STEM: men go to Mars, women go to Venus

Tim Hunt, the Nobel prizewinning UCL biologist recently chased out of his professorship by a baying mob for joking that women scientists cause problems by falling in love with their male counterparts and crying if you criticise them, was recently asked if he thought the relative dearth of women in harder sciences was a problem. He said, in the purest crimethink:

I’m not sure there is really a problem, actually. People just look at the statistics. I dare, myself, think there is any discrimination, either for or against men or women. I think people are really good at selecting good scientists but I must admit the inequalities in the outcomes, especially at the higher end, are quite staggering. And I have no idea what the reasons are. One should start asking why women being under-represented in senior positions is such a big problem. Is this actually a bad thing? It is not immediately obvious for me … is this bad for women? Or bad for science? Or bad for society? I don’t know, it clearly upsets people a lot.

Although not directly responding to Hunt, prominent Spanish language website Politikon has a piece up (kindly translated for me) pre-emptively denying that the sorts of relevant sex differences that might cause these differences ‘legitimately’ exist.

Of course everyone accepts that there are huge differences between men and women in some domains. For example no one thinks than men’s thicker jaws or higher basal metabolic rates are socially constructed. No one thinks the fact that even athletically-trained women are much weaker than normal men is down to society.

But some people, including author Guido Corradi, do think that social construction is responsible for men and boys being judged better at mathematical subjects. He attacks Simon Baron-Cohen as a main progenitor of this view, and suggests the perspective is speculative and lacking evidentiary backing. He accepts that men are stronger at visuospatial skills (e.g. 3Dmentalrotation), but not that they are stronger mathematically overall.

More recent studies (Lindberg, et al. 2010) support the hypothesis that there are no mathematical skill differences. It has to be mentioned that since they started to be tracked, differences in general mathematical achievement have been decreasing. In a seminal meta-analysis by Hyde (1990) this tendency is observed.

Lindberg et al. do seem to convincingly show us that girls and boys are equally good at maths on average. But this doesn’t mean that things are the same the whole way along the scale, because men may differ more widely than women. Corradi appears to know that this possibility exists, but completely dismisses the point without considering it seriously.

Lindberg et al. find a small variability ratio, of 1.08, but other studies suggest this is still enough for a substantial gap at the top end. Johnson et al. (2008) at the highest level of mental ability, there tends to be a ratio of two men to each woman. Deary et al. (2007) find, in a sibling study to control for genes and environment, that when you get to the top 2%, there are also about two times as many men as women.

We can see how this opens up a wedge when you start selecting particularly talented groups, e.g. SAT-takers:

And it widens by the time you get to GRE:

This explains part of the different attainment between men and women in Science Technology Engineering and Mathematics, but ratios at the level of tenure track positions are somewhere in the range of 7:1 to 14:1, leaving a lot left over. Is this down to discrimination? Stereotyping? Social construction? Different preferences?

One large part of the gap is down to the distribution of skills. Women who have high mathematical skills are more likely than men to also have high verbal skills, opening up a number of extra options the men at that level don’t have. Those with high verbal skills tend to take these options. This explains a fraction of the remaining gap.

On the other hand stereotype threat, which Corradi alludes to, is much in vogue. I myself, I must admit, promoted one of the studies suggesting that male-female mathematics differences could be down to stereotyping. But it doesn’t seem like these results have held up over replication (e.g.this meta-analysismoremoremore).

By contrast, men and women do seem to have starkly different preferences about how their lives should go. For example, women tend to like different kinds of relationships (one-on-one 'dyadic’ pairings vs. gregarious multipolar groupings), and they tend to do more child-rearing.

Women (even the most talented women) tend to want to work less and more flexibly; neither of which fit with the long blocks of hours expected at the top of STEM professions or in STEM academia. Goldin (2014) explains how this leads to no gender wage gaps in industries with constant returns to hours, and large ones in industries where 60 hours work in a week is more than double as productive as 30 hours.

Su & Rounds (2015) review 52 samples between 1964 and 2007, including209,810 male and 223,268 female respondents and find large differences in interests.

We found gender differences in interests to vary largely by STEM field, with the largest gender differences in interests favoring men observed in engineering disciplines (d = 0.83–1.21), and in contrast, gender differences in interests favoring women in social sciences and medical services (d = −0.33 and −0.40, respectively). Importantly, the gender composition (percentages of women) in STEM fields reflects these gender differences in interests.

Overall the evidence seems to tell us that though men and women are equally smart, men are more prominent on both tails: they are more likely to be very dull and very bright. This variance isn’t huge overall, but when you start selecting for the top 0.01% or the top 0.0001%, like Fields Medallists, Nobel Prizewinners, or Harvard Professors the differences become overwhelming. The women who do have these incredible quantitative skills often also have excellent verbal skills, giving them alternatives they prefer.

While there may be residual discrimination, there is substantial evidence that on top of differing variance and skill distribution, men and women also have different preferences. Women tend to prefer to do less hours and focus more on the other important things in life. Men want to compete, earn lots of money, and work with objects.

Corradi makes a rash and unwarranted leap: there is good evidence for multifarious sex differences—not just in cognitive ability but in interests and preferences—that make complete and exact similarity between men and women in STEM a mirage.