Glenn Loury on “color-blind” affirmative action programs

Diligent readers will recall the argument made here earlier that when an institution selecting a group of people from an applicant pool attempts to serve “diversity” goals indirectly, through “color-blind” mechanisms, it must make greater sacrifices in terms of whatever other characteristics it’s trying to optimize than would be the case if it pursued that goal directly.

My friend Glenn Loury — we were friends even when he was still hanging out with the neocons, and I’m still grateful to him for rubbing my nose in a couple of truths that neoconservatism asserted and I that didn’t want to believe — just gave a talk here at UCLA in which he put some numbers around that conceptual argument.

Loury’s statistical strategy was extremely clever. He started with the “College and Beyond” dataset, which gave him both the pre-admissions picture and the college transcripts of the 1989 entering classes at a bunch of elite colleges and universities. He then created for each institution a hypothetical problem: selecting, from the students it actually matriculated, a class half that size, constrained to have at least as high a proportion of black students as the actual class. That is, he pretended that the admitted class was an applicant pool, and that the college was trying to throw away the least-promising half of it. The advantage of running the selection exercise on the actual classes is that he had each student’s actual college GPA to use as an outcome measure.

He assumed that the hypothetical institution he had invented would want to maximize the GPA of the admitted class, subject to the diversity constraint imposed, and that it could do so either with “race-conscious” measures or “race-neutral” ones.

Obviously the job can be done with “race-neutral” measures: randomly selecting from the actually admitted class for the smaller class would be expected to reproduce the proportion of blacks. But selecting for likely GPA alone (using measures such as grades, class ranks, and test scores), ignoring race, would lead to an under-representation of blacks, reflecting the “affirmative action” component of the original selection. So to maintain diversity the school would have to find characteristics of applicants that correlated with being black and give those greater weight than they would otherwise deserve in the selection process. (That’s the logic of plans such as the Texas “10% plan” that admit anyone with a sufficiently high class rank, no matter how poor his test scores.)

Since random selection clearly isn’t optimal, that leaves two strategies:

1) Divide the class into blacks and whites, and pick the “best” (i.e., most likely to get high grades, based on their pre-admission profile) half of each group to make your new class.

2) Invent some other selection process that leads to the racial composition you want, while doing as well as you can on predicted GPA, without mentioning race explicitly in the selection of any individual student.

Loury’s question: What price race-neutrality? That is, how much do you have to sacrifice on the GPA measure to meet your diversity goal without using the race of the applicant directly? Answer: several times as much as you have to sacrifice if you ignore high school grades altogether and use test scores only.

Conclusion: while it’s true, as the Bush Administration argues, that the University of Michigan could serve its admitted “state interest” in diversity without explicitly considering race as a factor, doing so would have very substantial costs in the quality of the student body. [On the other hand, it would also result in increasing the social-class diversity among the white students, because any “color-blind” system designed to favor blacks will also pick up some less-advantaged whites.]

That seems to me a very convincing argument, and I hope it convinces the Supreme Court. (Loury and some other academics, including my friend and colleague Tom Kane, will be submitting an amicus brief based on this study.)

But it also seems to me, as the whole debate seems to me, to be missing, or rather ducking, the hard question: given that “diversity” has costs in terms of “quality,” how much are we willing to pay? How much are we willing to expand the left-hand tail of the distribution of academic aptitude and preparation? (Everyone who has ever been in front of a classroom knows how much the classroom experience is influenced by the hardest-to-teach students.) And how great a gap are we willing to have between the average black student and the average white student? (Or teacher, or doctor, or cop?) [Note: One rather perverse advantage of “race-neutral measures” is that, by selecting a less-than-optimal group of whites, they tend to lessen that gap.]

The logic of “color-blindness” is that the right answer must always be zero, in both cases. Make no sacrifice, accept no gap, just select the best and let the racial chips fall where they may. But I don’t think that’s really the right answer, and that’s not the answer in the Administration’s brief. At the opposite extreme is the demand for “representativeness” on a population basis, which, given the one-standard-deviation gap between blacks and whites on most academic measures, strikes me as a recipe for disaster.

The terms of the quality-diversity tradeoff aren’t happy ones, and the least-bad solution is likely to leave you feeling just a little bit sick both about how badly blacks are under-represented in your student body on the one hand and how weak your weakest students are and how big the gap is on the other. And as Loury and his collaborators have just demonstrated, doing it “color-blind” style is going to make you considerably sicker.

However we do this, it isn’t going to be any fun. But I wish we were arguing frankly about the terms of the tradeoff, rather than making the whole thing into another exercise in Constitutional Talmudism. On a strictly personal level, I’m grateful that the lawyers have agreed to take the brunt of this. If we were having the real argument I’d be on the side saying that, at least in elite academia, the current level of racial preference is somewhat too aggressive, which wouldn’t make me any friends.

But the attempt to make a social question into a legal one is likely to produce some very strange answers, and the resulting policies are likely to be both more expensive and less successful than they could be.

Author: Mark Kleiman

Professor of Public Policy at the NYU Marron Institute for Urban Management and editor of the Journal of Drug Policy Analysis. Teaches about the methods of policy analysis about drug abuse control and crime control policy, working out the implications of two principles: that swift and certain sanctions don't have to be severe to be effective, and that well-designed threats usually don't have to be carried out. Books: Drugs and Drug Policy: What Everyone Needs to Know (with Jonathan Caulkins and Angela Hawken) When Brute Force Fails: How to Have Less Crime and Less Punishment (Princeton, 2009; named one of the "books of the year" by The Economist Against Excess: Drug Policy for Results (Basic, 1993) Marijuana: Costs of Abuse, Costs of Control (Greenwood, 1989) UCLA Homepage Curriculum Vitae Contact: