During the past several years, we have been working intently on a field experiment with smallholder farmers in Uganda randomly assigned to get access to information in the form of short engaging videos. Because women and men are both involved in farming in the typical Ugandan farm household, we took a particular interest in the gender dimensions of our experiment.
We have a heap of data to work with. After spending months combing through it, we wondered whether our emphasis on gender was misplaced. After all, responses to our survey questions posed to male household heads clearly indicated that women have equal voice in most strategic decisions made in the household, while men contribute to reproductive actives and household chores equally. It all looks like the optimal unitary household a la Becker (1991) .
But the moment we dig into survey responses from women within the household, and when we talk to women in private focus group discussions, our image of the perfect household quickly vanishes: Women respond that they have little say in agricultural decision making, yet provide the bulk of the work, both on the field and in the household.
This kind of disagreement between what husbands and wives claim is well known among social scientists, and specifically among researchers working on women’s empowerment and intra-household power dynamics. But our observations do draw attention to a specific element of the problem: Using proxies to report on other household members’ choices and actions. The problem is particularly acute when husbands are interviewed to obtain information on their wives, although discord can appear along other dimensions as well, such as age or other observable and even non-observable individual characteristics.
But if male and female co-heads provide widely differing answers to the same questions, who should we believe? What is the cause of this discord? Is it just measurement error, or is there also signal in the noise? Does this reflect differences in aspirations? Can spousal disagreement tell us something about power relations within the household? And is disagreement correlated with welfare outcomes?
In a recent paper, we explore two potential explanations for disagreement with our detailed sex-disaggregated data on decision-making, labor-time allocations, and marketing choices among maize-farming households in Uganda.
First, we examine whether spouses strategically withhold information from each other. There is both theory and evidence to support this hypothesis: In non-cooperative bargaining models of the household, members have their own preferences which may not always align, and so spouses may hide decisions, activities, and actions from each other. For example, they may decide to use a lower seed rate than what the other spouse thinks is appropriate; they may work less than what the other thinks; or they may sell part of the harvest behind the other spouse’s back.
Second, we test if disagreement between spouses is perhaps due to the fact that spouses are biased by traditional gender norms; that is, they respond in line with what they think society expects from them. In particular, as maize cultivation is culturally seen as an activity largely in the male domain, men are likely to exaggerate their own contributions relatively more and play down the role of the women, leading to differences in answer patterns between male and female co-heads.
We test both explanations for spousal disagreement through a field experiment. Central to this field experiment is a short engaging video shown to households.
To answer the first question, in one random subset of our sample of households, a video is shown to only the (male) co-head within the household. This is the control group, and corresponds to “business as usual,” where the (male) co-head within the household holds a monopoly over knowledge. In the treatment group, the same video is shown, but now to both spouses together, to make sure both spouses start off with the same information.
The reasoning behind this treatment is this: All else being equal, if the video reduces differences in information between spouses, then it should increase their ability to monitor each other, which in turn should reduce over-extraction from the household’s common pool resources—in this case, its skills, assets, and resources allocated to maize cultivation.
In a second treatment, a (control) group of households is shown a video in which all maize-related activities are performed by a man only, reinforcing the prevailing view in society that maize cultivation is a largely male activity, whereas women are relegated to crops such as cassava and sweet potato. In the treatment group, the same video is shown, with the only difference being that now the male and female co-head perform all maize cultivation activities together as a couple. The idea is that the household in the video serves as a role model, leading spouses to respond more in line with reality, and so men may be more likely to give credit to women where credit is due.
When asking about who makes decisions, we find substantial disagreement between spouses. For instance, we find that co-heads exaggerate their own decision-making power on almost one in four maize plots of land that they cultivate. However, we also find that promoting a cooperative approach to maize farming reduces the likelihood that co-heads overestimate their own influence on decision-making. This suggests spouses respond in line with what is expected from them.
We find that co-heads seem to overstate their own individual labor contributions. We further observe that, to some extent, a reduction in information asymmetries between spouses tends to reduce the likelihood that a co-head overstates his or her own labor contribution, indicating an increase in monitoring capacity within the household.
We also find that reducing information asymmetries between spouses reduces discord about who makes key decisions related to maize cultivation. This suggests that spouses also hide certain decisions from each other.
Overall, our research suggests multiple reasons for spousal disagreement in survey responses. Measurement error is clearly still a reason, as is the influence of customary and culturally determined gender roles. But some of it may also be driven by information-hiding behavior.
This leads us to the somewhat sobering conclusion that sometimes, sex-disaggregated data may be less useful than many assume. Indeed, without knowing what happened in reality, sex-disaggregated answers to the same questions amount to little more than perceptions, and it seems difficult to really know whose response is correct and whose is not.
This should not discourage us from collecting sex-disaggregated data; rather, it should encourage us to do more to understand the nature of discord in survey response between members of the household and potential solutions to it. After all, the reason why we collect sex-disaggregated data is precisely because we are investigating dynamics that are not easily observed.
Our research reminds us to not take survey data at face value, as bias may arise for a variety of reasons, and gender norms are only one potential source of bias. Examples of other factors that have been shown to lead to particular patterns in the data include response fatigue or length of recall period. Survey data should thus always be interpreted with a solid understanding of how it was collected and the cultural context. If possible, gender disaggregated data should be used alongside data that was collected though joint interviews of spouses. Innovative data collection methods, such as the use of vignettes or list experiments for particularly sensitive decisions or actions may also help in uncovering intra-household dynamics.
Bjorn Van Campenhout is a Research Fellow with IFPRI’s Development Strategy and Governance Division; Els Lecoutere is the Science Officer of the CGIAR GENDER Platform; David Spielman is a Senior Research Fellow and Leader of IFPRI’s Rwanda Strategy Support Program.
This research was supported by Digital Green and the U.S. Agency for International Development (USAID) under Feed the Future’s Developing Local Extension Capacity (DLEC) project; the IFPRI-led CGIAR Research Program on Policies, Institutions, and Markets (PIM); the CGIAR Research Program on Maize (MAIZE), led by the International Maize and Wheat Improvement Center (CIMMYT); and the International Development Research Centre, Ottawa, Canada, through the CGIAR Collaborative Platform for Gender Research.
This post first appeared on the EnGendering Data blog, a forum for researchers, policymakers, and development practitioners to pose questions, engage in discussions, and share resources about promising practices in collecting and analyzing sex-disaggregated data on agriculture and food security. If you are interested in writing for EnGendering Data, please contact blog editor Katrina Kosec.
