OK, let’s put two blog posts in a pot and stir. In a previous post on measuring consumption, Jed gave us some food for thought, while over on Aid Thoughts, Matt is talking about how a respondent is seeing the enumerator on the sly to conceal land that he doesn’t want his wife to know about. Put it together, and what do you have? Spouses who lie to one another about what they have and what they spend.
So to start with, they don’t know about each other’s spending habits. In a “long-run” working paper that I am working on with Michael Boozer and Tavneet Suri, we look at data from Ghana where Chris Udry and I asked expenditure by asking each individual within a couple to report on their own expenditure, but also that of their spouse. This gives you three potential measures of consumption: what he says they spent, what she says they spent, and the sum of what she says she spent plus what he says he spent (it was actually much less confusing for the respondents for once). The upshot: there is a set of expenditure such as meals taken outside of the household, alcohol and the like that the spouse knows next to nothing about. Among other things, this has implications for how we measure poverty – the poverty rate using one individual’s report of household expenditure are significantly higher than using the combination of their own reports of their own expenditure (and there is little to no double counting). The usual approach to measuring consumption by asking “who does most of the shopping?” doesn’t help us here – because they don’t know much at all about the private expenditure, there are a fair number of couples in which they both think they are the major shopper.
OK, so why do they do this – this is where things get really interesting. Some newer work is trying to answer this by getting at how private information influences intra-household allocation. Nava Ashraf’s paper on the Philippines finds that when a husband can conceal money, he puts it in his own account. When it’s public, he spends it. And communication can change things. One critical, interesting twist on her results are that the underlying gender roles matter – the treatment effects are more pronounced in households where the woman is in charge of savings decisions. Nava also has an interesting paper with Erica Field and Jean Lee where they show that family planning take up among women is much higher if they can do so without their husband being present and a reduction in unwanted births occur only when the woman access family planning in private (in Zambia). Here is Berk's longer discussion of this paper.
Some other interesting new work in this vein is starting to come out from a large cross-country data set of experiments by a multi-disciplinary team at the University of East Anglia. They too are experimenting with information – in the context of a set of games. Some initial results from Ethiopia are in a recent paper. First, expectations seem to be off – husbands expect wives to contribute more than they do, and wives expect husbands to contribute less than they actually do. Second, information matters somewhat – in some games it alters the outcome (at a 10% level of significance) while in others it doesn’t matter. So this work suggests that the structure of the interaction also may matter for the role of private information.
So we’re seeing some interesting, and deep scratches into the surface of this. But this work suggests that while private information between spouses can be a measurement issue (with attendant implications for how we do surveys), the underlying behavioral reasons for this are quite important for policy. Thoughts? Other examples?