Feminist Economics
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/rfec20
Joint Forces: The Impact of Intrahousehold
Cooperation on Welfare in East African Agricultural
Households
Els Lecoutere & Bjorn Van Campenhout
To cite this article: Els Lecoutere & Bjorn Van Campenhout (2023) Joint Forces: The Impact
of Intrahousehold Cooperation on Welfare in East African Agricultural Households, Feminist
Economics, 29:1, 266-297, DOI: 10.1080/13545701.2022.2120206
To link to this article:  https://doi.org/10.1080/13545701.2022.2120206
© 2022 The Author(s). Published by Informa
UK Limited, trading as Taylor & Francis
Group
View supplementary material 
Published online: 02 Oct 2022.
Submit your article to this journal 
Article views: 758
View related articles 
View Crossmark data
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=rfec20
Feminist Economics, 2023
Vol. 29, No. 1, 266–297, https://doi.org/10.1080/13545701.2022.2120206
JOINT FORCES: THE IMPACT OF INTRAHOUSEHOLD
COOPERATION ON WELFARE IN EAST AFRICAN
AGRICULTURAL HOUSEHOLDS
Els Lecoutere and Bjorn Van Campenhout
ABSTRACT
In low- and middle-income countries, poor cooperation between members
of smallholder agricultural households may lead to inefficient allocation
of productive resources. This study estimates the causal mediating effects
of cooperation between spouses on household welfare and public goods
provision in Ugandan and Tanzanian monogamous smallholder coffee farming
households. The random encouragement to participate in an intensive
training program coaching couples in farming as a household enterprise and
participatory intrahousehold decision making, which stimulates cooperation
and, in turn, household welfare and public goods provision, enables estimating
causal mediating effects while avoiding challenges of endogeneity. Spousal
cooperation has positive mediating effects on household welfare, measured by
total household income per capita and food security, and on household public
goods provision, measured by the adoption intensity of agronomic practices
and use of improved seed for food crops. Spousal cooperation has larger effects
on total household income per capita with longer duration of marriage.
KEYWORDS
Causal mediation analysis, spousal cooperation, household welfare, household
public goods provision, agricultural households, East Africa
JEL Codes: Codes: D13, O12, Q12
HIGHLIGHTS
• In Uganda and Tanzania, the Gender Household Approach program
aims to improve gender relations by promoting spousal cooperation.
• Participatory decision making implies strengthening women’s voice and
ability to include their claims in a household.
• GHA presents a concept of women’s empowerment that avoids backlash
by promoting shared control of resources and agency.
© 2022 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis
Group
This is an Open Access article distributed under the terms of the Creative Commons
Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/
licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and
reproduction in any medium, provided the original work is properly cited, and is not
altered, transformed, or built upon in any way.
ARTICLE
• Programs that promote spousal cooperation can improve the welfare
and public goods provision of agricultural households.
INTRODUCTION
Smallholder farming continues to be the most important way to sustain a
livelihood for most households in Sub-Saharan Africa. By now, a large body
of literature suggests that many smallholder household farms produce well
below capacity, partly due to a lack of cooperation and unequal bargaining
power between the main decision makers in agricultural households,
which leads to under-provisioning of household public goods (Udry 1996;
Iversen et al. 2011; Doss 2013; Doss and Meinzen-Dick 2015; Fiala and He
2017; Munro 2018).1 The conceptualization of agricultural households as
a group facing collective action problems similar to the ones arising in
common pool resource (CPR) settings, as proposed by Cheryl Doss and
Ruth Meinzen-Dick (2015), provides a theoretical basis for the potential
benefits of intrahousehold cooperation for household outcomes. While the
extent to which members of (agricultural) households cooperate has been
studied extensively, there is less evidence of the implications of improved
intrahousehold cooperation for household welfare and household public
goods provision (Doss and Quisumbing 2020).
The key research objective of this article is to estimate the impact of
cooperation between spouses in agricultural households on household
welfare and household public goods provision. We do so through causal
mediation analysis. The contributions of this approach to studying the
impact of cooperation are threefold. First, it addresses the challenges
that cooperation is likely to be endogenous (Doss 2013) and difficult to
directly manipulate (hence, randomly assign). Second, in comparison to
more “traditional” mediation analyses, causal mediation analysis accounts
for some of the bias following from potentially confounded mediator-
outcome association (Pearce and Vandenbroucke 2016). Third, it responds
to an increasing interest in understanding the extent to which effects on a
particular outcome are mediated through a hypothesized mechanism. This
is useful for understanding how interventions work and, in this case, also
to empirically test a theory-based hypothesis that cooperation will improve
efficiency and welfare; both of which have policy relevance (VanderWeele
2016).
To address the challenges related to endogeneity and random
assignment of cooperation, Esther Duflo and Christopher Udry (2004)
used an instrumental variables approach, and Nancy McCarthy and
Talip Kilic (2017) used fixed effects panel data regressions. To our
knowledge, apart from Jessica Heckert, Deanna Olney, and Marie Ruel
(2019) who use structural equation modeling to examine mediation effects
of communication between spouses on children’s nutritional outcomes,
267
INTRAHOUSEHOLD COOPERATION
causal mediation analysis has not been applied to examine the effect of
spousal cooperation on household outcomes. Causal mediation analysis is
used to examine mediation through hypothesized mechanisms in other
examples, among others, by Leyla Karimli et al. (2021) who look into
causal mediation of increased access to assets on women’s intrahousehold
decision-making power, Alan de Brauw et al. (2018) who examine
mediation of nutrition information on the adoption of bio-fortified crops,
and Fenella Carpena and Bilal Zia (2020) who investigate mediation
of financial literacy on financial behavior. In this study, we exploit the
random encouragement to participate in a couples’ training intervention,
hypothesized to stimulate spousal cooperation that, in turn, is expected to
lead to increased household welfare and public goods provision, to estimate
causal mediating effects of cooperation.
As such, this article addresses the gap in robust evidence of causal effects
of spousal cooperation on household welfare and public goods provision.
We find that, in monogamous smallholder coffee farming households in
Uganda and Tanzania, over 10 percent of the total positive intention-to-
treat (ITT) effect of a randomly encouraged couples’ training intervention
on total household income per capita as a measure of household welfare
is mediated by spousal cooperation. Approximately 45 percent of the total
positive ITT effect on the likelihood of household food security is mediated
by cooperation. We also show causal mediation by cooperation on measures
of household public goods provision. Spousal cooperation mediates over 25
percent of the positive ITT effect of the couples’ training intervention on
adoption intensity of agronomic practices, and almost 20 percent of the
positive effect on the use of improved seed for food crops.
THE RELATIONSHIP BETWEEN INTRAHOUSEHOLD
COOPERATION AND HOUSEHOLD WELFARE AND PUBLIC
GOODS PROVISION
Theory and prior evidence
Different theoretical models lay out how agricultural households
(co-)operate. The unitary household model, which assumes aligned
preferences by household members and cooperative bargaining models
that do not but predict Pareto efficient household outcomes (Alderman
et al. 2003),2 are not always empirically supported (Udry 1996; Iversen et al.
2011; Fiala and He 2017). Non-cooperative bargaining models recognize
that households can persist in non-cooperation (Duflo and Udry 2004;
Doss 2013). The separate spheres bargaining model allows for household
outcomes, including public goods provision, to be inefficient (Lundberg
and Pollak 1993; McCarthy and Kilic 2017).
268
ARTICLE
The theory and evidence on intrahousehold allocation implies that
impact will depend on which household member is supported through
the intervention (Doss and Quisumbing 2020). As there is also growing
evidence of positive associations of women’s empowerment and limited
gender disparities in the household with a range of household and
children’s development outcomes (Quisumbing and Maluccio 2003;
Sraboni et al. 2014; Seymour 2017; Santoso et al. 2019; Quisumbing et al.
2021), many interventions have focused on improving on women’s versus
men’s bargaining power to achieve better outcomes. Cheryl Doss and Agnes
Quisumbing (2020), however, argue that, in some cases, interventions may
have missed opportunities for more effectively improving development
outcomes by ignoring potential areas for cooperation within households.
Agricultural households can be conceptualized as a group facing
collective action problems (Doss and Meinzen-Dick 2015). In parallel with
the CPR literature, cooperation implies the absence of opportunism by
household members regarding the provision of the household’s commons
and the appropriation of resources produced in the household, which
is associated with more efficient and sustainable outcomes (Baland and
Platteau 1998; Agarwal 2000; Agrawal 2003). As in CPR settings, in turn,
the experience of less opportunistic provision and consumption can be
expected to strengthen incentives for cooperative behavior (Ostrom 1990;
Agrawal 2003; Doss and Meinzen-Dick 2015). The latter brings in a
dynamic perspective. This is relevant because households are characterized
by repeated interactions between members over longer periods of time.
Household members may seek maximization of their utility over a series
of interactions, and possibly even take into account mutual obligations or
interdependence (Doss and Meinzen-Dick 2015; Munro 2018).
