An Experimental Investigation of Intra-Household Resource Allocation in Rural India

Abstract

This study aims to investigate intra-household bargaining outcomes elicited in an artefactual field experiment design where participants completed a purchase task of real commodities. Married couples separately expressed their initial preferences over commodities. The bargaining process in the experiment was exogenously introduced by sharing information about partners’ preferences in the treatment group. We hypothesized that the spouse with weaker bargaining position at the household level would consider the information of their partner’s preferences while making own consumption decisions more compared to their partner. Therefore, they may deviate from their own preferences when purchasing commodities. More than 230 married couples from two villages in the Tamil Nadu state of India participated in the experiment. It was observed that information about partners’ spending preferences resulted in reduced final allocations for female
participants. However, the deviation was not significantly different from the original intention to spend. Therefore, information about partners’ preferences may not be an effective medium to elicit bargaining power in the context of jointly-consumed household commodities. Subgroup analyses were performed to identify any heterogeneous treatment effects.

Full text

i
own
Universite Laval
!
An experimental investigation of intra-household
resource allocation in rural India
working paper
2016-20
October 2016
Savita Kulkarni
Anirudh Tagat
Hansika Kapoor

!
An experimental investigation of intra-household
resource allocation in rural India!
Abstract
This study aims to investigate intra-household bargaining outcomes elicited in an artefactual field
experiment design where participants completed a purchase task of real commodities. Married
couples separately expressed their initial preferences over commodities. The bargaining process
in the experiment was exogenously introduced by sharing information about partners’
preferences in the treatment group. We hypothesized that the spouse with weaker bargaining
position at the household level would consider the information of their partner’s preferences while
making own consumption decisions more compared to their partner. Therefore, they may deviate
from their own preferences when purchasing commodities. More than 230 married couples from
two villages in the Tamil Nadu state of India participated in the experiment. It was observed that
information about partners’ spending preferences resulted in reduced final allocations for female
participants. However, the deviation was not significantly different from the original intention to
spend. Therefore, information about partners’ preferences may not be an effective medium to
elicit bargaining power in the context of jointly-consumed household commodities. Subgroup
analyses were performed to identify any heterogeneous treatment effects.
JEL: C93; D13; J12; J16; O15
Keywords: intra-household bargaining, gender, artefactual field experiment, women’s
empowerment, welfare schemes
!
Authors
Savita Kulkarni
Assistant Professor,
Symbiosis School of Economics
Pune, India
kulkarnisavita@gmail.com
!
Anirudh Tagat
Research Author, Department of Economics,
Monk Prayogshala
Mumbai, India
at@monkprayogshala.in!
Hansika Kapoor
Research Author, Department of Psychology,
Monk Prayogshala
Mumbai, India
hk@monkprayogshala.in
!
!
Acknowledgements
!
This research work was carried out with financial and scientific support from the Partnership for
Economic Policy (PEP) (www.pep-net.org) with funding from the Department for International
Development (DFID) of the United Kingdom (or UK Aid), and the Government of Canada through
the International Development Research Center (IDRC). The authors are grateful to participants
at the 12th and 13thAnnual PEP Conference, in particular Maria Laura Alzua and Maria Adelaida
Lopera. We are grateful to Hari K. Nagarajan and G Palanithurai for advice and institutional
support. We also thank volunteers from Gandhigram Rural Institute, Dindigul for assistance with
field work. All errors are solely attributable to the authors.

!
List of tables
Table 1: Randomization and balance table …………………………………………………………….12
Table 2: Individual and household characteristics ……………………………………………………..13
Table 3: Gender-wise differences in preferences for commodities …………………………..…….13
Table 4: Mean of experimental outcomes by treatment ……………………………………………..15
Table 5: Result of Logit Regression for correct guesses …………………………………………...…...17
Table 6: Impact of information on experimental outcomes …………………………………….…...17
Table 7: Information and final allocation of spouse ……………………………………………..…….19
Table 8a: Heterogeneous treatment effects on final allocations, difference between final and
intent-to-spend, and absolute deviations ……………………………………………………………….20
Table 8b: Heterogeneous treatment effects of MGNREGS participation on difference between
final order and intent-to-spend ………………………………………………………………………..…..22
Table A1: Role of information in choice convergence …………………………………………….…33
List of figures
Figure 1: Treatment-wise deviations ............................................................................................. …31
Figure 2: Difference between expectation and final order ………………………………………….32
Figure 3: Power analysis ……………………………………………………………………………………..32
Figure 4: Quantile plot of intent and final allocation ……………………………………………..…...32
Table of contents
Executive summary p.1
I. Introduction p.3
II. Conceptual framework p.5
2.1. Theoretical framework
2.2. Experimental design
III. Data analysis and results p.15
IV. Heterogeneous treatment effects p.20
V. Conclusions and policy implications p.23
References p.24
Appendices p.27

!
1!
!
Executive summary
Decisions made within the household affect household outcomes. For instance, deciding to
allocate more resources toward food rather than recreation yields positive benefits toward
overall health. The manner in which husbands and wives arrive at such allocation decisions is
therefore of interest to researchers. Given the differences in preferences between men and
women with respect to such consumption decisions, the manner in which they allocate resources
within the household was examined in this study. First, whether such preferences played a role
toward allocation was studied; and second, whether information about each other’s preferences
played a role toward allocation was investigated. By contextualizing this research problem within
intra-household dynamics, particularly of bargaining, the researchers assessed whether women
held a weaker bargaining position in the household with respect to resource allocation decisions,
as compared to men. The sharing of information was motivated by the assumption that such
disparities in information about each other’s intentions toward spending behaviour may be
associated with bargaining positions. Therefore, this study sought to examine the role of
information sharing in intra-household resource allocation through a field experiment in rural
India.
Methodology
The experiment took place in two villages in rural India: Thethoor (Madurai District) and
Mallanampatti (Dindigul District) in the Indian state of Tamil Nadu. Husbands and wives from 231
households participated in the study, bifurcated into two conditions: one where they received
information about their spouse’s intentions regarding purchase decisions (n = 102 couples) and
the second where they did not receive such information (n = 129 couples). Couples participated
in a real-world purchase task, wherein they were provided with Rs. 100 worth of token money
(divided equally between the spouses), and made decisions to spend it on the following
commodities: rice, salt, paracetamol, pain relief balm, notebooks, pens, soap, and toothpaste.
These commodities were selected on the basis of household consumption between men and
women via focus group discussions beforehand. A pilot study was conducted in August, 2015 to
test a few commodities, which were then finalized for the main study. The eight commodities
represented food, health, education, and sanitation. Thus, this task allowed couples to
demonstrate how they would allocate resources toward such commodities (and thereby
household outcomes), under conditions of information sharing (knowing what their spouse would
decide) and no information sharing (not knowing what their spouse would decide). This is an
application of the concessions and claims model that is outlined in Muthoo (1992, 1996) and
considers a revocation cost of an initial decision when information sharing is added to the model.

!
2!
!
Key findings
Results showed that (a) spouses were aware of their partners’ preferences and were able to
correctly predict their partners’ choices; (b) Providing information about partner preferences only
influenced women’s final allocations, and information was not very effective in examining how
intra-household resource allocation operated; and (c) Intra-household resource allocations were
associated with individual and household characteristics, such as the difference in years of
education between spouses, as well as which member participated in a welfare program.
There were differences between men and women in their preferences for commodities: on
average, women were almost twice as likely to prefer spending on rice as compared to men. In
contrast, men were 1.2 times more likely to spend on pens than women. More than 85% of the
individuals did not change their initial preferences, potentially due to status-quo bias or effects
of existing intra-household dynamics. Indeed, spouses who had information about their partner’s
preferences were 36% more likely to correctly guess on what their partner would spend. Thus,
giving spouses information about each other’s preferences did not lead to any statistically
significant differences between their intentions and final decisions. Women reduced their final
allocations by about 1.5 times on preferred commodities when they were given information and
knew their husband’s preferences. These effects appear to vary by heterogeneity of household
characteristics such as education, age, and participation in a welfare program. Less educated
women, compared to their husbands, lowered their final allocations on preferred commodities
by 0.3 times. With respect to participation in a government welfare program, MGNREGS, women
in households where only the woman participated reduced their final allocation, whereas men in
households where both spouses participated increased their final allocation.
In this setting, we found correlational interactions between intra-household bargaining dynamics
and a large-scale employment guarantee programme (MGNREGS). A well-designed welfare
scheme with the goal of empowering women may provide financial resources to women but may
not obtain desirable results if their bargaining position is low. Our results indicate that intra-
household bargaining positions are associated with participation in the employment guarantee
scheme. Causal relations may not be drawn based on such a quasi-field experiment, but our
study sets the agenda for future research in terms of developing methodology to elicit the intra-
household bargaining process and investing the impact of welfare schemes on the bargaining
positions of the beneficiaries. Given that gendered preferences are directly associated with
household outcomes, it is recommended that policy stakeholders take cognizance of existing
intra-household dynamics when targeting benefits to households.

