After Affirmative Action, Meritocracy?

In cases against Harvard and the University of North Carolina, the U.S. Supreme Court appears poised to end the use of race in college admissions. But when the decisions come down, colleges will not merely drop racial preferences and otherwise continue the admissions practices that they had before.

Some colleges may discreetly defy the ruling and continue to give better admissions chances to applicants from preferred racial groups.[1] But many more will change their practices openly, placing greater emphasis on socioeconomic factors and other variables in order to increase diversity.

There are two interesting things about these ostensibly “race-neutral alternatives” to affirmative action.

One is that, while they are race-neutral in the sense that they do not involve the explicit consideration of race, they are anything but race-blind in their construction. To a surprising extent, the goal of these policies is to maintain the racial balance that obtained at each school under race-based affirmative action. Their parameters are calibrated to achieve that end, and many judge them as a success or failure depending on whether they do so. To affirmative-action supporters, they are an end run around bans; to affirmative-action opponents, they are a way of proving that diversity is possible without the use of race.

Simple college admissions preferences based on parental income, for example, do a poor job of targeting underrepresented minorities, so administrators often seek out alternatives. Richard Kahlenberg, who wrote expert reports in the UNC and Harvard cases, laying out how race-neutral alternatives can achieve racial diversity, in a 2015 paper[2] forthrightly referred to variables such as wealth and neighborhood-level poverty as “racialized SES [socioeconomic status] factors” because they capture forms of disadvantage that are more tightly linked to race: black families tend to live in poorer neighborhoods and to have less wealth than white families with the same current income. A “percent plan,” to take another example, grants admission to state schools to all students in the top 10% (or any other chosen number) of their high school classes, which essentially leverages segregation[3] to bring in students from minority neighborhoods who might not win admission otherwise. This policy is already in effect, with bipartisan support, in states such as Texas and Florida.

The second interesting thing about race-neutral alternatives is that everyone across the political spectrum seems fine with them. While current Supreme Court precedents require schools to show that reasonable race-neutral alternatives cannot achieve the diversity that they seek before resorting to the use of race,[4] even most affirmative-action opponents would maintain race-neutral alternatives as an option.

Kahlenberg’s work in the current cases included actively proposing race-neutral alternatives for the schools being sued and touting their demographic effects. This included the “Simulation D” for Harvard that was extensively discussed at oral argument[5] (originally “Simulation 7” in Kahlenberg’s reports).[6] At oral argument, lawyers for the plaintiffs also affirmed that schools could, for example, consider race in the context of essays in which applicants voluntarily discuss their racial identities or experiences with discrimination.[7]

For those who want race-blind, academically meritocratic, admissions, how much of an improvement will this be? Time will tell. But based on past experience, academic research, federal data, and current trends, we can make some observations and educated guesses.

Race-neutral alternatives are popular for very good reasons. They limit the unseemly practice of directly handing out valuable slots at selective colleges based on skin color. They focus preferences on applicants facing real obstacles, whereas traditional affirmative-action beneficiaries are often quite privileged in nonracial ways; even “racialized” socioeconomic factors target concrete disadvantages that often historically stem from racial oppression. And race-neutral alternatives force schools to make a trade-off that we might find desirable: if they want to play out their racial obsessions in crafting their student bodies, they’ll have to help disadvantaged kids of all races, at least to some extent, along the way.

These policies can also break down some of the advantages of elite “feeder” high schools, in favor of kids at less prestigious schools who perform well. One study found that kids pulled in by the “Texas Ten Percent Plan”[8] actually had higher test scores than the feeder-school kids who were rejected,[9] though it analyzed the Texas Ten Percent Plan in isolation rather than in conjunction with the end of affirmative action—and thus did not evaluate the students who previously would have been offered admission to schools on the basis of race.

But there are two major pitfalls to bear in mind—one of them mathematical and the other more legal and ethical.

Racial Quotas Through Other Means

The mathematical problem is this: if the actual goal is to create a certain racial balance, the most efficient way to do that is simply to have a quota. If a school wants, say, 15% of its students to be black, it can reserve 15% of the seats for black applicants, and then—separately, within the reserved and non-reserved groups of seats—just use its normal admissions process, maximizing the other things that it values, such as academic qualifications. Current Supreme Court precedent prevents schools from being quite that blatant about it, but colleges can get close by providing amorphous “boosts” and “tips” based on race in a so-called holistic process designed to achieve a vision of “diversity.”[10]

By contrast, when a school tries to achieve a certain racial balance but must ignore precisely the variable that it is trying to calibrate, it has to admit on the basis of preference a wider swath of applicants to succeed. To admit the 15% of the student body it actually wants, school admissions officers need to give preference to more than 15% of the total.[11] Those concerned about “mismatch” between students admitted through preferences and the rest of the school should find this very troubling.[12] Other preferences—for example, for socioeconomic status—pose this problem just as much as racial affirmative action does, and (if the goal is really racial balance, after all) will actually require heavier use to get the same result.

