Monday, April 04, 2005

A note on affirmative action and university admissions

Brad DeLong points to UC Berkeley chancellor Robert Birgeneau's call for the repeal of Prop. 209. In general I agree; however, I would like to point out one fact that is, perhaps, not as dispositive as it may originally appear. Birgeneau writes (emphasis added):

Proposition 209 assumed that considering race or ethnicity in the admissions process would allow undeserving students into Berkeley. But it is significant that the graduation rates of African Americans before and after the proposition's passage have stayed virtually the same. Far from weeding out students who could not succeed, the elimination of race as a consideration in admissions has actually prevented many of California's most able students from the opportunity of a Berkeley education.

Now, that the graduation rates of African Americans remained the same does indicate exactly what Birgeneau says: that many able students are being denied opportunity. However, this is actually a trivial observation, because of two empirical facts:

  • The pool of qualified applicants to any truly top-tier educational institution vastly exceeds the number of open slots.
  • The information available to admissions committees constitutes a highly noisy and imperfect channel.

These facts are definitely true, for example, of the [EDIT: my school]'s computer science graduate program, and, to only a slightly lesser extent, also of [my school]'s undergrad CS major (to which students must apply; admission is highly competitive). Therefore, the job of an admissions committee at a top institution is not to select "the best qualified N applicants" for some suitable value of N, as some naïve commentators on affirmative action seem to assume. Rather, there is some broad equivalence class of applicants who are "probably good enough" --- the committee usually doesn't have enough information to be more precise than that [0] --- and the committee's job is to pick a subset of this class of students that embodies some desirable balance. For a graduate computer science program, this would mean ensuring that you admit students whose expressed primary research interests are in a variety of subfields; that students do not all come from the same handful of undergraduate institutions (and especially that only a very small fraction come from your own undergraduate program); and, most likely these days, that your students come from a variety of personal/cultural backgrounds. It seems to me that these concerns, including the last one, are are all entirely appropriate when you're trying to compose an incoming class.

But the fact remains that, no matter which subset you pick, you will always have to reject a whole bunch of students who could probably have succeeded in your program. So this particular aspect of Birgeneau's statement is rather trivial. Graduation rate can function as a sanity check that you are not reducing the quality of the acceptance pool. But it's also important to realize that any "sane" acceptance strategy --- including the pre-Prop-209 strategy --- will inevitably have "prevented many of California's most able students from the opportunity of a Berkeley education", and repealing Prop 209 will not change this fact.

Fortunately, because America has many excellent universities, and because no two admissions committees will make exactly the same decisions, it's basically the case that (setting aside the very serious issue of financial hardship) anybody who's in the top equivalence class can get a very high-quality education.

It's also worth noting that, at top undergraduate programs, the empirical effect of affirmative action is roughly as follows: the fraction of qualified African Americans, Latinos, and middle Americans who get accepted increases; the fraction of qualified whites, Asian-Americans, Jews, and bicoastal urbanites who get accepted decreases. It's rather telling that the complaints against affirmative action almost always focus on the racial/cultural components of this selectiveness, and not the geographical components. I might take the meritocratic moral outrage of white conservatives from middle America somewhat more seriously if they complained as vociferously about the tremendous advantages that Nebraskans and Montanans enjoy over New Yorkers and Angelenos as they do about the advantages of African-Americans and Latinos over whites. As it is, it's pretty clear that most of them have no greater social good in mind, and speak purely from self-interest.

[0] OK, this is only true to a first approximation. If you want to be more precise, there are basically two equivalence classes at the top end. (A) Superstars, whom you must admit because they're just too good to refuse. At the undergraduate level, these are the 4.0 GPA varsity letter athletes who have also won Westinghouse or written W3C specs or played at Carnegie Hall or published a bestselling novel before they graduate from high school. (B) People who are "good enough", among whom you choose in order to create a balanced incoming class. These are the students who "merely" get stellar grades and have lots of extracurricular activities. The number of people in category (A) is so small as to be relatively insignificant in the grander scheme of things.

No comments:

Post a Comment