You Don’t Need Our Education – We Must Have Cohort Control

Race-conscious admissions face their D-Day

In the next few days, the Supreme Court is expected to deliver its decision in two of this year’s highest profile cases, potentially banning racial discrimination by colleges in admission decisions. The cases pit activist group Students for Fair Admissions, whose membership includes thousands of students and parents who claim that they or their children have been rejected from elite colleges due to their race, against Harvard (from which I received my PhD last year) and the University of North Carolina.

Law (What is it good for?)

Compared to the thorny disputes that often characterize cases that reach the Supreme Court, the disputes in these two cases are refreshingly simple. SFFA alleges, backed by copious statistical evidence, that Harvard’s undergraduate admission process discriminates against Asian-American applicants compared to otherwise equivalent White applicants. SFFA argues that this places Harvard in violation of Title VI of the Civil Rights Act of 1964, which states that “No person in the United States shall, on the ground of race, color, or national origin… be denied the benefits of… any program or activity receiving Federal financial assistance.” Supposing Harvard does use race in admission decisions, and rejects Asian-American applicants who would be accepted if only their race differed, it is hard to imagine how the violation of the plain meaning of the statute could be clearer. Earning a degree is a core benefit provided by universities to the public, which someone rejected from admission to the university is denied. And like almost every college in the country, Harvard receives federal funding. Universities, even including the most elite institutions such as Harvard, are not above the law.

The lawsuit against UNC-Chapel Hill is similar: SFFA alleges that UNC exhibits an overwhelming tendency to admit underrepresented minority students at the expense of White and Asian-American applicants with superior academic credentials. Since UNC is a public university, SFFA argues that in addition to violating Title VI, these admission procedures are unconstitutional under the Equal Protection Clause of the Fourteenth Amendment – as in Brown v Board of Education, which found racial segregation in public schools illegal, arms of the government are not generally permitted to discriminate based on race.

Harvard (and UNC), however, take a very different view. They point to over four decades of Supreme Court precedent from the Bakke decision onward, through which the Court has upheld race-conscious policies that consider race as one factor among many in admissions. Harvard argues that it has a compelling interest in obtaining the benefits that stem from having a student body that is diverse along many dimensions including race, and that it would be unable to achieve a comparably diverse class without considering race as a factor in admissions.

In practice, this is indeed the current operative law of the land. But it is also subject to fierce opposition from those, like Justice Thomas, who maintain that “a state’s use of race in higher education admissions decisions is categorically prohibited by the Equal Protection Clause” and “that constitutional imperative does not change in the face of a ‘faddish theor[y]’ that racial discrimination may produce ‘educational benefits.’” Consistent with this, opponents of race-conscious admissions argue that these practices which deny people equal treatment because of their race are unfair and immoral. They further argue that the precedents upon which race-conscious admissions rely are a legally incoherent mess stemming from justices imposing their political preferences over the plain text of law. 

Indeed, it is difficult to read the jurisprudence regarding race-conscious admissions as being anything other than logically incoherent. In the Bakke case, the Supreme Court declared that racial quotas were illegal, but that other forms of race-conscious admissions, such as assigning bonus points based on race in an admissions formula, were permitted. Even leaving aside how this could possibly be faithful to either Title VI or the 14^th amendment, this makes little sense. As economics students are taught in entry-level classes, for any quota system, there is an equivalent set of ‘prices’ (typically taxes, or here, point penalties) that yields an identical result. Indeed, a few decades later, in Gratz, the Supreme Court declared that mechanical and predetermined point penalties based on race in admissions formulas were also unconstitutional. However, on the same day, in Grutter, the Court said that using race in admissions was lawful provided race was merely one factor used among many. In other words, the Court gave its blessing to colleges that used race-conscious formulas to decide admissions provided that the formulas were never written down in public. Later decisions in Fisher I and Fisher II essentially left this subterfuge intact. While universities were ostensibly not allowed to implement racial quotas, in practice the law only required that they self-attest to knowing that more minority students were needed to achieve the benefits of diversity, and that the requisite racial composition (heaven forbid anyone call it a quota) could not be costlessly achieved by race-neutral means, in order to legitimize use of these unwritten race-conscious admissions formulas. 

