As Jason Richwine noted yesterday, Harvard is being sued for discriminating against Asian applicants, and it’s defending itself by claiming that Asian applicants score lower on non-academic measures — measures that Harvard is almost forced to use because so many applicants have perfect scores. As Richwine explains, while many applicants do have perfect math scores or perfect verbal scores or perfect GPAs, it’s unlikely very many have perfect scores across all the different measures. Harvard is using non-academic criteria by choice, which of course is not illegal but raises the question of whether the criteria themselves are being selected with the goal of engineering a certain racial balance.
Another aspect of the case I find fascinating, though, is a battle royale between the economists David Card (handling data analysis for Harvard) and Peter Arcidiacono (for the plaintiffs). Both have submitted reports in the case (here and here) as well as responses to each other’s reports (here and here) analyzing whether, all else equal, Asians are less likely to be admitted than whites. It’s a pretty striking illustration of how researchers can make statistical results appear and disappear by subtly changing the way they crunch the numbers.
To me, in terms of simple facts, the most damning evidence against Harvard actually comes from Card’s own report. Asian applicants outperform whites academically and in terms of extracurriculars; where they actually fall behind is in athletics and an amorphous “personal” category that the plaintiffs say is biased in itself. (As the New York Times puts it, this measures “traits like ‘positive personality,’ likability, courage, kindness and being ‘widely respected,’” and Asians tend to get lower ratings.)
Card claims his statistical model accurately predicts Harvard’s admissions decisions and produces no evidence of bias; he alleges Arcidiacono’s model leaves out crucial control variables. Arcidiacono says that, actually, it’s Card’s model that isn’t set up correctly — e.g., it includes applicants who were shoo-ins on the basis of donor parents and athletics or sure rejections because they fell far short of Harvard material, fails to account for “interactions” between racial and economic preferences, and includes that controversial “personal” category.
This is par for the course in terms of academic debates. It’s just that a lawsuit hangs in the balance this time.