WORK IN PROGRESS – PLEASE DON’T QUOTE
Experimental measures of household
bargaining power
Ragnhild Haugli Bråten, Frisch Centre, NorwayA
Peter Martinsson, University of Gothenburg, SwedenB
This version 2012-05-22
Abstract
In this paper we experimentally investigate risk and uncertainty preferences among peasant
couples in Peru, first individually and then jointly to determine intra-household bargaining
power. We find high degrees of extreme risk and uncertainty aversion. Further, individual
preferences are more important than gender in determining bargaining power, implying that
the most risk or uncertainty averse spouse possesses the highest bargaining power. However,
when controlling for individual attitudes towards risk and uncertainty, men exhibit higher
bargaining power than women. We find that female bargaining power is associated with lower
education, women’s organization level, higher altitudes and husbands sometimes residing out
of home. The experimental measures of bargaining power correlate with self-reported
measures of decision power regarding household and school specific expenditures, but not
with decision power in agricultural investments.
Key-words: Risk; Ambiguity; Bargaining power; Female empowerment; Experiment; Peru.
Acknowledgments: This research is financed by Norwegian Research Council though the Latin America
program grant no. 196328 and Environment 2015 and is a part of the Peru Land and Gender research project. We
particularly thank Henrik Wiig and Daniela Fuentes for invaluable discussions and survey design, Lene Sandvik,
Carlos De los Ríos and Oscar Madaleingoita for excellent assistance and discussions, and Kjell Arne Brekke for
helpful comments.
A
Frisch Centre, Gaustadalléen 21, NO-0349 Oslo, Norway; Ph +47 2295 8824; Fax +47 2295 8825; E-mail
[email protected]
B
Department of Economics, University of Gothenburg, Box 640, 405 30 Göteborg, Sweden; Ph +46 31 786 52
55; Fax +46 31 786 10 43; E-mail peter.martinsson @economics.gu.se
1 1. Introduction
Many important household decisions are not made by individuals alone, but are rather
products of a joint decision making process. Collective models of household decision making
assumes that the joint decision is a result of a bargaining process based on the individual
preferences of the spouses (Manser and Brown 1980; Lundberg and Pollak 1993; Browning
and Chiappori 1998; Bourguignon et al. 2009), which is different to the frequently used
unitary model assuming that the household consists of one unique decision-maker.1 The
wife’s bargaining power is believed to increase when she obtains larger resources in terms of
income and assets, by increasing her threat point utility in case of divorce or non-cooperative
behaviour within the household (Lundberg and Pollak 1993; Browning and Chiappori 1998).
Empirical studies have indeed found that transferring income and assets to women from men
may shift household consumption patterns in line with assumedly female preferences,
specifically towards greater investments in nutrition and childcare, and particularly for
daughters (Thomas 1990; Hoddinott and Haddad 1995; Lundberg, Pollak et al. 1997; Phipps
and Burton 1998; Thomas et al. 1999; Duflo 2003). These findings are taken as a support for
collective household bargaining models. They also emphasize the importance of women’s
bargaining power for children’s outcomes and hence countries’ future growth prospects.
These issues are particular relevant in countries with large gender differences and assumedly
low female bargaining power (World Development Report 2012).
Female empowerment is defined by increasing women’s capabilities to make strategic life
choices. Kabeer (1999) identifies this empowerment concept by three inter-related aspects;
possession of resources, agency (including decision making) and well-being achievements.
When women hold larger shares of assets or receive wage income it may increase women’s
empowerment though evolved social norms and perceptions, and not necessarily only by
increasing their threat point in the bargaining process (Agarwal 1997). And just as well as
increased access to resources may push her bargaining power in the household, increased
bargaining power may in turn improve her access to resources (Basu 2006). These interrelated aspects of women’s empowerment are all important policy goals for their own rights.
However, they are often difficult terms to identify empirically. Different approaches have
been taken to measure empowerment, including survey stated decision power, household
1
The collective decision model has been empirically supported over the unitary model (Lundberg et al. 1997;
Browning and Chiappori 1998; Cherchye et al. 2009).
2 consumption allocations, resources and outcomes of fertility, health or education. Several
empirical studies have found that women’s self reported decision power in different aspects of
life increases with participation in credit programs or paid labour, income or land rights
(Hashemi et al. 1996; Friedberg and Webb 2006; Allendorf 2007; Anderson and Eswaran
2009)2. However, to investigate the household decision process and bargaining power as
determined by collective bargaining models, individuals’ preferences need to be elicited
explicitly and compared to the joint decision.
The objective of this paper is to investigate bargaining power among spouses in agricultural
communities in highland Peru in the case of risky and uncertain decisions using an
experimental approach, and to compare experimental and self-reported measures of
bargaining power. By first conducting the experiments with the spouses individually and
afterwards jointly, we determine intra-household bargaining power as a binary variable being
positive for the spouse whose individual decision lies closer to the joint decision.3 The
strength of an experimental method is the possibility to keep a strict control over the decision
process, while revealed data might have confounded effects and unclear causality. There is a
growing literature on household decision making using economic experiments (Peters et al.
2004; Bateman and Munro 2005; Munro et al. 2008; Carlsson et al. 2009; Ashraf et al. 2010;
de Palma et al. 2011). Previous bargaining experiments have focused on risky decisions4,
taking into account that the majority of important household decisions involve some aspects
of uncertainty. Bateman and Munro (2005) compared individual and joint decisions of
spouses to investigate whether different axioms of the expected utility were fulfilled in a
study with 76 British couples. Their results indicate that joint decisions are on average more
risk averse than when the spouses’ make the choices individually. In a related study, de Palma
et al. (2011) investigated the decision power among 22 cohabitating German couples. They
found that husbands typically have more influence on joint decisions, but when wives had a
2
Another approach to investigate decision power is to analyze household surveys on consumption behavior (e.g.,
Mazzocco (2004) and Donni (2003)).
3
We take this simplistic approach partly because decisions under risk and uncertainty may be somewhat shifted
from an individual to a joint setting (shown theoretically by Mazzocco (2004) and experimentally in a tangential
literature, see e.g. Shupp and Williams (2008), Masclet et al. (2009) and Baker et al. (2008)). Overall, the
experimental results seem to indicate that groups are more risk-averse in situations with a low probability of
winning, while opposite holds for high probabilities.
4
The typical findings in risk experiments among individuals are that women are more risk averse than men (see
overview in Croson and Gneezy (2009)).
3 larger degree of influence when they controlled the computer keyboard. Carlsson et al. (2010)
similarly found men to have higher bargaining power in their study of 117 rural Chinese
couples, but that women with higher education, higher share of household income and
communist party membership were more likely to control the joint decision.
To our knowledge, there is no previous experimental study of household bargaining power
under uncertainty, i.e., decisions involving unknown probabilities rather than known
probabilities. Although risky behaviour is easier to measure, uncertain outcomes are more
relevant for real life situations since probabilities of different outcomes are often unknown.
Ellsberg (1961) argued that people prefer a risky situation over an uncertain situation in a
situation where indifference is predicted by expected utility (see also Savage (1954)). Many
experiments have shown support for Ellsberg’s claim, recognizing such preferences as
ambiguity aversion (for recent contributions in this area see e.g. (Borghans et al. 2009;
Trautmann and Zeckhauser 2010; Abdellaoui et al. 2011; Trautmann et al. 2011)). As in most
experiments, there is a limited knowledge about behavior beyond university students, but the
few exemptions we are aware of using samples of professionals generally confirm ambiguity
aversion (Viscusi and Chesson 1999; Maffioletti and Santoni 2005; Cabantous 2007). We
included uncertainty decisions in this analysis to increase the potential comparability of
experimental findings and everyday investment behavior reported by participants. In general,
there is a growing interest in validating behavior in laboratory experiments with field behavior
(Levitt and List 2007; Levitt and List 2007; Camerer 2011). In a large study in Germany,
Dohmen et al. (2011) find positive correlation between risk attitudes elicited experimentally
and self-reported risk behavior in different domains of life.
We measure intra-household bargaining power in the Peruvian highlands by means of lottery
tasks eliciting preferences over risk and uncertainty. Our sample consists of 287 couples
residing in small-scale peasant communities, nearly all households with farming as main
activity. Our results show that a majority of the sample are extremely risk and uncertainty
averse, in line with previous studies of famers in developing countries (Yesuf and Bluffstone
2009; Akay et al. 2011), and that individuals in general are ambiguity averse. Women are
more uncertainty averse than men, but there is no significant difference in risk aversion
between genders. Joint decisions are on average more risk and uncertainty averse than the
individual decisions, implying that the more averse spouse in general has higher bargaining
power in the joint decision. When controlling for this drift towards risk aversion in the joint
4 setting, we find that men have higher bargaining power than women, although this gender
effect is weaker than the effect of individual preferences. Moreover, women are more likely to
have the higher bargaining power when they are organized, their husbands sometimes resides
outside the home, or the couple resides at higher altitudes (communities at higher altitudes are
generally less economically developed). On the other hand, women reporting verbal
mistreatment by their husbands or in couples with higher education reveal lower bargaining
power in our experiments.