This brings us to the present day, with Harvard and SFFA presenting very different claims about whether Harvard discriminates based on race, and whether its procedures are legal. Harvard axiomatically declares it does not discriminate, because race is only ever used as a ‘plus factor’ in admissions, and no-one is ever harmed due to their race. This is nonsense - the sort of obviously ludicrous claim that only an Ivy League college can utter with a straight face. Clearly, in a zero-sum game like college admissions where relative position matters, giving a bonus to one group is identical to giving all others a penalty. Indeed, Harvard complains that banning racial preferences would be disastrous because it would cause the number of Black and Hispanic students it admits to decline sharply. Yet, if this occurred, other applicants would obviously be admitted instead; applicants who presently are being rejected because they have the wrong skin color. Furthermore, few would accept this ‘plus factor’ argument in any other context. If a firm were to hire unqualified White workers but refuse to hire qualified Black workers, a claim that it was not discriminating, but was merely giving White applicants a bonus in hiring, would be rightly scoffed at. Simply put, as commonly understood, discrimination is inherently a claim of disparate treatment compared to some other group. Whether or not discrimination is occurring cannot be altered by merely switching which group is treated as the reference point.

Under Harvard’s ‘plus factor’ logic, it could of course have attempted to rebuff SFFA’s claims by declaring that what SFFA saw as discrimination against Asian-American applicants was merely some or many White applicants receiving a racial bonus. After all, Grutter’s blessing of race-conscious admissions never held that Title VI or the 14^th Amendment stipulated that all racial groups must be treated at least as well as Whites. But while university officials openly admitted to giving such racial ‘tips’ to Black and Hispanic applicants, when forced to defend itself in court, Harvard made a more restrictive argument by flatly claiming that it treats Asian-American and White applicants equally, implicitly accepting that discriminating against Asian-Americans would be illegal. It is interesting to speculate why. Of course, Harvard may truly believe the claims that it discriminates against Asian-American applicants are meritless. But several other possibilities are intriguing. Perhaps Harvard views discriminating against Asian-Americans as a necessary evil to get the class composition it desires but, just as Ivy League colleges are seemingly embarrassed by their historical discrimination against Jewish applicants in the early-to-mid 20^th century (the original raison d'être for ‘holistic admissions’ practices), it is embarrassed to admit it. Perhaps some things must be done but do not bear talking about, especially among those who live in deeply progressive communities where yard signs professing “Hate Has No Home Here” and “Stop Asian Hate” abound. Another possibility: proponents of affirmative action have generally lauded Grutter as allowing underrepresented racial minorities such as Black, Hispanic and Native American students to achieve greater representation at elite colleges, and decried opposition to this as racist. Openly using Grutter to benefit Whites at the expense of Asian students would risk undermining support for affirmative action and making the Supreme Court precedents that allow it politically easier to overturn.

Admissions by the Numbers

Irrespective, since Harvard and SFFA implicitly seemed to agree that it would be unlawful for Harvard to discriminate against Asian-American applicants relative to Whites in admissions, whether Harvard in-fact does is the core of SFFA’s lawsuit and was the main focus of the lawsuit’s lower court trials. Indeed, the evidence presented in the case largely revolves around two competing statistical analyses of Harvard’s admissions procedures, where both Harvard (via its expert witness, University of California, Berkeley Economics Professor David Card) and SFFA (via its expert witness, Duke University Economics Professor Peter Arcidiacono) focus on a simple empirical question – does Harvard discriminate against Asian-Americans in admissions relative to otherwise identical White applicants?

