Affirmative Action, Transparency, and the SFFA v. Harvard Case
Affirmative action in college admissions for underrepresented minorities provokes strong emotions. These strong emotions are guided by two competing principles. One of these principles is the desire to treat individuals as individuals. For many, it is galling that race plays an explicit role in college admissions decisions. And in an ideal world it would not. On the other side is the fact that we do not live in an ideal world, but rather a world that has treated some minority groups—and especially Blacks—appallingly. Given the current reckoning over the mistreatment of Blacks, there is a strong desire to set things right, or to at least move in that direction.
For those with strong positions on either side of the affirmative-action debate, there is no need to study the effects of affirmative-action policies. For some, the effects are irrelevant: one’s race should never play a role in admissions, and race-based preferences should be outlawed. For others, any outcome that deviates from proportional representation is by definition unjust given the lack of genetic differences across races. All that is needed for this group is for the racial shares among matriculants to line up with their respective shares in the population.
While affirmative action is often presented as a binary issue, some may support race-based preferences at certain levels but not at other levels. For this group—at least until Students for Fair Admissions v. Harvard—it would be nearly impossible to form a position on whether current admissions practices go too far or not far enough. The reason for this is simple: universities tightly guard their data, and so actual knowledge of the effects, or even the extent, of race-based preferences is infeasible.
In this Essay we make three points. First, constructing affirmative-action programs requires nuance. Conditional on supporting some form of affirmative action, the optimal level of affirmative action will depend on the weight one places on things like representation of minorities at top schools, representation of minorities in particular majors, the value of cross-racial interaction, and the benefits to underrepresented minorities themselves. And the optimal level will also depend on what universities are willing to reveal to their students. As we will illustrate, the utility benefits that underrepresented minorities receive could be enhanced by universities being up-front about how students’ academic backgrounds translate into educational outcomes at their schools.
Possible reasons universities are not up-front with students include perverse competitive incentives but also the legal environment in which affirmative action operates. And this leads to our second point: the Supreme Court’s decisions in two milestone affirmative-action cases have pushed universities to be less transparent. By ruling that using race in a points system is not allowed but that using race holistically is, the Court has given universities an incentive to hide what they are doing. As we will show, holistic admissions systems can be expressed as an implicit formula where one of the components is not observed. A natural consequence of this is that a points-based system can appear holistic by simply not reporting one of the components and not making the formula public.
And much could be learned from universities making their admissions process more transparent. Our third point is that, by pulling back the veil of how Harvard’s admissions operates, the SFFA v. Harvard case showed how little we actually know about how admissions works. For example, it was understood that underrepresented minorities as well as legacies and athletes receive preferences in the admissions process. But the evidence that was uncovered shows that the strength of these preferences is striking.
I. Optimal Affirmative Action
We begin by making the case that nuance is needed when considering optimal affirmative action. We discuss the subtleties around three questions related to affirmative action: (1) when and how does academic background matter?; (2) what do students know when making their college decisions?; and (3) where are the diversity benefits? These are by no means all the factors that should be taken into account when evaluating affirmative-action policies. But these three questions illustrate the importance of nuance in the affirmative-action debate.
A. When and How Does Academic Background Matter?
More selective colleges tend to have more resources, and more resources generally produce better outcomes. Graduation rates at the most selective schools tend to be higher than at less selective schools even after accounting for selection. But it is not clear that increasing college quality always leads to better outcomes as the match between the school and the student may be important.
To illustrate, the pace and assumed knowledge of those taking courses at an elite school may be different than those at lower ranked institutions. This is particularly relevant for those in majors that build on previous coursework, such as in STEM fields where the monetary returns to college are especially high. And here the level of academic preparation does appear to matter. Arcidiacono, Aucejo, and Spenner (2012) document that, at Duke University, over 54 percent of Black men who express an initial interest in STEM or economics end up choosing a major in humanities or one of the other social sciences; the similar share for White men was less than 8 percent. The authors show that these large racial differences disappear once one accounts for differences in academic preparation.
These results are further confirmed by Arcidiacono, Aucejo, and Hotz (2016), who studied the University of California system. Attending a more selective UC school generally increases five-year graduation rates. But in the sciences, the match between the school and the student is important. Those with the strongest academic backgrounds are more likely to graduate in the sciences at the top schools but this is not true for those with relatively weaker academic backgrounds. And the effects here can be substantial: the authors estimate that underrepresented minority students (URMs) who attend Berkeley and are interested in the sciences would see their probability of graduating in the sciences in five years increase by 7.2 percentage points (off a base of 27.5 percent) had they instead attended UC Riverside. But this is for all URMs. The academically strongest URMs would not see such an increase; the large gains occur for those URMs at Berkeley with relatively weaker academic backgrounds.
