Another Study Fails to Justify Affirmative Action

There’s nothing wrong with the first sentence of the National Bureau of Economic Research’s new report, “Affirmative Action and Human Capital Development,” which defines affirmative action as “the practice of granting preferential treatment to under-represented (UR) demographic groups,” but it’s down hill from there. The descent begins in the second sentence, which states that “It was first mandated by the Kennedy Administration in the 1950s….”

Of course there was no Kennedy Administration in the 1950s, and the affirmative action required by Executive Order 11925 — issued by President Kennedy March 6, 1961— not only did not mandate but actually prohibited preferential treatment by requiring government contractors to “take affirmative action to ensure that applicants are employed, and that employees are treated during employment, without regard to their race, creed, color, or national origin” [emphasis added].

The descent continues in the first sentence of the second paragraph, which asserts that “the rationale” for affirmative action is “to help compensate for the disadvantage” caused by “the lingering effects of past institutionalized racism” when in fact promoting “diversity” is the only currently accepted legal rationale for the preferential treatment defended by the authors.

That the three authors — all economists: Christopher Cotton, Queen’s University (Kingston, Ontario); Joseph P. Price, Brigham Young; Brent R. Hickman, University of Chicago — evidently have little interest in the history or legal environment of affirmative action does not necessarily weaken their methodologically dense defense of preferential treatment, but necessarily or not it does weaken it. Indeed, even if their statistically intricate argument promoting the effectiveness of quota-based race norming were persuasive, no court following current legal doctrine would approve such a policy. That’s right: the authors conclude that quota-based race norming “significantly improves disadvantaged students’ exam scores, while significantly narrowing gaps between them and the advantaged group.”

They attempt to demonstrate that affirmative action admissions — “giving special consideration for race in allocation of college seats” — has a positive effect on the “behavior prior to college admissions” of preferred students by providing incentives “to develop pre-college human capital,” i.e., studying and doing their homework. Their research involved a “field experiment” using 5th to 8th graders in Utah that they claim (unpersuasively, I believe) created “a microcosm of the college admissions market.”

Their “experiment,” the authors explain, involves

paying 5th through 8th grade students based on their relative performance on the American Mathematics Competition 8 (AMC8), a national mathematics exam. In order to provided a clean test of theory, we use age/grade cohort as our demographic delimiter. This distinction mirrors racial differences in that our disadvantaged students have, on average, less mathematics training and practice, while filtering out cultural differences which could confound the effects we seek to test.

Students are divided into two treatments for a math competition with real cash prizes, a “color-blind” control treatment and an AA treatment. In the control treatment, students compete against others in their own grade and an adjacent grade for prizes (we ran one contest for 5th and 6th graders and another for 7th and 8th graders), and test scores alone determine one’s pay-off. In a second treatment, prizes are reserved for allocation to disadvantaged students, meaning that competition occurs only within each demographic group, while the distribution of prizes is left unchanged. This treatment represents a “quota” AA policy, an [sic] prize allocation rule which assures equal division of prizes across the disadvantaged and advantaged groups.

All the students were provided access to a test preparation web site, and a test was given after 10 days, and the “disadvantaged” students were found to have taken greater advantage of the web site, i.e., put more effort into developing their “human capital.”

However “clean” the authors’s theory may have been made by using age and grade as proxies for race and ethnicity, it is highly doubtful that the “disadvantage” 5th and 6th graders have compared to culturally similar “advantaged” 7th and 8th graders comes close to mirroring the academic disadvantage entering black and Hispanic students have, on average, compared to whites and Asians, either in the source, nature, or size of that disadvantage. Indeed, even the authors admit, in a footnote, that

[t]he difference in average preparation between our disadvantaged and advantaged groups is likely to be lower than national differences between black and white students in the US. Using performance figures from the National Assessment of Educational Progress (NAEP) exam, we estimate that by 8th grade, the average black student performs roughly as well on the mathematical portion of the NAEP exam as the average white student would have performed in 6th; this represents a performance gap consistent with two years education difference.

Also highly doubtful is whether “individuals’ labor-leisure trade-offs” in a ten-day test with the possibility of a cash reward at the end resembles the long haul of delayed gratification throughout high school that admission to selective colleges requires.

What is, however, perhaps most the most serious flaw in the authors’ research design — or at least the most serious limitation on its relevance to affirmative action admissions in our current legal world — is that it creates and then endorses the efficiency of what can only be described as race-norming (although that term appears nowhere in the report), even though that practice was specifically barred by the Civil Rights Act of 1991.

All students in the “experiment” were told in advance “the group that they were assigned to, how many students they would be competing against, the score distribution of the students in their group based on the practice test, and the prize structure.”

In the color-blind treatments, students competed against others in their own grade and in an adjacent grade. For example, 7th grade students in a color-blind treatment needed to score within the top 30 percent of all 7th and 8th grade students in their treatment to receive a prize.

In the quota treatment, students competed against others in their own grade only. For example, 7th grade students in the quota contest only had to score within the top 30 percent of 7th graders in their treatment to receive a prize.

Roger Clegg has noted other ways the “market” created in this study fails to resemble affirmative action admissions. It does not embody a zero-zum game — “that is, no one is disadvantaged by the advantage given to the preferred students” — and “no attention is given to the costs of preferential treatment (divisiveness, resentment, stigmatization, mismatching, etc.”

This “prize structure,” in which “disadvantaged” students had a much greater chance of winning the prize, does, however, “mirror” the way affirmative action works in practice in one crucial respect. Shorn of the “holistic” camouflage colleges use to disguise what they’re doing, we can see it for the race-norming it is.