it is important to identify the sources of
improvement in decision-making overtime. For this purpose I use the Oaxaca-Blinder
decomposition technique (Oaxaca 1973; Blinder 1973) applied to the analysis of the
probability of participating in decision-making, which I estimate with linear probability
models (see chapter 3 for a detailed discussion).
Let D
i
denote a dummy variable equal to one; if the respondent participates in i =
decisions regarding her own health; large purchases; daily purchases; visits to
65
family/friends; and child health care. Then the model for a given dependent variable
according to Equation 3.4 in chapter 3 can be written as:
y
2007
– y
1999
= #x!
1999
+ #!x
1999
+ #x#!…...........Equation 3.4
= M + C + CM
D
i2007
- D
i1999
= [X
i2007
-X
i1999
]
B
i1999
+ [B
i2007
- B
i1999
]
X
i1999
+
[(X
i2007
-X
i1999
)B
i1999
*(B
i2007
-B
i1999
)X
i1999
)]…... Equation 4.2
The first component [X
i2007
-X
i1999
]B
i1999
in Equation 4.2 can be interpreted as the
contribution of the differences in mean of the predictors between 1999 and 2007. The
second component [B
i2007
- B
i1999
]X
i1999
measures the contribution of difference in the
coefficients (including differences in the intercept).
9
The third component is an
interaction term [(X
i2007
-X
i1999
)B
i1999
*(B
i2007
-B
i1999
)X
i1999
)], which is the residual effect.
In the next section, I report the results of the descriptive analysis and logistic
regressions (see Tables 4.1 through 4.8). Decomposition results are summarized in Table
4.9. Note that, some explanatory variables (e.g. household assets, woman’s work) maybe
jointly determined by education level of the woman and her husband. Therefore one
should be very cautious in interpreting the regression results in a causal way. Most of the
included regressors are, in fact, likely to be correlated with unobserved heterogeneity in
preferences, cultural norms, or other location specific characteristics that may also have a
9
The intercept subsumes the effects of group differences in unobserved explanatory variables (Jann 2008).
66
direct impact on the dependent variables. For these reasons, the interest of these results
likely lie more in their descriptive content that in their causal meaning.
5. Results
5.1 Descriptive Statistics
Table 4.1 presents the descriptive statistics of the explanatory variables. The last column
shows the results of the two-sample tests, which indicates whether the variable has
changed significantly over time. The proportion of women without any education
decreased from 46 percent to 34 percent from 1999 to 2007, while secondary education
increased from 21 percent to 30 percent. Further, the proportion of women who work
increased from 23 percent to 32 percent, and membership in microcredit organizations
such as the grameen bank increased from 11 percent to 31 percent. These are the
improvements in women’s socioeconomic characteristics that I expect to be associated
with the increase in women’s participation in decision-making in Bangladesh. In the
second half of table 4.1, changes in control variables are presented. In contrast to
women’s education, men’s education has shown modest improvements. This is possibly
due to heightened interest in improving girls’ school enrollment rather than boys’ school
enrollment, and also because boys school enrollment were at higher levels in 1999 to
being with. On a scale of 0-3, household assets have increased from 1.14 to 1.57. This
shows that more respondents are able to purchase household items in 2007 compared
with 1999, indicating an increase in wealth. The percentage women who live in rural
household have declined by 3 percentage points; this is expected as Bangladesh is more
67
Table 4.1: Summary Statistics of the Explanatory Variables, Bangladesh in 1999 and 2007
Table 4.1: Summary Statistics of the Explanatory Variables, Bangladesh in 1999 and 2007
1999 2007 Two Sample Test
N=10,140 N=10,955 sig
Woman's Characteristics
Woman's age in categories
15-24 years 0.33 0.33
25-34 years 0.35 0.33 ***
35-49 years 0.32 0.35 ***
Woman's education level
No Education 0.46 0.34 ***
Primary Education 0.28 0.30 **
Secondary Education 0.21 0.30 ***
Higher Education 0.04 0.06 ***
Woman works 0.23 0.32 ***
Woman is the household head 0.05 0.08 ***
Woman is a member of a grameen bank
a
0.11 0.31 ***
Controls
Partner's Education
No Education 0.40 0.36 ***
Primary Education 0.23 0.26 ***
Secondary Education 0.24 0.26 **
Higher Education 0.12 0.12
Household Assets (Scale 0-3) 1.14 1.57 ***
Rural Residence 0.80 0.77 ***
Religion
Islam 0.88 0.91 ***
Hinduism 0.11 0.08 ***
Buddhism 0.009 0.006 ***
Christianity 0.003 0.002
Division
Barisal 0.07 0.06 ***
Chittagong 0.19 0.18
Dhaka 0.31 0.31 ***
Khulna 0.12 0.13 **
Rajashani 0.26 0.25 +
Sylhet 0.06 0.06 ***
Source: BDHS
a
This variable includes membership in grameen banks, BRAC, BRDB, mother's club and other
micro credit organizations
Notes: All means are weighted according to the survey design;
***p<0.001, **p<0.01, *p<0.05, +p<0.10
Explanatory Variables
68
industrialized and market-oriented in 2007 than it was in 1999, thus more rural residents
who used to work in agriculture migrate to the city in search of employment, creating an
influx of urban residents. The distribution of religion in Bangladesh has also significantly
changed from 1999 to 2007, where percentage of Muslim women has increased by 3
percentage points. This may be a result of intermarriage between religions, where
conversion to Islam is a strict requirement for a non-Muslim to marry a Muslim.
5.2 Regression Analysis
Table 4.2 presents the results of the pooled bivariate regressions of the five binary
outcomes and the combined decision-making scale. The preliminary observations that
emerge from this table indicate that women who are older tend to participate more in
decision-making than women who are younger, and women with higher education are
more likely to participate in decision-making than women with no education. Similarly
employed women and women who have membership in a microcredit organization are
more likely to participate in decision-making than women who do not work and who do
not belong to any microcredit organization. Women who are household heads are more
likely to participate in decision-making than women who are not household heads. All the
coefficients are in the direction that I expected, and are consistent with theory and past
empirical studies.
The control variables indicate that if the partner has higher education, the woman
is more likely to participate in decision-making than partners with no education. Women
who live in households with more assets participate in decision-making more than
69
Table 4.2: Pooled Sample Bivariate Regression: Women’s Participation in Decision-Making
Predicted by Women’s Characteristics and Control Variables
Woman's Characteristics
Woman's age in categories
15-24 years (Reference)
25-34 years
0.56 *** 0.41 *** 0.51 *** 0.60 *** 0.53 *** 0.44 ***
35-49 years
0.67 *** 0.50 *** 0.58 *** 0.72 *** 0.65 *** 0.58 ***
Woman's education level
No Education (Reference)
Primary Education
0.04 -0.03 -0.02 0.07 + 0.02 0.08 *
Secondary Education
0.10 ** 0.04 -0.02 -0.06 0.05 0.23 ***
Higher Education
0.70 *** 0.67 *** 0.51 *** 0.40 *** 0.60 *** 0.95 ***
Woman works
0.52 *** 0.38 *** 0.54 *** 0.60 *** 0.43 *** 0.58 ***
Woman is the household head
1.49 *** 2.15 *** 1.68 *** 2.44 *** 2.13 *** 2.49 ***
0.31 *** 0.16 *** 0.32 *** 0.38 *** 0.27 *** 0.38 ***
Controls
Partner's Education
No Education (Reference)
Primary Education
-0.01 0.01 -0.06 -0.07 -0.03 0.05
Secondary Education
0.07 + 0.06 -0.07 + -0.04 0.09 * 0.15 **
Higher Education
0.45 *** 0.39 *** 0.30 *** 0.26 *** 0.46 *** 0.56 ***
Household Assets (Scale 0-3)
0.16 *** 0.14 *** 0.10 *** 0.08 *** 0.13 *** 0.21 ***
Rural Residence
-0.33 *** -0.30 *** -0.28 *** -0.21 *** -0.32 *** -0.35 ***
Religion
Islam (Reference)
Hinduism
-0.04 -0.01 -0.11 * -0.05 0.04 -0.05
Buddhism
0.20 -0.06 -0.04 0.44 + 0.23 0.63 *
Christianity
0.82 *** 0.75 * 0.68 + 0.70 * 1.12 ** 2.21 ***
Division
Barisal (Reference)
Chittagong
0.21 *** 0.21 *** 0.22 *** 0.22 *** 0.13 * 0.07
Dhaka
0.33 *** 0.14 * 0.43 *** 0.45 *** 0.21 *** 0.20 **
Khulna
0.35 *** 0.15 * 0.50 *** 0.40 *** 0.22 *** 0.15 *
Rajashani
0.45 *** 0.24 *** 0.58 *** 0.52 *** 0.40 *** 0.27 ***
Sylhet
-0.28 *** -0.18 ** -0.08 -0.22 ** -0.24 *** -0.42 ***
N Observations
Notes: All coefficients are weighted according to the survey design;
Source: BDHS
21,355 19,762
***p<0.001, **p<0.01, *p<0.05, +p<0.10
a
This variable includes membership in Grameen Banks, BRAC, BRDB, mother's club and other micro credit
organizations
21,365 21,358 21,354 19,738
Table 4.2: Pooled Sample Bivariate Regression: Women's Participation in Decision-Making
Predicted by Women's Characeteristics and Control Variables
Woman is a member of a
grameen bank
a
Woman's
Health
Large
Purchases
Daily
Purchases
Visits to
Family and
Friends
Child
Health Care
Bivariate Regression Coefficients
Explanatory Variables
Decision
Making
Additive
Scale
70
women with less household assets. Rural women are less likely to participate in decision-
making compared with urban women. Christianity stands out among the four major
religions in Bangladesh where women have higher decision-making power than women
who are Muslim. Finally, the six divisions of Bangladesh have varying influences on
women’s status. The Sylhet division in particular has a negative impact on all five
decision-making variables compared with Barisal division; however women in the
remaining four divisions have relatively high participation in household decision-making.
As explained in section 3 of this chapter, these controls are in the expected direction and
are consistent with theory. In the next section, I analyze the marginal effect of each
variable by running a multivariate logistic analysis for each decision-making outcome.
Multivariate logistic regression results are presented in table 4.3 through 4.8.
10
The combined decision-making scale (table 4.3) summarizes the key findings. In a setting
such as Bangladesh, age confers authority and status. Therefore, while holding other
variables constant, women who belong to an older age category are more likely to
participate in all decision-making domains, than women who are younger. This
relationship between women’s age and decision-making has not changed from 1999 to
2007.
Education at all levels (primary, secondary and higher education) has a positive
impact on decision-making compared with having no education. This is expected as
education levels are enhanced; women will have increased agency as well as increased
10
The pooled sample chi-square test (see Appendix - A1) indicates whether coefficients in 1999 are jointly
significantly different from coefficients in 2007. Significant differences in individual coefficients are
indicated by " in the regression tables. The results indicate that the regression coefficient vector differs
systematically between the 1999 and 2007 (where p<0.05).
71
negotiating power both at home and in the community. It should be noted that this is an
independent effect, which means regardless of one’s employment status, higher levels of
education give women more authority to influence household decision-making (since
most men are better educated than their wives: when holding men’s education constant,
women’s higher level of education may also imply a reduction in the education gap
between spouses). The coefficients of education categories have not changed from 1999
to 2007, which means their predictability has not changed overtime.
As explained earlier, work empowers women and allows them to influence
household decisions, as they no longer have to depend on their husbands for resources.
This is evident in the multivariate analyses, where women who work have greater odds of
participating in decision-making than women who don’t work. However, the effect of
work has weakened from 1999 to 2007 (0.398 and 0.214 respectively in Table 4.3), this is
especially true for decisions regarding woman’s health and large purchases. What this
may substantially mean is that as women’s employment becomes more widespread the
strength of the variable becomes weaker, because it is no longer a novel attribute in
women.
Membership in organizations such as the Grameen Bank has a positive
independent effect on all five decision-making outcomes. As explained earlier,
membership in microcredit program benefits women in two ways: (1) it provides women
credit to engage in self-employment activities, and (2) it gives access to a large social
network, which allows women to share information and knowledge. This is an
encouraging finding, especially because microfinance programs were first introduced in
72
Bangladesh by Muhammad Yunus (1983) to empower women by giving them access to
credit, skills and social capital. It is wise to continue such programs as means of
empowering women in less-developed countries. It should be noted that the effect of
membership in a microcredit program is significantly stronger in 2007 on decisions
regarding children’s health care (odds ratio is 1.02 in 1999 and 1.25 in 2007). Thus, it can
be said, these programs not only benefit women’s wellbeing, but also their children’s
wellbeing.
Household headship is the strongest predictor of decision-making regardless of
woman’s age, education level and employment. In a patriarchal country like Bangladesh,
women rarely assume household headship unless her spouse is absent for a prolong
period, or if her spouse is deceased or ill. According to these results it is clear that women
assume household headship, they strongly influence child health, than on any other
decision-making sphere.
The control variables, presented in the second half of each multivariate regression
table indicate that men’s education does not have an independent effect on decision-
making in any of the outcome variables, even the joint square significant tests does not
yield significant results (see joint significance test in Appendix-A2). This can be partly
explained by the correlation between women’s higher education and her partner’s higher
education ($ = 0.54***). I decided to keep this variable in the multivariate model because
controlling for partner’s education, the wife’s education is interpreted as relative
education (or a measure of relative power).
73
The asset scale that I created using household items (almirah, television and
table/chair) is positive and significant at p<0.05 in 1999 for three outcomes – woman’s
own health, visits to family/friends and child health. However, it does not have a
significant effect in 2007. This may be due to relative increase in household wealth from
1999 to 2007 where most households can afford to buy the above items, which may have
reduced the variation in asset ownership between the rich and the poor.
Rural areas are less progressive and are more likely to hold on to traditional social
norms and traditions (e.g. purdah) that disempower women. Thus, the results consistently
show that women who live in rural areas are less likely to make household decisions
compared with their urban counterparts. The relationship between this variable and
decision-making has not changed overtime, which means, rural residence is still strongly
predictive in 2007 as it was in 1999.
Out of the four religion variables, only Buddhism has a positive independent
effect on decision-making compared with Islam. This is perhaps due to the gender
egalitarian thoughts promoted by Buddhist teachings (Seth 2001). But this effect
disappears in 2007. This could be due to the significant reduction in the proportion of
Buddhist in Bangladesh which fell from 0.9 to 0.6 percent from 1999 to 2007, or it could
be because the association between religious beliefs and gender norms no longer exist in
Bangladesh. Whatever the reason may be, this is an interesting finding because for the
longest time scholars attributed women’s lower status to certain religious beliefs in less-
developed countries. But these results clearly show that as nations develop, the impact of
religion fades away.
74
The final control variable captures division-level variations that affect women’s
participation in household decision-making. Women in Dhaka, Khulna and Rajashani
consistently have higher odds-ratios of decision-making compared with Barisal division,
this is expected because Barisal division which is closest to the Bay of Bengal is subject
to constant environmental disasters such as cyclones and flooding, therefore there is little
economic development in this division compared with Dhaka where much of
Bangladesh’s industrialization and infrastructural development has taken place. Sylhet,
however, seems to lag behind in almost all the decision-making domains. As explained
earlier, Sylhet has the lowest HDI index in the nation, and it has the worst social
indicators compared with the national average. Therefore it is not surprising that women
in this division are less likely to participate in household decision-making compared with
their counterparts in other divisions. In addition, these divisional variations in decision-
making indicates the differences in community level variables such as religiosity of the
community, literacy level, availability of school for girls, availability of employment
opportunities for women and the period of exposure to women’s full time employment,
which are not directly measured in BDHS. Thus, the division variable captures some of
the unobservable variables that affect women’s decision-making power in the household.
Another observation that is worth mentioning is the significant increase in these
coefficients form 1999 to 2007. This indicates that there have been significant changes in
these divisions that differentiate them even more from Barisal division. Or, it could be
because Barisal has become even more under-developed compared with other divisions
due to the 2007 super cyclone Sidr that devastated Barisal division (UNICEF 2007).
75
People suffered from diarrhea and waterborne diseases, crops and livestock were
destroyed causing a food shortage and starvation. According to one Dhaka newspaper,
Barisal is still severely affected by the natural calamity, and is yet to fully recover from
the losses and damage to their crops and livestock (The Daily Star 2012). Although
natural disasters don’t directly influence women’s participation in decision-making, when
resources are scarce, it is natural for household members to conserve. Therefore, it is
possible that, men in the household make decisions on what to purchase for food and
when to take the children to the health clinic, instead of consulting their wives during
difficult time periods.
The multivariate results indicated that women’s age, education, work, household
headship and membership in microcredit programs have a positive impact on their
participation in household decision-making. However, the strength of the coefficients in
some variables have weakened overtime indicating that certain attributes that used to
predict women’s decision-making power no longer have the same impact it used to have.
In the next section I discuss the results of the Oaxaca-Blinder decomposition, which
explores the sources of improvements in women’s decision-making from 1999 to 2007.
76
Table 4.3: Woman’s Participation in Decision Making: Combined Scale (0-5)
Table 4.3: Woman's Participation in Decision Making: Combined Scale (0-5)
Coeff Coeff
Women's Characteristics
Woman's age in categories
15-24 years (Reference)
25-34 years 0.496 *** (0.052) 0.357 *** (0.052)
35-49 years 0.584 *** (0.055) 0.462 *** (0.054)
Woman's education level
No Education (Reference)
Primary Education 0.126 * (0.054) 0.072 (0.052)
Secondary Education 0.158 * (0.072) 0.172 ** (0.063)
Higher Education 0.467 *** (0.115) 0.512 *** (0.101)
Woman works 0.398 *** (0.049) 0.214 *** (0.042)
!
Woman is the household head 1.464 *** (0.056) 1.293 *** (0.040)
!
0.116 + (0.064) 0.150 *** (0.041)
Controls
Partner's Education
No Education (Reference)
Primary Education -0.006 (0.057) -0.063 (0.053)
Secondary Education 0.010 (0.063) -0.107 + (0.058)
Higher Education 0.148 (0.091) 0.096 (0.084)
Household Asset Scale (0-3) 0.096 *** (0.025) 0.010 (0.022)
!
Rural Residence -0.132 ** (0.047) -0.223 *** (0.040)
Religion
Islam (Reference)
Hinduism 0.028 (0.064) 0.002 (0.065)
Buddhism 0.764 *** (0.172) -0.292 (0.358)
!
Christianity 0.496 + (0.285) 0.573 (0.375)
Division
Barisal (Reference)
Chittagong -0.032 (0.082) 0.259 *** (0.071)
!
Dhaka 0.063 (0.079) 0.549 *** (0.067)
!
Khulna 0.145 + (0.082) 0.549 *** (0.070)
!
Rajashani 0.285 *** (0.081) 0.648 *** (0.068)
!
Sylhet -0.484 *** (0.094) 0.025 (0.091)
!
Constant 2.330 *** (0.096) 2.642 *** (0.089)
N Observations
F Statistic
R-squared
Source: BDHS
Robust S.E.
1999 2007
Ordinary Least Square Regression
Robust S.E.
! =
1999 and 2007 coefficients significantly different from one another (p <= 0.05)
Notes: Coefficients and robust standard errors are weighted according to the sample design
***p<0.001, **p<0.01, *p<0.05, +p<0.10
a
This variable includes membership in Grameen Banks, BRAC, BRDB, mother's club and other micro
credit organizations
0.084
Woman is a member of a grameen bank
a
9,727
83.15
0.098
9,764
75.63
77
Table 4.4: Woman’s Participation in Decision Making Regarding Her Own Health
Odds
Ratio
Coeff
Odds
Ratio
Coeff
Women's Characteristics
Woman's age in categories
15-24 years (Reference)
25-34 years 1.43 0.359 *** (0.054) 1.33 0.288 *** (0.061)
35-49 years 1.52 0.420 *** (0.058) 1.40 0.334 *** (0.064)
Woman's education level
No Education (Reference)
Primary Education 0.99 -0.015 (0.058) 1.06 0.060 (0.064)
Secondary Education 1.01 0.011 (0.078) 1.18 0.165 * (0.079)
Higher Education 1.54 0.430 ** (0.139) 1.83 0.606 *** (0.139)
Woman works 1.46 0.378 *** (0.057) 1.11 0.100 (0.055)
!
Woman is the household head 7.16 1.969 *** (0.164) 8.24 2.109 *** (0.149)
0.96 -0.040 (0.072) 1.13 0.119 * (0.054)
Controls
Partner's Education
No Education (Reference)
Primary Education 0.98 -0.019 (0.061) 1.00 0.004 (0.065)
Secondary Education 0.94 -0.058 (0.068) 0.97 -0.033 (0.072)
Higher Education 1.09 0.087 (0.101) 1.06 0.062 (0.107)
Household Asset Scale (0-3) 1.12 0.114 *** (0.027) 1.01 0.014 (0.027)
!
Rural Residence 0.92 -0.081 (0.052) 0.75 -0.294 *** (0.053)
!
Religion
Islam (Reference)
Hinduism 1.07 0.064 (0.070) 1.00 -0.002 (0.082)
Buddhism 2.16 0.771 ** (0.269) 0.32 -1.128 * (0.507)
!
Christianity 1.80 0.588 (0.474) 1.69 0.524 (0.505)
Division
Barisal (Reference)
Chittagong 1.07 0.069 (0.088) 1.16 0.151 + (0.082)
Dhaka 1.01 0.014 (0.084) 1.18 0.169 * (0.079)
Khulna 1.01 0.013 (0.088) 1.13 0.280 ** (0.084)
!
Rajashani 1.25 0.225 * (0.087) 1.30 0.261 ** (0.082)
Sylhet 0.71 -0.339 ** (0.098) 0.97 -0.030 (0.093)
!
Constant -0.280 ** (0.103) 0.110 (0.107)
N Observations
Wald Chi-square
Pseudo R-square
Source: BDHS
Logistic Regression
Robust S.E.
Table 4.4: Woman's Participation in Decision Making Regarding Her Own Health
1999 2007
Robust S.E.
! = 1999 and 2007 coefficients significantly different from one another (p <= 0.05)
Notes: Coefficients and robust standard errors are weighted according to the sample design
Woman is a member of a
grameen bank
a
***p<0.001, **p<0.01, *p<0.05, +p<0.10
a
This variable includes membership in Grameen Banks, BRAC, BRDB, mother's club and other micro
credit organizations
457
0.045 0.047
399
10,140 10,955
78
Table 4.5: Woman’s Participation in Decision Making Regarding Large Purchases
Odds
Ratio
Coeff Odds Ratio Coeff
Women's Characteristics
Woman's age in categories
15-24 years (Reference)
25-34 years 1.51 0.411 *** (0.055) 1.56 0.445 *** (0.062)
35-49 years 1.62 0.483 *** (0.059) 1.64 0.496 *** (0.066)
Woman's education level
No Education (Reference)
Primary Education 1.18 0.163 ** (0.059) 1.04 0.035 (0.066)
Secondary Education 1.24 0.215 ** (0.079) 1.16 0.149 + (0.079)
Higher Education 1.81 0.592 *** (0.145) 1.51 0.410 ** (0.139)
Woman works 1.56 0.446 *** (0.059) 1.31 0.269 *** (0.057)
!
Woman is the household head 4.76 1.560 *** (0.149) 5.04 1.617 *** (0.131)
1.18 0.169 * (0.074) 1.25 0.224 *** (0.056)
Controls
Partner's Education
No Education (Reference)
Primary Education 0.95 -0.046 (0.061) 0.94 -0.064 (0.067)
Secondary Education 0.97 -0.026 (0.069) 0.80 -0.228 ** (0.073)
!
Higher Education 1.13 0.125 (0.103) 0.99 -0.011 (0.110)
Household Asset Scale (0-3) 1.05 0.045 (0.027) 1.01 0.015 (0.027)
Rural Residence 0.87 -0.145 ** (0.053) 0.80 -0.224 *** (0.054)
Religion
Islam (Reference)
Hinduism 0.93 -0.068 (0.071) 0.87 -0.135 + (0.081)
Buddhism 1.96 0.671 * (0.265) 0.55 -0.595 (0.462)
!
Christianity 0.90 -0.104 (0.460) 3.29 1.192 + (0.668)
Division
Barisal (Reference)
Chittagong 0.98 -0.024 (0.088) 1.41 0.342 *** (0.081)
!
Dhaka 1.17 0.154 + (0.085) 1.96 0.674 *** (0.079)
!
Khulna 1.30 0.261 ** (0.089) 2.11 0.745 *** (0.084)
!
Rajashani 1.55 0.436 *** (0.088) 2.14 0.760 *** (0.083)
!
Sylhet 0.73 -0.320 ** (0.098) 1.29 0.252 ** (0.093)
!
Constant -0.220 * (0.104) -0.277 * (0.108)
N Observations
Wald Chi-square
Pseudo R-square
Source: BDHS
Table 4.5: Woman's Participation in Decision Making Regarding Large Purchases
Logistic Regression
1999 2007
Robust S.E. Robust S.E.
Woman is a member of a
grameen bank
a
Notes: Coefficients and robust standard errors are weighted according to the sample design
***p<0.001, **p<0.01, *p<0.05, +p<0.10
a
This variable includes membership in Grameen Banks, BRAC, BRDB, mother's club and other micro credit
organizations
! = 1999 and 2007 coefficients significantly different from one another (p <= 0.05)
10,135 10,953
491 557
0.044 0.054
79
Table 4.6: Woman’s Participation in Decision Making Regarding Daily Purchases"
Odds
Ratio
Coeff
Odds
Ratio
Coeff
Women's Characteristics
Woman's age in categories
15-24 years (Reference)
25-34 years 1.61 0.476 *** (0.056) 1.76 0.565 *** (0.064)
35-49 years 1.69 0.524 *** (0.060) 2.04 0.713 *** (0.068)
!
Woman's education level
No Education (Reference)
Primary Education 1.22 0.196 ** (0.060) 1.23 0.206 ** (0.071)
Secondary Education 1.18 0.164 * (0.079) 1.15 0.139 + (0.083)
Higher Education 1.53 0.425 ** (0.143) 1.39 0.332 * (0.145)
Woman works 1.44 0.362 *** (0.060) 1.45 0.370 *** (0.061)
Woman is the household head 8.59 2.150 *** (0.195) 12.38 2.516 *** (0.206)
1.27 0.241 ** (0.076) 1.17 0.160 ** (0.059)
Controls
Partner's Education
No Education (Reference)
Primary Education 0.98 -0.020 (0.062) 0.92 -0.085 (0.071)
Secondary Education 1.06 0.056 (0.070) 0.85 -0.165 * (0.077)
!
