Some thoughts and evidence on racial preferences in admissions

They're probably effective, but Americans really don't like them, and they might be illegal. So what do we do?

Well, it looks like America’s national conversation about race is about to get another topic of discussion: The Supreme Court has decided to hear challenges to racial preferences in college admissions, starting in October. The last time SCOTUS adjudicated this topic was in 2003, when it declared that racial preferences could stay, but only if colleges used them as one of many factors, and with an eye toward ensuring a diverse student body. Now, with a more conservative set of justices on the bench, many experts expect the court to overturn that ruling and just ban racial preferences outright.

It’s worth noting that unlike with abortion, where SCOTUS’ recent moves to weaken Roe v. Wade go against the general lean of public opinion, this would probably be a popular move. A strong majority of Americans of all races and political parties opposes the use of race as a factor in college admissions:

This is not because Americans are against using policy to increase the number of Black and minority students on campus. In fact, they are strongly in favor of that idea:

It’s only explicitly using race as a factor in admissions that Americans oppose. Gallup surveys confirm this dichotomy — people are in favor of “affirmative action”, but oppose explicit racial preferences:

Americans' top-of-mind reactions to the term affirmative action are generally positive. Gallup asks a straightforward question about affirmative action without a definition or explanation -- "Do you generally favor or oppose affirmative action programs for racial minorities?" -- and as of Gallup's last asking in 2018, 61% of Americans were in favor, while 30% were opposed. Support has increased from the range of 47% to 50% who were in favor in 2001, 2003 and 2005…

Americans are also solidly behind the broad concept of equal opportunity and improving the position of racial minorities in society -- the underlying rationale for affirmative action…

Despite this majority consensus for the general concept of affirmative action and the need for more action on reducing racial inequities, public support appears significantly lower when questions ask about policies that explicitly take race into account to achieve these objectives…72% of U.S. adults oppose giving preference to Black Americans in hiring and promotion, including 43% who say they oppose strongly…[T]hese results highlight the complexities of public opinion when considerations of affirmative action get down to specifics.

A cynic might interpret this disparity as a form of hypocrisy — people wanting to think they’re in favor of racial equality but shying away when the rubber hits the road. But I think the fact that Black Americans also display this divergence in poll responses suggests that something subtler — and more reasonable — is at work here. I’ll speculate more about this later on, and suggest some ideas for how to square the circle.

But for now, let’s think about the policy itself. Social science research can’t tell you what’s right and wrong, nor can it adjudicate the law. But it can offer some information about what’s at stake — about what effects different policies have. Americans want policies to increase opportunity for historically disadvantaged minorities, but they don’t want explicit racial preferences. So let’s take a look at the evidence on affirmative action programs, including racial preferences, and see what works and what doesn’t work.

Racial preferences in college admissions: The evidence

One theory of racial preferences is that they improves minority achievement by spurring minority kids to greater efforts. If Black kids, for example, feel that the deck of the education system is stacked against them, they may not try hard in school; if they know there are programs out there to help them get into top colleges, they may take their schooling more seriously.

Akhtari, Bau & Laliberté (2020) investigate this idea. They use the fact that the Supreme Court ruling in 2003 that allowed racial preferences in college admissions ended up overruling a 1996 lower court ruling against the policy. The authors look at the SAT scores and college applications of high school students in Texas, and find that the reinstatement of the policy has a salutary effect:

Our estimates suggest that the re-instatement of affirmative action helped close the racial gap in applications to selective schools by 13%, the gap in math SAT scores by 5%, the gap in secondary school grades by 18%, and the gap in secondary school attendance by 8%. These positive effects are concentrated among high school students in the upper-half of the 6th grade test score distribution, who are more likely to be on the margin of admission to selective Texas public universities.

Second, we find no evidence that pre-college human capital investment fell for [under-represented minorities] anywhere in the test score distribution. The reintroduction of affirmative action policies in Texas had no detectable disincentivizing effect on effort by URMs. Third, we find that white students also experience small increases in their SAT scores in response to the policy, consistent with either positive spillovers from URMs or whites also responding to increased returns to effort due to greater competition for admissions.

In other words, affirmative action made Black and Hispanic kids try harder, and that made White kids try harder in order to keep up. The authors also find a decrease in the college graduation gap, though they caution that this could have other explanations.

An interesting 2019 working paper by Thibaud suggests that preferences work via the mechanism described above — students from disadvantaged minority groups underestimate their own ability to succeed within the system, and racial preferences send a signal that corrects this underestimation. Thibaud finds that in France, the implementation of racial preferences at some selective schools increased minority applications to, and enrollment in, other selective schools. This suggests that the signal sent by these preferences is important, and that signal increases effort and ambition.

Cotton, Hickman & Price have a 2014 experimental paper that suggests the same conclusion. They give a math test, and pay students to do well on it — the intention being to mimic the very real incentives of getting a college education. They then investigate whether dividing the prize winners by grade level has an effect on people’s scores on the test — the idea being that younger students are at a disadvantage that roughly mimics the effect of being from a disadvantaged minority group. And in fact, they find that dividing up the prizes does increase both effort and achievement! In other words, once they know that they won’t be competing directly against advantaged students, the disadvantaged students try harder and do better. And the older students don’t decrease their effort or do any worse.