!
3!
!
I. Introduction
Household decision-making and resource allocation are critical for economic and human
development. Traditional economics viewed the household as a collection of individuals who
behave in consensus to allocate time and resources for individual and collective wellbeing.
However, within households, many factors like age, gender, marital status, income level, and
education influence the dynamics of intra-household decision making. In the context of
gender and intra-household resource allocation, recent empirical studies have indicated that
gender differences exist in household preferences, which may have important welfare
implications. For instance, studies have shown that resources, when entrusted to women in
the household, are better used for overall family welfare improvements (Quisumbing &
Maluccio, 2000; Udry, Hoddinott, Alderman & Haddad, 1995; Quisumbing, 1996). Similarly,
women endowed with income are more likely to invest in education, children's nutrition, and
housing than men (Thomas 1990; 1994; Hoddinott & Haddad 1995; Duflo 2003). Various
countries including the UK and Mexico have designed policies to direct aid, such as food
coupons, towards women instead of men. This implies that endowing women in the
household with greater decision-making power will have positive spillover effects on the
wellbeing of the household and society at large.
The existing literature therefore implies that men and women may have qualitatively
different preferences for various goods and services consumed jointly at household level.
Preference-consistent consumption, however, depends significantly on an individual’s
bargaining power in the household. Implicitly, it also assumes that the individual is able to
monitor preferences of other members or has information about others’ preferences and is
aware of the relative bargaining position of each member of the household.
Bargaining interactions between spouses may not be perfectly observable, but
household decisions can be considered an outcome of the intra-household bargaining
process. Weak bargaining positions may be reflected in preference-inconsistent choices at
an individual level if the wife is aware of the conflicting preferences between her and her
partner. In the present study, we investigate the intra-household bargaining process by
sharing information about spouses’ preferences in an experimental setting.
While a survey-based approach may not reveal the precise structure of spousal
relationships, household-level secondary data on consumption may reflect a post-bargaining
consensus. Therefore, both primary surveys and secondary data may not precisely reveal the
dynamics of intra-household bargaining. Thus, a game theoretic bargaining model may be
more appropriate to elicit intra-household bargaining interactions. We adapt the two-person
sequential bargaining game from Muthoo (1992, 1996) considering the case of intra-
household resource allocation between two major household decision-makers: the husband
and wife. An experimental methodology was adopted to test such a game theoretic

!
4!
!
bargaining model in a more controlled environment in the form of an artefactual field
experiment (Levitt & List 2007). Our experiment was administered in a more controlled
environment compared to naturally-occurring situations. We manipulated information about
spousal preferences, while keeping other things constant. Therefore, it allowed us to establish
causal links between information and gender-specific preferences in a contextual
environment.
In our novel experimental setting, individuals’ preferences for jointly-consumed real
commodities were elicited. These were shared with their partners in the treatment group,
while the control group did not receive any information. Sequentially, individuals made their
consumption decisions separately. Sharing spouses’ preferences in the absence of any direct
communication or enforcement mechanism was expected to bring more changes in the
choices of the individual with relatively weak bargaining power compared to the group having
no such information.
The experiment was carried out with 231 randomly-selected married couples from two
villages in the state of Tamil Nadu in India. For women, we observed that the information
about their partners’ initial allocation preferences brought about a reduction in the females’
final consumption decision, compared to those who had no such information. However, we
find that such a change was not significantly different from their original intent-to-spend,
pointing toward the potential ineffectiveness of information in simulating bargaining
dynamics. We argue that this may be because partner preferences were already accounted
for by individuals in their own intent-to-buy. Therefore, sharing information in this context did
not bring any significant changes in final consumption choices. However, this evidence does
not imply the absence of bargaining and imbalanced power structure within households.
Our results show that (a) spouses appear to have good knowledge of their partners’
preferences and can correctly predict their partners’ consumption choices. This suggests that
existing intra-household dynamics may be driving behaviour in the experiment. (b) Providing
information about their partners’ preferences only influences women’s final allocations, and
may not be a thoroughly effective medium to initiate the bargaining processes between
household members in the case of jointly-consumed commodities. (c) Intra-household
resource allocations are associated with individual and household characteristics, such as the
difference in years of education between spouses, as well as which member participates in a
welfare program.
These results reflect bargaining dynamics in the context of consumption decisions in the state
of Tamil Nadu.1 Given the specificity of our sample, future policy research in this area should
focus on assessing similar behaviour in a non-subsidy-based state such as Bihar or Gujarat in
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 The State government of Tamil Nadu offers extensive subsidies to all households (but particularly poorer households, who may
also participate in MGNREGS) for rice, oil, school uniforms (for children studying in government schools), notebooks, mixer-
grinders, ceiling fans, among many other agricultural subsidies (Leena, 2014).

!
5!
!
India. This will help in eliminating the potential role of state-sponsored consumption
subsidies on consumption behaviour.
Another direction that policy can take in light of our study’s results is to offer incentives
to facilitate coordination between spouses for the consumption demand of specific
commodities or services. This is similar to conditional cash transfers where a cash benefit is
provided to households under the stipulation that it will be allocated to a particular use (e.g.
payment of school fees or vaccination fees). This may be implemented by offering transfers
with tied and untied components. A specific portion of the transfer could be in the form of
redeemable coupons for nutritious food (e.g. fruits and vegetables) or doctors’ fees. Such a
measure will ensure that for all sets of household preferences, certain outcomes (perceived
to be beneficial for the entire household) will be facilitated.
The remainder of the paper is organized as follows: the first section describes the
conceptual framework that guides the experimental design. The second section outlines the
experimental setting with subsections on treatments and experimental outcomes. The third
section discusses data analysis and results while the subsequent section concludes the paper.
II. Conceptual framework
!
The proposed research design and methodology aims to build on current empirical
literature on intra-household bargaining in India. In order to assess the impact of changes in
intra-household bargaining, our field experiment draws on empirical methods that have only
recently gained ground, while also laying emphasis on seminal economic theories of
households and families such as Basu (2006) and Browning and Chiappori (1998). We first
tackle the literature of direct importance: studies that employ field experiments to address
questions of intra-household bargaining. Mani (2011) uses an investment game between
spouses of the same household (in rural Andhra Pradesh) to investigate the relative
importance of key factors (return on investments and informational awareness) that influence
the efficacy of household investment decisions. In finding that household members are willing
to trade off lower efficiency for more control over decisions, she makes an important case for
factors such as identity that may be spurring such ‘spiteful’ intra-household dynamics. Ashraf
(2009) conducts a similar experiment to test the impact of treatment variables like information
and communication on making decisions to save money in the Philippines. It was observed
that when decision-making was private, men put money in their personal accounts but spent
this amount for their own benefits when choices were observable. When it was required to
communicate the choices, men put more money in their wives’ accounts. The author has
drawn an important inference that men (women) whose wives (husbands) control household
savings respond more strongly to the treatment. Thus, this confirms that gender-specific
differences exist in household decision-making which may be influenced by information. For

!
6!
!
a more recent study of the influence of information asymmetry on intra-household allocations,
see Castilla and Walker (2013).
Dasgupta and Mani (2015) show that private consumption commodities are more
likely to be preferred by men when they exert effort in order to obtain earnings. Our
experiment will differ in that there will be no hypothetical effort-demanding task; instead
exogenous variations enter via information about spouses’ intention to buy. Furthermore, we
expect that the proposed artefactual field experiment will be among the first to
experimentally investigate consumption choices in the rural Indian population using real
commodities (see Beblo, Beninger, Cochard, Couprie & Hopfensitz, 2015). Other studies that
examine intra-household dynamics using experimental procedures include Cochard, Couprie
and Hopfensitz (2014), Munro, Kebede, Tarazona-Gomez and Verschoor (2014), and Yang
and Carlsson (2012).
2.1 Theoretical framework
!
The experiment was designed to investigate the bargaining outcome between spouses in
two situations – one in which information about partners’ preferences is shared and another
in which they are not.
Household model and predictions
The experimental design was based upon the two-person sequential bargaining game
designed by Muthoo (1992, 1996) and Binmore (1998). Under conditions of risk-neutrality,
the two players (Husband (H) and Wife (W)) bargain over the allocation of fixed endowments
! "#$ %& to consumption of the '() commodity. At time*+ , %, both players decide their
intention-to-spend, -., / , 01$ 23, for commodities on display. This initial preference may
already take into account existing intra-household dynamics as well as any reference points
over consumption given that participants are aware that their spouses are playing the same
game. At time*+ , 4, the husband-wife pair is randomly assigned to receive a treatment of
either no information or information (I = 0, 1) by the experimenter. Under the information
condition, each individual receives information about the intent-to-spend of their spouse,
providing a signal of the initial claim that the partner would like to make over allocation of
resources for consumption. Individuals assigned to the no information condition receive no
such signals about their partners’ intent-to-spend.2
The resulting bargaining game plays out as follows: W makes an offer to H (and H to
W) that signals her/his preferred allocation distribution for the '() commodity. The players
then simultaneously decide their final allocation, 5.* ! "#$ %& for consumption with this
information. The final payoffs for the /() *individual as 6.* - *form a strategy pair - , 7-8$ -9:.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2 Ashraf (2009)’s ‘Public’ treatment condition provides a parallel for this experiment’s ‘no information’ condition, however they
learn each other’s decisions only after they leave the lab. Ashraf’s ‘Negotiation’ treatment condition is similar to our ‘Information’
condition; however, participants are not informed of each other’s decisions but only of their intentions.