Higher Education 1.13 0.121 (0.103) 1.18 0.163 (0.114)
Household Asset Scale (0-3) 1.05 0.050 + (0.028) 0.95 -0.051 + (0.029)
!
Rural Residence 0.91 -0.090 + (0.054) 0.87 -0.140 * (0.056)
Religion
Islam (Reference)
Hinduism 0.96 -0.041 (0.072) 1.05 0.046 (0.083)
Buddhism 2.57 0.943 ** (0.280) 1.16 0.147 (0.492)
Christianity 1.20 0.183 (0.462) 1.97 0.679 (0.663)
Division
Barisal (Reference)
Chittagong 1.04 0.036 (0.089) 1.38 0.326 *** (0.084)
!
Dhaka 1.29 0.254 ** (0.086) 2.07 0.726 *** (0.083)
!
Khulna 1.35 0.297 ** (0.090) 1.76 0.566 *** (0.088)
!
Rajashani 1.49 0.396 *** (0.089) 2.13 0.757 *** (0.087)
!
Sylhet 0.70 -0.357 *** (0.099) 1.05 0.050 (0.097)
!
Constant -0.281 ** (0.105) -0.191 (0.112)
N Observations
Wald Chi-square
Pseudo R-square
Source: BDHS
Table 4.6: Woman's Participation in Decision Making Regarding Daily Purchases
Logistic Regression
1999 2007
Robust S.E. Robust S.E.
Woman is a member of a grameen
bank
a
Notes: Coefficients and robust standard errors are weighted according to the sample design
***p<0.001, **p<0.01, *p<0.05, +p<0.10
a
This variable includes membership in Grameen Banks, BRAC, BRDB, mother's club and other micro credit
organizations
! = 1999 and 2007 coefficients significantly different from one another (p <= 0.05)
10,131 10,953
495 580
0.049 0.074
80
Table 4.7: Woman’s Participation in Decision Making Regarding Visits to Family and Friends
Odds
Ratio
Coeff
Odds
Ratio
Coeff
Women's Characteristics
Woman's age in categories
15-24 years (Reference)
25-34 years 1.55 0.441 *** (0.055) 1.59 0.461 *** (0.062)
35-49 years 1.80 0.588 *** (0.060) 1.69 0.522 *** (0.066)
Woman's education level
No Education (Reference)
Primary Education 1.14 0.135 * (0.060) 1.08 0.079 (0.067)
Secondary Education 1.18 0.168 * (0.079) 1.18 0.167 * (0.080)
Higher Education 1.60 0.471 ** (0.148) 1.50 0.402 ** (0.146)
Woman works 1.35 0.297 *** (0.059) 1.20 0.181 ** (0.058)
Woman is the household head 7.37 1.998 *** (0.181) 8.00 2.080 *** (0.154)
1.20 0.182 * (0.076) 1.17 0.159 ** (0.056)
Controls
Partner's Education
No Education (Reference)
Primary Education 1.02 0.019 (0.061) 0.91 -0.090 (0.066)
Secondary Education 1.03 0.034 (0.070) 1.00 -0.001 (0.074)
Higher Education 1.18 0.165 (0.104) 1.31 0.272 * (0.115)
Household Asset Scale (0-3) 1.09 0.085 ** (0.028) 0.97 -0.029 (0.028)
!
Rural Residence 0.85 -0.160 ** (0.054) 0.75 -0.294 *** (0.055)
Religion
Islam (Reference)
Hinduism 1.07 0.063 (0.072) 1.07 0.066 (0.082)
Buddhism 1.83 0.603 ** (0.270) 1.07 0.069 (0.467)
Christianity 1.85 0.618 (0.522) 3.13 1.142 + (0.687)
Division
Barisal (Reference)
Chittagong 0.94 -0.064 (0.091) 1.17 0.160 + (0.083)
Dhaka 0.96 -0.043 (0.087) 1.57 0.452 *** (0.081)
!
Khulna 1.16 0.146 (0.091) 1.41 0.346 *** (0.085)
Rajashani 1.24 0.216 * (0.089) 1.98 0.686 *** (0.085)
!
Sylhet 0.62 -0.486 *** (0.100) 1.09 0.085 (0.095)
!
Constant -0.112 (0.106) -0.005 (0.110)
N Observations
Wald Chi-square
Pseudo R-square
Source: BDHS
Table 4.7: Woman's Participation in Decision Making Regarding Visits to Family and Friends
Logistic Regression
1999 2007
Robust S.E. Robust S.E.
Woman is a member of a
grameen bank
a
Notes: Coefficients and robust standard errors are weighted according to the sample design
***p<0.001, **p<0.01, *p<0.05, +p<0.10
a
This variable includes membership in Grameen Banks, BRAC, BRDB, mother's club and other micro
credit organizations
! = 1999 and 2007 coefficients significantly different from one another (p <= 0.05)
10,134 10,951
490 516
0.048 0.057
81
Table 4.8: Woman’s Participation in Decision Making Regarding Children’s Health Care
Odds
Ratio
Coeff
Odds
Ratio
Coeff
Women's Characteristics
Woman's age in categories
15-24 years (Reference)
25-34 years 1.46 0.379 *** (0.057) 1.31 0.271 *** (0.074)
35-49 years 1.66 0.507 *** (0.062) 1.49 0.398 *** (0.079)
Woman's education level
No Education (Reference)
Primary Education 1.14 0.134 * (0.061) 1.04 0.043 (0.078)
Secondary Education 1.20 0.184 * (0.083) 1.28 0.249 * (0.096)
Higher Education 1.84 0.609 *** (0.157) 2.21 0.794 *** (0.197)
Woman works 1.49 0.400 *** (0.062) 1.38 0.321 *** (0.070)
Woman is the household head 9.66 2.268 *** (0.217) 12.45 2.522 *** (0.252)
1.02 0.019 (0.076) 1.25 0.220 ** (0.066)
!
Controls
Partner's Education
No Education (Reference)
Primary Education 0.99 -0.007 (0.063) 0.99 -0.008 (0.078)
Secondary Education 1.02 0.018 (0.071) 0.95 -0.053 (0.088)
Higher Education 1.15 0.136 (0.109) 1.21 0.193 (0.139)
Household Asset Scale (0-3) 1.11 0.104 *** (0.029) 1.04 0.043 (0.033)
Rural Residence 0.87 -0.134 * (0.056) 0.77 -0.259 *** (0.066)
Religion
Islam (Reference)
Hinduism 1.05 0.051 (0.074) 0.98 -0.023 (0.098)
Buddhism 2.62 0.965 ** (0.313) 1.74 0.556 (0.558)
Christianity 14.92 2.703 ** (1.031) 3.39 1.220 (0.903)
Division
Barisal (Reference)
Chittagong 0.86 -0.151 (0.095) 1.07 0.068 (0.094)
Dhaka 0.93 -0.073 (0.091) 1.62 0.483 *** (0.096)
!
Khulna 0.93 -0.075 (0.094) 1.54 0.432 *** (0.101)
!
Rajashani 1.07 0.067 (0.093) 1.80 0.590 *** (0.100)
!
Sylhet 0.60 -0.518 *** (0.103) 0.79 -0.242 * (0.104)
Constant 0.093 (0.110) 0.464 *** (0.128)
N Observations
Wald Chi-square
Pseudo R-square
Source: BDHS
Table 4.8: Woman's Participation in Decision Making Regarding Children's Health Care
Logistic Regression
1999 2007
Robust S.E. Robust S.E.
Woman is a member of a
grameen bank
a
Notes: Coefficients and robust standard errors are weighted according to the sample design
***p<0.001, **p<0.01, *p<0.05, +p<0.10
a
This variable includes membership in Grameen Banks, BRAC, BRDB, mother's club and other micro credit
organizations
! = 1999 and 2007 coefficients significantly different from one another (p <= 0.05)
9,784 9,731
450 404
0.050 0.062
82
5.3 Oaxaca-Blinder Decomposition
In table 4.9A and 4.9B, I summarize the decomposition results for all five decision-
making domains including the combined scale. Table 4.9A presents the predicted change
in decision-making if means or coefficients in 1999 were at 2007 levels. Table 4.9B
presents the predicted percentage difference in decision-making if means or coefficients
in 1999 were at 2007 levels. For instance, the last column presents the results for
decisions on child health. The proportion of women who participate in this decision-
making sphere has increased by 13.7 percentage points from 1999 to 2007. If women in
1999 had same socioeconomic characteristics as women in 2007, there would have been a
2.8 predicted percentage point increase (0.028) in women’s decision-making on child
health in 1999 (see 21% increase in the last column of table 4.9B), if women’s
coefficients were at 2007 levels, there would have been a 4.3 predicted percentage point
decrease (-0.043) in 1999 (see 32% decrease in the last column of table 4.9B). The latter
finding indicates that the relationship between women’s characteristics and decision-
making (i.e., coefficients) have weakened in 2007 compared with 1999. This was
discussed in detail in the previous section.
Since the five decision-making spheres are measuring different domains of
household decision-making, we see varying levels of predicted values. For example if
women in 2007 had the same means as women in 1999. There would be a 63% increase
in the proportion of women who participate in decisions regarding large purchases (see
third column from the left in table 4.9B), and there would be a 52% increase in the
proportion of women who participate in decision regarding visits to family and friends
83
(see fifth column from the left in table 4.9B). This shows that the improvement in
women’s socioeconomic characteristics influence certain decision-making domains more
than others.
Two key findings emerge from table 4.9A and 4.9B. First, improvements in
women’s education, work, membership in microcredit programs and household headship
accounts for on average 3-percentage point increase in decision-making. This partly
supports my hypothesis that improvements in women’s characteristics affect
improvements in decision-making. The second finding is that, the change in intercepts
accounts for a large proportion of the change in decision-making (except for large
purchases), but its impact is counteracted by the reduction in coefficients due to the
weakening of the association between women’s characteristics and decision-making. The
intercepts subsume the effects of differences in unobserved predictors (Jann 2008). Such
unobserved predictors may include, changes in norms and attitudes towards women,
omitted development variables, or other omitted characteristic that impacts women’s
decision-making power. In Bangladesh it could be changes to the law and government
programs. For example, the Prime Minister of Bangladesh issued a memorandum in 2003
urging all government officials to work towards abolishing the practice of dowry. Dowry
is closely related to age at first marriage, as parents are forced to marry their daughters at
young age to avoid an expensive dowry. If the dowry system is abolished, girls can
continue schooling and join the labor force before getting married. Further, direct
intervention from the mass media, which shows workingwomen’s images through
dramas, advertisements and feature films, has a profound impact on the culture of the
84
Bangladesh. Moreover, as modernization theory suggests, the rapid globalization of the
economy fosters a more progressive attitude towards women. Therefore, the intercept
may have subsumed these changes in norms and attitudes. However, I cannot rule out
other development variables and women’s characteristics that may strongly predict
women’s decision-making, but are omitted from the model. Thus, I cannot confirm that
the large differences in intercept are due to unobservable social norms without empirical
evidence.
85
Table 4.9A: Oaxaca-Blinder Decomposition: Predicted Changes in Decision-Making
Table 4.9B: Oaxaca-Blinder Decomposition: Percentage Differences in Decision-Making if
Characteristics in 1999 were at 2007 Levels
Decision
Making
Combined
Scale (0-5)
Woman's
Own Health
Large
Purchases
Daily
Purchases
Visits to
Family and
Friends
Child Health
Care
Change in outcome (Y 2007-Y 1999) 0.452*** 0.072*** 0.052*** 0.084*** 0.058*** 0.137***
Change due to means (X 2007-X 1999)B 1999 0.194*** 0.029*** 0.037*** 0.038*** 0.038*** 0.037***
Due to women's characteristics 0.158*** 0.019*** 0.033*** 0.033*** 0.030*** 0.028***
Due to control variables 0.035** 0.010*** 0.005+ 0.005+ 0.008** 0.008**
Change due to coefficients (B 2007-B 1999)X 1999 0.311*** 0.044*** 0.021** 0.063*** 0.036*** 0.107***
Due to women's coefficient -0.146* -0.013 -0.022 0.000 -0.022 -0.043*
Due to control variables' coefficients 0.145 -0.040 0.053+ 0.028 0.028 0.045
Due to intercept 0.312* 0.100** -0.010 0.036 0.030 0.105**
Overall interation (X 2007-X 1999)*(B 2007-B 1999) -0.052* -0.002 -0.007 -0.017** -0.016** -0.007
Interaction - women's mean and coefficient -0.021 0.004 -0.005 -0.008 -0.006 -0.002
Interaction - control's mean and coefficient -0.031* -0.006 -0.001 -0.009** -0.010** -0.005
Note: ***p<0.001, **p<0.01, *p<0.05, +p<0.10
Source: BDHS
Decision
Making
Combined
Scale (0-5)
Woman's
Own Health
Large
Purchases
Daily
Purchases
Visits to
Family and
Friends
Child Health
Care
Change in outcome (Y 2007-Y 1999) 0.452
a
* 7.2** 5.2** 8.4** 5.8** 13.7**
Change due to means (X 2007-X 1999)B 1999 43%* 41%** 71%** 45%** 65%** 27%**
Due to women's characteristics 35%* 27%** 63%** 39%** 52%** 21%**
Due to control variables 8%* 14%** 9%** 6%** 14%** 6%**
Change due to coefficients (B 2007-B 1999)X 1999 69%* 62%** 41%** 75%** 62%** 78%**
Due to women's coefficient -32%+ -18%* -42%* 0%* -37%* -32%*
Due to control variables' coefficients 32% -55%* 102%* 33%* 48%* 33%*
Due to intercept 69% 135%* -20%* 43%* 51%* 77%*
Overall interation (X 2007-X 1999)*(B 2007-B 1999) -12%* -2%** -12%** -20%** -27%** 5%**
Interaction - women's mean and coefficient -5%* 6%** -10%** -9%** -10%* -1%**
Interaction - control's mean and coefficient -7%* -8%** -2%** -11%** -17%* -4%**
Note: ***p<0.001, **p<0.01, *p<0.05, +p<0.10
Source: BDHS
86
6. Conclusion
The purpose of this chapter was to examine whether: (1) the relationship between
women’s characteristics and participation in household decision-making has changed
over time, and (2) to identify the sources of improvement in women’s participation in
decision-making in Bangladesh. I analyzed five decision-making spheres available in the
Bangladesh DHS.
The multivariate regression results confirms past empirical work on women’s
status. As expected women’s age, education, employment, household-headship and
membership in microcredit programs have a significant impact on decision-making,
while controlling for partner’s education, household assets, rural residence, religion and
division. However, the strength of the coefficients of some variables (e.g., woman’s
work, household asset scale and religion) decreased from 1999 to 2007, while the effect
of division increased. Therefore, it can be said that variables that used to predict women’s
decision-making in 1999 do not have the same predictability in 2007. This is an
interesting finding because it means that relative importance of variables that used to
serve as proxies for women’s power have weakened, or are no longer significant in
predicting decision-making in 2007.
The second part of the analyses was focused on investigating the sources of
improvement in women’s decision-making power in the household. On average, about 3-
percentage point increase (i.e., 35% increase) in decision-making is accounted for by the
improvements in women’s characteristics from 1999 to 2007. Although changes in the
means have a substantial impact on decision-making, the majority of the change is due to
87
the differences in the intercept, which encompass omitted variables. We can assume that
these are unobservable variables such as shifts in social norms and attitudes towards
women or omitted development variables. However, without further investigation I
cannot confirm the cause for the large differences in the intercept.
As with any analysis based on cross-sectional survey data, this study has some
limitations. The explanatory variables leave much of the actual change in the dependent
variable unexplained as indicated by the large contribution of the differences in
intercepts. These results can be further improved by including community level variables,
such as religiosity in the community, proportion of women in the work force, female
literacy level and attitudes towards women. Nevertheless, since the goal of this chapter
was to assess whether gains in women’s socioeconomic characteristics account for
improvements in women’s status as measured by five variables, I acknowledge this
shortcoming but do not think it detracts from the findings shown here.
88
Chapter Five
The Role of Social Context in Household Decision-Making in
Bangladesh and Indonesia
1. Introduction
Cross-country comparisons provide an important way to gain insights into spatial
variations in women’s status. As explained in chapter 2, both Bangladesh and Indonesia
have shown massive improvements in women’s socioeconomic characteristics from the
late 1990s to 2007. But these two countries differ from one another in two aspects: (1)
family organization and (2) economic development. It is useful to investigate how
differences in these factors, along with other socioeconomic characteristics, may or may
not affect women’s decision-making power and whether or not it improves over time.
In this chapter I accomplish several goals. First, I repeat the previous analysis that
was conducted for Bangladesh (see chapter 4) by using data from the Indonesian Family
Life Surveys (IFLS). I examine how women’s characteristics and participation in
household decision-making have changed from 1997 to 2007 in Indonesia. Then I
explore the sources of improvement in women’s participation in decision-making. In the
second part of the analyses, I directly compare Bangladesh with Indonesia to investigate
how the association between the determinants and decision-making outcomes differ
between the two countries. Finally, I explore how much of the difference in decision-
making is accounted for by the differences in the population composition (i.e., means),
and how much is accounted for by the difference in association (i.e., coefficients).
89
The data derive from the 1997 and 2007 Indonesian Family Life Surveys (IFLS)
and the 2007 Bangladesh Demographic and Health Surveys (BDHS). Both surveys are
nationally representative and are appropriate for examining nationwide trends and
patterns in women’s status. They include information on women’s demographics,
decision-making, partner’s characteristics, etc. IFLS, in particular, includes detailed
information on women’s share of household assets at the time of the survey and women’s
parents’ relative positions in the society compared with their in-laws at the time of
marriage. The first part of the chapter investigates the change in coefficients between the
two survey years and exploits the unusually detailed information on women’s status
available in the IFLS. The second part of the chapter takes advantage of the different
types of family organization and traditions in Bangladesh and Indonesia to examine
women’s status in the household and its impact on decision-making.
2. Background
2.1 A Comparison between Bangladesh and Indonesia
Although I already discussed the social and economic context of Bangladesh and
Indonesia in chapter 2, it is worth reiterating the key similarities and dissimilarities that
make Bangladesh and Indonesia suitable candidates for comparison. In the past, these
two countries were remarkably similar on a number of background features such as
religion and organization of the agricultural sector. In the 1960s, both countries had very
low life expectancy (about 40 years), and high fertility (about 6 children per women over
her reproductive years). Large fractions of children between ages 6 and 12 were not
90
enrolled in primary school (Arthur and McNicoll 1978; Hugo et al. 1987; World Bank,
1982). However, after the 1960s, Indonesia developed at a faster pace than Bangladesh
on most dimensions. By the mid-1990s, Indonesia was at a considerably higher level of
socioeconomic development than Bangladesh. During the fiscal year 2007-08
Bangladesh’s per capita income was at US$ 599, whereas Indonesia’s was at US$ 3,900.
At first glance one might think this difference in socioeconomic development is the key
reason for the relative high status of Indonesian women compared with Bangladeshi
women. But part of the difference lies in the patters of family organization between the
two countries.
Bangladesh is a Muslim country; the vast majority of the population practices
Islam. According to Islamic tenets, responsibility of one’s parents is in the hands of adult
children. In Bangladesh this responsibility largely falls to sons (Kabir, Szebehely and
Tishelman 2000; Mahmood 1992; Rahman 1999). As a result, sons are preferred over
daughters, and this difference in the relative importance of sons is an integral part of the
patriarchal family system. After marriage, a woman is expected to move in with her
husband, his parents, his brothers and their wives. Her own parents no longer have any
claim on her labor. The low status of women in Bangladesh is a result of this rigid family
system coupled with strong traditions of purdah. According to the tradition of purdah,
respectable women do not engage in trade or fieldwork or leave the family for other than
traditionally specified visits to relatives (Arthur and McNicoll 1978); this means they
cannot find employment outside of their homes. Thus, separation and divorce are real
threats to the wellbeing of Bangladeshi women.
91
Although the majority of the Indonesian population is also Muslim, Islamic tenets
and traditional law obligates children of both sexes to care for their older parents
(Mahmood 1992; Frankenberg, Lillard and Willis 2002; Keasberry 2001). Therefore,
differences in the relative importance of sons and daughters do not exist in the Indonesian
culture. Further, purdah is not at all practiced in Indonesia; women have complete
freedom of movement. In fact, Indonesian women have traditionally played a prominent
role in both the public and domestic realms. It is not uncommon for couples to work
together, not only as part of an economic survival strategy (Koentjaraningrat 1967), but
also as an interactive decision-making unit (Bangun 1981). Wives are given an equal say
in the determination of important household matters, particularly those involving control
over financial resources (Geertz 1961; Mangkuprawira 1981). Women work in
agriculture, sometimes with their husbands and sometimes on their own, and they engage
in trade. Women also own property separate from their husbands. Since women are not
dependent on their husbands economically, divorce is not a major threat to their survival.
If divorced, women usually are able to rely on their parents to provide them and their
children with a place to live (Heaton, Cammack and Young 2001; Jones 1994). Further,
Indonesian society is extremely diverse in terms of ethnicity; therefore there are a variety
of traditions known as adat
11
with respect to the organization of family and community
life. This variation in community norms and ethnic traditions regarding the family gives
rise to different levels of power for different household members and is useful in
determining differences in the causes of women’s decision-making power.
11
adats are defined the customary laws regarding marriage, women’s rights, divorce, inheritance and
residence.
92
2.2 Determinants of Women’s Status in the IFLS
As explained in the previous chapter, past models of the household decision-making
assumed that all household members have identical preferences (unitary model). But in
reality, each individual has his or her own preference and resources. Through intra-
household bargaining and negotiation, the person who has relatively more power, or a
higher status in the household, exercises his or her preferences (collective model).
However, measuring relative status has proven to be a difficult task. Household members
may derive higher status from multiple sources, especially through education and
economic resources. The main focus of this dissertation is the gendered nature of
household preferences and decision-making, where women display dissimilar preferences
towards food, health, and childcare compared with their husbands.
In the previous chapter, we found that age, education, employment, access to
microcredit and household headship have net positive effects on women’s decision-
making in Bangladesh. But employment and therefore income can be also thought of as
outcomes of a bargaining process between husbands and wives. For instance, if time
allocation choices (including the time spent at work) are part of a negotiation between
husbands and wives, it is reasonable to suppose that the subsequent distribution and
spending of the individual income will also be part of that negotiation. Treating
employment as predetermined in women’s decision-making models may not be
appropriate as the estimates of the effect of employment on household decisions will be
subject to simultaneity bias. Studies that use individual labor income as a proxy for a
woman’s employment status are even more strongly prone to these biases (Gage 1995;
93
Mason 1996; Miles-Doan and Brewster). The same argument can be made with regard to
membership in microcredit organizations, as the decision to participate in such programs
may also be an outcome of the household bargaining process.
To avoid this problem of simultaneity bias, some studies have used non-labor
income or assets brought to marriage (e.g. dowry and inheritance), or the value of assets
owned at the time of marriage to measure relative power (McElroy and Horney 1981;
Schultz 1990; Thomas 1990; Quisumbing 1994; Thomas et al. 1997). However,
measurement-error is a real concern with collecting retrospective information on assets
brought to a marriage. Measurement-error may include recall bias in the value of the
assets, and the date of marriage, as well as error due to the respondent’s tendency to
either hide resources or inflate their status. It may also be difficult for respondents to
report the real value of the assets in present currency, as it may seem too low after many
years of inflation (Frankenberg and Thomas 2003).
Therefore, share of current household assets owned by the woman may serve as a
better indicator of relative status. The IFLS includes detailed questions on ownership of
household assets at the time of the survey. Collecting such data is not standard practice in
broad-purpose socioeconomic surveys, but was implemented in the IFLS in an attempt to
measure the relative economic position of husbands and wives. IFLS collects information
from each spouse about the value and ownership of all the assets owned by any member
of the household. For assets for which some portion was owned by the husband and wife
(or both), each respondent was asked to report the proportion that each partner owned.
The common assets in an Indonesian household include the house they occupy, the land,
94
jewelry, livestock, furniture, appliances, vehicles, etc. In this chapter, I will be using
‘share of assets’ in place of woman’s work. However, by using this variable, I am making
a central assumption that the distribution of assets is an indicator of power over decision-
making. One can contend that a woman who has titular ownership of assets may not have
real control over them. An understanding of the cultural context Indonesia may help in
making this argument.
The ethnographic literature indicates that resources brought to marriage by a
woman tend to be held under her control; gold and jewelry are commonly cited as
examples of such assets. In an event of marriage dissolution, these assets remain with her
and revert to her family if she dies and leaves no heirs. Further, Hart (1978) and Wolf
(1991) state that assets acquired by Javanese women through their own employment also
remain under their own control. In a community survey conducted as part of the IFLS,
adat experts report that under the traditional law, a woman is allowed to own land or a
field by herself after marriage. Women are also allowed to own their businesses. In the
event of a divorce, both the husband and wife leave the marriage with the assets they had
owned prior to marriage; the assets acquired after the marriage are either split evenly or
divided up according to who obtained the assets. However, whether reported ownership
of assets reflects control over decision-making is an empirical issue. Beegle,
Frankenberg, and Thomas (2001) analyzed the association between ownership of assets
and the use of prenatal and delivery care in Indonesia. They find that, compared with a
woman with no assets she perceives as being her own, a woman with some share of
household assets has greater influence over reproductive health decisions. In my analysis,
95
I hope to find similar relationships between women’s share of household assets and her
control over decision-making regarding child health and large expensive purchases.
In the bargaining model of the household, if the threat point is divorce, a person’s
power depends on the resources he or she could rely on in the event of divorce (Pollak
2005). In addition to one’s earning potential and assets taken from the marriage, a
person’s family would also be an important source of support and assistance in the event
of a divorce. Therefore, one’s family background plays an important role in moderating
power within the household. These power relations are formed at the time of marriage
and tend to influence marital dynamics (Beegle, Frankenberg and Thomas 2001).
Frankenberg and Thomas (2003) describe the general agreement found among
participants of focus groups in Indonesia regarding the influence of family’s relative
social status. The focus group participants indicated that large differences in the
socioeconomic status of parents could cause problems because the spouse with higher
family status might look down on or try to dominate the other spouse.
The IFLS includes a questionnaire module that attempts to provide insight into
relative status of husbands and wives within the household. These questions focus on the
family backgrounds of husbands and wives at the time of marriage. Each respondent was
asked to evaluate the relative position of his or her parents in relation to his or her
spouse’s parents at the time of their marriage. Questions include six categories on which
the comparisons are made: the father’s job and education, the mother’s education, the
family position in the community, the quality of housing, and levels of earnings and
assets. Frankenberg and Thomas (2003) use status of the respondent’s family relative to
96
the spouses’ family at the time of their marriage to predict household decision-making.
