John Quiggin follows up on my post about occupational segregation. He says:

I’d argue that the bulk of the explanation can be found in high school or earlier. Girls do relatively well in language and boys in mathematics. Although I have no data, the disproportion seems to be higher the further up the performance scale you go, so that the very best students (the future PhDs) are highly gender-segregated. Given the incredible power of social pressures in high school, I think it’s reasonable to assume that this outcome is generated by social stereotypes rather than by differences in genes.

These differences are progressively amplified:

At the graduate level, lousy prose will be forgiven but inadequately formalised arguments will not … So the forces of comparative advantage encourage bright women to leave economics and move to fields where their skills are better rewarded.

This is a plausible mechanism. I should have put in a point in my original post about how deep-rooted these processes are and how they are constantly reproducing themselves over time. So high-school and before obviously matter a great deal for the reasons he describes.

This can’t explain everything, though. Two things stand out. First, we’re still left with the fine-grainedness of intra-disciplinary segregation. Even the lower-status fields in Economics (e.g., those that involve looking at data of any sort, hem hem) require a very high degree of competence in formal methods. Why should women be disproportionately found in these subfields? Second, as Brian Weatherson points out, there are technical fields where women are much better represented. Brian’s example of cognitive science is consistent with the idea that newer fields will tend to be less segregated.

All of this may be six of one, half dozen of the other. There’s no reason to think that there’s just one mechanism at work here. Early socialization and high-school sorting is the ground-bass of the process, but disciplines still bear a good deal of responsibility for the outcomes we observe, and can do a lot to correct them.