While the existing literature comes from a limited number of universities, it is important to note that some affirmative action may still be beneficial for increasing STEM majors. Dillon and Smith (2020) use nationally representative data to show that varying the quality of college that a student attends has little effect on his or her majoring in STEM, on average. A possible explanation is that match effects are important at higher-quality colleges but that at lower levels of college quality, schools are vertically differentiated in terms of student ability. In this case, getting students into Riverside from lower-ranked schools would be beneficial from the perspective of graduating in STEM even if moving them all the way to Berkeley would not.
B. What Do Students Know When Making their College Decisions?
But even if affirmative action at elite institutions results in fewer Black STEM majors and STEM majors pay especially well in the labor market, it does not follow that affirmative action makes Black students worse off. Indeed, if students are as informed as universities about how their pre-college skills translate into success in different school-major combinations, then affirmative action can only be welfare-enhancing for its beneficiaries. The reasoning is simple. When a student is fully informed, expanding the choice set can only increase one’s welfare. Consider a student who, without affirmative action, would choose college A. If, under affirmative action, college B is added to the choice set, the individual would only choose college B if their payoff was higher than choosing college A.
As shown by Arcidiacono et al. (2011), this argument breaks down when students are not fully informed and when the university has private information about how well the student is likely to do at their school. In this case, adding an option can decrease welfare as the student may not recognize that, in fact, college B is a worse choice for them than college A. The authors show that, at the school they study, the university does have private information about how well students can expect to perform in the classroom, thereby satisfying a necessary (though certainly not sufficient) condition for mismatch.
This information asymmetry would be easily resolved if universities provided information to students about what their likely educational outcomes would be given what universities know about the student. Returning to the example from Duke, if Black students were informed about, for example, their probability of success in STEM fields and still chose to come to Duke, then there would be little scope for affirmative action to reduce welfare for its intended beneficiaries.
And here is where we begin to see the costs of opacity. It is our view that universities have a moral imperative to provide students with accurate information about their prospects of success. Examples include basic information such as expected grades conditional on their major, probability of persisting in their intended major, and the probability of dropping out. At each university there are plenty of professors available who could easily assemble this sort of information. This information would be especially useful to those for whom accurate information is harder to obtain. And it is even more important for students whose academic credentials are far from the average of their university so that they can make informed decisions about their schooling choices.
C. Where Are There Diversity Benefits?
The legal rationale for whether race can be taken into account in university admissions, however, is based on the benefits achieved from having a more diverse student body. A more diverse student body can positively impact student experiences not only in college but also later in life. When weighing these benefits, it is important to keep in mind that affirmative action primarily affects where students enroll in college, not whether they enroll at all. Indeed, Arcidiacono, Khan, and Vigdor (2011) show that the most selective schools have on average a greater share of Black students than moderately selective schools: diversity at the top comes at the expense of diversity one tier down.
Why is this relevant? It is well established that increased diversity increases cross-racial interaction. However, there is also strong evidence that similarity in academic background matters for cross-racial interaction. With affirmative action leading to larger within-school racial gaps in academic background, more interaction may be occurring at elite institutions at the expense of even more interaction at less elite institutions. And the evidence is actually quite striking. Using data from the National Longitudinal Survey of Freshmen (NLSF), Arcidiacono, Aucejo, Hussey, and Spenner (2013) show that the share of other-race friends for Blacks is roughly the same in high school as in college. But the high schools Black students attend have a much larger share of Blacks than the colleges they attend so one would have expected Blacks to have more other-race friends in college. Indeed, they show that the share of same-race friends is positively related to the share of same-race representation at the school, which implies that colleges are actually worse at promoting cross-racial friendships than high schools. And part of the explanation is differences in academic background: increasing the SAT scores of Black students raises the share of other-race friends for Blacks; decreasing the SAT scores of White and Asian students increases their share of Black friends.
