Via Tyler Cowen we learn there’s a new paper from Duke’s Peter Arcidiacono — the economist who served as an expert witness for the plaintiffs in the recent lawsuit against Harvard. (I’ve written about Arcidiacono’s work before here and here.) In this one, Arcidiacono and his coauthors detail a rather bizarre pattern in the school’s admissions data.
Basically, starting with the class of 2009 — in the wake of some Supreme Court activity that mildly limited affirmative action — the school began seeking out lots more applications from black students, leading the black share of the applicant pool to grow from about 6 percent to about 10 percent. Some of this effort seems to have taken the form of recruitment letters sent out on the basis of test scores, with far lower cutoffs for members of underrepresented groups.
Yet the new applicants apparently had little prayer of getting in. “The share of admits who were African American remained unchanged,” the authors write. “At the same time, the average SAT score of African American applicants fell by 33 points (on an 800-point scale)” between the classes of 2008 and 2012, something that did not happen for other racial groups.
Indeed, it’s not much of an exaggeration to say that many black applicants to Harvard have no chance of getting in:
African Americans account for only 11% of the applicant pool [for the classes of 2014 through 2019], but 41% of the applicants in the bottom decile of the academic index. The admit rate for non-[athlete, legacy, dean’s list, or staff] African Americans in the bottom decile was 0.03%: only 2 of the 5,921 applicants in this decile were admitted over the six year period.
Why beef up the applicant pool by adding a bunch of kids you know won’t be accepted from a single racial group? Arcidiacono et al. float a solid theory: Doing this helps to conceal the school’s affirmative-action policies in the data.
Take a look at this chart. Through 2008, admit rates fanned out by race exactly the way you’d expect at a school with strong affirmative action. Black applicants were the most likely to get in, followed by Hispanics, whites, and Asians.
But now think about this a little more carefully. Those early numbers, in and of themselves, don’t really prove the existence of affirmative action. In theory, we would also see this pattern if students from some groups were more promiscuous with their applications than others. Maybe blacks were more conservative, spending their time and money to apply only when they had really strong academic records, while Asians were more likely to apply even if their scores and grades were borderline at best in the context of Harvard.
In other words, admit rates depend not just on a school’s admissions policies, but also on students’ propensity to apply. The paper’s theory is that the school used this concept to reverse the pattern in the chart above.
If you solicit a ton of applications from black students who have no prayer whatsoever of getting in, it makes it look like you’re being more selective with blacks when you turn around and reject those kids — even though you’ve changed nothing else about your process. Indeed, since 2010, whites have actually had higher admit rates than blacks, despite the use of racial preferences to help blacks but not whites.
You have to admit it’s kind of clever.