On paper, what this graph seems to communicate is fairly simple: women tend to go into lower IQ majors and/or women are stupid (hurr durr) - with a few caveats.
The first caveat is that the true correlation probably isn’t that high. Based on the project talent data, the correlation is only -.45, in comparison to the original graph’s correlation of -.76.
One potential issue is that men tend to outscore women on IQ tests, and this difference increases in college students because of a collider bias, where universities select for being female and intelligent, so being female and intelligent correlate more negatively than what you would expect in the general population. However, this cannot explain the whole relationship - I tried calculating the average IQs by major controlling for the effect of sex on IQ within college students, and the correlation only shrank by 25% (r = -.35, p = .03).
The sex difference within the whole dataset (not just the college students) was 1.95 points in favor of men, which is a little lower than the typical difference that is observed (~3 points), so I tried adding 1.05 IQ points to the men in the dataset, and the correlation decreased to -0.57.
So the sex difference in IQ clearly amplifies this relationship to a degree, but it cannot explain the whole picture (meaning I was wrong).
The most likely explanation I see is that women go into less prestigious majors, but not because they are less intelligent. Using the same data without adjusting the IQ scores at all, I find that women who go into the same major as the men tend to be more intelligent:
This only leaves the selection theory (women tend to go into lower IQ majors) left as a plausible explanation. The proportion of students of a major who are female and the average IQ of the men/women within that major correlates highly (-0.51 for women, -0.59 for men), which is stronger than the raw correlation observed between the average IQ of a major and the percent of women it has.
Notes: detail on how I calculate average IQs by college major are here
Edit and no. It’s not GMV either. Forcing men and women to the same distribution of intelligence does the same thing as adjusting for the mean. And there is GMV in intelligence within the project talent
One interesting question I long ago noticed on these charts is why is math such a sharp deviation for women?
If left to guess many would say medicine, I suspect.
Men are over-represented in highly cognitively demanding fields because they have a higher standard deviation...