The gender gap in math scores, particularly at the high end of the distribution, has been well documented and debated, with some suggesting that genetics might be a big reason for the differences we see.
Although some recent research has shown that the average girl and boy in the U.S. perform similarly, when you move away from the center, a surprisingly large gap appears. This chart from a new NBER paper by Glenn Ellison and Ashley Swanson at MIT shows how the percentage of women getting high scores on the SAT and the AMC (a math competition held at about 3,000 high schools annually around the country) declines sharply with scores:
Ellison and Swanson find that at the highest percentiles for AMC scores, the ratio of males to females increases from about to 2:1 to 10:1. And while high-achieving boys can come from most any background, "the top-scoring girls are almost exclusively drawn from a remarkably small set of super-elite schools: as many girls come from the top 20 AMC schools as from all other high schools in the U.S. combined."
Ellison and Swanson believe that societal rather than genetic reasons are the more likely culprit and that -- astonishingly -- "almost all girls who would be capable of achieving extremely high scores do not do so."
My first post here at The Stash pointed to some data which backs up Ellison and Swanson. If you look at math test scores in other countries, you see that the gender gap at the high end is not a universal phenomenon: In Iceland, Thailand, Indonesia, and the U.K., girls and boys score at about the same levels in the 95th and 99th percentiles: (click chart for larger image)
Now, Steve Levitt of the Univeristy of Chicago and Roland Fryer of Harvard offer some more details on the universality of the math score gender gap. Here
is a chart from their new paper showing the relative performance of girls versus a "gender equality" index:
Again it looks like there is a pretty consistent trend of males doing better than females, but the gap narrows in societies with greater gender equality.
But look what happens when countries from the Middle East like Jordan, Cyprus, Bahrain, Saudi Arabia, Iran, etc... are added:
The strong correlation disappears. So, what's going on? These countries aren't particularly well-known for having liberal attitudes towards women. Levitt and Fryer think it has something to do with same-sex schooling:
Having lived in one of those Middle Eastern countries -- Iran -- for six years, I think this theory does have some merit. Iranian women are no slouches when it comes to education, particularly post-Islamic Revolution, and perhaps the lack of sexual freedom (and imagery) for both men and women also plays a role in how discrimination against women impacts math scores in those countries.