Whose voice matters? An examination of gender bias in intra-household decision making This version: September 2015 Abstract It has been suggested that offering microcredit to women to empower them may be ineffective as women borrowers hand over the control of loans to their husbands. We thus conduct a lab-in-the-field experiment to examine whether gender bias exists in intra- household decision making in rural Bangladesh. The experiment mimics a real-life scenario where microcredit was offered to either the wife or the husband in a household and the borrower could decide whether to make his/her own investment choice or to transfer the decision-making to the spouse. We find that women are more likely to let their spouses make decision, compared with their male counterparts. Different treatments in the experiment also allow us to test and quantify the underlying causes of the bias. Our findings show women’s decision to transfer the decision-making is driven both by their lower decision-making power and their belief that their spouses are more capable of making financial decisions. We also look at subjects’ control over use of earnings from the investment and find offering credit to women did not improve their control, irrespective of whether or not they let their spouses make the investment decision. Keywords: Gender bias, intra-household bargaining, field experiment. JEL Classification: C91, C93, D13, O12
49
Embed
Whose voice matters? An examination of gender …Whose voice matters? An examination of gender bias in intra-household decision making This version: September 2015 Abstract It has
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Whose voice matters? An examination of gender bias in intra-household decision making
This version: September 2015
Abstract
It has been suggested that offering microcredit to women to empower them may be ineffective as women borrowers hand over the control of loans to their husbands. We thus conduct a lab-in-the-field experiment to examine whether gender bias exists in intra-household decision making in rural Bangladesh. The experiment mimics a real-life scenario where microcredit was offered to either the wife or the husband in a household and the borrower could decide whether to make his/her own investment choice or to transfer the decision-making to the spouse. We find that women are more likely to let their spouses make decision, compared with their male counterparts. Different treatments in the experiment also allow us to test and quantify the underlying causes of the bias. Our findings show women’s decision to transfer the decision-making is driven both by their lower decision-making power and their belief that their spouses are more capable of making financial decisions. We also look at subjects’ control over use of earnings from the investment and find offering credit to women did not improve their control, irrespective of whether or not they let their spouses make the investment decision.
Keywords: Gender bias, intra-household bargaining, field experiment.
JEL Classification: C91, C93, D13, O12
1. Introduction
Microfinance programs in developing countries have been commonly targeted at
lending to women, with the twofold objective of sustainable repayment rate and
women empowerment. According to Duflo (2012), there is a two-way relationship
between economic development and women empowerment. In one direction,
economic development can reduce gender inequality and in the other direction,
empowering women can stimulate economic growth. In developing areas like rural
Bangladesh, women in most poor households remain inactive, not participating in any
productive activities. This seems a paradox in the context of Bangladesh, where there
is no significant gender gap in education and many women are even better educated
than their spouses, thanks to the various programs targeting girls in rural areas such as
female stipend program to secondary school girls in rural areas (Begum, Islam and
Smyth 2014). The fact that women’s income-earning potential has not been realized is
usually attributed to social gender bias that restricts women from access to economic
resources and opportunities. Microfinance is thus claimed to provide women with
better access to economic resources, so they can participate in income-earning
activities and contribute to household’s income. At the same time, microfinance also
aims to empower women to improve their mobility, bargaining power and decision
making, so reducing the social gender inequality.
Whether those objectives of microfinance can be achieved in reality remains a big
question. The findings from empirical studies are mixed. Some provided positive
outcomes of women empowerment, such as increase in women’s participation in
decision making, ownership of household assets, freedom and mobility, political
2009), and ownership of assets (Agarwal, 1994, Folbre, 1984, Kabeer, 1999) also
dictates the inferior position of women in the decision-making. In the context of rural
Bangladesh societies women are mostly confined by social and cultural norms to
limited choices of occupation within their households. Specifically, the institution of
purdah promotes the seclusion of women and enforces their exclusion from public
spaces, thus preventing them from employment opportunities outside their households
(Amin, 1997, Kabeer, 2001). Moreover, women in Bangladesh commonly possess
less unearned assets than men. In Bangladesh Islamic law specifies daughters have the
right to inherit half a son’s share of the father’s property (Cain, 1978). However, in
practice daughters receive substantially less than what they are entitled to, mostly in
the form of jewellery from their dowries while sons mostly inherit land (Anderson &
Eswaran, 2009). These factors are thus expected to reinforce the inferior role of
women in intra-household decision making. We thus test the following hypothesis:
H1: Women feel obliged to transfer the control of their loan to their spouses due to
their lower decision-making power. Women thus transferred the decision-making
when there is no asymmetric information between them.
However, other anecdotes and theoretical studies in the literature also suggest
women’s decision to transfer the control of the loan to their spouses is voluntary.
Kabeer (2001) presented testimonies from women who consider conformity with
purdah as “a voluntary adherence to status norms rather than as a direct manifestation
of male control” and women who chose to transfer control over their loans to male
household heads in recognition of their responsibility for the collective welfare of the
household. The paper also pointed to the effect of financial and entrepreneurial
competence in women’s decision-making. Montgomery et. al. (1996) and Goetz &
Gupta (1996) argued that women transferred control of loans to men to have greater
expenditures on their own or consumption needs of the children and the whole family.
A number of theoretical studies showed women concede control of loan to their
spouses for strategic reasons or for the welfare of the household. For example, Tassel
(2004) developed a dynamic bargaining model where loan repayment is required for
continuing borrowing in the next period and borrowers have two investment options
with same expected payoff but different risk profile. In this model, female members
are faced with limited income generating opportunities, thus always choose safer
investment projects, in order to improve their bargaining power in the household. On
the contrary, male members prefer riskier project to protect their bargaining power
against their spouses, even when aggregate household consumption is expected to
increase by the same amount in both investment options. However, in the equilibrium
women would transfer control of their loan to their husband to ensure that he would
help with loan repayment and credit access is not terminated. This transfer of control
actually benefits them even if the investment ends up not being her first choice.
Similarly, but the model by Ligon (2011) based on the assumption of risk averse
agents who aimed at consumption smoothing instead of risk neutrality in Tassel
(2004). To explain the stylized fact that women usually relinquish their microcredit
loans to their husbands, the model shows that women can benefit from letting their
husbands undertake most of the risk from investing and repaying loans. A more recent
paper by Ngo & Wahhaj (2012) introduced a sphere of joint production and
household public good and found heterogeneous impacts across households. Access
to credit may not improve women’s bargaining power if they have limited skills to
engage in an autonomous productive activities, or if they have sufficient skills to do
so but their husband wants to appropriate the loan to maintain their own bargaining
power. We are motivated by these findings to develop the following hypotheses to
test under our experimental design:
H2: Women transferred the control of their loans to their spouses voluntarily. Women
thus transferred the decision-making even under asymmetric information, where they
could hide their options and decisions from their spouses.
H2a: Women transferred the control of their loans to their spouses voluntarily due to
their perceived inferior financial competence.
H2b: Women transferred the control of their loans to their spouses voluntarily to have
greater expenditures on their own or on children and common uses.