To explore this question, it is useful to first take a ten-thousand-foot view of how the Harvard admission process works. Since Harvard uses a ‘holistic’ admissions procedure, it collects a broad set of data (including demographic characteristics such as race and legacy status) and creates ratings along a number of dimensions for each applicant. For example, each applicant is given an academic rating which incorporates performance on standardized tests such as the SAT, high school grades and class rank among other factors. Extracurricular and athletic ratings are assigned in a similar manner. In addition, applicants typically have high school teachers and counselors submit letters to Harvard attesting to their personal and academic qualities, each of which is converted by Harvard into a rating. Further, promising applicants are usually interviewed by a Harvard graduate, who assigns the applicant an alumni personal and alumni overall rating. Harvard then uses this information to assign each applicant a personal and overall rating. All of this information is then considered when making admission decisions.

Under a holistic procedure like this, race-based discrimination may arise in either of two ways. First, otherwise identical applicants of different races, who have the same characteristics and ratings on various dimensions, may have systematically different probabilities of admission. Second, the ratings individuals are assigned on specific dimensions may themselves be tainted by racial bias. For example, the same teacher recommendation letter may be assigned a different score by Harvard depending on the race of the applicant.

While it is impossible to briefly summarize every facet of the statistical back-and-forth, SFFA’s expert, Professor Arcidiacono, makes three key arguments, arguing that Asian-American applicants suffer from both forms of discrimination. First, Asian-American applicants are admitted at lower rates than applicants from any other major ethnic/racial groups (White, Black, or Hispanic) despite having the strongest performance on objective characteristics. For example, Asian-American applicants were admitted at the lowest rate in every year from the classes of 2000 through 2019, despite having the highest average SAT scores in every single year. The incongruity between admission rates and applicant strength is even more stark in the 2014-2019 admission cycles where Harvard provided more detailed data. Here, focusing on the vast majority of applicants who are ‘typical’ – main round applicants excluding those belonging to a special subcategory such as recruited athletes and children of faculty or alumni, Asian-American applicants simultaneously had the lowest admission rate despite on average having the strongest academics, extracurriculars and overall ratings from alumni interviews - and close to equal highest ratings from school teachers and counselors. In response, Professor Card, Harvard’s expert, can only quibble: for example, he complains that this is misleading because Asian-American applicants have lower athletic ratings and have by some margin the lowest personal ratings on average. Since Harvard assigns low importance to the athletic ratings outside of recruited athletes, and the personal rating is assigned by Harvard admission officers who - unlike the school teachers, counselors, and alumni interviewers - do not meet the applicants, this is not exactly a compelling response.

Second, Arcidiacono provides compelling evidence that Harvard was implementing a racial quota for Black applicants for the classes of 2017 – 2019, despite quotas having been explicitly illegal in college admissions for over 40 years. For three years in a row, Arcidiacono shows that the admission rate for Black applicants coincided almost exactly with the admission rate of all domestic applicants, a pattern that he says would arise by chance only 0.2% of the time if Harvard were not engaged in explicit racial balancing. To see the intuition behind this point, consider flipping a fair coin 50 times. On average, we expect ‘heads’ to occur 25 times, but random variation means there would typically either be fewer (e.g. 21, or 23) or more (e.g. 26, or 29) heads – getting exactly 25 is quite unlikely. Put differently, the suspicious pattern in Harvard’s admission rate for Black applicants is akin to getting exactly 25 heads three iterations in a row. Fortuitously, I studied a related question during my PhD studies, and I can apply a method I developed to the Harvard admission data to verify Arcidiacono’s claims. If anything, I find his claim to be conservative – I estimate that such a pattern would arise by chance roughly one in a thousand times if Harvard is not illegally operating a quota. Professor Card dismisses the claims of an illegal quota, but does so without providing any meaningful evidence or reason to doubt Arcidiacono’s analysis. Rather, he merely asserts that there is no reason for Harvard to do so. Perhaps he should talk to Harvard. While the university denies using racial quotas or racial balancing, it admits to monitoring admission rates by race in order to guard “against inadvertent drop-offs in representation of minority applicants”. You say quo-ta-toe, I say quo-tah-toe.