2. Design The experimental session was performed in the couple’s home, in the majority of cases with
one instructor performing both the two individual and the joint parts in order to keep all
factors within the household constant5. Only female instructors were used, 8 in total, to keep
possible instructor gender effects constant.
Due to high fractions of analphabets in the rural highlands, we conducted the experiment
orally. The experimental session started with a short introduction held for both spouses at the
same time. The couple was told they were going to make choices over economic outcomes
and that the money they would earn would depend on their decisions. They would make the
choices individually, and could choose themselves who should start. The starting spouse
stayed in the room, while the other left. When the individual part was conducted with the first
spouse, the second was called as the first left the room. After the second spouse was finished,
the former was called again, and the joint part was conducted. Participants were not told that
there would be a joint part until they both had finished the individual parts, to prevent them
from thinking about the preferences of the other spouse. Both the individual and the joint part
each consisted of three games; a trade game and a public good game in addition to the risk
and ambiguity game explained here. The order of the games was randomized across and
constant within households. Each game was paid separately, hence each individual was paid
for 6 games. When both spouses had finished the individual parts, the first spouse was called
5
This was not possible in 65 out of 287 households due to time constraints of the participating couples, and two
instructors had to perform the individual games at the same time in different rooms. In these cases, both
instructors were present during the joint part in order to reduce the possible instructor bias. We test whether
women’s bargaining power is affected by whether the lead instructor in the joint part had been instructing the
husband or the wife in the individual part, and find no such effects.
5 again, and they were presented with the three games once again and asked to make joint
decisions in all of them. Couples were specifically not instructed to agree or discuss, as
normal joint decision making doesn’t necessarily contain such elements. Instructors were
trained not to interfere in their decision making process and pay similar attention to both
spouses. A final answer from either spouse was taken as the joint decision, unless the other
spouse expressed disagreement or answered differently. In such cases the instructor left the
couple alone until they agreed upon an answer, in order to obtain a more natural setting for
the joint decision making.
The risk and uncertainty game consisted of two Ellsberg choice list tasks. The experiment was
set up such that incentives for the individual should be exactly the same in the individual and
the joint part, such that we could use the choices made in the individual part as the
individual’s preferences in the joint part. In the individual part it was emphasized that the
decision would not be revealed to the spouse afterwards, in order to avoid participants
thinking about what their spouse would have wanted them to do. In the risk game the
respondent was asked to make 20 choices using a choice list (the list is included in the
appendix, table A1). In the first choice situation, the participant chose between 0.5 soles for
sure and a lottery with a 50 % chance of winning 10 soles6. The amount of soles in the certain
option was increased by increments of 0.5 soles in each of the 19 remaining choice situations,
until it reached 10 soles in the last choice situation. Thus, we expect that the participant
choose the lottery in the beginning and then at some later stage switch to the certain amount.
The earlier a participant switches to the certain amount, the more risk averse he or she is. The
uncertainty experiment consisted of the same 20 choice situations, only with the difference
that the probabilities of the lottery were unknown. We define risk aversion as the number of
safe choices in the risk experiment and uncertainty aversion as the number of safe choices in
the uncertainty experiment, both ranging from 0 to 20. A risk aversion of 0 implies that the
participant chose the lottery for all options, even when the safe amount equals the lottery
price, exhibiting extreme risk loving behavior.
The lotteries were described and conducted in the following way. When explaining the risk
experiment, the participant was shown a brown bag consisting of 5 red and 5 blue chips. The
participant would determine the outcome of the risk lottery by picking a winning color, and
6
10 soles equaled around $3.6 US at the time. A standard daily wage was around 20 soles.
6 then draw a chip. He or she would win the price of 10 soles if the drawn chip was of the
winning colour, and nothing if it was not. When explaining the uncertainty lottery the
participant was shown a black bag and told it consisted of 10 chips, with an unknown mix of
red and blue chips. The outcome of the ambiguous lottery was determined by the participant
in the same manner as the risky lottery. The participant chose a winning color, and then drew
from the black bag. He or she won the price of 10 soles if the drawn chip was of the specified
color. Note that the participant could decide the winning color himself/herself, implying that
they could take subjective probabilities into account. If they believe there are 7 red and 3 blue
chips, picking red as the winning color would be a better deal than the 50-50 lottery in the risk
experiment. Assuming that participants form subjective expectations about the number of red
and blue chips in the black bag, the ambiguous lottery is then always at least as good as the
risky lottery in expectation.
Instructions were held in the language preferred by the participant, either the local Quechua
language or Spanish, due to language differences in the sample. The instructions were
explained thoroughly to each participant in the individual part, to maximize the understanding
of the game, and the instructor wrote down the stated answers of the participant for each
choice situation when she was certain that the participant understood the rules of the game. In
order to reduce the problem of participants switching inconsistently between the safe amount
and the lottery, we used the method of Dohmen et al (2010). At the first choice situation the
participant chose the safe amount, the instructor asked “Does that mean that you want the
safe option in all the choices with a higher safe amount as well?”. If the participant answered
affirmatively, she filled out all further options accordingly. If not, she continued asking in
each choice situation.
Only one of choice situations was paid for real from the individual part and also one from the
joint part. The participant determined the paid situation by drawing from a pile of 40 cards,
each representing one choice situation. The drawing of outcomes was done after both the
individual and joint part was finished, in order to prevent learning and wealth effects. When
the couple had finished the joint part, one spouse left the room while the other drew a card to
determine the paid situation for his or her individual part. If a situation was drawn in which
the lottery had been chosen, the participant then decided for a color and drew a chip from the
right bag to determine the lottery outcome. The drawing of individual outcomes was done
without the presence of the other, in order to keep the individual decisions private. After both
7 spouses had drawn their outcome in the individual game they joined again, and one of them
drew a card to determine their joint outcome. If they then for example drew a card
representing choice situation 10 in the risk game and they had chosen the safe amount of 5
soles in that situation, they both got 5 soles for the joint part. If they had chosen the lottery,
one of them decided for a winning color and drew a chip from the brown bag. If the chip was
of the winning color, both of them won 10 soles. If not, none of them won anything for the
joint part. Their individual earnings from the joint part were added to what they earned from
the individual part. Payments were made shortly after all couples in the community had
conducted the experiments, distributed in private to each individual in a sealed envelope.
Instructions for the introduction and the risk and uncertainty games can be found in the
appendix.
The experiments were conducted in 15 communities, selected from two distinct regions in the
southern highlands of Peru; Cusco and Apurimac. In each community, the data collection
started with a survey part, followed by the experimental part several days later and performed
by a different research team. The survey team entered the community in accordance with the
community leaders and randomly drew between 10 and 30 couples for participation, with the
conditions that they were married or cohabiting and that at least one of the spouses possessed
land. Couples that were not available or rejected participation were replaced with the
neighboring household. The survey was performed in the couple’s home. It started with a
joint part answered by both spouses together, regarding general household and land issues,
and continued with two equal parts answered by each spouse privately, amongst other about
household decision making. These questions were asked without the presence of the spouse.
After all sampled couples were surveyed in one community the experimental team entered
and offered the same couples the possibility of participating in economic experiments as well.
32 out of the 320 surveyed couples either rejected participation in the experiments or were not
accessible at this later stage, and were not replaced.
3. Results 287 couples participated in the experiments and survey in October, November and December
2010.
8 3.1. Sample description
The participating couples all reside in peasant communities in the rural highlands, at altitudes
ranging from 2300 to 4000 m.a.s. Depending on the altitude they grow mainly potatoes,
vegetables and maize. The harvest is mostly used for household consumption, while any
surplus is traded at local markets. Most participants mainly have plots in their residing
community, in which most daily interaction is based. Communities are organized by a
community assembly, in which all land-possessing households are represented. Since the man
is seen as the household head, he traditionally represents the household in the assembly and
public work. However, women are to an increasing extent gaining access to these arenas. The
division of labour within the household is additionally very gender specific. Agriculture is
seen as a male responsibility, although women typically assist during labour intensive periods
of the year. Meanwhile, women are responsible for children and housework, and often
additionally for animal breeding if they hold livestock.
Our sample consists of 269 couples, after omitting 18 couples on rationality grounds due to
inconsistent switching from the safe to the risky option by either the man, the woman or the
couple together (11 men, 6 women and 4 couples switched in this manner in one or both
games).