In CPR settings, reduced information asymmetry, credible commitments
to, and mutual monitoring of mutually agreed upon “rules” have been
shown to be instrumental for cooperation (Ostrom 1990; Bardhan 2000;
Agrawal 2003). The effective participation in rule and decision making and
monitoring of compliance with those rules not only increases cooperation
and compliance, but also the likelihood that the rules are fair (Doss and
Meinzen-Dick 2015). Indirectly, reduced inequality could contribute to
efficiency by alleviating constraints to opt for the most efficient options
(Slootmaker 2014), and because fairness in the allocation of benefits from
common resources is conducive for sustained cooperation (Ostrom 1990;
Baland and Platteau 1999).
Despite the theoretical basis for the benefits of intrahousehold
cooperation, robust evidence of its effects on household welfare and public
goods provision remains limited. Duflo and Udry (2004) show that, in
Cote D’Ivoire, jointly controlled income as an indicator of cooperation,
which is proxied by rainfall patterns favoring crops normatively used
to the benefit of the household, has a significant positive impact on
269
INTRAHOUSEHOLD COOPERATION
household public goods expenditures. McCarthy and Kilic (2017) find
that, in rural Malawi, cooperation, proxied by the share of jointly
controlled household income, is positively related to total household
income and consumption expenditures per capita, as well as the share
of household consumption devoted to public goods. Heckert, Olney, and
Ruel (2019) find that communication between spouses, as a proxy for
spousal cooperation, mediates the positive impact of a nutrition-sensitive
agriculture program in Burkina Faso on nutritional outcomes of infants.
Increased spousal communication is significantly associated with decreased
wasting in infants, more strongly so than other mediators that relate to
women’s empowerment.
Other evidence of the relationship between intrahousehold cooperation
and household outcomes relies on mixed methods analysis combining
lab-in-the-field-experiments with observational data. Els Lecoutere and
Laurence Jassogne (2019) show that, in Ugandan smallholder coffee
farming households, lab-measured intrahousehold decision making
supportive of cooperation is significantly positively associated with
subjective household well-being and food security, as well as investment in
household public goods such as sustainable crop-intensification practices.
Jessica Hoel et al. (2017) find that, in Senegalese pastoralist dairy farming
households who exhibit more cooperative behavior in the lab, female-
owned cows produce more milk than male-owned cows and the gender
gap in productivity is smaller. In Burkina Faso, Richard Akresh, Joyce J.
Chen, and Charity T. Moore (2016) find that yield gaps between plots
cultivated by different co-wives in polygynous households are smaller than
yield gaps between plots cultivated by wives and (male) household heads
in monogamous households. The authors relate this to the fact that, in the
latter case, cooperation is inhibited by greater degrees of altruism, which
limits the scope for punishment or inspires non-cooperative choices.
Promoting intrahousehold cooperation through the gender household
approach
The intervention promoting intrahousehold cooperation that is studied
here is a package of intensive couples’ coaching activities that is part of
a program called the Gender Household Approach (GHA).
The context
The GHA is implemented among smallholder coffee farming households
in Mubende district in central Uganda and Mbeya rural and Mbozi districts
in southern Tanzania by the Hanns R. Neumann Stiftung (HRNS), a non-
profit organization working with coffee farmers.
270
ARTICLE
The smallholder coffee farming households included in this study
typically produce food crops for household consumption, of which excess
harvest is sold, and some cash crops – mostly coffee in this case – for
marketing. In Mubende, mainly Robusta coffee is grown, in Mbeya rural
and Mbozi Arabica coffee is the main variety. The household farm system
comprises of productive resources such as land, (family) labor, financial
and other assets, from which agricultural produce and income are derived.
In these contexts, typically, men have most decision-making power over
agricultural production, especially over marketable crops such as coffee,
and marketing of crops produced on the household farm (Fisher and
Carr 2015; Meier zu Selhausen 2016; Lecoutere and Jassogne 2019). This
is tied to their (informal) ownership of the land, which women lack as
they tend to marry into the community with limited own resources. Labor
for agricultural production, coffee production in particular, tends to be
provided by both spouses, sometimes also by other household members.
Some tasks are divided along gender-specific roles, and some tasks are
performed together (Meier zu Selhausen 2016; Lecoutere and Jassogne
2019). Women do most of the domestic and reproductive care activities.
Men generally have greater decision-making power than women over the
allocation of benefits of household farming and expenditures of household
income (Lecoutere and Jassogne 2019). As in western Uganda (Mayoux
2012; Meier zu Selhausen 2016), in Mubende (Uganda) and Mbeya-Mbozi
(Tanzania), there is anecdotal and qualitative evidence of non-cooperative
(opportunistic) behavior whereby one spouse sells part of the coffee
without the other one knowing or whereby one spouse, often the husband,
excessively consumes farm income rather than invest it in the farm or the
household (Lecoutere and Chu 2021; Lecoutere and Wuyts 2021).
The households in our study population are member of mixed-gender
agricultural producer organizations (POs) that focus on coffee production
and are linked to HRNS. The POs require abidance by by-laws and
member fee payment and include twenty-five to seventy-five members.
In both countries, loosely organized (coffee) farmer groups or farmers
from the same village organized themselves in POs. Standard interventions
organized by HRNS for the POs and their members include farmer
field schools, training in sustainable agronomic intensification practices,
climate change mitigation, good post-harvest practices for coffee, and
joint marketing via coffee collection points. The POs do not provide
microfinance services.
Characteristics of the intervention
The GHA fits in the type of interventions commonly referred to as
“household methodologies” (FAO, IFAD, and WFP 2020). These are
gender-transformative approaches that address gender relations within
271
INTRAHOUSEHOLD COOPERATION
households, avoid empowering women at the expense of men, and
project a common vision and goal for the household to achieve (Mayoux
2012; Farnworth et al. 2018). The GHA projects agricultural production,
including coffee production, as a household farm enterprise where all
household members, specifically the spouses, can contribute to and equally
benefit from. The GHA promotes cooperation within the household,
sharing resources, and a more participatory way of intrahousehold
decision making, whereby spouses consult each other and come to mutual
agreement on strategic farm and household decisions, and women, de
facto, become more involved. The GHA avoids exclusively empowering
women in households to avoid opposition by men and the wider
community where a patriarchal family model prevails, as well as to avoid
overburdening women if they would take over responsibilities and roles that
are normally shared by spouses. HRNS piloted the GHA in selected areas
where limited cooperation and women’s empowerment in households are
prevailing challenges.
The GHA is introduced in the communities and POs by first informing
and sensitizing the PO leaders and overarching committees. Couple seminars,
organized at the PO level, make up the first GHA activity for couples
and aim to enhance awareness of gender role divisions and gender
imbalances in contributions to and benefits from household farming.
After invitation by the HRNS gender officers and PO leaders, self-
identified (men–women) couples heading smallholder (PO member)
coffee-farming households self-select into the couple seminars. During
the couple seminars, couples are guided through a self-assessment of
the division of roles and responsibilities and control over resources in
their households with the help of activity profiles, control over resources
matrices, and a subsequent group discussion.
The next stage of the GHA for couples who participated in couple
seminars is the Intensive-Coaching Package for couples consisting of four
activities. First, during a one-day seminar, the concept of a household
farm enterprise is operationalized. Couples are coached on how to consult
each other and come to mutual agreement and how to put participatory
planning and decision making into practice. Together, spouses draft a
plan and draw up a budget for household and farm to achieve their
envisioned goals. This plan and budget are also used as communication
tools throughout the coaching package. After that, each couple receives a
home visit by the HRNS gender officer who continues the coaching and
follows up on the way the spouses collaborate, share decision making, and
implement their farm plan and budget. The third activity is a women’s
leadership training to strengthen women’s participation and leadership
skills within farmer groups and their household. The final activity is
a half-day follow-up workshop in which couples share experiences and
self-evaluate the coaching package and their development.
272
ARTICLE
METHOD
Our identification strategy relies on the randomized encouragement for
the intensive-coaching package across treatment and control groups.
The random encouragement for the intensive-coaching package, which
is hypothesized to stimulate spousal cooperation, which, in turn, is
hypothesized to affect household welfare and public goods provision, offers
a solution to the challenge of the potential endogeneity of cooperation and
the outcomes of interest. It allows estimating the causal mediation effect of
cooperation between spouses on household welfare and household public
goods provision, which is our key objective.
We will first discuss the data, the intervention’s pathways of change, and
the research hypotheses, after which we will elaborate on the identification
strategy, causal mediation analysis, and heterogenous mediation effects.
Thereafter, we will define the indicators of spousal cooperation, household
welfare, and public goods provision.
Data
The implementation of couple seminars and the intensive-coaching
package started in November 2016 in Mubende district (Uganda) and in
January 2017 in Mbeya rural and Mbozi districts (Tanzania). At baseline,
the data collection took place after couple seminars and before the
encouragement for the intensive-coaching package. Endline data was
collected from mid-January until mid-March 2018 in Mubende, from March
until May 2018 in Mbeya rural and Mbozi. Base- and endline interviews were
done in approximately the same order with on average one year between
interviews.
Both at base- and endline, we conducted individual surveys separately
with the men and women spouses of the couples in our study population.
Presence of the other spouse or other people (except infants) during the
individual interview was avoided. Descriptive statistics of key characteristics
of our study population measured at baseline are included in Table 1.