In addition to intra-household dynamics as discussed, the literature on gender
difference provides an alternative explanation for women’s transfer of control over
loans. Gender differences have been widely studied in lab experiments, among which
most related to our studies include those that examined differences in risk-taking
behaviour1 and self-confidence2. Women are generally found to be more risk averse
and less self-confident. In our context where women are given an opportunity to make
an investment decision that involves risk, women’s avoidance of risk and lower self-
1 See Eckel & Grossman (2002) for a review of literature on gender difference in risk-taking behavior. 2 See Niederle & Vesterlund (2011) for a review of literature on gender difference in competiveness and self-confidence.
confidence could make women inclined to transfer the control of loans to men in
general but not necessarily their spouses. This motivates our next hypothesis:
H3: Women transferred the control of their loans to their spouses voluntarily due to
their risk avoidance and lower self-confidence. Women thus transferred the decision-
making to any men but not necessarily their spouses.
Aside from women’s control over their loans, another aspect of women empowerment
through microfinance is the improvement of women’s control over household
expenditures. Pitt & Khandker (1996) and Hashemi, Schuler, & Riley (1996) are
among studies that provide empirical evidence on positive impact of offering credit to
women on women’s purchasing power. However, these findings remain questionable
due to the potential selection bias related to microfinance program placement and
participation choice. We thus develop the following hypothesis to test whether
offering credit to women improve their decision-making in household’s spending,
irrespective of their decision-making in loan uses.
H4: Women who are offered credit to invest are more likely to keep control of
household expenditure rather than concede the control to their spouses, compared
with those who are not offered credit.
We are also interested to look at whether there is any correlation between women’s
control over loan uses and their control over household expenditures. As Kabeer
(2001) suggested, conflicting conclusions about the impact of credit in empowering in
the literature reflect differences in the questions asked by different studies,
particularly between those that focused on processes of loan uses and those that
focused on outcomes associated with access to loans. We thus test the following
hypothesis:
H5: Women who are offered credit to invest and make their own investment decision
are more likely to keep control of household expenditure rather than concede the
control to their spouses, compared with those who are offered credit but transferred
the control of investment to their spouses. As a result, household expenditures are
spent more towards women’s consumption or children and household common uses.
3. Experimental Design
3.1. Experimental setting
The experiment was conducted with 826 married couples of 18-55 years old in 26
rural villages in three upazilas (Assasuni, Koyra, and Paikgacha) of the Khulna
district in June-July 2014. The map of the villages is shown in Figure 1. Recruiters
randomly went door to door and invited 832 respondents and their spouses3 to a study
on the understanding of financial matters. Each subject would receive a 100 taka
(approximately USD1.5 and an adult’s average daily wage) fee for show up and have
the opportunity to earn more money. The recruiters also conducted a household-level
survey that gathers information on general household characteristics.
In each village we selected a local school as the experiment venue. Only one
experiment session was conducted in each village, to avoid any contamination of the
experiment through information leakage. We conducted the experiment on two
treatment groups (of six treatments in total) at the same time in each session. The
treatment groups were randomly assigned across villages. Each session consisted of
31-32 couples, so each treatment group in each village has 15-16 couples. An
individual-level survey was conducted for each spouse privately and separately after
the experiment was completed so that the survey questions did not prime subjects
3 There were six couples who were invited and surveyed but did not show up on the experiment day. Since the attrition rate is only 0.7%, we see no need to perform any attrition test.
about the purpose of the experiment. The survey questions include subjects’
awareness and mobility, earnings and assets, household finance and decision-making
matters, understanding of risk, and other individual preferences.
3.2. Experiment procedure
The game involves each subject being matched with another subject of the opposite
gender to be his/her game partner and both having to make separate investment
choices for the chance of earning money for both himself/herself and his/her partner.
The game procedure is illustrated in the game tree in Figure 2. Each subject was
endowed with 300 taka at the beginning of the game. In each pair one partner played
the role of the “borrower” and the other partner the “spender”. The borrower was
given the opportunity to invest his/her own 300 taka and his/her partner’s 300 taka.
They could either invest in a Safe lottery or a Risky lottery4. The lotteries’ payoffs
and risk are presented in table 1. The lotteries have the same set of events: TRIPLE,
KEEP, and LOSE, in order to ensure any outcome from choosing a lottery would be
possible to be obtained by choosing the other lottery. This feature is critical for the
design of our treatments, which is discussed in section 3.3. TRIPLE means the payoff
triples the initial investment, so each player would receive 900 taka. KEEP means the
payoff remains the same as the initial investment, which is 300 taka. LOSE means the
player would receive zero, losing their initial investment. The risky lottery has higher
expected payoff but also higher risk. Given the lotteries are relatively complicated to
our subject pool, we provided visual demonstration and elaborate training and
practices before the subjects made decision. We explained the lotteries by showing
4 In the real experiment, we called the Safe lottery Paan and the Risky lottery Supari. Paan and Supari are the betel leaf and betel nut that are usually chewed together by the locals in Bangladesh (and other South Asian countries) for stimulant effect. Since they are generally consumed together, using their names as the lottery names could prevent subjects from having preference for one lottery over the other irrespective of their payoffs and risk profiles.
see-through bottles that contained balls in three different colours to reflect three
different payoffs.
After making decision on which lottery to invest in, the borrower was asked to make
decision on whether he/she wants to use his/her own lottery choice regardless of what
his/her partner chose or to use the partner’s choice if the partner’s choice was
different from his/her own choice. The borrower had to make this transfer decision
without discussing with the spender or knowing the spender’s choice, therefore we
would know for sure by whom the final decision was made. Since intra-household
interaction and communication between spouses are extremely complex and generally
unobservable, this knowledge would be impossible to obtain outside our controlled
experiment environment. These instructions were literally the same as asking them to
choose between using their own lottery choice or using their partners’ choice
regardless of what their partners chose. However, we intentionally phrased it that way
to highlight the potential conflict where the partners made different lottery choices
from each other. Subjects might not take the decision to transfer the investment
seriously if they were not made aware of the potential conflict, expecting the spouses
would make the same decision as theirs.
At the same time the spender was also asked which lottery he/she would choose if
he/she was to invest his/her endowment and his/her partner’s endowment. The final
payoff, which was based on the actual result of the lotteries, was determined
according to these choices from both partners and distributed equally between them.
The post-game survey included a question that asks on which purpose the subjects
planned to spend the game payoff. The responding options are: (1) to keep for
him/herself for later use, (2) to give to his/her spouse, (3) to buy something for
him/herself, (4) to buy something for his/her spouse, (5) to buy something for children
or common use, and (6) others. We also did home visit to all subjects’ two weeks
after the experiment day and asked them on which purpose they actually spent the
game payoff. The options are: (1) keeping for him/herself for later use, (2) gave to
his/her spouse, (3) bought something for him/herself, (4) bought something for his/her
spouse, (5) bought something for children or common use, and (6) others.
