A few days ago, Kevin Drum of Mother Jones was wondering if college students really did study much less than in the past. A physics professor wrote in to say that during his career teaching introductory college physics, he had always given his classes a standard pretest and post-test to see how much they learned. From 1990-1998, he taught at Coastal Carolina, a "fourth tier" public college in the Carolina golf resort metropolis of Myrtle Beach, then gone into research, then resumed teaching in 2004 at private Spelman College in Atlanta. The Spelman students seem much less like slackers than the Coastal Carolina students, yet his test showed they learned much less physics from him. What had changed between 1998 and 2004?

I of course pointed out the obvious. The professor later commented that the professors at Spelman were more interested in value-added analyses of teaching effectiveness than elsewhere in his experience. I could well believe this is true, precisely because Spelman is racially and sexually segregated (it's the best-known women-only historically black college). Value-added analysis and diversity are, in practice, antagonistic to each other.

Everybody is in favor of in-depth analysis of educational statistics in theory, but when they actually finally look at them, the obvious leaps out: Holy cow, compared to everything else, race really matters. You can't adjust it away (except by using cynical proxies for race the way Steven Levitt tried to do a few years ago: e.g., favorite soda flavor is Grape or the rough equivalent). And when it comes to physics and a few other subjects, sex matters.

And so the reformers give up on honest analysis of educational statistics in despair, boredom, and cynicism. On the other hand, if you are teaching at Spelman, race and sex differences don't get in the way, so trying to look at subtler matters of the impact of teaching style and the like is less psychologically devastating.

Still, there are things to be learned. For example, here are the counts of every high school student who got the maximum score of five on the Physics C Advanced Placement test:

5 on AP Physics C 2008
	Male	Female	Sex Ratio	% of Total
Total	9,017	2,010	4.5	100.0%
White	5,705	1,127	5.1	62.0%
Asian	2,414	708	3.4	28.3%
Other	316	69	4.6	3.5%
Not Stated	199	57	3.5	2.3%
Other Hispanic	177	32	5.5	1.9%
Mexican American	98	7	14.0	1.0%
Black	78	6	13.0	0.8%
Puerto Rican	18	3	6.0	0.2%
American Indian	12	1	12.0	0.1%

These are quite striking figures. Leaving aside the ambiguous "Other Hispanic," "Not Stated," and "Other" categories for the moment, only 17 of the 11,027 high school students in America to max out the Physics C AP test in 2008 were specified Non-Asian Minority girls, or 0.15%.

And that's with a massive national effort to get NAM girls interested in science. If we doubled that effort and there were no diminishing marginal returns, we'd be all the way up to 0.3%. If we doubled it again with no fall-off in return, we'd be up to 0.6%.

The sex ratio column is interesting. It's highest for NAMs and lowest for Asians. This is the opposite of scores on less difficult tests where black females often outdo black males.

My theory is that girls are more conformist, so in a culture that obsesses over academic credentials, such as Asians, they do well, but in cultures that don't, they don't. Thus, only 6 black girls in the entire U.S. in 2008 got 5's on Physics C, and only 7 Mexican-American girls. In contrast, boys are more individualistic, so 78 black youths and 98 Mexican-American youths just went with their inner physics nerdishness and pulled down 5s. Good for them.

Another thing to note is that the Other Hispanic category shows up large compared to Mexican-American on most tests of high achievement. For example, the number of LSAT test-takers who identify themselves as Mexican Americans is remarkably tiny.

Some of these Other Hispanics are of course Cubans, some are elite immigrants from the capitals of Latin America, some are middle-class Central American refugees' children. And others who check "Other Hispanic" represent what I might call "non-homeboy Latinos," the strivers. If your Mexican-American dad went off to college and there he met your Guatemalan mother, then "Other Hispanic" would be natural for you to check. But, the key here is that your parents left their neighborhoods for their educations. In contrast, the homies who are pure Mexican by descent come from a culture that validates loyalty to family ties over individual advancement.