Our study sample includes a Treatment group (T) that is composed of
couples who were randomly selected out of the couples who participated
in couple seminars to be encouraged for participation in the intensive-
coaching treatment that was subsequently organized. Our study sample also
includes a Control group (C) composed of couples randomly selected out of
the couples who participated in couple seminars not to be encouraged for
the intensive-coaching treatment.3 The random allocation of couples from
couple seminar attendance lists to the treatment and control group was
done using random number generation.
The encouragement consisted of a personal phone call by the HRNS
gender officer and a personalized invitation letter by HRNS for the first
273
INTRAHOUSEHOLD COOPERATION
274
Table 1 Descriptive statistics and test of balance in key baseline characteristics across treatment and control groups
Control group (CB) Treatment group (T) Test of difference
T and CB
Sig. based on Sig. based on
Avg SE Avg SE p-value (t-test) FDR q-value N
Age wife 36.926 0.958 37.159 0.582 462
Age husband 43.667 1.097 44.590 0.688 462
Age difference 6.741 0.705 7.431 0.366 462
Number of children living with the 3.941 0.236 3.318 0.130 ∗∗ 462
household (reported by wife (W))
Acerage of land (reported by husband 5.359 0.302 4.997 0.174 443
(HB))˚
Number of cattle owned by the 1.585 0.265 1.618 0.156 462
household (HB)
Number of small livestock owned by 4.222 0.495 2.872 0.343 ∗∗ 462
the household (HB)
Sig. based on Sig. based on
Proportion SE Proportion SE p-value (z-test) FDR q-value N
Wife has at least some secondary 0.141 0.030 0.138 0.019 462
education
Husband has at least some secondary 0.193 0.034 0.232 0.023 462
education
Husband reported owning a bicycle 0.593 0.042 0.511 0.028 462
Husband reported having earned 0.215 0.035 0.306 0.026 ∗∗ 462
off-farm income in last 3 months
(Continued).
ARTICLE
275
Table 1 Continued.
Control group (CB) Treatment group (T) Test of difference
T and CB
Avg SE Avg SE Sig. based on p- Sig. based on FDR N
value (t-test) q-value
Wife reported having received 0.059 0.020 0.052 0.012 462
remittances in last 3 months
House has iron roof (reported by W) 0.993 0.007 0.979 0.008 462
House is built with fire-baked bricks 0.867 0.029 0.746 0.024 ∗∗∗ † 462
(reported by W)
Household is food secure (reported by 0.830 0.032 0.758 0.024 ∗ 462
W)
Notes: Acerage topped at fifteen acres (95 percent of sample). Avg: Average; SE: Standard error. ∗∗∗, ∗∗, ∗ denote significance at the 1, 5, and 10 percent levels,
respectively. †††, ††, † denote significance based on FDR (Sharpened False Discovery Rate) q-values at the 1, 5, and 10 percent levels, respectively. Indicator
definitions and details of survey measurement in Online Appendix B.1. Descriptive statistics and balance tests by Mubende and Mbeya-Mbozi subsamples in Table
B in Online Appendix B.2.
INTRAHOUSEHOLD COOPERATION
activity of the intensive-coaching package, accompanied by a folder with
a blank notebook and two pens.4 Couples that showed up for the first
activity were closely monitored and, if needed, additional invitation efforts
and a second chance to participate if they missed an activity ensured they
participated in all four activities of the coaching package. The personalized
encouragement makes it unlikely that the non-interference assumption
is violated through passing the invitation on to neighbors, family, or
friends.
For this study, we excluded all couples in which either spouse reported
that they are a polygamous household because spousal cooperation may
be subject to other dynamics than in monogamous couples. In Mubende,
the couples to be part of the treatment or control group were randomly
selected out of couples who participated in couple seminars and who
were known not to be polygamous. Crosschecking with base- and endline
data collected from both spouses confirmed the absence of polygamous
households in this subsample. In Mbeya-Mbozi, we randomly selected
couples to be part of the treatment or control group regardless of whether
they were monogamous or polygamous. We oversampled by 25 percent
which was the expected proportion of couples in polygamous relations in
this area. For this study, we excluded the couples in polygamous relations
ex-post (based on whether either spouse reported to be in a polygamous
relationship at base- or endline).
The sampling implies that the external validity of the results pertains to
self-identified (male-female) couples in monogamous relationships who are
member of agricultural POs focused on coffee production linked to HRNS
and who self-selected into couple seminars raising awareness about farming
as a household enterprise and intrahousehold decision making.
The determination of samples sizes was informed by ex-ante power
calculations, of which a discussion is included in the Supplemental Online
Appendix A. These were based on the estimated effects of spousal
cooperation on total household income per capita and expenditures
devoted to household public goods in McCarthy and Kilic (2017) and
estimated proportion of variation in outcome variables explained by
covariates in a prior study in Masaka district, Uganda.
The sample size of our study population amounts to 462 couples, of which
327 are in the treatment group (including fourteen non-compliers to their
encouragement status) and 135 in the control group (Table A in Online
Appendix A). Both treatment and control groups include approximately
equal numbers of couples from Mubende (Uganda) and Mbeya-Mbozi
(Tanzania). The couples are member of seventy-four different POs in
Mubende, of 110 POs in Mbeya-Mbozi. Our effective sample size for causal
mediation analysis amounts to 345 households (reduced from the original
sample size partly due to listwise omission of missing cases in indicators
included in indices of cooperation).
276
ARTICLE
Pathways of change and hypotheses
The pathways of change of the intensive-coaching package in the GHA,
visualized in Figure 1, are compatible with the conceptualization of
agricultural households as groups managing a set of common resources
where members are mutually interdependent on the others’ individual
decisions about provision and appropriation of (the benefits of) those
resources (Doss and Meinzen-Dick 2015).
First, the framing of agricultural production as a household farm
enterprise emphasizes its collective nature and is expected to encourage
collaboration.5 It is also expected to promote participation in rule and
decision making by both spouses, as is the more participatory way of spousal
decision making about farm and household. Incentives for cooperation and
compliance with mutual decisions are expected to be strengthened.
Participatory intrahousehold decision making is also expected to
reduce information asymmetries between spouses about one another’s
contributions to generating farm income and managing the household,
as well as the allocation of benefits generated through the household
farm, hence enable mutual monitoring. Consequently, the likelihood of
opportunistic behavior, such as, for instance, overconsumption or side-
selling of farm produce, is expected to reduce in favor of cooperation
and provision of the commons in the form of, for instance, greater labor,
financial, or in-kind investments in farm and household. These changes
are anticipated to improve household welfare, incrementally so when
the experience of less opportunism strengthens incentives for continued
cooperative behavior.
A participatory way of intrahousehold decision making implies
strengthening women’s voice and ability to include their claims in
household “rules” and enforce compliance with sharing rules, which are
normally limited in this context with a prevailing patriarchal family model.
This reduces the likelihood of unfair outcomes which is motivational for
sustained cooperation. Reduced inequities are also expected to alleviate
constraints to choosing the most household efficient options, such as, for
instance, investing in sustainable intensification of food production or
more nutritious crop varieties (Slootmaker 2014; Gilligan et al. 2020).
Based on these pathways of change, we hypothesize that the intensive-
coaching package of the GHA will have a positive effect on spousal
cooperation (Hypothesis H1; Figure 1: Upper left curved light blue arrow)
and a positive effect on household welfare and household public goods
provision (H2; Figure 1: Right curved white arrow). We hypothesize that
spousal cooperation will prove a significant mediator of the effects of
the intensive-coaching package on household welfare and public goods
provision (H3; Figure 1: Lower left curved dark blue arrow), which is the
key hypothesis of interest in this study.
277
INTRAHOUSEHOLD COOPERATION
Figure 1 Pathways of change of the Gender Household Approach
intensive-coaching package for couples.
Notes: Dark blue box: Intensive-coaching package; Light blue box: Aspects
of spousal cooperation; Medium blue box: Outcomes in terms of household
welfare and household public goods. Curved arrows: Hypothesized effects of
the intensive-coaching package on cooperation (light blue; Hypothesis H1); on
household welfare and public goods mediated through cooperation (dark blue;
H3); on household welfare and public goods (white; H2).
Heterogeneous mediation effects of cooperation on household welfare
and public goods provision may exist. Former qualitative work revealed
that the most likely heterogeneous mediation effects are linked to the
education level of the spouses, age difference, age of the husband, and
length of marriage (Lecoutere and Jassogne 2019; Lecoutere and Chu
2021; Lecoutere and Wuyts 2021). We hypothesize that the mediating
278
ARTICLE
effect of cooperation on household welfare and public goods provision will
be larger among couples in which one or both spouses have a relatively
high education level (H4a) and smaller if the age difference between
spouses is relatively large (H4b) and the husband is relatively old (H4c).
If the wife is relatively old, we anticipate a larger mediating effect of
cooperation because it implies a longer duration of marriage, which is said
to be conducive for household welfare through accrued cooperation and
harmony in the household (H4d).