That similarity in academic background matters for cross-racial friendships is also seen in Carrell, Hoekstra, and West (2019). Using data from the Air Force Academy, they show that White students who are randomly assigned a Black roommate as freshmen are substantially more likely to have a Black roommate in the future. However, this effect is driven entirely by Black roommates in the top two terciles of the high school performance distribution: the authors found no increases for those assigned a Black roommate from the bottom tercile.
More promising are the later-life benefits from diversity, though the evidence is limited. The most compelling evidence comes from the medical profession. In Alsan, Garrick, and Graziani (2019), Black subjects were recruited for free health screenings where they were randomly assigned either a Black or White doctor. The Black and White doctors gave similar prescriptions following the health screenings, but Black subjects were significantly more likely to follow the advice if they were assigned a Black doctor. Hence a diverse set of doctors may be important for health outcomes. Further benefits could also accrue through the types of communities where Black doctors choose to serve.
But here too, it is important to come back to the effects on major choice where mismatch effects may lower the probability of majoring in STEM. To the extent that being a STEM major increases the probability of going to medical school, then increasing diversity at elite undergraduate institutions may come at the cost of a less diverse medical profession. At the same time, if going to a more elite undergraduate school increases the probability of going to medical school, then the effects may go in the other direction.
The evidence cited above emphasizes that studying the effects of affirmative action requires treating the subject with nuance. And doing affirmative action well requires good research on the topic, necessitating that universities make their data available to researchers and their prospective students. Yet, Supreme Court precedent related to affirmative action encourages universities to do precisely the opposite.
II. The University of Michigan Cases and Perverse Incentives
While holistic admissions criteria have been employed at selective universities in the United States for at least the past one hundred years, the Supreme Court decisions in Gratz v. Bollinger and Grutter v. Bollinger further encouraged universities to use them.
The key issue in Gratz was the University of Michigan’s use of race in a points system that determined admission. URM applicants to Michigan’s undergraduate program were automatically awarded twenty points, where one hundred points were needed to gain admission. The Court ruled that this points-based system was unconstitutional on the grounds that “the diversity contributions of applicants cannot be individually assessed” because the “automatic distribution of 20 points has the effect of making ‘the factor of race . . . decisive’ for virtually every minimally qualified underrepresented minority applicant.”
At the same time, the Supreme Court ruled in Grutter that the University of Michigan Law School’s “narrowly tailored use of race in admissions decisions to further a compelling interest in obtaining the educational benefits that flow from a diverse student body” was constitutional. The law school had employed holistic criteria, and the Court held such an approach to be permissible.
The two court cases emphasized that a transparent and formulaic application of minority preferences is unconstitutional. Instead, the Court argued, the use of race needs to be “narrowly tailored” so that the racial preference is not “decisive.”
With no explicit definition of “narrowly tailored,” universities in the United States have obfuscated their admissions criteria as they relate to race. This obfuscation has had the unfortunate consequence of hiding other admissions practices, like favoring legacy applicants or applicants connected to donors. The 2019 college admissions scandal exposed the various ways applicants from well-off families were able to circumvent the admissions process at many elite universities, including by taking advantage of favorable admissions policies for recruited athletes.
In addition to hiding other admission advantages, the use of holistic criteria to support a “narrowly tailored” system encourages the use of other arbitrary criteria like the rating of personal qualities. Documents in SFFA v. Harvard—hereafter, the Harvard case—revealed that, even after comparing applicants on equal grounds, Asian-American applicants were rated as having worse personal qualities, while URM applicants were rated as having better personal qualities. Deposition evidence from the Harvard case revealed that the personal qualities measured by this rating are things like whether the applicant is “a very attractive person to be with and have in your school community and widely respected.” Furthermore, Harvard’s admissions director admitted that the criteria used to produce this rating are “not terribly helpful.”
Finally, the Supreme Court decisions emphasize the importance of racial/ethnic diversity as a worthy goal of educational institutions. However, there is evidence that elite universities may be interested only in diversity reflected in skin color as opposed to lived experiences. For example, Massey et al. (2007) show that first- and second-generation immigrants of African descent make up an outsized share of Black students at selective universities, and especially among Ivy League institutions. They also show that these students are more likely to come from two-parent households and have higher-educated fathers than Blacks who descend from slaves.
