The collapsing math genius gender gap

Via Allison Schrager, Andy McAfee, and others, here comes a chart of the gender gap in upper-tail SAT scores:

As you can see, the ratio has drifted downward at all levels, but at the 0.01% level (about 3.7 standard deviations of the combined score distribution, if you assume normality) it has fairly dramatically collapsed. When Ronald Reagan was elected, girls made up only 7% of the top performers. By the time Clinton took office, it was 20%. As of Obama's reelection, it was 28%.

The key piece of information here isn't the current level of the gender gap, it's the recent rate of change. This sort of rapid shift implies that there's lots going on here that's not genetic - unless biological gender differences can change radically from decade to decade.

It's highly unlikely to be a change in the SAT math section. The SAT math section is the same damn thing year after year after year. "Do you know the ratio of sides of a 30-60-90 triangle?" "Can you factor a quadratic expression, using only a paper clip and bubblegum?" "What was the color of Napoleon's white horse?" And so on.*

If it ain't biology, and it ain't a measurement problem, it must be environmental forces that are causing the gender gap to collapse. And whatever these environmental forces are, they're strong stuff.

Knowing that strong environmental forces are present doesn't rule out inherited ability differences. It could be that there are two forces at work, and that the ratio will eventually settle into its "natural," biologically determined perfectly meritocratic level. But Occam's Razor - or really, any penalty term for model complexity - says that a one-factor model beats a two-factor, all else equal. In other words, if strong environmental forces were still present in 2006, it's a decent bet that they weren't completely gone as of 2011. So this doesn't kill the genetic-determinist position, but it weakens it.

OK, so who cares? Well, this is relevant to a recent Journal of Economic Perspectives article by Amanda Bayer and Cecilia Rouse, entitled "Diversity in the Economics Profession: A New Attack on an Old Problem". Bayer and Rouse show that the percentages of PhDs and undergrad degrees awarded to women in the econ field have remained flat for the past two decades:

What this says is that an econ degree isn't just an upper-tail IQ test. If it is, the gender gap would have shrunk since the days of Soundgarden and flannel shirts.

So what is causing the gender gap in econ? Preferences are one possibility. Maybe women just don't like econ as much as men. SAT scores are a great proxy for IQ, but they're a pretty terrible proxy for the desire to spend your life selling a bunch of skeptical colleagues on the merits of your own mildly challenging variation on a highly stylized benchmark model with poor data fit, or sifting through the historical record for plausibly exogenous natural experiments. There's no reason to expect math-geniuses to desire that life in strict proportion to their math-geniusy-ness.

Sociology is another possibility. The econ boys' club might be shutting out the ladies. Perhaps they're afraid that a girl will accidentally brush their arms at a seminar, infecting them with incurable icky girl-cooties. Knowing certain economists, I wouldn't be surprised if this is exactly their fear. Or maybe some women think econ has way too high a fraction of creepy libertarian dudes who will subtly try to sink their careers in a misguided attempt at revenge against the cheerleaders who wouldn't sleep with them in high school. Knowing certain economists, I wouldn't be surprised if this is exactly their fear.

So anyway, yes, I guess at the end of the day this is just another "Girls RULE!" post. Are we tired of "Girls RULE!" posts yet? Hell no. You know me. I never get tired of anything.

*Answers: 1. 1/2/sqrt(3), 2. Trick question, as we're all out of bubblegum, and 3. C