The socio-economic survey data reveal a sample that is generally poor and low educated, with
significant gender differences. Men have more education than women, and only 14 % of men
are illiterate, compared to 40 % of women. 54 % of households classify themselves as poor or
very poor. Couples were jointly asked to estimate the labour contribution of each spouse to
household income. The mean answer is a 60 % contribution level of the man. This gender
difference in perceived contribution is related to the gender separated working spheres; 92 %
of men report that agricultural work is their main activity, which is also the most important
income producing activity in the sample, while only 32 % of women report the same.
Women’s main activities are otherwise animal breeding (54 % of households) and housework
(7 % of households). A vast majority reports that the woman keeps control over household
money.
Women are underrepresented in all community assemblies, and it is rarely normal that they
take the word in these traditionally male arenas. Household violence is frequent, with 13 % of
women reporting to be physically mistreated by their husbands (1 % of men report to be
9 beaten by their wives). Sample descriptive variables by gender are summarized in the
appendix.
3.2. Experimental results
Figure 1 shows the number of safe choices made by the subjects both in the individual as well
as in the joint experiments. Notably, large fractions are extremely risk averse, choosing the
safe option in all 20 choices (42 % of men, 44 % of women and 58 % of couples). The
proportion of participants with extremely averse attitudes are even greater in the uncertainty
game, in which the shares choosing the safe option in all 20 choices are 51 % of men, 60 % of
women and 64 % of couples. On the other end, there are participants choosing the risky
option in all 20 choices, which is more common among men (5 % in both games) and couples
(4 % in both games) than among women (1 % in both games).
The frequent behaviour of extreme risk and uncertainty aversion we observe is difficult to
explain with a standard expected utility theory. Neither does it fit well with prospect theory,
as it would imply losses to weight ca 40 times more than gains (Kahneman and Tversky
calibrates losses to weight 2.25 times as much as gains according to Benartzi and Thaler
(1995)). Considering that our participants are highly exposed to severe climatic uncertainty
and that coping with this uncertainty historically has been a crucial survival condition in the
Andes, the risk-as-feelings perspective from psychological literature7 offers a more plausible
explanation of our findings, also in a cultural aspect.8 Note that the emotional explanation
does not imply lower understanding of the game rules. In the trade-off between participant
understanding and controlled instructions our experimental design emphasized the former
aspect with individual instructions and illustrations, hence we are confident in that the
extremely averse behaviour we observe is not due to misunderstandings.9
7
This literature emphasizes how risky prospects may induce emotional anxiety and fear of the worst possible
outcomes, resulting in extreme resistance often unrelated to the actual probabilities and outcome values. People
with previous experiences of devastating risky outcomes are more likely to respond to risk with such emotional
opposition (Loewenstein et al. 2001).
8
The emotional resistance to risk and uncertainty as a rule-of-thumb is supported by participant’s comments
preceding the experiments. When asked why they chose the safe amount in all options, a recurring argument
was “I dislike risk and I always avoid it when I can”. Some used the example of last year’s harvest, which was
severely destructed by heavy rains in all sampled areas, to argue how important it is to secure outcomes when
one can. This recent experience of a bad natural outcome may have intensified the aversion.
9
Further, participants with extreme risk or uncertainty aversion are not less educated than others, and they do not
perform worse on a simple math test.
10 In the further analysis we refer to risk aversion and uncertainty aversion as the number of safe
choices made in the risk and uncertainty game respectively. These are discrete variables
ranging from 0 to 20. We find that women are the more risk averse gender, with a mean of
15.6 safe choices in the risk game compared to 14.6 among men, however this difference is
not significant with a Mann-Whitney test (p-value 0.201). The gender difference is larger and
significant in the uncertainty game, with a mean of 16.6 safe choices among women and 14.9
among men (p-value 0.004 with Mann-Whitney test).
The average behaviour of couples differs from that of individuals. Increased frequencies of
extreme risk and uncertainty aversion from the individual to the joint part imply that couples
are significantly more risk averse (p-value 0.0001) and uncertainty averse (p-value 0.014).
Additionally, both individuals and couples are on average more averse in the uncertainty
game than in the risk game.
60
40
Percent
20
20
20
Percent
Percent
40
40
60
60
Figure 1. Number of safe choices in risk and uncertainty experiment, by women, men and
couples.
0
0
5
10
15
Female safe choices, risk experiment
20
20
Men, risk experiment
0
5
10
15
Couple safe choices, risk experiment
20
Couples, risk experiment
40
Percent
20
20
20
Percent
Percent
40
40
60
60
60
Women, risk experiment
5
10
15
Male safe choices, risk experiment
0
0
5
10
15
Female safe choices, ambiguity experiment
Women, uncertainty
experiment
20
5
10
15
Male safe choices, ambiguity experiment
Men, uncertainty
experiment
20
0
5
10
15
Couple safe choices, ambiguity experiment
20
Couples, uncertainty
experiment
3.3. Bargaining power
The next question concerns the distribution of bargaining power in the household. In the risk
and uncertainty game, 76 and 100 couples reveal exactly the same individual preferences. A
main reason is the high share of extremely risk and uncertainty averse individuals. 63 and
11 93couples, in the two games respectively, are both extremely averse in the individual part, i.e.
they chose the safe option in all 20 choices. For the households with different individual
preferences, 40 % of the joint decisions in the risk experiment are closer to the man’s
preference, while 32 % are closer to the woman’s preference. This implies an overall, but
statistically insignificant, gender difference in bargaining power. More striking is the
bargaining power of the most risk averse spouse. The joint decision lies closer to the more
risk averse spouse in 50% of households and closer to the least risk averse in 22 % of
households, implying significantly higher bargaining power of the most risk averse.
Tables 1 and 2 tabulate the direction of the joint decisions relative to individual decisions
according to two dimensions; i) the individual’s gender, and ii) the individual’s preferences.
From table 2 we see that the overall picture in the uncertainty experiment is similar to the
findings in the risk experiment, namely that a majority of couples make joint decisions closer
to the most uncertainty averse spouse, while gender seems less important in determining
bargaining power.
Table 1. Bargaining power in the risk experiment.
Joint decision closer to
Woman
Man
the…
Least risk averse
20 (7%)
38 (14%)
Most risk averse
66 (25%)
68 (25%)
Equal individual decision .
.
Total
86 (32%)
106 (40%)
Total
58 (22%)
134 (50%)
76 (25%)
268
Note: Only one couple made joint decision exactly on the middle point between male and female individual
decision (equal bargaining power), and is hence excluded here.
Table 2. Bargaining power in uncertainty experiment.
Joint decision closer to
Woman
Man
the…
Least uncertainty averse
15 (6%)
44 (16%)
Most uncertainty averse
60 (22%)
49 (18%)
Equal individual decision .
.
Total
75 (28%)
93 (35%)
Total
59 (22%)
109 (41%)
100 (37%)
268
Note: Only one couple made joint decision exactly on the middle point between male and female individual
decision (the same couple which show equal bargaining power also in the risk game), and is hence excluded
here.
The relative decisions within each household are shown in detail in figure 2 for the risk and
uncertainty games respectively. Each household is plotted according to the absolute difference
12 between couple and female number of safe choices (x-axis), and the absolute difference
between male and female number of safe choices (y-axis). In all observations above the
horizontal line, the man is more risk/uncertainty averse than the woman. This holds for 33 %
of households under risk and 24 % of households under uncertainty, while the woman is more
averse in 39 % of households, both under risk and uncertainty (not necessarily the same
households). The x-axis identifies couples’ risk aversion relative to the woman’s risk
aversion. Households plotted on the vertical line imply joint decision equal to woman’s
individual decision, hence full decision power to the woman, applying to 41 households in the
risk game and 39 households in the uncertainty game. Households plotted on the rising
diagonal imply joint decision equal to the man’s individual decision, hence full decision
power to the man. This holds for 58 households in the risk game and 56 households (33 %) in
the uncertainty game.
Man's no of safe choices relative to woman, risk
-20
-10
0
10
20
Man's no of safe choices relative to woman, uncertainty
-20
-10
0
10
20
Figure 2: Scatter plots of joint decisions versus man and woman decisions, in each
experiment (N=269).
-20
-10
0
10
Couple's no of safe choices relative to woman, risk
20
-20
-10
0
10
Couple's no of safe choices relative to woman, uncertainty
20
Note: The observational unit is the household. In observations above the horizontal line, the man is more
risk/uncertainty averse than the woman, and opposite under the horizontal line. Observations on the vertical line
imply joint decision equal to woman’s decisions, hence full decision power to the woman (except the origo
observations, in which man and woman choose equally and decision power cannot be interpreted). Observations
on the increasing diagonal imply joint decision equal to man’s decision and full decision power to the man
(except origo observations). The clustering at full female decision power under the horizontal line and full male
decision power over the horizontal line implies full decision power to the most risk/uncertainty averse.