They employ multinomial logit regressions to predict whether decisions are made by: (1)
the wife alone; (2) the husband alone; or (3) the wife and husband jointly, for five
decision-making spheres.
12
Their results indicated that men from higher status families
than their wives were more likely to make decisions about the health and education of
their children, purchasing large durable items, and using contraceptives than men from
lower status families.
I extend the current literature by examining the change in the means of the
predictors and the change in the strength of the predictors from 1997 and 2007—a time
period in which Indonesia has undergone dramatic economic and social change. I also
perform Oaxaca-Blinder decompositions to identify the sources of variation in decision-
making from 1997 to 2007, and between Bangladesh and Indonesia. In addition, I use the
share of household assets owned by women as a predictor of women’s decision-making –
a variable not considered in Frankenberg and Thomas (2003). Finally, I simplify their
analysis by looking at a binary outcome variable instead of a three-category outcome
variable, which makes multivariate and decomposition results easier to interpret.
3. Data
This study is based on the 2007 Bangladesh Demographic and Health Surveys (please
refer to chapter 4 for a description of the BDHSs), and 1997 and 2007 Indonesian Family
Life Surveys. The IFLS is a large-scale, multipurpose household panel survey. RAND
12
The five spheres include decisions regarding expenditure on food eaten at home, child education,
provision of health care for kids, purchasing large expensive items and contraceptive use
97
corporation conducted the first IFLS in 1993-94 in collaboration with Lembaga
Demografi, University of Indonesia. The sampling design for the first wave is the
primary determinant of the sample in subsequent waves. The IFLS1 sample was stratified
on provinces, and then randomly sampled within provinces. Provinces were selected to
maximize representation of the population, capture the cultural and socioeconomic
diversity of Indonesia, and be cost-effective to survey given the size and terrain of the
country. The sample included 13 of Indonesia’s 26 provinces containing 83 percent of the
population. Over 7,000 households were surveyed in 1993-94. The IFLS2 was conducted
between August 1997 and February 1998, and 94 percent of the 7,224 households that
were contacted in IFLS1 were re-interviewed (excluding household where all members
died), along with split-off households. IFLS4 was conducted between November 2007
and April 2008, and 93.6 percent of the IFLS1 households were re-contacted. My sample
includes 5,490 married women aged 15-49 years in 1997, and 9,823 married women aged
15-49 years in 2007. Although IFLS is a longitudinal survey, cross-sectional person-
weights allow us to get estimates for Indonesian population living in 1997 and 2007. The
IFLS survey scientists constructed these weights by raking the IFLS2 and IFLS4 samples
to external samples from the 1997 and 2007 SUSENAS (Survei Sosial Ekonomi Nasional
- National Socio-Economic Survey) in the 13 IFLS provinces, this is after having made
adjustments for sample attrition from 1993 to 2007 (see IFLS User Guide for details).
98
3.1 Decision-making
The IFLS has 17 items in the household decision-making module. Five of these decision-
making items increased significantly from 1997 to 2007. However, for the purpose of
comparison with Bangladesh, I use two decision-making variables that are comparable
across the two surveys: (1) child health and (2) large expensive purchases.
Figure 5.1 presents the percentages of Indonesian women who participated in
decision-making in 1997 to 2007. Because women are the primary caregivers of their
children, it is natural for them to be involved in issues that concern them, whereas
decisions on large purchases (which require a significant financial investment) are
typically made by husbands; as they are more likely to be primary earners in these
households. Therefore, we see a stark contrast between women’s decision-making
regarding child health and decisions regarding large purchases, where involvement in the
latter is significantly lower, at least in 1997.
It is evident from Figure 5.1 that women’s participation in decision-making is
quite high in Indonesia even in 1997 (84 and 77 percent respectively). Despite the high
levels in 1997, significant increases in these decision-making spheres still occurred from
1997 to 2007. For instance, women’s participation in decisions regarding child health
increased from 84 to 89 percent, while women’s participation in decisions regarding large
purchases increased from 77 to 86 percent from 1997 to 2007. Figure 5.2 presents the
difference in decision-making between Indonesia and Bangladesh in 2007. It is very clear
that Bangladesh lags behind in both decision-making spheres. Women’s participation in
decisions regarding child health is about 11-percentage points higher in Indonesian than
99
in Bangladesh, the difference is even greater on decisions regarding large purchases,
which is about 20-percentage points.
86%
89%
77%
84%
0% 20% 40% 60% 80% 100%
Large Purchases
Child Health
Notes: The percentage differences between 1997 and 2007 are
statistically signficant at 0.05 level.
Source: IFLS
Figure 5.1: Weighted Percentage of Women who
Participate in Decision Making in Indonesia from
1997 to 2007
1997 2007
86%
89%
66%
78%
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Large Purchases
Child Health
Notes: The percentage differences between Indonesia and
Bangladesh are statistically signficant at 0.05 level.
Source: IFLS and BDHS
Figure 5.2: Weighted Percentage of Women who
Participate in Decision Making - Indonesia and
Bangladesh in 2007
Bangladesh Indonesia
100
3.2 Explanatory Variables
The first part of the analysis explores the increase in decision-making in Indonesia from
1997 to 2007. For this purpose I use five characteristics that influence women’s status in
the household: woman’s age, education, share of household assets, household headship
and woman’s parents’ status relative their husbands’ parents’ at the time of marriage. To
isolate the effect of these variables I control for partner’s age and education, rural/urban
residence, religion and ethnicity. Ethnicity provides some additional insight as to how
various family systems (i.e. bilateral, patrilineal and matrilineal) and traditions affect
decision-making over time.
The second part of the analysis examines the differences in decision-making in
Bangladesh and Indonesia. The harmonization of the two data sets (IFLS and BDHS)
resulted in four comparable variables that influence decision-making (woman’s age,
education, work and household headship), and three control variables (partner’s
education, rural residence and religion).
4. Empirical Specification
This chapter extends previous research by examining the relative impact of improved
women’s socioeconomic status versus changing coefficients in explaining changes in
household decision-making from 1997 to 2007. For this purpose, I first specify a
multivariate logistic regression model for 1997 and 2007. (Please refer to chapter 3 for an
explanation of the method and the rationale for using it in this analysis). Then I use the
Oaxaca-Blinder decomposition method to identify the sources of improvements in
101
decision-making in Indonesia from 1997 to 2007. (Please refer to chapter 3 for an
explanation of the method and the rationale for using it in this analysis).
In the second part of this chapter I look at two social contexts—Bangladesh and
Indonesia in one time period. For this purpose, I follow a similar empirical strategy
described in the previous paragraph where I first run multivariate logistic regression
models to analyze the differences in coefficients, and then decompose the differences in
decision-making by using the Oaxaca-Blinder decomposition method. The following
equation represents the multivariate logistic regression model.
…… Equation 5.1
where y denotes a binary variable which equals to one if the respondent participates in
decision-making, else zero, WC—are women’s characteristics described in section 3.2;
and C—are control variables described in section 3.2.
The Oaxaca-Blinder decomposition takes the following format. Let D
i
denote a
dummy equal to one if the respondent participates in i = decisions regarding child health
or large expensive purchases. Then the model for a given dependent variable can be
written as:
D
i2007
- D
i1997
= [X
i2007
-X
i1997
]
B
i1997
+ [B
i2007
- B
i1997
]
X
i1997
+
[(X
i2007
-X
i1997
)B
i1997
*(B
i2007
-B
i1997
)X
i1997
)]…..................Equation 5.2
102
The first component [X
i2007
-X
i1997
]B
i1997
in Equation 5.2 can be interpreted as the
predicted increase in decision-making in 1997 if women’s characteristics were at the
2007 levels. The second component [B
i2007
- B
i1997
]X
i1997
can be interpreted as the
predicted increase in decision-making in 1997 if the coefficients of women’s
characteristics were same as 2007. The third component is an interaction term
[(X
i2007
-X
i1997
)B
i1997
*(B
i2007
-B
i1997
)X
i1997
)], which accounts for the residuals (see chapter
3 for a detailed discussion on the Oaxaca-Blinder decomposition method).
D
iI
-D
iB
= [X
iI
-X
iB
]
B
iB
+[B
iI
-B
iB
]X
iB
+ [(X
iI
-X
iB
)B
iB
*(B
iI
-B
iB
)X
iB
)]..……..Equation 5.3
Equation 5.3 can be interpreted in the same manner, where [X
iI
-X
iB
]B
iB
is the predicted
increase in decision-making in Bangladesh if it had the same levels of women’s
characteristics as Indonesia. The second component [B
iI
-B
iB
]X
iB
is the predicted increase
in decision-making in Bangladesh if the coefficient of women’s characteristics were same
as in Indonesia. The last component is the interaction term [(X
iI
-X
iB
)B
iB
*(B
iI
-B
iB
)X
iB
)]
(see chapter 3 for a detailed discussion on the Oaxaca-Blinder decomposition method).
5. Results
5.1 Comparison between 1997 and 2007 in Indonesia
5.1.1 Descriptive Statistics
Table 5.1 presents the descriptive statistics of the explanatory variables; the last column
shows the results of the two-sample t-test and test of proportions to indicate the
103
significant changes (if any) from 1997 to 2007. The results indicate that the age
distribution in 1997 is significantly different from 2007, there are more women aged 15-
24 years and aged 35-49 years in 2007 compared with 1997. Women with higher
education (grade 7 and above) increased from 52 to 63 percent. Further, women who own
25-50% of household assets increased from 33 to 53 percent, while women with no
household assets decreased from 13 to 5 percent. There was no change in the proportion
of women who are household heads from 1997 to 2007, and woman’s parents’ relative
position in the community compared with her parents-in-law at the time of marriage also
remained unchanged from 1997 to 2007. It is interesting to see however, that 60 percent
of women reported that their parents and their spouse’s parents had equal positions in the
community, and only about 11 percent reported a higher position in the community, and
only about 10 percent reported a lower position in the community. The focus group
discussion conducted by Frankenberg and Thomas (2003) revealed that most women in
Indonesia feel that it is important to marry someone from an equal family background to
avoid one person dominating the other; this socioeconomic homogamy is evident from
these descriptive statistics.
In this analysis, I include partner’s
13
age in years and education as controls. Both
variables significantly changed over time; partner’s age increased by 2.62 years and
women with partners with no education declined from 12 to 2 percent. Further, rural
residence decreased from 63 to 56 percent; this is an indication of rapid urbanization and
industrialization in Indonesia. The religion composition in the population significantly
13
I use the word partner, husband and spouse interchangeably to refer to woman’s husband as all women in
this sample are married
104
Table 5.1: Summary Statistics of the Explanatory Variables in Indonesia
Table 5.1: Summary Statistics of the E xplanatory Variables in Indonesia
1997 2007
Two Sample
N=5,490 N=9,823 sig
Woman's Characteristics
Woman's age
15-24 years 0.15 0.17 ***
25-34 years 0.40 0.34 *
35-49 years 0.45 0.49 ***
Woman's education level
No grade school 0.18 0.12 ***
Some grade school (1-6 grade) 0.30 0.25 ***
Completed grade school or higher (7+) 0.52 0.63 ***
Woman's share of household assets
0% of household assets 0.13 0.05 ***
1-25% of household assets 0.28 0.22 ***
25-50% of household assets 0.33 0.53 ***
50% or more of household assets 0.01 0.01
Household members do not own assets 0.26 0.19 ***
Woman's position in the household
Household head 0.01 0.01
Spouse of the household head 0.85 0.89
Other household member 0.15 0.10 *
Woman's parent's position in community
compared to her parents-in-law at marriage
1=Higher 0.005 0.007
2=High 0.11 0.11
3=Equal 0.60 0.57
4=Low 0.10 0.10
5=Lower 0.003 0.004
0.17 0.20
Controls
Partner's age in years 43.21 45.83 ***
Partner's education
No grade school 0.12 0.02 ***
Some grade school (1-6 grade) 0.29 0.24 ***
Completed grade school or higher (7+) 0.57 0.54 *
Spouse's education missing 0.02 0.20 ***
Rural Residence 0.63 0.56 ***
Religion
Islam 0.92 0.93 ***
Protestant 0.03 0.03
Catholicism 0.02 0.02
Hinduism 0.02 0.02
Other Religion 0.006 0.002 ***
Ethnicity
Javanese 0.51 0.52 *
Sundanese 0.19 0.14 ***
Balinese 0.02 0.02 *
Batak 0.03 0.03
Bugis 0.02 0.02
Minang 0.03 0.03
Banjar 0.02 0.03 **
Betawi 0.03 0.03
Other Sothern Sumatran 0.03 0.03
Other Ethnicity 0.13 0.15 ***
Notes: All means are weighted according to the survey design
***p<0.001, **p<0.01, *p<0.05
Source: IFLS
Parents not alive at marriage / unwilling /
don't know
Explanatory Variables
105
changed from 1997 to 2007, but the degree of change is not substantial. Finally, women’s
ethnicity also changed significantly over time – the Sundanese proportion fell from 19 to
14 percent whereas the Javanese proportion increased from 51 to 52 percent.
5.1.2 Bivariate Regression Results
Table 5.2 presents the bivariate relationship between the explanatory variables and the
two decision-making outcomes. By looking at the four columns, it is evident that
women’s characteristics are more strongly correlated with decisions regarding children’s
health than with decisions regarding large purchases in both years
The bivariate results consistently show that older women are more likely to
participate in decision-making than younger women. Women who have some education
are more likely to participate in household decision-making, compared with those who
have no education. As explained in the introductory chapters, education gives women
more decision-making power in several ways. For instance, education increases their
chances of obtaining outside employment, it teaches them important skills such a
communication and negotiation, and it also provides them with more knowledge and
information. The share of household assets variable is in the expected direction, where
women who own at least 1% of the household assets have higher participation in
decision-making than those who do not own any assets. The effect becomes stronger as
the woman’s share of assets increases. Further, women who are not the household head or
the spouse of the household head (i.e., other household member) are less likely to
participate in decision-making. This is more likely to happen in households where three-
106
Table 5.2: Bivariate Regression Results for 1997 and 2007 in Indonesia
Table 5.2: Bivariate Regression Results for 1997 and 2007 in Indonesia
Woman's Characteristics
Woman's age in categories
15-25 years (Reference)
25-34 years 1.52 *** 1.46 *** 0.39 ** 0.66 ***
35-49 years 1.37 *** 1.73 *** 0.40 ** 0.77 ***
Woman's education level
No education (Reference)
Some grade school (1-6 grade) 0.61 *** 0.16 0.25 * 0.12
Completed grade school or higher (7+) 0.35 *** 0.22 + 0.28 ** 0.33 **
Woman's share of household assets
0% of household assets (Reference)
1-25% of household assets 0.54 *** 0.67 *** 0.47 *** 0.66 ***
25-50% of household assets 1.16 *** 1.60 *** 1.00 *** 1.49 ***
50% or more of household assets 1.51 * 3.47 ** 1.59 ** 0.85 *
Household members do not own any assets 0.92 *** 1.42 *** 0.59 *** 1.14 ***
Household headship
Household head (Reference)
Spouse of the household head -0.16 -0.07 0.20 0.12
Other household member -1.31 * -1.46 *** -0.48 -1.05 **
Woman's parent's position at marriage
1=Higher (Reference)
2=High -0.79 0.48 -0.87 1.18 ***
3=Equal -0.89 0.30 -0.66 0.98 **
4=Low -0.98 0.17 -0.79 0.76 *
5=Lower -1.89 * 0.37 -0.53 0.45
Parents not alive at marriage / unwilling / DK -1.18 + 0.27 -0.75 1.02 **
Controls
Partner's age in years 0.00 0.01 ** 0.00 0.00
Partner's education
No education (Reference)
Some grade school (1-6 grade) 0.56 *** 0.07 0.37 ** -0.12
Completed grade school or higher (7+) 0.49 *** -0.11 0.41 *** 0.01
Spouse's education missing 0.39 -0.15 0.19 -0.04
Rural Residence -0.10 -0.09 0.00 -0.15 *
Religion
Islam (Reference)
Protestant 0.40 + 0.11 0.19 0.51 **
Catholicism 0.62 + -0.36 0.35 0.88 **
Hinduism 0.21 -0.24 -0.48 *** -0.07
Other Religion -0.07 1.12 -0.67 * 0.87
Ethnicity
Javanese (Reference)
Sundanese -0.19 + -0.07 -0.52 *** -0.17 +
Balinese 0.27 -0.14 -0.85 *** -0.04
Batak 0.56 * 0.20 -0.28 0.31 +
Bugis -0.19 -0.18 -0.92 *** 0.05
Minang 0.64 * -0.26 + 0.33 -0.08
Banjar -0.32 -0.10 -0.56 ** 0.36 +
Betawi -0.42 * -0.32 + -0.74 *** -0.31 *
Other Sothern Sumatran 0.20 0.22 -0.39 * 0.35 *
Other Ethnicity 0.19 0.24 * -0.30 * 0.33 **
Notes: Coefficients and robust standard errors are weighted according to the sample design
***p<0.001, **p<0.01, *p<0.05, +p<0.10
Source: IFLS
Explanatory Variables
1997 2007 1997 2007
Child Health Large Purchases
Bivariate Regression
107
generations co-reside, and older parents may make decisions on behalf of the younger
married couple. However, there is no significant difference in the decision making power
between women who are household heads and those that are the spouse of the household
head. A woman’s parent’s relative position has a strong effect on her participation in
decisions regarding large purchases in 2007. However there is no significant effect on
decisions regarding child health in either survey year, or on decisions regarding large
purchases in 1997. These bivariate results provide insight as to how women’s decision
making differs according to various characteristics, I turn now to the multivariate
analysis, which show the effects of women’s characteristics net of each other, and of
other factors.
5.1.3 Multivariate Regression Results
Tables 5.3 and 5.4 present the results of the multivariate regressions for decision-making
on child health and large purchases respectively. The first column for each model
provides the odd-ratios, the second column provides coefficients and the third column
provides robust standard errors. Women aged 25-34 years have about 2-3 times greater
odds of participating in decision-making regarding child health compared with women
aged 15-25 years. The effect is weaker for women aged 35-49 years in 1997 compared
with 2007. Woman’s education does not have a strong independent effect on decisions
regarding child health; even the joint significance test did not yield significant results (see
Appendix B2). This means the effect of education is completely explained by other
explanatory variables. A woman’s share of household assets has a positive effect on
108
decision-making regarding child health. For instance, compared with women who don’t
own any household assets, women who own 1-25% of household assets have 1.5 greater
odds of participating in decisions regarding child health in 1997. Similarly, compared
with women who don’t own any household assets, women who own 50% or more assets
have 7.9 greater odds of participating in decisions regarding child health in 2007. In
general, as the woman’s share of household assets increases, her participation in
decisions regarding child health also increases. Household headship has no effect on
women’s decision-making power in 1997 or in 2007. Woman’s parents’ status in the
community has a significant influence on decisions regarding child health in 1997.
Relative to women who had a “higher” position in the community at the time of marriage
compared with their husbands, women who had a “lower” position have negative odds of
participating in decision-making. However, this effect disappears in 2007.
Woman’s partner’s increasing age has a positive impact on the outcomes. This
means, keeping women’s age constant, women who have older partners are more likely
to participate in decisions than women with younger partners. The effect of partner’s
higher education has a positive impact in 1997, but the relationship becomes insignificant
in 2007. Rural residence does not have an independent effect on women’s participation in
decision-making. This is an interesting finding as we saw a strong negative relationship
between rural residence and decision-making in Bangladesh in the previous chapter. This
shows that depending on the social context, certain attributes do not predict women’s
status or decision-making. In a country like Indonesia this is especially true as women
traditionally had more authority in the household even before urbanization took place.
109
Table 5.3: Woman’s Participation in Decision Making Regarding Children’s Health in Indonesia
Table 5.3: Woman's Participation in Decision Making Regarding Children's Health in Indonesia
Odds
Ratio
Coeff
Odds
Ratio
Coeff Robust S.E.
Woman's Characteristics
Woman's age in categories
15-25 years (Reference)
25-34 years 2.56 0.941 *** (0.159) 3.21 1.167 *** (0.126)
35-49 years 1.41 0.340 + (0.206) 2.59 0.951 *** (0.207)
!
Woman's education level
No education (Reference)
Some grade school (1-6 grade) 1.51 0.414 * (0.185) 0.96 -0.037 (0.221)
Completed grade school or higher (7+) 1.30 0.263 (0.183) 1.32 0.278 (0.212)
Woman's share of household assets
0% of household assets (Reference)
1-25% of household assets 1.51 0.412 ** (0.150) 1.15 0.143 (0.152)
25-50% of household assets 2.53 0.930 *** (0.168) 2.06 0.720 *** (0.169)
50% or more of household assets 1.57 0.451 (0.807) 7.91 2.068 * (1.037)
Household members do not own any assets 1.71 0.536 ** (0.164) 1.87 0.629 ** (0.184)
Household headship
Household head (Reference)
Spouse of the household head 0.79 -0.238 (0.697) 0.75 -0.284 (1.018)
Other household member 0.40 -0.924 (0.709) 0.43 -0.856 (1.025)
Woman's parent's position at marriage
1=Higher (Reference)
2=High 0.34 -1.086 (0.703) 1.23 0.205 (0.502)
3=Equal 0.31 -1.184 + (0.690) 1.23 0.210 (0.488)
4=Low 0.28 -1.284 + (0.703) 1.21 0.190 (0.499)
5=Lower 0.06 -2.889 ** (0.869) 1.18 0.163 (0.733)
!
Parents not alive at marriage / unwilling / DK 0.24 -1.443 * (0.697) 0.88 -0.122 (0.504)
Controls
Partner's age in years 1.03 0.028 ** (0.009) 1.03 0.031 ** (0.010)
Partner's education
No education (Reference)
Some grade school (1-6 grade) 1.70 0.531 ** (0.195) 0.91 -0.098 (0.285)
Completed grade school or higher (7+) 1.79 0.581 ** (0.193) 0.76 -0.274 (0.271)
!
Spouse's education missing 2.10 0.741 + (0.447) 0.48 -0.726 * (0.364)
!
Rural Residence 1.01 0.010 (0.111) 1.09 0.083 (0.093)
Religion
Islam (Reference)
Protestant 0.75 -0.289 (0.326) 0.66 -0.417 (0.340)
Catholicism 3.01 1.104 (0.765) 0.72 -0.331 (0.364)
Hinduism 0.95 -0.056 (0.721) 0.62 -0.471 (0.342)
Other Religion 1.19 0.176 (0.729) - - -
Ethnicity
Javanese (Reference)
Sundanese 0.87 -0.145 (0.138) 0.87 -0.139 (0.138)
Balinese 1.71 0.536 (0.733) 1.63 0.488 (0.357)
Batak 1.81 0.595 + (0.339) 1.32 0.277 (0.348)
Bugis 1.20 0.179 (0.292) 1.23 0.204 (0.238)
Minang 2.47 0.905 ** (0.338) 0.60 -0.509 * (0.210)
!
Banjar 0.74 -0.300 (0.258) 1.06 0.060 (0.216)
Betawi 0.60 -0.503 * (0.248) 0.81 -0.208 (0.227)
Other Sothern Sumatran 0.94 -0.059 (0.280) 1.44 0.366 (0.235)
Other Ethnicity 1.45 0.373 * (0.172) 1.65 0.502 *** (0.142)
Constant 0.364 (1.068) -0.066 (1.213)
N Observations
Wald Chi-square
Pseudo R-square
Notes: Coefficients and robust standard errors are weighted according to the sample design
Source: IFLS
! = 1997 and 2007 coefficients significantly different from one another (p <= 0.05)
***p<0.001, **p<0.01, *p<0.05, +p<0.10
Logistic Regression
1997 2007
0.109 0.128
Robust S.E.
4,075 6,617
298 542
110
In 1997, the women in the matrilineal Minang ethnic group had 2.47 greater odds of
making decisions than women in the bilateral Javanese ethnic group. This provides some
evidence on the relatively higher status of women whose descent is traced through the
mother and maternal ancestors. However, this effect reverses in 2007; it could be due to
improvements in Javanese women’s education and employment compared with Minang
women in 2007.
Table 5.4 presents the multivariate regression results predicating women’s
participation in decisions regarding large purchases. In 1997, women’s age did not have
an independent effect on decisions regarding large purchases, whereas in 2007 it did.
Woman’s education does not have an independent effect on decisions regarding large
purchases; even the joint significance test did not yield significant results (see Appendix
B2). Woman’s share of household assets has a positive effect on decision-making
regarding large purchases in both 1997 and 2007. For instance, compared with women
who don’t own any household assets, women who own 25-50% assets have 1.87 greater
odds of participating in decisions regarding large purchases in 1997. Similarly, compared
with women who don’t own any household assets, women who own 25-50% or more
assets have 2.69 greater odds of participating in decisions regarding large purchases in
2007. Compared with those who have more assets, those with fewer assets have lower
decision-making power, and this effect grows stronger as the proportion of assets
increases. Similar to previous results on child health, household headship has no effect on
women’s decision-making power on large expensive purchases in 1997 or in 2007.
Woman’s parents’ status in the community has a significant influence on decisions
111
Table 5.4: Woman’s Participation in Decision Making Regarding Large Expensive
Purchases in Indonesia
Table 5.4: Woman's Participation in Decision Making Regarding Large E xpensive Purchases in Indonesia
Odds
Ratio
Coeff
Odds
Ratio
Coeff
Woman's Characteristics
Woman's age in categories
15-25 years (Reference)
25-34 years 1.05 0.050 (0.150) 1.38 0.326 ** (0.114)
35-49 years 0.91 -0.090 (0.189) 1.45 0.372 * (0.172)
Woman's education level
No education (Reference)
Some grade school (1-6 grade) 1.20 0.185 (0.148) 0.83 -0.190 (0.193)
Completed grade school or higher (7+) 1.29 0.254 + (0.153) 1.03 0.028 (0.189)
Woman's share of household assets
0% of household assets (Reference)
1-25% of household assets 1.34 0.291 * (0.134) 1.51 0.414 ** (0.140)
25-50% of household assets 1.87 0.624 *** (0.143) 2.69 0.988 *** (0.152)
50% or more of household assets 2.18 0.779 (0.674) 1.20 0.183 (0.481)
Household members do not own any assets 1.32 0.278 + (0.142) 2.80 1.029 *** (0.168)
!
