Carolina Castilla is an Assistant Professor of Economics at Colgate University.
“Soon we saw that money going to women brought much more benefit to the family than money going to the men. So we changed our policy and gave a high priority to women. As a result, now 96% of our four million borrowers in Grameen Bank are women.” tweet
-Muhammad Yunus tweet
The 2012 World Development Report highlighted that non-experimental evidence has repeatedly found women to allocate more resources, in terms of food, education, and health expenditures, to children than their male counterparts. These findings have greatly influenced policy through calls for increased targeting of women in social transfer programs. However, recent evidence from behavioral economics using randomized controlled trails contradicts these findings; instead finding that women contribute an equal or lesser amount than men to joint savings accounts and household investment opportunities, and that wives are less likely than husbands to give money to their spouses and to other community members in dictator games even when this behavior reduces overall household or community earnings.
I reconcile these divergent findings using a lab-in-the-field experimental design with spouses in Uttarakhand, India in 2013. Participant couples played a trust game in which each spouse was randomly assigned into sender and receiver role. Senders were given an endowment of Rs. 75 and decided how much to keep and how much to transfer to her or his partner. The transferred amount was tripled prior to reaching the receiver who then decided how much to keep and how much to return to his or her spouse. The details of the game were common knowledge to the couple.
The non-cooperative, individually optimal strategy of this game when played between strangers is for the receiver to keep the entire transfer and thus, in anticipation, the sender will transfer nothing. In contrast, the jointly optimal strategy that maximizes household earnings is for the sender to transfer the entire amount as the transfer earns 300% interest. In a household where spouses jointly make decisions over resource allocations, there is no incentive to transfer less than the full amount. On the other hand, non-cooperative households in which spouses do not share information, do not have sufficient bargaining power to attain their preferred allocations, or control money separately will transfer less than the maximum amount and the household will jointly earn less money, implying a loss in efficiency.
I find that senders only transfer 57% of the endowment, which is greater than the amount transferred when the trust game is played between strangers (here, here, here, and here), but is much lower than the household income maximizing action of transferring 100% to their spouse. Only 3% of senders transferred the entire amount. Most interestingly, when women were in the sender role they transferred less money, and also returned less money in the receiver role, than men. This behavior makes sense if women believe keeping the money is equally or less costly than transferring it to their husbands when they expect husbands to spend some (or all) of the money on items the wives do not support.
I examine this potential mechanism using survey data on individual expenditures collected following the conclusion of the experiments. I find a negative and significant relationship between tobacco expenditures by the husband and the proportion of the endowment sent by the wife, but no relationship with the proportion sent by the husband. If the costs of tobacco consumption are as large as the foregone return from not transferring the full endowment, then this behavior does not imply a loss in allocative efficiency. I examine this hypothesis by looking at returning behavior and find that husbands in households where the share of expenditure on tobacco exceeds the 85th percentile return less money to their wives.
These results suggest that women’s lower level of sharing in experiments is consistent with frictions in informal intra-household resource allocation contracts. The results of this paper, however, should be interpreted with caution as the sample was not drawn randomly from the population. Nonetheless, while the sample of spouses who participated in this experiment is not representative of the Indian population, my results suggest that we should examine the intra-household decision making environment more closely prior deciding whether to target women or men. Understanding how spouses interact with each other, how they manage money, and what drives that behavior is important to make sure development interventions (whether they involve cash transfers or something else) indeed have the intended effect on the outcomes of interest.