At many leading universities in states that have banned affirmative action—nine states currently do so, and court decisions resulted in a Texas ban that lasted from 1996 to 2003[13]—black and Hispanic shares did, in fact, return all the way to their pre-ban levels, as Kahlenberg and Halley Potter detailed in a 2012 study,[14] which Potter followed up with a 2014 book chapter.[15] (Three exceptions are UCLA, UC-Berkeley, and the University of Michigan, discussed below.)

Meanwhile, Simulation D, based on far more detailed data than are publicly available, nicely illustrates the trade-offs involved here—if at a highly unusual and extremely selective school. Under Simulation D, Harvard would completely wipe out its racial, as well as its legacy, dean’s list and children-of-faculty-and-staff (LDC) preferences, and instead grant a large preference based on socioeconomic status (SES). Those current preferences take up a lot of space: nearly a fifth of all Harvard students in 2019 were estimated to be LDCs (though some would have been admitted even without preferences);[16] racial preferences boost admissions chances four times for black applicants and 2.4 times for Hispanic ones.[17] SES would be gauged not only at the level of individual families but also at the high school and census-tract level, where it would be measured by equally weighting three factors: parental income, parental education, and percentage of students speaking a language other than English at home.[18]

By killing LDC preferences, Harvard would lose things that it values, though few outside the institution will shed tears for the school’s legacy admits or $50 billion-plus endowment.[19] Yet by giving such weight to socioeconomic factors—which would boost admissions of students considered “disadvantaged” a whopping 30 percentage points[20]—Harvard would also take a hit to its academic excellence. The number of admits with the highest academic indexes would fall about 17%.[21] And the typical SAT score of Harvard students would fall 64 points on a 2,400-point scale, roughly from the 99th to the 98th percentile.[22] In more dramatic terms, instead of demonstrating 1-in-100 talent on the test, students would generally exhibit less extraordinary 1-in-50 performance; in more technical terms, the difference is about a quarter of a standard deviation on a bell curve.

At oral argument, downplaying the SAT drop, a plaintiffs’ lawyer likened it to moving “from Harvard to Dartmouth.”[23] But Harvard is not Dartmouth; it’s Harvard. And even by becoming Dartmouth, Harvard still wouldn’t quite get back the minority shares that it had under affirmative action.

The “Hispanic and other” share of enrollment would rise 5 percentage points, presumably in part thanks to the use of foreign languages in doling out the massive SES boost, but the black share would fall 4 points, or nearly a third, from 14% to 10%. Not to worry, Kahlenberg argued in his filings: wealth and family structure, for instance, were missing from the data used in the simulation, but Harvard has access to them and could grant preferences based on low wealth and single parents, too, to better target African Americans.[24]

Legal and Ethical Debates on Racial Preferences

This brings us to the legal and ethical dimensions of the problem. Many affirmative-action opponents profess to believe that racial preferences for blacks and Hispanics in college admissions are morally comparable to old-school racial discrimination in the other direction. But with anti-black or anti-Hispanic discrimination, few would tolerate race-neutral alternatives extensively calibrated to achieve the same result by other means.

Quite to the contrary: in employment law, a major focus of the fight against discrimination, the Supreme Court in the 1970s concocted a “disparate impact” doctrine holding that race-neutral policies could be held discriminatory even if they were not intended to achieve a particular racial result. If a challenged policy has disproportionate impacts by race, even with zero evidence that these impacts occurred by design, the burden is on the employer to show that the policy has a legitimate business purpose.[25] In the early 1990s, when the Court appeared likely to eliminate or curtail this doctrine, Congress wrote it directly into Title VII, the portion of the Civil Rights Act of 1964 that deals with employment.[26]

Similar court precedents apply (via the same act’s Title VI) to programs that receive federal funding, which include most colleges,[27] and Department of Education regulations contain a disparate-impact rule, too, barring funded entities from using “criteria or methods of administration which have the effect of subjecting individuals to discrimination because of their race, color, or national origin.”[28] There are also assorted prohibitions, of course, on intentionally discriminating by race through the use of facially neutral criteria.[29]