Household headship
Household head (Reference)
Spouse of the household head 1.82 0.601 (0.453) 1.44 0.367 (0.600)
Other household member 0.95 -0.053 (0.471) 0.72 -0.327 (0.609)
Woman's parent's position at marriage
1=Higher (Reference)
2=High 0.38 -0.967 (0.716) 3.06 1.118 ** (0.371)
!
3=Equal 0.49 -0.710 (0.709) 2.72 0.999 ** (0.355)
!
4=Low 0.43 -0.835 (0.717) 1.98 0.685 + (0.365)
5=Lower 0.44 -0.813 (0.985) 1.60 0.471 (0.590)
Parents not alive at marriage / unwilling / DK 0.46 -0.766 (0.716) 2.54 0.930 * (0.369)
!
Controls
Partner's age in years 1.00 0.002 (0.007) 0.99 -0.006 (0.007)
Partner's education
No education (Reference)
Some grade school (1-6 grade) 1.28 0.248 (0.168) 1.08 0.080 (0.247)
Completed grade school or higher (7+) 1.38 0.325 + (0.169) 1.30 0.261 (0.239)
Spouse's education missing 1.23 0.209 (0.352) 1.27 0.239 (0.338)
Rural Residence 1.01 0.015 (0.091) 0.90 -0.104 (0.084)
Religion
Islam (Reference)
Protestant 1.14 0.134 (0.265) 1.22 0.203 (0.282)
Catholicism 1.21 0.192 (0.374) 2.57 0.944 * (0.474)
Hinduism 0.93 -0.075 (0.502) 1.07 0.070 (0.339)
Other Religion 0.42 -0.860 + (0.439) 0.87 -0.135 (0.883)
Ethnicity
Javanese (Reference)
Sundanese 0.60 -0.506 *** (0.117) 0.72 -0.329 ** (0.119)
Balinese 0.55 -0.603 (0.509) 0.96 -0.037 (0.349)
Batak 0.66 -0.421 + (0.249) 0.86 -0.151 (0.253)
Bugis 0.48 -0.737 *** (0.209) 1.16 0.149 (0.206)
!
Minang 2.14 0.762 ** (0.273) 0.84 -0.175 (0.193)
!
Banjar 0.48 -0.743 *** (0.206) 1.58 0.460 + (0.235)
!
Betawi 0.52 -0.651 ** (0.209) 0.60 -0.509 ** (0.186)
Other Sothern Sumatran 0.58 -0.548 * (0.223) 1.42 0.349 + (0.206)
!
Other Ethnicity 0.80 -0.224 (0.136) 1.41 0.343 ** (0.130)
!
Constant 0.865 (0.900) -0.292 (0.793)
N Observations
Wald Chi-square
Pseudo R-square
Notes: Coefficients and robust standard errors are weighted according to the sample design
Source: IFLS
! = 1997 and 2007 coefficients significantly different from one another (p <= 0.05)
***p<0.001, **p<0.01, *p<0.05, +p<0.10
Robust S.E.
Logistic Regression
1997 2007
0.040 0.063
Robust S.E.
4,075 6,629
154 296
112
regarding large purchases in 2007. Relative to women who had a “higher” position in the
community at the time of marriage compared with their husbands, women who had a
“high” position have greater odds of participating in decision-making regarding large
purchases in 2007, however there is no significant effect in 1997. Woman’s partner’s age
has no impact on decisions regarding large purchases, nor does partner’s education. Rural
residence does not have an independent effect on women’s participation in decision-
making. In 1997, compared with Javanese women, women from Sundanese, Bugis,
Banjar, Betawi and other Southern Sumatran ethnic groups are less likely to participate in
decisions regarding large purchases, whereas women in matrilineal Minang ethnic group
are more likely to participate in decisions regarding large purchases (odds ratio = 2.14).
This effect becomes weaker and insignificant in 2007.
5.1.4 Oaxaca-Blinder Decomposition Results
The next phase of the analysis focuses on the sources of change in women’s decision-
making from 1997 to 2007. Table 5.5 illustrates the summary of the Oaxaca-Blinder
decomposition. The first row presents the change in the outcome variable. Between 1997
and 2007, the proportion of women who participate in decisions regarding child health
increased by 5.1 percentage-points while women’s participation on decisions regarding
large purchases increased by 8.6 percentage points. If women’s characteristics in 1997
were at 2007 levels, there would have been a 1.0 (0.010) percentage-point predicted
increase (i.e. 20% increase) in decision-making regarding child health. Similarly, for
large purchases the predicted percentage-point increase would have been 1.9 (i.e., 0.019
113
= 22%). Overall, however, the majority of the predicted percentage-point increase in
decision-making is due to the changes in coefficients and intercepts (4.1 and 6.2 for child
health and large purchases respectively). This suggests that important unobservable
variables are subsumed in the intercept. Although it is easier to conclude that these
unobservable are changes in social norms, without empirical evidence I cannot confirm
this finding. Further these unobservable variables could represent omitted development
variables that were not specified in the regression models. Therefore it is best to refrain
from making any assumptions about the intercept.
Table 5.5: Oaxaca-Blinder Decomposition: Indonesia in 1997 and 2007
Table 5.5: Oaxaca-Blinder Decomposition: Indonesia in 1997 and 2007
% Change % Change
Change in outcome (Y
2007
-Y
1997
) 0.051*** 0.086***
Change due to means (X
2007
-X
1997
)B
1997
38% 0.020*** 30% 0.026***
Due to women's characteristics 20% 0.010* 22% 0.019***
Due to control variables 19% 0.010*** 8% 0.007*
Change due to coefficients (B
2007
-B
1997
)X
1997
79% 0.041*** 72% 0.062
Due to women's coefficient 220% 0.113 282% 0.241
Due to control variables' coefficients -251% -0.128* -49% -0.042
Due to intercept 110% 0.056 -160% -0.137
Overall interation (X
2007
-X
1997
)*(B
2007
-B
1997
) -18% -0.009 -2% -0.002
Interaction - women's mean and coefficient 2% 0.001 -1% -0.001
Interaction - control's mean and coefficient -20% -0.010** -1% -0.001
Note: ***p<0.001, **p<0.01, *p<0.05, +p<0.10
Source: IFLS
Child's Health Large Purchases
114
5.2 Comparison between Bangladesh and Indonesia in 2007
5.2.1 Descriptive Statistics
The second part of this chapter examines the differences in decision-making in
Bangladesh and Indonesia. Summary statistics are presented in table 5.6. It can be seen
that Indonesian women have a significantly different age distribution than Bangladeshi
women. The education variable indicates that Indonesian women are more educated than
Bangladeshi women; this is consistent with country-level data that show that Indonesia
has a superior secondary school enrollment level for girls compared with Bangladesh.
Further, Indonesian women are more likely to work, than Bangladeshi women (32 and 56
percent respectively). This difference can be explained in part because of the purdah
system that is practiced in Bangladesh, which restricts women’s freedom of movement,
thus denying opportunities to engage in gainful employment outside their homes. Lower
levels of education also contribute to this factor. In both countries, only 1 percent of
women claim to be the household head (only women who live with their spouses were
included in this analysis). Partner’s secondary or higher education is 54 percent in
Indonesia and only 36 percent in Bangladesh. This is consistent with country-level data,
and it is a direct result of the differences in industrialization and economic growth
between the two countries. Bangladesh is largely an agrarian society compared with
Indonesia; this is reflected in the proportion of women who live in rural areas (77% in
Bangladesh and 56% in Indonesia). As explained in the previous chapter, urbanization
and industrialization affect social norms towards women; therefore we should see a
negative impact of rural residence on household decision-making. Religion is another
115
variable that is often considered to be predictive of women’s status; because both
countries are predominantly Muslim there is only a small variation in the proportion of
women who are Muslim between Indonesia and Bangladesh (93 and 90 percent
respectively).
Table 5.6: Summary Statistics of the Explanatory Variables: Bangladesh and Indonesia in 2007
5.2.2 Bivariate Regression Results
Table 5.7 presents the bivariate relationship between the explanatory variables and the
two outcomes variables for Bangladesh and Indonesia. Older women have a positive
significant effect on decision-making in both countries compared with younger women.
Bangladesh Indonesia
Two Sample Test
N=9,021 N=9,822 sig
Woman's Characteristics
Woman's age in categories
15-25 years 0.33 0.17 ***
25-34 years 0.33 0.34 ***
35-49 years 0.34 0.49 ***
Woman's education level
No education 0.34 0.12 ***
Primary education (1-6 grade) 0.31 0.25 ***
Secondary or higher (7+ grade) 0.36 0.63 ***
Woman works 0.32 0.56 ***
Woman is the household head 0.01 0.01
Controls
Partner's education
No education 0.37 0.02 ***
Primary education (1-6 grade) 0.27 0.24 ***
Secondary or higher (7+ grade) 0.36 0.54 ***
Spouse's education missing 0.00 0.20 ***
Rural Residence 0.77 0.56 ***
Muslim 0.90 0.93
Notes: All means are weighted according to the survey design
***p<0.001, **p<0.01, *p<0.05
Source: IFLS & BDHS
Table 5.6: Summary Statistics of the Explanatory Variables: Bangladesh and
Indonesia in 2007
Explanatory Variables
116
On decisions regarding child health, women who have secondary or higher education are
more likely to participate in decision-making compared with women with no education in
Bangladesh (#=0.23**), we see a similar effect for decisions regarding large purchase in
Indonesia (#=0.33**). Woman’s work has a positive impact on decision-making for both
countries and for both decision-making outcomes, while woman’s household headship is
only significant in Bangladesh. Rural residence has a negative effect on decision-making
(except for child health in Indonesia). Finally, being a Muslim woman has a negative
effect on decisions regarding large purchases but not on child health. Now I turn to
multivariate analysis to investigate the net effect of these explanatory variables on
decisions regarding child health and large purchase.
Table 5.7: Bivariate Regression Results: Bangladesh and Indonesia, 2007
Table 5.7: Bivariate Regression Results: Bangladesh and Indonesia, 2007
Woman's Characteristics
Woman's age in categories
15-25 years
25-34 years 0.28 *** 1.46 *** 0.38 *** 0.66 ***
35-49 years 0.36 *** 1.73 *** 0.45 *** 0.77 ***
Woman's education level
No education
Primary education (1-6 grade) 0.02 0.16 -0.10 0.12
Secondary or higher (7+ grade) 0.23 ** 0.22 + -0.04 0.33 **
Woman works 0.44 *** 0.31 *** 0.38 *** 0.48 **
Woman is the household head 1.18 * 0.29 1.11 ** 0.05
Controls
Partner's education
No education
Primary education (1-6 grade) -0.07 0.07 -0.17 * -0.12
Secondary or higher (7+ grade) 0.20 ** -0.11 -0.10 0.01
Spouse's education missing -1.17 -0.15 -0.34 -0.04
Rural Residence -0.39 *** -0.09 -0.30 *** -0.15 *
Muslim 0.01 0.10 0.19 * -0.37 **
N observations 9,021 9,822 9,021 9,822
Notes: Coefficients and robust standard errors are weighted according to the sample design
Source: IFLS and BDHS
***p<0.001, **p<0.01, *p<0.05,
Explanatory Variables
Bivariate Regression
Child Health Large Purchases
Bangladesh Indonesia Bangladesh Indonesia
117
5.2.3 Multivariate Regression Results
Similar to the previous analyses, I first run multivariate regressions, and then decompose
the differences in decision-making variables using Oaxaca-Blinder decomposition
method. In table 5.8 and 5.9, I examine the socioeconomic and cultural differences
between Indonesia and Bangladesh, and their impact on women’s participation in
decision-making. In both countries, woman’s age has a positive impact on decision-
making; however the effect is significantly stronger in Indonesia than in Bangladesh.
Women with secondary or higher education have greater odds of decision-making
compared with women with no education in Bangladesh, similarly women who work and
who are household heads have greater odds of decision-making compared with women
who don’t work or who are not household heads. By contrast, Indonesian women’s
education, work and household headship have no independent effects on decision-making
regarding child health. Rural residence has a negative effect on decisions regarding child
health in Bangladesh, but not in Indonesia. However, rural residence has a negative effect
on decisions regarding large purchases in both countries. Overall the multivariate results
indicate that variables that strongly influence decision-making in Bangladesh do not have
a strong impact (or no impact at all) on decision-making in Indonesia.
118
Table 5.8: Woman’s Participation in Decision Making Regarding Children’s Health: Comparison
Between Indonesia and Bangladesh in 2007
Odds
Ratio
Coeff
Odds
Ratio
Coeff Robust S.E.
Woman's Characteristics
Woman's age in categories
15-25 years (Reference)
25-34 years 1.34 0.291 *** (0.077) 4.21 1.437 *** (0.094)
!
35-49 years 1.53 0.426 *** (0.081) 5.85 1.766 *** (0.105)
!
Woman's education level
No education (Reference)
Primary education (1-6 grade) 1.13 0.122 (0.080) 1.04 0.042 (0.198)
Secondary or higher (7+ grade) 1.44 0.362 *** (0.095) 1.42 0.348 (0.187)
Woman works 1.57 0.449 *** (0.071) 1.05 0.049 (0.085)
!
Woman is the household head 2.94 1.078 * (0.479) 1.64 0.492 (0.672)
Controls
Partner's education
No education (Reference)
Primary education (1-6 grade) 0.93 -0.072 (0.080) 0.95 -0.052 (0.285)
Secondary or higher (7+ grade) 1.03 0.034 (0.090) 0.69 -0.374 (0.272)
Spouse's education missing 0.36 -1.018 (0.742) 0.82 -0.193 (0.282)
Rural Residence 0.71 -0.342 *** (0.066) 0.99 -0.013 (0.081)
!
Muslim 1.04 0.041 (0.101) 1.23 0.208 (0.138)
Constant 0.889 *** (0.140) 0.702 * (0.346)
N Observations
Wald Chi-square
Pseudo R-square
Notes: Coefficients and robust standard errors are weighted according to the sample design
Source: IFLS & BDHS
! = 1997 and 2007 coefficients significantly different from one another (p <= 0.05)
Table 5.8: Woman's Participation in Decision Making Regarding Children's Health: Comparison
Between Indonesia and Bangaldesh in 2007
Logistic Regression
Bangladesh Indonesia
Robust S.E.
8,107 8,039
137 411
0.019 0.081
***p<0.001, **p<0.01, *p<0.05,
119
Table 5.9: Woman’s Participation in Decision Making Regarding Large Purchases: Comparison
Between Indonesia and Bangladesh in 2007
Odds
Ratio
Coeff
Odds
Ratio
Coeff Robust S.E.
Woman's Characteristics
Woman's age in categories
15-25 years (Reference)
25-34 years 1.44 0.365 *** (0.065) 1.80 0.588 *** (0.092)
!
35-49 years 1.60 0.472 *** (0.068) 2.02 0.701 *** (0.096)
Woman's education level
No education (Reference)
Primary education (1-6 grade) 1.06 0.059 (0.070) 0.87 -0.144 (0.176)
Secondary or higher (7+ grade) 1.27 0.242 ** (0.080) 1.08 0.079 (0.167)
Woman works 1.41 0.345 *** (0.059) 1.49 0.401 *** (0.076)
Woman is the household head 2.73 1.003 * (0.426) 0.91 -0.094 (0.411)
Controls
Partner's education
No education (Reference)
Primary education (1-6 grade) 0.87 -0.134 (0.070) 1.11 0.106 (0.240)
Secondary or higher (7+ grade) 0.83 -0.186 * (0.075) 1.20 0.180 (0.231)
Spouse's education missing 0.90 -0.107 (0.815) 1.32 0.278 (0.241)
Rural Residence 0.74 -0.298 *** (0.055) 0.86 -0.147 * (0.074)
Muslim 1.23 0.206 * (0.085) 0.81 -0.206 (0.130)
!
Constant 0.292 * (0.119) 1.178 *** (0.315)
N Observations
Wald Chi-square
Pseudo R-square
Notes: Coefficients and robust standard errors are weighted according to the sample design
Source: IFLS & BDHS
! = 1997 and 2007 coefficients significantly different from one another (p <= 0.05)
Table 5.9: Woman's Participation in Decision Making Regarding Large Purchases: Comparison
Between Indonesia and Bangaldesh in 2007
Logistic Regression
Bangladesh Indonesia
Robust S.E.
9,004 8,039
150 128
0.017 0.022
***p<0.001, **p<0.01, *p<0.05,
120
5.2.4 Oaxaca-Blinder Decomposition Results
Table 5.10 summarizes the decomposition results. The first row presents the differences
in the outcome variables between Bangladesh and Indonesia. On decisions regarding
child health there is a 12.3 percentage point difference between Indonesian women and
Bangladeshi women; the difference is even greater for decisions regarding large
purchases—20.8 percentage points. If Bangladeshi women had the same socioeconomic
characteristics as Indonesian women, then there would be a 4.8 percentage point (i.e.,
39%) increase on decisions regarding child health, and 5.2 percentage point (i.e., 25%)
increase on decisions regarding large purchases. But if we look at the predicted increase
due to coefficients, women’s decision-making on child health would increase by 9.3
percentage points (75%) if Bangladesh had the same coefficients and intercept as
Indonesia, similarly, women’s decision-making on large purchases would increase by
16.3 percentage points (79%). The main source of the difference in decisions on large
purchases originates from the difference in the intercept (0.191), which subsumes
differences in unobserved variables.
121
Table 5.10: Oaxaca-Blinder Decomposition: Bangladesh and Indonesia in 2007
Table 5.10: Oaxaca-Blinder Decomposition: Bangladesh and Indonesia in 2007
% Difference % Difference
Difference in outcome (Y
I
-Y
B
) 0.123*** 0.208***
Difference due to means (X
I
-X
B
)B
B
15% 0.019 27% 0.056
Due to women's characteristics 39% 0.048 *** 25% 0.052 ***
Due to control variables -23% -0.029 2% 0.004
Difference due to coefficients (B
I
-B
B
)X
B
75% 0.093*** 79% 0.163***
Due to women's coefficient 52% 0.063 ** -19% -0.039
Due to control variables' coefficients 35% 0.043 5% 0.011
Due to intercept -11% -0.013 92% 0.191 ***
Overall interation (X
I
-X
B
)*(B
I
-B
B
) 9% 0.011 -6% -0.011
Interaction - women's mean and coefficient -7% -0.009 -10% -0.021 *
Interaction - control's mean and coefficient 16% 0.020 5% 0.010
Note: ***p<0.001, **p<0.01, *p<0.05, +p<0.10
Source: IFLS & BDHS
Child's Health Large Purchases
122
6. Conclusion
In this chapter, I attempted to investigate sources of improvements in women’s decision-
making in Indonesia by looking at two decision-making variables: child health and large
purchases. In addition to women’s age and education, I used two women’s status
variables that are rarely seen in household surveys: woman’s share of household assets
and woman’s parents’ status in the community at the time of marriage.
Multivariate analysis showed that Indonesian women’s education does not have a
significant impact on decision-making. This may suggest that although education had a
significant independent effect in patriarchal and less-developed nations like Bangladesh,
it does not have an independent effect on a more gender egalitarian and relatively more
developed society like Indonesia. It also implies that certain attributes that are deemed to
be important predictors of women’s status are not relevant in certain contexts.
Women’s age, share of household assets and family status at the time of marriage
significantly influences decision-making. However, household headship has no effect on
both outcome variables. It can be surmised that in Indonesia because women traditionally
have a higher status, there is no distinction between male household heads and their
wives when making household decisions. Women are able to exert an equal amount of
power as if they were the household heads themselves. Further, it was found that women
from Minang matrilineal ethnic group are more likely to participate in decision-making
compared with bilateral and partilineal ethnic groups.
The Oaxaca-Blinder decomposition indicates that on average about three-quarters
of the increase in decision-making between 1997 and 2007 derives from the change in
123
coefficients and intercept. According to modernization theory this can be explained as
unobservable shifts in attitudes towards women from 1997 and 2007 due to economic
development. However, this cannot be proved empirically. Other unobserved variables
such as omitted development variables may well be the reason behind the large difference
in intercept.
The second part of the analysis examines the differences in decision-making
between Bangladesh and Indonesia. Overall, the multivariate results indicate that
variables that strongly influence decision-making in Bangladesh do not have a strong
impact (or no impact at all) on decision-making in Indonesia. The multivariate analysis
further suggests that the underlying cultural context in Indonesia and Bangladesh plays an
important role in women’s status regardless of their socioeconomic characteristics (i.e.,
population composition). Further investigation using Oaxaca-Blinder decomposition
confirms these findings.
The decomposition results imply that, the difference in women’s participation in
decision-making regarding large purchases between Bangladesh and Indonesia is mostly
due to the large difference in intercepts, whereas the difference in women’s participation
in decision-making regarding child health is mostly due to the stronger coefficients in
Indonesia compared with Bangladesh. What this qualitatively means is that, the
difference in women’s participation in decision-making between the two countries is only
partly due to the relatively higher socioeconomic status of Indonesian women compared
with Bangladeshi women. The majority of the difference is attributed to differences in the
strength of the relationships and unobservable variables. Although I cannot empirical
124
prove that the intercept subsumes unobservable variables that are attributable to cultural
norms and family systems, the anthropological literature that I discussed earlier makes a
strong argument on the impact of differential cultural norms on women’s status in
Bangladesh and Indonesia.
This chapter contributes to the current literature in three ways. First it employs
richer data on women’s status that are not available in traditional household surveys,
second, it attempts to quantify the sources of change in decision-making overtime in
Indonesia, and finally, it examines the role of the social context in household decision-
making by comparing two predominantly Muslim countries in Asia. In the next chapter I
analyze how changes to women’s socioeconomic characteristics at population-level
affects child nutrition in Bangladesh.
125
Chapter Six
Women’s Status and Child Nutrition in Bangladesh
1. Introduction
Since the 1990s researchers have noted massive improvements in women’s education,
employment, total fertility rate and median age at first marriage in Bangladesh.
According to human capital literature, these improvements should have a profound
impact on child nutrition. This chapter employs cross-sectional data from the 1999 and
2007 Bangladesh Demographic Health Surveys and uses multivariate regression and
Oaxaca-Blinder decomposition methods to test whether increases in women’s status, as
measured by improvements in socio-demographic characteristics, are associated with
gains in child nutrition. This chapter is organized as follows. Section 2 presents the
background. Section 3 describes the data. Section 4 discusses the empirical approach
used in this chapter. The results are presented in section 5, followed by the discussion in
section 6 and the conclusion in section 7.
2. Background
2.1 The Context
With over 140 million people squeezed into an area of 145,000 km
2
, Bangladesh is one of
the most densely populated countries in the world (UNFPA 2011). Until 1971,
Bangladesh was part of Pakistan (1947-1971) and before 1947, part of India. Like its
neighbors, Bangladesh has deeply rooted patriarchal ideals and relatively low women’s
status when compared with other less-developed countries. High levels of female
126
infanticide, domestic violence against women and low resource allocation for girls are
indicators of women’s relative low status (Chen, Huq and D’Souza 1981).
According to Islamic tenets, responsibility for one’s parents is in the hands of
adult children. In Bangladesh this responsibility largely falls onto sons (Kabir, Szebehely
and Tishelman 2000; Mahmood 1992; Rahman 1999). In fact, sons have been described
as the best risk insurance available to women (Arthur and McNicoll 1978; Cain, Khanam
and Nahar 1979). This difference in the relative importance of sons and daughters is
deeply rooted in the patrilineal family system where extended family is the predominant
system of organization. Marriage of sons and daughters is a highly planned part of a
family’s economic strategy; it is an arrangement between the couple’s parents (Arthur
and McNicoll 1978). Although daughters do useful household work, they do not bring in
outside wages like sons. Under such circumstances, families tend to marry off daughters
at an early age but hold sons as long as possible. This contributes to the large age
difference between brides and grooms (Arthur and McNicoll 1978).
After marriage, the woman moves in with her husband, his parents, his brothers
and their wives. Her own parents no longer have any claim on her labor. This practice of
village exogamy usually takes a woman some distance from her own parent’s home when
she marries (Frankenberg and Khun 2004). According to Cain (1991), Bangladesh has
both a patriarchal and gerontocratic (i.e. group governed by old people) family system,
where young women who move in with their husband’s families are usually at the bottom
of the family hierarchy with little or no control over household resources. Such family
systems provide the greatest scope of structural bias based on age and sex (Skinner
127
1997). In addition, early marriage of women enables extended families to further
dominate the young bride (Arthur and McNicoll 1978).
The low status of women in Bangladesh is a result of this rigid family system
coupled with strong traditions of purdah.
14
According to the tradition of purdah,
respectable women do not engage in trade or fieldwork or leave the family for other than
traditionally specified visits to relatives (Arthur and McNicoll 1978). This means they
cannot find employment outside their homes. Women work mostly on tasks that can be
done at home, such as processing harvested rice and producing handicrafts (Balk 1997;
Cain, Khanam and Nahar 1979; Zaman 1995). However, men are responsible for selling
these products; as a result women have little or no control over the profits gained by these
products (Frankenberg and Kuhn 2004). For these reasons, separation and divorce are
real threats to women, especially if they don’t have sons. Therefore, fulfilling duties to
one’s husband and in-laws take on a special importance, as does producing sons for
security in later years. As a result, women find themselves carrying out roles dictated by
the interests of husband, sons or in-laws (Arthur and McNicoll 1978).
Despite continuous international and domestic efforts to improve Bangladesh’s
economic and demographic prospects, it is still among the world’s least developed
nations. According to the United Nations Development Program’s Human Poverty Index
(2007), Bangladesh ranks 93
rd
poorest of 108 developing countries and its average per
capita income is $440 per year. However, current trends show that the poverty level in
Bangladesh has fallen over the past decade from 58 percent to 49 percent (UNFPA 2011).
14
A curtain or screen used to keep women separate from men.
128
The country has graduated from the status of “low economic potential” to an “emerging
market economy,” defying the gloomy predictions made by many in the mid-1970s.
Furthermore, in the past decade Bangladesh’s key demographic and health indicators
have improved. The total fertility rate (TFR) fell from 6.3 in 1975 to 3.3 in 1994; the
current TFR in Bangladesh is 2.36 births per woman (UNFPA 2011). Between 1994 and
2004, the contraceptive prevalence rate increased from 45 percent to 58 percent. Infant
mortality decreased from 87 deaths per 1,000 live births in 1994 to 65 deaths per 1,000
live births in 2004. Similarly, child mortality (under five years) decreased from 50 deaths
per 1,000 live births in 1994 to 24 deaths per 1,000 live births in 2004 (UNFPA 2011).
