Sex differences in math-intensive fields

I did not actually set out to create a 'women in computer science' blog, and I do, in fact, do interesting technical work at my day job. But I can't talk about my day job, and people keep sending me interesting papers like this one (by Stephen Ceci and Wendy Williams). One sentence summary: the reason women are under-represented in math intensive fields is not due to biology or 'hard' discrimination (my term, not theirs) but because of 'choices' made by, and preferences held by, women as a whole. I put the work 'choice' in quotes because I get the feeling that they are using the word 'choice' in some technical sense, and not one that that implies 'conscious thought' or 'good options.' But let's don our asbestos blogging suit and delve into the paper.

The paper is structured around an a priori trichotomy: there are relatively few women in math-intensive fields because:

• Biological differences between men and women,
• Discrimination on the part of hiring committees, program committees, etc., or
• Choices and preferences on the part of women themselves.

Note that the third category, choices/preferences, is really just a catch-all that holds everything not in first two categories. Not only does it contain such things as 'maybe women would prefer to be lawyers, all other things being equal' but also such things as 'maybe women would prefer not to be sexually harassed by their colleagues, kthnxbye.' Although I see a huge difference between these two possibilities, they are both regarded as 'choices' or 'preferences' by this paper.

As frustrating as that is to me, however, this actually does make sense in context.¹ The point of this paper is to review lots of other studies/research through the prism of this trichotomy, and conclude that there is no evidence for the first two possibilities. Specifically:

• There do seem to be sex-related differences in some aspects of mathematical ability. In particular, men exhibit higher variance on standardized tests (like the GRE) and so will outnumber women on both ends of the bell curve. Even when you take this into account, however, it does not come close to explaining the number we actually see in real life: the bell-curve hypothesis would predict a 2:1 men-to-women ratio, and we actually see a 10:1 to 20:1 ratio among tenure-track professors. Oh-- and this effect disappears entirely when you look at grades instead of test scores.
• As for discrimination, the evidence seems to be mixed: there are a handful of studies showing discrimination on the part of admission/funding/hiring committees, but other studies which looked for the same phenomenon and found nothing. (It is probably worth mentioning here that I think the paper focused entirely on the United States.) And even if the first studies (the ones that did find discrimination) are true, the measured effects are still too small to explain the numbers we actually see in real life. (And yes, discrimination does explain historical under-representation, but doesn't seem to explain it any more.)

Okay, so what's left? Preferences and choices on the part of women... and here's where I start to get irritated at this paper. I'm actually glad to hear that the problem is not biological differences, because I can't do anything about biological differences. And I'm glad to hear that the problem is not due to (overt) discrimination on the part of institutions/committees, because such discrimination would be both wrong and hard to eradicate. And so I'm glad to learn that the under-representation of women is due to choices and preferences, because this means that I might be able to do something about it.

Okay, great. But what? Why are women choosing to leave or avoid math- intensive fields? What are the operational preferences here? Are women concluding that CS labs are 'male' environments? Are they choosing to avoid CS because they've concluded (falsely, I assert) that CS does not advance communal goals? What is going on?

Unfortunately, this particular paper isn't much help. The purpose of this paper was to eliminate 'biology' and 'discrimination', not to shed more light on 'choices'. It does mention a few possibilities, though:

• There may be sex-related differences on where men and women fall in 'people vs. things' interests.
• Women with high mathematical aptitudes tend also to have high verbal aptitudes, where the same is not true for men. Therefore, such women have more choices than such men.
• Women, more than men, sacrifice their careers for family reasons---particularly to care for children. (This has a wildly disproportionate effect on careers when there's a tenure-clock ticking.)

And here's where I get irritated again, but not necessarily at the paper this time. Remember, my goal is to increase the number of computer scientists in the world, and this paper tells me that we're missing 45% or more of the top- notch computer scientists we might have in a perfect world. Why? Well, it could be due to honest-to-God preferences on the part of women. Women prefer 'people' careers over 'thing' careers more than men do? Okay, so be it. We can do more to raise awareness of the (massive) social and communal benefits that CS still promises to produce for humanity, but CS is still fundamentally about things and not social relationships.

But for every 'legitimate' reason like this one I come across, I hear two or three possibilities that are solely about women being... not men. Women conclude that undergraduate CS labs are 'male' environments and than women are not welcome? Women give up tenure-track positions because the tenure clock doesn't stop for maternity leave? Women are told that math is for boys, and that they shouldn't display mathematical aptitude if they want to be cool in high school? These have nothing to do with what women want and everything to do with the environment surrounding mathematical fields. If we can fix these, then we'd get more top-notch computer scientists...

...I assert. And I'd love to have some numbers to back me up here. Unfortunately, I'm going to have to look elsewhere for them-- this paper doesn't have them. It does a great job of countering the 'biology' and 'discrimination' hypotheses, but it doesn't help me get any insight into the inner workings of the all-inclusive 'choice/preferences' category.