In June 2023, the Supreme Court of the United States outlawed affirmative action in the landmark decision Students for Fair Admissions v. President and Fellows of Harvard College}(“SFFA”), ending 62 years of the most divisive admissions policy in modern American history.1 In the majority opinion, Chief Justice John Roberts wrote, “eliminating racial discrimination means eliminating all of it.”2 Harvard said it would comply; indeed, as a recipient of federal funds, anything short of complying would be unlawful.3
Harvard’s initial numbers after SFFA appear like compliance. They are not. The university’s behavior indicates an institution with something to hide: delayed releases, a manipulated methodology for reporting race, missed deadlines, and general antagonism. In a landscape where the Justice Department has opened another university’s files and found systematic evasion of the Court’s ruling, Harvard has an imperative to demonstrate that it is dedicated to the individual applying, not their skin color.
In the aftermath of SFFA, the thumb came off the scale, and the numbers moved, though not on the schedule anyone expected. In the Class of 2028, the first admitted without illegal racial preferences, African American enrollment fell from 18 to 14 percent. Hispanic enrollment rose, from 14 to 16 percent. Asian American enrollment did not move at all, holding at 37 percent. The following year Asian American enrollment climbed to 41 percent, African American enrollment fell to 11.5 percent, and Hispanic enrollment fell to 11 percent. But even then, the statistics are not fully trustworthy. This is because the two statistics are fundamentally incomparable. For the Class of 2027, Harvard reported each race as a share of the entire freshman class. Starting with the Class of 2028, however, it reported each race as a percentage of those who answered the question, inflating the numbers. Suspiciously, the new method was implemented in the same years as when SFFA came into effect.
To recap on the trial records: Harvard’s files, forced into evidence, showed Asian Americans getting systematically lower “personal” ratings for courage and likability, despite the strongest academics and extracurriculars of any group.4 Harvard’s own researchers found the same tilt in 2013, in findings that never left secretive administrative hands. The plaintiffs’ expert modeled the odds: an Asian American with a 25 percent shot of admission would have a 35 percent chance as a Caucasian student, 75 percent as a Hispanic, and 95 percent as an African American student.5 The Supreme Court struck this race-based model down.
When the Supreme Court decided that racial discrimination was banned by the Civil Rights Act, the Crimson editorial board mourned that “diversity seems to have been offered a spot on the waitlist.” When Harvard renamed its DEI office, columnist (and now President of the Crimson) Matteo Diaz wrote: “So much for defiance.” When the DEI website came down, an op-ed said Harvard “cares neither for its own stated values nor its student body.” The president of the Black Men’s Forum said Harvard has “fail[ed] to support Black voices while courting conservative ones” after the university created a conservative institute, as if the two races affected by SFFA were “black” and “conservative.” This griping is baseless because it is divorced from a real conception of merit. The person who benefits never comes up in these articles, and neither does any of the evidence that their success was unjust.
We know who was favored and why they were favored: Neither fact favors Harvard. First, take the beneficiaries. The numbers the litigation pried out are the most recent Harvard has let anyone see. Seventy-one percent of its African American, Hispanic, and Native American admits counted as “advantaged” under Harvard’s own measure, while only about a third of these populations are socio-economically disadvantaged. That means many of those classified as disadvantaged based on their race are advantaged socio-economically. A post-trial study modeled six years of Harvard’s applicant files, holding grades, scores, and ratings constant and asking how each applicant would fare if they were Caucasian. If African American and Hispanic applicants were affluent, racial preference would affect their application substantially more than if they were working class.6 Harvard’s policy effectively rewarded wealthy African Americans and Hispanics at the expense of impoverished ones.