Data from the Demographic and Health Surveys indicate that child stunting declined
from 50 percent in 1999 to 44 percent in 2007. In addition, Bangladesh has achieved
gender parity in primary education and nearly removed the gender gap in secondary
education. Although female labor force participation is still low, it increased from 23.9
percent in 2000 to 29.2 percent in 2006 (UNFPA 2011). All of these improvements were
achieved despite low per capita income. Due to these demographic and socioeconomic
changes, Bangladesh makes an interesting research site to analyze trends in child
malnutrition and gender inequality.
2.2 Theory of Household Resource Allocation
In the past, the most common model of household resource allocation assumed that all
household members have identical preferences or that the preferences of one member
determine resource allocation – unitary model. However, such models came under
129
scrutiny because in reality each individual has their own preferences and resources. As a
result researchers started considering more general models (collective models), which
take individuals as the basic element and treat household decisions as the outcomes of
interactions and bargaining among the members (McElroy and Horney 1981; Manser and
Brown 1980).
Scholars who study development have had a long-standing interest in how a
woman's preference relative to that of her domestic partner affects behaviors and
outcomes related to household welfare. Since differential preferences do not necessarily
mean that the woman will be able to exercise her preferences, the status of a woman
relative to her partner plays a central role in household decisions. In fact, most of the
literature has focused especially on male-female equity in intra-household decision-
making power and allocation of resources, and on the economic and social benefit of
educating girls and women as a form of human capital investment. Women may derive
higher status from multiple sources, especially through education, employment and
assets. Settings of power such as customs and norms regarding marriage and family life
are also sources of social status.
Education has long been considered to be a source of power for women. Educated
women have an enhanced ability to process information that is transmitted through
newspapers, radio or television. They may have a better understanding of their
community and society, and are able to negotiate effectively with their husbands and
navigate social institutions. More importantly education increases the chances of gaining
130
outside employment for women, which means they will be able to support themselves in
the event of separation or divorce.
Lundberg and Pollak (1993) emphasized the different roles that men and women
play in the household and the implications of these “spheres of interest” (such as
expenditures on food or on child health) for models of household behavior. Who
influences each sphere has consequences for the overall welfare of the household. For
instance, if women control expenditures on food, they may buy nutritional food for the
children, which improves their growth, and makes them less susceptible to illness
(Schultz 1999).
2.3 The Determinants of Child Malnutrition
Clinically, malnutrition is characterized by inadequate intake of protein, energy, and
micronutrients; and frequent infections, gastrointestinal parasites and other childhood
diseases. Children are also malnourished if they are unable to utilize fully the food they
eat, for example, due to diarrhea or other illnesses (secondary malnutrition). Malnutrition
in all its forms increases the risk of disease and early death (WHO 2000). As a result of
malnutrition, children may become stunted i.e. low “height-for-age”, wasted i.e. low
“weight-for-height” or underweight i.e. low “weight-for-age” (Mishra and Retherford
2000). Stunting is a long-term indicator of a child’s nutritional status, which means that
the child did not received adequate nutrition or was subject to frequent infections and
diseases throughout his/her life. Stunted children are at a higher risk of early death. They
are also more likely to be stunted as adults and have low productivity compared with
131
adults who were not stunted during childhood (Strauss and Thomas 1998; Hoddinott et al.
2008). Therefore, stunting has serious consequences for the economic growth of a
country and the overall health of the population.
The South Asian region by far has the highest number and prevalence of
malnutrition in the world. It is home to half of all underweight children under five years
in the developing world. Sub-Saharan Africa, where roughly one child out of every three
is underweight, has the second highest rate. But ironically South Asia appears to be doing
better than Sub-Saharan Africa in terms of national per capita income, per capita food
supplies, and education levels. Ramalingaswami, Jonsson, and Rohde (1996) argue that
malnutrition in South Asia is due to extreme inequality between men and women, which
leads to widespread destitution. They explain that low birth weight is the best single
predictor of malnutrition; birth weights below 2,500 grams have been found to be very
closely associated with poor growth, not just in infancy, but also throughout childhood.
Approximately one third of all babies in India are born with low birth weight, in
Bangladesh the proportion is one half, and in sub-Saharan Africa the proportion is about
one sixth. Low birth weight indicates that the infant was malnourished in the womb
and/or that the mother was malnourished during her own infancy, childhood,
adolescence, and pregnancy. The proportion of babies born with low birth weight
therefore reflects the condition of women, and particularly their health and nutrition, not
only during pregnancy, but also during their childhood and young lives (Ramalingaswami
et al. 1996). During the pregnancy itself, the average woman should gain about 10 kilos
132
in weight. Most women in Africa come close to that figure, whereas most women in
South Asia gain little more than 5 kilos (WHO 1995).
Even if children are not born with low birth weight, poor feeding practices and
inadequate medical care may lead to stunting. Since women are the primary caregivers,
they play an important role in preventing these conditions. T.P. Schultz (1993; 1999)
formulated a human capital model that explains the determinants of children’s health. He
argued that at the household level, mother’s education has a positive effect on the health
of her child, whereas father’s education does not. The importance of maternal education
has been well documented in the child survival literature as well (Caldwell 1979; Schultz
1980; Mensch, Lentzner and Preston 1986). Schultz (1999) explains that education
increases the chances of gaining outside employment for a woman. This means she will
earn an income, which she can use at her discretion, thus giving her the financial power
to purchase nutritional food, medicine, and visits to the doctor for her children as well as
for herself.
A more direct mechanism by which education influences women’s status is
through the ability to process information regarding child care, nutrition, health, and
diseases (Rosenzweig and Schultz 1983; Kesarda, Billy, and West 1986). For example,
educated women are able to comprehend and follow instructions to give correct amounts
of medicine at the right time. Further, educated mothers understand the nutritional aspects
of certain types of food and select the best available types to feed their children. They are
also more likely to exclusively breastfeed their infants in the early months of life which
offers considerable protection against diseases (both because of breast milk’s inherent
133
immunological properties and because breastfeeding minimizes the chances of infection
through unclean water and contaminated foods). Educated mothers have access to more
information on proper parenting techniques, nutritional guidelines, healthy hygiene
practices and proper health care behavior (Thomas, Strauss and Henriques 1991). They
are able to understand how infections and diseases are borne and take measures to
prevent them (e.g. use of mosquito nets to prevent malaria and dengue, boiling and
straining water before consumption to prevent diarrheal disease, typhoid and cholera).
And, they are also more likely to know how to navigate social institutions such as schools
and health care facilities, and how to interact successfully with the staff (Casper and
Kitchen 2008). Therefore, women’s education, employment and overall status are the
main source of variation in child malnutrition in developing countries.
2.4 Previous Research
Many cross-sectional empirical studies have shown the impact of women’s status on
children’s health and nutritional status. Dyson and Moore (1983) found evidence that
women’s autonomy promotes child survival by comparing the northern states of India
(where kinship structure is very patriarchal), to the eastern and southern states. Caldwell
(1986) compared Muslim and non-Muslim populations in the world, where seclusion of
women is common. He showed that child mortality rates are higher in Muslim countries
than in non-Muslim countries. Using data from the 1986 Brazil Demographic and Health
Survey, Thomas, Strauss and Henriques (1990) showed that mother’s education has a
large and significant impact on child height (conditional on household income and after
134
controlling for intercommunity heterogeneity). They also show that almost all the impact
of maternal education is explained by access to information—reading papers, watching
television, and listening to the radio. In another study in Brazil, Thomas (1997) showed
that income accruing to women has a statistically significant and larger positive impact
on child nutrition’s status than income accruing to men. Smith and Haddad (2000) using
cross-country data from 63 developing countries during 1970-96, measured women’s
status as a ratio of female-to-male life expectancy. Their results showed that women’s
status has a negative effect on the percentage of children who are underweight. In a study
in Egypt, Kishor (2000) used a variety of measures of women’s “empowerment”
available in Egypt Demographic and Health Survey 1995-96 to show that women’s status
is positively associated with child survival. Smith et al. (2003) studied the relationship
between women’s status and children’s nutrition in 36 countries in three developing
regions: South Asia, Sub-Saharan Africa, and Latin America and the Caribbean. They use
two measures of women’s status, women’s decision-making power relative to that of
their male partners and the degree of equality between women and men in their
communities. Their results confirm that women’s status impacts child nutrition because
women with higher status have better nutritional status themselves, are better cared for,
and provide higher quality care to their children. In a more recent study, Bhagowaila and
her colleagues (2010) used the BDHS 2007 to examine the relationship between gender
inequality and nutrition using direct indicators of empowerment such as mobility,
decision-making power and attitudes towards verbal and physical abuse. Their results
indicate that a greater degree of women’s empowerment is associated with better long-
135
term nutritional status of children. Tarozzi and Mahajan (2007) analyzed the
improvements in child nutrition in India from 1992-93 to 1998-99. They found that
mother’s schooling had a positive impact on the reduction in child stunting. But they also
found that nutritional status improved substantially more for boys than for girls, and this
was especially true in rural areas and areas of North India where son preference is
widespread.
In this chapter, I exploit the significant change in child stunting and women’s
socioeconomic characteristics in Bangladesh to measure how much of improvements
child stunting can be explained by increases in women’s socioeconomic status. I also
investigate whether the determinants of child stunting have changed over time.
In addition to women’s status, other factors such as household wealth and
infrastructure play an important role in child stunting. For example, if the household
doesn’t have enough monetary resources to purchase nutritional food or to pay for
doctor’s visits, then even if the mother has high status in the household, a child’s
nutrition will be jeopardized. Similarly, not having access to infrastructure such as
electricity, a sanitary environment or health centers also penalizes a child’s nutrition
regardless of the mother’s status. Therefore, it is important to account for these household
and community level changes.
2.5 Conceptual Framework
The UNICEF’s conceptual framework on child malnutrition (UNICEF 1998) and the
subsequent extensions of that framework (Engle, Menon and Haddad 1997) discuss the
136
role of direct and indirect determinants of child malnutrition. Bhagowalia and her
colleagues (2010) further refined the UNICEF model by including variables such as
attitudes towards domestic violence, control over cash and decision-making.
Figure 6.1A and 6.1B illustrate a simple conceptual framework and its
corresponding operational framework. Figure 6.1A shows the linkage between culture,
economic development, women’s socioeconomic characteristics, and child stunting. The
extended UNICEF model highlights the importance of the physical and mental wellbeing
of women, support from family and community, and education all of which directly
impact their ability to care for children. The culture and traditions of the community such
as purdah, economic development and modernization determine the relative status of
women in a given time period. The status of women in turn can determine food intake,
preventive care and visits to the doctor by participating in household decision-making
(Bhagowalia et al., 2010).
Figure 6.1A: Conceptual Framework
Social
Context
Culture
Traditions
Development
Women’s
Characteristics
Household &
Community
Nutritional
Food
Diarrheal
diseases
Health center
Stunting
137
Figure 6.1B: Operational Framework
Figure 6.1B illustrates the operational framework used in this study. Education of the
woman (net of partner’s education) is often used as a measure of relative status between
husband and wife in the household (Frankenberg and Thomas 2003). Birth number of the
child, mother’s age at first marriage, and her exposure to media, are considered to be
pathways in which women’s education affects child nutrition, however these variables
also have an independent effect. For instance, if the birth number of a child is high, we
can assume that the mother had depleted energy/nutrition during gestation due to repeated
pregnancies, this may lead to low birth weight which in turn leads to childhood
malnutrition. Thus, birth number serves as a proxy for the total fertility rate of the
mother. Further, households with many children need to share limited resources such as
food, healthcare and parent’s attention. A child with many siblings will receive a smaller
Women’s
Characteristics
Education
Birth number (TFR)
Age at 1
st
marriage
Media exposure
Household &
Community
Household wealth
Electricity
Drinking water
Sewerage system
Stunting
138
share of these resources compared with a child with fewer siblings. Therefore birth
number is an important predictor of child stunting.
Further, in a setting such as Bangladesh, age confers authority and status, age at
first marriage, in particular, leads to greater authority because the age difference between
the partner and the woman is likely to be smaller if the woman marries at a later age.
Media exposure provides women with useful information about children’s health and the
ways in which they can prevent common diseases that cause malnutrition.
My operational model also accounts for community level characteristics that
were not included in Bhagowalia et al. (2010); source of drinking water and modern
sewerage systems in particular have a direct impact on the children’s health. Household
wealth is an indication of the family’s economic capability to purchase nutritional food or
pay for visits to the health center (if they have one the in the community).
2.6 WHO Child Growth Standards
Previous research on the variation in early childhood growth and on the effects of
experimental nutritional supplementation has shown that inadequate growth in young
children in less-developed countries is generally the consequence of infectious diseases
and of low nutrient in take (Martorell and Habicht 1986). These studies also indicate that
genetic variation among ethnic groups plays a relatively minor role in early childhood
growth patterns compared with factors related to diet and infection such as social class
and economic status (Habicht et al. 1974; Martorell and Habicht 1986). Ethnic
139
differentiation in growth potential appears to be more important during adolescence, than
during infancy and early childhood (Martorell and Habicht 1986)
In 1977, the National Center for Health Statistics (NCHS) developed a child
growth reference standard using a sample consisting of primarily white middle-class,
formula-fed infants from southwestern Ohio. This reference was widely used to compare
the nutritional status of populations and to assess the growth of individual children
throughout the world. In 2006, the World Health Organization (WHO) published child
growth standards for attained weight and height to replace the previously recommended
1977 NCHS child growth reference. These new standards are based on breastfed infants
and appropriately fed children of different ethnic origins raised in optimal conditions and
measured in a standardized way. Six countries were used to create this reference: Brazil,
Ghana, India, Norway, Oman and the United States.
The new WHO growth standards confirm earlier observations that the effect of
ethnic differences on the growth of infants and young children in populations are small
compared with the effects of the environment. Studies have shown that there may be
some ethnic differences among groups, just as there are genetic differences among
individuals, but for practical purposes they are not considered large enough to invalidate
the general use of the WHO growth standards population as a standard in all populations.
These new standards have been endorsed by international bodies such as the United
Nations Standing Committee on Nutrition, the International Union of Nutritional
Sciences and the International Pediatric Association. And, they have been adopted in
more than 90 countries (WHO and UNICEF 2009). Therefore, I use the WHO 2006
140
reference growth standard to calculate height-for-age z-scores for Bangladeshi children
under-5 years.
3. Data
The Bangladesh Demographic and Health Survey
15
is a nationally representative cross-
sectional data set collected under the auspices of the ICF (Inner City Fund) International
and United States Agency for International Development. It is suitable for examining
nationwide trends and patterns in child malnutrition and how they correlate with
women’s status. All eligible women in each selected household were surveyed which
includes women who were currently or ever married. The surveys are based on two-stage
sample designs. In the first stage, enumeration units or “clusters” were selected from
larger regional units within a country. Next, households were randomly selected within
clusters. Detailed information on women, their male partners (if they had one), and their
children under age five were gathered. My analytical sample comprises children less than
five years of age with full anthropometric data, resulting in sample sizes of 5,313 in 1999
and 5,328 in 2007. The corresponding number of mothers with children under age five is
4,335 in 1999 and 4,530 in 2007; therefore on average each mother has 1.2 children
under five.
15
The BDHS covers the entire population residing in private dwelling units in Bangladesh. The country is divided into
six administrative divisions: Barisal, Chittagong, Dhaka, Khulna, Rajashani and Sylthet. Each division is divided into
zilas and each zila into upazilas. Rural areas in an upazila are divided into union parishads (UPs), and UPs are further
divided into mouzas. Urban areas in an upazila are divided into wards, and wards are subdivided into mahallas. These
subdivisions are used as the Primary Sampling Units (PSUs) for the surveys and each enumeration area consists of
about 100 households, on average, and is equivalent to a mauza in rural areas and to a mohallah in urban areas.
141
3.1 Explanatory Variables
This study uses data on women’s education, age at first marriage, television viewership
(as a proxy for media exposure), and partner’s education. Mother’s television viewership
was recoded as ‘1’ if she watched television at least once a week, else ‘0’. Household
income and infrastructure are important determinants of child nutrition. But unfortunately
BDHSs do not include household income or consumption data.
16
However, the surveys
include detailed information on household assets such as chairs/tables, almirahs (i.e.
cupboard), radios, televisions, bicycles, etc. According to principal component factor
analysis the almirah, television and chair/table measure a common construct, however,
since television viewership was already included in the analysis and is highly correlated
with ownership of a television (!=0.889) I only use chair/tables and almirah as a measure
for household assets.
17
As part of the 1999 BDHS, a Service Provision Assessment (SPA)
survey collected information on the infrastructure and socioeconomic characteristics of
communities as well as information on the accessibility and availability of health and
family planning services. However, the SPA survey was not conducted in 2007.
Therefore, I use the household questionnaire in the BDHS to extract information on the
availability of electricity.
18
Additionally, I use accessibility to a good source of drinking
water and advanced sewerage systems, as proxies for community infrastructure. A good
source of drinking water includes pipes inside or outside the dwelling or a tube well, an
16
The recoded wealth indexes provided in the BDHS data sets are not comparable across surveys (Rutstein 2008).
17
Principal component factor analysis was performed separately for the 1999 and 2007 data sets. The eigen values and
factor loadings of chair/table and almirah indicated good internal consistency within the factor.
18
This may or may not be a community level variable as some wealthy households may use a generator to produce
electricity even if the village does not have a power line.
142
unsuitable source of drinking water includes open water sources such as a lake, pond or
unprotected well. An advanced sewerage system includes a septic tank/toilet, water
sealed/slab or flush toilets, as opposed to an open pit, bush or no facility.
3.2 Measuring Child Stunting
A child is classified as “stunted” if his height conditional on his age and gender z-score
(haz) is below -2 standard deviations from the median of the 2006 World Health
Organization international growth reference.
The normal curves in Figure 6.2 show the prevalence of stunting among
Bangladeshi children in 1999 and 2007. The solid line at x=0 indicates the median of the
well-nourished reference population and the dashed line at x=-2 indicates the threshold
for stunting (i.e., two standard deviations below the reference median). Children whose z-
scores fall to the left hand side of the dashed line are stunted. In 1999, 50 percent of
Bangladeshi children under age five were stunted; in 2007 stunting declined to 44
percent. Not only did the percentage of children who were stunted decrease, but so did
the percentage who were severely stunted, as noted by the excess height in the red curve
compared with the blue curve to the left of the dashed line.
143
Figure 6.2: Height-for-Age Z-Score Distribution of Children Under 5 Years – Normal Curves
In Figure 6.3, the Cumulative Distribution Functions (CDFs) of the two height-for-age
(haz) distributions are illustrated. It is clear that F
2007
(haz) ! F
1999
(haz) for all haz; the
CDF of 2007 is to the right of the CDF of 1999 in an ascending plot. This means the 2007
distribution of haz scores are superior to 1999 haz scores because for any cumulative
probability value, 2007 yields a higher haz score. This is known as “stochastic
dominance” which indicates the superiority of one distribution over another. Further, the
Kolmogorov-Smirov test for equality of distribution functions also indicates that the 1999
and 2007 height-for-age z-scores are not equal at 0.001 confidence level (D=0.0686, p-
value=0.000). Therefore these results show that not only has the proportion of children
who are stunted declined over time, but also the overall distribution of child haz scores
have improved over time in Bangladesh.
0
0
0 .1
.1
.1 .2
.2
.2 .3
.3
.3 Density
Density
Density -6
-6
-6 -5
-5
-5 -4
-4
-4 -3
-3
-3 -2
-2
-2 -1
-1
-1 0
0
0 1
1
1 2
2
2 3
3
3 4
4
4 5
5
5 6
6
6 Z-Scores
Z-Scores
Z-Scores y1999
y1999
y1999 y2007
y2007
y2007 Two standard deviations below the median
Two standard deviations below the median
Two standard deviations below the median Median of the well nourished population
Median of the well nourished population
Median of the well nourished population Figure 6.2B Height-for-Age Z-Score Distribution of Children Under 5 Years - Normal Curve
Figure 6.2B Height-for-Age Z-Score Distribution of Children Under 5 Years - Normal Curve
Figure 6.2B Height-for-Age Z-Score Distribution of Children Under 5 Years - Normal Curve
144
Figure 6.3: Cumulative Distribution Functions of Height-for-Age Z-scores in 1999 and 2007
4. Empirical Strategy
For the multivariate analysis, I specify a multivariate logistic regression model:
u HCC
2
b W
1
b
0
b y + + + = C
……………………..Equation 6.1
where y denotes a binary variable which equals one if the child has a z-score less than -2
(i.e., stunted), else zero. WC—are women’s characteristics including education, partner’s
education, birth number of the child, age at first marriage and television viewership; and
HCC— are household assets, availability of electricity, access to a good source of
drinking water and access to an advanced sewerage system, and u is the error term.
0
0
0 .2
.2
.2 .4
.4
.4 .6
.6
.6 .8
.8
.8 1
1
1 Cumulative Probability
Cumulative Probability
Cumulative Probability -6
-6
-6 -5
-5
-5 -4
-4
-4 -3
-3
-3 -2
-2
-2 -1
-1
-1 0
0
0 1
1
1 2
2
2 3
3
3 4
4
4 5
5
5 6
6
6 Height-for-Age Z-score
Height-for-Age Z-score
Height-for-Age Z-score c.d.f. of 1999
c.d.f. of 1999
c.d.f. of 1999 c.d.f. of 2007
c.d.f. of 2007
c.d.f. of 2007 Figure 6.3 Cumulative Distribution Functions of Height-for-Age Z-scores in 1999 and 2007
Figure 6.3 Cumulative Distribution Functions of Height-for-Age Z-scores in 1999 and 2007
Figure 6.3 Cumulative Distribution Functions of Height-for-Age Z-scores in 1999 and 2007
145
This study investigates the changes in women’s characteristics at the macro level,
and compares two groups: one group that lived in an era where women had less
education, fewer employment opportunities and lower access to credit and social capital
(i.e. 1999 survey); and another group that has enjoyed the fruits of economic
development, including gender parity in education, establishment of garment factories
with female employees, more self-employment opportunities through microfinance, and
fewer children to care and provide for. Thus, it is suitable to use Oaxaca-Blinder
19
decomposition (Oaxaca 1973; Blinder 1973) to compare these two groups. The Oaxaca-
Blinder decomposition is represented in equation 6.2 where D denote a dummy variable
equal to one if the child is stunted, X denotes the mean of the explanatory variable(s), and
B represents the coefficient:
D
2007
- D
1999
= [X
2007
-X
1999
]
B
1999
+ [B
2007
- B
1999
]
X
1999
+ $X$B………Equation 6.2
The first component [X
2007
-X
1999
]B
1999
in Equation 6.2 can be interpreted as the
predicted reduction in stunting in 1999 if women’s characteristics were at 2007 levels
(i.e., contribution of the mean differences in the predictors between 1999 and 2007). The
second component [B
2007
- B
1999
]X
1999
can be interpreted as the predicted reduction in
stunting in 1999 if the coefficients were same as 2007 (i.e., the contribution of the
difference in the coefficients, including differences in the intercept).
20
The third
19
See chapter 3 for a detailed discussion on the Oaxaca-Blinder decomposition method.
20
This component subsumes the effects of group differences in unobserved explanatory variables (Jann 2008).
146
component $X$B is an interaction term [(X
i2007
-X
i1999
)B
i1999
*(B
i2007
-B
i1999
)X
i1999
)],
which accounts for the residuals.
Note that, some explanatory variables (e.g. household assets, age at first marriage)
may be jointly determined by education level of the woman and her husband. Therefore,
one should be very cautious in interpreting the results in a causal way. Most of the
included predictors are, in fact, likely to be correlated with unobserved heterogeneity in
preferences, cultural norms, or other location specific characteristics that may also have a
direct impact on the dependent variables. For these reasons, the interest of these results
lies more in their descriptive content that in their causal meaning.
5. Results
5.1 Descriptive Statistics
Table 6.1 reports the summary statistics of the explanatory variables for each survey year;
the results of the two sample tests indicate the significance of the change over time. All
variables indicate significant improvements from 1999 to 2007. The proportion of
mothers with primary and secondary education increased from 29 to 32 percent and 21 to
35 percent, respectively. Birth number of the child declined from 2.93 to 2.57; this is an
indication of mother’s lower fertility rate as well. Age at first marriage rose from 14.99 to
15.41 years. The household asset scale increased from 0.84 to 1.14. Similar patterns can
be seen for access to electricity and sewerage systems, with access to good source of
drinking water becoming nearly universal in 2007. The dependent variable (haz) as
discussed in section 3.2 declined from 50 to 44 percent over this time period.
147
Table 6.1: Summary Statistics of the Explanatory Variables: Child Stunting
5.2 Regression Results
The results of the multivariate logistic regression for stunting are presented in Table 6.2.
The first and third columns (i.e. Model 1 and 3) do not include birth number, age at first
marriage and television viewership because these variables are correlated with education
levels. As expected, parent’s education remains significant in 1999 and 2007 even after
controlling for household and community characteristics (Model 1 and 3). Controlling for
partner’s education, mother’s education can be interpreted as relative education (or as a
measure of relative power). Since most men are better educated than their wives, when
Table 6.1: Summary Statistics of the Explanatory Variables: Child Stunting
2007
N=5,313 N=5,328
Mother's Characteristics (age 15-49 years)
Mother's education level
No Education 0.46 0.27 ***
Primary Education 0.29 0.32 *
Secondary Education 0.21 0.35 ***
Higher Education 0.04 0.06 ***
Partner's education level
No Education 0.43 0.35 ***
Primary Education 0.25 0.28 ***
Secondary Education 0.22 0.26 ***
Higher Education 0.10 0.11 +
Birth number of the child 2.93 2.57 ***
Mother's age at first marriage 14.99 15.41 ***
Mother watches TV at least once a week % 0.31 0.45 ***
Household & Community Characteristics
Household furniture as a proxy for assets (scale 0-2) 0.84 1.14 ***
Household has electricity % 0.29 0.45 ***
Access to a good source of drinking water % 0.71 0.97 ***
Advanced sewerage system % 0.09 0.23 ***
Notes: All parameters are weighted according to the survey design
Source: BDHS
1999
Explanatory Variables
***p<0.001, **p<0.01, *p<0.05, +p<0.10
Two Sample
Test
148
holding men’s education constant, higher levels of education among women imply a
reduction in the gap (Frankenberg and Thomas 2003). However, in model 4 the effect of
mother’s education disappears, while father’s education remains significant. This implies
that the effect of mother’s education is completely explained by child’s birth number, age
at first marriage and television viewership in 2007. These variables are considered to be
pathways in which mother’s education affects child stunting, and they seem to matter
more in 2007 than in 1999.