3.3. Experiment treatments
The recruited subjects were randomly assigned into one of six treatment groups,
which are different in two dimensions. In the first dimension, the treatments differ by
the roles of each gender: (1) the female partner plays the role of the borrower and the
male partner the role of the spender or (2) the female partner plays the role of the
spender and the male partner the role of the borrower.
There are three conditions in the second dimension. In the No secret condition,
subjects and their spouses were in the same room but were separated: all men were on
one side and all women were on the other side of the room. Therefore, instructions for
each partner were fully disclosed to the other. Subjects were told that their spouses
would be their game partners. They were also informed that all the choices they made
would be revealed to their spouses after they all made decision. However, men and
women were strictly prohibited from talking to each other, thus no discussion was
allowed between spouses and spouses made decision without knowing the choice(s)
of each other. In the Secret condition, men were in one room and women were in
another room. The instructors gave different instructions to men and women
separately. The subjects were not informed of any instructions given to their spouses
and any decisions their spouses made. They were also told that all their options and
their decisions would be kept confidential and private from their spouses. Even
though it might be impossible to hide the money received after the game finished in
the household context, borrower in the Secret condition could always hide his/her
choices from the spender due to two features of our experiment: (1) the lotteries have
the same possible payoffs and only differ in the possibility of each payoff and (2) the
lotteries were carried out separately to determine each couple’s payoff by the
instructors without the presence of the couple.
To examine whether women are more likely than men to pass the control of the
investment to their spouse, we compare the female borrower’s decision on whose
lottery choice to use with that of the male borrower under the ‘no secret’ condition
and the ‘secret’ condition, separately. Moving from the secret condition to the no-
secret condition can show us the net effect of imbalance in decision-making power
between wife and husband. If one has a lower decision-making power than his/her
spouse and fears of the spouse being aware of him/her disregarding the spouse’s
choice, he/she would be more likely to transfer the control of the investment under the
no-secret condition than under the secret condition. However, if the borrower
transferred the control of the investment even under the secret condition, he/she must
have believed that his/her spouse would make better investment decision and thus
would have been willing to transfer the control. Therefore, any gender difference in
the borrower’s decision to transfer the control of the investment could suggest the
effect of imbalance in financial capacity between spouses.
While the ‘secret’ and ‘no secret’ conditions allow us to isolate the effect of
imbalance in decision-making power and imbalance in financial capacity, we could
not determine whether the bias is between-spouse specific or generally between men
and women. Men and women might be naturally different and in our game context
one may expect a subject of the other gender can make better investment decision. Or
social norm could build the expectation that making such important decisions as
where money should be invested is the role of one gender but not of the other. We
thus introduced another condition where each subject was randomly and anonymously
matched to another subject of the opposite gender that was not his/her real-life
spouse. The subjects were made fully aware of this matching condition. Other than
this condition the additional treatments have the same conditions as the secret
treatments. This treatment therefore can control for any effect of the general gender
bias that exists outside the household.
4. Descriptive statistics
Table 2 summarizes the demographic characteristics among secret, no-secret, and
random-couple treatments. The treatments are generally balanced in most of the
characteristics, except for a slight difference in the number of household members
between the secret and no-secret treatments. The households are relatively poor
compared with the average Bangladesh rural population, since we target on
households that are targeted by microfinance programs. The annual income per capita
in the three treatments range between 15,000 and 16,000 taka, compared with the
national level of 25,560 taka (HIES Survey Report, 2010). The schooling gap between
husband and wife is relatively small.5
In table 3, we report a number of indicators of subjects’ social capital and mobility by
gender. All of the indicators show a significant gender gap. Women seem less
exposed to social media and social gatherings than men. Women are less frequent at
public places and places outside their villages and are also more likely to seek
5 This is partly attributed to the Female Secondary School Stipend Program, which was introduced nationwide in 1994 (see, for example, Begum, Islam and Smyth 2014), and other programs targeting girls in rural Bangladesh to address the gender imbalance in schooling at that time.
permission of their spouses to go to these places. While most women (82%) earn
income from home activities, only 13% of them do any income-earning works outside
their home in the last year. On the contrary 65% of men have income from work
outside home. Women also earned much less than men.
Figure 3 shows who is/are the main decision maker in different family matters, based
on subjects’ responses to our survey questions. Although the questions are asked to
wife and husband individually and separately, there is generally not much discrepancy
between their answers. The decision-making role seems to be dominated by men in all
aspects, except in the spending for poultry. As poultry farming is commonly women’s
main and only income-earning activity, this could be the only area where women can
have more control than their spouse. Women do not have much say even in the use of
their own earnings, in which only 16% of women can make their own decision.
5. Results
5.1. Decision making in investment
We first look at the likelihood to transfer the decision making of the lottery choice by
gender in the three treatments. Table 4 reports the transfer rate by gender and
treatment at the village level. The male (female) transfer rate is defined as the number
of male (female) subjects in the village who played as ‘the borrower’ and chose to
transfer the decision making of the lottery choice to their partners divided by the total
number of participating couples in that village. The female-male transfer rate is the
difference between the female’s and the male’s transfer rate. We run the Fisher two-
sample randomization test to compare the transfer rates between treatments. We find
the female transfer rate is significantly higher than the male transfer rate in both the
non-secret and secret treatment but not in the random-couple treatment. The average
female transfer rate is relatively high at 65.3% in the non-secret treatment and 43% in
the secret treatment, compared with the male transfer rates of 21.3% and 25.2%
respectively. There is no statistically significant difference between males and
females in the random-couple treatment and both transfer rates are less than 10%. The
female transfer rate and the female-male transfer rate both are significantly higher in
the non-secret treatment than in the secret treatment while the male transfer rate is not
different between the two treatments.
Table 5 shows the probabilities of transferring the decision making at the individual
level. We perform Fisher exact test to compare the difference between treatments.
The probabilities and the one-sided p-values from the test reflect the same results as
the village-level findings. In table 6, we report results from the probit6 regressions that
control for various household characteristics and district fixed effect. We run the
where 𝑇𝑟𝑎𝑛𝑠𝑓𝑒𝑟!" is a dummy variable which equals to one if subject i transferred the
decision making of the lottery choice to his/her partner, and equals to zero otherwise.
𝐹𝑒𝑚𝑎𝑙𝑒! is a dummy variable that indicates the gender of subject i. 𝑁𝑜𝑆𝑒𝑐𝑟𝑒𝑡! and
𝑅𝑎𝑛𝑑𝑜𝑚! are dummy variables that indicate whether subject i was in the no-secret
treatment and the random-couple treatment respectively. The secret treatment is the
base value and thus the coefficient 𝛽 captures the gender effect on decision-making
transferring in the secret treatment. The coefficients 𝜃 and 𝜇 show the difference in
the gender effect of the no-secret treatment versus the secret treatment and the
random-couple treatment versus the secret treatment respectively. 𝑋! is a vector of
subject i’s household characteristics, which are wife’s age, husband’s age, wife’s 6 We also run OLS regressions and the results are similar to those from the probit regressions.
schooling, husband’s schooling, religion, number of household members, and
household annual income per capita. 𝑣! captures the village fixed effect.
