Scan the administrative ranks of US colleges and universities, and more likely than not you will find a Chief Diversity Officer charged with building a diverse, equitable, and inclusive campus community. University commitments to these initiatives can include scores of administrative positions and upward of ten million dollars in financial support. Many universities (including my own) now place diversity, equity, and inclusion (DEI) on equal footing with knowledge production, education, and service to society as a core mission. In light of this undeniable trend, one might reasonably ask whether these initiatives align with widely held social values or whether they are the result of administrative capture by a liberal elite forcing their ideology on a polarized society.
Enter John M. Carey, Katherine Clayton, and Yusaku Horiuchi. In this exemplary book, they undertake an earnest, innovative approach to add to our understanding of student and faculty attitudes toward campus composition. They ask a relatively straightforward set of questions: Who do students want in their cohorts, who do students want as faculty, and who do faculty want as colleagues? And, is there a consensus for diversity, or do campuses resemble our patchwork of polarized red states and blue states?
The book’s impressive contribution owes much to the authors’ limited scope. They measure preferences, and they measure them well. They leave to others to delve into the psychological, sociological, and historical contributions of those preferences. They touch on the economic, political, and legal implications of the policies that universities have adopted, but recognize that those broader questions already spawn enough PowerPoint decks, STATA analyses, legal briefs, tweets, and takes to fill your Gmail inbox quota thrice over. So, why not get some facts?
Take a moment and think about their central question: Are preferences for diverse campuses widely held? This is not so easy to answer; that is, unless you possess the capacious hubris of say, David Brooks, enabling you to grasp the zeitgeist of college campuses through osmosis while popping by on a book tour. The old school approach, a survey, suffers from likely larger-than-normal sample bias and social desirability bias. And there is no natural experiment to be found, a cause for concern or, dare I say, a concern for causality.
The authors, wisely, apply conjoint analysis, a tool more familiar to marketing than political science. Conjoint analysis offers subjects multiple alternatives described as collections of attributes, and subjects then choose their preferred alternative. Here, the alternatives consist of two potential admits to a college, and the attributes include the applicant’s grade, SAT score, gender, race, family income, and extracurriculars. In brief, everything an admissions officer might see except the essay and the letters of recommendation.
By varying which levels of attributes people see—for example, top 5% SAT, top 2% SAT—conjoint analysis can infer how much weight people attach to changes in SAT scores. The method produces an average marginal component effect (AMCE), which is the average effect of an attribute on the probability of choosing a student. An AMCE of 0.3 for high SAT scores implies that, given two applicants who are identical across all other attributes (averaged across all possible combinations of attributes)—the one with the higher SAT scores will be chosen 30% more often than the baseline level; in this case, an SAT score in the bottom 25%. These inferences are possible, because conjoint analysis presents combinations of alternative values in uncorrelated bundles.
AMCE values are defined relative to the mean of all attribute values (the centroid to be precise). They make sense in a specific case, say for a female applicant from a high-income family, provided there are no interaction effects. Although interaction effects do exist (see intersectionality), their form and magnitude are such that we can restrict attention to marginal effects.
As a methodological exercise, the book surpasses expectation. Methods instructors might well consider including it as an exemplar. The scale and scope of the empirical project merit accolades as well. The experiments sample thousands of students and faculty from a half-dozen schools. Although not a random sample, the schools—which include the University of Nevada–Reno, Dartmouth, and the University of North Carolina—are broadly representative.
The analysis shows that scholarly achievement matters most. Getting an SAT score in the top 2% has an AMCE of 40% relative to being in the bottom 25%. Graduating in the top 1% of one’s high school class has an AMCE of 25% relative to being in the bottom 60%. Race matters: African Americans and Latinx get nearly 10% bumps. So, too, does social class: if your parents make a cool half-million, you get an AMCE of minus 10%.
Turning to the point of the study, the data also reveals, as given away by the book’s title, a broad consensus in favor of diversity. The authors characterize the degree of consensus by comparing AMCEs (the marginal value of attributes) across subpopulations. Strong consensus means that two group’s AMCE values have significant signs in same direction and the statistical differences in those values are insignificant. White students give African American students about an 8% bump, and African American students give a 10% bump. Strong consensus.
Consensus can also be weak; that is, the AMCEs can be significant in the same direction but significantly differ in their values. As might be expected, some groups prefer larger benefits for themselves. Finally, the authors define polarization as groups having significantly different signs; for example, one group favors legacies, and the other does not.
Although conjoint analysis can construct groups endogenously, the authors fixed their groups ahead of time based on school, cohort (freshman through senior), race, gender, family income, and political ideology. This was an appropriate choice given their objectives. The analyses find nonexistent school and cohort effects. Dartmouth students weight diversity, income, and legacy status pretty much the same as students at Nevada–Reno. Seniors think like freshmen, a finding that undermines claims of liberal indoctrination.
They also find that people of all races, genders, and income level agree on direction, although many of the AMCE values differ significantly (weak consensus). As we might expect, underrepresented minorities give more weight to race than whites, and lower-income students discriminate against wealth more than wealthy students.
Variations in magnitude notwithstanding, the signs almost always align. Even students who oppose affirmative action give a slight nod to minorities and lower-income applicants. The only groups that do not weigh such factors differently for underrepresented minorities or lower- income applicants are Republicans and people with high racial resentment. Both groups show favoritism to legacies and first-generation students, while discriminating against people of nonbinary gender. This last finding represents the lone instance of polarization. Democrats give nonbinary gender applicants a leg up.
In sum, there is a good deal of consensus all around. We generally all get along. Well, with two additional caveats. First, the authors assumed their groupings. Within groups of Republicans and Democrats there may exist coherent subgroups who disagree. This could be discovered by allowing for endogenous groups. Second, agreement on directional effects (weak consensus) need not imply agreement on actual admission decisions where applicants have correlated attributes. The analysis suggests that African American students would be far more likely than white students to admit a lower-income, African American applicant than a rich, white applicant with slightly higher SAT scores. Similarly, nonwhite students would advocate for more nonwhite faculty than would white students, female students for more women faculty, and so on, and so on.
Thus, even though the study reveals almost universal consensus, we can still look forward to lively campus debates about admissions criteria, with no shortage of people lining up on opposite sides of admissions and hiring decisions. Even so, how wonderful to know that though we may differ in the strength of our advocacy for diversity and inclusion, we believe in a common direction—forward.