Then there is what Harvard’s racial preference bought. Harvard sold its racial prejudice as economically beneficial, something that would help out the consulting and banking firms for which classes train Harvard’s students. But this was a dishonest sell. The most-cited business case Harvard relied upon to defend racial discrimination, four McKinsey studies tying diverse executives to higher profits, fell apart when two accounting professors tried to reproduce the studies and concluded they “should not be relied on” due to methodological errors.7 A meta-analysis of 146 studies found that racial and gender diversity has basically no effect on team performance; what moves innovation is the training and knowledge of the person, not their skin color.8\ The Court found Harvard’s claimed diversity benefits “not sufficiently coherent” to be measured under strict scrutiny, not exactly a vote of confidence.9
Research does, however, show that at least a certain kind of diversity is important: diversity of viewpoint. Six social scientists, Jonathan Haidt and Philip Tetlock among them, wrote the landmark paper: viewpoint diversity drives “creativity, discovery, and problem solving,” and academia spent fifty years destroying that intllectual diversity.10 Harvard included. In the Crimson’s 2025 faculty survey, just nine percent of faculty called themselves conservative while 63 percent called themselves liberal. In FIRE’s 2024 and 2025 free speech rankings, Harvard finished dead last, at 0.00. Government professor Harvey Mansfield made the point in the Crimson: “affirmative action only allowed for the admission of its supporters.” Harvard is now courting ten-million-dollar gifts for individual “viewpoint diversity” professorships, money that could, efficiently spent, pay for twenty to thirty chairs under a conservative provost. So the diversity the campus defends is the useless kind. Meanwhile, the kind that actually helps is bought by Harvard for ten million dollars a chair, when, if the administration took it seriously, they could have it for free.
If that were the whole story, it would have ended with the Court’s opinion. It did not, because Harvard refuses to give up on racial discrimination. The university broke nearly seventy years of precedent and sat on the Class of 2029 demographics until a federal deadline pried them loose. Harvard changed its math too. Starting with the Class of 2028, it counted race only among students who chose to disclose it. Eight percent did not, double the prior year. Multiracial students got tallied in every box they checked.11 The reported shares therefore run higher than the real ones, and experts told the Crimson they could not reconcile or explain the figures. Harvard’s administration, the only people who could explain the data, declined.
Harvard’s stated reason for withholding this data is student privacy. In February 2026 the Crimson’s editorial board backed the refusal on this ground, under the headline “Student Privacy Is Worth Protecting.” Privacy, of course, is a right. But it should be noted that the brave defenders of freedom at the Crimson only seem to recognize this right when it is consistent with their defense of racial discrimination. Eighteen months earlier, the pressure had run the other way, and the Crimson was not so defensive of our rights. Back then, the African and African American Resistance Organization launched a Release the Data campaign, with nearly thirty members and affiliates emailing the dean and Harvard’s President for the same demographic data. An organizer wrote in the Crimson that Harvard kept leaving them “in the dark.” Same records, same students, same disclosure. To the Crimson, public data is a right when it might preserve racial discrimination, but it becomes a “privacy threat” when it might engender fairness.
The stakes could not be higher, because Harvard is not the only school that was asked for its data, it was just the only school intent on lying about it. Duke, Yale, and Princeton showed their data: African American enrollment held; Asian enrollment fell. SFFA sent all three warning letters: “your racial numbers are not possible under true race neutrality.” Harvard’s first class invites the same question. The trial record identified Asian Americans as the group affirmative action harmed most while African Americans and Hispanics were the students whose applications were most aggressively boosted under the previous, explicitly race-based system. However, in the first class admitted without racial discrimination, it was instead Hispanic enrollment that rose, while Asian American enrollment did not move at all. The reason for this is clear—covert racial discrimination.
This can happen because Roberts’ opinion leaves a university free to weigh what an applicant writes about how race shaped his life and a university free to consider that race-based story. Thus, a class that barely moves is not legal proof that anything improper happened. Maybe, the school can facetiously argue, racism just ballooned in the 2023-2024 period. They had no choice. Whether Harvard discriminated based on race is, according to the new guidelines, a result no one outside Harvard can check, and the year it happened is, as it just-so-happens, the year Harvard changed how it counts race.