Third, Arcidiacono presents the results of formal statistical modeling that estimates the effects of race on the probability of being admitted to Harvard, controlling for a large number of other variables that Harvard may consider when deciding who to admit. Under some assumptions (and with a sufficiently large sample), these statistical regression models have an incredible ability – by comparing only the attributes of those whom Harvard admits to those it rejects, the relative weight that Harvard gives to different factors can be accurately estimated. Consistent with having lower admission rates despite being stronger on average in many of Harvard’s ratings, Arcidiacono presents evidence that Asian-Americans are subject to statistically-significant discrimination compared to White, Hispanic and Black applicants. Notably, Arcidiacono’s findings are consistent with an internal report by Harvard’s Office of Institutional Research which, analyzing admission decisions for the 2007-2016 classes, likewise found Harvard was subjecting Asian-Americans to a penalty in the admissions process. The magnitudes of the effects Arcidiacono estimates are, frankly, stunning. For example, he considers a non-disadvantaged Asian male who has characteristics that produce a 25% chance of admission. Changing only his race to White, the probability of admission would jump to 36%. And if treated as Hispanic or Black, the probability of admission would jump to 77% and 95% respectively. Not only is this evidence of discrimination against Asian-Americans, it is evidence that race is frequently a determinative factor – an Asian applicant who is a longshot becomes an almost certain admit if treated as if Black.

It is here that Card and Harvard mount the most strenuous objection to Arcidiacono and SFFA’s claims. Card presents his own array of statistical analyses, where he both criticizes a number of modeling choices made by Arcidiacono and argues that his superior analysis shows there is no clear evidence that Harvard penalizes Asian-Americans. While the models of the two experts vary in a large number of ways, a careful reading shows that several key choices drive the results.

For example, Harvard gives applicants from socioeconomically disadvantaged backgrounds a boost in admissions. However, Arcidiacono observes that while this boost is substantial for White and Asian applicants, it is far more modest for Hispanics, and approximately non-existent for Black applicants. A common narrative among proponents is that affirmative action is primarily about redressing differences in racial composition caused by economic disadvantage. But in actuality, relative to how it treats applicants of other races, Harvard appears to favor affluent Black applicants over disadvantaged ones. This is unsurprising; previous research has documented that a large share (approximately 40%) of Black students at Ivy League colleges are first and second-generation immigrants, who on average come from more educationally-rich families. To reflect this, in his model, Arcidiacono allows the effect of disadvantage on admission probability to vary by race. Card, however, does not. Instead, he argues that Arcidiacono’s use of race by disadvantaged status ‘interaction’ effects is arbitrary, lacks theoretical justification, and is thus inappropriate; but for the reasons mentioned above, this strains credulity. This causes the boost given to Asian-American and White applicants who are disadvantaged to be underestimated in Card’s model. Since the former are more likely to be disadvantaged, in aggregate this works to hide some of Harvard’s apparent discrimination against Asian-Americans.

Second and most importantly, Arcidiacono’s statistical analysis of Harvard’s personal rating suggests it is racially biased and thus serves as a vehicle for obscuring racial preferences. He accordingly excludes the personal rating from his model of admissions decisions. Card, however, includes it in his model, assuming that the personal ratings aren't racially biased. What reasoning does Card offer to support this crucial assumption? Well, Harvard declares that the personal ratings aren't racially biased! And since this clearly can be taken as gospel, he argues that the lower personal ratings Harvard assigns to Asian-Americans are most likely not actually due to racial bias, but instead presumptively reflect legitimate personal attributes that Harvard considers and where Asian-Americans score worse but that Arcidiacono’s statistical models do not observe. This is dubious at best. Arcidiacono’s statistical analysis of the personal rating accounts for approximately 200 different variables, yet the unexplained penalty against Asian-Americans is almost unchanged in aggregate. It is not clear why we should expect further statistical controls to dramatically upend this. As Arcidiacono notes, no-one in the Harvard admissions office has supported the view that Asian-American applicants tend to have worse unobserved qualities that explain their lower personal rating. If Card believes he - presumably without having met them - possesses evidence that Asian-American applicants are weaker than other groups on such qualities, he could provide clear examples. Yet, he has not. Furthermore, the addition of the controls already in Arcidiacono’s model reveal a robust, large, and generally growing bias in the personal rating in favor of Black and Hispanic applicants. Such a bias contradicts Harvard’s claim that the personal rating is race-neutral. In turn, this increases the plausibility that it is also biased against Asians, and irrespective of this, implies that it should not be used as an explanatory variable when trying to understand the role of race in admissions.

