Replies

In their 1994 book The Bell Curve, Richard Herrnstein and Charles Murray report that in 1990, the average student admitted to Harvard scored 697 on the verbal SAT and 718 on the math SAT, as compared to 674 and 695 for legacy students admitted; these disparities are miniscule in comparison to those that exist between white and black admittees.

This statistic could prove very misleading if the average SAT scores against which the lagatees are compared includes the scores of black admitees. In other words, if you compare the average SAT scores of all white legatees who are admitted to the average SAT scores of all white non-legatees who are admitted, then I bet the differences are quite significant.

This statistic could prove very misleading if the average SAT scores against which the lagatees are compared includes the scores of black admitees.

I was thinking the same way. You have to control for race and sex when evaluating the effect of legacy. And I suppose you should control for legacy and sex when evaluating the effect of race--but that would tend to make the effect of race stronger if the legacy and sex bring the average white SAT down.
But if you reflect on it, the proportion of black students will be on the order of 10% if that's the proportion in the general population--and if legacy admittees are also only 10%, the effect of a 200 point SAT gap in blacks would be about 20 points lower overall average score, a little higher if you control for race when evaluating legacy. But 20 points is after all not insignificant--it's apparently bigger than the male-female or the legacy gap would be.