But few seem enthusiastic about applying restrictions on race proxies to university admissions policies. During oral argument in the UNC case, a plaintiffs’ lawyer even helpfully highlighted for worried administrators a legal argument that public schools could employ in a scenario where they considered three race-neutral admissions policies and picked the one that admitted the largest number of black students. Under existing case law, governments accused of making decisions based on illegal criteria such as race can argue that they would have made the same choice even without the alleged wrongful motives.[30]

To be sure, comments made at oral argument are not binding legal precedent, and future cases may explore limits for race-neutral alternatives to affirmative action. Indeed, that process is already under way at selective public high schools, where courts have historically been less friendly toward the use of race. One case currently working its way through the system challenges the policies of the elite Thomas Jefferson High School in Northern Virginia, whose leaders lamented the demographic results of its academics-heavy admissions process and then switched to a system that admitted fewer Asian students.[31] A district court sided with the plaintiffs; an appeals court recently sided with the school.[32]

Nonetheless, in her discussion of Simulation D at oral argument in the Harvard case, Justice Sonia Sotomayor made an important point:[33]

All of the alternatives, whether it’s the 10 percent plan, whether it’s socioeconomic [preferences], they’re all subterfuges to reaching some sort of diversity in race. You’re touting them as race-neutral, but none of them are race-neutral. You’re doing them because you believe in racial diversity. I just don’t understand why considering race as one factor but not the sole factor is any different than using any of those other metrics.

Opponents of affirmative action might ask themselves the same thing.

As David Card, Harvard’s expert economist in the case, pointed out in one of his reports,[34] it’s probably not a coincidence as to which flagship schools had not managed to restore their previous racial balances under state affirmative-action bans when Kahlenberg and Potter looked at the numbers about a decade ago—three of which, again, were UCLA, UC-Berkeley, and the University of Michigan.[35] These are unusually selective public schools. In fact, they take the top three slots nationwide in the current U.S. News & World Report rankings of public colleges and universities.[36] 

Kahlenberg argued that such schools are in a broader geographic competition for talent than less elite, more locally focused, public schools. Therefore, under state affirmative-action bans, these schools were still competing against colleges that used preferences.[37] That dynamic probably did play a role and will not be an issue if preferences are barred nationwide.

But perhaps these schools also would have had to make particularly large changes to their admissions practices to re-achieve their old ethnic mixes, and were not willing to do so, given their high standards, the fact that racial gaps in academic qualifications tend to be more pronounced at the extremes,[38] and the fact that affirmative action can cause a cascade effect, where the most elite schools admit the students who would normally get in and also poach students from one rung down the ladder, which, in turn, means that lower-level schools have to scramble even more to meet their diversity goals. (This last effect could play out in reverse under an affirmative-action ban: the very top schools lose their ability to poach but also do not gain access to new students who otherwise would have been poached from them, since there are no schools above them from which such students might trickle down.)[39]

An unwillingness to lower standards could thus be a major brake on elite schools’ behavior if affirmative action is banned nationwide. Maybe Harvard is telling the truth when it says that Simulation D would undermine its selectivity too much to be a reasonable alternative.

Or maybe this time will be different because a lot has changed since most of the states banning affirmative action passed their policies. Most strikingly, a movement to reduce or eliminate the role of standardized tests—which are a strong predictor of college performance and, contrary to allegations of bias, do not underpredict the college performance of racial minorities[40]—has gained ground. The group FairTest reports that more than 1,700 schools are test-optional, and more than 80 schools are entirely test-blind, which means that they do not consider test scores even if students submit them.[41] Harvard itself has announced that it will remain test-optional through the 2023–26 admissions cycles.[42]

Relatedly, the Covid-19 pandemic forced an abrupt change in the way colleges do business, including shutting down some testing sessions that incoming 2021 freshmen would normally have used. A survey from the federal Integrated Postsecondary Education Data System (IPEDS) asks colleges about their test-score requirements—albeit with a poorly worded question that will be improved starting with the yet-unreleased 2022 data[43]—and the numbers reveal the extremely fast deterioration of testing requirements. For Figure 1, I pulled the numbers for schools that mostly grant BAs and above (not including schools that grant only graduate degrees). The question is not asked of open-admissions schools, which grant enrollment to all applicants.[44]

One can only guess what the chart above will look like updated through, say, 2025. Some schools, such as the Massachusetts Institute of Technology, have rushed to reinstate their testing requirements, noting how effective the scores are at predicting whether students are ready to handle intense course work.[45] Others have decided to make their testing changes permanent, and in the case of the University of California, a court settlement eliminated the use of the ACT and SAT and administrators decided not to pursue other options.[46] But the pending affirmative-action rulings will likely give schools another incentive to drop tests, and (analogously to the pandemic’s effect on remote work) Covid-19 forced many schools to at least give it a shot, acclimate to the idea, and invest in making test-optional admissions work, whether they liked it or not.