Suspiciously, the pattern of results is exactly what would be expected if Harvard were trying to hide discrimination in admissions. The easiest way to do so is to inject bias into one of the component ratings that in turn affects admission decisions. And by virtue of being highly subjective, doing so through the personal rating would be the obvious choice. Previous analysis of admissions to highly selective private colleges has found strongly suggestive evidence that Asians are penalized relative to Whites, Hispanics and Blacks in turn by increasingly large amounts. Here, in the Harvard data, the personal rating appears biased in exactly the same manner: in favor of Blacks relative to Hispanics, Hispanics relative to Whites, and Whites relative to Asians. The coincidences abound.

Nonetheless, Card employs one last rhetorical trick to argue for retaining the personal rating. When analyzing the determinants of Harvard’s academic rating, some of Arcidiacono’s specifications find a small unexplained gap in favor of Asian-Americans. Card uses this to suggest that Arcidiacono’s models of the various Harvard ratings are unreliable and cannot be trusted. This would be a clever argument if it were not so misleading. On average, Asian-Americans have substantially higher academic ratings than White, Hispanic and Black applicants in turn. Controlling for a basic set of objective measures of academic performance – such as scores on external standardized tests and high school GPA – explains fully 100% of the gap between Asian-American and White applicants’ academic ratings. It is only when subjective metrics that Harvard itself creates – such as ratings by teachers and the racially-biased personal rating – are included as controls that a slight unexplained racial gap favoring Asian-Americans in the academic rating appears, which Card then complains about. That is, relative to their (Harvard-scored) teacher and personal ratings, Asian-Americans applicants have slightly unexpectedly high academic ratings. But since the high average academic rating that Asian-American applicants receive is fully explained by objective academic measures, rather than being evidence that Arcidiacono’s models are flawed, a much simpler explanation is that either the teachers writing recommendation letters or Harvard, in subjectively turning these letters into scores, discount academic achievement of Asian applicants. In other words, on balance this is evidence of more, not less, bias against Asians. What it most certainly does not do is justify retaining the personal rating in analysis of admissions.

And indeed, Card begrudgingly admits that if the personal rating is removed from his statistical model, then he too finds a statistically significant penalty against Asian-Americans in admissions. Furthermore, even retaining the personal rating, Card’s model finds the scale of racial preferences towards other groups to be large. In particular, his model implies (although he takes pains to avoid stating) that almost 70% of admitted Black students would not qualify for admission under Harvard’s current criteria if racial preferences were removed. In aggregate, the balance of evidence is overwhelming; Harvard appears to be discriminating against Asian-Americans in admissions, and racial-preferences appear to be determinative for many of the underrepresented minority students that Harvard admits.

After Grutter?

So, what happens next? The Supreme Court is widely expected to rule against Harvard (and UNC), and there is a strong possibility that it will overrule Grutter and issue a landmark judgment outlawing the use of race as a factor in college admissions. Supposing this occurs, what will follow?