The figure indicates a clustering of households at either full decision power to the man or the
woman. This clustering reveals two important structural facts about the data. Firstly, the joint
decisions tend to land exactly on either spouse’s individual preference, and to a lower extent
as a middle point between the two. Note that this is partly an implication of the high shares of
13 extremely risk/uncertainty averse among individuals and couples. Couples choices tend to fall
to extreme aversion, and the spouse who initially revealed extreme aversion gets full decision
power. In the risk game, 81 % of the 115 couples on the rising diagonal and 90 % of the 98
couples on the vertical line revealed extreme joint risk aversion. For the uncertainty game,
these numbers are 82 % of 142 on the diagonal and 94 % of 125 on the vertical line.
Secondly, the clustering, and hence bargaining power, is not symmetric around individual
preferences. Especially, the clustering around full decision power to the woman is mostly
found when she is more risk/uncertainty averse than her husband, while men may obtain full
bargaining power independently of preferences. One interpretation is hence that women only
get to decide when they opt for higher risk/uncertainty aversion, while men to a larger extent
can push the joint decision in whichever direction they want.
When excluding all couples in which one of the spouses are extremely averse, i.e. choosing
the safe amount in all of the 20 options, these patterns are less obvious. Scatter plots based on
this limited sample are included in the appendix, figure A1. In the risk game it still indicates a
clustering around male preferences or the more risk averse direction, seemingly independent
of the woman’s preferences.
3.4. Household variation in women’s bargaining power
In order to analyse household variation in bargaining power, we generate a household specific
measure of female bargaining power for the risk and uncertainty experiment respectively,
with the following definition
Female bargaining power 
Couple Safe Choices  Man Safe Choices
.
Woman Safe Choices  Man Safe Choices
For households in which the joint decision lies outside the interval between the two individual
decisions, female bargaining power is truncated to 1 if the joint decision is closer to the
woman’s initial decision or 0 if it is closer to the man’s initial decision. This applies for 35 %
and 31 %, in the risk and uncertainty experiment respectively, of the couples who revealed
different individual preferences 10. Female bargaining power is hence a continuous variable in
the interval [0,1]. The distributions in the two experiments are shown in figure 3, confirming
10
In the risk game, 19 % of the couples with different individual preferences ended up with a joint decision
which was more risk averse and 16 % made a joint decision which was less risk averse than both individuals. In
the ambiguity game these numbers were respectively 15% and 11%. The lower shares in the ambiguity game is
due to a higher share of individuals making 20 safe choices, leaving less room to exceed the scope of couples’
individual decisions.
14 that a vast majority of households reveal full bargaining power to either one of the spouses
and that this goes in the favour of men somewhat more often than of women.
To test the significance of this gender difference, we run logit regressions on individual level
bargaining power as a dummy identifying the individuals for which the joint decision lies
closer to their, rather than their spouse’s, individual decision. We control for individual
preferences by including a dummy for being the more averse spouse in the individual game,
which has a large and significant effect on bargaining power. Further, when controlling for
individual preferences we also observe that women have significantly lower bargaining power
than men. These regressions are presented in table A4 in the appendix.
0
0
20
20
Frequency
40
60
Frequency
40
60
80
80
100
100
Figure 3: Female bargaining power in risk (left, N=193) and uncertainty (right, N=169) game
0
.2
.4
.6
Female barg. power, risk game
.8
1
0
.2
.4
.6
Female barg. power, uncertainty game
.8
1
Note: The grey bars identify couples in which the joint preference lies outside the two individual preferences,
either on the woman’s side (female bargaining power=1) or on the man’s side (female bargaining power=0).
Next we analyse the characteristics of households with higher female bargaining power by
regressing the female bargaining power dummies on relevant survey variables. In the further
analysis we refer to female bargaining power as a dummy for the joint decision being closer
to the woman’s than the man’s individual decision. It is defined for the 192 couples in the risk
game and 168 couples in the uncertainty game with different individual preferences and joint
decision not lying exactly on the middle point between the two spouses. The results from logit
regressions are shown in table 3, columns 1 and 2. In order to control for the general drift
towards more risk/uncertainty aversion in the joint parts, we include a dummy for the woman
being more averse than the man. As already shown, it has a strong effect on female bargaining
power in both experiments. Bargaining power could in principle depend on the level of initial
disagreement between the spouses, hence we also control for the absolute difference between
the man and the woman’s number of safe choices. This initial disagreement has a positive
effect on woman’s bargaining power in the risk game, but no significant effect in the
uncertainty game. Readers who are worried about additional implications of individual risk
15 and uncertainty preferences on the bargaining power measure can study how household
characteristics are related to individual preferences in regressions presented in the appendix,
table A5.11
All the non-experimental variables are from the pre-experimental survey. Descriptions and
means of these variables are found in the appendix. We find no effects of perceived poverty,
women’s income contribution, age or relative education. However, couples mean education
level has negative effects in both games. Educational levels are unrelated to individual
preferences and are only to a small extent correlated with bargaining power of the least risk
averse, hence the relationship between education levels and lower female bargaining power
seems relatively robust.
Related to this is the positive effect of residential altitude on women’s bargaining power.
Altitude is an important determinant of economic welfare in the highlands, as higher altitudes
imply difficult agricultural conditions and often geographical isolation. One might thus expect
that women at higher altitudes would be less empowered, following a conventional belief that
economic development increases empowerment. That we actually find the opposite
relationship, seen together with the negative relationship between education and woman’s
bargaining power, indicates that such a supposition may be inappropriate.
Female empowerment is often assumed to reduce fertility due to family planning. Indeed we
see that women’s bargaining power is negatively related to the household size, which might
have indicated that stronger women bear less children. However, when breaking the
household size variable down to number of children and number of adults in the household,
we see that the negative relationship mainly comes from adults and not children. In addition
to the principal couple there may be adult children or other relatives residing in the household,
which may take decision power from the woman. There is a common view that women
residing with other and older women, and particularly family-in-law, have less decision power
in the household, and this may be the effect we observe here.
Women’s level of organization12 has a positive effect on her bargaining power in the risk
game, confirming our initial notion that organized women are more empowered. The causality
might go both directions, both through stronger women being recruited into organisations and
organizational membership itself empowering women. Patterns of everyday life might also
11
For the regressions presented in table 3 we have also controlled for the direct effect the spouses’ level of
risk/uncertainty aversion (by controlling for the mean number of safe choices of the two individuals), which does
not alter the results noteworthy.
12
Organization level is a categorical variable counting the number of organizations she is a member of among
the male dominated community defence committee and irrigation committee.
16 influence women’s bargaining power. When the man travels and resides out of the house, the
woman might gain more power in the household as she gets used to ruling alone. Indeed we
find that households in which the man sometimes resides out of home13 have higher
likelihood for female bargaining power. Further, working outside of the home sphere is often
thought to empower women, hence we might expect that a woman’s main activity would be
related to her bargaining power. However, we find no such effects.
Social signalling is a potentially strong driver of female empowerment. The extent to which
women participate in the community assembly is both an indicator of the general
empowerment level, and a potential driver of empowerment both of the women present
themselves and other men and women in the same community through updating social norms
regarding women’s rights and abilities. Indeed we find that female participation at community
level is related to higher female bargaining power also at the household level.
Taking a step back, we turn to the households in which the spouses reveal the same individual
preferences and have thus not been included in the analysis until now. Remember that this
holds for 25 % of households in the risk game and 37 % of households in the uncertainty
game, of which a majority reveals the highest level of risk/uncertainty aversion possible.
Looking at the different characteristics of these households compared to the sample used in
the bargaining power analysis, sheds some light on the total sample. It might be argued that
women who reveal exactly the same preference as their husbands are less empowered than
others, if it is a result of not making up a personal opinion but rather choosing according to
the husbands assumed preferences. Thus we run logit regressions identifying the couples with
equal preferences, compared to the rest of the sample. The results are presented in table 3,
columns 3 and 4. We control for the spouses’ mean number of safe choices, since a large part
of this equal-preference-behavior seems to be driven by extreme risk/uncertainty aversion.
The results does give any further support to the hypothesis that weak women mimick their
husband’s preferences. First, few variables have any significant effects. Since we do not find
any specific characteristics of the couples with equal preferences, it indicates that there is
nothing special with these couples, and that the equality of preferences is purely by chance.
13
This implies staying the night other places, normally due to work commuting.
