Social Science Statistics Blog

You state: "Differences in mean SATs among accepted students have little or nothing to do with affirmative action!!"

Even given your simple example, that isn't likely to be true. Affirmative action doesn't explain 100% of the black-white SAT score gap but it likely explains some of it. If the cutoff for blacks is lower than that of whites, in expectation, the black-white differential for matriculating students will be greater than under the equal standard regime.

You make a good point. Few people seem to understand that affirmative action does not account for all of the black-white score gap. But its disingenous to imply that, on the basis of this example, affirmative action explains little or none of it.

You note that "the distribution for blacks is shifted a bit to the left." Is "a bit" Harvard-talk for "more than 1 standard deviation"?

Just curious.

By the way, it is beyond dispute that Harvard (and other elite schools) practice significant amounts of affirmative action. See the briefs in the 2004 Supreme Court cases for evidence. Whether this is good or bad depends on your point of view.

In this context, I think it is worth keeping in mind familiar results on selection bias. From Achen (1986), we know that in censored samples, as admissions committees do better at selecting on factors that are unobserveable from the point of view of the data analyst, their apparent performance on observeable factors (say, SAT scores) will decline. The example of two distinctive SAT distributions is important to keep in mind, but so too is the problem that we only observe some of the factors on which committees make their decisions.

I am not sure what selection bias or censored samples has to do with this discussion.

This is an empirical question and Harvard is, theoretically, interested in the truth. We can just ask Gary (Ah! The wonders of tenure . . .) to request the data. Harvard has admissions data on all its students as well as measures of academic performance. We would find is that students with strong high school records and test scores do much better at Harvard than students with less strong records. There is a huge literature on this, almost all of which shows that things like GPA are predictable.

Although Felix is correct that, in theory, a mean difference in SAT scores does not demonstrate that affirmative action plays a significant part in admissions, we know from the universities own testimony that it does. There is no reason to rely on guess work.

If you don't think that affirmative actions "dilutes academic standards" --- defined as letting in a large number of students who you would not have let in if they were, for example, Chinese American --- then you don't know what you are talking about.

But anyone looking for a tenure track job should not discuss this issue in a public forum. There is little upside.