Identification strategy
First, we test balance across the randomized treatment and control
groups based on baseline data in the entire sample. We use t-test to
assess differences in continuous variables and z-test to test differences in
proportions. We correct the p-values of tested differences for multiple
hypotheses testing by calculating (sharpened) False Discovery Rate (FDR)
q-values using the method by Michael Anderson (2008) described in
David McKenzie (2021). Based on FDR q-values, the tests demonstrate that
the treatment and control group are balanced on most of the baseline
characteristics in the total sample, except the proportion with a fire-
baked brick house (Table 1). There is balance on most characteristics in
the Mubende and Mbeya-Mbozi subsamples as well (Table B in Online
Appendix B.2).
Our identification strategy relies on the randomized encouragement for
the intensive-coaching package across treatment and control groups. We
use endline data to estimate ITT effects of the intensive-coaching package
on outcome variable Yi , with the encouragement status as the (binary)
treatment variable (Ei) for couple i (with error term εi ; Equation 1; Testing
H2). We test the hypothesis of the absence of an ITT effect, β1 = 0.
Rejecting that hypothesis favors the alternative hypothesis of an ITT effect
that is statistically significantly different from zero.
Yi = β0 + β1Ei + β2Ai + β3{Xi} + ε1i (1)
In each regression model estimating treatment effects, we include a
dummy variable for Mubende to control for area fixed effects (Ai). The
presence of sources of imbalance, even if limited, warrants a cautionary
note on potential selection bias. Taking a conservative approach, we
estimate ITT effects with inclusion of each of the covariates in which
imbalance may exist in either of the subsamples in the regression model
(that is, where p-values uncorrected for multiple hypotheses testing point
to significant differences [Table 1, Table B in Online Appendix B.2]). The
vector {Xi} includes the following control variables measured at baseline:
House is built with fire-baked bricks, number of small livestock owned by
279
INTRAHOUSEHOLD COOPERATION
the household, household is food secure, number of children, husband
reported having earned off-farm income in the last three months. We
use ordinary least squares (OLS) regression for continuous and Probit
regression for binary outcome variables.
Causal mediation analysis
The key research objective is estimating the impact of cooperation
on household outcomes, more specifically, testing H3 whether spousal
cooperation is a significant mediator of the effects of the (encouragement
for the) intensive-coaching package on household welfare and public goods
provision. Hence, we are particularly interested in the average (intention-
to) treatment effect through cooperation as a mediating variable Mi(Ei)
(Equation 2), which, itself, is affected by the (randomized encouragement
for) treatment Ei (Equation 3).
Yi = β0 + β1Ei + β2Ai + β3{Xi} + β4Mi + ε2i (2)
Mi = γ0 + γ1Ei + γ2Ai + γ3{Xi} + ε3i (3)
The causal mediation effect δi(Ei) of spousal cooperation is equal to the
change in outcome Yi through the change in the mediator variable from
the control (Ei = 0) to the treatment condition (Ei = 1) while holding the
effect of the treatment otherwise constant (Equation 4; Imai et al. 2011; de
Brauw et al. 2018). Averaged over couples i, the average causal mediation
effect (ACME) is δ̄(E).
δi(Ei) = Yi(Ei , Mi(1)) − Yi(Ei , Mi(0)) (4)
The direct (intention-to-)treatment effect ζi(Ei) is what remains after the
indirect (mediated) effect is estimated (Equation 5). The average direct
effect (ADE) is ζ̄(E). The total effect is the sum of ACME and ADE.
ζi(Ei) = Yi(1, Mi(Ei)) − Yi(0, Mi(Ei)) (5)
The counterfactual Mi(0) for couples receiving the (encouragement for)
treatment must be estimated as it cannot be observed. Estimating the
ACME, as well as the ADE, relies on two assumptions for identification. The
first assumption that, conditional on pretreatment covariates, the treatment
is statistically independent of mediator and outcome variables holds due to
randomization (Imai, Keele, and Tingley 2010; de Brauw et al. 2018).
The second assumption of sequential ignorability (or mediator
exogeneity assumption) requires that, conditional on actual treatment
status and pretreatment covariates, the mediator is statistically independent
280
ARTICLE
of the potential outcome. If there would be unobserved confounders jointly
affecting the mediator and the outcome, estimates of the ACME could be
biased (de Brauw et al. 2018). Making this assumption allows estimating
ACME and ADE. The robustness of the estimates to this assumption can
be evaluated through an analysis of the sensitivity of the estimates to
unobservable confounders that might be correlated with both mediator
and outcome. If the assumption of sequential ignorability would hold,
the correlation ρ between error terms ε3i and ε2i would be zero. The
sensitivity analysis consists of, on the one hand, an estimated ρ at which
ACME would be zero (the smaller the value, the more sensitive the ACME
estimate is to a potential violation of sequential ignorability; in other words,
a small unobserved confounder could overturn conclusions obtained under
sequential ignorability) and, on the other hand, a visualization of ACME
estimates for varying ρ (Yamamoto 2016).
In the causal mediation analysis procedure, we apply OLS regression for
continuous and Probit regression for binary outcome variables. We use OLS
regression for the mediation part of the analysis as our mediator variables
are continuous. The causal mediation analysis procedure estimates (ITT)
effects of the treatment on the mediating variable as well (Equation 3;
Testing H1).
As we evaluate ACME on multiple outcomes, the likelihood of finding a
significant effect by chance increases. We account for that by calculating
FDR q-values which correct p-values for multiple hypothesis testing using
the Anderson (2008) method (McKenzie 2021). We adjusted for testing
ten hypotheses of statistical significance of the ACME on five outcome
indicators mediated through two different indices of spousal cooperation.
With an effective sample size for causal mediation analysis of 345 couples
while taking into account unequal treatment allocation, ex-ante power
calculations indicate that minimum detectable standardized effect size
(MDES) for 80 percent power and 0.05 significance level amounts to 0.19
(resp. 0.31) when a high (resp. low) proportion of variation is explained by
covariates (Online Appendix A.3). We additionally calculate ex-post MDES
with 80 percent power and 0.05 significance by multiplying the standard
error of estimated ACME by 2.8, as recommended by David McKenzie and
Owen Ozier (2019) and applied for causal mediation analysis in Carpena
and Zia (2020). In the result section, we assess statistically insignificant
ACME estimates against ex-post MDES in endnotes.
Testing for heterogeneous causal mediation effects
To test for the presence of heterogeneous causal mediation effects of
cooperation (Testing H4), we run the causal mediation analysis for couples
in which one or both spouses have a relatively high education level (at least
some secondary education) and for couples where this is not the case. We
281
INTRAHOUSEHOLD COOPERATION
then test the statistical significance of difference in the ACME using t-tests.
We do the same for couples where the age difference between spouses
is relatively large (that is, larger than the average of seven years) versus
not large; where the husband is relatively old (that is, 52 years, the 75th
percentile age, or older) versus not old; and where the wife is relatively old
(that is, 44 years, the 75th percentile age, or older) versus not old.
Indicators
Indicators for cooperation between spouses
Following good practice of avoiding testing a large number of hypotheses
and running risks of false discovery, we construct a first index indicator of
cooperation between spouses (Cooperation index) based on six measures of
spousal cooperation measured at endline using the method described in
Anderson (2008).6 The first measure is the share of total seasonal income
earned from coffee, as the household’s (second) most important cash
crop, received jointly by husband and wife (Share of total coffee income jointly
received by spouses; Detailed definitions of measures in Online Appendix
B.3). The second measure is the share of jointly owned tropical livestock
units (TLU) in total household owned TLU (Share of total TLU jointly owned
by spouses). The third measure takes the value one if husband and wife
agree with the statement that husband and wife should have an equal say
about the household’s cash crops and their production (Spouses agree on
having equal say about cash crops). The fourth measure is the proportion
of decisions about the adoption of eight possible sustainable agronomic
intensification practices that have been made jointly (agreed upon by
husband and wife; Share of adoption decisions jointly made by spouses).7 The
fifth and sixth measures take the value one if husband and wife agree they
jointly manage (that is, make the majority of agricultural decisions) their
most important food crop grown in the household farm (Spouses agree they
jointly manage food crop), respectively their most important cash crop (Spouses
agree they jointly manage cash crop). The mean value of the cooperation index
is 0.751 (standard deviation S.D. 1.062; Descriptive statistics of measures and
indices in Table 2; Correlation matrix of measures in Table C, Online
Appendix C).
A second index for cooperation between spouses (Alternative
cooperation index) is based on two measures, the Share of total coffee
income jointly received by spouses and the Share of total TLU jointly
owned by spouses. We assume these measures and the resulting alternative
cooperation index, solely based on joint control over common resources
and not on joint decision making, are not incompatible with specialization
as a way of cooperation (See Endnote 5). The alternative cooperation index
has a mean value of 0.532 (S.D. 1.152).
