Statistics and the SAT:

The the L.A. Times anti-SAT op-ed, which I criticize in the post below, has one especially noteworthy statistical claim: “[W]hite students score 206 points higher on average than nonwhites” on the SAT, “according to Psychology Today.”

     To begin with, this seems to be a misquote of the Psychology Today article, which says “Whites outscore African Americans on average by 206 points.” African Americans are fewer than half of nonwhites in America, and less than 40% of the nonwhites taking the SAT (note that the SAT statistics treat Hispanics as nonwhite).

     Second, it is whites and Asians who score much higher on average than blacks and, to a lesser extent, Hispanics. According to Time, Oct. 27, 2003, the racial breakdown of SAT averages was:

1. Asians: 1083.
2. Whites: 1063.
3. Mexican-Americans: 905.
4. Other Hispanics: 921.
5. Blacks: 857.

(I assume that this is the same data on which the Psychology Today account is based, since it also shows a 206-point gap between whites and blacks.)

     The Times op-ed quote is still correct as to blacks (rather than “nonwhites”), and it does make an important point. But don’t the numbers have a somewhat different impact — one much less compatible with the “evil whites oppressing nonwhites” subtext that one often hears in connection with the SAT — when one sees that the gap is between Asians and whites on one side and blacks and Hispanics on the other, rather than between whites and nonwhites?

UPDATE: Reader Michelle Dulak reports:

I was curious what the mean “white”/”nonwhite” score comparison would look like, so tried to find out. The College Board site gave the 2002 breakdown of test-takers as 65% white, 11% black, 10% Asian, 5% “other” Hispanic, 4% Mexican, 4% “other,” 1% American Indian. I left out the American Indian and “other” categories, because they weren’t in the data you quoted, and tried weighting them by fraction to get an overall mean. The mean for those 95% of test-takers was 1026. The mean for “non-white” test-takers (counting Hispanics as “non-white”) was 950. The white/non-white gap is therefore 113 points, not 206 — if you count Asians as “non-white.”

Of course putting it like that obscures the obvious fact that the “non-white” distribution curve is shaped like a two-humped camel. Still . . .

I haven’t checked the math here, but it sounds about right to me.