The regression results are similar across different specifications and offer the same
implication as the results from the Fisher tests. Women in the secret treatment were
14-15% more likely to transfer decision-making than the male counterparts. In the
secret treatment where each spouse’s decision is not disclosed to the other spouse, we
expect almost no effect of the power imbalance (or at least less than in the no-secret
treatment), thus women who let their spouses decide on the lottery choice in the secret
treatment most likely were willing to do so in the absence of any threat. This finding
is consistent with H2. The gender gap is wider in the no-secret treatment than in the
secret treatment by 21%. There is no difference between men in the secret treatment
and those in the non-secret treatment, thus the condition of information between
spouses only affects decision to transfer made by the wife but not the husband. The
wife was more likely to let her husband make the investment decision once all her
options and decisions would be revealed to her husband. This suggests the effect of
the decision-making power imbalance, in which women, but not men, feel threatened
by their spouses being aware of them making own decision. This thus supports H1.
Both men and women were less likely to transfer decision-making in the random-
couple treatment than in the secret treatment. And the gender gap was absent in the
random-couple treatment as shown by the coefficient 𝜇. This proves the difference in
the transfer decision between the wife and the husband is not driven by the belief that
men are generally better than women in making investment decision but most likely
due to household-specific factors. H3 is thus rejected.
To test the hypothesis that women’s willingness to transfer decision making to their
spouses is driven by their belief in the capacity imbalance between themselves and
their spouses, we run the following regression in the secret treatment and the non-
where 𝐺𝑎𝑝! is either the measure of social capital gap between spouses or the
earnings gap between them. The social capital variable is a measure of how much the
subject is exposed to social media and social network and is based on subject’s
responses to our three survey questions that ask whether the subject (1) reads the
newspaper at least once a week, (2) listens to the radio or watch TV at least once a
week, and (3) participates in any club/committee/meeting group at least once a month.
Each positive response adds one-third to the social capital measure, thus the measure
can take the value between zero and one. The social capital gap variable is the
difference between subject’s social capital measure and that of his/her spouse. The
social capital gap between the husband and his wife has a mean of 0.15 and a standard
deviation of 0.4. Similarly, the earnings gap variable is the difference between
subject’s annual earnings and that of his/her spouse. The earnings gap between the
husband and his wife has a mean of 61,500 taka and a standard deviation of 33,900
taka. Thus in general husbands earn more and have a higher level of social capital
than their wives. 𝑀𝑜𝑡ℎ𝑒𝑟𝑖𝑛𝑙𝑎𝑤! is a dummy variable that equals to one if the
husband’s mother stays with the couple and equals to zero otherwise. In
approximately 24% of the households in our experiment the husband’s mother lives
with the couple. We interact this dummy variable with the gender dummy variable
since we expect no similar effect of the wife’s mother (or father) staying with the
couple and the wife’s parent(s) stays with the couple in only less than 1% of the
households.
Tables 9A and 9B report the results from regressions (3) and (4) in the no-secret
treatment and the secret treatment separately. There is a significant correlation
between the social capital gap and the transfer decision in the no-secret treatment but
not in the secret treatment. Thus the social capital gap is correlated more with the
intra-household imbalance in decision-making power than with the intra-household
imbalance in capability. In particular, an increase of one-point in the social capital gap
between the subject and his/her spouse reduced the subject’s probability to let his/her
spouse make the lottery decision by 2.3%-points. This means a decrease of one
standard deviation in the husband-wife social capital gap could reduce the wife’s
transfer probability by 2.7%-points. Since this effect is relatively small and adding the
social capital gap variable does not change the coefficient on the gender variable
considerably, the higher social capital of the husband relative to the wife does not
explain much the gender difference in the transfer decision. We also find a significant
correlation between the earnings gap and the transfer decision in the no-secret
treatment. A decrease of one standard deviation in the husband-wife earnings gap
could reduce the wife’s transfer probability by 6.8%-points. Adding the earnings gap
variable also reduced the gender effect drastically in magnitude and statistical
significance. This supports a high correlation between the earnings gap and the
transfer decision in the no-secret treatment but not in secret treatment, thus suggesting
the earnings gap can fully explain the gap in decision-making power but not the gap
in capacity. The coefficient of the mother-in-law variable is significant in both the
secret and no-secret treatments. Having the mother-in-law live in the same household
increased the probability the wife let her spouse decide on the lottery option by 13%
in the secret treatment and 14% in the no-secret treatment. Since the coefficient is not
statistically larger in the no-secret treatment than in the secret treatment, we may
claim that the mother-in-law affects the wife’s decision-making mostly through her
belief on her capacity relatively to her spouse’s capacity. However, this may not
reflect a causal relationship but simply a correlation where households in which the
mother-in-law lives with the couple are also households in which the wife has lower
financial capacity.
5.2. Risk taking and Decision making in investment
We examine whether there is gender difference in risk taking, using subjects’ choice
between the safe lottery and the risky lottery. Women are commonly found to be more
risk averse than men in the literature. We however find men and women are not
statistically different in the choice of lottery in the whole sample (table 10). We only
find gender difference in the secret treatment: women were about 17% less risk taking
than men. The percentage of women choosing the risky lottery is higher than that of
men in the no-secret treatment, however the difference is not statistically significant.
This suggests the effect of spousal observability on subject’s risk-taking. Since
subjects in the no-secret treatment expected their choices to be disclosed to their
spouses, they might not reveal their true preference of lottery but instead choose what
they think their spouse would choose. Meanwhile subjects in the secret treatment do
not have the same incentive to hide their true preference. Assuming the lottery choice
in the secret treatment reflects subject’s own risk-taking preference, we find on
average women expect their spouses to be more risk taking and men expect their
spouses to be less risk taking than themselves (statistically significant at 1% and 5%
respectively). We also find women in the no-secret treatment are significantly more
likely to choose the risky lottery than those in the secret treatment and women’s guess
of their spouse’s lottery choice is not significantly different from their own choice.
Although the percentage of men taking the risky lottery is lower in the no-secret
treatment than in the secret treatment, the difference is not statistically significant.
This suggests the effect of spousal observability on risk-taking applies stronger to
women than to men. Men and women in the random-couple treatment were not
different in risk taking and both were less likely to choose the risky lottery than in the
real-couple treatments.
We then examine whether subject’s risk-taking behaviour affects his/her decision to
transfer the decision-making of the lottery choice. In table 11 we report the
percentage of men and the percentage of women transferring the decision-making to
their partners by their risk-taking behaviour in each treatment. In the secret and the
no-secret treatment risk-loving men were less likely to let their spouses make decision
than risk-averse men. As previously discussed men expected their spouses to be less
risk-taking than themselves, men who were risk lovers were thus more driven to stick
to their own decision rather than to let their spouse decide. We also ran the Fisher
exact one-sided test to test whether men’s choice of lottery and their guess on their
spouse’s choice are less likely to be the same for risk-loving men than for risk-averse
men. The test p-value confirms the hypothesis at 1% confidence level. We do not find
similar effect among women. There is no difference in women’s transfer decision
between risk-loving and risk-averse women.