But, it should be noted that the evidence on this particular mechanism is far from a slam-dunk. Card & Krueger (2006) find that while the elimination of racial preferences in California and Texas schools in the late 1990s did substantially decrease enrollment of Black and Hispanic students at selective public universities, it didn’t decrease the rate at which Black and Hispanic students sent their SAT scores to those universities. And Antonovics and Backes (2014) find that California’s 1998 ban on affirmative action at public colleges didn’t result in a measurable decline in Black and Hispanic academic achievement or in a widening of racial achievement gaps.

But Card and Krueger’s result shows that while the theorized mechanism of greater ambition and effort is still in doubt, the blunt effect of racial preferences — increasing Black and Hispanic representation at selective schools — isn’t really in doubt. Other papers, such as Hinrichs (2012), confirm this. Whether it’s due to increased effort and ambition or simply to the effects of explicit discrimination, racial preferences do mean more Black and Hispanic students at selective universities.

But that doesn’t mean that Black and Hispanic students necessarily always benefit from these policies. There’s another strand of research on the possible negative effects of racial preferences on minority students. For example, Arcidiacono, Aucejo, Coate & Hotz (2012) find that when the University of California dropped racial preferences after the 1998 ban, UC graduation rates rose. They attribute a modest fraction of this effect to matching — to a modest degree, racial preferences were putting some minority students into college programs they weren’t ready for, causing them to drop out. But the bulk of the increased graduation rate can’t actually be explained by better student-to-school matching, or by UC schools picking more talented students; some unidentified other factor is at work.

In general, evidence on the mismatch hypothesis is patchy. Sander (2004) looks at law school outcomes and argues that Black law students do much worse as a result of being directed toward schools they aren’t prepared for. But Rothstein & Yoon (2009) take a look at the data on law schools and conclude that this mismatch is actually pretty moderate. They conclude that if racial preferences at law schools were removed, the number of Black lawyers would significantly decline — again, not that unintuitive or controversial of a finding.

So to sum up, the evidence is strong that racial preferences do affect the racial composition of elite schools, and almost certainly of professional jobs that recruit graduates from elite schools. In other words, racial preferences are effective at their most basic task. But what’s not as clear yet is how much of this effect comes via lifting effort and ambition, and how much comes from simple brute-force discrimination in favor of disadvantaged minorities. I find the data encouraging in terms of the former, but it’s absolutely not settled science yet.

Meanwhile, there’s evidence that shifting to race-neutral systems comes at a cost. Ellison & Pathak (2016) study a shift in Chicago from race-based to place-based preferences in admission to elite high schools, and find that the place-based system is only about a quarter as effective as a race-based system in increasing minority representation at those schools. And Arcidiacono & Lovenheim (2015) find that Texas state schools’ “percent plans” that automatically admit anyone in the top grade percentiles at their high school — essentially a placed-based policy — are less effective than racial preferences at ensuring minority representation. This strongly suggests that class-based preferences — a commonly suggested alternative — would also do less for Black representation.

In other words, we do see a tradeoff here. Race-based preferences increase the representation of Black (and, to a lesser degree, Hispanic) students at elite schools. Popular alternatives to racial preferences are less effective at this task. So eliminating racial preferences — as most Americans want to do, and as the Supreme Court might force schools to do — will almost certainly decrease the number of Black students at elite schools to some degree.

How to square the circle

So what do we do? Americans want “affirmative action”, and they want policies to put more Black people on campus, but they oppose using race as an admissions criterion — which, in practice, is by far the most effective way of putting more Black people on campus. How do we square that circle?

As I said above, it’s unlikely that hypocrisy or insincerity is the main explanation for the diverging poll results, since Black people show the divergence as well. One possibility is that these seemingly inconsistent responses are all about tone — that “racial preferences” sounds punitive and zero-sum, while “affirmative action” sounds happy and positive-sum. Fryer and Loury (2005) suggest something like this is at work, arguing that basic misunderstandings about affirmative action lie at the heart of public attitudes.

But I have another hypothesis. Perhaps what Americans want is informal racial preferences, but not formal ones. Maybe what they want is for admissions officers to quietly favor Black applicants, but not to make this an explicit rule.

Why might Americans want this? One reason is that racial policies aren’t just about outcomes — they send a message about the nature of our society. If we write race into rules and regulations, it sends a message that our society does not comprise a single, united policy, but is irrevocably balkanized into separate racial groups. More racial division is something that Americans generally do not want — including Black Americans. Drawing bright legal lines around racial groups is innately scary and distasteful to most people.

Another reason is that formal rules are persistent in ways that individual decisions and attitudes are not. Hopefully, affirmative action policies and the passage of time will eventually remedy racial disparities — in fact, for Hispanic Americans, that day may not be far off. If and when racial disparities diminish, so does the need for affirmative action; it was always intended to be a temporary boost to counteract a history of exclusion. But if racial preferences get written into formal rules, it will be politically and logistically difficult to get rid of them if and when changing conditions warrant it. At that point, systematic racial unfairness will be ingrained into our rules. But if racial preferences are quiet and informal, they can be quietly and informally abandoned when appropriate.

So although this may sound a bit wacky, maybe what Americans want is actually just individual bias in favor of disadvantaged minority groups. And if the Supreme Court rules formal racial preferences illegal, that may be exactly what ends up happening. Admissions officers are still going to read applicants’ essays, and look at information on their background, take into account the disadvantages that those applicants faced — including racial disadvantages. At that point, there won’t be much the opponents of racial preferences can do, because proving bias in court is very hard to do. They will just have to live with the knowledge that admissions officers believe in the importance of historic, persistent racial disadvantage.

In other words, maybe this time Americans will get what they want when it comes to racial policy.

---