Conventional wisdom tends to run towards giving it to the woman. But a recent paper (ungated) by Laurens Cherchye, Pierre-Andre Chiappori, Bram De Rock, Charlotte Ringdal and Frederic Vermeulen tells us it might not matter (at least in terms of nutrition).
Let’s start with some background. There is early work by folks like Duncan Thomas in the early 1990s that suggested favoring the woman as a beneficiary. Fast forward to 2016 and Akresh and co-authors randomized cash transfers to men versus women in Burkina Faso. Their answer: it depends on the outcome you are focused on and the context (more nuance in the paper and this nice post by Damien de Walque). And Kathleen blogged about this issue here. Altogether not an unambiguous “always give it to mothers.”
Then, last year, a paper came out by Armand and co authors looking at transfers to mothers versus fathers in Macedonia. They find that transfers to women give you significantly higher expenditure on food, and a shift to more nutritious food among the poorer households.
This new work takes a different approach to this question. Given that the authors include some of the top thinkers on household models, they take us away from programs in the field to a lab setting in Kenya. This is going to enable them to give us more insight into the process that might be going on.
Cherchye and co. recruit 424 couples with children between 6 and 12 from two of Nairobi’s larger informal settlements. A large fraction of folks living in these settlements are quite poor and rates of child malnutrition are high.
Individual men and women are asked to choose 11 times, with different budgets and then amongst 3 to 8 bundles of cash for themselves and (nutritious) meals for one of their kids. The identified child is chosen at random when they have more than one kid. This gives Cherchye and co. price variation. Overall, the budgets range from about $2 to around $9.
After they make their individual decisions, the couple gets together and makes the same 11 decisions together where now they have to divide the cash between the two of them (and choose that versus meals for a kid). One decision per couple is randomly chosen for payout but to protect individual choices, there is a 10% chance the computer just chooses a random payout.
Cherchye and co. complement the game with a survey that captures income, household decision making, time-use and family networks.
So what happens?
In the simple descriptives, men allocate 37 percent of the experimental pot to meals for their child and women allocate 32 percent. Later in the paper (after some modeling, more on that in a minute) Cherchye and co. also look at this through the lens of Lindahl prices and they find no significant difference between men and women. Their bottom line: “these results suggest that women do not have a stronger preference than men do; rather they either have similar preferences or men have stronger preferences for children’s meals.” (Italics in the original)
Now, one thing that Cherchye and co. stress throughout their results is the strong degree of heterogeneity. And so here they look at what variables predict the wife or husband allocating more to the child. There are a bunch of variables here, but one that stood out to me was that the probability the wife allocates more to the child is decreasing when the chosen child is a girl. On the other hand, the probability that the husband allocates more to the child is increasing when the chosen child is a girl.
When husband and wives get together to make the decision, 35% gets allocated to child meals (in the simple descriptive stats). And, amongst the cash, the wife averages 36% and the husband 30% (statistically significantly different). Given my priors on bargaining power (e.g. women in this sample have earnings about half of those of men), this was a surprise.
This is where the modeling work comes in. Cherchye and co. work the theory and related tests and the first result is that when look at sole decision making, individuals are acting rationally (a lot of the time). What about as couples? Here Cherchye and co. set about testing the cooperative model of the household and come down in its favor – couples are acting efficiently, and there seems to be a better fit to the collective model than the unitary model.
Cherchye and co. also model and estimate the bargaining power, with the relative sharing rule indicating the on average women receive 53 percent of the couple’s expenditure and men coming in at 47 percent. How does this experimental measure line up with other measures that we commonly pull from survey data? There is a positive correlation with the number of decisions in which the wife says she has a say. However, there is no significant correlation with her income share. Some food for thought here.
So this is a really interesting paper. It gives us further food for thought in what is a nascent literature about how income coming in may be handled differently or not depending on the sex of the recipient. And, intriguingly, we get to see the diversity of households behind the average within a single setting – as Cherchye and co. note “while the average sharing rule sits around ½ (indicating equal sharing) for all decisions, the standard deviation is about ¼ in all cases, suggesting that, in a number of couples the allocation is grossly unbalanced.” Understanding this diversity, and how to effectively design policy in the face of it, is exciting.
Join the Conversation