Further evidence on this point comes from Paul Tough, who documents a dispute at Cornell University where a Black student group protested the lack of representation of Blacks with more than two generations in the United States.1 These patterns are also reflected in the Harvard case, which shows that, while White and Asian American applicants who are socioeconomically disadvantaged receive an admissions boost, Blacks who are disadvantaged receive no such boost.2
III. Evidence about Affirmative Action from SFFA v. Harvard
The lack of transparency in college admissions has made it difficult to understand the importance of affirmative action in determining admissions outcomes. Gratz outlawed the use of a points system, meaning that universities must now rely on a less transparent approach to implement affirmative action. In this setting, the most straightforward way to determine the impact that race has on admissions is to compare admissions outcomes across applicants of different races who are otherwise observably identical. The richer the admissions data in terms of applicant attributes, the more precise these comparisons across race can be.
However, colleges and universities are reluctant to publicly release their admissions data, and as a result, even the best data sources often lack key applicant details. Estimates of the impact of affirmative action on college admissions are then criticized since they cannot account for all of the relevant factors affecting admissions. As part of the SFFA v. Harvard lawsuit, Harvard was required to release detailed applicant-level data for all applicants to the classes of 2014 through 2019 to two expert witnesses for analysis.
Harvard’s applicant-level data is extensive, capturing information about an applicant’s academic, social, and family background, in addition to the final admissions outcome. (For additional details, see Arcidiacono, Kinsler, and Ransom, 2020.) A unique feature of the data is the availability of Harvard’s own internal rating of each applicant on a host of dimensions, including an academic, extracurricular, personal, and overall rating, among others. Using this detailed data allows estimation of a model of admissions outcomes where the impact of race can be estimated holding fixed hundreds of other applicant attributes. In other words, the model implicitly compares the predicted admissions probability for White and Black applicants who have similar SAT scores, high school GPAs, family backgrounds, intended major, internal Harvard ratings, and many other observed attributes.
The key advantage of using the Harvard admissions data to estimate the scope of affirmative action is the detailed nature of the applicant data. However, when assessing the impact that race has on admissions outcomes, it is imperative not to control for applicant attributes that already embed racial preferences. At least two of Harvard’s internal ratings suffer from this issue: the overall and personal rating. The overall rating is a direct measure of an applicant’s likelihood of admission, and thus incorporates Harvard’s preferences over race and other applicant attributes. The inclusion of the overall rating in an admissions model will thus lead to an understatement of the impact of race in the admissions decision. There is also ample evidence that racial preferences influence the personal rating. As a result, the most reliable admissions model (for the purpose of measuring the extent of racial preferences) should exclude the personal rating.
In addition to excluding certain applicant attributes, attention should also be paid to the sample of data used to estimate the model. For example, applicants of special interest to Harvard—those who are recruited athletes, legacies, applicants on the dean’s list, or children of faculty and staff (ALDC)—receive large preferences in admissions. One could expect racial preferences to work differently among this group of applicants whose predominant characteristic is belonging to one of these special groups. By focusing only on typical applicants, the comparison between similarly situated applicants is more appropriate. Note that 97 percent of Black and Hispanic applicants are typical (non-ALDC) applicants.
Estimating a logistic model of admissions for six Harvard admissions cycles that includes hundreds of applicant attributes shows that affirmative action has a large and statistically significant impact on admissions probabilities for typical applicants. In the absence of affirmative action, the average Black or Hispanic applicant would have admissions probabilities equal to 2.25 percent and 2.29 percent, respectively. While these numbers seem small, the admit rate for typical White applicants is also small at only 5.5 percent. However, with affirmative action in place, the admissions probabilities for the average Black or Hispanic applicant rise to 9.54 percent and 7.14 percent, respectively. Thus, affirmative action quadruples the admissions chances of the average Black applicant and triples the chances for the average Hispanic applicant.
To put the magnitude of affirmative action in context, the admissions bonus associated with being a Black applicant relative to being White is only slightly smaller than the admissions bonus for receiving the highest possible academic rating relative to receiving the median academic rating. The highest academic rating category is reserved for “genuine scholar(s)” with “near-perfect scores and grades” and “evidence of original scholarship.” The median academic rating is given to “very good student(s) with excellent grades and mid-600 to low-700 [SAT] scores.” Only 0.43 percent of all applicants receive the highest academic rating. Thus, belonging to an underrepresented minority group confers nearly the same benefit as being labeled an academic superstar or the equivalent of being among the one hundred academically strongest applicants to Harvard each year.