!
7!
!
In the case that*-8; * -9* < %, there will be no deviations from the initial claim and that 5.* =
*-. and the utility derived for each player will be >.5. from obtaining 5. share of the allocation
of endowments to the '() commodity. In the case of incompatible claims (-8; * -9* ? %), at
least one player must revoke his/her initial claim. If player / receives a share*5.* @ * -., then the
cost of revoking the initial claim (i.e. making a concession) is given by*A.*75.$ -.:. Thus, if*5.* =
*-., then *A., #, but if 5.* @ * -., then*A.? #. The cost-of-revoking function (A.) here represents
the bargaining power that player / exercises over the allocation decisions for both members
of the household. To summarize, the utility of the household is therefore given by:
>.5.$ -., * >858********************* ; ******************* >959*$*********************************% = *5.* = * -.
">858B *A8758$ -8:& * ; * ">959B *A9759$ -9:&$*********-.* = 5 = # *
Broadly, his model suggests that in any bargaining game between two players, there are
significant costs to deviating (‘revoking’) their original decisions, which vary with the
information each player has regarding the strategy of the other player. In the context of intra-
household dynamics, we assume that the deviation from stated preferences will impose some
‘revocation cost’, which determines each spouse’s household allocation. The bargaining
power can therefore be inferred from these deviations (Muthoo, 1996, p. 145). For instance,
the extent to which a player deviates from his or her initial preferences, will represent the
magnitude of bargaining power; a higher deviation (from initial preferences) implies a lower
cost of revoking one’s decision, and hence lower bargaining power, while a lower deviation
implies higher cost of revoking one’s decision, and hence higher bargaining power. Both
players may choose to concede or retain their preferences, making the outcome of
bargaining an impasse, respectively.
We attempted to find deviation in individuals’ final consumption choice (allocation in
terms of purchase quantity or price) from their initial preferences in the replica of a real-world
household-level bargaining situation in the experimental setting. In our experimental setup,
the deviation from stated preferences was elicited through a purchase task of real
commodities. Spouses first expressed their preference for a set of commodities they intended
to purchase with the available endowment. These intentions were likely to represent their
own preferences as unawareness about their spouses’ preferences would not drive them to
predict and reconcile their partners’ preferences3. The bargaining process in the experiment
was exogenously introduced by providing the information of the partners’ intention to buy.
Couples were randomly assigned to a ‘full-information’ treatment or a ‘no-information’
treatment. In the former group, individuals’ intention-to-buy list was shared with their spouses
and then spouses were allowed to alter or retain their own intention-to-buy commodity list.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
3 Note, however, that initial intentions may already take into account past consumption behaviour or unobservable existing intra-
household dynamics. We argue that such existing intra-household dynamics should also be manifested in the real purchase task
of the experiment.

!
8!
!
The no-information group did not receive such information. Nonetheless, this no-information
group also received the chance to alter their intention-to-buy list while placing the final order.
This provision maintained the symmetry in the decision-making process in both groups and
also captured time-wise changes in preferences. Deviations from the intention-to-buy in the
no-information group serve as a benchmark for estimating the bargaining power in the full-
information condition. Conditioned on information availability, it is possible that being aware
of the spouse’s intention-to-spend alters the final allocation for any particular commodity. We
further attempted to investigate the bargaining positions of spouses in the household by
inquiring about expectations of their partners’ choices in the final-order.
Both spouses arrived to the experiment together and took decisions in the same
room, though separately. The endowment of Rs.100 was framed as a collective endowment
with equal division between the spouses.4 This arrangement was expected to emphasize a
sense of household-level collective utility. As a result, the intra-household power relationship
was expected to be reflected in the decision-making.
Absolute differences between intentions and final-order forms are considered to be
an unambiguous indicator of the bargaining power as it is arguably more context-neutral
compared to any other form of personal interaction (face-to-face or telephone conversations).
It is important to note that such differences may also be on account of coordination efforts
by both individuals to maximize household utility (e.g. both members of households may not
wish to purchase the same commodity). However, the coordination problem can be
overcome using expectations of spouses’ endowments. If information serves as a
coordination mechanism (rather than a bargaining facilitator), then it may lead to either a
divergence or convergence of allocation choice (that is, purchasing the same commodities or
complementary consumption bundles). We provide a test for convergence of allocation
choices in appendix C.
Given that information-sharing plays a key role in our experimental setting to simulate
bargaining, it is important to understand what the deviations between household allocations
in the no information condition and information condition represent. In the case of a collective
household model (as noted above), households may use information on spouses’ allocations
differently, which in turn determine their decisions that are taken at the household level.
However, in the case of a unitary model, information may not have a role to play, since
spouses will have a tendency toward a cooperative allocation that maximizes the utility of the
household without considering their individual utility functions. Sen (2001) suggests that there
may be various factors that determine the position of a woman in intra-household bargaining
dynamics – such as education and employment opportunities. We account for these in our
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
4 A number of studies have also studied the role of initial endowments in intra-household resource allocation (Doss 2013;
Browning & Bourguignon, 1994). Since this was not the core focus of the study, endowments were kept equal.

!
9!
!
analyses and provide correlational insights using heterogeneous treatment effects in section
III.
Another important merit of our design is the involvement of real monetary endowment
and real commodities. Blumenschein, Johanesson, Blomquist, Liljas and O'Conor (1998)
argue that individuals may treat hypothetical decisions differently from real decisions. The
experimental setting in this way was a replica of a real-life situation. Introduction of real
money and real commodities helped provide a real-life bargaining situation in the
experimental setup. One concern was that a real purchase task may not be incentive-
compatible, since households may decide to stock or sell the commodities that they receive
as payoffs from the experiment. There are two compelling context-specific reasons against
this argument: (a) Stocking or selling in rural India comes with additional costs in terms of
time and effort that may not outweigh the benefits of household consumption5, given the
utility ascribed to these specific commodities6. Under the assumption of present bias,
individuals are rational to prefer present consumption over future consumption and are
therefore disinclined to stock commodities; (b) By the process of backward induction, if
couples can predict that there will be another bargaining game while redeeming
commodities, they will potentially argue over: (1) who will retain commodities and who will
redeem; or (2) whether both will redeem. In the former case, the member who dictates her
terms in the household-decisions (i.e. has higher bargaining power), is more likely to redeem.
This is mainly because the experimental endowment (50 rupees) was assigned to each
individual, who could use it as per their preferences. Therefore, bargaining outcomes during
experiment and post-experiment through such uncontrolled processes may not be
qualitatively different. In the latter case, it implies that they have equal bargaining power or
do not care about collective household utility. Considering these, it was presumed that
participants would make their choices carefully.
The list of real commodities were selected after focus-group discussions to determine
the most common choice set for individuals who would be the potential participants. On the
basis of these discussions (held during the last week of September, 2015 separately in both
villages where the experiment was administered), it was decided to offer two commodities
per ‘category’ for individuals to choose from. Care was also taken to ensure that these took
into account existing subsidies provided by the local or State governments. These
commodities were offered at market prices.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
5 In addition, participants may anticipate risk of selling below the market price by predicting that many
experimental participants may go for this option. Therefore, local shopkeepers may trade for below the market
prices and they may face a loss.
6 Selling these commodities to a local shopkeeper would involve some negotiation given that households are not
aware of the precise source of purchase of experimental commodities.

!
10!
!
2.2 Experimental design
Experimental setting
The experiment took place between 2nd and 4th October 2015 in the school buildings of Local
Government in two villages: Thethoor (Madurai District) and Mallanampatti (Dindigul District)
in the Indian state of Tamil Nadu. The Social and Economic Profile of Rural India (SEPRI)-2014
collected by the Institute of Rural Management Anand (IRMA), a nationally-representative
dataset that contains detailed information on household consumption, was used for recruiting
households for the experiment. Of these, 70% of the participants were successfully tracked
while the remaining were recruited randomly on field. We selected these villages on account
of logistical convenience and local partnerships that supported the implementation of the
experiment.
Recruiters invited participants and their spouses to the study for which they would
each receive Rs. 100 for participation along with the opportunity to obtain additional
compensation in the form of commodities. At the time of recruitment, a questionnaire was
administered to gather information about household consumption of various commodities.
In addition, individuals’ time and risk preferences were measured using a separate
questionnaire.
A total of 30 experimenters (including two lead experimenters) who were fluent in the
local language as well as English were trained to administer the experiment. The instructions
were translated into the local language so as to ensure uniformity in the delivery of
instructions by these experimenters. Prior to this, the consistency of translation was checked
through back-translation into English.
The experiments were run with 231 married couples. Once they arrived at the site,
participants gave informed consent to participate in the experiment. They were endowed
with Rs. 507 (10 five-rupee tokens) per head to conduct a decision-making task. Both spouses
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
7 Rs. 100 may serve as an incentive given that it represents an average household expenditure when visiting the store in
rural Tamil Nadu. This is also on account of the high level of consumption subsidies that many households avail. We
summarize below a table of average monthly consumption expenditure (with standard errors in parentheses) at the
household level for the commodities from SEPRI (or where specific commodity data not available, group of the
commodity).
Commodity (or group of commodities)
Average monthly
expenditure (Rs.)
Rice (from non-subsidized sources)
1009.95 (4695.03)
Salt
159.89 (436.62)
Personal Care (includes spectacles, torch, umbrella,
lighter, etc.)
2807.72 (43395.76)
Toilet Articles (includes toothpaste, hair oil, shaving
blades, etc.)
101.14 (109.08)
Household Items (electric bulb, tube-light, glassware,
bucket, washing soap agarbatti, etc.)
95.26 (83.84)
Medical expenses (out-of-pocket)
502.29 (945.41)
School Books & Other Educational Articles (newspaper,
library, stationery, etc.)
242.41 (1096.94)
Source: Authors’ own calculations using SEPRI data (2014)

!
11!
!
were sent to two different experimenters sitting in different corners of the room. Each
experimental space was arranged at a sufficient distance and in the opposite direction to
maintain privacy of the participants’ decisions. The adequate physical distance between them
prevented any strategic communication between spouses.
However, both were informed that their partner was also engaged in the same
decision task. The experimenter gave instructions for each stage of decision-making
sequentially (see Appendix A for instructions and experimental forms). Participants were
asked to observe the sample product of commodities displayed on a table with respective
price tags (written in the local language) and were allowed to gauge the quality of the
products by handling them (photo in appendix D).
The available commodities consisted of:
a) Food
i. Rice (Rs. 15) in half-kilogram bags of superior quality
ii. Salt (Rs.10) in half-kilogram bags, of superior quality
b) Health
i. Paracetamol (Rs. 15) in one strip of ten tablets
ii. Pain Relief Balm (Rs. 10) in 5 sachet packs
c) Education
i. Notebooks (Rs. 15) per unit
ii. Pens (Rs. 10) per unit
d) Sanitation
i. Soap (Rs. 10) per unit
ii. Toothpaste (Rs. 10) per unit
Participants were first asked to state their intent-to-buy or first-hand preferences over
commodities, given the endowment allocated to them. Care was taken not to prompt or
guide the participants in any way; experimenters helped them only with calculation to ensure
that expenses did not exceed the endowment. Similarly, they were informed that any unspent
amount was not redeemable in cash. To facilitate the decision-making and calculation,
participants were instructed with an example as shown in the instructions in the appendix.
Within the time interval of 2-3 minutes, participants were asked to place the final order after
which they exchanged tokens with the experimenter. They were informed that revisions in
the intention-to-buy were allowed. After this final-order was placed, participants were also
asked to state their expectations about spouses’ final order. This information was recorded
in the expectation form (Appendix A).