17 Table 3. Logit regressions: Female bargaining power and decision equality
Perceived poverty household
Share income contribution woman
Mean age spouses
Age difference, years man older than woman
Mean education level, couple
Education difference, man's levels above woman
Household members <16 yrs
Household members >=16 yrs
Agriculture main activity, woman
Housework main activity, woman
Other main activity, woman
Paid work main activity, man
Man sometimes resides out
Organisation level, woman
Woman verbally mistreated by man
Woman physically mistreated by man
Man verbally mistreated by woman
Altitude km
Women in community assembly
Market distance
Region Apurimac, dummy
Woman most risk averse
Man's no of safe choices relative to woman, risk
Woman most uncertainty averse
Man's no of safe choices relative to woman,
uncertainty
Couple's mean no of safe choices, risk
Female barg.
power, risk
0.199
(0.64)
-0.126
(-0.10)
-0.0106
(-0.52)
0.0389
(1.00)
-0.931**
(-2.33)
0.169
(0.88)
-0.123
(-0.68)
-0.381**
(-2.03)
0.344
(0.69)
0.0766
(0.08)
0.495
(0.57)
0.0237
(0.03)
0.863
(1.23)
1.070***
(3.37)
-0.0333
(-0.07)
0.211
(0.35)
-1.350*
(-1.65)
-0.573
(-0.87)
0.0430
(0.21)
-0.00430
(-0.55)
0.374
(0.54)
3.532***
(4.79)
0.0809**
(2.31)
Female barg.
power, uncertainty
-0.298
(-0.80)
-0.812
(-0.59)
0.00183
(0.09)
-0.0165
(-0.42)
-0.888**
(-2.30)
0.0770
(0.38)
-0.0173
(-0.08)
0.230
(1.24)
0.327
(0.69)
-1.186
(-0.86)
0.0950
(0.10)
-0.288
(-0.44)
2.127***
(2.92)
0.441
(1.43)
-1.396**
(-2.43)
0.906
(1.37)
1.558*
(1.92)
1.794**
(2.20)
0.369*
(1.76)
0.0127
(1.45)
1.854**
(2.28)
Spouses same
decision, risk
-0.436
(-1.35)
-1.048
(-0.86)
-0.0240
(-1.18)
0.0167
(0.44)
-0.132
(-0.35)
-0.0764
(-0.48)
0.0943
(0.61)
0.0614
(0.36)
0.386
(0.77)
0.666
(1.01)
0.411
(0.46)
-1.045
(-1.21)
-1.590
(-1.51)
-0.520
(-1.58)
0.286
(0.51)
0.0346
(0.04)
-0.381
(-0.48)
-0.508
(-0.79)
0.180
(0.91)
-0.00369
(-0.52)
0.642
(1.02)
Spouses same
decision, uncertainty
0.169
(0.57)
-0.343
(-0.29)
-0.00986
(-0.50)
0.0373
(0.99)
-0.193
(-0.50)
0.232
(1.35)
-0.0582
(-0.36)
-0.315*
(-1.68)
-0.847*
(-1.75)
0.726
(1.01)
0.364
(0.41)
-0.0357
(-0.04)
-0.395
(-0.50)
-0.351
(-1.15)
0.0553
(0.09)
-1.047
(-1.37)
0.159
(0.20)
0.370
(0.61)
0.107
(0.56)
-0.00441
(-0.65)
0.745
(1.24)
2.258***
(2.90)
0.0179
(0.48)
0.443***
(6.17)
Couple's mean no of safe choices, uncertainty
0.408***
(6.39)
Constant
1.021
-8.414**
-4.957*
-7.750***
(0.35)
(-2.45)
(-1.65)
(-2.66)
Observations
192
168
269
269
Pseudo R2
0.273
0.259
0.354
0.391
t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01. ‘’ Reference category is animal breeding as main
activity, which is the largest category for women in the sample.
18 3.5. Self-reported household investments and decision power
This section summarizes survey data on household investments and self-reported decision
power by the two spouses separately. We test whether our experimental measures of risk and
uncertainty aversion correlates with survey reported investment behavior of the household, as
well as whether our experimental measures of bargaining power correspond to survey
reported decision structures.
Reported agricultural expenditures and credit uptake by the household the last 12 months are
listed in table A6 in the appendix, together with correlation coefficients with revealed risk and
uncertainty preferences of men, women and couples. The returns on these investments are all
highly uncertain. Agricultural outcome depends to a large extent on uncontrollable climatic
conditions, and all investments may literally be washed away in case of excessive rains or dry
up in the opposite case. Likewise, livestock investments are uncertain due to risk of diseases.
Hence, we expect to find lower levels of risk and uncertainty aversion among households with
larger such expenditures. Indeed this is also what we observe; all particular expenditures are
negatively correlated with number of safe choices made in the experiments, both among men,
women and couples. Similarly, households which have taken loans the last 12 months choose
significantly less number of safe choices in the joint parts of both games. These correlations
support the relevance of lab experiments for field behaviour in general and particularly
increase our confidence in that the preferences induced function as measures for more general
preferences of our participants, both at the individual and joint level.
We now turn to the experimental measure of woman’s bargaining power. Since bargaining
power in the experiment is highly affected by individual preferences for risk/uncertainty, we
redefine this measure somewhat. We know that being the most risk/uncertainty averse gives
higher bargaining power per se. Thus we now define the women who are less risk/uncertainty
averse than their husbands, but in spite of this manages to pull the joint decision in their
direction, as the powerful (female bargaining power of 1). Likewise, when the man is the least
risk/uncertainty averse spouse and manages to pull the joint decision in his direction, he is
categorized as the more powerful (female bargaining power of -1). The households in which
the joint decision lies closer to the most risk/uncertainty averse are ranged in the middle
(female bargaining power of 0). Note that this way of determining female bargaining power is
similar to our binary variable in the last section when a control for the woman being the most
averse spouse is included, as we did in the regressions in table 3. This new preference19 adjusted bargaining power measure hence picks up the same variation and regressing it on
household characteristics, without controlling for individual level preferences, gives similar
results. The two measures are highly correlated. Of 143 couples with non-equal preferences in
both games, 106 have the same value of preference-adjusted female bargaining power. In
order to get more observations into one variable, we construct an index variable of the two. It
takes the value 1 if female bargaining power is 1 in at least one of the two games, -1 if female
bargaining power is -1 in at least one game, and otherwise 014. All three experimental female
bargaining power variables are tabulated in table 4 and correlated with indexed surveyreported decision structures in the household, which are described in the following. The
reported decision structures for different household investments, reported by the woman and
the man individually, are tabulated in tables A7 and A8 in the appendix. Further, both the man
and the woman have reported who they believe should make different household decisions,
which are tabulated in table A9 in the appendix. In general, both men and women report a
high extent of joint decision making, both as the actual and preferable case. However, to the
extent that either the man or the woman is reported to decide alone, the pattern follows the
expected gender spheres in activities and responsibilities. Men are to a larger degree than
women reported to decide in agricultural issues, while women are reported to decide about
smaller household consumption items. The more important decision, the more likely it is that
they report to make the decision jointly. We categorize these answers according to female
decision power, giving the value 1 if the man decides alone, 2 if they decide jointly, and 3 if
the woman decides alone. Then we generate decision-power-indexes in 5 different decisions
areas, by computing the mean female decision power reported by the woman15 in each of the
following categories; 1) agricultural and house investments (all investments listed in table
A7); 2) school and economic expenditures of the household (all listed in table A8 except beer
and alcohol); 3) beer and alcohol; 4) perceptions about minor economic spending; and 5)
perceptions about agricultural investments. These 5 indexes are listed in table 4, together with
correlations with each of the three experimental measures of preference adjusted female
bargaining power. We find positive and significant correlations between experimentally
revealed female bargaining power and reported female decision power regarding school and
14
4 households reveal female bargaining power equal to1 in one game and -1 in the other game, and these are
given value 0 for the indexed experimental bargaining power.
15
We base the indexes on decision structure reported by the women, as this seems to be the convention in the
literature.
20 economic spending, as well as with reported perceptions about who should decide regarding
minor economic spending. Perceptions about who should decide in agricultural investments
are only slightly significantly correlated with the indexed experimental bargaining power,
while actual reported decision structure for agricultural investments and alcohol is not
correlated with either of the experimental measures. These patterns suggest that the
experimental context is interpreted as a decision regarding minor monetary spending of the
household, in which women are normally responsible and reported to decide, rather than as a
major agricultural investment. As one man commented after the experiment when asked about
who decided and why: “In this area it is the women who keep control over the household
economy, therefore she decided.” This interpretation may explain why women have relatively
high bargaining power in the experiment. However, it does not necessarily imply that women
enjoy high bargaining power in other aspects of life as well, since the experimental measures
do not correlate with reported decision power in the important area of agricultural
investments. This lack of correlation might be partly due to little variance in the reported
measure of agricultural decision power, as a majority of household report to make these big
decisions jointly. However, given that reported decision power is found to be very task
specific, it also suggests that what we have measured in the experiment translates only to
some specific areas of household decision making, and not all.