The regression results of the household and community variables are less
conclusive. In 1999, availability of electricity, a good source of drinking water and an
advanced sewerage system had a negative impact on child stunting, however, in 2007, the
drinking water and sewerage variables are not significant. This is because in 2007, the
availability of a good source of drinking water was nearly universal (97%), thus
providing little variation for estimation. However, the non-significance of the sewerage
variable is difficult to explain, it could be due its modest correlation ($ = 0.49***) with
the availability of electricity in 2007. The household furniture variable (a proxy for
household assets) does not have a significant effect on child stunting in 1999, although it
has a weak but significant effect in 2007. Although, some coefficients seem to differ in
1999 compared with 2007 (e.g., birth number of the child), the pooled sample chi-square
test indicated that overall, the coefficients did not change significantly over-time. This
implies that, the reduction in stunting from 1999 to 2007 is mainly due to the changes in
the means of the variables, and not due to the change in the relationship between the
variables and stunting.
149
Table 6.2: Multivariate Logistic Regressions: Predicting Child Stunting in Bangladesh, 1999
and 2007
Table 6.2: Multivariate Logistic Regressions: Predicting Child Stunting in Bangladesh, 1999 and 2007
Mother's Characteristics (age 15-49 years)
Mother's education level
No education (Reference)
Primary Education -0.139 + -0.137 + 0.022 0.084
(0.076) (0.078) (0.090) (0.093)
Secondary Education -0.380 *** -0.346 ** -0.252 * -0.145
(0.102) (0.107) (0.103) (0.112)
Higher Education -0.797 ** -0.672 ** -0.404 * -0.307
(0.233) (0.241) (0.204) (0.218)
Partner's education level
No education (Reference)
Primary Education -0.140 + -0.141 + -0.079 -0.074
(0.080) (0.080) (0.089) (0.089)
Secondary Education -0.299 ** -0.285 ** -0.281 ** -0.275 **
(0.094) (0.094) (0.101) (0.101)
Higher Education -0.685 *** -0.655 *** -0.756 *** -0.766 ***
(0.156) (0.156) (0.167) (0.167)
Birth number of the child -0.011 0.054 *
!
(0.016) (0.022)
Mother's age at first marriage -0.025 * 0.008
(0.013) (0.015)
Mother watches tv at least once a week -0.141 + -0.071
(0.079) (0.082)
Household & Community Characteristics
Household furniture as -0.059 -0.055 -0.096 + -0.101 *
a proxy for assets (scale 0-2)
(0.049) (0.049) (0.051) (0.051)
Household has electricity -0.238 ** -0.175 * -0.277 ** -0.243 **
(0.077) (0.082) (0.081) (0.087)
Access to a good source of drinking water -0.152 * -0.158 * -0.053 -0.044
(0.067) (0.068) (0.184) (0.184)
Advanced sewerage system -0.287 * -0.264 * -0.104 -0.103
(0.123) (0.124) (0.091) (0.092)
Constant 0.554 *** 0.967 *** 0.334 + 0.027
N Observations
Wald Chi-square
Pseudo R-square
Pooled Sample F-test (Model 2 & 4): 13.45 (p value =0.4138)
Joint Tests of Significance: Model 2 Model 4
Mother's Education 12.98** 7.08+
Partner's Education 19.22*** 23.08***
Notes: All coefficients are weighted according to the survey design; roubst standard errors are presented in parentheses
Source: BDHS
Logistic Regression Coefficients
1999 2007
Model 1 Model 2 Model 3 Model 4
5,304 5,323 5,323
252.58 258.81 179.67 190.42
5,304
0.040 0.042 0.036 0.038
***p<0.001, **p<0.01, *p<0.05, +p<0.10
! = 1999 and 2007 coefficients significantly different from one another (p <= 0.05)
150
In a separate regression (not shown) I explored whether child’s gender has any impact on
the relationship between parent’s characteristics and stunting. The results indicated that
the effect of having a father who has primary education has a negative impact on boy’s
stunting, however, there is no significant effect for girls. Similarly, mothers who watch
television at least once a week have a negative impact on boy’s stunting, but not on girls.
These results are congruent with earlier studies on son preference in Bangladesh. I
explore this further in the next section.
5.3 Oaxaca-Blinder Decomposition
In Table 6.3, I present the decompositions for all children under-5 years in Bangladesh.
The proportion of stunted children decreased from 49.9% to 44.0% from 1999 to 2007;
this is a 5.9 percentage point reduction (-0.059). Overall, about half (in bold 48%; -0.029)
of the predicted reduction in stunting can be accounted for by the improvements in
parent’s education, birth number of the child, age at first marriage and television
viewership. The other half (in bold 48%; -0.028) can be attributed to the improvements in
household wealth and community infrastructure from 1999 to 2007. Mother’s education
accounts for the largest predicted reduction in child stunting (-0.016) followed by access
to a good source of drinking water (-0.010). The main finding in this analysis is that,
improvements in women’s status are responsible for about half of the predicted reduction
in child stunting.
21
21
The second column labeled as [B
2007
- B
1999
]X
1999
does not have a significant cumulative effect on child stunting.
However, if you look at the individual coefficients, you can see that the change in size of the coefficients in 2007
positively contributes to child stunting (0.197), but this is counteracted by the negative effect of the constant (-0.224).
The interaction term $X$B (i.e., residual) does not contribute to the overall reduction of stunting.
151
Table 6.3: Oaxaca-Blinder Decomposition: Child Stunting – All Children
Gender differences in child stunting are explored in Table 6.4A and 6.4B. Overall,
improvements in mothers’ characteristics from 1999 to 2007, account for 56% (-0.034) of
the predicted reduction in stunting for boys. But for girls, improvements in mother’s
characteristics account for only 39% (-0.022). Further, improvements in maternal
education have a significant impact on both boys and girls nutritional status; but
improvements in paternal education only impact boys’ nutrition. The majority of the
reduction in stunting for girls derived from improvements in household wealth and
community infrastructure (59%; -0.034), which are generally available for all children in
the household regardless of gender. This implies that although Bangladesh has progressed
Table 6.3 . Oaxaca-Blinder Decomposition: Child Stunting -- All Children
Mother's Characteristics (age 15-49 years) 48% -0.029 *** -331% 0.197 ** -18% 0.011
Mother's education -0.016 *** 0.029 0.010
Partner's education -0.006 ** 0.003 0.000
Birth number of the child 0.001 0.046 * -0.006 *
Mother's age at first marriage -0.002 * 0.115 0.003
Mother watches tv at least once a week -0.005 0.005 0.002
Household & Community Characteristics 48% -0.028 *** -13% 0.008 -10% 0.006
Household furniture as a proxy for assets -0.004 -0.009 -0.003
Household has electricity -0.007 * -0.005 -0.002
Access to a good source of drinking water -0.010 * 0.019 0.007
Advanced sewerage system -0.008 * 0.003 0.005
Constant 376% -0.224 **
Change in stunting: -0.059 ***
Due to change in all means (X2007-X1999)B1999 -0.057 ***
Due to change in all coefficients (B2007-B1999)X1999 -0.019
Due to change in all interactions
!
X
!"
-0.017
Notes: All coefficients are weighted according to the survey design
Source: BDHS
(X2007-X1999)B1999 (B2007-B1999)X1999
!
X
!
B
***p<0.001, **p<0.01, *p<0.05, +p<0.10
152
in terms of women’s education, employment, access to credit and other demographic
indicators; parents still favor sons over daughters.
6. Discussion
As with any analysis based on cross-sectional survey data, this study has some
limitations. The explanatory variables leave much of the actual change in the dependent
variable unexplained as indicated by the low Pseudo R
2
of the multivariate logistic
regression model. I suspect that this may be due in part to a lack of information on village
level variables such as the number of hospitals, health clinics, pharmacies, midwives,
etc., factors which directly impact child health and nutrition. As the goal of this chapter
was to assess whether gains in women’s status account for improvements in children’s
nutrition, I acknowledge this shortcoming but do not think it detracts from the findings
shown here.
In addition to stunting, I attempted to investigate other indictors of child nutrition,
including wasting i.e., weight-for-height (a short term indicator of nutrition) and birth
weight as additional dependent variables. But wasting in Bangladesh increased by five
percentage points from 1999 to 2007 and none of the variables in my model were able to
explain this increase. This finding seems counterintuitive. However, in 2007, Bangladesh
and two other South Asian countries experienced heavy monsoonal rains, which led to
what is being described as the worst flood in living memory (UNICEF 2007). People
suffered from diarrhea and waterborne diseases, crops and livestock were destroyed
causing a food shortage and starvation. I assume that the increase in wasting might be due
153
Table 6.4A: Oaxaca-Blinder Decomposition: Child Stunting -- Boys
Table 6.4B: Oaxaca-Blinder Decomposition: Child Stunting -- Girls
Mother's Characteristics (age 15-49 years) 56% -0.034 *** -254% 0.156 -13% 0.008
Mother's education -0.020 ** 0.033 0.011
Partner's education -0.009 ** 0.018 0.003
Birth number of the child 0.001 0.037 -0.004
Mother's age at first marriage -0.001 0.076 0.002
Mother watches tv at least once a week -0.005 -0.008 -0.004
-0.004
Household & Community Characteristics 37% -0.023 ** 55% -0.034 14% -0.009
Household furniture as a proxy for assets -0.003 -0.018 -0.006
Household has electricity 0.000 -0.014 -0.007
Access to a good source of drinking water -0.013 * -0.007 -0.002
Advanced sewerage system -0.007 0.004 0.007
Constant 206% -0.127
Change in stunting: -0.062 ***
Due to change in all means (X 2007-X 1999)B 1999 -0.057 ***
Due to change in all coefficients (B 2007-B 1999)X 1999 -0.004
Due to change in all interactions !X!" 0.000
(X 2007-X 1999)B 1999 (B 2007-B 1999)X 1999 !X!B
Mother's Characteristics (age 15-49 years) 39% -0.022 ** -387% 0.222 * -22% 0.013
Mother's education -0.013 * 0.023 0.009
Partner's education -0.002 -0.013 -0.002
Birth number of the child 0.001 0.053 -0.007
Mother's age at first marriage -0.003 * 0.142 0.004
Mother watches tv at least once a week -0.004 0.018 0.009
Household & Community Characteristics 59% -0.034 *** -82% 0.047 -36% 0.020
Household furniture as a proxy for assets -0.005 -0.001 0.000
Household has electricity -0.013 ** 0.004 0.002
Access to a good source of drinking water -0.006 0.042 0.016
Advanced sewerage system -0.009 0.002 0.003
Constant 530% -0.304 *
Change in stunting: -0.057 ***
Due to change in all means (X 2007-X 1999)B 1999 -0.056 ***
Due to change in all coefficients (B 2007-B 1999)X 1999 -0.035
Due to change in all interactions !X!" 0.033
Notes: All coefficients are weighted according to the survey design
Source: BDHS
***p<0.001, **p<0.01, *p<0.05, +p<0.10
(X 2007-X 1999)B 1999 (B 2007-B 1999)X 1999 !X!B
154
to this natural disaster because child’s weight can temporarily fluctuate due to illness or
famine and wasting is a measure of short-term malnutrition. Birth weight, which is the
best single predictor of malnutrition, is closely associated with poor growth. Birth weight
indicates the condition of a woman during pregnancy and the quality of her life before the
pregnancy. However, the BDHS data sets do not include information on the exact birth
weight of the child.
7. Conclusion
This chapter attempted to analyze the sources of improvement in child nutrition in
Bangladesh. I use child stunting, i.e. height conditional on age z scores (haz) as my
measure of child nutrition. Although malnutrition rates are generally high in Bangladesh,
I found significant reduction in stunting from 1999 to 2007. To explore the sources of this
reduction in stunting, I first identified the predictors of stunting that have changed over
time. Then I use Oaxaca-Blinder decomposition to analyze the key variables that have
contributed to the reduction in stunting from 1999 to 2007.
The results indicated that women’s education, their partner’s education, age at
first marriage, birth number of the child, and television viewership improved significantly
from 1999 to 2007 in Bangladesh. Additionally, indicators of economic development
such as household assets, and availability of electricity, good sources of drinking water
and advanced sewerage systems have also significantly improved in the same period.
Multivariate logistic regression showed that mother’s education, father’s education,
household assets and community infrastructure have independent effects on child
155
stunting. Birth number of the child, mother’s age at first marriage and television
viewership are pathways in which mother’s education affects child nutrition. Therefore,
when these variables are added to the logistic models, the effect of the mother’s
education diminished in 1999 and completely disappeared in 2007.
The results from the Oaxaca-Blinder decomposition revealed that almost half of
the reduction in stunting is due to changes in women’s status. The other half is due to
changes in household assets and community infrastructure. In a separate analysis I
explored the differences in the decomposition results by child’s gender. For boys,
improvements in parent’s education accounted for a large portion of the reduction in
stunting, but for girls the effect was smaller. In fact, the majority of the reduction in
stunting for girls comes from improvements in household assets and community
infrastructure, which are usually available for all children regardless of their gender.
These results confirm two things: improvements in women’s education (net of partner’s
education), age at first marriage, fertility (thus birth number of the child) and access to
information through television viewership is partly responsible for the overall
improvement of child nutrition. However, because parents still favor sons over daughters
in Bangladesh, the effect of parent’s education is weaker for girls than for boys.
From a welfare policy point of view, these results are encouraging, as
empowering women through mandatory school enrollment, microcredit programs and
distribution of contraceptives seem to have benefited the nutritional status of children,
which will ultimately impact the economic growth of the country. However, there is still
156
work to be done in terms of improving norms and attitudes towards female children in
Bangladesh.
157
Chapter Seven
Discussion and Conclusion
1. Background
“When women are healthy, educated and free to take the opportunities life affords them,
children thrive and countries flourish, reaping a double dividend for women and
children” (UNICEF 2007). The goal of my dissertation was to test several hypotheses that
are embedded in the above statement.
Past research studies have shown that when women are on a more equal footing
with men, they are able to influence household decision-making and child nutrition.
Women’s decision-making power varies by individual-level characteristics such as
woman’s age, education, employment, income and assets, and by community-level
characteristics such as social norms, religion, public policies and infrastructure.
International aid organizations such as the World Bank and local governments in
developing countries have taken important steps to improve women’s socioeconomic
characteristics as a pathway to economic development. As a result, there have been
massive improvements in girls’ school enrollment, women’s participation in microcredit
programs (e.g., Grameen Banks) and women’s employment opportunities.
This dissertation explores the change in women’s socioeconomic characteristics
over time and how the relationship between women’s characteristics and decision-making
has evolved over time. I also use the Oaxaca-Blinder decomposition method to quantify
the sources of improvements in women’s participation in decision making and child
nutrition. This method allows me to investigate whether improvements are attributed to
158
changes in means, or changes in coefficients. This is one of the first studies in the
development literature that tracks improvements in women’s status over time.
I analyze two predominantly Muslim countries in South and South-East Asia—
Bangladesh and Indonesia. These two countries are remarkably similar in a number of
contextual features such as religion and the organization of the agricultural sector, which
makes cross-country comparison possible. In the 1960’s both countries had very low life
expectancy (about 40 years), and high fertility (about 6 children per woman over her
reproductive years lifetime). A large fraction of children between ages 6 and 12 were not
enrolled in primary school (Arthur and McNicoll 1978; Hugo et al. 1987; World Bank,
1982). However, after the 1960’s Indonesia developed at a faster pace than Bangladesh
on most dimensions. And by the mid-1990s, Indonesia was at a considerably higher level
of socioeconomic development than Bangladesh. During the fiscal year 2007-08
Bangladesh’s per capita income was US$ 599, whereas Indonesia’s was US$ 3,900. In
addition, the two countries differ in the systems of family organization; relationships are
far more patriarchal in Bangladesh than in Indonesia.
With over 140 million people squeezed into an area of 145,000 km
2
(i.e., about
the size of the state of Iowa in the US), Bangladesh is one of the most densely populated
countries in the world (UNFPA 2011). According to the United Nations Development
Program’s Human Poverty Index (HPI), in 2007, Bangladesh ranked 93
rd
poorest of 108
less-developed countries. Despite the gloomy predictions made by many in the mid-
1970s, Bangladesh has graduated from the status of “low economic potential” to an
“emerging market” economy. Current trends show that the poverty level in Bangladesh
159
has fallen over the past decade from 58 percent to 49 percent (UNFPA 2011). In addition,
human development indicators such as infant the mortality rate, the child mortality rate,
contraceptive use, the total fertility rate, primary and secondary school enrollment levels,
women’s labor force participation and median age at first marriage for women, all have
shown rapid improvements from the late 1990s to late 2000s. Although, Indonesian
citizens generally enjoy better human conditions than Bangladeshi citizens, between 1997
and 2007 Indonesia also made significant progress in human development indicators and
economic growth. I use these population-level changes in women’s socioeconomic
characteristics to predict change in decision-making and child stunting.
2. Methods Used
Throughout this dissertation, my analysis plan consisted of multivariate logistic
regression models to investigate change in coefficients over time, and Oaxaca-Blinder
decomposition models to identify sources of improvement in decision-making and child
stunting. The rationale for using maximum likelihood estimates (MLE) for the
multivariate analysis is straightforward. MLE selects values of the model parameters that
produce a distribution (logistic or probit) that gives the observed data the greatest
probability. MLE methods are preferred for binary response models over ordinary least
squares (OLS) because the predicted values fall within (0, 1), estimates of the marginal
effects are consistent, and variance is constant (homoscedasticity).
Economists have used the Oaxaca-Blinder decomposition method to study wage
discrimination in the labor market since the 1970s. More recently, O’Donnell et al.
160
(2008) used it to analyze health inequalities by poverty status, and Tarozzi and Mahajan
(2007) used it to analyze improvements in child nutrition in India from 1992-93 to 1998-
99. Similar methods have been used in sociological and demographic research under
various headings (see Kitagawa 1955; Casper, McLanahan and Garfinkel 1994; Preston,
Heuveline and Guillot 2001). This method is suitable for analyzing two distinct groups. It
provides predicted probabilities for an outcome estimating results of one group if it had
the means or coefficients generated by the models of the other group. Therefore this
method was useful to make comparisons between two survey years (i.e., 1999 and 2007)
and two countries (i.e., Bangladesh and Indonesia) in my dissertation.
Although I used MLE for the multivariate analysis to investigate the change in
coefficients, for the purpose of decomposition, I specified a Linear Probability Model
(LPM). I chose to use LPM even if the dependent variable is binary because a linear
model makes the decomposition results easier to interpret. Moreover, the slopes
estimated with linear probability models were very close to the marginal effects estimated
with logit and probit. This approach is acceptable, as long as the predicted values fall
comfortably between 0 and 1, which in my case they did. Further, heteroskedasticity,
which is an inherent problem with the LPM, was resolved by specifying robust standard
errors (Nielsen 1998; Tarrozi and Mahajan 2007).
161
3. Results
3.1 Women’s Participation in Decision-Making in Bangladesh
In chapter 4, I attempted to answer two questions: (1) have determinants of women’s
participation in decision making changed from 1999 to 2007? and (2) do improvements
in women’s socioeconomic characteristics contribute to increases in women’s
participation in household decision-making? The multivariate logistic results showed that
in Bangladesh, while holding other variables constant, women who belong to an older
age category are more likely to participate in all decision-making domains than women
who are younger. This relationship between women’s age and decision-making did not
change from 1999 to 2007. Education at all levels (primary, secondary and higher
education) has a positive impact on decision-making compared with having no education.
This is expected; as education levels are enhanced, women will have increased agency as
well as increased negotiating power both at home and in the community. It should be
noted that this is an independent effect, which means regardless of one’s employment
status, higher levels of education give women more authority to influence household
decision-making. The coefficients of education categories did not change from 1999 to
2007, which means the manner in which they predict decision-making and their strength
in doing so has not changed over time.
Work empowers women and allows them to influence household decisions,
because they are not as likely to have to depend on their husbands for resources. This is
evident in the multivariate analyses, where women who work have greater odds of
participating in decision-making than women who don’t work. However, the effect of
162
work has weakened from 1999 to 2007, this is especially true for decisions regarding
woman’s health and large purchases. What this may mean substantively is that as
women’s employment becomes more widespread, the strength of the variable becomes
weaker, because it is no longer a novel attribute in women.
Membership in organizations such as the Grameen Bank has a positive
independent effect on all five decision-making outcomes. Membership in microcredit
program benefits women in two ways: (1) it provides women credit to engage in self-
employment activities, and (2) it gives access to a large social network, which allows
women to share information and knowledge. This is an encouraging finding, especially
because microfinance programs were first introduced in Bangladesh by Muhammad
Yunus (1983) to empower women by giving them access to credit, skills and social
capital. It is wise to continue such programs as a means of empowering women in less-
developed countries. It should be noted that the effect of membership in a microcredit
program is significantly stronger in 2007 than in 1999 on decisions regarding children’s
health care. Thus, it can be said, these programs not only benefit women’s wellbeing, but
also their children’s wellbeing.
Household headship is the strongest predictor of decision-making, regardless of
women’s age, education level and employment. In a patriarchal country like Bangladesh,
a woman rarely assumes household headship unless her spouse is absent from the
household. According to these results, it is clear that when women head households, they
strongly influence child health, and that headship matters more for child health than any
other decision-making sphere. The multivariate results indicated that women’s age,
163
education, work, household headship and membership in microcredit programs have a
positive impact on their participation in household decision-making. However, the
strength of the coefficients for some variables has weakened overtime, indicating that
certain attributes that used to predict women’s decision-making power no longer have the
same impact.
The Oaxaca-Blinder decomposition results indicate that the proportion of women
who participate in decision-making regarding child health increased by 13.7 percentage
points from 1999 to 2007. If women in 1999 had same socioeconomic characteristics as
women in 2007, there would have been a 2.8 predicted percentage point increase in
women’s decision-making on child health in 1999 (i.e., 21% increase), if women’s
coefficients were at 2007 levels, there would have been a 4.3 predicted percentage point
decrease in 1999 (i.e., 32% decrease). The latter finding indicates that the relationship
between women’s characteristics and decision-making (i.e., coefficients) weakened in
2007 compared with 1999 as indicated by the multivariate analyses.
Because the five decision-making spheres are measuring different domains of
household decision-making, we see varying levels of predicted values. For example, if
women in 1999 had the same means as women in 2007, there would be a 63% increase in
the proportion of women who participate in decisions regarding large purchases, and
there would be a 52% increase in the proportion of women who participate in decisions
regarding visits to family and friends. This shows that the improvement in women’s
socioeconomic characteristics influences certain decision-making domains more than
others.
164
Two key findings emerge from the Oaxaca-Blinder decomposition. First,
improvements in women’s education, work, membership in microcredit programs and
household headship accounts for on average a 3-percentage point increase in decision-
making. This partly supports my hypothesis that improvements in women’s
characteristics affect improvements in decision-making.
The second finding is that, the change in intercepts accounts for a large proportion
of the change in decision-making (except for large purchases), but its impact is
counteracted by the reduction in coefficients due to the weakening of the association
between women’s characteristics and decision-making. The intercepts subsume the
effects of differences in unobserved predictors (Jann 2008). Such unobserved predictors
may include, changes in norms and attitudes towards women, omitted development
variables, or other omitted characteristics that impacts women’s decision-making power.
In Bangladesh, such omitted variables might include changes to the law and government
programs. For example, the Prime Minister of Bangladesh issued a memorandum in 2003
urging all government officials to work towards abolishing the practice of dowry. Dowry
is closely related to age at first marriage, as parents are forced to marry their daughters at
young age to avoid an expensive dowry. If the dowry system is abolished, girls can
continue schooling and join the labor force before getting married. Further, direct
intervention from the mass media, which shows working women’s images through
dramas, advertisements and feature films have a profound impact on the culture of the
Bangladesh. Moreover, as modernization theory suggests, the rapid globalization of the
economy fosters a more progressive attitude towards women. Therefore, the intercept
165
may have subsumed these changes in norms and attitudes. However, we cannot rule out
other development variables and women’s characteristics that may strongly predict
women’s decision-making, but are omitted from the model.
3.2 Household Decision-Making in Bangladesh and Indonesia
In the second paper, I repeated the above analysis in Indonesia—a more gender
egalitarian and economically developed Muslim society. In addition to using women’s
age and education as predictors of women’s decision-making, I used two other women’s
status variables that are rarely seen in household surveys: woman’s share of household
assets and woman’s parents’ status in the community at the time of marriage.
Multivariate analysis showed that Indonesian women’s education does not have an
independent effect on decision-making, however, women’s age, share of household assets
and family status at the time of marriage significantly influences decision-making.
Women aged 25-34 years have about 2-3 greater odds of participating in decision-making
regarding child health compared with women aged 15-25 years. The effect is weaker for
women aged 35-49 years in 1997. Woman’s education does not have a strong
independent effect on decisions regarding child health; even the joint significance test did
not yield significant results. This means the effect of education is completely explained
by other explanatory variables. It also reveals that although education had a significant
independent effect in a patriarchal and less-developed nation like Bangladesh, it does not
have an independent effect on a more gender egalitarian and relatively more developed
society like Indonesia. This also emphasizes that certain attributes deemed to be
166
important predictors of women’s status are not relevant in certain contexts. Woman’s
share of household assets has a positive effect on decision-making regarding child health.
For instance, compared with women who don’t own any household assets, women who
own 50% or more assets have 7.9 greater odds of participating in decisions regarding
child health in 2007. The effect becomes stronger as the woman’s share of assets
increases. Household headship has no effect on women’s decision-making power in 1997
or in 2007. Woman’s parents’ status in the community has a significant influence on
decisions regarding child health in 1997. Relative to women who had a “higher” position
in the community at the time of marriage compared with their husbands, women who had
a “lower” position have negative odds of participating in decision-making. However, this
effect disappears in 2007. Woman’s partner’s increasing age has a positive impact on the
outcomes. This means, keeping women’s age constant, women who have older partners
are more likely to participate in decisions than women with younger partners. The effect
of partner’s higher education has a positive impact in 1997, but the relationship becomes
negative and insignificant in 2007.
Indonesia makes an interesting research site for social scientists because of its
diverse ethnicity and various adat laws. In 1997, compared with Javanese women,
women from Sundanese, Bugis, Banjar, Betawi and other Southern Sumatran ethnic
groups were less likely to participate in decisions regarding large purchases, whereas
women in the matrilineal Minang ethnic group were more likely to participate in
decisions regarding large purchases (odds ratio = 2.14). This effect becomes weaker and
167
insignificant in 2007; this is possibly due to changes to traditional adat laws in Java
where attitudes towards women resemble the values of the matrilineal Minang group.
