Individual statistics
Posted by aogWednesday, 25 June 2003 at 11:30 TrackBack Ping URL
Michael Kinsley attacks the recent Supreme Court decision on racial preferences because of its logical incoherence. While I agree with his main point Kinsley doesn't do much better in terms of coherence. Kinsley writes
Admission to a prestige institution like the University of Michigan or its law school is what computer types call a "binary" decision. It's yes or no. You're in, or you're out. There is no partial or halfway admission. The effect of any factor in that decision is also binary. It either changes the result or it doesn't. It makes all the difference, or it makes none at all. Those are the only possibilities.
His statement about getting in to school is correct. But not the statement about the "effect of any factor". It's the equivalent of discussing which vital organ keeps you alive. Remove any one and you're dead (that's what makes them "vital"). More precisely, if one is accepted by scoring one point over the minimum, then every single factor is vital – remove any of them and no acceptance. But it makes no sense to say that multiple different factors each made "all the difference". A truly determinative factor would be one that had a 100% correlation with acceptance. In other words, if every member of an "officially recognized minority" (to quote the Michigan Law School) who applied was accepted, then one could argue that race was a determinative factor. But that's clearly not true.

Kinsley goes on to write

The law-school dean testified that "the extent to which race is considered in admissions … varies from one applicant to another." It "may play no role" or it "may be a determinative factor." O'Connor cites this approvingly, but it is nonsense on several levels. First, "no role" and "determinative factor" are in fact the only possible options: There cannot be an infinite variety of effects on a yes-or-no question.
Actually, there can, because in the statistical scheme of things there are an infinite number of amounts by which the statistical outcome can be changed. Kinsley does a number of switches between statistics and individuals to confuse the issue. But consider this scenario: an applicant needs 100 points to be accepted. He gets as 20 point minority bonus but still only scores 97 points. But the next day, it's discovered that his ACT scores were reported incorrectly and he gets another 5 points for his high scores. Now he has 102 points and gets accepted. What was the determinative factor? If, as Kinsley claims, the answer is race, then that's arguing that the ACT scores played no role. That's as incoherent as the Supreme Court decision.