Harvard learns discrimination is complicated

Reading through the expert economists’ submissions in the discrimination suit that Asian-American applicants have brought against Harvard University, I’m afraid what struck me most was their compensation, which they evidently must disclose early in their reports.

Students for Fair Admissions, Inc., claims Harvard’s affirmative-action policies discriminate against AsianAmerican students by taking spots from qualified AsianAmericans to give to less-qualified students from those racial groups targeted by the policy (for instance, African-Americans and Latinos). The plaintiffs paid Duke University economics professor Peter Arcidiacono US$450 per hour for a paper supporting their claims of discrimination against Harvard, while Harvard paid University of California at Berkeley economist David Card, a Canadian, his “standard billing rate” of US$750 per hour for his analysis supporting Harvard’s defence. (Note to any reader interested in acquiring my services: I am available at either rate.)

In The Wealth of Nations, Adam Smith observed how the common custom of paying lawyers and clerks of court “according to the number of pages which they had occasion to write” had led them “to multiply words beyond all necessity, to the corruption of the law language of … every court of justice in Europe.” Hourly billing is different. Still: Professor Arcidiacono’s report to the court is 168 pages, Professor Card’s is 198 pages. Given their rates, maybe that’s not surprising.

Why did Card get US$300 more per hour than Arcidiacono? It’s complicated, as compensation often is. Arcidiacono is a very talented labour economist with many high-quality publications. Card, however, is something of an economics superstar, winner in 1995 of the John Bates Clark Medal, given to the best American economist under 40 (where “American” includes foreigners working at U.S. universities). There’s also the fact that with its US$37-billion endowment, Harvard has deeper pockets than Students for Fair Admissions. Again, like many outcomes in labour markets and other social processes, it’s complicated.

In fact, that is essentially Harvard’s defence in the admissions case: It’s complicated. In his report for the plaintiffs, Arcidiacono lays out some pretty damning statistics. Asian students’ Scholastic Aptitude Test (SAT) scores are on average “24.9 points higher than white applicants; 153.9 points higher than Hispanic applicants; and 217.7 points higher than African-American applicants.” In fact, he notes, the test scores among all Asian-American applicants (not necessarily those admitted) were higher on average than those among the African-American or Hispanic students that were actually let into Harvard.

He also runs some regressions — equations correlating admission with different possible explanatory variables — and finds that being Asian-American shows up negatively and in a statistically significant way. That is, it shows up in a way that’s too big to be just a random tick of the numbers. On average, Asian-Americans have a 3.95-per-cent chance of getting in to Harvard. But, with their test scores and GPAs, if they were treated the same as whites, their chance would be 4.7 per cent. If they were treated the same as Hispanics, their chance would be 12.3 per cent, and as African-Americans, 24.2 per cent. If the admissions process were race-blind, Arcidiacono calculates, Asian-Americans would be more than half of Harvard’s undergraduates, not their recent average 22 per cent.

But hold on, says Professor Card: it’s more complicated than that. Harvard is swamped with academic over-achievers. It admitted 1,800 U.S.-based students to the graduating Class of 2019. But for those 1,800 slots, it received more than 37,000 applications, including 8,000 with grade point averages of a perfect 4.0 and 3,500 with perfect math SAT scores.

If Harvard relied only on high school marks and test scores, it would have to turn away thousands of students with essentially identical, essentially perfect records, Card argues. Not surprisingly, for this and other reasons, it puts more into its admissions pot than simply academics. It also explicitly scores students on their extracurricular and sports activity (not just for potential varsity athletes). And, in a process that seems very labour-intensive, time-consuming and self-important, it seeks “distinguishing excellences” in lots of other things: family background, hardships overcome, maturity, artistic qualities, the degree of parental involvement, leadership, unusual “life experience,” and so on.

When Professor Card includes all the additional variables that he could plausibly quantify in Professor Arcidiacono’s explanatory equation, the variable “Asian-American” loses its statistical significance. But even with a greatly extended set of variables, he explains, you can’t perfectly predict who gets in and who doesn’t. Almost certainly, your econometric efforts suffer from the “omitted variables” or “missing data” problem. As a general rule, when the processes being examined are complex, attempts to emulate them are all but certain to miss out at least some important pieces of the explanation.

It is complicated. These are both good papers. I don’t envy the judge who must decide between them.

But whether Harvard wins or loses, the world’s most influential university officially is arguing — on record, in court — that questions of discrimination are very hard to untangle. What looks like slam-dunk racial bias actually may be quite innocent. Universities, including Harvard, have been spawning-grounds of the socially corrosive belief that discrimination is pervasive in the rich democracies, even “systemic.” If a lawsuit or two makes them think more deeply about that, then bring on more lawyers.

WHAT LOOKS LIKE SLAM-DUNK RACIAL BIAS MIGHT NOT BE.