5.3. Decision making in spending of earnings
We first examine whether giving credit to women instead of their spouses would
change women’s control of earnings. This analysis is only conducted for the secret
treatment and no-secret treatment but not the random-couple treatment. We report the
results using the combined sample of the secret treatment and the no-secret treatment
in table 12. Table 12A reports how subjects intended to use their earnings from the
game. We classify subjects by player types: the ‘borrower’ vs. the ‘spender’, and
gender. The most common use of the earnings is to spend on purchasing for children
or household’s common purposes. We run the Fisher exact text to compare the
difference in each usage type between male borrowers and male spenders, female
borrowers and female spenders, female borrowers and male borrowers, and female
spenders and male spenders. We also run the two-sample Kolmogorov-Smirnov (K-S)
test to test the difference in distribution of all usages. The K-S test is useful in this
case where there are more than two usage types. There is no significant different
between female borrowers and female spender, in all usages separately and in the
whole distribution. This suggests giving credit to women may not affect their intended
use of the earnings, particularly whether they planned to control the earnings by
themselves or let their spouses take control. There is a significant difference in the
distribution of earning usages between female borrowers and male borrowers. The
Fisher exact test shows the difference lies mostly in the ‘keep/purchase for self’ and
the ‘give to spouse’ categories. The female borrowers were more likely to give their
spouses their earnings and less likely to keep or purchase for themselves. In another
word, the female borrowers were more likely to let their spouses control their
earnings rather than make their own decision. However, we find no difference in the
distribution of earning usages between female spenders and male spenders. We only
find they are different in the probability of giving the earnings to spouses and the
probability of purchasing for spouses. The female spenders were more likely to give
the earnings to their spouses rather than purchasing for their spouses, relatively to the
male spenders. These are consistent with the findings that male spenders are different
from male borrowers. The male spenders were more likely to purchase for children or
common uses or purchasing for their spouses than to keep or purchase for themselves.
While women’s tentative usages of earnings were not affected by which spouse was
given the opportunity to invest, men were more likely to be less ‘selfish’ (tend to
purchase for other household members or household common uses than to keep or
purchase for themselves) when they were not directly given the investing opportunity.
This suggests men relate the control of earnings to the source of investment while
women do not.
Table 12B show what subjects actually did to their earnings from the game. The
information was collected two weeks after the day the experiment was conducted.
There is no significant difference between male borrowers and male spenders and
female borrowers and female spenders, in all usages separately and in the whole
distribution of usages. Thus the actual decision on usages of the earnings from the
game was not affected by who received the opportunity to invest. There is a
significant gender difference among both borrowers and spenders. The K-S test p-
value is below 5% for both the difference between female borrowers versus male
borrowers and the difference between female spenders versus male spenders. The
main difference lies in the ‘give to spouse’ and ‘purchase for children/common use’
categories. Women were more likely to give the earnings to their spouses while men
were more likely to use the earnings to purchase for children or household common
uses. This suggests women irrespective of their role in the game were more likely to
let their spouses control their earnings. We thus reject H4.
Tables 13 report the borrowers’ intentional uses and actual uses by gender and
whether they transferred the decision making to their spouses. Female borrowers are
not different in the distribution of intended uses of earnings, although the Fisher exact
test p-value shows women who did not transferred the investment control had a lower
probability of intentionally giving earnings to spouse and a higher probability of
intentionally purchasing for children or common uses, compared to those who
transferred the control. This might suggest women who make their own decision on
investment can have more control in spending. However, there is no significant
difference in the actual uses of earnings between female borrowers who transferred
the investment control and those who did not. Female borrowers irrespective of their
transfer decision were more likely to give their earnings to spouses and less likely to
purchase for children or common uses, compared to male borrowers. Therefore, H5 is
rejected. We also do the same analysis on the restricted sample that exclude female
borrowers who did not transfer the decision making and lose money from the
investment (final earnings were zero) to look at the difference between women who
were given opportunity to and actually earned money and those who were not given
the opportunity. The results are similar to those of the unrestricted sample.
The transitional matrices in appendix tables A1-A4 show how subjects changed
between intentional uses and actual uses of their earning from the game. In general
these matrices suggest the effect of commitment behaviour and unforeseen post-
experiment interactions. There is a large shift towards actually keeping the earnings or
purchasing for self from other intentional uses among women. The shift is not
pronounced among men. This could be explained by the local context where women
have more options of indulgence goods (for example, clothes & jewelleries) than
men, who were restrained by religion and social norms from bad indulgences like
drinking and gambling. Another notable shift is women who planned to use the
earnings in children or household common uses turned out giving the earnings to
spouses. On the contrary the percentage of men giving the earnings to spouse was not
much different from that of men planning to do so. These two main shifts led to a
higher percentage of women keeping the earnings or purchasing for self, a higher
percentage of those giving the earnings to spouse and a lower percentage of those
spending on children or common uses, compared with the percentages of women who
planned to do so.
We also analyse the uses of earnings for the secret treatment and the no-secret
treatment separately. Since the ‘borrowers’ in the secret treatment could hide both
their choices and available options from their spouses, the ‘spenders’ may not be
aware of the difference in the role of the ‘borrowers’ and the ‘spenders’. Therefore,
we may not find any effect of offering credit to women in the secret treatment even if
the effect exists. We however find the results for the secret treatment and the no-
secret treatment separately are similar to each other and to those of the combined
sample. In general, there is no difference between female spenders and female
borrowers and between those who transferred the decision-making of the lottery
choice and those who did not.
5.4. Robustness check on microfinance membership
Our subject pool is drawn from relatively low-income households, who are targeted
clients by microfinance institutions. However, to make sure our findings apply to
microfinance borrowers we run a robustness check on a smaller sample of households
who are currently microfinance members and households who reported to have plan
of getting microcredit loan in near future. We call this sample ‘potential microfinance
members’ hereafter 7 . Appendix table A5 show statistics of basic household
characteristics for potential microfinance members and those who are not separately.
The t-test p-values show no statistical difference between these two samples, except
for the number of household members and household annual income per capita. The
potential microfinance members are relatively poorer and have more household
members.
We reproduce table 3 using the smaller sample of potential microfinance members
and the results are reported in appendix table A6. The findings are similar to those of
the full sample. The Fisher randomization test p-values show significant gender
difference in the village-level transfer rate in the secret and no-secret treatments but 7 To address the concern that our estimates of the gender effect might capture the effect of being a microfinance member, we also run another robustness check on the sample of ‘potential microfinance members’ but excluding microfinance members whose memberships are longer than one year.
not in the random-couple treatment. There is also a significant difference in the
female’s transfer rate and female-male transfer rate between the secret and the no-
secret treatment.
