Looking for Class Preferences to Replace Racial Ones

Socioeconomic preferences can be a better proxy for race than race preferences, according to an Inside Higher Ed report this morning on a new study to be published this summer in the Harvard Law & Policy Review.

More precisely, the authors, Matthew N. Gaertner, a researcher at Pearson’s Center for College and Career Success and Melissa Hart, associate professor of law at the University of Colorado, argue that properly constructed class-based preferences can lead to more racial diversity, i.e., a larger number of underrepresented minorities (URMs) being admitted, than current race-based preferences. A preliminary version of the article is available here.

Analyzing random selection of applicants admitted and rejected at the University of Colorado in 2008 and 2010, the authors describe a complex class-based construct built on a highly complex “disadvantage index” and “overachievement index.” The latter is relatively simple: it measures the degree an applicant’s grades or standardized test scores exceed those typically earned by those in their socioeconomic group.

The disadvantage index, however, “derived from two prediction equations” calculating the marginal increase or decrease in the probability of enrolling in a four year college based on an expansive list of measures of socioeconomic status, is far from simple. Here, for example, is the diversity index the authors provide for a hypothetical applicant, “James,” based on a whole slew of disadvantage factors such as his parent or parents’ income and educational history, language spoken at home, number of siblings, his high school and the performance of its graduates, etc.:

exp(−2.07 − 0.07(1) − 0.06(3) + 0.39 − 0.15(1) − 0.003(70) − 0.03(15) + 0.0001(100) + 0.86(2.7) + 0.6(−0.21))
1 + exp(−2.07 − 0.07(1) − 0.06(3) + 0.39 − 0.15(1) − 0.003(70) − 0.03(15) + 0.0001(100) + 0.86(2.7) + 0.6(−0.21))

= 0.391

exp(−2.07 − 0.07(1) + 0.11(2) + 0.71 − 0.15(0) − 0.003(15) − 0.03(18) + 0.0001(400) + 0.86(2.7) + 0.6(−0.21))
1 + exp(−2.07 − 0.07(1) + 0.11(2) + 0.71 − 0.15(0) − 0.003(15) − 0.03(18) + 0.0001(400) + 0.86(2.7) + 0.6(−0.21))

= 0.636

This is not the place and I am not the person (I dropped out of math before  arithmetic got hard) to evaluate these and other similar measures. Readers will have to go to the study itself to judge the variables that went into these equations and the results that came out. What is important here is the authors’ conclusion that the “privileging” of their complex class-based identifications produced more URM admits than CU’s straight race preferences.

Holding constant high school GPA and standardized test scores, URMs are 1.4 times more likely than non-URMs to be admitted under CU’s race- based policy. By contrast, under the Disadvantage and Overachievement Indices (again holding constant high school GPA and standardized test scores) applicants identified for primary factor consideration are 5.7 times more likely to be admitted.

If this study survives the analysis it deserves, it could undermine the argument that overt racial preferences are necessary to achieve racial diversity, but it is not without its own problems (which perhaps will be addressed when the final study is published). First, since it is so clearly intended to produce an admissions system that will admit significant numbers of URMs, it could easily run afoul of any standard of discrimination that relies on intent.

Another question concerns the identity of the expanded numbers of URMs. The preliminary study defines URMs as “Blacks, Latinos, and Native Americans,” but it doesn’t disaggregate them. Since some of the disadvantage characteristics would favor Hispanics over blacks in Colorado and elsewhere (such as language spoken at home and rural domicile and high school), it would be useful to see a breakdown of the URMs by race and ethnicity.

This class-based affirmative action is built on “privileging” intricately determined socioeconomic identity even more than race preferences “privilege” race, and the authors recognize that the resulting admission of students with lower academic qualifications than their non-preferred peers makes it vulnerable to “mismatch” criticism. They recognize the criticism, but are not at all persuasive in refuting it. They argue, for example, that “replacing race-based affirmative action with a class-based approach will not substantially affect aggregate measures of academic qualifications” (emphasis added), but that’s only because the class-based affirmative action doesn’t admit sufficient numbers to lower the entire class. For individuals, however, the “mismatch” problems could be substantial.

Regarding the hypothetical “James” referred to above, for example:

while his raw academic credentials may not understate his potential, when James enters CU he will be able to draw on life experiences that most of his undergraduate peers will not. Thus, James should bring views and perspectives to the University that would be absent were he refused admission.

If the mismatch between James and his peers is too great, he will probably be at the very bottom of his class and may well not graduate, but at least he can provide “diversity” to others as long as he lasts. This would not seem to be a good or even moral bargain.

Indeed, one of the authors’ judgments that I believe casts a threatening cloud over their entire study is their argument, expressed in a footnote, that “the mismatch theory has been subject to significant and credible challenge.” Any proposed admission system that relies on rejecting the accumulating evidence supporting the work of Richard Sander, Stuart Taylor, and others — which in my view has not been credibly challenged — has a long, steep, uphill path to climb.