282
ARTICLE
Table 2 Descriptive statistics of endline indicators of household welfare, household
public goods provision, and cooperation
Avg/Proportion SD N
Indicators of cooperation between spouses
(i) Share of total coffee income jointly received 0.523 0.467 385
by spouses
(ii) Share of total TLU jointly owned by spouses 0.307 0.412 461
(iii) Spouses agree on having equal say about cash 0.377 0.485 462
cropsb
(iv) Share of adoption decisions jointly made by 0.449 0.388 394
spouses
(v) Spouses agree they jointly manage food cropb 0.801 0.400 462
(vi) Spouses agree they jointly manage cash cropb 0.630 0.483 462
Cooperation index 0.751 1.062 345
(based on indicators (i) through (vi))
Alternative cooperation index 0.532 1.152 384
(based on indicators (i) and (ii))
Outcome indicators of household welfare and public goods provision
Income per capitaa 181.443 257.152 396
Evolution coffee incomea − 30.600 633.308 332
Foodsecure at endlineb 0.623 0.485 462
Adoption intensity of agronomic practices 3.714 2.633 462
Used improved seed for food cropsb 0.411 0.493 462
Notes: aIn USD, exchange rate on March 15, 2018, 1 USD = 3600 UGX; 1 USD = 2250 TSH.
bDummy variable. Avg: Average; SD: Standard Deviation.
Indicators for the outcomes of household welfare and household public goods
provision
As a first indicator for household welfare at endline, we use total household
income per capita at endline (Income per capita; Descriptive statistics in
Table 2; Detailed indicator definitions in Online Appendix B.3).8 A second
monetary indicator for household welfare is the evolution between base-
and endline of seasonal coffee income reported by the husband (Evolution
coffee income). The third indicator takes the value one if the wife reported
the household did not have to eat less preferred foods nor did it have to
reduce the number or size of meals in the course of three months prior to
the endline interview (Foodsecure at endline).
A first indicator of household public goods provision9 is the adoption
intensity of sustainable agronomic intensification practices for coffee
production and improved seed(lings) for other crops, which has a
283
INTRAHOUSEHOLD COOPERATION
maximum value of eight and is based on husband’s reporting (Adoption
intensity of agronomic practices). To check the extent to which investments
are made into food production, which can be assumed to align more with
women’s priorities (Quisumbing and Maluccio 2003; Njuki et al. 2011), we
construct an indicator taking the value one if the husband reported to have
used improved seed(lings) for a staple food crop, maize in most cases (Used
improved seed for food crops).
RESULTS
Intention-to-treat effects of intensive coaching on indicators of spousal
cooperation
Results in Table 3 Panel A show a statistically significant positive ITT
effect of the intensive-coaching package on the cooperation index as
the mediator amounting to 0.622; an increase equivalent to double the
cooperation index value in the control group (ceteris paribus).10 The
statistically significant positive ITT effect on the alternative cooperation
index as the mediator amounts to 0.582; an increase equivalent
to one time the alternative cooperation index value in the control
group.
Intention-to-treat effects of intensive coaching on outcome indicators of
household welfare and public goods provision
Results in Table 3 Panel B demonstrate that the intensive-coaching package
had a statistically significant positive ITT effect of 96 USD on total
household income per capita and an effect of 273 USD on the evolution
in coffee income between base- and endline (Columns 1–2).11 These
are respectively an approximately 60 percentage point increase and 75
percentage point increase versus the control group (ceteris paribus).12 The
ITT effect on the likelihood that the household is food secure at endline
is 0.580, equivalent to a marginal effect of an increase of the probability by
0.173 versus the control group (Column 3).
Similarly, there are statistically significant and positive ITT effects on
indicators of household public goods provision. The adoption intensity
of agronomic practices increased by about one practice (0.909), an
approximately 20 percentage point increase versus the control group
(Column 4). The ITT effect on the likelihood of having used improved
seed for food crops is 0.674; equivalent to a marginal effect of
an increase of the probability by 0.179 versus the control group
(Column 5).
284
ARTICLE
285
Table 3 Intention-to-treat effects of intensive-coaching on indicators of spousal cooperation and household welfare and public goods
provision
(1) (2) (3) (4) (5)
Panel A: ITT estimates of intensive coaching on mediator variables
Alternative
Cooperation cooperation
Mediator variable index index
Average ITT effect on mediator 0.622 0.582
SE 0.122 0.111
Sig. based on p-value ∗∗∗ ∗∗∗
N 345 384
Panel B: ITT estimates of intensive coaching on outcome variables
Adoption
intensity of Used
Income per Evolution Foodsecure at agronomic improved seed
Outcome variable capita coffee income endline practices for food crops
Average ITT effect 96.08 272.958 0.580 0.909 0.674
SE 21.468 79.424 0.152 0.151 0.162
Sig. based on p-value ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
R2/Pseudo R2 0.189 0.046 0.200 0.703 0.300
N 396 332 462 462 462
Covariates Yes Yes Yes Yes Yes
Notes: Columns 1, 2, and 4 ordinary least squares regression estimates for continuous outcome variables; Columns 3 and 5 probit regression estimates for binary
outcome variable. ITT effect: Intention-to-treat effect; SE: Robust standard error. ∗∗∗, ∗∗, ∗ denote significance at the 1, 5, and 10 percent levels, respectively.
INTRAHOUSEHOLD COOPERATION
Causal mediation effects of spousal cooperation on outcome indicators of
household welfare and public goods provision
The average causal mediation effects (ACME) of spousal cooperation on
outcome indicators of household welfare and public goods provision are
presented in Table 4 Panel A, sensitivity analyses in Panel B.
Causal mediation effects of cooperation on indicators of household welfare
The ACME through the alternative cooperation index on total household
income per capita is statistically significant (at 11 percent based on FDR
q-values) and positive. It amounts to about 12 USD and covers a share of
12.9 percent of the total effect (Column 2). The ACME has a relatively
high degree of sensitivity to the mediator exogeneity assumption as the
estimated correlation in error terms ρ at which ACME would be zero is
0.084 and the visual sensitivity analysis shows that the confidence interval at
ρ = 0 nearly includes ACME = 0 (Figure B in Online Appendix E).13
The ACME through spousal cooperation measured by the cooperation
index, as well as the alternative cooperation index, on the likelihood
that the household is food secure at endline is statistically significant
(based on FDR q-values) and positive (Columns 5–6). The share of the
total effect mediated through cooperation is estimated at 43.19 and 42.47
percent (respectively for cooperation and alternative cooperation index).
Sensitivity analyses indicate (low to) moderate sensitivity of the ACME
estimates to violation of the mediator exogeneity assumption for both
cooperation indices (ρ at which ACME would be zero is 0.300 for both
indices and the visualized confidence intervals at ρ = 0 do not include
ACME = 0 [Figures E and F in Online Appendix E]).
Causal mediation effects of cooperation on indicators of household public goods
provision
The ACME through spousal cooperation on adoption intensity of
agronomic practices is statistically significant and positive and estimated at
0.152 if measured by the cooperation index, respectively 0.215 if measured
by the alternative cooperation index (Table 4 Panel A Columns 7–8). The
share of the total effect that is mediated through cooperation is estimated
at respectively 25.82 and 28.27 percent. The sensitivity analyses point to
a moderate sensitivity to the mediator exogeneity assumption in case of
the cooperation index (ρ at which ACME would be zero of 0.195 in Panel
B Column 7; Figure G, Online Appendix E); and to a (low to) moderate
sensitivity in case of the alternative cooperation index (ρ at which ACME
would be zero of 0.251 in Panel B Column 8; Figure H, Online Appendix
E).
286
ARTICLE
287
Table 4 Causal mediation effects of spousal cooperation on indicators of household welfare and public goods provision
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Panel A: ACME through mediator on outcome variables
Alternative Alternative Alternative Alternative Alternative
Cooperation cooperation Cooperation cooperation Cooperation cooperation Cooperation cooperation Cooperation cooperation
Mediator variable index index index index index index index index index index
Adoption Adoption Used Used
intensity of intensity of improved improved
Income per Income per Evolution Evolution Foodsecure at Foodsecure at agronomic agronomic seed for food seed for food
Outcome variable capita capita coffee income coffee income endline endline practices practices crops crops
ACME 1.238 11.855 − 3.345 33.589 0.061 0.072 0.152 0.215 0.028 0.015
SE 7.25 7.139 19.728 28.776 0.019 0.02 0.053 0.063 0.015 0.014
Sig. based on p-value ∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗
Sig. based on sharpened [ ∼ ] ††† ††† ††† ††† †
(FDR) q-value
Share of total effect mediated 1.32% 12.97% − 1.26% 13.14% 43.19% 42.47% 25.82% 28.27% 18.17% 8.99%
ADE 92.964 79.722 268.3 225.536 0.081 0.098 0.445 0.554 0.128 0.151
SE 26.366 22.837 94.982 105.921 0.069 0.061 0.183 0.183 0.062 0.054
Sig. based on p-value ∗∗∗ ∗∗∗ ∗∗∗ ∗∗ ∗∗ ∗∗∗ ∗∗ ∗∗∗
Total effect 94.202 91.577 264.956 259.125 0.142 0.170 0.597 0.769 0.156 0.166
SE 23.521 21.29 88.183 87.978 0.067 0.057 0.17 0.176 0.057 0.05
Sig. based on p-value ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗ ∗∗∗
N 345 384 287 313 345 384 345 384 345 384
Covariates Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
INTRAHOUSEHOLD COOPERATION
288
Panel B: Sensitivity analysis
ρ at which ACME = 0 0.012 0.084 − 0.005 0.079 0.300 0.300 0.195 0.251 0.100 0.100
Variance in mediator and
outcome explained by
hypothesized unobserved
confounder:
Proportion of residual 0.01% 0.70% 0.00% 0.62% 9.00% 9.00% 3.80% 6.30% 1.00% 1.00%
variance explained
Proportion of total 0.01% 0.37% 0.00% 0.37% 5.35% 3.35% 0.96% 1.10% 0.54% 0.36%
variance explained
Degree of sensitivity to High High High High Moderate Moderate Moderate Moderate High High
mediator exogeneity
Notes: Columns 1–4 and 7–8 ordinary least squares regression estimates for continuous outcome variables; Columns 5–6 and 9–10 probit regression estimates for
binary outcome variable. SE: Robust standard error. ∗∗∗, ∗∗, ∗ denote significance at the 1, 5, and 10 percent levels, respectively. †††, ††, † denote significance based
on FDR (Sharpened False Discovery Rate) q-values at the 1, 5, and 10 percent levels, respectively. [ ∼ :sharpened q-value 0.108]. ACME: Average causal mediation
effect; ADE: Average direct effect; ρ: Correlation between error terms ε3i and ε2i .