6. Conclusion
The present paper analyses findings from a lab-in-the-experiment that was designed to
study the intra-household decision-making process between spouses in rural villages
in Bangladesh. We focus on the control over loan use and household expenditures in
the context of microfinance where a microcredit loan is offered to either spouse and
they have the opportunity to invest in a risky and profitable project. We are motivated
by anecdotes in the literature to explain the common phenomenon that female
borrowers concede control over loan use to their husbands. We find evidence that
gender difference exists in intra-household decision making, which prevents women
from taking control of their own loans. Women were more likely to let their spouses
make decision on which project to invest in and their decision was driven by both
voluntary and involuntary reasons. Women voluntarily transferred the control over
loan use due to their belief that their spouses were more capable of making financial
decision, thus they conceded the control even under asymmetric information
condition. However, when information were fully disclosed between spouses
women’s tendency to transfer the decision making increased, suggesting women felt
obliged to let their spouses make decision, due to their relatively lower decision-
making power. We however do not find evidence for other explanations that have
been suggested in the literature such as women’s strategic behaviour for gaining more
control over household expenditures or differences in gender nature like risk taking
and self-confidence. We also find women’s decision to transfer the control over loan
use is significantly correlated with the difference in social capital and earnings
between spouses and the presence of husband’s mother in the household.
These findings have affirmed that offering credit to women is not directly translated
into improving women’s access to economic resources and stimulating their
participation in household’s income-earning activities and decision-making process.
More importantly, the findings have shed light on the driving factors that hinder
women from taking control over their own opportunity. Most of the literature has
been focused on the imbalance in bargaining power between spouses that underpins
the household decision-making process and suggested the key role of asymmetric
information between spouses in determining household outcomes (for example,
Anderson & Eswaran, 2009; Ashraf, 2009; Ashraf et. al., 2014). Our findings have
provided important evidence for the effect of intra-household imbalance in financial
capacity in household’s financial decision making. We thus propose policies that
target household outcomes in general and microfinance policies in particular take into
account of both the gender difference in decision-making power and the difference in
capacity, with financial capacity as one example. However, this is deemed to be not a
simple task, especially in the context of rural Bangladesh where social norms and
religion strongly dictate the role of women in the family, making these two factors
even intertwined. Since women are obliged to obey their husbands, restricted on
mobility, and confined to household domain, they are restrained from acquiring
knowledge and skills necessary for financial management. On the other direction,
women’s lower level of knowledge and skills would reinforce their lower bargaining
power in the household. Field, Jayachandran, & Pande (2010) conducted a
randomized evaluation of a business training program in India and found positive
impact on business income for Upper Caste women, but not for the least restricted
group of Lower Caste women or the most restricted group of Muslim women. They
explained this finding by the non-monotonic effect of social norms on women’s
ability to acquire and apply business knowledge: Upper Caste women might have had
more to learn than Lower Caste women due to their prior more restricted mobility but
Muslim women had too little mobility to put the knowledge to use. There are however
positive evidences that providing necessary skills and information to women could
improve women’s participation in income-earning activities. Bandiera, Burgess,
Goldstein, Buehren, Gulesci, Rasul, & Sulaiman (2014) found simultaneously
providing vocational training and information on sex, reproduction and marriage
significantly improved both the economic and social position of adolescent girls in
Uganda. They were 72% more likely to engage in income generating activities and
their monthly consumptione expenditures increased by 41%. In the context of
Bangladesh, Bandiera, Burgess, Das, Gulesci, Rasul, & Sulaiman (2013) found
transfers of assets and skills had permanent and positive impact on the occupational
choice and earnings of the poorest women.
The paper also offers an important insight into intra-household decision making in
allocating household expenditures. Our analysis of intentional use and actual use of
the earnings from the loan investment shows women were more likely to let their
spouses take control over household expenditures, irrespective of which spouse was
offered the opportunity to invest and irrespective of whether or not they transferred
the control of loan use. This thus brings reservation to the argument that increasing
women’s earned income could improve their control over household expenditures, at
least in the short time frame and the local context as in our experiment setting.
Offering credit to women does not necessarily translate into improvement in women’s
control over household expenditures, and even more disappointedly women’s control
over loan use does not guarantee their control over earnings from using the loan.
Table 2: Demographic characteristics across treatments Mean t-test p-value
No secret Secret Random No secret-Secret
Secret-Random
No secret-Random
no. couples 319 316 191 muslim 0.80 (0.40) 0.78 (0.41) 0.75 (0.43) 0.65 0.47 0.26 no. household members
Safe TRIPLE 1/10 900 330 66.41 KEEP 8/10 300 LOSE 1/10 0 Risky TRIPLE 4/10 900 420 128.69 KEEP 2/10 300 LOSE 4/10 0 Note: Risk is calculated as the standard deviation from the expected payoff.
Table 4: Male and female transfer rate at village level Treatment Mean Std Fisher randomization test one-sided p-value
Non secret-Secret
Secret-Random
Male Non Secret 0.213 0.107 0.827 Secret 0.252 0.098 0.017 Random 0.085 0.077 Female Non Secret 0.653 0.190 0.028 Secret 0.430 0.283 0.0002 Random 0.073 0.047 Female-male Non Secret 0.440 0.241 0.001 0.019 Secret 0.179 0.277 0.038 0.0003 Random -0.012 0.047 0.875
Table 3: Social capital and mobility %/Mean(Std)
Men Women Do you read the newspaper at least once a week? 26.27 10.90 Do you listen to the radio or watch TV at least once a week? 50.97 34.02 Do you participate in any club/ committee/ meeting group at least once a month?
37.89 27.00
Do you go on your own at least once a month to: public places in your village 79.78 48.55
places outside your village 71.31 36.44 Do you need to seek permission of your spouse to go to: public places in your village 14.16 89.71 places outside your village 14.89 87.17 Did you do any income-earning work outside home in the last one year?
65.01 12.59
Did you do any income-earning work on your own at home in the last one year?