Does dropping test-score requirements change a school’s demographic composition? Given the racial skew in test scores, one would certainly expect that to be the case, but the existing research is sparse and mixed. One study, for example, found about a 10% (or 1 percentage point) rise in combined black, Hispanic, and Native American enrollment from test-optional policies, with little difference by the institutions’ selectivity;[47] another failed to find any effect at all.[48]

My own analysis of the 2019–21 IPEDS data, presented in more detail in the Appendix, suggests a modest effect of ending test requirements on black first-time enrollment that may be stronger at more demanding schools—in the ballpark of a 5% boost for schools in the top half of selectivity—though no such effect is measurable for Hispanic enrollment. My results also suggest that black and Hispanic enrollment shifted toward more selective colleges in general in 2021, regardless of

whether the schools reported dropping testing requirements; an obvious theory is that summer 2020’s “racial reckoning,” spurred by the murder of George Floyd, made admissions officers more sympathetic to black applicants.

These data have the advantage of being very recent and also of covering a period with much abrupt change in test-score requirements, creating a sort of natural experiment. But they have the disadvantage of being drawn from the pandemic years, which were utter chaos, as testing sessions were canceled[49] and international enrollment plummeted and then rebounded.[50]

Perhaps most important, none of the periods covered by past research represents what we’ll see in the future, when affirmative action will likely be curtailed nationwide in the new environment of lax attitudes toward standardized testing. If, before the pandemic, a school had already achieved its desired racial balance by combining test scores with racial preferences, and then the school suspended its testing requirement only because the pandemic forced it to, the school may not have been overly tempted to exploit the situation to shift its demographics even further. The results may be different if a school finds its racial balance threatened and chooses to ditch testing as a way to preserve it.

A better glimpse of the future may lie within the University of California system, where racial preferences have been banned since voters approved a ballot referendum in the mid-1990s, where the top two schools struggled mightily to regain their prior demographic balances, and where admissions officers started ignoring test scores entirely in 2021, though, of course, the pandemic and other changes to admissions practices make any analysis of the system’s enrollment only suggestive.

Figure 2 illustrates the black share of applicants, admits, and enrollees at all nine UC campuses, calculated from data released by the system itself (and running through 2022, a year past the most recent IPEDS numbers).[51] UCLA is the most striking example: some 10–15 years ago, it managed to push its black enrollment share up after a decade of decline; and in 2022, soon after it started ignoring test scores, it admitted black students more or less the way it had before the 1996 ban.

Figure 3 illustrates Hispanic enrollment, with generally less dramatic results but still with clear changes at some schools.

The Supreme Court can ban the direct use of race in admissions, but it cannot dictate schools’ admissions processes more broadly. Therefore, the ultimate impact of such a decision is hard to predict. Schools have many options for racial engineering beyond outright preferences—race-neutral does not mean race-blind. The plaintiffs challenging UNC and Harvard even elaborated on those options, including admissions preferences for socioeconomic status, percent plans, and consideration of essays that note students’ racial identities. Each of these options will present many thorny trade-offs. And so the fights over race and enrollment in higher education will be far from over when the Court rules.

Test-Score Requirements and Black and Hispanic Enrollment

As in Figure 1, I begin with IPEDS data from non-open-enrollment schools that primarily grant BAs or above for 2019, 2020, and 2021.[52] Because 2019 test scores are available only for schools for which testing was “required” or “considered but not required” in that year, the data are limited to those schools. I use the NIPALS algorithm to create a combined index of 25th-percentile SAT math, SAT reading/writing, ACT math, and ACT reading scores to measure how selective schools were in 2019; this measure of schools’ selectivity is a time-invariant baseline, applied to the 2020 and 2021 data as well. I match these testing variables to first-time, full-time fall undergrad enrollment by race. IPEDS records racial categories only for U.S. students, so foreign students are excluded from the analysis. Further limiting the data to schools with the needed admissions and testing-requirement data for all three years, I have about 1,200 schools in total, or 3,600 observations over the three years.

