Sun 18 Sep 2011
Marc Bellemare at Duke has been using Delivering Development in his development seminar this semester. On Friday, he was kind enough to blog a bit about one of the things he found interesting in the book: the finding that women were more productive than men on a per-hectare basis. As Marc notes, this runs contrary to most assumptions in the agricultural/development economics literature, especially some rather famous work by Chris Udry:
Whereas one would expect men and women to be equally productive on their respective plots within the household, Udry finds that in Burkina Faso, men are more productive than women at the margin when controlling for a host of confounding factors.
This is an important finding, as it speaks to our understanding of inefficiency in household production . . . which, as you might imagine given Udry’s findings, is often assumed to be a problem of men farming too little and women farming a bit too much land. So Marc was a bit taken aback to read that in coastal Ghana the situation is actually reversed – women are more productive than men per unit area of land, and therefore to achieve optimal distributions of agricultural resources (read:land) in these households we would actually have to shift land out of men’s production into women’s production.
I knew that this finding ran contrary to Udry and some other folks, but I did not think it was that big a deal: Udry worked in the Sahel, which is quite a different environment and agroecology than coastal Ghana. Further, he worked with folks of a totally different ethnicity engaged with different markets. In short, I chalked his findings up to the convergence of any number of factors that had played out somewhat differently in my research context. I certainly don’t see my findings as generalizable much beyond Akan-speaking peoples living in rural parts of Ghana . . .
All of that said, Marc points out that with regard to my findings:
Of course, this would need to be subjected to the proper empirical specification and to a battery of statistical tests . . .
Well, that is an interesting question. So, a bit of transparency on my data (it is pretty transparent in my refereed pubs, but the book didn’t wade into all of that):
- The data was gathered during the main rainy season, typically as the harvest was just starting to come in. This required folks to make some degree of projection about the productivity of their fields at least a month into the future, and often several months into the future
- The income figures for each crop, and therefore for total agricultural productivity, were self-reported. I was not able to cross-check these reported figures by counting the actual amount of crop coming off each farm.
- I also gathered information on expenses, and when I totaled up expenses and subtracted them from reported income, every household in the village was running in the red. I know that is not true, having lived there for some 18 months of my life.
- There is no doubt in my mind that production figures were underestimated, and expenses overestimated, in my data – this fits into patterns of income reporting among the Akan that are seen elsewhere in the literature.
- Therefore, you cannot trust the reported figures as accurate absolute measures of farm productivity.
- The data was replicated across three field seasons. The first two field seasons, I conducted all data collection with my research assistant. However, in the final year of data collection, I lead a team of four interviewers from the University of Cape Coast, who worked with local guides to identify farms and farmers to interview – in the last year, we interviewed every willing farmer in the village (nearly 100% of the population).
- It turns out that my snowball sample of households in the first two years of data collection actually covered the entire universe of households operating under non-exceptional household circumstances (i.e. they are not samples, they are reports on the activities of the population).
- In other words, you don’t have to ask about my sampling – there was no sampling. I just described the activities of the entire relevant population in all three years.
- This removes a lot of concerns people have about the size of my samples – some household strategies only had 7 or 8 households working with them in a given year, which makes statistical work a little tricky Well, turns out there is no real need for stats, as this is everyone!
- The only exception to this: female-headed households. I grossly underinterviewed them in years 1 and 2 (inadvertently), and the women I did interview do not appear to be representative of all female-headed households. I therefore can only make very limited claims about trends in these households.
- Even with completely new interviewers who had no preconceived notions about the data, the income findings came in roughly the same as when I gathered the data. That’s replicability, folks! Well, at least as far as qualitative social science gets in a dynamic situation.
- Though the data was gathered at only one point in the season, at that point farmers were already seeing how the first wave of the harvest was doing and could make reasonable projections about the rest of the harvest.
I’m probably forgetting other problems and answers . . . Marc will remind me, I’m sure! In any case, though, Marc asks a really interesting question at the end of his post:
Assuming the finding holds, it would be interesting to compare the two countries given that Burkina Faso and Ghana share a border. Is the change in gender differences due to different institutions? Different crops?
The short answer, for now, has to be a really unsatisfying “I don’t know.” Delivering Development lays out in relatively simple terms a really complex argument I have building for some time about livelihoods, that they are motivated by and optimized with reference to a lot more than material outcomes. The book builds a fairly simple explanation for how men balanced the need to remain in charge of their households with the need to feed and shelter those households . . . but I have elaborated on this in a piece in review at the Development and Change. I will send them an email and figure out where this is in review – they have been struggling mightily with reviewers (last I heard, they had gone through 13!?!) and put up a preprint as soon as I am able. This is relevant here because I would need a lot more information about the Burkina setting to work through my new livelihoods framework before I could answer Marc’s question.