We have additionally tested whether women revealing the same individual preferences as
their husbands differ to the rest of the sample in regards of reported bargaining power, to test
the hypothesis of disempowered women mimicking their husbands. Since a vast majority of
couples revealing equal preferences are extremely risk/uncertainty averse, we need to control
for the level of aversion in this analysis. Hence we run logit regressions on dummies
indicating equal preferences with each indexed reported measure of bargaining power as
explanatory variables in separate regressions, controlling for mean number of safe choices as
in the regressions of table 3. None of the indexed reported bargaining power measures give
significant coefficients in these regressions. Thus we are confined that mimicking women do
not pose a severe problem to this analysis.
During the joint household survey, interviewers noted the percentage talked by each spouse
when asked about different household issues, with the perception that answering on behalf of
the couple in an interview situation was a sign of empowerment. The percentage talked is
positively and significantly correlated with female bargaining power in both games as well as
21 the game index. Note that these interviews were performed several days before the
experiments, hence the correlations should be explained by some permanent household
characteristics.
Table 4. Preference adjusted female bargaining power and correlations with indexed reported
decision structure (reported by the woman).
Pref. adjusted Pref. adjusted
Pref.
No of
adjusted
female
female
N decisions
female
bargaining
bargaining
in index bargaining
power,
power, games
power, risk
uncertainty
index
N
193
169
219
Joint decision closer to...
Least averse woman (1)
20
15
21
Most averse spouse (0)
135
110
141
Least averse man (-1)
38
44
57
Reported female decision power
Agricultural and house investments
249
6
-0.0555
-0.0641
-0.0096
(0.4449)
(0.4089)
(0.8874)
K
School materials/fees and household
269
5
0.1477*
0.0932
0.1243*K
spendings
(0.0404)
(0.2283)
(0.0663)
Alcohol
201
2
-0.0872
(0.2938)
-0.0527
(0.5452)
-0.0113
(0.8848)
The extent to which women should
decide
Minor spendings
269
2
Agricultural decisions
269
1
0.0912
(0.2074)
0.0674
(0.3520)
0.1717* S K
(0.0256)
0.1142
(0.1394)
0.1237* S
(0.0676)
0.1121*
(0.0981)
Note: Experimentally revealed bargaining power classified as a categorical variable, taking value 1 if joint
decision is closer to female and she is the least risk averse spouse, -1 if joint decision is closer to male and he is
the least averse spouse, and 0 if joint decision is closer to the most averse spouse (missing if spouses reveal same
individual preferences). Reported decision indexes are averages of the decision structures reported by women
listed in tables A4-A6 in the appendix, which are categorical variables (1-man decides, 2-couple decides
together, 3-woman decides). Some households are missing in the agricultural/house and alcohol indexes since
these investments were not reported.
Correlations coefficients (and p-values in parenthesis) are based on t-tests, with * indicating significance at a
90% level. Note also that both Spearman’s and Kendall’s rank correlation tests are performed in each case.
Coefficients significant at a 90% level are indicated with S or K for these tests respectively.
22 4. Conclusions In this paper we have investigated intra-household bargaining power within households using
an experimental approach. In addition, we compare our results with other types of investment
behaviour and self-assessed decision power in the household. Our results show that
experimental bargaining power is highly related to individual preferences, and less so to
gender. The joint decision lies closer to the more averse spouse in the majority of
observations. However, when controlling for the most averse spouse, we also find that joint
decisions are significantly closer to the preferences of the man than the woman. This is in line
with a general view of men as main decision makers in Peruvian highland communities.
In addition to a risk experiment similar to what others have used to investigate intrahousehold decision making, we also conducted an uncertainty experiment motivated by the
literature on ambiguity aversion. The results show that intra-household bargaining power in
risky and uncertain decisions do not differ substantially, which supports the possibility to
generalize previous results regarding risky situations to uncertain situations.
This paper has aimed at concretizising the somewhat abstract concept of intra-household
bargaining power. We find that bargaining power is very task specific, and that bargaining
power in different areas of decision making does not necessarily correlate. The experimental
measure of bargaining power under risk and uncertainty correlates with reported bargaining
power over household expenditures, but not with reported bargaining power over agricultural
investment decisions. It suggests that it is difficult to view bargaining power as a stable
concept across choice situations, as it seems to be highly related to gender segregated spheres
of responsibilities in every-day decision making. Not all these areas are equally important in
determining female empowerment, constituted by the capability of women to make strategic
life choices. Thus, if one wants to be able to evaluate policies directed at increasing women’s
bargaining power in the household particularly, it seems necessary to be specific about the
areas of bargaining power aimed for. Whether experimentally revealed bargaining power can
contribute to this as a reliable measure is still an open question.
23 References World Bank (2012). World Development Report 2012: Gender Equality and Development, The World Bank. Abdellaoui, M., A. Baillon, et al. (2011). The Rich Domain of Uncertainty: Source Functions and Their Experimental Implementation. The American Economic Review 101(2): 695‐723. Agarwal, B. (1997). ''Bargaining'' and Gender Relations: Within and Beyond the Household. Feminist Economics 3(1): 1‐51. Akay, A., P. Martinsson, et al. (2011). Attitudes toward uncertainty among the poor: an experiment in rural Ethiopia. Theory and Decision: 1‐12. Allendorf, K. (2007). Do women's land rights promote empowerment and child health in Nepal? World Development 35(11): 1975‐1988. Anderson, S. and M. Eswaran (2009). What determines female autonomy? Evidence from Bangladesh. Journal of Development Economics 90(2): 179‐191. Ashraf, N., D. Karlan, et al. (2010). Female empowerment: Impact of a commitment savings product in the Philippines. World Development 38(3): 333‐344. Baker, R. J., S. K. Laury, et al. (2008). Comparing Small‐Group and Individual Behavior in Lottery‐
Choice Experiments. Southern Economic Journal 75(2): 367‐382. Basu, K. (2006). Gender and Say: A Model of Household Behaviour With Endogenously Determined Balance of Power. Economic Journal 116(511): 558‐580. Bateman, I. and A. Munro (2005). An Experiment on Risky Choice Amongst Households*. The Economic Journal 115(502): C176‐C189. Benartzi, S. and R. H. Thaler (1995). Myopic Loss Aversion and the Equity Premium Puzzle. The Quarterly Journal of Economics 110(1): 73‐92. Borghans, L., B. H. H. Golsteyn, et al. (2009). Gender Differences in Risk Aversion and Ambiguity Aversion. Journal of the European Economic Association 7(2‐3): 649‐658. Bourguignon, F., M. Browning, et al. (2009). Efficient Intra‐Household Allocations and Distribution Factors: Implications and Identification. Review of Economic Studies 76(2): 503‐528. Browning, M. and P. A. Chiappori (1998). Efficient Intra‐Household Allocations: A General Characterization and Empirical Tests. Econometrica 66(6): 1241‐1278. Cabantous, L. (2007). Ambiguity Aversion in the Field of Insurance: Insurers’ Attitude to Imprecise and Conflicting Probability Estimates. Theory and Decision 62(3): 219‐240. Camerer, C. (2011). The promise and success of lab‐field generalizability in experimental economics: Acritical reply to levitt and list. papers.ssrn.com. 24 Carlsson, F., P. Martinsson, et al. (2009). Household decision making and the influence of spouses' income, education, and communist party membership: a field experiment in rural China. Working Paper. Gotherburg, University of