The Gender/Math Gap

An item mentioned over in Discover GNXP on the Male-Female Math Gap got me looking a little more closely at SAT scores. The gap in high-level achievement in math has been narrowing, at least as measured by the SAT's. But there's reason to suspect that the narrowing may not be quite as straight-forward a trend as it may appear.

The SAT's - until last year - have published the number of male and female test takers by 50-point score ranges for the Verbal and Math tests. Thus we can compare the number of male vs. female students who have scored at 750 and above on the Math test to see how the gap has been fairing.

The SAT also publishes average test scores by sex along with their respective standard deviations. Thus we can also calculate the expected number of each sex who should score above 750 each year. The chart below compares the actual achievement gap vs. the predicted.

The actual/expected stayed within fairly tight bounds, but in 2007 and 2008 there was a noticeable shift where the actuals have dropped significantly below expected. The SAT is not universally taken. While in most states it's considered mandatory, it's certainly plausible that in traditional ACT states high-achieving girls have made a concerted effort to take the SAT in recent years, thus pumping up the 750+ scores over what would be expected based on the distribution of the scores generally.

Here's the same chart with expected male-female ratios for those scoring 800. Again it does not suggest quite the narrowing that has actually occurred in the 750+ scores.


Looking at the actual numbers of SAT test takers by sex who scored 750 and above compared to the expected number based on their means and standard deviations, it would appear that there is indeed a tendency for more high achievers to be more likely to take the SAT in recent years - and the female effect looks larger.

(Note the shamlessly sexist color scheme.)