!
12!
!
Experimental treatment
Among all participants, a randomly selected group of 129 couples (258 individuals or
56% of the experimental sample) followed the experimental procedure described above. We
call this group the no-information treatment group. The rest of the participates were
randomly selected to be part of what we call the full-information treatment group. This group
followed the same procedure explained before, except that spouses’ intention-to-buy list was
shared with the partner during the time interval. After reading their spouses’ intention to buy,
participants were free to alter their intention and make the final consumption decision.
Similarly in the no-information treatment, they were allowed to change their intentions while
making final choices during the same time span.
Participants in both treatment groups submitted an intention-to-buy form, a final-
order form and an expectation-form at the central table. They received their order along with
the showup fee after signing the receipt. Table 1 shows that individuals assigned to the
information group did not differ significantly in basic household characteristics from those in
the control group.
Table 1: Randomization and balanced table
Mean#
Control Group (N=258)
Treatment Group
(N=204)
T statistics
(p value)
Observations
Age
46.1
44.45
0.953
(0.342)
402
Household size
3.63
3.85
-1.139
(0.256)
399
Years of education
5.74
6.12
-0.786
(0.433)
298
Caste identity
226
177
0.494^ (0.482)
391
Scheduled Caste
33
32
Other Backward Castes
181
145
Note. # adjusted for cluster at household level ^ Pearson chi-square; chi-square was not used for Scheduled Tribe
and Other Caste since expected cell values were below five (McHugh, 2013).
Results of t-test for equality of means between these two groups shows the absence of
statistically significant differences for the characteristics under consideration. Table 2 further
describes the differences between males and females within the treatment and control group
on individual and household characteristics. We also report measures of subjective well-
being, happiness, risk aversion, and impatience. Of these, only data on risk aversion was used
in further analyses owing to superior data reliability of these measures.

!
13!
!
Table 2: Individual and household characteristics
Variable
Information
Obs
No Information
Obs
Male
Female
Male
Female
Age
47.92
(11.97)
41.11
(11.69)
93
49.22
(13.34)
42.66
(12.10)
120
Years of Education
7.37
(4.63)
7.89
(5.10)
84
7.52
(4.84)
8.28
(5.51)
111
Household size
3.82
(1.42)
3.83
(1.48)
93
3.66
(1.28)
3.63
(1.26)
120
Percentage Scheduled Caste (SC)
15.69
(36.54)
15.69
(36.54)
102
13.18
(33.95)
13.95
(34.78)
129
Percentage Other Backward Caste (OBC)
68.62
(46.63)
69.61
(46.22)
102
67.44
(47.04)
66.67
(47.32)
129
Percentage Risk Averse a
80.49
(39.87)
77.38
(42.08)
82
83.81
(37.01)
87.62
(33.09)
105
Percentage Happy b
85.39
(35.55)
83.53
(37.31)
85
84.82
(36.04)
82.14
(38.47)
112
Subjective Well-being c
2.50
(0.94)
2.51
(0.88)
85
2.49
(0.54)
2.38
(0.96)
112
Percentage Impatient d
50.84
(50.42)
46.15
(50.24)
59
60.22
(49.22)
54.54
(50.07)
88
Note:
a Response to a hypothetical lottery question: Choice between option 1 that guarantees you an income of Rs. 50,000
per month (risk averse) and option 2: an equal chance of receiving either Rs. 1 lakh per month or Rs. 25,000 per month,
depending on how lucky you are (risk-loving).
b Coded as 1 if ‘very happy’ or ‘happy’ was chosen from “Taken all things together how would you say things are
these days - would you say you were very happy, pretty happy, or not too happy?” and zero otherwise.
c Answer to the question “Please imagine a six-step ladder where on the bottom (the first step), stand the poorest
people, and on the highest step (the sixth step), stand the richest people. On which step are you today?”
d Takes a value of 1 if questions SI21-22(A-C) were answered with choice 1; zero otherwise
Table 3 shows gender-wise differences in the preferences for these commodities, after
controlling for some important demographic characteristics. The preferences are derived
from intent-to-buy expressions of individuals. We observed that women preferred more rice
compared to men whereas men intended to buy pens more than women. These gender-wise
differences were significant at the 5 percent level. This difference in the preferences was also
observed when we combined commodities according to their types. We observed that
women preferred the food bundle (consisting of rice and salt) more than men. In contrast,
men preferred education-related items such as pens and notebooks more. These findings
indicate that there are gender-specific consumption preferences for these commodities,
similar to studies in the past (Duflo, 2003; Van den Bold, Quisumbing & Gillespie, 2013).
Table 3: Gender-wise differences in preferences for commodities
Intention to spend
VARIABLES
Rice
Salt
Soap
Toothpaste
Balm
Tablet
Pen
Notebook
Age (years)
0.0507
(0.0389)
0.0187
(0.0248)
0.0252
(0.0235)
-0.0498**
(0.0248)
0.0409
(0.0250)
0.0368
(0.0286)
-0.065***
(0.0238)
-0.072***
(0.0261)
Years of
education
-0.0139
(0.0826)
-0.0187
(0.0526)
-0.0274
(0.0499)
0.0261
(0.0527)
0.0718
(0.0531)
0.0283
(0.0607)
-0.0931*
(0.0506)
-0.0159
(0.0554)
Household size
-0.436
(0.395)
0.192
(0.251)
0.142
(0.238)
0.105
(0.252)
-0.664***
(0.254)
-0.0117
(0.290)
0.316
(0.242)
0.709***
(0.265)

!
14!
!
Monthly
consumption
expenditure
(Food), Rs.
(SEPRI)
0.0003
(0.0015)
0.0001
(0.0009)
0.0002
(0.0009)
0.0005
(0.0009)
-0.0008
(0.0009)
0.0014
(0.0011)
0.0003
(0.0009)
-0.0017*
(0.0009)
Gender (Male =
1; Female = 0)
-1.787**
(0.902)
0.0454
(0.575)
-0.830
(0.545)
0.619
(0.576)
-0.542
(0.581)
0.432
(0.663)
1.196**
(0.553)
0.583
(0.605)
Constant
12.97***
(2.732)
0.801
(1.740)
7.649***
(1.649)
8.635***
(1.744)
4.697***
(1.758)
-1.153
(2.008)
7.754***
(1.673)
3.249*
(1.831)
Village fixed
effects
Yes
Observations
316
R-square
0.034
0.017
0.018
0.021
0.057
0.021
0.085
0.070
Note. Standard errors in parentheses
*** p < 0.01, ** p < 0.05, * p < 0.1
Experimental outcomes
The bargaining outcome for each individual is interpreted as deviation from the initial
intended amount during the final consumption for each of eight commodities. The
experimental data was constructed in the panel format where each individual had eight
decision-nodes (462*8 = 3,696), along with their demographic variables and applicable
treatment. The variable ‘deviation’ ranges between [-50, 50]. The lowest extreme value of the
variable for a commodity indicates that the individual changed his or her initial preference of
spending the whole amount on that commodity completely and did not buy it at all. Zero
value of the deviation indicated that individuals continued with their initial preferences for a
particular commodity. If an individual had no intention-to-buy a particular commodity but
finally placed an order for it worth Rs. 50, the deviation indicator takes the value 50. All the
interim values are feasible. In order to control for price effects, we also checked for quantities
purchased at the intention, final order, and expectation stages of the experiment. Finally, we
also computed the extent to which the individual’s expectation of their spouse’s final
allocation differed from actual allocations of the spouse. This indicates a parameter of spousal
knowledge on the basis of past consumption decisions or understanding of existing intra-
household dynamics. To test if choices were made randomly and test for incentive
compatibility, we provide a quantile plot in appendix C. We show that allocation choices were
not random by plotting the allocation decisions (Intent and Final) against a discrete uniform
distribution using quantile plots. The graphs show that allocation choice across commodities
is not equally distributed or equally likely across allocation amounts.
Table 4 summarizes the experimental outcomes by commodity. These are the
Intention-to-spend (purchase allocation and quantities), Final allocation (purchase allocation
and quantities), Deviation between final and intention-to-spend (purchase allocation and
quantities), absolute values of Deviations, and the difference between the expectation of
spouses’ final allocation and their actual allocation (Difference (E)). We find no significant