Gothenburg. Cherchye, L., B. D. Rock, et al. (2009). Opening the Black Box of Intrahousehold Decision Making: Theory and Nonparametric Empirical Tests of General Collective Consumption Models. Journal of Political Economy 117(6): 1074‐1104. Croson, R. and U. Gneezy (2009). Gender Differences in Preferences. Journal of Economic Literature 47: 448‐474. de Palma, A., N. Picard, et al. (2011). Individual and couple decision behavior under risk: evidence on the dynamics of power balance. Theory and Decision 70(1): 45‐64. Dohmen, T., A. Falk, et al. (2010). Are Risk Aversion and Impatience Related to Cognitive Ability? The American Economic Review 100: 1238‐1260. Dohmen, T., A. Falk, et al. (2011). INDIVIDUAL RISK ATTITUDES: MEASUREMENT, DETERMINANTS, AND BEHAVIORAL CONSEQUENCES. Journal of the European Economic Association 9(3): 522‐550. Donni, O. (2003). Collective household labor supply: nonparticipation and income taxation. Journal of Public Economics 87(5–6): 1179‐1198. Duflo, E. (2003). Grandmothers and granddaughters: Old‐age pensions and intrahousehold allocation in South Africa. The World Bank Economic Review 17(1): 1. Ellsberg, D. (1961). Risk, ambiguity and the Savage axioms. Quarterly Journal of Economics 75: 643–669. Friedberg, L. and A. Webb (2006). Determinants and consequences of bargaining power in households. NBER Working Paper. Hashemi, S. M., S. R. Schuler, et al. (1996). Rural credit programs and women's empowerment in Bangladesh. World Development 24(4): 635‐653. Hoddinott, J. and L. Haddad (1995). Does female income share influence household expenditures? Evidence from Côte d'Ivoire. Oxford Bulletin of Economics and Statistics 57(1): 77‐96. Kabeer, N. (1999). Resources, Agency, Achievements: Reflections on the Measurement of Women's Empowerment. Development and Change 30(3): 435‐464. Levitt, S. D. and J. A. List (2007). Viewpoint: On the generalizability of lab behaviour to the field A propos de la possibilité de généraliser les comportements de laboratoire à ce qui se passe sur le terrain. Canadian Journal of Economics/Revue canadienne d'économique 40(2): 347‐370. Levitt, S. D. and J. A. List (2007). What Do Laboratory Experiments Measuring Social Preferences Reveal About the Real World? The Journal of Economic Perspectives 21(2): 153‐174. Loewenstein, G. F., E. U. Weber, et al. (2001). Risk as Feelings. Psychological Bulletin 127. 25 Lundberg, S. and R. A. Pollak (1993). Separate spheres bargaining and the marriage market. Journal of Political Economy 101(6): 988‐1010. Lundberg, S. J., R. A. Pollak, et al. (1997). Do husbands and wives pool their resources? Evidence from the United Kingdom child benefit. Journal of Human Resources 32(3): 463‐480. Maffioletti, A. and M. Santoni (2005). Do Trade Union Leaders Violate Subjective Expected Utility? Some Insights From Experimental Data. Theory and Decision 59(3): 207‐253. Manser, M. and M. Brown (1980). Marriage and household decision‐making. A bargaining analysis. International Economic Review 21(1): 31‐44. Masclet, D., N. Colombier, et al. (2009). Group and individual risk preferences: A lottery‐choice experiment with self‐employed and salaried workers. Journal of Economic Behavior & Organization 70(3): 470‐484. Mazzocco, M. (2004). Saving, Risk Sharing, and Preferences for Risk. The American Economic Review 94(4): 1169‐1182. Munro, A., I. J. Batemen, et al. (2008). The Family Under the Microscope: An Experiment Testing Economic Models of Household Choice, Available at SSRN: http://ssrn.com/abstract=1138969. Peters, H. E., A. S. Ünür, et al. (2004). Free‐Riding and the Provision of Public Goods in the Family: A Laboratory Experiment*. International Economic Review 45(1): 283‐299. Phipps, S. A. and P. S. Burton (1998). What’s Mine is Yours? The Influence of Male and Female Incomes on Patterns of Household Expenditure. Economica 65(260): 599‐613. Savage, L. J. (1954). The Foundations of Statistics. New York, Wiley. Shupp, R. S. and A. W. Williams (2008). Risk preference differentials of small groups and individuals*. The Economic Journal 118(525): 258‐283. Thomas, D. (1990). Intra‐Household Resource Allocation: An Inferential Approach. The Journal of Human Resources 25(4): 635‐664. Thomas, D., D. Contreras, et al. (1999). Distribution of power within the household and child health. Santa Monica, CA, RAND. Trautmann, S. T., M. V. Ferdinand, et al. (2011). Preference Reversals for Ambiguity Aversion. Management Science 57(7): 1320‐1333. Trautmann, S. T. and R. Zeckhauser (2010). Blindness to the Benefits of Ambiguity: The Neglect of Learning Opportunities. Working Paper, Harvard University. Viscusi, W. K. and H. Chesson (1999). Hopes and Fears: the Conflicting Effects of Risk Ambiguity. Theory and Decision 47(2): 157‐184. Yesuf, M. and R. A. Bluffstone (2009). Poverty, Risk Aversion, and Path Dependence in Low‐Income Countries: Experimental Evidence from Ethiopia. American Journal of Agricultural Economics 91(4): 1022‐1037. 26 Appendix Table A1. Choice list
Mark the decisions in each situation. Brown bag (risk)
Safe
amount
Safe
option
Black bag (uncertainty)
Draw
from the
bag
Safe
amount
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
6
6.5
7
7.5
8
8.5
9
9.5
10
Safe
option
Draw
from
the bag
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
6
6.5
7
7.5
8
8.5
9
9.5
10
Table A2: Household descriptive variables
Variable name
Perceived poverty
Estimated income
Income stability
Altitude
Market distance
Women in community
assembly
Variable explanation
1-not poor, 2-more or less poor, 3-poor,
4-very poor
1-very unstable, 2-more or less
unstable, 3-stable
Kilometers above sealevel at household
residence
Minutes by car from district center to
community center
Categorized share of women in the
community assembly; 1-none, 2-a few
(~10%), 3-some(~25%), 4-half(~50%),
5-more than half (50-99%), 6-all
Mean (Standard deviation)
2.6 (0.62)
14 261 (18 303)
1.7 (0.61)
3.2 (0.38)
26 (32)
3.5 (1.17)
27 Table A3: Individual descriptive variables by gender
Variable name
Description
Women
No education
Primary school
Secondary school
Literacy
Spanish speaker
Age
Contribution to
couple’s income
Agriculture main
activity
Livestock main
activity
Trade main activity
Paid main activity
Housework main
activity
Sometimes resides
out of home
Organisation
Verbally mistreated
by spouse
Physically
mistreated by spouse
No education or
incompleted primary school
Primary school or
incompleted secondary
school
Completed secondary
school or higher
Men
Chi 2 p-values
testing gender
differences (ttest values in
italics)
0.000
0.66 (0.48)
0.48 (0.50)
0.28 (0.45)
0.35 (0.48)
0.114
0.06 (0.24)
0.17 (0.38)
0.000
0.60 (0.49)
0.83 (0.38)
44 (14)
0.40 (0.17)
0.86 (0.35)
0.93 (0.26)
48 (14)
0.60 (0.17)
0.000
0.000
0.002 (t-test)
0.000 (t-test)
0.32 (0.47)
0.92 (0.27)
0.000
0.54 (0.50)
0.003 (0.06)
0.000
0.03 (0.17)
0.01 (0.11)
0.07 (0.26)
0.003 (0.06)
0.07 (0.26)
0 (0)
0.019
0.000
0.000
0.03 (0.16)
0.09 (0.29)
0.002
Counted membership in;
defense committee and
irrigators committee
One or more incidents last
12 months
0.59 (0.67)
0.97 (0.67)
0.000
0.25 (0.43)
0.07 (0.25)
0.000
One or more incidents last
12 months
0.13 (0.34)
0.01 (0.11)
0.000
Spouses’ individual share of
work contributions to
household incomes,
estimated jointly
28 Man's no of safe choices relative to woman, risk
-20
-10
0
10
20
Man's no of safe choices relative to woman, uncertainty
-20
-10
0
10
20
Figure A1. Scatter plots of joint decisions versus male and female decisions, in each
experiment. Sample is restricted to couples in which none of the spouses revealed extreme
risk/uncertainty aversion individually (i.e. neither chose the safe amount in all 20 options).
-20
-10
0
10
Couple's no of safe choices relative to woman, risk
Risk game: N=99
20
-20
-10
0
10
Couple's no of safe choices relative to woman, uncertainty
20
Uncertainty game: N=65
Table A4. Logit regressions: Individual level bargaining power
Individual
Individual
bargaining power, bargaining power,
risk
uncertainty
Woman
-0.672**
-0.874**
(-2.06)
(-2.45)
Most risk averse spouse
1.776***
(5.44)
1.494***
(4.19)
Most uncertainty averse
spouse
Constant
Observations
Pseudo R2
-0.552***
(-2.70)
384
0.133
-0.310
(-1.56)
336
0.092
t statistics in parentheses
*
p < 0.1, ** p < 0.05, *** p < 0.01
Observations at individual level, standard errors clustered at household level.