I divide the schools into four equal-size categories based on 2019 test scores and seek, using negative binomial regression to account for overdispersion, to test the relationship between the count of black enrollment and the presence of a hard test-score requirement (i.e., the “required” category, which dominated the data for the 2019 year but had faded by 2021). The model includes controls for total non-international enrollment (as black enrollment is expected to rise when total enrollment does) and the interaction of test-score category and year (to allow schools of varying selectivity to have different trends in black enrollment), as well as school fixed effects.

Dropping test-score requirements appears to slightly boost black enrollment at more selective schools. A nearly 6% estimated increase in black enrollment at top-quartile schools that drop testing requirements (“Not required x top-quartile test-score index”) is marginally statistically significant, and combining the top two quartiles into the top half gives a clearly significant result. The results also suggest greater black enrollment in top-half schools in 2021 in general.

Table 1.

Model 1: Black Enrollment After Dropping Test-Score Requirements

Variable	Estimate (SE)
Log of total non-international enrollment	1.066*** (0.0481)
Test scores not required	–0.0025 (0.0239)
Not required x second-quartile test-score index	0.0113 (0.0357)
Not required x third-quartile test-score index	0.0500 (0.0334)
Not required x top-quartile test-score index	0.0593• (0.0326)
Bottom quartile x 2020	–0.0119 (0.0134)
Second quartile x 2020	0.0198 (0.0156)
Third quartile x 2020	0.0221 (0.0149)
Top quartile x 2020	0.0138 (0.0128)
Bottom quartile x 2021	–0.0372• (0.0206)
Second quartile x 2021	0.0138 (0.0211)
Third quartile x 2021	0.0655** (0.0239)
Top quartile x 2021	0.0534** (0.0205)

An alternative approach is to include the 2019 test-score index as a continuous variable (scaled to have mean 0 and standard deviation 1), which generates loosely similar results. In this approach, the general effect of dropping test scores on black enrollment is nearly 3% and possibly a bit higher at more selective schools, though the interaction is not statistically significant. Once again, we see a boost in black enrollment concentrated at more selective schools in 2021.

Table 2.

Model 2: Black Enrollment After Dropping Test-Score Requirements

Variable	Estimate (SE)
Log of total non-international enrollment	1.072*** (0.0492)
Test scores not required	0.0281* (0.0117)
Not required x scaled test-score index	0.0147 (0.0118)
2020	0.0105 (0.0075)
2021	0.0233* (0.0112)
Scaled test-score index x 2020	0.0076 (0.0064)
Scaled test-score index x 2021	0.0344** (0.0114)

As noted in the main text, results for Hispanic enrollment are less clear, but here are the same two models with the outcome variable changed to the count of Hispanic enrollment. In these models, Hispanic enrollment generally grows over time, and there again appears to be a “2021 effect” in which enrollment seems to shift to more selective schools in that year, but there is no measurable effect of dropping test-score requirements.

Table 3.

Model 1: Hispanic Enrollment After Dropping Test-Score Requirements

Variable	Estimate (SE)
Log of total non-international enrollment	1.041*** (0.0249)
Test scores not required	–0.0027 (0.0169)
Not required x second-quartile test-score index	0.0064 (0.0240)
Not required x third-quartile test-score index	0.0064 (0.0220)
Not required x top-quartile test-score index	0.0174 (0.0223)
Bottom quartile x 2020	0.0300** (0.0111)
Second quartile x 2020	0.0419*** (0.0105)
Third quartile x 2020	0.0463*** (0.0101)
Top quartile x 2020	0.0341** (0.0105)
Bottom quartile x 2021	0.0090 (0.0165)
Second quartile x 2021	0.0410* (0.0165)
Third quartile x 2021	0.0698*** (0.0146)
Top quartile x 2021	0.0643*** (0.0144)

Table 4.

Model 2: Hispanic Enrollment After Dropping Test-Score Requirements

Variable	Estimate (SE)
Log of total non-international enrollment	1.046*** (0.0243)
Test scores not required	0.0063 (0.0079)
Not required x scaled test-score index	~0.0000 (0.0083)
2020	0.0394*** (0.0055)
2021	0.0458*** (0.0079)
Scaled test-score index x 2020	–0.0047 (0.0056)
Scaled test-score index x 2021	0.0195* (0.0083)

Please see Endnotes in PDF

Photo: APCortizasJr/iStock