A standard concern when modeling racial differences in college admissions or other economics outcomes, such as wages or hiring, is that the estimated effect of race is overstated as a result of unobserved applicant characteristics. For example, in the context of Harvard admissions, the large positive impact on admissions associated with being Black may reflect unobserved differences between Black and White applicants. While this concern is somewhat muted since the Harvard data set includes so many applicant attributes, it is impossible to completely incorporate all of the subtle details of an applicant’s file into statistical data. However, it is possible to explore whether Black and Hispanic applicants tend to be positively selected on the observable characteristics associated with admissions, since this might suggest that they are also positively selected on the unobserved determinants of admissions. Our model of Harvard’s admissions decisions reveals the opposite pattern, showing that Black and Hispanic applicants are significantly weaker than White applicants on the other observable factors that predict admissions. Thus, there is little reason to believe that Black and Hispanic applicants have unobserved attributes that make them more attractive than White applicants. On the contrary, if anything, the estimated magnitude of Harvard’s racial preferences is likely to be understated.
IV. SFFA v. Harvard Opened the Black Box
Whether the application of affirmative action observed at Harvard is too weak, too strong, or on target remains an open question. What is indisputable is that SFFA v. Harvard provided an unprecedented glimpse inside the college admissions process. Admissions criteria and decision-making are tightly guarded secrets in higher education, and never before have researchers been able to study admissions outcomes in such great detail. The patterns uncovered at Harvard suggest that there are many other admissions practices in addition to affirmative action that require additional scrutiny.
One such practice is the favored treatment athletes receive in college admissions. Although large state universities tend to dominate the most commercialized college sports, small elite private colleges actually field more varsity teams. For example, the University of Georgia fields nineteen varsity teams, while Williams College, which enrolls less than one-tenth the number of students as the University of Georgia, fields thirty-two varsity teams. Harvard is an extreme example of this, offering a nation-leading forty-two Division I intercollegiate sports teams. In fact, recruited athletes account for 10 percent of the admitted class to Harvard.
The emphasis on athletics at the most prestigious academic institutions in the country seriously distorts the admissions process. In Arcidiacono, Kinsler, and Ransom (2020), we use domestic applicants and admits to Harvard to highlight three specific features of recruited athletes that are worthy of attention. First, admitted athletes are significantly weaker academically than their non-athlete peers. For the Harvard classes of 2014 through 2019, just 25 percent of admitted recruited athletes receive an academic rating in the top two categories as defined by the Harvard admissions office. The corresponding number for non-athlete admits is 81 percent. Second, recruited athlete admits at Harvard are overwhelmingly White and economically advantaged relative to their peers. 69 percent of athlete admits are White, while the White share of all domestic admits is only 45 percent. Finally, 26 percent of recruited athletes who attend Harvard come from families earning more than $500,000 per year, while the corresponding number is only 15 percent among all enrollees.
Recruited athletes are not the only advantaged applicants to receive preferential treatment in university admissions. Our research reveals that applicants whose parents attended Harvard, donate to Harvard, or work at Harvard (LDC applicants) receive large boosts to their admissions chances. LDC applicants are admitted to Harvard at a rate of 34 percent, while for typical applicants (non-LDC and non-athlete) the admit rate is only 5 percent. Similar to recruited athletes, LDC applicants are overwhelmingly White and come from high income families. For example, nearly 70 percent of legacy applicants are White as compared with only 40 percent of typical applicants. Two percent of LDC applicants to Harvard are labeled as disadvantaged by the admissions staff, and yet 13 percent of typical applicants are disadvantaged.
The admissions advantage that athletes and LDC applicants receive is enormous. We estimate that 75 percent of White athlete and LDC admits would have been rejected absent these preferences. While this number suggests significant inequality in access to elite universities, more alarming is the fact that preferences for special applicants have been growing over time. In another paper, we demonstrate that recruited athlete and legacy applicants for the Harvard class of 2000 were four times more likely to be admitted than their non-athlete, non-legacy counterparts. By the class of 2017, recruited athlete and legacy applicants were nine times more likely to be admitted. Despite the growing number of highly competitive non-athlete, non-legacy applicants, Harvard has essentially kept the share of legacies and athletes in its admitted class fixed over time. This implicitly means a growing admissions advantage for these groups.
