gender roles and labor: path dependence

Alberto Alesina, Paolo Giulliano, and Nathan Nunn have a recent paper on the origins of gender roles.

They test the hypothesis that the cross-country variation in the cultural presence of gender norms today can be explained by looking at how different countries arranged their agricultural production prior to their Industrial Revolutions. Their framework is simple: one choice of technique did not require specific physical attributes, while the other was more capital-intensive and also required significant physical strength. Specifically, the latter technique was the plough, and given that the plough required "significant upper body strength", leading to the need for men to "specialize" in working the plough, the hypothesis is that societies that used the plough developed clear gender roles. (For agricultural tools with no biological need for specialization, women were also able to work in the fields and thus no specialization needed to occur.)

Their initial empirical strategy is to do a brute-force test of the hypothesis given country and ethnographic data on gender roles. On a crude first-pass through the data, they find a positive association between use of the plough (vs. the less-specialized and less-capital-intensive technique) and gender inequalities today. Given that this story is not really robust (gender discrimination most likely has its origins even further back than early industrialization, and indeed choice of technique could be endogenous to gender inequality -- i.e. gender inequalities prior to adoption of the plough may have influenced the economic trajectory of the region), they move on to a other possible tests of the hypothesis.

One other possible test they do is to use geographic factors as instrumental variables for the use of the plough. Ploughs are not appropriate for all types of crops, so the question becomes whether pseudo-random (based on geography) assignment of ploughs vs. other agricultural methods still ends up explaining gender inequalities. If so, then they can more confidently say that it was the plough itself, and not pre-existing gender norms favoring the adoption of the plough, which led to the gender inequalities we see today.

This second test of the thesis (i.e. the geographic origins thesis) is much more interesting and nuanced than the first. Doing so requires moving beyond OLS. Simple OLS estimates are not appropriate because plough use is most likely endogenous to existing social conditions (i.e., a more male-dominant society is probably highly correlated with the use of certain technologies and production methods). So what of these second (IV) estimates?

The authors begin by citing Engels' (1902) discussion of gender norms and agriculture. In Engels' discussion, according to the authors, intensification of agriculture led to the emergence of private property which, in turn, led to a male-dominated institutional regime overturning matriarchy. This is essentially how Engels argued it in Origin of the Family, Private Property, and the State. Intensified agriculture led to an agricultural surplus, leading to the need for market relations and private property, which gave rise to male power. Thus for Engels, it is not the capital-intensive or the use of specific physical exertion in agricultural technique which leads to male domination -- rather, imposing private property for accumulation of a surplus creates unequal power in a society.

The authors address Engels' alternative argument by including a set of controls which are intended to consider the possibility that historically patri-local marriage rules are the real driver of the variation. In the OLS estimates these "patriarchy varialbes" are not found to have an impact on the statistical significance of the main variable (plough use) -- though the estimate on plough use does drop by roughly 30% with the inclusion of these controls, highlighting that they may have some explanatory power. In other words, it seems that Engels' theory is validated, at least partially, in this analysis -- and even to the extent that the authors want to throw out Engels' theory, doing so is not justified simply through the use of OLS.

The more important point is, of course, the endogeneity issue: won't plough use be correlated with gender norms that exist prior to its adoption? The geographical IV is employed to solve the issue -- seeing whether geographies more conducive to the plough give rise to gender roles more strongly than geographies not conducive to heavy plough use. But there are problems with this strategy too: plough use according to the tables (pp. 35-36) is highly correlated with industrialized countries today. In other words, I am sure that plough use is a much better predictor for modern development than it is for gender inequalities. Still, the authors find a statistically significant impact of geography on cultural beliefs, showing some evidence for their hypothesis.

Overall, it is the presence of the significant correlation between plough use and modern economic development which sticks strongest with me as a key weakness in the authors' study. The main hypothesis is just too simple -- that the need for specialization in agriculture in particular is what led to the formation of gender inequality. Perhaps the authors simply lose a bit too much by going on such a macro-level scale.
And path dependence, as I've mentioned before, is not the most rigorous of historical methods. More about that in the next post.