Results from Oaxaca-Blinder decompositions show that on average, about 20
percent of the increase in decision-making is accounted for by the improvements in
women’s characteristics from 1997 and 2007; however, about 75 percent of the increase
in decision-making derives from the changes in coefficients and the intercept.
Qualitatively, these results are consistent with what I found in Bangladesh, i.e., important
unobservable variables are subsumed in the intercept, which may or may not represent
developmental variables.
In the second part of this analysis, I examine the differences in decision-making
between Bangladesh and Indonesia. In both countries, woman’s age has a positive impact
on decision-making; however the effect is stronger in Indonesia than in Bangladesh.
Further, women with secondary or higher education have greater odds of decision-
making compared with women with no education women in Bangladesh. Similarly
Bangladeshi women who work and who are household heads have greater odds of
decision-making compared with women who don’t work or who are not household heads.
By contrast, Indonesian women’s education, work and household headship have no
independent effect on decision-making regarding child health, but women who work have
greater odds of decision-making on large purchases than women who don’t work.
Overall, the multivariate results indicate that variables that strongly influence decision-
making in Bangladesh do not have a strong impact (or no impact at all) on decision-
making in Indonesia. This suggests that the underlying cultural context plays an
168
important role in women’s status regardless of their socioeconomic characteristics.
Further investigation using Oaxaca-Blinder decomposition confirms these findings.
The decomposition results indicated that on decisions regarding child health there
is a 12.3 percentage point difference between Indonesian women and Bangladeshi
women; the difference is even greater for decisions regarding large purchases—20.8
percentage points. If Bangladeshi women had the same socioeconomic characteristics as
Indonesian women, there would be a 4.8 percentage point increase on decisions regarding
child health, and 5.2 percentage point increase on decisions regarding large purchases.
But if we look at the predicted increase due to coefficients, women’s decision making on
child health would increase by 9.3 percentage points if Bangladesh had the same
coefficients and intercept as Indonesia, similarly, women’s decision-making on large
purchases would increase by 16.3 percentage point. The main sources of change for
decisions on large purchases originate from the differences in intercepts, which subsumes
differences in unobserved variables. The main sources of change for decisions on child
health originate from the differences in coefficients. What this qualitatively means is that,
the difference in women’s participation in decision-making between the two countries is
only partly due to the relatively higher socioeconomic status (i.e., population
composition) of Indonesian women compared with Bangladeshi women, this could be a
result of the differences in economic development between the two countries. But the
majority of the difference is attributed to coefficients and intercepts. Although I cannot
empirically prove that these unobservable variables are attributable to cultural norms and
169
family systems, the anthropological literature makes a strong argument for the impact of
differential cultural norms on women’s status in Bangladesh and Indonesia.
According to Islamic tenets, the responsibility for one’s parents is in the hands of
adult children. In Bangladesh this responsibility largely falls to sons (Kabir, Szebehely
and Tishelman 2000; Mahmood 1992; Rahman 1999). In fact, sons have been described
as the best risk insurance available to women (Arthur and McNicoll 1978; Cain, Khanam
and Nahar 1979). This difference in the relative importance of sons and daughters is
deeply rooted in the patrilineal system of family organization and is exacerbated where
the joint family (i.e. extended family) takes precedence over the conjugal family.
Marriage of sons and daughters is a highly planned family strategy; it is an arrangement
between the couple’s parents (Arthur and McNicoll 1978). Although daughters do useful
household work, they do not bring in outside wages like sons. Under such circumstances,
families tend to marry off daughters at an early age but hold sons as long as possible.
This contributes to the large age difference between brides and grooms, where brides are
much younger than grooms (Arthur and McNicoll 1978).
The low status of women in Bangladesh is a result of this rigid family system
coupled with strong traditions of purdah. According to the tradition of purdah,
respectable women do not engage in trade or fieldwork or leave the family for other than
traditionally specified visits to relatives (Arthur and McNicoll 1978). This means they
cannot find employment outside their homes. Women work mostly on tasks that can be
done at home, such as processing harvested rice and producing handicrafts (Balk 1997;
Cain, Khanam and Nahar 1979; Zaman 1995). However, men are responsible for selling
170
these products, thus women have little or no control over the profits gained by these
products (Frankenberg and Kuhn 2004). For these reasons, separation and divorce are
real threats to women, especially if they don’t have sons. Therefore, fulfilling duties to
one’s husband and in-laws take on a special importance, as does producing sons as
security for later years. As a result, women find themselves carrying out roles dictated by
the interests of their husbands, sons, or in-laws (Arthur and McNicoll 1978).
In contrast, according to Islamic tenets and traditional law (adat) in Indonesia,
children are obligated to care for their older parents (Mahmood 1992; Frankenberg,
Lillard and Willis 2002). But unlike in Bangladesh, the obligation is assigned to children
of both genders (Keasberry 2001). Therefore, the difference in the relative importance of
sons and daughters does not exist in the Indonesian culture. Further, purdah is not at all
practiced in Indonesia; women have complete freedom of movement.
Indonesian women have traditionally played a prominent role in both the public
and domestic realms. It is not uncommon for couples to work together, not only as part of
an economic survival strategy (Koentjaraningrat 1967), but also as an interactive
decision-making unit (Bangun 1981). Wives are given an equal say in the determination
of important household matters, particularly those involving control over financial
resources (Geertz 1961; Mangkuprawira 1981). Women work in agriculture, sometimes
with their husbands and sometimes on their own, and they engage in trade. Women also
own property separate from their husbands and participate in household management and
decision-making—sometimes to the point of domination. Because women are not
dependent on their husbands economically, divorce is not a major threat to their survival.
171
If divorced, women usually can rely on their parents to provide them and their children
with a place to live (Heaton, Cammack and Young 2001; Jones 1994).
These differences in social norms and attitudes towards women may have played
an important role in the differences in women’s participation in decision-making.
Although I cannot empirically prove that the unobserved variables measure differences in
social norms, the above-mentioned anthropological literature helps us to understand why
we see such large differences between two predominantly Muslim countries.
3.3 Women’s Status and Child Nutrition in Bangladesh
The final paper of this dissertation analyzes the sources of improvement in child nutrition
in Bangladesh. I use child stunting, (i.e. height conditional on age) z scores (haz) as my
measure of child nutrition. Although malnutrition rates are generally high in Bangladesh,
I found significant reduction in stunting from 1999 to 2007. To explore the sources of this
reduction in stunting, I first identified the predictors of stunting that have changed over
time.
The results of the multivariate logistic regression showed that parent’s education
remains significant in 1999 and 2007 even after controlling for household and community
characteristics. Controlling for partner’s education, the mother’s education can be
interpreted as relative education (or as a measure of relative power). Because most men
are better educated than their wives; when holding men’s education constant, higher
levels of education among women imply a reduction in the gap (Frankenberg and
Thomas 2003). However, the effect of mother’s education disappears when controlling
172
for age at first marriage, birth number and television viewership, while father’s education
remains significant. This implies that the effect of mother’s education is completely
explained by child’s birth number, age at first marriage and television viewership in
2007. These variables are considered to be pathways through which mother’s education
affects child stunting, and they seem to matter more in 2007 than in 1999.
The regression results of the household and community variables are less
conclusive. In 1999, availability of electricity, a good source of drinking water and
advanced sewerage system had a negative impact on child stunting, however, in 2007 the
drinking water and sewerage variables a not significant. This is because in 2007,
availability of a good source of drinking water was nearly universal (97%), thus
providing little variation for estimation. However, the non-significance of the sewerage
variable is difficult to explain, it could be due its modest correlation with the availability
of electricity in 2007. The household furniture variable (a proxy for household assets)
does not have a significant effect on child stunting in 1999, although it has a weak but
significant effect in 2007. Although, some coefficients seem to differ in 1999 compared
with 2007 (e.g., birth number of the child), the pooled sample chi-square test indicated
that overall, the coefficients did not change significantly over time. This implies that, the
reduction in stunting from 1999 to 2007 is mainly due to the changes in the means of the
variables, and not due to the change in the relationship between the variables and
stunting, an implication that was confirmed using Oaxaca-Blinder decomposition.
In the second part of the analysis, I use Oaxaca-Blinder decomposition to analyze
the key variables that have contributed to the reduction in stunting from 1999 to 2007.
173
The results revealed that almost half of the reduction in stunting is due to mean changes
in women’s characteristics. The other half is due to mean changes in household assets
and community infrastructure. There has been little or no change in the coefficients,
which indicates that the improvement in stunting is purely due to improvements in the
means. In a separate analysis, I explored the differences in the decomposition results by
child’s gender. For boys, improvements in parent’s education accounted for a large
portion of the reduction in stunting, but for girls the effect was smaller. In fact, the
majority of the reduction in stunting for girls comes from improvements in household
assets and community infrastructure, which are usually available for all children
regardless of their gender. These results confirm two things: improvements in women’s
education (net of partner’s education), age at first marriage, fertility (birth order of the
child) and access to information through television viewership are partly responsible for
the overall improvement of child nutrition. However, because parents still favor sons over
daughters in Bangladesh, the effect of parent’s education is weaker for girls than for
boys. From a welfare policy point of view, these results are encouraging, as empowering
women through mandatory school enrollment, microcredit programs and distribution of
contraceptives have benefited the nutritional status of children, which will ultimately
impact the economic growth of the country. However, there is still work to be done in
terms of improving norms and attitudes towards female children in Bangladesh.
In addition to stunting, I attempted to investigate other indictors of child nutrition,
including wasting i.e. weight-for-height (a short term indicator of nutrition) and birth
weight as additional dependent variables. But wasting in Bangladesh increased by five
174
percentage points from 1999 to 2007 and none of the variables in my model were able to
explain this increase. This is possibly because in 2007 Bangladesh and two other South
Asian countries experienced heavy monsoonal rains, which led to what has been
described as the worst flood in living memory (UNICEF 2007). People suffered from
diarrhea and waterborne diseases, crops and livestock were destroyed causing a food
shortage and starvation. I assume that the increase in wasting might be due to this natural
disaster because child’s weight can temporarily fluctuate due to illness or famine.
4. Conclusion
The empirical evidence presented in this dissertation supports the notion that when
women are educated and are free to pursue employment and other gainful activities, they
are able to make decisions that affect their wellbeing as well as their children’s
wellbeing.
Women’s status has significantly improved over time in Bangladesh and in
Indonesia, as indicated by their increased participation in decision-making. The most
important factors that have influenced these improvements include education and
economic activity (measured by employment, micro-credit program participation or share
of household assets). These factors have not only improved women’s agency within their
households, but they are also responsible for improvements in children’s nutrition.
Therefore, this dissertation provides empirical evidence to bolster UN Secretary
General’s argument that “When women are healthy, educated and free to take the
175
opportunities life affords them, children thrive and countries flourish, reaping a double
dividend for women and children” (UNICEF 2007).
176
Bibliography
Agarwal, B. 1997. “Bargaining and Gender Relationships: Within and Beyond the
Household.” Feminist Economics 3(1): 1-51
Arthur, W.B., and G. McNicoll. 1978. “An Analytical Survey of Population and
Development in Bangladesh.” Population and Development Review, 4(1), 23-
80.
Badan Pusat Statistik (BPS) and ORC Macro. 2003. Indonesia Demographic and
Health Survey 2002-2003. Calverton, Maryland, USA: BPS and ORC Macro.
Badan Pusat Statistik (BPS) and Macro International. 2008. Indonesia Demographic
and Health Survey 2007. Calverton, Maryland, USA: BPS and Macro International.
Badan Pusat Statistik (BPS). 2012. “Population by Region and Religion.” Accessed
online at http://sp2010.bps.go.id/index.php/site/tabel?tid=321&wid=0 on
January 23
rd
, 2012
Balk, D. 1994. “Individual and Community Aspects of Women’s Status and Fertility
in Rural Bangladesh.” Population Studies, 48:1, 21-45
______. 1997. “Defying Gender Norms in Rural Bangladesh: A Social Demographic
Analysis.” Population Studies, 51(2), 152-172.
Bangladesh Bureau of Statistics (BBS). 2008. Statistical Y earbook of Bangladesh
2006. Dhaka, Bangladesh: BBS.
Bangun, M. 1981. "Functional education and credit facilities." In The Endless Day:
Some Case Material on Asian Rural Women Ed. T. S. Epstein and R. A.
Watts,. New York: Pergamon Press.
Barosso, C., and J. L. Jacobson. 2000. “Population Policy and Women’s
Empowerment: Challenges and Opportunities.” In Women’s Empowerment and
Demographic Processes, ed. H.B. Presser and G. Sen. Oxford: Oxford
University Press.
Becker, H.S. 1963. Outsiders: Studies in the Sociology of Deviance. New York: Free
Press.
Becker S., F. Fonseca-Becker and C. Schenck-Yglesias. 2006. “Husbands' and wives'
reports of women's decision-making power in Western Guatemala and their
effects on preventive health behaviours.” Social Science & Medicine, 62(9):
2313-2326
177
Beegle, K., E. Frankenberg, and D. Thomas. 2001.”Bargaining power within couples
and reproductive health care use in Indonesia.” Studies in Family Planning, 32(2),
130-146
Bernasek, A. 2003. “Banking on Social Change: Grameen Bank Lending to Women.”
International Jouranal of Politics, Culture and Society.
Bhagowalia, P., P. Menon, A. R. Quisumbing and V. Soundrararjan. 2010.
“Unpacking the Links Between Women’s Empowerment and Child Nutrition:
Evidence Using Nationally Representative Data From Bangladesh.” Selected Paper
prepared for presentation at the Agricultural & Applied Economics Association
2010. AAEA, CAES, & WAEA Joint Annual Meeting, Denver, Colorado, July 25-
27.
Blau, F. D. and L.M. Kahn. 1994. “Rising Wage Inequality and the U.S. Gender
Gap.” The American Economic Review 84(2)
Blinder, A. S. 1973. “Wage Discrimination: Reduced Form and Structural Estimates.”
The Journal of Human Resources 8: 436–455.
Blumberg, R. L. 1988. “Income under female versus male control: Hypothesis from a
theory of gender stratification and data from the Third World.” Journal of Family
Issues 9(1): 51-84.
Boserup, E. 1970. Woman’s Role in Economic Development, London: George Allen
and Unwin.
Bourguignon, F., M. Browning, P. Chiappori and V. Lechene. 1994. “Incomes and
Outcomes: A Structural Model of Intrahousehold Allocation.” Journal of
Political Economcy 102(6): 1,067-1,096.
Cain, M., S. R. Khanam and S. Nahar. 1979. “Class, patriarchy, and women’s work in
Bangladesh.” Population and Development Review, 5(3), 405-438.
Cain, M. 1991. “The activities of the elderly in rural Bangladesh.” Population Studies,
45, 189-202.
Caldwell, J.C. 1979. “Education as a Factor in Mortality Decline: An Examination of
Nigerian Data.” Population Studies 33: 395-413.
___________. 1982. Theory of fertility decline. London: Academic Press
___________. 1986. “Routes to Low Mortality in Poor Countries.” Population and
Development Review 12: 171-220.
178
Casper, L. M., S. S. McLanahan and I. Garfinkel. 1994. “The Gender-Poverty Gap:
What We Can Learn from Other Countries.” American Sociological Review
59(4), 594-605
Casper, L. M. and P. J. Kitchen. 2008. “Demographic Factors.” in Encyclopedia of
Infant and Early Childhood Development, edited by M. M. Haith and J. B.
Benson. Oxford: Elsevier Science.
Central Bureau of Statistics (CBS), State Ministry of Population/National Family
Planning Coordinating Board, Ministry of Health, and Macro International.
1998. Indonesia Demographic and Health Survey 1997. Calverton, Maryland,
USA: CBS and Macro International.
Chen, L., A. Chowdhury, and S. Huffman. 1980. “Anthropometric Assessment of
Energy-Protein Malnutrition and Subsequent Risks of Mortality Among
Preschool Aged Children.” American Journal of Clinical Nutrition 33: 1836-
45.
Chen, L.E., E. Huq and S. D’Souza. 1981. “Sex-bias in the Family: Allocation of
Food and Health Care in Rural Bangladesh.” Population and Development
Review 7(1): 54-70.
Contreras, D., D. Thomas, and E. Frankenberg. 1997. “Child health and the
distribution of household resources at marriage.” Manuscript. RAND
Corporation, Santa Monica, Calif., U.S.A.
Corner, L. 2008. “Women’s Role in Economic Development: Review.” European
Review of Agricultural Economics 35(4): 598-601.
Cotton, J. 1988. “On the Decomposition of Wage Differentials.” Review of Economics
and Statistics 70(2): 236–43.
Curtin, L. B. 1982. Status of Women: A Comparative Analysis of Twenty Developing
Countries. Population Reference Bureau, Washington.
Curtin, L.B., C.N. Johnson, A.B. Kantner, and A. Papilaya. 1992. Indonesia’s
National Family Planning Program: Ingredients of Success. Occasional Paper
No. 6. Washington, DC: Population Technical Assistance Project.
Dey, J. 1981. Gambian Women: Unequal Partners in Rice Development Projects?
Journal of Development Studies, 17:109-122.
179
Dixon, R. 1976. “The Roles of Rural Women: Female Seclusion, Economic
Production and Reproductive Choice.” In Population and Development, Ed. R.
Ridker, Johns Hopkins University Press: Baltimore.
Domingo, L., and E. King. 1992. “The role of the family in the process of entry to
marriage in Asia.” In Family systems and cultural change (Eds) E. Berquo & P.
Xenos. Oxford, England: Clarendon Press
Dyson, T., and M. Moore. 1983. “On Kinship Structure, Female Autonomy. And
Demographic Behavior in India.” Population and Development Review 9(1):
35-60
England, P. 2000. “Conceptualizing Women’s Empowerment in Countries of the
North.” In Women’s Empowerment and Demographic Processes, ed. H. Pressed
and G. Sen. Oxford: Oxford University Press.
Engle, P. L., P. Menon and L. Haddad. 1997. Nutrition: Concepts and Measurements.
International Food Policy Institute.
Frankenberg, E. and D. Thomas. 2000. "The Indonesia Family Life Survey (IFLS):
Study Design and Results from Waves 1 and 2." March 2000. RAND, Santa
Monica, CA. DRU-2238/1-NIA/NICHD.
_______________. 2001. “Women’s Health and Pregnancy Outcomes: Does Access
to Services Make a Difference?” Demography, 38(2): 253-265
Frankenberg, E., Lillard, L and R. Willis. 2002. “Patterns of Intergenerational
Transfers in Southeast Asia. Journal of Marriage and Family, 64, 624-641.
Frankenberg, E. and R. S. Kuhn. 2003. “The Role of Social Context in Shaping
Intergenerational Relations in Indonesia and Bangladesh.” In Annual Review
of Gerontology and Geriatrics: Focus On Intergenerational Relations Across
Time and Place, ed M. Silverstein and K. W. Schaie, vol 24, Springer
Publishing Company.
Frankenberg, E. and D. Thomas. 2003. "Measuring Power", in A. Quisumbing Ed.
Household Decisions, Gender and Development, Washington, D.C.: IFPRI.
Gage, A. 1995. “Women’s Socioeconomic Position and Contraceptive Behavior in
Togo.” Studies in Family Planning 26(5): 264-277.
Geertz, H. 1961. The Javanese Family. New York: Free Press of Glencoe.
Gerth, H.H. and C.W. Mills. 1946. From Max Weber: Essays in Sociology. New
York: Oxford University Press.
180
Goffman, E. 1959. The Presentation of Self in Everyday Life. Garden City, N.Y:
Doubleday.
Goldsheider, F. K. 1995. “Interpolating Demography with Families and Households.”
Demography 32:471-80
Goode, W. J. 1963. World revolution and family patterns. New York: Free Press of
Glencoe.
Habicht, J., R. Martorell, C. Yarbroughm R, Malina, and R. Klein. 1974. “Height and
Weight Standards for Preschool Children. How Relevant are Ethnic
Differences in Growth Potential?” Lancet April 6: 611-614
Haddad, L., J. Hoddinott, and H. Alderman. 1997. Intra-household Resource
Allocation in Developing Countries: Models, Methods, and Policy. Baltimore,
Maryland: Johns Hopkins University Press.
Hart, G. 1978. “Labor Allocation Strategies in Rural Javanese Households.”
University of California, Berkeley. Doctoral Dissertation.
Heaton, T., Cammack, M., and L. Young. 2001. “Why is the divorce rate declining in
Indonesia?” Journal of Marriage and Family 63, 480-490.
Hirschman, C. 1985. “Pre-marital socioeconomic roles and the timing of family
formation.” Demography, 22, 35-59.
Hoddinott, J., and C. Adam. 1998. Testing Nash bargaining household models with
time series data: Divorce law reform and female suicide in Canada. Food
Consumption and Nutrition Division Discussion Paper 52. International Food
Policy Research Institute, Washington, D.C.
Hoddinott, J., J. A. Maluccio, J. R. Behrman, R. Flores and R. Martorell. 2008.
“Effect of a Nutrition Intervention During Early Childhood on Economic
Productivity in Guatemalan Adults.” Lancet 371: 411-416.
Hoodfar, H. 1995. “Population Policy and Gender Equity in Post-Revolutionary Iran”
in Family, Gender, and Population in the Middle East, Ed by C.M. Obermeyer.
Cairo: American University in Cairo Press.
Hughes, E. C. 1945. “Dilemmas and Contradictions of Status.” American Journal of
Sociology 50.
Hugo, G., T. Hull, V. Hull and G. Jones. 1987. The Demographic Dimension in
Indonesian Development. Singapore: Oxford University Press.
181
Hull, T. H. 1981. "The 'positive relation' and future Indonesian fertility." Research
note prepared for the International Population Dynamics Program, Department of
Demography, Australian National University, Canberra.
Jann, B. 2008. “A Stata Implementation of the Blinder-Oaxaca Decomposition.” ETH
Zurich Sociology Working Paper No.5.
Jay, R. R. 1969. Javanese Villagers: Social Relations in Rural Modjokuto.
Cambridge, MA: MIT Press.
Jejeebhoy, S. J. 2000. Women’s autonomy in rural India: Its dimensions,
determinants, and the influence of context. In Women’s empowerment and
demographic processes, ed. H. Presser and G. Sen. Oxford: Oxford University
Press.
Jejeebhoy, S.J. and Z. Sathar. 2001. “Women’s Autonomy in India and Pakistan: The
Influence of Religion and Region.” Population and Development Review
27:687–712.
Jin, H. 1995. “A study of rural women's decision making power on reproduction and
fertility.” Chinese Journal of Population Science, 7:241-257
Jones, G. 1994. Marriage and Divorce in Islamic Southeast Asia. Kuala Lumpur.
Oxford University Press.
Kabeer, N. 1999. “Resources, Agency, Achievement: Reflections on the measurement
of women’s empowerment. Development and Change 30(3): 435-464
Kabir, Z., Szebehely, M., and C. Tishelman. 2002. “Support in old age in the
changing society of Bangladesh.” Ageing & Society, 22,615-636.
Kamal, N. and K. M. Zunaid. 2011. “Education and Women’s Empowerment in
Bangladesh.” Accessed online at http:/centers.iub.edu.bd in December 2011
Katz, E. 1997. “The intrahousehold economics of exit and voice.” Feminist
Economics 3 (3): 25–46.
Keasberry, I. N. 2001. “Elder care and intergenerational relationships in rural
Yogyakarta, Indonesia.” Ageing & Society, 21, 641-665.
Kennedy, E. and P. Peters. 1992. “Household Food Security and Child Nutrition: The
Interaction of Income and Gender of Household Head.” World Development 20
(8): 1077-1085.
182
Kesarda, J.D., J. O.G. Billy and K. West. 1986. Status Enhancement and Fertility:
Reproductive Response to Social Mobility and Educational Opportunity.
Orland, Florida. Academic Press.
Kevane, M. 2000. “Extrahousehold norms and intrahousehold bargaining: Gender in
Sudan and Burkina Faso.” In Women farmers and commercial ventures:
Increasing food security in developing countries, ed. A. Spring. Boulder,
Colo., U.S.A., and London: Lynne Rienner Publishers.
Kishor, S. 1999. “Women’s Empowerment and Contraceptive Use in Egypt.” Paper
presented at the Annual Meetings of the Population Association of America,
March, New York
________. 2000. “Empowerment of Women in Egypt and Links to the Survival and
Health of their Infants.” In Women’s Empowerment and Demographic
Processes, ed. H. Pressed and G. Sen. Oxford: Oxford University Press.
Kitagawa, E. M. 1955. “Components of a Difference between Two Rates,” Journal of
the American Statistical Association, 50(272): 1168-94.
Koentjaraningrat, R. M. 1960. “The Javanese of South Central Java.” In G. P.
Murdock (ed), Social Structure in Southeast Asia (pp. 88-115). Chicago:
Quadrangle.
__________________. 1967. "Tjelapar: A village in South Central Java." In Villages
in Indonesia. Ithaca, NY: Cornell University Press.
Kritz M. M. and P. M. Adebusoye. 1999. “Determinants of women's decision-making
authority in Nigeria: the ethnic dimension.” Sociological Forum 14:399-424.
Lundberg, S., and R. Pollak. 1993. “Separate spheres, bargaining, and the marriage
market.” Journal of Political Economy 101 (6): 988–1010.
Lundberg, S., R. A. Pollak, and T. J. Wales. 1997. "Do Husbands and Wives Pool
Their Resources? Evidence from the U.K. Child Benefit," Journal of Human
Resources, Vol. 32, No.3, (Summer), 463-480.
Mahamood, T. 1992. “Law and the Elderly in the Islamic Tradition.” In Aging in East
and Southeast Asia, ed. D. R. Philips, (pp. 227-235). London, England:
Edward Albert.
Malhotra, A. 1997. “Gender and the Timing of Marriage: Rural-Urban Differences in
Java.” Journal of Marriage and Family 59(2), pp. 434-450
183
Mandelbaum , D. G.1974. Human Fertility in India: Social Components and Policy
Perspectives. London, England: University of California Press.
Mangkuprawira, S. 1981. "Married women's work pattern in rural Java." In The
Endless Day: Some Case Material on Asian Rural Women (Ed) T. S. Epstein
and R. A. Watts, New York: Pergamon Press.
Manser, M. and M. Brown. 1980. “Marriage and household decision making: a
bargaining analysis”International Economic Review, 21.1:31-44.
Martorell, R., and J. Habicht. 1986. “Growth in Early Childhood in Developing
Countries.” In Human Growth: A Comprehensive Treatise, ed, F. Falkner and
J. Tanner, vol 3, 2
nd
edition, New York: Plenum Press.