!
15!
!
differences between the information and no information group on these outcomes across
commodities, except in two cases. First, where participants were more accurately able to
guess their spouses’ final allocation in the case of balm in the information condition; and
second, where participants have greater absolute deviations from originally stated intentions
when deciding the final allocation toward purchasing a pen. The first finding suggests that
information about intent-to-spend made a difference to the expected spousal allocations.
The second finding, however, implies that allocation toward purchase of pens was influenced
by information.
Table 4: Mean of experimental outcomes by treatment
Commodity
Intent
Intent (Q)
Final
Final (Q)
Info
No Info
Info
No Info
Info
No Info
Info
No Info
Rice
15.116
15.074
1.008
1.005
16.512
16.544
1.101
1.103
Salt
3.682
3.627
0.368
0.363
3.721
3.480
0.372
0.348
Soap
5.000
5.000
0.500
0.500
5.233
4.706
0.523
0.471
Toothpaste
2.907
2.794
0.194
0.186
2.674
2.941
0.178
0.196
Balm
7.442
7.451
0.744
0.745
6.977
7.157
0.698
0.716
Tablet
7.054
7.157
0.705
0.716
6.357
6.716
0.636
0.672
Pen
3.605
3.480
0.360
0.348
3.682
3.284
0.368
0.328
Notebook
1.977
2.132
0.132
0.142
1.919
2.132
0.128
0.142
Commodity
Difference
Difference (Q)
Absolute Deviation
Difference (E)
Info
No Info
Info
No Info
Info
No Info
Info
No Info
Rice
1.395
1.471
0.093
0.098
2.093
2.353
-2.965
-1.985
Salt
0.039
-0.147
0.004
-0.015
1.047
1.127
0.233
0.490
Soap
0.233
-0.294
0.023
-0.029
1.473
1.961
-0.620
-0.882
Toothpaste
-0.233
0.147
-0.016
0.010
1.163
1.324
0.233
1.299
Balm
-0.465
-0.294
-0.047
-0.029
1.240
1.275
0.039
1.225**
Tablet
-0.698
-0.441
-0.070
-0.044
1.395
1.422
-0.349
0.294
Pen
0.078
-0.196
0.008
-0.020
0.775
1.471**
-1.202
-0.686
Notebook
-0.058
0.000
-0.004
0.000
0.872
0.588
-0.252
0.074
Note. ** significant at 5% level using t-test for differences in means. Intent refers to Intention-to-spend (purchase
allocation and Intent (Q) refers to Intention-to-spend quantities, Final refers to Final purchase allocation and Final (Q)
refers to quantities of final allocation commodities, Difference refers to deviation between final and intention-to-spend
in terms of purchase allocation. Difference (Q) refers to the difference between final and intention-to-spend quantities.
Absolute Deviation refers to absolute values of Deviations, and Difference (E) represents the difference between the
expectation of spouses’ final allocation and their actual allocation.
III. Data analysis and results
As evident in Figure 1 (appendix B), the proportion of individuals who continued with
their initial preferences was substantially higher than those who changed. This is true for both
groups. Approximately 89.4% of the observations in the no-information group underlined
that the final order was consistent with their initial preferences. This proportion was marginally
lower (87.5%) for the full-information group. The mean deviation from intent-to-buy while

!
16!
!
placing the final order was Rs.0.036 for the full-information treatment. It was marginally higher
than the mean deviation of Rs.0.031 for the no-information treatment. The t-test result (p
value = 0.96) also indicates a non-significant difference in mean value of deviations between
these two groups. Similarly, the variations in the deviation for these two groups were slightly
different at 4 and 4.2 respectively.
It is important to note that both spouses entered the experimental room together and
had information that their partner was also participating in the same experimental task. This
might have elicited a sense of collective utility. The spouses are assumed to maximize the
collective utility either by purchasing all possible commodities jointly or by increasing the
quantity of the most important commodities. In the absence of communication and
information about intent-to-buy, individuals may have formed expectations about their
partners’ final choices and would have aligned their own choices accordingly. If expectations
about their partner’s final order are well-accounted in the initial preferences by the individual,
he/she may not change his/her preferences while placing the final order for the commodities.
This will result in zero deviation. We gathered information about the individual’s expectations
about their partner’s final order after they placed their own final order. We calculated
commodity-wise difference between the individuals’ expectations about their spouses’ final
orders and their spouses’ final orders. Lower frequency of such differences represent greater
accuracy of prediction of spouses’ final orders and therefore a higher chance of adopting it
in their own intentions. This is strongly plausible for the no-information group, whose initial
expectations about their partners’ final order were not influenced by the information-sharing.
As shown in Figure 2 (appendix B), around 62.16 percent of observations in the no-
information group were valued at zero. This implies that commodity-wise expectations
matched with those in the final order for relatively more observations. The full-information
group observed 70.34% observations being zero. The higher proportion of correct guesses
(as indicated by zero deviations) underlines that information about partners’ initial
preferences improved expectations.
These results are substantiated by the Logit regression result, as shown in Table 5. It
unambiguously indicates that individuals have a better notion of their spouses’ and own
bargaining position in the household. It shows that commodity-wise the full-information
group were 36% more likely to report more correct expectations about their partner’s final
order compared to the no-information group.

!
17!
!
Table 5: Result of Logit Regression for correct guesses
Guesses
Coefficient
Robust Standard
Errors
P value
Treatment =1 for Full-information ; 0
otherwise
0.3673
0.0947
0.000 ***
Constant
0.4964
0.0536
0.000***
Number of obs = 3,696
Wald chi2(1) = 15.04
Prob > chi2 = 0.0001
*** significant at 1 percent
Deviations between final and intent-to-order stage
We use a simple OLS model to determine the impact of the information treatment on
experimental outcomes. We also use an interaction with gender and individual characteristics
(such as age, years of education, and program participation) to better understand
heterogeneous effects. The model is estimated as:
C
.D
E, *F ; * GHI.; * GJ*I.* K L.; * GMI.* K N.; * GON.; * P
.; * Q.D* (01)
Where, C
.D
E is the outcome of interest (k = deviations in purchase allocations,
quantities, and absolute deviations by the ith participant for the jth commodity), I. is a dummy
variable that takes a value of 1 if the participant belonged to the full-information group, and
zero if belonging to the no-information group. Individual and household characteristics (age,
squared age, years of education, household size, and risk preferences) enter through a linear
term N. as well as an interaction term with information. P
. is the village fixed-effect. All
regressions are clustered at the level of the household.
Table 6 shows regression results for deviation from initial preferences while placing
the final order. Statistically non-significant coefficient for the treatment variables confirms that
sharing information about spouses’ initial preferences, after controlling for various
demographic and experimental variables (such as commodities), did not bring any systematic
changes in individuals’ final orders (compared to the no-information group).
Table 6: Impact of information on experimental outcomes
(1)
(2)
(3)
VARIABLES
Difference between final
and intention order
Difference between final and
intention order (Quantities)
Absolute
deviations
Information
0.0188
-0.000119
0.0790
(0.0304)
(0.00418)
(0.290)
Information * Female
-0.0739
-0.00775
-0.0254
(0.0804)
(0.00681)
(0.355)
Female
0.125
0.000935
0.634
(0.0808)
(0.00908)
(0.420)
Rice
1.504***
0.105***
1.456***
(0.362)
(0.0249)
(0.288)
Salt
-0.0190
0.000598
0.337*
(0.243)
(0.0233)
(0.194)
Balm
0.0437
0.00674
0.940***
(0.270)
(0.0280)
(0.229)
Tablet
-0.0470
-0.0344
0.489**

!
18!
!
(0.286)
(0.0221)
(0.240)
Soap
-0.369
-0.0528**
0.510**
(0.239)
(0.0232)
(0.196)
Paste
-0.553**
0.00281
0.660***
(0.267)
(0.0242)
(0.195)
Pen
0.00467
0.00352
0.334*
(0.215)
(0.0191)
(0.188)
Difference between expectation
and final order (of spouse)
0.0179
0.00167*
-0.00222
(0.0118)
(0.000979)
(0.00970)
Constant
-0.275
0.00327
-1.683
(0.298)
(0.0309)
(1.208)
Village fixed effects
Yes
Household and individual controls
Yes
Observations
3,696
3,696
3,696
R-squared
0.020
0.016
0.042
Note. Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1
Standard errors are adjusted for 230 clusters at household level.
Missing values for variables were replaced by the mean value of the respective variable. Commodity fixed-effects are
relative to base variable of purchase allocation toward Notebook.
We find that the average deviation8 for individuals in the full-information group was
marginally higher than the no-information group, but it was not statistically significant. This
result is in accordance with t-test results discussed above. This result did not change even
when considering the quantities purchased. The interaction effect of information with gender
suggests that female participants reduced their final allocation from their original intention,
although the effect was not statistically significant.
We further conducted a post hoc power analysis to check whether our non-significant
results were caused by lack of statistical power. The observed treatment effect was relatively
low at 0.019 and the power was 0.43. It implies that with the given sample size, there is 43%
probability of detecting a treatment effect of 0.019 correctly. As the magnitude of the effect
was so small, the observability of the treatment effect remains low. If the effect were at least
0.5, we would have 90% probability of detecting it. It is evident form Figure 3 in Appendix B.
The regression results for absolute deviation further confirms that sharing partners’
initial preferences did not bring any systematic change in the average deviation of two
groups. However, deviations in allocations vary by commodity significantly. This is potentially
due to the fact that men and women have different preferences for each commodity, but also
that individuals considered some commodities of greater (or lesser) importance when they
had to make a conclusive purchase decision. In these cases, it was found that absolute
deviations were more likely across all commodities. While individual and household
characteristics did not significantly influence deviations in purchase decisions, there were
some statistically significant impacts on absolute deviations: the absolute deviations reduce
(by 4%, p < 0.1) with an increase in education, and increase linearly with an increase in
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
8 The treatment effect was not statistically significant (t-test = 1.11 ; p value = 0.27)) even when only non-zero
deviations (423 observations) were analysed. There were 219 non-zero deviations for the no-information group
and 204 observations for the full-information group.

