There has been a rapid expansion of safety net programs in response to COVID-19. Countries continue to introduce new programs and expand existing ones. Two weeks ago, Berk blogged about the challenge of targeting households for such programs (in general, as well as, quickly in the case of emergencies or pandemics). There is an additional layer to targeting. Once a household is identified to receive assistance, there is still a decision about which household member(s) will receive the cash or other forms of assistance. Intrahousehold dynamics then come on the scene.
We have moved well beyond the days where we don’t consider how who gets the benefits matters in terms of impact on the household or among household members. Many cash transfer programs target a female adult in the household either because of her greater vulnerability or in an instrumental way towards another objective (or both). Specifically, in the latter case, program design is often based on an assumption about how the program will impact outcomes for children differentially if the mother (woman) receives the assistance than the father (man). I frequently read studies or program design documents attesting to the evidence that giving the cash to the woman will tilt spending in ways that benefit children -- such as education spending – to support this design.
One frequently-cited study on this is Thomas (1990) from Brazil, which finds that nonearned income in the hands of mothers (compared to fathers) significantly improves child health outcomes. The paper assumes exogeneity of nonearned income, though noting that this income reflects past endogenous labor supply decisions -- as well as other endogenous income sources.
Thirty years later, and with the large number of impact evaluations of cash transfers (RCTs and others), what additional evidence on this do we have? In their detailed review of evidence of program impacts, Bastagli et al (2016) find “No conclusive evidence… on the role of transferring the cash to women versus men as main recipients.” Studies of cash transfers finding no difference in impacts by sex of the recipient have been done in Morocco (Benhassine et al), in Kenya (Haushofer and Shapira), and in Burkina Faso (Akresh et al).
Ringdal and Sjursen also examine this, though not in the form of an impact evaluation. Their work tries to differentiate between two explanations for findings that it makes no difference who gets the cash: that cash transfers are not large enough (to improve women’s bargaining power and shift spending towards kids) or that men and women have similar preferences about spending on their children. They conduct a between-subject lab experiment with 287 couples in Dar es Salaam.
Couples are given three incentivized tasks, illustrated with laminated pictures of 500 and 1000 Tanzanian shilling (TZS) notes (so not requiring literacy). In the distributive task, each participant allocates a TZS15,000 endowment between his/herself, his/her spouse, and one randomly selected (and named) child. Time and risk preferences are elicited using two separate tasks (see details on these in the paper). The participants were paid for the tasks in the experiment. In the distributive task, the payout for spending on the child comes in the form of a tuition certificate, on the basis of a pre-experiment survey of 120 households where both men and women expressed demand for tutoring for their children. Other payouts are in cash.
There are four treatment areas in the distributive task: husband/wife and dictator/bargaining. A randomly selected dictator decides how to allocate the TZS15,000, placing the laminated notes into cups. Cups are shown to the dictator’s spouse in a separate room. The bargaining treatment is a Rubinstein shrinking-pie game. In this treatment, the first proposer makes a proposal on allocation which is shown the spouse in a different room. If the spouse does not agree, s/he makes a counter-proposal but the endowment is reduced by TZS500. This continues until an agreement is reached or there is no money left. The four treatments are designed to exogenously increase the wife’s bargaining power: Husband Dictator ≤ Husband Bargaining ≤ Wife Bargaining < Wife Dictator.
They find that a large increase in the wife’s bargaining power (from husband as dictator to the wife) increases the share allocated to her and decreases the share allocated to her husband.
Increasing the wife’s bargaining power does not increase the amount allocated to the child’s education. Moreover, an intermediate increase in the wife’s bargaining power (from husband dictator to wife as lead bargainer) significantly reduces the share allocated to the child’s education. This nonmonotonic change is consistent with a non-cooperative framework. One explanation is that the wife as bargainer wants to avoid that the husband rejects her proposal. And she may underestimate her husband’s preferences for investments in the child, giving more to him (less to the child) than she would as dictator. This is consistent with studies showing that spouses have inaccurate beliefs about each other’s preferences.
The effect of the wife’s bargaining power is different depending on the sex of the child. While a large increase in the wife’s bargaining power reduces the allocation for couples where the child is male, it increases the allocation for couples where the child is female. But these effects are not statistically significant. So bargaining power to the wife may impact gender equality between children, if not levels of investment on average.
Differences in patience matters. When the wife is as or more patient than her husband, a large increase in bargaining power increases the share allocated to the child. But when he is more patient, a shift in bargaining power toward the wife decreases the allocation. The intrahousehold difference in risk preferences makes no difference to the effect of a large increase in the wife’s bargaining power.
There are important caveats to these results. Among them, it is possible that wives would spend a larger share of the money they allocated to themselves on their children compared to husbands in real life, which they can’t explore in this study. Second, wives with low bargaining power in real life might over-allocate to the child in expectation of the husband undoing the cash allocation after the experiment, since the tutoring certificate is harder to un-do than cash. Robustness checks based on survey measures of decision-making reporting and employment status, however, are not consistent with this alternative explanation.
The RCTs mentioned above and this lab experiment find that it does not matter whether the wife or husband gets the cash in terms of child education inputs or outcomes (sometimes for child health too). Cash transfers that target women, especially those conditional on school attendance or health program participation of children, can add to the many burdens of poor mothers. And they can reinforce traditional gender roles in the household. Programs assume a benefit of such targeting for children (for which there is scant evidence), but it is not clear they also consider these potential negative consequences for the mother. Of course, the perceived benefit to children is not the only reason to give the cash to women instead of men in a beneficiary household. And it may not be the main reason. This lab experiment in Tanzania clearly shows that bargaining power matters for the allocation of cash between the husband and the wife. Extra cash in the hands of women might further empower women in their households. Though the mixed evidence of this could be the topic of another blog.
Join the Conversation