Mason, K. 1985. The Status of Women: A Re-view of its Relationships to Fertility and
Mortality. New York: Rockefeller Foundation
______. 1986. “The status of women: Conceptual and methodological issues in
demographic studies.” The Eastern Sociological Society 1 (2): 284–300.
______. 1987. "The impact of women's social position on fertility." Sociological
Forum 4: 718-745.
______. 1993. “The impact of women’s position on demographic change during the
course of development.” In Women’s position and demographic change, ed.
N. Federici, K. Mason, and S. Sogner. Oxford: Clarendon Press.
______. 1996. “Women’s Domestic Economic Power in Five Asian Counties.”
Honolulu: East-West Center. Unpublished.
McDonald, P. 1985. “Social organization and nuptiality in developing societies.” In
Reproductive change in developing countries, (Eds) J. Cleland & J. Hobcraft.
New York: Oxford University Pres
McElroy, M., and M. Horney. 1981. “Nash-bargained Household Decisions: Toward
a New Generalization of the Theory of Demand.” International Economic
Review 22(2): 333-349.
Mensch, B., H. Lentzner and S. Preston. 1986. Socio-economic Differentials in Child
Mortality in Developing Countries. United Nations, New York.
Miles-Doan, R. and K. L. Brewster. 1998. “The Impact of Type of Employment on
Women’s Use of Prenatal-Care Services and Family Planning in Urban Cebu,
the Philippines.” Studies in Family Planning 29(1): 69-78.
184
Mishra, V. K. and R. D. Retherford. 2000. “Women's Education Can Improve Child
Nutrition in India.” National Family Health Survey Bulletin No.15.
Morgan, P. and K. Hiroshima. 1983. “The persistence of extended family residence in
Japan: Anachronism or alternative strategy? “American Sociological Review,
48, 269-281.
Nag, Moni, N. F. White, and R. C. Peet. 1980. "An anthropological approach to the
study of the economic value of children in Java and Nepal." In Rural
Household Studies in Asia, Eds. H. P. Binswanger et al Pp. 248-288, Singapore:
Singapore University Press.
Nath, S. R., M. Kabir, K. S. Ahmed, G. Roy, A. Hossain, S. M. N. Alam, F. K.
Chowdhury and A. Mahbub. 2012. “Exploring Low Performance in
Education: The Case of Sylhet Division” Accessed online at
http://www.bdeduarticle.com/component/content/article/207 on March 22nd 2012
National Institute of Population Research and Training (NIPORT), Mitra and
Associates, and Macro International. 2007. Bangladesh Demographic and Health
Survey 2007. Dhaka, Bangladesh and Calverton, Maryland, USA: National
Institute of Population Research and Training, Mitra and Associates, and Macro
International.
Neumark, D. 1988. “Employers’ Discriminatory Behavior and the Estimation of
Wage Discrimination.” The Journal of Human Resources 23: 279–295.
Nielsen, H. S. 1998. “Discrimination and Detailed Decomposition in a Logit Model.”
Economic Letters 61:115–20.
Oaxaca, R. 1973. “Male-Female Wage Differentials in Urban Labor Markets.”
International Economic Review 14(3): 693–709.
O’Donnel, O., A. L. Nicols and E. V. Doorslaer. 2009. “Growing richer and taller:
Explaining change in the distribution of child nutritionals status during
Vietnam’s economic boom.” Journal of Development Economic,s 88 (45-58).
Omi, M. and H.Winant. 1994. Racial Formation in the United States: From the 1960s
to the 1990s (2
nd
Edition). New York: Routledge.
Pasternak, B., C. Ember, and M. Ember. 1997. Sex, gender, and kinship: A cross-
cultural perspective. Upper Saddle River, N.Y., U.S.A.: Prentice Hall.
185
Pearson, R. 2007. “Gender and Development,” in D.A. Clark, eds., The Elgar
Companion to Development Studies. Massachusetts, USA: Edward Elgar
Publishing Inc.
Pollak, R. A. 2005. “Bargaining Power in Marriage: Earning, Wage Rates and
Household Production.” Paper presented at the AEA meeting in Philadelphia,
7-9 January 2005.
Powers, M. 1985. “Measuring the Situation of Women from Existing Data.” Paper
prepared for the Rockefeller Foundation Workshop on Women’s Status and
Fertility, Mount Kisco, NY
Preston, S. H., P. Heuveline and M. Guillot. 2001. Demography: Measuring and
Modeling Population Processes. Oxford: Blackwell Publishers.
Quisumbing, A. R. 1994. “Intergenerational Transfers in Philippine Rice Villages:
Gender Differences in Traditional Inheritance Customs.” Journal of
Development Economics 43(2): 167-195.
Quisumbing, A. R and J. A. Maluccio. 2003. “Resources at Marriage and Intra
household Allocation: Evidence from Bangladesh, Ethiopia, Indonesia and
South Africa.” Oxford Bulletin of Economies and Statistics 65(3): 283-328.
Rahaman, O. 1999. “Family matters: The impact of kind on the mortality of the
elderly in rural Bangladesh.” Population Studies, 53,221-225.
Ramalingaswami, V., U. Jonsson, and J. Rohde. 1996. “Commentary: The Asian
enigma.” in The Progress of Nations. New York: United Nations Children’s
Fund.
Reimers, C. W. 1983. “Labor Market Discrimination against Hispanic and Black
Men.” Review of Economics and Statistics 65(4): 570–79.
Riley, N. 1994. “Interwoven lives: Parents, marriage, and guaxi in China.” Journal of
Marriage and the Family, 56, 791-804.
_______. 1997. “Gender, power, and population change.” Population Bulletin 52 (1):
2–46.
Rosenzweig, M. and T. P. Schultz. 1983. “Estimating a Household Production
Function: Heterogeneity, the Demand for Health Inputs, and Their Effects on
Birth Weight.” Journal of Political Economy 91(5): 723-746.
186
Rutstein, S. O. 2008. “The DHS Wealth Index: Approaches for Rural and Urban
Areas.” DHS Working Papers No. 68
Safilios-Rothschild, C. 1982. Female power, autonomy and demographic change in
the Third World.” In Women’s roles and population trends in the Third World,
ed. R. Ankar, M. Buvinic, and N. Youssef. London: Croom Helm.
Sathar, Z. A. and S. Kazi. 2000. “Women’s Autonomy in the Context of Rural
Pakistan.” The Pakistan Development Review, 39: 2(Summer) pp 89-110
Schultz, T. P. 1980. “Assessing the Causes and Consequences of the Incidence of
Disease and Death: Interpretation of Relations among Health Status, Household
Economics and Health Environment.” in Proceedings of the Meeting on Socio-
Economic Determinants and Consequences of Mortality, Mexico City, 19-25 June
1979 (United Nations and World Health Organization, New York and Geneva).
__________. 1990. “Testing the Neoclassical Model of Family Labor Supply and
Fertility.” Journal of Human Resources 25(4): 599-634.
__________. 1993. “Returns to Women’s Education.” in Women’s Education in
Developing Countries: Barriers, Benefits and Policies edited by E. M. King
and M. A. Hill. Baltimore, MD: The Johns Hopkins University Press.
__________ .1999. “Health and Schooling Investments in Africa.” Journal of
Economic Perspectives 13(3): 67-88.
Seth, M. 2001. Women and Development: The Indian Experience. New Delhi: Sage
Publications India Pvt Ltd.
Sen, A. 1990. “More Than 100 Million Women Are Missing.” The New Y ork Review
of Books.
Sen, A. 1990. “Gender and cooperative conflicts.” In Persistent inequalities: Women
in world development, ed. I. Tinker. Oxford: Oxford University Press.
Sen, G., and S. Batliwala. 2000. “Empowering women for reproductive rights.” In
Women’s empowerment and demographic processes, ed. H. Presser and G. Sen.
Oxford: Oxford University Press.
Shorter, E. 1977. The making of the modern family. New York: Basic Books.
Skinner, G. 1997. “Family Systems and Demographic Processes.” In Anthropological
demography: Toward a new synthesis, (pp.53-95) ed D. I. Kertzer and T.
Fricks. Chicago: University of Chicago Press.
187
Smith, P. 1980. “Asian marriage patterns in transition.” Journal of Family History, 5,
58-96.
Smith, L.C. 1995. The Impact of Agricultural Price Liberation on Human Well-Being
in West Africa: Implications of Intra-household Preference Heterogeneity.
PhD Dissertation, University of Michigan, Ann Arbor, Mich., USA.
Smith, L., and L. Haddad. 2000. “Explaining child malnutrition in developing
countries. A cross-country analysis.” Research Report 111. Washington, D.C.:
International Food Policy Research Institute.
Smith, L.C., U. Ramakrishnan, A. Ndiaye, L. Haddad and R. Martorell. 2003. “The
Importance of Women’s Status for Child Nutrition in Developing Countries.”
International Food Policy Research Institute - Research Report 131
Soland, B. 2000. Becoming Modern: Y oung Women and the Reconstruction of
Womanhood in the 1920s. Princeton, New Jersey: Princeton University Press.
Strauss, J. and D. Thomas. 1998. “Health, Nutrition and Economic Development.”
Journal of Economic Literature, 36.2:766-817.
Strauss, J., K. Beegle, B. Sikoki, A. Dwiyanto, Y. Herawati and F. Witoelar. 2004.
"The Third Wave of the Indonesia Family Life Survey (IFLS): Overview and
Field Report", March 2004. WR-144/1-NIA/NICHD.
Strauss, J., F. Witoelar, B. Sikoki and A.M. Wattie. 2009. "The Fourth Wave of the
Indonesian Family Life Survey (IFLS4): Overview and Field Report". WR-
675/1-NIA/NICHD.
Tarozzi, A. and A. Mahajan. 2007. “Child Nutrition in India in the Nineties.”
Economic Development and Cultural Change 55(3): 441-486.
The Daily Star. 2012. “Sidr still haunts the coastal people.” Accessed online at
http://www.thedailystar.net/newDesign/news-details.php?nid=162672
on May 23
rd
, 2012
Thomas, D., J. Strauss and M.H Henriques. 1990. Child survival, height for age and
household characteristics in Brazil, Journal of Development Economics, 33:197-
234.
Thomas, D. 1990. “Intra-Household Resource Allocation: An Inferential Approach.”
Journal of Human Resources 25(4): 635-664.
188
Thomas, D., J. Strauss and M.H. Henriques. 1991. “How Does Mother’s Education
Affect Child Height? Journal of Human Resources 26(2): 183-211.
Thomas, D. 1997. “Incomes, expenditures, and health outcomes: Evidence on
intrahousehold resource allocation.” In Intrahousehold resource allocation:
Methods, models, and policy, ed. L. Haddad, J. Hoddinott, and H. Alderman.
Baltimore, Md., U.S.A.: Johns Hopkins University Press for the International
Food Policy Research Institute.
Thomas, D., D. Contreras and E. Frankenberg. 1997. “Child Health and the
Distribution of Household Resources at Marriage.” Sanara Monica, CA:
RAND. Unpublished.
Thomas, D., E. Frankenberg and B. Skoki. 1999. “The Measurement of Power and its
Implications for Understanding Inter-Household Decision-making.” Prepared
for the 1999 Population Association of America Meeting, March 25-27, New
York.
Thomas, D., D. Contreras and E. Frankenberg. 1999. “Distribution of Power Within
the Household and Child Health.” A draft paper intended to transmit preliminary
results of IFLS research.
Thornton, A. and H. S. Lin. 1994. “Social change and the family in Taiwan.”
Chicago: University of Chicago Press.
Tinker, I. 2007. “Boserup, Ester (b. 1910),” in D.A. Clark, eds., The Elgar
Companion to Development Studies. Massachusetts, USA: Edward Elgar
Publishing Inc.
United Nations. 1988. First marriage: Patterns and determinants. New York: United
Nations.
United Nations. 2010. Population Division of the Department of Economic and Social
Affairs of the United Nations Secretariat, World Population Prospects: The
2008 Revision. Accessed online at http://esa.un.org/unpp
United Nations Children’s Fund (UNICEF). 1998. The State of the World’s Children.
Accessed Online at http://www.unicef.org/sowc98/fig5.htm
United Nations Children’s Fund (UNICEF). 2007. The State of the World’s Children:
Women and Children -The Double Dividend of Gender Equality. UNICEF,
New York.
189
United Nations Children’s Fund (UNICEF). 2012. “At a Glance: Indonesia.”
Accessed online at
http://www.unicef.org/infobycountry/indonesia_statistics.html#73 on January
23
rd
, 2012
United Nations Development Program (UNDP). 2007. Human Development Report
2007-2008. New York: UNDP.
United Nations Development Program (UNDP). 2011. Human Development Report
Accessed online at http://hdr.undp.org/en/statistics/ on December 22
nd
2011.
United Nations Population Fund (UNFPA). 2011. “Reproductive Health and Rights is
Fundamental for Sound Economic Development and Poverty Alleviation.”
Accessed online at http://www.unfpa-bangladesh.org/pdf/success_05.pdf on May
2nd, 2011
United States Department of State. 2012. “Bureau of East Asian and Pacific Affairs-
Countries and Other Areas-Indonesia.” Accessed online at
http://www.state.gov/p/eap/ci/id/ on January 23
rd
, 2012
Visaria, Pravin M. 1969. “The Sex Ratio of the Population of India” Census of India
1961 - 1(10).
Waterlow, J., R. Buzina, W. Keller, J. Lane, M. Nichaman, J. Tanner. 1977. “The
Presentation and Use of Height and Weight Data for Comparing the Nutritional
Status of Groups of Children under the Age of Ten Years.” Bulleting of the
World Health Organization 55:489-98.
Williams, L. B. and P. Guest. 1985. "Female status as a determinant of contraceptive
use." Paper presented at the 1985 meeting of the Population Association of
America, Boston (28-30 March).
Williams, L. B. 1989. “Postnuptial Migration and the Status of Women in
Indonesia.” Journal of Marriage and Family, 51(4), pp. 895-905.
Wolf, D. 1991. Factory Daughters: Gender, Household Dynamics and Rural
Industrialization in Java. Berkeley: University of California Press.
World Bank. 1982. World Bank Development Report. New York: Oxford University
Press.
World Bank. 2001. Engendering development through gender equality in rights,
resources, and voice. World Bank Policy Research Report. Washington, D.C.
190
World Health Organization (WHO). 1995. “Physical Status: The Use and
Interpretation of Anthropometry. “ Report of a WHO Expert Committee,
World Health Organization Technical Report Series No. 854. Geneva.
World Health Organization (WHO). 1995. “Maternal Anthropometry and Pregnancy
Outcomes: A WHO Collaborative Study.” Bulletin of the World Health
Organization 73 (Suppl).
World Health Organization (WHO). 2000. “Turning the Tide of Malnutrition:
Responding to the Challenge of the 21st Century.” Geneva: WHO/NHD/00.7.
Accessed online at www.who.int/mip2001/files/2232/NHDbrochure.pdf on
October 1st, 2011
World Health Organization (WHO) and United States Agency for International
Development (UNAIDS). 2007. Consultation on STI Interventions for
Preventing HIV: Appraisal of the Evidence. Geneva: World Health Organization.
World Health Organization (WHO) and United Nations Children’s Fund (UNICEF).
2009. “WHO Child Growth Standards and the Identification of Severe Acute
Malnutrition in Infants and Children” Accessed online at
http://www.who.int/nutrition/publications/severemalnutrition/9789241598163/
en/index.html on March 23rd 2012
World Health Organization (WHO). 2011. Indonesia and Family Panning: An
Overview. Accessed online at
http://www.searo.who.int/LinkFiles/Family_Planning_Fact_Sheets_indonesia.pdf
Zaman, H. 1995. “Patterns of Activity and Use of Time in Rural Bangladesh: Class,
Gender and Seasonal Variations.” The Journal of Developing Areas, 29, 371-388.
191
Appendix-A: Additional Tables for Chapter Four
Table A1: Pooled Sample F test / Chi-square statistics (p values in parentheses)
Table A2: Joint Tests of Significance for Categorical Variables for Table 4.3 to 4.8
Decision
Making
Additive
Scale (0-5)
Woman's
Own
Health
Large
Purchases
Daily
Purchases
Visits to
Family and
Friends
Child
Health
Care
F-Statistic 3.57*** 55.28*** 55.82*** 40.11** 46.75*** 46.22***
P value (0.0000) (0.0000) (0.0000) (0.0048) (0.0006) (0.0008)
Notes: ***p<0.001, **p<0.01, *p<0.05, +p<0.10
Source: BDHS
1999
Decision
Making
Additive
Scale (0-5)
Woman's
Own
Health
Large
Purchases
Daily
Purchases
Visits to
Family and
Friends
Child
Health
Care
Woman's age in categories 64.19*** 62.51*** 80.95*** 99.12*** 107.93*** 74.23***
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)
Woman's education 5.89*** 12.54** 19.46*** 15.14** 12.2** 16.44***
(0.0005) (0.0058) (0.0002) (0.0017) (0.0067) (0.0009)
Partner's education 1.22 3.11 3.69 2.22 2.64 1.94
(0.3000) (0.3745) (0.2971) (0.5280) (0.4509) (0.5850)
Religion 7.51*** 10.35* 7.49+ 11.92** 6.97+ 16.66***
(0.0001) (0.0158) (0.0577) (0.0077) (0.0729) (0.0008)
Divisions 21.15*** 48.52*** 102.8*** 105.04*** 82.96*** 51.97***
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)
2007
Woman's age in categories 37.55*** 32.42*** 72.18*** 127.14*** 77.27*** 26.54***
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)
Woman's education 8.61*** 19.39*** 9.6* 10.73* 8.35* 18.93***
(0.0000) (0.0002) (0.0223) (0.0133) (0.0394) (0.0003)
Partner's education 3.3* 1.13 12.83** 13.55** 11.39** 3.91
(0.0196) (0.7701) (0.0050) (0.0036) (0.0098) (0.2714)
Religion 1.00 6.04 7.63* 1.42 3.41 2.87
(0.3902) (0.1099) (0.0543) (0.7000) (0.3331) (0.4120)
Divisions 29.15*** 21.52*** 136.21*** 137.98*** 91.4*** 99.05***
(0.0000) (0.0006) (0.0000) (0.0000) (0.0000) (0.0000)
Notes: ***p<0.001, **p<0.01, *p<0.05, +p<0.10
Source: BDHS
192
Appendix-B: Additional Tables for Chapter Five
Table B1: Pooled Sample Chi-Square test for Table 5.3 and 5.4
Table B2: Joint Tests of Significance for Categorical Variables for Table 5.3 and 5.4
Children's Health Large Purchases
Chi-Square
47.45* 82.95***
P value [0.0297] [0.0000]
Notes: ***p<0.001, **p<0.01, *p<0.05
Source: IFLS
Child Health 1997 2007
Woman's age in categories 49.51*** 89.32***
[0.0000] [0.0000]
Woman's education level 5.15 7.71*
[0.0762] [0.0211]
Woman's share of household assets 31.13*** 31.58***
[0.0000] [0.0000]
Household headship 22.38*** 19.35***
[0.0000] [0.0001]
Woman's parent's position in community 16.56** 5.70
[0.0054] [0.3367]
Partner's Education 9.78* 6.86
[0.0205] [0.0766]
Religion 3.03 3.55
[0.5528] [0.3148]
Ethnicity 24.11** 28.74***
[0.0041] [0.0007]
Large Expensive Purchases 1997 2007
Woman's age in categories 1.50 8.39*
[0.4732] [0.0151]
Woman's education level 2.76 4.48
[0.2518] [0.1063]
Woman's share of household assets 21.03*** 60.67***
[0.0003] [0.0000]
Household headship 24.99*** 37.78***
[0.0000] [0.0000]
Woman's parent's position in community 5.61 17.62**
[0.3459] [0.0035]
Partner's Education 3.70 3.96
[0.2954] [0.2656]
Religion 4.43 4.28
[0.3511] [0.3695]
Ethnicity 54.98*** 36.45***
[0.0000] [0.0000]
Notes: ***p<0.001, **p<0.01, *p<0.05
Source: IFLS
193
Table B3: Pooled Sample Chi-Square test statistics for Tables 5.8 and 5.9
Table B4: Joint Tests of Significance for Categorical Variables for Table 5.8 and 5.9
Children's
Health
Large
Purchases
Chi-Square 156.46*** 23.71*
P value (0.0000) (0.0140)
Notes: ***p<0.001, **p<0.01, *p<0.05
p-values are in parenthesis
Source: IFLS
Bangladesh
Children's
Health
Large
Purchases
Woman's age in categories 28.94*** 53.43***
(0.0000) (0.0000)
Woman's education level 14.90*** 10.30**
(0.0006) (0.0058)
Partner's education 3.60 6.73
(0.3081) (0.0809)
Indonesia
Woman's age in categories 359.35*** 60.59***
(0.0000) (0.0000)
Woman's education level 10.51** 6.05*
(0.0052) (0.0486)
Partner's education 9.71* 2.70
(0.0212) (0.4398)
Notes: ***p<0.001, **p<0.01, *p<0.05
p-values are in parenthesis
Source: IFLS
Abstract (if available)
Abstract
The goal of this dissertation is (1) to analyze how the determinants of women’s decision-making power and child stunting have changed over time and (2) to identify the sources of improvement in women’s decision-making and child health in two less-developed Muslim countries. According to past literature, the agents of change in women’s decision-making at the household-level should derive from changes in women’s education and wage employment. Sources of change in child health should derive from parental education, household wealth and community infrastructure. ❧ This dissertation contributes to the current literature by looking at change in women’s status and its determinants over time, and its implications for child health in one of the most gender biased and malnourished societies in the world—Bangladesh. It also examines the role of social context on women’s status by comparing two predominantly Muslim societies—Bangladesh and Indonesia that had similar origins in terms of economic and social development. ❧ In the first paper, I ask two questions: (1) what factors predict women’s decision-making in Bangladesh in 1999 and 2007? and (2) do increases in women’s socioeconomic characteristics contribute to improvements in women’s decision-making in Bangladesh from 1999 to 2007? I use logistic regression and Oaxaca-Blinder decomposition to answer these questions. The results indicate that, the strength of the relationships between predictors and women’s decision-making has changed over time. And the mean increase in women’s socioeconomic status only has a modest effect on decision-making. Much of the difference between 1999 and 2007 is accounted for by the intercept, which may have subsumed unobserved variables that represent change in norms and attitudes towards women and economic development. ❧ In the second paper, I attempt to answer the same questions in Indonesia. The results indicate that improvements in women’s decision-making from 1997 to 2007 are partly due to improvements in women’s socioeconomic status, but similar to Bangladesh, the majority of the improvement derives from the differences in intercepts. In the second part of this paper, I harmonize data from Bangladesh and Indonesia to examine the role of social context on women’s decision-making. The results suggest that the underlying cultural context in Indonesia and Bangladesh plays an important role in determining women’s status regardless of their socioeconomic characteristics. Ethnographic evidence from Indonesia and Bangladesh on family systems and traditions support these results. Therefore, it can be concluded that social context is an important factor (net of socioeconomic characteristics) in women’s decision-making. ❧ In the third paper, I look at the impact of women’s socioeconomic status on child stunting in Bangladesh, where child malnutrition is widespread and detrimental to development. In recent years, the number of children who are stunted has been significantly reduced. The results from Oaxaca-Blinder decompositions reveal two things: 1) improvements in women’s education (net of partner’s education), delayed first marriage, reduction in fertility (i.e. the lower birth order of the child) and access to information through television viewership account for about half of the decline in child stunting. The other half is accounted for by improvements in household wealth, availability of electricity, drinking water and advanced sewerage systems. Because parents favor sons over daughters in Bangladesh, the effect of parent’s education on stunting is weaker for girls than for boys. In fact, the majority of the reduction in stunting for girls comes from improvements in household assets and community infrastructure that are usually available for all children in the household regardless of their gender. ❧ These results partly support the notion that increases in women’s socioeconomic characteristics have an impact on improvements in decision-making and child nutrition in Bangladesh and Indonesia. But other community level changes such as norms and attitudes towards women, improvements in infrastructure and wealth, which usually accompany modernization and economic development, also play an important role in the improvements of women’s decision-making and child nutrition. From a welfare policy point of view, these results are encouraging
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Diverging fertility patterns? Racial and educational differences in fertility behaviors and their implications for socioeconomic mobility
PDF
Essays on education: from Indonesia to Los Angeles, CA
PDF
Intergenerational transmission of values and behaviors over the family life course
PDF
Essays on health economics
PDF
The measurement, life course patterns, and outcomes of intergenerational ambivalence among parent-adult child dyads
PDF
Essays on the empirics of risk and time preferences in Indonesia
PDF
Three essays on economics of early life health in developing countries
PDF
Essays on family and labor economics
PDF
Essays in health economics: evidence from Medicare
PDF
Behabioral and neural evidence of state-like variance in intertemporal decisions
PDF
Marital matching in West Africa: an examination of interethnic and interreligious first marriages in Benin
PDF
An empirical evaluation of breast cancer treatment decision making
PDF
Life on the sidelines: the academic, social, and disciplinary impacts of male high school sports participation in California
PDF
The effects of time and space on health status of older adults in China
PDF
Gender, family, and Chinese nation: ""leftover women,"" ""foreign f women,"" and ""second wives""
PDF
Sequential decisions on time preference: evidence for non-independence
PDF
The color of success: African American and Japanese American physicians in Los Angeles
PDF
Social determinants of physiological health and mortality in China
PDF
Where will your path lead? Military services, career paths, and life course outcomes: implications for social mobility
PDF
The global market for wombs: a study of the transnational surrogacy industry in Mexico
Asset Metadata
Creator
Jayasundera, Radheeka Ranmali
(author)
Core Title
Improvements in women's status, decision-making and child nutrition: evidence from Bangladesh and Indonesia
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Sociology
Publication Date
06/27/2012
Defense Date
06/14/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
Bangladesh,child nutrition,child stunting,Decision making,Indonesia,OAI-PMH Harvest,women's status
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Lynne Casper (
committee chair
), Biblarz, Timothy (
committee member
), Emeka, Amon (
committee member
), Silverstein, Merril (
committee member
), Strauss, John A. (
committee member
)
Creator Email
jayasund@usc.edu,radikajay@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-50211
Unique identifier
UC11290053
Identifier
usctheses-c3-50211 (legacy record id)
Legacy Identifier
etd-Jayasunder-908.pdf
Dmrecord
50211
Document Type
Dissertation
Rights
Jayasundera, Radheeka Ranmali
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
Tags
child nutrition
child stunting
women's status