!
19!
!
participant age (by 9.7%, p < 0.05). It implies that with education, individuals are likely to
continue with their own preferences.
Changes in spouses’ final allocations
We further explore the role of information in altering spousal decisions on purchase
allocations by regressing the intention of the spouse on the final allocation of the participant.
When interacted with information, this will tell us about the incremental impact of providing
this information explicitly (as opposed to situations where they may be internalized due to
previous bargaining experiences in their regular lives as in the case of the no-information
group). We use the following model:
R.D , *S ; * THUV.D ; * TJUV.D K* I.; * TMI.; * TON.; P
.; *W* (02)
Where, R.D is the final allocation decided by the ith participant on the jth commodity,
UV.D is the intent-to-spend of the spouse of the ith participant on the jth commodity, and I. is
a dummy variable as in equation 1. Table 7 shows the results of estimating equation 2 by
OLS.
Table 7: Information and final allocation of spouse
(1)
(2)
(3)
(4)
VARIABLES
Final allocation
of female
Final allocation
of male
Final allocation
of female
(Quantities)
Final allocation
of male
(Quantities)
Spouse’s intention to spend
0.271***
0.221***
0.725***
0.777***
(0.0446)
(0.0429)
(0.0369)
(0.0418)
Information * Spouse’s intention to spend
-0.147***
-0.0291
-0.0208
-0.0342
(0.0523)
(0.0492)
(0.0413)
(0.0471)
Information
0.969***
0.169
0.0121
-0.00279
(0.328)
(0.301)
(0.0222)
(0.0281)
Rice
13.09***
10.46***
0.355***
0.299***
(0.901)
(0.844)
(0.0512)
(0.0427)
Salt
1.077**
0.389
0.0461
0.0499
(0.487)
(0.530)
(0.0281)
(0.0303)
Balm
2.445***
1.592***
0.0780**
0.105***
(0.513)
(0.535)
(0.0374)
(0.0340)
Soap
3.707***
2.977***
0.0889***
0.136***
(0.516)
(0.563)
(0.0332)
(0.0341)
Paste
3.269***
2.694***
0.0952***
0.0753**
(0.502)
(0.583)
(0.0318)
(0.0332)
Pen
0.660
1.074**
0.0119
0.0867***
(0.483)
(0.523)
(0.0267)
(0.0327)
Notebook
-0.201
-0.758
-0.0204
0.00440
(0.499)
(0.500)
(0.0219)
(0.0252)
Difference between expectation and final
order (of spouse)
0.186***
0.175***
0.00325**
0.00403***
(0.0308)
(0.0284)
(0.00144)
(0.00136)
Constant
1.432**
2.668***
0.0810**
0.0292
(0.580)
(0.546)
(0.0368)
(0.0310)
Village fixed effects
Yes
Household and individual controls
Yes
Observations
1,848
1,848
1,848
1,848
R-squared
0.443
0.381
0.656
0.676
Note. Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1 Standard errors clustered at household
level.

!
20!
!
We note that information has a significant impact on the final purchase allocation of
females in the sample. If the final allocation decision was indeed influenced by the level of
the intent-to-spend of her husband, then making this information available reduces the
female’s overall allocation. When her husband's claim is made known, she is more likely to
concede by reducing her claim in the final allocation. Thus, she has a lower cost of revocation,
implying a weaker bargaining position. The convergence on allocation tests is included in
appendix C. We also find that the intention decision of the spouse has a statistically significant
and positive impact on the final allocation decision of the participant, regardless of whether
this information was available to her/him. This may be indicative of preference-congruence
in the case of some commodities between the spouses, a finding first mentioned in Table 3.
It also implies that the spouse’s intention significantly predicts a participant’s final allocation
decision and that commodities offered for purchase in the experiment had strong demand
within the households.
IV. Heterogeneous treatment effects
We also investigate the potential heterogeneity of the treatment impact of providing
information on various experimental outcomes. We hypothesize that the treatment effects
may vary by differences in age, years of education, and program participation between
spouses. These were tested using interactions with the information treatment variable in
equations 1 and 2. Note that these are not to be treated as causal impacts but rather as
correlational in nature and are described in Tables 8a and 8b.
Table 8a: Heterogeneous treatment effects on final allocations, difference between final and
intent-to-spend, and absolute deviations
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
VARIABLES
Final
allocation
of female
Final
allocation
of male
inal
allocation
of female
F Final
allocation
of male
inal
allocation
of female
Final
allocation
of male
Difference
between
final and
intention
order
Absolute
deviations
Difference in
age
-0.0109
-0.00321
-0.00121
(0.00913)
(0.0112)
(0.00507)
Information *
difference in
age
0.00919
0.0153
0.00416
(0.0149)
(0.0145)
(0.00749)
Difference in
years of
education
0.00239
-0.00610
0.00280
(0.0108)
(0.00932)
(0.00581)
Information *
difference in
years of
education
-0.0339**
0.0182
-0.0116
(0.0162)
(0.0139)
(0.00876)
Both
participate
0.104
0.357***
(0.126)
(0.132)
Wife
participant
0.0668
0.0288
(0.130)
(0.105)

!
21!
!
Note. Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1
Standard errors clustered at household level.
Differences in age. Table 8a shows the impacts of differences in age on experimental
outcomes. A greater difference in ages between spouses is meant to approximate
experience. We find no statistically significant impact on deviations (purchase and quantities)
as well as absolute deviations. There was also no impact found on the final allocations of the
spouse when the decision took into account spouses’ intent-to-spend.
Differences in years of education
.
A larger difference in years of education
completed between spouses was hypothesized to have an impact on allocation decisions.
Table 8a shows no statistically significant impact on the deviations; the final allocation
decisions of a participant when taking into account their spouse’s intention to spend varied
inversely with differences in years of education. In the full information group, a greater
difference in years of education resulted in a lower final allocation by female participants (-
0.034, p < 0.05).
Program participation. Within the empirical literature on intra-household bargaining
and allocation of resources, a large number of studies focus on the impact of participation in
welfare programs for households (e.g., Ferro, Kassouf, and Levison 2010). Welfare programs
such as PROGRESA (Programa de Educación, Salud y Alimentación–Education, Health and
Nutrition Program)-Oportunidades in Mexico and Bolsa Familia in Brazil have found varied
impacts on intra-household resource allocation. Handa, Peterman, Davis and Stampini (2009)
find that income earned from PROGRESA is not spent differently from general income, with
husbands and wives having common preferences with respect to consumption expenditure.
While they do find benefits for child nutrition, healthcare, and food consumption
expenditure, they do not attribute it to having a female-targeted beneficiary. Their results
are thus not representative of how middle-class households typically react to exogenous
changes in female non-labour income.
Husband
participant
0.00130
0.207
(0.154)
(0.205)
Both
participate *
Information
-0.193
-0.238
(0.181)
(0.165)
Wife
participant *
Information
-0.361*
-0.112
(0.189)
(0.145)
Husband
participant *
Information
-0.0525
-0.111
(0.258)
(0.254)
Constant
1.409**
2.804***
1.397**
2.779***
1.432**
2.668***
-0.329
-0.348
(0.571)
(0.584)
(0.574)
(0.578)
(0.580)
(0.546)
(0.367)
(0.352)
Observations
1,848
1,848
1,848
1,848
1,848
1,848
3,696
3,696
R-squared
0.443
0.380
0.443
0.380
0.443
0.381
0.020
0.020

!
22!
!
Table 8b: Heterogeneous treatment effects of MGNREGS participation on difference between
final order and intent-to-spend
(1)
(2)
VARIABLES
Difference between Final
Order and Intent-to-spend
Absolute Deviation
Information
0.0188
0.0790
(0.0304)
(0.290)
Information * Female
-0.0739
-0.0254
(0.0804)
(0.355)
Both participate * Female
0.0460
-0.0181
(0.137)
(0.453)
Both participate * Male
1.111**
(0.545)
Female participant * Female
-0.148*
-0.242
(0.0817)
(0.369)
Female participant * Male
0.0818
(0.427)
Male participant * Female
0.144
1.261*
(0.192)
(0.686)
Male participant * Male
0.626
(0.601)
Constant
-0.275
-1.683
(0.298)
(1.208)
Observations
3,696
3,696
R-squared
0.020
0.042
Note. Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1
Standard errors clustered at household level.
In our study, a large group of households (N = 194) had at least one participant who
was part of a government-sponsored welfare scheme. This programme (The Mahatma
Gandhi National Rural Employment Guarantee Scheme, or MGNREGS) aims to improve the
welfare of men and women independently by offering them temporary employment for a
fixed duration of 100 days and for standard wage between genders (variable between Rs.
100 – Rs. 150 per day; USD 1.87 on average). Indeed, the major objective of this programme
is to create a strong social safety net for vulnerable groups (which include women). This
improvement in social and economic security may be manifested in the form of better quality
and quantity of consumption, potentially due to independent decision-making ability within
the household. Participants were categorized under four groups – the first group where only
the wife participated, the second group where only the husband participated, the third group
where both participated and the fourth group where none of them participated (taken as the
control group).
Table 8b shows the impacts of participation in MGNREGS on experimental outcomes.
Male participants in households where both individuals take part in MGNREGS (1.11; p <
0.05), and female participants in households where the male is the sole MGNREGS
participant (1.26, p < 0.1) have larger absolute deviations. For females in households where
she is the sole MGNREGS participant, we find reductions in final allocations relative to the
original intention to spend (-0.15, p < 0.1). In terms of spousal final allocations, we see that
the male members’ final allocations (across commodities) are significantly higher in
households where both participate (0.357, p < 0.01). Conditioned on information availability,

!
23!
!
there is a significant reduction in the female’s final allocation in households where she is the
sole MGNREGS participant (-0.36, p < 0.1). These indicate that participation in MGNREGS
may be associated with intra-household resource allocation decisions, and that these
associations are mediated by gender.
V. Conclusions and policy implications
We administered a quasi-field experiment to elicit intra-household bargaining power with
respect to jointly-consumed commodities at the household level. The existing literature
points out gender-wise differences in the allocation decisions; women were observed to be
more concerned about the collective welfare of household and were spending more on
education and health services. Such expenditure generates positive externalities from the
development perspective. However, women’s weak bargaining power in the household may
prove to be a serious impediment. This has implications on the welfare of society.
We attempted to initiate the bargaining process by sharing information about
spouses’ preferences for jointly-consumed household commodities with the individual in the
treatment group. We observed sharing information about the spouse’s initial preferences did
not initiate the bargaining interaction. We also observed that individuals can form better
expectations about their partners’ choices even in the absence of information about the
preferences. This finding implies that individuals have knowledge about their partners’
preferences and increases efficient intra-household resource allocation. However, mere
awareness about partner preferences does not ensure that efficient allocation will always be
obtained. Instead, it will be subject to the asymmetry in the intra-household bargaining
power. Individuals with low bargaining power may deviate more frequently from their initial
preferences while making final choices.
In this setting, we found correlational interactions between intra-household
bargaining dynamics and a large-scale employment guarantee programme (MGNREGS). A
well-designed welfare scheme with the goal of empowering women may provide financial
resources to women but may not obtain desirable results if their bargaining position is low.
Our results indicate that intra-household bargaining positions are associated with
participation in the employment guarantee scheme. Causal relations may not be drawn based
on such a quasi-field experiment, but our study sets the agenda for future research in terms
of developing methodology to elicit the intra-household bargaining process and investing
the impact of welfare schemes on the bargaining positions of the beneficiaries.