63.32 81.84
How much is your annual earnings in the last one year? (taka) 69,831 (33,446) 8,340 (27,218)
Table 5: Male and female transfer rate at individual level Treatment Mean SE/Std Fisher exact test one-sided p-value Non secret-
Secret Secret-Random
Male Non Secret 0.213 0.410 0.244 Secret 0.252 0.435 0.001 Random 0.0842 0.261 Female Non Secret 0.654 0.477 0.000 Secret 0.427 0.496 0.000 Random 0.073 0.261 Female-male Non Secret 0.442 0.050 0.000 Secret 0.175 0.053 0.001 Random 0.011 0.039 0.491
Table 7: Probit Regressions: determinants of decision-making, using different measures of capacity gap
Table 6: Probit Regressions: determinants of decision making
(1) (2) (3) (4) (5) (6) (7)
female 0.22*** 0.14*** 0.15*** 0.14*** 0.15*** 0.15** 0.15** (0.027) (0.043) (0.043) (0.043) (0.043) (0.064) (0.070) no-secret 0.078** -0.038 -0.043 -0.038 -0.045 -0.045 -0.045 (0.032) (0.046) (0.046) (0.047) (0.047) (0.047) (0.045) random -0.32*** -0.21*** -0.21*** -0.22*** -0.22*** -0.22*** -0.22*** (0.044) (0.063) (0.062) (0.063) (0.063) (0.060) (0.058) female*no-secret 0.21*** 0.21*** 0.21*** 0.21*** 0.21** 0.21** (0.061) (0.062) (0.061) (0.061) (0.082) (0.081) female*random -0.17* -0.18** -0.16* -0.17* -0.17** -0.17* (0.090) (0.089) (0.090) (0.089) (0.078) (0.093) +demographics controls x x x x + upazila (sub-district) FE x x x x + corrected standard error for village cluster
x
+ corrected standard error for session cluster
x
N 826 826 826 826 826 826 826 This table shows results from probit regressions where the dependent variable is the probability of transfering the decision making of lottery choice in the game. Demographic controls are: wife’s age, husband’s age, wife’s schooling, husband’s schooling, religion, number of household members, annual income per capita. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
N 316 316 316 316 316 This table shows results from probit regressions where the dependent variable is the probability of transfering the decision making of lottery choice in the game. All specifications control for upazila fixed effect and correct standard error for village cluster but do not include demographic controls. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 8: Women’s intentional and actual uses of game payoff in the secret treatment
Intentional uses Actual uses Female
borrower, transfering
Female spender
Fisher exact test one-sided p-value
Female borrower, transfering
Female spender
Fisher exact test one-sided p-value
keep/purchase for self
25.00% 14.29% 0.070 11.76% 26.40% 0.024
give to spouse 13.46% 8.73% 0.244 29.41% 19.20% 0.102 purchase for spouse
1.92% 3.97% 0.434 1.96% 0.80% 0.497
children/common use
59.62% 73.02% 0.058 56.86% 53.60% 0.411
Kolmogorov-Smirnov test
0.303 0.374
Table 9: Probit Regressions: determinants of decision-making
A. No-secret treatment (1) (2) (3) (4) female 0.39*** 0.37*** 0.14 0.36*** (0.050) (0.053) (0.096) (0.037) social capital gap -0.068* (0.041) earnings gap -0.0020*** (0.00074) female*husband’s mother 0.14* (0.084) N 319 319 319 319 B. Secret treatment (1) (2) (3) (4) female 0.17** 0.18** 0.12* 0.14* (0.082) (0.088) (0.072) (0.074) social capital gap 0.017 (0.048) earnings gap -0.00043 (0.00067) female*husband’s mother 0.13*** (0.044) N 316 316 316 316 This table shows results from probit regressions where the dependent variable is the probability of transfering the decision making of lottery choice in the game. All specifications control for upazila fixed effect and correct standard error for village cluster but do not include demographic controls. Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 10: Probability of choosing riskier lottery (%) Men Women Fisher exact
test one-sided p-value
All 31.36 31.11 0.479 No-secret 32.92 37.62 0.123 Secret 36.39 31.01 0.089 Random 20.42 20.42 0.550 Secret vs. no-secret 0.201 0.047 Secret vs. random 0.000 0.006 No-secret vs. random 0.001 0.000
Table 11: Male and female transfer rate by risk-taking behaviour Treatment Risk adverse Risk lover Fisher exact test
one-sided p-value Male Non Secret 0.257 (0.044) 0.136 (0.045) 0.051* Secret 0.300 (0.044) 0.143 (0.051) 0.025** Random 0.091 (0.033) 0.056 (0.056) 0.529 Female Non Secret 0.686 (0.046) 0.593 (0.067) 0.160 Secret 0.404 (0.047) 0.479 (0.073) 0.240 Random 0.053 (0.026) 0.143 (0.078) 0.175
Table 12: Intentional and actual uses of game payoff by gender and partner types
Percentage Fisher exact test one-sided p-value Male
G1 There is a lottery where the chance of winning a 1,000 taka prize is 1 in 100. If 1000 people buy the lottery tickets and each buys only 1 ticket, how many people will win the 1,000 taka prize?
G2 Suppose you are offered either 1 of 2 lottery bags: The Paan bag has 2 balls with 500 taka value each and 2 balls with 100 taka value each. The Supari bag has 2 balls with 400 taka value each and 2 balls with 100 taka value each. You will pick randomly 1 ball from the bag you choose and that will be your prize. Which bag would you choose?
Paan (1) Supari (2)
Either Paan or Supari (they are the same to me) (3)
Don’t know (4)
G3 Suppose you are offered either 1 of 2 lottery bags: The Paan bag has 2 balls with 500 taka value each and 2 balls with 100 taka value each. The Supari bag has 3 balls with 500 taka value each and 2 balls with 100 taka value each. You will pick randomly 1 ball from the bag you choose and that will be your prize. Which bag would you choose?
Paan (1) Supari (2)
Either Paan or Supari (they are the same to me) (3)
Don’t know (4)
G4 Suppose you are offered either 1 of 2 lottery bags: The Paan bag has 2 balls with 500 taka value each and 2 balls with 100 taka value each. The Supari bag has 5 balls with 500 taka value each and 5 balls with 100 taka value each. You will pick randomly 1 ball from the bag you choose and that will be your prize. Which bag would you choose?
Paan (1) Supari (2)
Either Paan or Supari (they are the same to me) (3)
Don’t know (4)
H. Personal opinions: H6 In general are men or women better in working
with numbers and making financial decisions? Men (1)
Women (2) They are the same (3)
Don’t know (4)
H7 Are you or your wife/husband better in working with numbers and making financial decisions?