While athletes and other special applicant groups receive large and growing admissions preferences, there is also evidence that typical Asian-American applicants face a higher admissions bar relative to other typical applicants (non-athlete, non-LDC). Race-based affirmative action necessarily disadvantages Asian-American applicants relative to Black and Hispanic applicants. However, it is also the case that Asian-American applicants are disadvantaged in admissions relative to White applicants. As a group, Asian-American applicants are significantly stronger than White applicants in terms of academic preparation and extracurricular participation, and are also more likely to come from disadvantaged backgrounds. Yet, typical Asian-American applicants are nearly 20 percent less likely to be admitted than observationally equivalent White applicants. Harvard’s explanation for this gap is that typical Asian-American applicants are less well-rounded than their White counterparts, pointing primarily to their lower athletic and personal ratings. Yet, the personal rating itself appears to be a primary channel through which Harvard implements race-based and other demographic preferences.
The finding that typical Asian-American applicants are disadvantaged relative to their White counterparts is relevant for the ongoing debate over affirmative action since it makes clear that race can have a detrimental impact on admissions chances. Harvard’s view, as communicated over a dozen times at trial—including by its expert witness—is that an applicant’s race is never considered a negative factor in admissions. This is a patently absurd statement because college admissions is a zero-sum game: an advantage given to one racial group is necessarily a disadvantage for a different racial group. The fact that race can negatively impact an applicant’s admissions chances does not immediately disqualify using race in admissions since many applicant attributes share this feature. However, there needs to be a compelling case for why one racial group should be favored relative to another.
On this count, the detailed admissions data revealed as part of the Harvard case also provides unique insight. A common rationale for race-based preferences is to promote a diversity of backgrounds and opinions on campus. Diversity comes in many forms, including income, geography, religion and race, to name a few. While universities may pursue diversity across all of these dimensions, it appears that diversity among its Black and Hispanic admitted students is less valuable relative to other racial groups. The data show that disadvantaged White and Asian-American applicants receive large admissions preferences relative to their more advantaged counterparts. However, disadvantaged Black and Hispanic applicants are treated identically to their more advantaged counterparts. This is an odd pattern since there is no reason to believe that the diversity of experiences across income groups among Black and Hispanic applicants is any smaller than for White and Asian-American applicants. Harvard’s use of race, specifically for Black and Hispanic applicants, is instead more consistent with race simply being a plus factor that is incorporated regardless of context. If this is how race is going to be utilized in a holistic process, a more transparent approach would be to simply assign points based on race.
Affirmative action has been a primary focus of policy and litigation involving college admissions. Numerous states have banned the use of race in admissions (Washington, Florida, Michigan, Arizona, etc.), while other states have implemented alternative policies, such as percent plans (Texas and California). Yet, the admissions data from Harvard reveal other practices that are worthy of our attention. Admissions preferences for athletes, legacies, and donors seem at odds with a social system built on the idea of meritocracy. Moreover, penalizing Asian-American applicants because of their race is discriminatory and inconsistent with the notion of diversity.
In the conclusion of the court’s ruling in the Harvard case, the judge declared that Harvard employs “a very fine admissions program.” The evidence cited above indicates that Harvard’s admissions process is littered with special interests that would be condemned in most other settings (e.g. firm hiring). Rather than holding up Harvard’s process as a beacon for other schools, we should be working to eliminate these special interests and return to a more transparent and less arbitrary system.
One step in this direction has come very recently. In 2020, the U.S. Department of Justice alleged that Yale University employs racially discriminatory admissions practices. While the litigation process is still early on, this investigation has revealed admissions patterns that are strikingly similar to Harvard’s. For example, admissions rates by race among those applicants in the top 10 percent academically are nearly identical across the two universities. Additionally, Yale and other elite universities utilize a multistage review process that resembles Harvard.
Conclusion
In this Essay, we have argued three points. First, we claimed that optimally constructing affirmative action in higher education requires nuance. Second, we contended that an unintended consequence of the two most recent Supreme Court rulings is that there is much less transparency about how universities make admissions decisions. This lack of transparency makes it nearly impossible to know if affirmative-action programs are fulfilling their goals. Third, we showed the value of the data exposed by SFFA v. Harvard in bringing to light many questionable admissions practices.
While obtaining access to Harvard’s admissions data was valuable to understanding the key questions we have investigated, so much more could be done to evaluate the effects of affirmative-action programs. Much of the information needed to do so comes from outcomes that occur after the admissions decision: grades, majors, social interactions, graduation, and post-college labor market outcomes. We emphasize that it is well within the means of almost all major universities to publicly provide this information. But until such information is provided, the affirmative-action debate will continue without sufficient information to do it well.