Table A5. OLS regressions: Risk and uncertainty aversion by gender
Risk aversion
women
Uncertainty
aversion women
Risk aversion
men
Uncertainty
aversion men
29 Perceived poverty household
0.319
(0.53)
0.330
(0.56)
-1.189*
(-1.66)
0.330
(0.56)
Share income contribution woman
0.0182
(0.01)
0.942
(0.41)
6.647**
(2.42)
0.942
(0.41)
Mean age spouses
-0.0359
(-0.95)
0.0210
(0.56)
-0.00590
(-0.13)
0.0210
(0.56)
Age difference, years man older than
Woman
-0.109
(-1.54)
-0.0141
(-0.20)
0.0494
(0.58)
-0.0141
(-0.20)
Mean education level, couple
-0.323
(-0.47)
-0.525
(-0.77)
-0.0672
(-0.08)
-0.525
(-0.77)
Education difference, man's levels above
Woman
-0.0204
(-0.06)
0.131
(0.41)
-0.0591
(-0.15)
0.131
(0.41)
Household members <16 yrs
0.163
(0.51)
0.381
(1.22)
0.516
(1.36)
0.381
(1.22)
Household members >=16 yrs
-0.0283
(-0.08)
-0.563*
(-1.66)
0.0628
(0.15)
-0.563*
(-1.66)
Agriculture main activity, woman
-0.504
(-0.55)
-0.105
(-0.12)
-0.337
(-0.31)
-0.105
(-0.12)
Housework main activity, woman
1.521
(1.08)
1.690
(1.21)
3.546**
(2.10)
1.690
(1.21)
Other main activity, woman
-1.158
(-0.63)
-0.0780
(-0.04)
0.0842
(0.04)
-0.0780
(-0.04)
Paid work main activity, man
-0.859
(-0.62)
2.019
(1.48)
-2.326
(-1.41)
2.019
(1.48)
Man sometimes stays out of home
-2.409*
(-1.77)
-3.928***
(-2.93)
0.172
(0.11)
-3.928***
(-2.93)
Organisation level, woman
-0.555
(-0.99)
-0.670
(-1.21)
-1.078
(-1.61)
-0.670
(-1.21)
Woman verbally mistreated by man
0.827
(0.84)
-0.00315
(-0.00)
-1.980*
(-1.69)
-0.00315
(-0.00)
Woman physically mistreated by man
-4.848***
(-3.98)
-2.867**
(-2.39)
0.715
(0.49)
-2.867**
(-2.39)
Man verbally mistreated by woman
-0.491
(-0.34)
-0.550
(-0.39)
1.435
(0.84)
-0.550
(-0.39)
Altitude residence, household
-0.000624
(-0.52)
-0.000553
(-0.47)
0.000309
(0.22)
-0.000553
(-0.47)
Women in community assembly
0.474
(1.31)
0.0294
(0.08)
0.899**
(2.08)
0.0294
(0.08)
Market distance
0.00372
0.0213
-0.00411
0.0213
30 (0.27)
(1.56)
(-0.25)
(1.56)
Region Apurimac, dummy
0.733
(0.60)
-0.638
(-0.53)
-0.895
(-0.61)
-0.638
(-0.53)
Constant
18.17***
(3.47)
269
0.147
18.45***
(3.57)
269
0.143
11.63*
(1.86)
269
0.102
18.45***
(3.57)
269
0.143
Observations
R2
t statistics in parentheses
* p < 0.1, ** p < 0.05, *** p < 0.01
31 Table A6. How reported household agricultural expenditures and credit uptake correlate with
number of safe choices in the risk and uncertainty games
Risk game
Uncertainty game
EXPENDITURES
Mean
expenditure last
1ast 12 months
Leasing land
75
Seeds
108
Fertilizers
165
Pesticides
45
Paid labour
291
Transport
31
Irrigation
27
Total agriculture
909
Livestock food
191
Livestock bought
324
Total livestock
592
CREDIT UPTAKE
Household loan uptake
last 12 months
Mean
0.15
Men
Women
Couples
Men
Women
Couples
-0.1580*
(0.0095)
-0.0032
(0.9589)
-0.0943
(0.1229)
0.0313
(0.6098)
-0.0139
(0.8204)
-0.1178*
(0.0537)
-0.0318
(0.6032)
-0.0933
(0.1267)
-0.0890
(0.1456)
-0.0249
(0.6842)
-0.0748
(0.2215)
-0.1398*
(0.0218)
0.0545
(0.3729)
-0.1132*
(0.0636)
0.0849
(0.1648)
-0.0800
(0.1906)
-0.1174*
(0.0545)
-0.0579
(0.3442)
-0.1098*
(0.0721)
-0.1608*
(0.0082)
-0.0694
(0.2565)
-0.1435*
(0.0185)
-0.1557*
(0.0106)
-0.0391
(0.5227)
-0.0794
(0.1944)
-0.0188
(0.7591)
-0.1234*
(0.0431)
-0.1287*
(0.0349)
-0.1045*
(0.0870)
-0.1537*
(0.0116)
-0.0125
(0.8388)
-0.0605
(0.3230)
-0.0546
(0.3723)
-0.1367*
(0.0249)
-0.0143
(0.8156)
-0.0448
(0.4647)
0.0511
(0.4037)
-0.0289
(0.6374)
-0.0810
(0.1854)
-0.0167
(0.7846)
-0.0818
(0.1812)
-0.0907
(0.1378)
-0.0802
(0.1899)
-0.1263*
(0.0384)
-0.1692*
(0.0054)
-0.0130
(0.8323)
-0.1047*
(0.0867)
0.0722
(0.2379)
-0.1231*
(0.0436)
-0.1156*
(0.0584)
-0.0841
(0.1689)
-0.1548*
(0.0110)
-0.0450
(0.4625)
-0.0571
(0.3506)
-0.0736
(0.2290)
-0.1439*
(0.0182)
-0.0561
(0.3590)
-0.0885
(0.1479)
0.0377
(0.5385)
-0.1159*
(0.0575)
-0.1665*
(0.0062)
-0.1400*
(0.0216)
-0.1609*
(0.0082)
-0.0119
(0.8464)
-0.0846
(0.1667)
-0.0785
(0.1996)
0.0328
(0.5921)
-0.0555
(0.3644)
-0.1191*
(0.0511)
-0.0507
(0.4073)
-0.0473
(0.4394)
-0.1176*
(0.0540)
Note: Correlation coefficients with significance levels in parentheses. Spearman’s rank correlation tests reveal
similar results, though not reported here.
32 Table A7. Self-reported decision power. Who makes final investment decisions? Total N=269.
Reported by:
House
construction
Buying
fertilizers
Hire labour
Renting ox
plow / tractor
Buying
pesticides
Calling for
public agr. work
Women Men
Women Men
Women Men
Women Men
Women Men
Women Men
N (report
investment made)
82
79
187
187
153
156
159
169
100
107
213
232
Woman
6%
4%
12%
7%
15%
10%
9%
6%
5%
3%
9%
3%
Man
22%
20%
38%
44%
32%
40%
32%
36%
32%
36%
36%
44%
Joint
72%
76%
50%
48%
53%
51%
58%
58%
63%
61%
54%
53%
Reject
p=0.006
Reject
p=0.002
Reject
p<0.001
Reject
p<0.001
Reject
p=0.002
Reject
p<0.001
Reject
p<0.001
Reject
p<0.001
Reject
p<0.001
Reject
p<0.001
Reject
p<0.001
Reject
p<0.001
H0: Men and
women decide to
an equal extent
H0: Spouses
report same
decision structure
No rejection
p=0.568
Reject
p=0.021
Reject
p=0.025
Reject
p=0.030
No rejection
p=0.127
Reject
p=0.005
Hypothesis testing is performed with t-tests.
33 Table A8. Self-reported decision power. Total N=269.
Reported by:
N (report
investment made)
Woman
Man
Joint
H0: Men and
women decide to
an equal extent
H0: Spouses report
same structure
Who decides
how household
money is
spent?
Buying school
materials
Buying school
uniforms
School fees for
boys
School fees for
girls
Buying beer
Buying other
alcohol
Women
Men
Women
Men
Women
Men
Women
Men
Women
Men
Women
Men
Women
Men
268*
269
201
202
189
192
159
157
151
151
111
153
178
215
28%
5%
67%
26%
7%
66%
25%
3%
72%
18%
13%
69%
25%
4%
71%
18%
14%
68%
25%
4%
70%
21%
15%
64%
26%
5%
69%
20%
12%
68%
13%
44%
43%
3%
60%
37%
15%
26%
59%
5%
47%
48%
Reject
p<0.001
Reject
p=0.0136
Reject
p<0.001
Reject
p<0.001
No rejection
p=0.286
Reject
p<0.001
Investments: Who makes final investment decisions?
Reject
p<0.001
Reject
p<0.001
No
rejection
p=0.205
Reject
p<0.001
Reject
p<0.001
No
rejection
p=0.250
Reject
p<0.001
Reject
p=0.004
No
rejection
p=0.182
Reject
p<0.001
No reject
p=0.083
Reject
p=0.024
Reject
p<0.001
Reject
p=0.004
Reject
p<0.001
* Note: One woman reports that another household member decides, this category is not included in the table.
Hypothesis testing is performed with t-tests.
34 Table A9. Self reported perceptions: Who should decide about economic decisions
Reported by:
N (report invested)
Woman
Man
Both
H0: Men and women
should decide to an
equal extent
H0: Spouses report same
decision structure
Minor spending of
the couple
Women Men
269
269
13%
56%
50%
13%
36%
31%
Minor spending of
the household
Women Men
269
269
43%
33%
2%
8%
54%
59%
Agricultural
investments
Women
Men
269
269
5%
3%
9%
12%
86%
84%
Reject
p<0.001
Reject
p<0.001
No rejection
p=0.105
Reject
p<0.001
Reject
p<0.001
Reject
p<0.001
Reject
p<0.001
Reject
p<0.001
Reject
p=0.039
Hypothesis testing is performed with t-tests.
35