ARTICLE
The ACME through the cooperation index on the likelihood of using
improved seed for food crops is statistically significant and positive
(Column 9). The share of the total effect mediated through cooperation
is estimated at 18.17 percent. The ACME has a relatively high sensitivity to
the mediator exogeneity assumption based on ρ at which ACME would be
zero of 0.100 (Panel B Column 9; Figure I, Online Appendix E).14
Heterogeneous causal mediation effects of cooperation between spouses
on indicators household welfare and household public goods provision
There is evidence of a heterogeneous ACME by age of the wife, which
we use as a proxy for length of marriage (Table E, Online Appendix F).
In line with expectations that a longer duration of marriage is conducive
for household welfare because of accrued cooperation and harmony, the
ACME through spousal cooperation, measured by the cooperation index,
on total household income per capita when the wife is relatively old is
statistically significantly larger than when she is not.15
DISCUSSION AND CONCLUSION
A lack of cooperation within agricultural households in low- and middle-
income country contexts is hypothesized to reduce the efficiency of
smallholder household farming and households’ well-being. This article
contributes to a debate about the virtues of cooperation between the main
decision makers in agricultural households for household welfare and the
provision of household public goods such as investments in the household’s
future livelihood. It addresses the gap in robust evidence of the causal
effects of spousal cooperation on household welfare and public goods
provision and applies causal mediation analysis. This method addresses
the empirical challenges of endogeneity or confounded association of
cooperation and household outcomes, as well as the difficulty of directly
manipulating cooperation (Doss 2013).
To provide evidence of the causal mediation effect of spousal
cooperation on household welfare and household public goods provision
in agricultural households in Uganda and Tanzania, we rely on the
random encouragement to participate in a couples’ training intervention,
which is hypothesized to stimulate spousal cooperation, which, in turn, is
hypothesized to promote household welfare and public goods provision.
We measure spousal cooperation by a cooperation index capturing the
extent of jointness in income and asset ownership, agricultural decision
making and management, and an alternative cooperation index capturing
jointness in income and asset ownership which is not incompatible with
specialization as a way of cooperation.
289
INTRAHOUSEHOLD COOPERATION
We demonstrate that spousal cooperation as measured through the
cooperation indices are statistically significant positive mediators of the
effects of the couples’ training intervention on household welfare and
public goods provision. This supports our key research hypothesis H3. The
causal mediation effect of spousal cooperation amounts to over 10 percent
of the total effect on total household income per capita, approximately 45
percent of the total effect on the likelihood of household food security, over
25 percent of the total effect on adoption intensity of agronomic practices,
and almost 20 percent of total effect on the likelihood of using improved
seed for food crops. We find no supporting evidence for mediating effects
of spousal cooperation on the evolution in coffee income between base-
and endline.
There is evidence of heterogeneous causal mediation effects (Hypothesis
H4). Spousal cooperation has a larger mediation effect on income per
capita with higher age of the wife, proxying a longer duration of marriage.
We also demonstrate positive intention-to-treat (ITT) effects of the couples’
training intervention on the indicators of spousal cooperation (Hypothesis
H1), as well as positive (total) ITT effects on indicators of household
welfare and household public goods provision (Hypothesis H2).
The study’s estimates of the impact of spousal cooperation are
externally valid for self-identified monogamous (men–women) couples in
smallholder coffee farming households, who are member of producer
organizations linked to the Hanns R. Neumann Stiftung (HRNS) in
Mubende district in Uganda and Mbeya rural and Mbozi districts in
Tanzania, who self-selected into initial awareness-raising couple seminars.
Exploring the relation between cooperation and household welfare and
public goods provision in agricultural households with other farming
systems and in other contexts could provide answers to whether the
relationship between cooperation and household welfare and public
goods provision is particular among households in these localities, among
coffee farming households, or among households that self-selected into a
program promoting farming as a household enterprise and problematizing
prevailing gender relations.
Based on the evidence provided in this article, policy and future
programs that promote cooperation between spouses through gender-
transformative household methodologies can be recommended for
improving welfare and public goods provision in agricultural households.
Our study results also support the suggestion that stimulating both
intrahousehold cooperation and women’s empowerment can be beneficial
for household outcomes in agricultural households, who manage a lot of
common resources but whose members do not necessarily have equal voice
in decisions (Doss and Quisumbing 2020). Interventions that (additionally)
promote cooperation may also be more appropriate in case women prefer
investing their increased agency in strengthening the household unit
290
ARTICLE
rather than gaining independence (Kabeer 1999; Molyneux and Thomson
2011; Doss 2021). In that sense, a conceptualization of the empowerment
of women in their households that acknowledges that women can be
empowered not only by increased exclusive control, but also by shared
control over resources and agency, captures aspects of cooperation in the
household (Malapit et al. 2014; Hanmer and Klugman 2016; Johnson et al.
2016; Doss, Kieran, and Kilic 2017; Ambler et al. 2021). Qualitative research
conducted in the study areas in Uganda and Tanzania, for instance,
revealed that women see their involvement in strategic household decision
making and controlling key household resources, such as coffee income,
as empowering (Lecoutere and Chu 2021; Lecoutere and Wuyts 2021).
Women did not mention personal income and personally owned assets
as goals for their empowerment. Besides, in these contexts, opportunities
for women to acquire personal income and assets remain challenging and
limited. Finally, promoting intrahousehold cooperation alongside women’s
empowerment may avoid some of the unintentional adverse effects that
can follow from promoting women’s exclusive control over resources
and agency while disregarding the (intrahousehold) social relations in
which women are embedded, such as confirmation of disadvantageous
gender roles and potential backlash in the form of jealously, distrust,
husbands withdrawing from household responsibilities, or domestic
violence (Mayoux 2001; Molyneux and Thomson 2011; Cornwall
2016).
Els Lecoutere
Consultative Group on International Agricultural Research
CGIAR GENDER Platform
Nairobi, Kenya
email: e.lecoutere@cgiar.org
http://orcid.org/0000-0002-1025-742X
Bjorn Van Campenhout
International Food Policy Research Institute
Washington, DC, USA
KU Leuven - LICOS Centre for Institutions and Economic Performance
Leuven, Flanders, Belgium
email: B.VanCampenhout@cgiar.org
http://orcid.org/0000-0003-2404-7826
ACKNOWLEDGMENTS
We would like to thank anonymous reviewers and participants of the
Centre for the Study of African Economies (CSAE) Conference 2019 in
Oxford, the Environment and Production Technology Brownbag seminar
291
INTRAHOUSEHOLD COOPERATION
at the International Food Policy Research Institute (IFPRI) in Washington
DC, and the ECON@UA meeting at the University of Antwerp for useful
comments.
NOTES ON CONTRIBUTORS
Els Lecoutere is Senior Scientist at the CGIAR GENDER Platform. Her
research focuses on the way unequal access to resources and power
relations, including those shaped by gender, define the efficiency and
equity of outcomes within households and common pool resource
systems in the context of low- and middle-income countries. She also
assesses the impact of interventions or policy that aim to create more
equal opportunities. Her PhD in political science and background in
development economics and demography allows her to approach her
research in a multidisciplinary way, applying qualitative and quantitative
methods, including field and lab-in-the-field experiments.
Bjorn Van Campenhout is Research Fellow in the Development Strategy
and Governance Division at the International Food Policy Research
Institute (IFPRI) and Associate Research Fellow at LICOS Centre for
Institutions and Economic Performance of the KULeuven, Belgium. His
research evolves around smallholder agriculture, especially in East Africa.
He has published on food market integration and smallholder market
participation. More recently, he has started investigating the potential
of Information and Communication Technologies to make agricultural
extension more inclusive and effective.
FUNDING
This project has received funding from the European Union’s Horizon
2020 Research and Innovation Program under the Marie Sklodowska-Curie
[grant agreement No 702964]. The Hanns R. Neumann Stiftung funded
the interventions.