!
24!
!
References
ASHRAF, N. (2009). Spousal Control and Intra-Household Decision Making: An
Experimental Study in the Philippines. American Economic Review, 99(4), 1245–1277.
doi:10.1257/aer.99.4.1245.
BASU, K. (2006). Gender and Say: A Model of Household Behaviour with Endogenously
Determined Balance of Power. The Economic Journal, 116(511), 558–580. doi:
10.1111/j.1468-0297.2006.01092.x
BEBLO, M., BENINGER, D., COCHARD, F., COUPRIE, H., & HOPFENSITZ, A. (2015).
Efficiency-Equality Trade-off within French and German Couples A Comparative
Experimental Study. Annals of Economics and Statistics, 117-118(June 2015), 233–252.
Retrieved from: http://ideas.repec.org/a/adr/anecst/y2015i117-118p233-252.html.
BINMORE, K. (1998). Game Theory and the Social Contract: Just Playing (1st ed.).
Cambridge, MA: MIT Press. Retrieved from: https://mitpress.mit.edu/books/game-theory-
and-social-contract-0.
BLUMENSCHEIN, K., JOHANNESSON, M., BLOMQUIST, G, C., LILJAS, B., &
O’CONOR, R, M. (1998). Experimental Results on Expressed Certainty and Hypothetical Bias
in Contingent Valuation. Southern Economic Journal, 65(1), 169-177. doi:10.2307/1061360.
BROWNING, M., & CHIAPPORI, P. A. (1998). Efficient Intra-Household Allocations: A
General Characterization and Empirical Tests. Econometrica, 66(6), 1241-1278.
doi:10.2307/2999616.
BROWNING, M., & BOURGUIGNON, F. (1994). Income and Outcomes: A Structural
Model of Intrahousehold Allocation. Journal of Political Economy, 102(6), 1067-1096.
Retrieved from: http://www.jstor.org/stable/2138780.
CASTILLA, C., & WALKER, T. (2013). Is Ignorance Bliss? The Effect of Asymmetric
Information between Spouses on Intra-Household Allocations. American Economic Review,
103(3), 263–268. doi:10.1257/aer.103.3.263.
COCHARD, F., COUPRIE, H., & HOPFENSITZ, A. (2014). Do Spouses Cooperate? An
Experimental Investigation. Review of Economics of the Household, 14(1), 1-26. doi:
10.1007/s11150-014-9276-3
DASGUPTA, U., & MANI, S. (2015). Only Mine or All Ours: Do Stronger Entitlements
Affect Altruistic Choices in the Household. World Development, 67(March), 363–375.
doi:10.1016/j.worlddev.2014.10.021.
DOSS, C. (2013). Intrahousehold Bargaining and Resource Allocation in Developing
Countries. World Bank Research Observer, 28(1), 52–78. doi:10.1093/wbro/lkt001.
DUFLO, E. (2003). Grandmothers and Granddaughters: Old Age Pensions and
Intrahousehold Allocation in South Africa. The World Bank Economic Review, 17(1), 1–25.
doi:10.1093/wber/lhg013.

!
25!
!
FERRO, A., KASSOUF, A. A., & LEVISON, D. (2010). The Impact of Conditional Cash
Transfer Programs on Household Work Decisions in Brazil. In R. K. Q. Akee, E. V. Edmonds,
& K. Tatsiramos (Eds.), Research in Labor Economics: Vol. 31. Child Labor and the Transition
between School and Work (pp.193–218). Bingley, UK: Emerald Group Publishing Limited.
doi:10.1108/S0147-9121(2010)0000031010.
HANDA, S., PETERMAN, A., DAVIS, B., & STAMPINI, M. (2009). Opening Up Pandora’s
Box: The Effect of Gender Targeting and Conditionality on Household Spending Behavior in
Mexico's Progresa Program. World Development, 37(6), 1129–1142.
doi:10.1016/j.worlddev.2008.10.005.
HODDINOTT, J., & HADDAD, L. (1995). Does Female Income Share Influence
Household Expenditures? Evidence from Côte d’Ivoire. Oxford Bulletin of Economics and
Statistics, 57(1), 77–96. doi:10.1111/j.1468-0084.1995.tb00028.x.
LEENA, S, B. (2014). Brand Amma’s Sops to the Poor. Mint, Retrieved from:
http://www.livemint.com/Politics/J55RkTaPjSFuDlmFsrWcAO/Brand-Ammas-sops-to-the-
poor.html.
LEVITT, S, D., & LIST, J. A., (2007). What Do Laboratory Experiments Measuring Social
Preferences Reveal about the Real World? Journal of Economic Perspectives, 21(2), 153–174.
doi:10.1257/jep.21.2.153.
MANI, A. (2011). Mine, Yours or Ours? The Efficiency of Household Investment
Decisions: An Experimental Approach. CAGE Online Working Paper Series, 2011(64), 1-45.
Retrieved from: http://wrap.warwick.ac.uk/id/eprint/57668
MCHUGH, M, L. (2013). The Chi-Square Test of Independence. Biochemia Medica,
23(2),143–149. doi: 10.11613/BM.2013.018
MUNRO, A., KEBEDE, B., TARAZONA-GOMEZ, M., & VERSCHOOR, A. (2014).
Autonomy and Efficiency. An Experiment on Household Decisions in Two Regions of India.
Journal of the Japanese and International Economies, 33(September), 114–133.
doi:10.1016/j.jjie.2013.10.009.
MUTHOO, A. (1992). Revocable Commitment and Sequential Bargaining. The
Economic Journal, 102(March), 378–387. doi:10.2307/2234522.
MUTHOO, A. (1996). A Bargaining Model Based on the Commitment Tactic. Journal of
Economic Theory, 69(1), 134–152. doi:10.1006/jeth.1996.0041.
QUISUMBING, A. R. (1996). Male-Female Productivity: Differences in Agricultural Issues
and Empirical Evidence. World Development, 24(10), 1579–1595. doi:10.1016/0305-
750X(96)00059-9.
QUISUMBING, A.R., & MALUCCIO, J. A. (2000). Intrahousehold Allocation and Gender
Relations: New Empirical Evidence from Four Developing Countries. IFPRI, 84. FCND
Discussion Paper. Retrieved from:
https://www.atria.nl/epublications/2000/Intrahousehold_allocation_and_gender_relations.p
df
SEN, A. K. (2001). Development as Freedom. Oxford: Oxford University Press.

!
26!
!
THOMAS, D. (1990). Intra-Household Resource Allocation: An Inferential Approach.
The Journal of Human Resources, 25(4), 635-664. doi:10.2307/145670.
THOMAS, D. (1994). Like Father, Like Son; Like Mother, Like Daughter: Parental
Resources and Child Height. The Journal of Human Resources, 29(4), 950–988.
doi:10.2307/146131.
UDRY, C., HODDINOTT, J., ALDERMAN, H., & HADDAD, L. (1995). Gender
Differentials in Farm Productivity: Implications for Household Efficiency and Agricultural
Policy. Food Policy, 20(5), 407-423. doi: 10.1016/0306-9192(95)00035-D
D. VAN DEN BOLD, M., QUISUMBING, A, R., & GILLESPIE, S. (2013). Women’s
Empowerment and Nutrition: An Evidence Review. IFPRI. Discussion Paper no. 01294.
doi:10.2139/ssrn.2343160.
YANG, X., & CARLSSON, F. (2012). Intra-Household Decisions Making on Intertemporal
Choices : An Experimental Study in Rural China, Working Papers in Economics No. 537.
Retrieved from: https://gupea.ub.gu.se/bitstream/2077/30274/1/gupea_2077_30274_1.pdf
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!