As a result, in a post-SFFA world, only file-level data can prove racial discrimination. Aggregate data is a portrait of a class after it has been assembled: how many students, from which groups, with what average scores. It can tell you what a class looks like. It cannot tell you how anyone in it got there. File-level data is the decision itself, the reader sheets and the ratings and the notes, and it is the only place where racial preference can be visible, because racial discrimination is not an aggregate statistic. It is a thumb that a person unjustly presses on an individual application at an individual time. But Harvard refuses to hand over this data. It has over two-thousand pages but refuses to hand over even one.
We all know why Harvard is engaging in this obfuscation. It’s because the alternative is to invite federal scrutiny and, possibly, to end racial discrimination. Just look at Duke Law School. While they tried to engage in the same obfuscation, eventually, that became impossible. The Justice Department examined Duke Law’s files and found intentional racial discrimination across three cycles: African American applicants in 2025 had roughly 3.5 times the odds of being admitted as Asian American applicants with the same credentials, and the median LSAT of rejected Caucasian and Asian American applicants was higher than the median LSAT of admitted African American applicants. Duke claims it is reviewing the findings and follows the law. From Cambridge, Duke’s story could be read as a warning: publish, and you invite the same scrutiny. As a result of this fear, Harvard refuses to release the same files that proved Duke was discriminating based on race. These do not seem like the protestations of an innocent institution.
Harvard cannot afford to destroy our limited public trust for the sake of unjust racial discrimination. Harvard’s administration has spent three years telling the public that Washington is acting in bad faith, and perhaps they are. But that defense is only available to an institution that is itself acting in good faith. But that would be an institution which discloses even when disclosure is inconvenient. That’s not Harvard. The administration kept its own researchers’ findings of bias secret until litigation pried them out. It broke seventy years of precedent to stop releasing real admissions data. The pattern is not that Harvard lies. It is that Harvard reports when reporting helps and goes quiet when it does not. If Harvard is following the law, it can report the data that proves it. But it hasn’t, and it won’t.
Merit is what the Court ordered Harvard to measure: one applicant at a time, as a person, not a shade of color.12 The first post-ruling numbers look like compliance. But the conduct looks like a school that would rather nobody check whether its numbers are real. The delayed releases. The friendlier denominator. The missed deadlines. The litigation. Three years on, the government is still asking Harvard for proof a confident school would have volunteered. The colors are on the brochure. The person is still waiting to be seen.
Students for Fair Admissions, Inc. v. President and Fellows of Harvard College, 600 U.S. 181 (2023).
SFFA v. Harvard, 600 U.S. at 206 (Roberts, C.J.).
Lawrence S. Bacow, Letter to the Harvard Community, June 29, 2023.
Findings of Fact and Conclusions of Law, SFFA v. Harvard, No. 14-cv-14176, Dkt. 672 (D. Mass. Oct. 30, 2019).
Expert report of Peter Arcidiacono, SFFA v. Harvard, as reported in BBC News, June 15, 2018.
Peter Arcidiacono, Josh Kinsler, and Tyler Ransom, “What the Students for Fair Admissions Cases Reveal about Racial Preferences,” Journal of Political Economy Microeconomics 1, no. 4 (2023): 615-668.
Jeremiah Green and John R. M. Hand, “McKinsey’s Diversity Matters/Delivers/Wins Results Revisited,” Econ Journal Watch 21, no. 1 (2024): 5-34.
Hans van Dijk, Marloes L. van Engen, and Daan van Knippenberg, “Defying Conventional Wisdom: A Meta-Analytical Examination of the Differences Between Demographic and Job-Related Diversity Relationships with Performance,” Organizational Behavior and Human Decision Processes 119, no. 1 (2012): 38-53.
SFFA v. Harvard, 600 U.S. 181 (2023).
José L. Duarte, Jarret T. Crawford, Charlotta Stern, Jonathan Haidt, Lee Jussim, and Philip E. Tetlock, “Political Diversity Will Improve Social Psychological Science,” Behavioral and Brain Sciences 38 (2015): e130.
”A Brief Profile of the Class of 2029,” Harvard College Admissions, race and ethnicity reporting note.
SFFA v. Harvard, 600 U.S. 181 (2023).