ETHICS APPROVAL
This study was granted ethical clearance by the Ethics Committee for the
Social Sciences and Humanities University of Antwerp (SHW_15_41), the
Uganda National Council for Science and Technology (A564), and the
Tanzania Commission for Science and Technology (2017-236-NA-2017-
249).
292
ARTICLE
NOTES
1 Household public goods are shared by or benefit all household members, such
as housing, investment in household production, and children’s education. The
consumption by one household member does not reduce other members’ benefits
derived from the goods (Doss 2013).
2 Pareto efficient, that is, impossible to improve for one person without making another
person worse off.
3 To avoid the risk of spillovers due to interaction between couples in the control
and treatment group, the control group couples in our study population are part
of randomly selected POs where the implementation of the intensive coaching
was postponed until after endline data collection. Correcting standard errors for
clustering by PO did not make a substantial difference as there are very few
observations per PO.
4 88 percent of households in our study population have at least one phone. If not
reachable by phone, the PO leader invited the couple.
5 The GHA’s projection of farming as a household enterprise does not promote
specialization whereby, for instance, one spouse takes full control over cash crop
farming and the other spouse over food production. Yet, depending on household
members’ preferences and production functions, some households may consider
specialization as the most efficient way of managing the provision of common
household farm resources and allocation of benefits (Doss 2021).
6 The summary index is a weighted mean of the different components in the index,
where each component is standardized using control group mean and control group
standard deviation. Weights are obtained from the inverse covariance matrix. We used
the Stata code developed by Cyrus Samii (2017) to construct the index.
7 By requiring husband and wife to agree on jointness, our indicators of joint decision
making are conservative and not subject to possible overestimation of joint decision
making based on women’s accounts (Acosta et al. 2019).
8 As a reference, average monthly household consumption expenditures (with average
household size of five) in 2016–17 in rural central Uganda amounted to 335600 UGX
(93 USD; UBOS 2018), and 350620 TSH (156 USD) in 2017–18 in Mbeya region
Tanzania (NBS 2019).
9 We did not collect the required data for constructing indicators of expenditures
on food and adult goods as in Duflo and Udry (2004) and McCarthy and Kilic
(2017).
10 In the model with evolution of coffee income as an outcome variable, the statistically
significant ITT effect of intensive coaching on the cooperation index is 0.604 (S.E.
0.127; N = 287) and 0.650 (S.E. 0.130; N = 313) on the alternative cooperation index.
The difference is due to missing data in the outcome variable.
11 ITT effects without covariates in Table D in Online Appendix D.
12 There are no obvious outliers for any of the (continuous) outcome indicators,
neither in the treatment nor control group that could have driven the relatively large
ITT effects. Nevertheless, we tested robustness of the IIT effect sizes to excluding
observations with the three highest values of income per capita, and the three lowest
values and five highest values of evolution in coffee income. Coffee prices fluctuated
slightly between base- and endline in Uganda and Tanzania but cannot be major
drivers of effects. Better and more complete reporting of coffee and other income
after treatment (rather than a real income effect) is not a likely reason for observed
effects since the extent of missing values in reports of coffee and other income is
not consistently lower or higher in treatment or control group, neither at base- nor
endline, and not among wives nor husbands.
293
INTRAHOUSEHOLD COOPERATION
13 The ACME through spousal cooperation measured by the cooperation index on
total household income per capita is not statistically significantly different from zero,
neither are the ACME through both cooperation indices on the evolution in coffee
income (Table 4 Panel A Columns 1, 3, and 4). While these could be null effects,
with our data, we cannot rule out the existence of effects smaller than the ex-post
MDES for the ACME which are relatively large (resp. 20.3, 55.2, and 80.6; mainly due
to noisy data).
14 The ACME through the alternative cooperation index on the use of improved seed
for food crops is not statistically significantly different from zero (Table 4, Panel A
Column 10). This could be a null effect but our data does not allow ruling out an
effect smaller than 0.039, the ex-post MDES for the ACME.
15 There is no evidence in support of heterogeneous causal mediation effects linked to
the education level of the spouses, the age difference, and the age of the husband.
Our data, however, may not have allowed detecting effects smaller than ex-post
MDES.
SUPPLEMENTAL DATA
Supplemental data for this article can be accessed online at https://doi.org
/10.1080/13545701.2022.2120206. Data, metadata, and code used in this
article can be accessed online at https://doi.org/10.17632/4N4CXKRGVZ.1.
REFERENCES
Acosta, Mariola, Margit van Wessel, Severine Van Bommel, Edidah Ampaire, Jennifer
Twyman, Laurence Jassogne, and Peter Feindt. 2019. “What Does It Mean to Make
a ‘Joint’ Decision? Unpacking Intra-Household Decision Making in Agriculture:
Implications for Policy and Practice.” Journal of Development Studies 56(6): 1–20.
Agarwal, Bina. 2000. “Conceptualising Environmental Collective Action: Why Gender
Matters.” Cambridge Journal of Economics 24(3): 283–310.
Agrawal, Arun. 2003. “Sustainable Governance of Common-Pool Resources: Context,
Methods, and Politics.” Annual Review of Anthropology 32(1): 243–62.
Akresh, Richard, Joyce J. Chen, and Charity T. Moore. 2016. “Altruism, Cooperation, and
Efficiency: Agricultural Production in Polygynous Households.” Economic Development
and Cultural Change 64(4): 661–96.
Alderman, Harold, John Hoddinott, Lawrence James Haddad, and Christopher Udry.
2003. “Gender Differentials in Farm Productivity: Implications for Household
Efficiency and Agricultural Policy.” In Household Decisions, Gender, and Development: A
Synthesis of Recent Research, edited by Agnes R. Quisumbing, 61–6. Washington, DC:
International Food Policy Research Institute.
Ambler, Kate, Cheryl Doss, Caitlin Kieran, and Simone Passarelli. 2021. “He Says,
She Says: Spousal Disagreement in Survey Measures of Bargaining Power.” Economic
Development and Cultural Change 69(2): 765–88.
Anderson, Michael. 2008. “Multiple Inference and Gender Differences in the Effects of
Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool, and Early
Training Projects.” Journal of the American Statistical Association 103(484): 1481–95.
Baland, Jean-Marie and Jean-Philippe Platteau. 1998. “Wealth Inequality and Efficiency
in the Commons, Part II: The Regulated Case.” Oxford Economic Papers 50(1): 1–22.
_____. 1999. The Ambiguous Impact of Inequality on Local Resource Management.”
World Development 27: 773–88.
294
ARTICLE
Bardhan, Pranab. 2000. “Irrigation and Cooperation: An Empirical Analysis of 48
Irrigation Communities in South India.” Economic Development and Cultural Change
48(4): 847–65.
Carpena, Fenella and Bilal Zia. 2020. “The Causal Mechanism of Financial Education:
Evidence from Mediation Analysis.” Journal of Economic Behavior & Organization 177:
143–84.
Cornwall, Andrea. 2016. “Women’s Empowerment: What Works?” Journal of International
Development 28(3): 342–59.
de Brauw, Alan, Patrick Eozenou, Daniel Gilligan, Christine Hotz, Neha Kumar, and J. V.
Meenakshi. 2018. “Biofortification, Crop Adoption and Health Information: Impact
Pathways in Mozambique and Uganda.” American Journal of Agricultural Economics
100(3): 906–30.
Doss, Cheryl. 2013. “Intrahousehold Bargaining and Resource Allocation in Developing
Countries.” World Bank Research Observer 28(1): 52–78.
_____. 2021. “Diffusion and Dilution: The Power and Perils of Integrating Feminist
Perspectives into Household Economics.” Feminist Economics 27(3): 1–20.
Doss, Cheryl, Caitlin Kieran, and Talip Kilic. 2017. Measuring Ownership, Control, and Use of
Assets. Policy Research Working Paper WPS 8146. Washington, DC: The World Bank.
Doss, Cheryl and Ruth Meinzen-Dick. 2015. “Collective Action Within the Household:
Insights from Natural Resource Management.” World Development 74: 171–83.
Doss, Cheryl and Agnes Quisumbing. 2020. “Understanding Rural Household Behavior:
Beyond Boserup and Becker.” Agricultural Economics 51(1): 47–58.
Duflo, Esther and Christopher Udry. 2004. “Intrahousehold Resource Allocation in
Cote D’Ivoire: Social Norms, Separate Accounts and Consumption Choices.” NBER
Working Paper No. 10498. National Bureau of Economic Research.
FAO, IFAD, and WFP. 2020. Gender Transformative Approaches for Food Security, Improved
Nutrition and Sustainable Agriculture. A Compendium of Fifteen Good Practices. Rome: FAO.
Farnworth, Cathy, Clare Stirling, Amon Chinyophiro, Andrew Namakhoma, and
Rebecca Morahan. 2018. “Exploring the Potential of Household Methodologies to
Strengthen Gender Equality and Improve Smallholder Livelihoods: Research in
Malawi in Maize-Based Systems.” Journal of Arid Environments 149: 53–61.
Fiala, Nathan and Xi He. 2017. “Unitary or Noncooperative Intrahousehold Model?
Evidence from Couples in Uganda.” World Bank Economic Review 30(Supplement 1):
S77–85.