You (1)
Your wife/husband (2)
We are the same (3)
Don’t know (4)
Table A1: Transitional matrix from intentional to actual uses for female borrowers
Intentional uses Actual uses
keep/purchase for self
give to spouse purchase for spouse
children/ common use
Total
keep/purchase for self 11.63% 20.93% 4.65% 62.79% 100.00% give to spouse 20.00% 30.00% 10.00% 40.00% 100.00% purchase for spouse 50.00% 12.50% 25.00% 12.50% 100.00% children/common use 23.53% 27.06% 1.18% 48.24% 100.00% Total 21.91% 25.90% 3.59% 48.61% 100.00%
Table A2: Transitional matrix from intentional to actual uses for male borrowers
Intentional uses Actual uses keep/purchase
for self give to spouse purchase for
spouse children/ common use
Total
keep/purchase for self 25.33% 4.00% 4.00% 66.67% 100.00% give to spouse 38.46% 0.00% 0.00% 61.54% 100.00% purchase for spouse 75.00% 0.00% 0.00% 25.00% 100.00% children/common use 27.04% 8.18% 4.40% 60.38% 100.00% Total 27.89% 6.37% 3.98% 61.75% 100.00%
Table A3: Transitional matrix from intentional to actual uses for female spenders
Intentional uses Actual uses keep/purchase
for self give to spouse purchase for
spouse children/ common use
Total
keep/purchase for self 41.46% 21.95% 0.00% 36.59% 100.00% give to spouse 21.43% 53.57% 0.00% 25.00% 100.00% purchase for spouse 30.00% 30.00% 0.00% 40.00% 100.00% children/common use 20.23% 21.97% 2.31% 55.49% 100.00% Total 24.21% 25.79% 1.59% 48.41% 100.00%
Table A4: Transitional matrix from intentional to actual uses for male spenders
Intentional uses Actual uses keep/purchase give to spouse purchase for children/ Total
for self spouse common use keep/purchase for self 41.67% 10.42% 4.17% 43.75% 100.00% give to spouse 20.00% 20.00% 20.00% 40.00% 100.00% purchase for spouse 15.00% 10.00% 10.00% 65.00% 100.00% children/common use 24.14% 5.75% 3.45% 66.67% 100.00% Total 26.72% 7.29% 4.45% 61.54% 100.00%
Table A5: Demographic characteristics: microfinance members vs. non members Mean (std) t-test p-value
Non microfinance members
Microfinance members
no. couples 246 580 muslim 0.77 0.42 0.78 0.41 0.70 no. household members 4.48 1.25 4.66 1.35 0.07* annual income per capital (taka) 16,815 11,546 15,452 7,982 0.06* no. years of marriage 16.25 8.09 17.05 7.37 0.17 wife's age 34.33 7.73 34.44 7.01 0.85 husband's age 40.70 8.19 40.57 7.27 0.81 wife's schooling 4.27 3.78 4.50 3.85 0.43 husband's schooling 4.40 4.30 4.82 4.36 0.20 age gap 6.37 2.75 6.13 2.22 0.19 schooling gap 0.13 3.13 0.32 3.03 0.41
Table A6: Male and female transfer rate at village level – Microfinance members Treatment Mean Std Fisher randomization test one-sided p-value
Non secret-Secret
Secret-Random
Male Non Secret 0.233 (0.126) 0.636 Secret 0.250 (0.092) 0.009 Random 0.113 (0.113) Female Non Secret 0.659 (0.200) 0.061 Secret 0.464 (0.322) 0.002 Random 0.063 (0.063) Female-male Non Secret 0.426 (0.275) 0.002 0.055 Secret 0.214 (0.287) 0.023 0.021 Random -0.049 (-0.049) 0.875
References
Allen, T., Armendáriz, B., Karlan, D., & Mullainathan, S. (2010). Inviting Husbands in
Women Only Solidarity Groups: Evidence from Southern Mexico. Working Paper.
Abbink, K., Irlenbusch, B., & Renner, E. (2006). Group size and social ties in microfinance
institutions. Economic Inquiry, 44(4), 614-628.
Agarwal, B. (1994). A field of one's own: Gender and land rights in South Asia (Vol. 58).
Cambridge University Press.
Amin, S., 1997. The poverty–purdah trap in rural Bangladesh: implications for women's roles
in the family. Development and Change, 28, 213–233.
Anderson, S., & Baland, J.M. (2002). The economics of roscas and intrahousehold resource
allocation. The Quarterly Journal of Economics, 117(3), 963-995.
Anderson, S., & Eswaran, M. (2009). What determines female autonomy? Evidence from
Bangladesh. Journal of Development Economics, 90(2), 179-191.
Armendáriz, B., & Morduch, J. (2010). The economics of microfinance. MIT press.
Ashraf, N. (2009). Spousal control and intra-household decision making: An experimental
study in the Philippines. The American Economic Review, 1245-1277.
Ashraf, N., Field, E., & Lee, J. (2014). Household Bargaining and Excess Fertility: An
Experimental Study in Zambia. The American Economic Review, 104(7), 2210-37.
Bandiera, O., Burgess, R., Das, N., Gulesci, S., Rasul, I., & Sulaiman, M. (2013). Can basic
entrepreneurship transform the economic lives of the poor? Working paper.
Bandiera, O., Burgess, R., Goldstein, M., Buehren, N., Gulesci, S., Rasul, I., & Sulaiman, M.
(2014). Women’s empowerment in action: evidence from a randomized control trial
in Africa. Working paper.
Cain, M. T. (1978). The household life cycle and economic mobility in rural Bangladesh.
Population and Development Review, 421-438.
Duflo, E. (2012). Women Empowerment and Economic Development. Journal of Economic
Literature, 50(4), 1051-79.
Eckel, C. C., & Grossman, P. J. (2002). Sex differences and statistical stereotyping in
attitudes toward financial risk. Evolution and human behavior, 23(4), 281-295.
Field, E., Jayachandran, S., & Pande, R. (2010). Do traditional institutions constrain female
entrepreneurship? A field experiment on business training in India. The American
Economic Review, 125-129.
Folbre, N. (1984). Household production in the Philippines: a non-neoclassical approach.
Economic Development and Cultural Change, 32 (2), 303–330.
Giné, X., Jakiela, P., Karlan, D., & Morduch, J. (2010). Microfinance games. American
Economic Journal: Applied Economics, 2(3), 60-95.
Goetz, A.M., & Gupta, R.S. (1996). Who takes the credit? Gender, power, and control over
loan use in rural credit programs in Bangladesh. World development, 24(1), 45-63.
Hashemi, S. M., Schuler, S. R., & Riley, A. P. (1996). Rural credit programs and women's
empowerment in Bangladesh. World development, 24(4), 635-653.
Kabeer, N. (1999). Resources, agency, achievement: reflections on the measurement of
women's empowerment. Development and Change, 30 (3), 435–464.
Kabeer, N. (2001). Conflicts over credit: re-evaluating the empowerment potential of loans to
women in rural Bangladesh. World development, 29(1), 63-84.
Kabeer, N. (2005). Gender equality and women's empowerment: A critical analysis of the
third millennium development goal 1. Gender & Development, 13(1), 13-24.
Ligon, E. (2011). Dynamic bargaining in households (with an application to Bangladesh).
Available at SSRN 1776810.
Mani, A. (2011). Mine, Yours or Ours? The Efficiency of Household Investment Decisions:
An Experimental Approach. CAGE Online Working Paper Series.
Montgomery, R., Bhattacharya, D., Hulme, D., & Mosley, P. (1996). Credit for the poor in
Bangladesh. The BRAC Rural Development Programme and the Government Thana
Resource Development and Employment Programme. Finance against poverty:
Volume 2., 94-176.
Niederle, M., & Vesterlund, L. (2011). Gender and competition. Annu. Rev. Econ., 3(1), 601-
630.
Ngo, T.M.P., & Wahhaj, Z. (2012). Microfinance and gender empowerment. Journal of
Development Economics, 99(1), 1-12.
Pitt, M. M., & Khandker, S. R. (1996). Household and intrahousehold impact of the Grameen
Bank and similar targeted credit programs in Bangladesh. Washington, DC: World
Bank.
Pitt, M.M., Khandker, S.R., & Cartwright, J. (2006). Empowering women with micro
finance: evidence from Bangladesh. Economic Development and Cultural Change,
54(4), 791-831.
Rahman, L., & Rao, V. (2004). The determinants of gender equity in India: examining Dyson
and Moore's thesis with new data. Population and Development Review, 239-268.
Schaner, S.G. (2011). The Cost of Convenience? Transaction Costs, Bargaining Power, and
Savings Account Use in Kenya: Mimeo.
Tassel, E.V. (2004). Household bargaining and microfinance. Journal of Development