!
27!
!
Appendix A
!
Instructions Real Purchase: No Information Condition
This is an experiment to study decision-making between husbands and wives in the
household.
The instructions are very simple. Please listen to them carefully.
• This experiment consists of you making some decisions. Your spouse is making the same
decisions separately.
• During the task, you will be given a certain amount of money. This money will be given as
five-rupee tokens. In total, there will be 20 five-rupee tokens divided between your spouse
and yourself equally. Thus, you and your spouse will get 10 coins each.
• There are no right or wrong decisions; please play the task as truthfully as possible.
Any questions? Yes – Ask me. No – Continue.
1.
Intention-to-Order Form
• First, you will be required to state your intention to purchase items from some
commodities displayed on the table, with their prices. [Items are on display] Any unspent
tokens will not be given to you, so you should attempt to use all the money you have.
Please tell me what and how much of each commodity you would like to purchase, and I
will make a note of this.
• We are not allowed to pay you directly with cash, so we have selected this group of items
from which you can freely choose up to an amount of Rs. 50 that has been allocated to
you.
Any questions? Yes – Ask me. No – Continue.
• Following is an illustration of how you can fill in the order form. Suppose you initially have
Rs. 50 (10 five-rupee tokens), then you can choose to allocate the tokens as under:
Sample Intention Order List
[Please indicate below what you
intend
to spend]
Item
Price (Rs.)
Number
Total Expenditure (Rs.)
Painkiller (crocin)
15
1
15
Notebooks
15
1
15
Pen
10
0
-
Rice (1/2 kg)
15
1
15
Zandy Balm (5 units)
10
0
-
Toothpaste
10
0
-
Salt (1/2 kg)
10
0
-
Soap
10
0
-
Total
95
3
45
• Now, please tell me what and how much you would like to purchase.”
[Participant completes Intention-to-Order form; experimenter notes decisions.]
2.
Final Order Form
• I will now give you two minutes to consider your decision and tell me what you would
finally like to purchase.

!
28!
!
• Once you tell me your final decisions, you will not be able to change them. I will
communicate this to the payment desk, and they will then arrange for the commodities
and give you the items. Please pay me with the tokens to purchase those items.
• There are no right or wrong decisions.
Any questions? Yes – Ask me. No – Continue.
• Following is an illustration of how you can fill in the order form. Suppose you initially have
Rs. 50 (10 five-rupee tokens), then you can choose to allocate the tokens as under:!
!
Sample Expenditure Order List
[Please indicate below what you will spend]
Item
Price (Rs.)
Number
Total Expenditure (Rs.)
Painkiller (crocin)
15
1
15
Notebooks
15
2
30
Pen
10
0
-
Rice (1/2 kg)
15
1
-
Zandy Balm (5 units)
10
0
-
Toothpaste
10
0
-
Salt (1/2 kg)
10
0
-
Soap
10
0
-
Total
95
3
45
If you have any questions, or need assistance of any kind, please ask me. We expect and
appreciate your cooperation. We assure you that the results of this experiment or any other
details will not be disclosed to anyone, and you will not be identified by name. The data
collected are strictly for the purposes of research.”
[Participants given the Final Order form.]
3. Expectations Form
Your spouse is making the same decisions separately. Please tell me how do you think
he/she spent the Rs. 50.
[Participant completes the expectations form; experimenter notes decisions.]
4. Payment and Receipt
• “Hello, we are not allowed to pay you directly with cash, so we have selected this group
of items from which have chosen up to an amount of Rs. 50 that has been allocated to you.
• I will now arrange for the commodities that you have chosen and you will receive them in a
bag. I will also now pay you Rs. 100 for your time in attending the experiment. Thank you
for your participation!”
Instructions Real Purchase: Full Information Condition
This is an experiment to study decision-making between husbands and wives in the
household.

!
29!
!
The instructions are very simple. Please listen to them carefully.
• This experiment consists of you making some decisions. Your spouse is making the same
decisions separately.
• During the task, you will be given a certain amount of money. This money will be given as
five-rupee tokens. In total, there will be 20 five-rupee tokens divided between your spouse
and yourself equally. Thus, you and your spouse will get 10 coins each.
• There are no right or wrong decisions; please play the task as truthfully as possible.
Any questions? Yes – Ask me. No – Continue.
5. Intention-to-Order Form
• First, you will be required to state your intention to purchase items from some
commodities displayed on the table, with their prices. [Items are on display] Any unspent
tokens will not be given to you, so you should attempt to use all the money you have.
Please tell me what and how much of each commodity you would like to purchase, and I
will make a note of this.
• We are not allowed to pay you directly with cash, so we have selected this group of items
from which you can freely choose up to an amount of Rs. 50 that has been allocated to
you.
Any questions? Yes – Ask me. No – Continue.
• Following is an illustration of how you can fill in the order form. Suppose you initially have
Rs. 50 (10 five-rupee tokens), then you can choose to allocate the tokens as under:
Sample Intention Order List
[Please indicate below what you intend to spend]
Item
Price (Rs.)
Number
Total Expenditure (Rs.)
Painkiller (crocin)
15
1
15
Notebooks
15
1
15
Pen
10
1
10
Rice (1/2 kg)
15
1
-
Zandy Balm (5 units)
10
0
-
Toothpaste
10
1
10
Salt (1/2 kg)
10
0
-
Soap
10
0
-
Total
95
4
50
Now, please tell me what and how much you would like to purchase.”
[Participant completes Intention-to-Order form; experimenter notes decisions.]
• Now, I will tell you how much your spouse has decided to spend on each of the same
commodities.

!
30!
!
[The intent-to-order forms are exchanged with the spouse’s experimenter and information is
provided. Read out-loud the intent-to order form from the spouse].
6. Final Order Form
• I will now give you two minutes to consider your decision as well as the information of your
spouse’s decisions just communicated to you and tell me what you would finally like to
purchase.
• Once you tell me your final decisions, you will not be able to change them. I will
communicate this to the payment desk, and they will then arrange for the commodities
and give you the items. Please pay me with the tokens to purchase those items.
• There are no right or wrong decisions.
Any questions? Yes – Ask me. No – Continue.
• Following is an illustration of how you can fill in the order form. Suppose you initially have
Rs. 50 (10 five-rupee tokens), then you can choose to allocate the tokens as under:
Sample Expenditure Order List
[Please indicate below what you will spend]
Item
Price (Rs.)
Number
Total Expenditure (Rs.)
Painkiller (crocin)
15
1
15
Notebooks
15
2
30
Pen
10
0
-
Rice (1/2 kg)
15
1
-
Zandy Balm (5 units)
10
0
-
Toothpaste
10
0
-
Salt (1/2 kg)
10
0
-
Soap
10
0
-
Total
95
3
45
If you have any questions, or need assistance of any kind, please ask the experimenter. We
expect and appreciate your cooperation. We assure you that the results of this experiment or
any other details will not be disclosed to anyone, and you will not be identified by name. The
data collected are strictly for the purposes of research.”
[Participants given the Final Order form.]
7. Expectations Form
• Your spouse is making the same decisions separately. Please tell me how do you think
he/she spent the Rs. 50.
[Participant completes the expectations form; experimenter notes decisions.]
8. Payment and Receipt Desk
• “Hello, we are not allowed to pay you directly with cash, so we have selected this group of

!
31!
!
items from which have chosen up to an amount of Rs. 50 that has been allocated to you.
• I will now arrange for the commodities that you have chosen and you will receive them in a
bag. I will also now pay you Rs. 100 for your time in attending the experiment. Thank you
for your participation!”
Appendix B
Figures
0.5 1
-40 -20 020 40 -40 -20 020 40
0 1
Fraction
Commodity-wise Difference between Final order (Rs.) and Intent-to-buy (Rs.)
Graphs by Information (Information = 1; No Info = 0)
Figure 1: Treatment-wise Deviations
Figure 1: Treatment-wise deviations

!
32!
!
Figure 3 : Power analysis
Figure 4: Quantile plot of intent and final allocation
0.2 .4 .6 .8
-40 -20 020 -40 -20 020
0 1
Fraction
Commodity-wise Difference between Expectation (Rs.) and Final order (Rs.)
Graphs by Information (Information = 1; No Info = 0)
Figure 3: Difference between Expectation and Final order
0.2 .4 .6 .8 1
power
0.05 .1 .15 .2
effect
Power as Function of Effect and N=3696
Figure 2: Difference between expectation and final order

!
!
Appendix C
Check for allocation convergence
We find that in 60.77 percent of all responses, choice allocations between
commodities converged. Participants in the full-information condition chose the same final
purchase allocations 59.19 percent of all responses, while participant choices in the no-
information group converged 62 percent of all responses. A proportion test for differences
showed a small but statistically significant effect of information on choice convergence
(0.028, p < 0.1). We also used convergence as a potential outcome variable in regressions
to indicate if household members made the same choices, controlling for a number of
factors as in equation (1). Though we find no statistically significant effect of information
on the likelihood of convergence, the negative sign of information on likelihood of
convergence indicates that participants in the treatment group were less likely to choose
the same commodities.
Table A1: Role of information in choice convergence
(1)
(2)
(3)
VARIABLES
OLS
Logit
Probit
Information
-0.0269
-0.118
-0.0745
(0.0242)
(0.107)
(0.0654)
Information * Male
-0.00142
-0.00684
-0.00359
(0.00468)
(0.0211)
(0.0126)
Rice
-0.125***
-0.571***
-0.346***
(0.0446)
(0.205)
(0.125)
Salt
-0.138***
-0.625***
-0.381***
(0.0441)
(0.203)
(0.123)
Balm
-0.247***
-1.075***
-0.660***
(0.0440)
(0.200)
(0.122)
Soap
-0.164***
-0.737***
-0.450***
(0.0451)
(0.206)
(0.125)
Paste
-0.167***
-0.746***
-0.458***
(0.0433)
(0.197)
(0.120)
Pen
-0.142***
-0.645***
-0.392***
(0.0454)
(0.208)
(0.127)
Notebook
0.0846**
0.477**
0.279**
(0.0390)
(0.222)
(0.130)
Difference between expectation of spouse’s
final allocation and actual allocation
0.00287***
0.0127***
0.00780***
(0.00102)
(0.00449)
(0.00262)
Constant
(0.140)
1.089*
0.657*
0.747***
(0.639)
(0.393)
Observations
3,696
3,696
3,696
R-squared
0.060
0.047
0.046
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

!
!
Appendix D
Photo: Experimental store