Fisher, Monica and Edward Carr. 2015. “The Influence of Gendered Roles and
Responsibilities on the Adoption of Technologies That Mitigate Drought Risk: The
Case of Drought-Tolerant Maize Seed in Eastern Uganda.” Global Environmental Change
35: 82–92.
Gilligan, Daniel, Neha Kumar, Scott McNiven, J. V. Meenakshi, and Agnes Quisumbing.
2020. “Bargaining Power, Decision Making, and Biofortification: The Role of Gender
in Adoption of Orange Sweet Potato in Uganda.” Food Policy 95: 101909.
Hanmer, Lucia and Jeni Klugman. 2016. “Exploring Women’s Agency and
Empowerment in Developing Countries: Where Do We Stand?” Feminist Economics
22(1): 237–63.
Heckert, Jessica, Deanna Olney, and Marie Ruel. 2019. “Is Women’s Empowerment a
Pathway to Improving Child Nutrition Outcomes in a Nutrition-Sensitive Agriculture
Program?: Evidence from a Randomized Controlled Trial in Burkina Faso.” Social
Science & Medicine 233: 93–102.
Hoel, Jessica, Melissa Hidrobo, Tanguy Bernard, and Maha Ashour. 2017. Productive
Inefficiency in Dairy Farming and Cooperation Between Spouses: Evidence from Senegal. IFPRI
Discussion Paper 1698. Washington, DC: International Food Policy Research Institute.
295
INTRAHOUSEHOLD COOPERATION
Imai, Kosuke, Luke Keele, and Dustin Tingley. 2010. “A General Approach to Causal
Mediation Analysis.” Psychological Methods 15(4): 309–34.
Imai, Kosuke, Luke Keele, Dustin Tingley, and Teppei Yamamoto. 2011. “Unpacking the
Black Box of Causality: Learning About Causal Mechanisms from Experimental and
Observational Studies.” American Political Science Review 105(4): 765–89.
Iversen, Vegard, Cecile Jackson, Bereket Kebede, Alistair Munro, and Arjan Verschoor.
2011. “Do Spouses Realise Cooperative Gains? Experimental Evidence from Rural
Uganda.” World Development 39(4): 569–78.
Johnson, Nancy, Chiara Kovarik, Ruth Meinzen-Dick, Jemimah Njuki, and Agnes
Quisumbing. 2016. “Gender, Assets, and Agricultural Development: Lessons from
Eight Projects.” World Development 83: 295–311.
Kabeer, Naila. 1999. “Resources, Agency, Achievements: Reflections on the Measurement
of Women’s Empowerment.” Development and Change 30(3): 435–64.
Karimli, Leyla, Els Lecoutere, Christine Wells, and Leyla Ismayilova. 2021. “More Assets,
More Decision-Making Power? Mediation Model in a Cluster-Randomized Controlled
Trial Evaluating the Effect of the Graduation Program on Women’s Empowerment in
Burkina Faso.” World Development 137: 105159.
Lecoutere, Els and Lan Chu. 2021. Changing Intrahousehold Decision Making to Empower
Women in Their Households: A Mixed Methods Analysis of a Field Experiment in Rural South-
West Tanzania. IOB Working Paper 2021.06. Antwerp: Institute of Development Policy
(IOB).
Lecoutere, Els and Laurence Jassogne. 2019. “Fairness and Efficiency in Smallholder
Farming: The Relation with Intrahousehold Decision-Making.” Journal of Development
Studies 55(1): 57–82.
Lecoutere, Els and Eva Wuyts. 2021. “Confronting the Wall of Patriarchy: Does
Participatory Intrahousehold Decision Making Empower Women in Agricultural
Households?” Journal of Development Studies 57(6): 882–905.
Lundberg, Shelly and Robert Pollak. 1993. “Separate Spheres Bargaining and the
Marriage Market.” Journal of Political Economy 101(6): 88–1010.
Malapit, Hazel, Kathryn Sproule, Chiara Kovarik, Ruth Meinzen-Dick, Agnes
Quisumbing, Farzana Ramzan, Emily Hogue, and Sabina Alkire. 2014. Measuring
Progress Toward Empowerment: Women’s Empowerment in Agriculture Index: Baseline Report.
Washington, DC: International Food Policy Research Institute.
Mayoux, Linda. 2001. “Tackling the Down Side: Social Capital, Women’s Empowerment
and Micro-Finance in Cameroon.” Development and Change 32(3): 435–64.
_____. 2012. Balanced Trees Grow Richer Beans: Promoting Gender Justice through Value Chain
Development in Western Uganda. Oxfam Novib.
McCarthy, Nancy and Talip Kilic. 2017. Stronger Together: Intra-Household Cooperation and
Household Welfare in Malawi. Policy Research Working Paper WPS 8043. Washington,
DC: World Bank.
McKenzie, David. 2021. “An Updated Overview of Multiple Hypothesis Testing
Commands in Stata.” World Bank Blogs, July 20. https://blogs.worldbank.org/impacte
valuations/updated-overview-multiple-hypothesis-testing-commands-stata.
McKenzie, David and Owen Ozier. 2019. “Why Ex-Post Power Using Estimated
Effect Sizes Is Bad, But an Ex-Post MDE Is Not.” World Bank Blogs, May 16.
https://blogs.worldbank.org/impactevaluations/why-ex-post-power-using-estimated-
effect-sizes-bad-ex-post-mde-not.
Meier zu Selhausen, Felix. 2016. “What Determines Women’s Participation in Collective
Action? Evidence from a Western Ugandan Coffee Cooperative.” Feminist Economics
22(1): 130–57.
296
ARTICLE
Molyneux, Maxine and Marilyn Thomson. 2011. “Cash Transfers, Gender Equity and
Women’s Empowerment in Peru, Ecuador and Bolivia.” Gender and Development 19(2):
195–212.
Munro, Alistair. 2018. “Intra-Household Experiments: A Survey.” Journal of Economic
Surveys 32(1): 134–75.
NBS (National Bureau of Statistics). 2019. Tanzania Mainland Household Budget Survey
2017–18. Key Indicators Report. Dodoma: National Bureau of Statistics.
Njuki, Jemimah, Susan Kaaria, Angeline Chamunorwa, and Wanjiku Chiuri. 2011.
“Linking Smallholder Farmers to Markets, Gender and Intra-Household Dynamics:
Does the Choice of Commodity Matter?” European Journal of Development Research 23(3):
426–43.
Ostrom, Elinor. 1990. Governing the Commons: The Evolution of Institutions for Collective
Action. New York: Cambridge University Press.
Pearce, Neil and Jan Vandenbroucke. 2016. “Causation, Mediation and Explanation.”
International Journal of Epidemiology 45(6): 1915–22.
Quisumbing, Agnes and John Maluccio. 2003. “Resources at Marriage and
Intrahousehold Allocation: Evidence from Bangladesh, Ethiopia, Indonesia, and
South Africa.” Oxford Bulletin of Economics and Statistics 65(3): 283–327.
Quisumbing, Agnes, Kathryn Sproule, Elena Martinez, and Hazel Malapit. 2021. “Do
Tradeoffs Among Dimensions of Women’s Empowerment and Nutrition Outcomes
Exist? Evidence from Six Countries in Africa and Asia.” Food Policy 100: 102001.
Samii, Cyrus. 2017. “Make_index.” https://github.com/cdsamii/make_index/blob/mas
ter/stata/make_index_gr.do.
Santoso, Marianne, Rachel Bezner Kerr, John Hoddinott, Priya Garigipati, Sophia
Olmos, and Sera Young. 2019. “The Role of Women’s Empowerment in Child
Nutrition Outcomes: A Systematic Review.” Advances in Nutrition 10(6): 1138–51.
Seymour, Greg. 2017. “Women’s Empowerment in Agriculture: Implications for
Technical Efficiency in Rural Bangladesh.” Agricultural Economics 48(4): 513–22.
Slootmaker, Christopher. 2014. “Three Essays on Gender Inequality, Dynamic
Bargaining, and Technology Adoption in Subsistence Agriculture.” PhD diss.,
Colorado State University.
Sraboni, Esha, Hazel Malapit, Agnes Quisumbing, and Akhter Ahmed. 2014. “Women’s
Empowerment in Agriculture: What Role for Food Security in Bangladesh?” World
Development 61: 11–52.
UBOS (Uganda Bureau of Statistics). 2018. Uganda National Household Survey 2016/2017.
Kampala: Uganda Bureau of Statistics.
Udry, Christopher. 1996. “Gender, Agricultural Production, and the Theory of the
Household.” Journal of Political Economy 104(5): 1010–46.
VanderWeele, Tyler. 2016. “Explanation in Causal Inference: Developments in
Mediation and Interaction.” International Journal of Epidemiology 45(6): 1904–8.
Yamamoto, Teppei. 2016. “Unpacking the Black Box of Causality: Learning about
Causal Mechanisms from Experimental and Observational Studies.” In Tutorial at the
19th Labor Economics Conference Massachusetts Institute of Technology/Stanford
University, September 17. http://web.mit.edu/teppei/www/teaching/OsakaLabor2
016.pdf.
297