Consider a very simple general equilibrium 2X2 example to illustrate my point:
Divide the population into two groups: call them "men" and "women". Divide all outcomes into two groups: call them "arts" and "science". Everybody is somewhere in the 2X2 matrix.
Here are four questions we could ask of the data:
1. Why are women under-represented in science?
2. Why are women over-represented in arts?
3. Why are men over-represented in science?
4. Why are men under-represented in arts?
Think up an answer for each of those 4 questions.
If you gave four different answers you got it wrong. If you gave two or more different answers you got it wrong. You got screwed by the framing effects. All four questions are logically equivalent. There is only one question being asked here. An answer to any one question is an answer to all four questions. WS/MS<WA/MA logically entails, and is logically entailed by, all three other inequalities.
Why are women under-represented in science? Maybe it's because arts are an especially hostile environment for men?
When boys and men do worse in school, and at university, to adopt the feminist partial equilibrium framing, and thus prejudice the question in favour of a feminist answer, is bigotry. "Why are women under-represented in economics?" is not the right way to ask the question. Maybe it's because men with low grades in economics have nowhere else to go, so stick with economics despite their low grades. (It's good to see Frances thinking general equilibrium.)
[And I wonder if Catherine Rampell "No, not because you might intimidate easily emasculated future husbands." has ever heard of female hypergamy? Maybe us men aren't always quite as stupid as she thinks we are?]
[Update: BTW, my point here is closely related to the Raven paradox, which econometricians, or anyone interested in Bayesian inference, should take a look at, if they haven't already.]