First, it is helpful to note what Harvard (and other elite colleges) claim will happen if they are forced to abandon racial preferences in admissions. Consistent with Professor Card’s implied result above, Harvard claims that the share of students who are from underrepresented racial minorities will be dramatically cut, that this cannot be acceptably remedied by using race-neutral alternatives, and that this “would severely compromise its ability to achieve the educational benefits” created by having a racially diverse student body. Indeed, Harvard argues that this harm alone provides a sufficient basis for leaving Grutter in place.

The careful reader may note several awkward tensions here. First, when a statute appears to say that an action is illegal, we typically do not take seriously arguments that the law should be ignored or allow for exemptions because people engaging in the action will be inconvenienced. Nor do we consider the outcome that could be achieved by breaking the law as the relevant benchmark against which alternatives should be judged. Consider, for a moment, a thief. Clearly, their livelihood and financial wellbeing – completely legitimate ends in the abstract - are threatened by laws that punish stealing. But we don’t consider that relevant and thus exempt thieves from laws against stealing. Nor do we act as though society has an obligation to provide alternatives to thieves that leave them at least as well off financially as they would be if allowed to thieve freely, because an entity has no moral entitlement to an outcome that they achieve by violating the law. Harvard apparently begs to differ. In Harvard’s view, race-conscious admissions must be allowed in spite of the clear language of Title VI, because otherwise Harvard will be harmed. And, despite definitely – pinky promise – not implementing racial quotas, race-neutral alternatives are insufficient because they fail to preserve the exact racial minority shares that Harvard desires and currently achieves using race-conscious preferences. Perhaps no-one put this better than Justice Sotomayor, who in oral arguments back in October, responding to a socioeconomic-status-conscious race-neutral admission scheme presented by SFFA that would increase the share of Asians, Hispanics and underrepresented minorities on campus, dismissed it with a single sentence: “Blacks wouldn’t increase”.

Second, recall that in its legal defense, Harvard persistently claims that race is only one minor consideration among many. However, when addressing the implications of banning racial discrimination, they predict a disastrous impact on black admissions. Clearly, this argument is self-contradictory - Harvard cannot have it both ways. Unsurprisingly, Justice Sotomayor amplified Harvard’s claim repeatedly during oral arguments, stating “Race is never the determinative factor…Race alone doesn't account for why someone's admitted or not admitted” and “if race is only one among many factors, how can you ever prove …that it's ever a determinative factor?” But if race were never determinative, then logic necessitates that banning its use as a factor should have no effect on admissions decisions, and universities would have no reason to care. The reality, of course, as Card finds but is loath to admit, is that race is determinative for a large fraction of minority admits. But race being determinative does not require that race is the only factor that enters admissions decisions, but rather that an individual is denied entry (or admitted) and would otherwise be admitted (rejected) if their race were different.

This reveals Harvard’s attempt to downplay the role of race in admissions by describing it as one factor among many to be pure sophistry. Fundamentally, out of all possible admissions systems, one of three things must be true. Either race must not be a factor (as SFFA desires), race must be one of multiple factors, or race must be the only factor. The latter is nonsensical; requiring all individuals of some races to be admitted, and of others to be rejected. Accordingly, any plausible racial preference, up to extremely large degrees of racial discrimination, is consistent with race being one factor among many. And contrary to the impression that elite universities often like to curate, the magnitude of racial preferences under affirmative action schemes is frequently large. For example, in the 1995-1997 admission cycles prior to the implementation of Proposition 209 in California, which banned affirmative action in Californian public universities, the median admitted underrepresented minority student at UC Berkeley had a composite measure of standardized test scores and high school GPA at the 7^th percentile of White admits, with fewer than 9 percent scoring above the median White admit. Similarly, an analysis of 2002 admissions data for the University of Michigan’s law school found that the median Black admit had an academic index at the second percentile of admitted White students.

As a result, if the Supreme Court bans affirmative action and elite universities halt using race as a factor in admissions but otherwise leave their current procedures intact, fewer Black and Hispanic and more White and Asian applicants will qualify for admission. But history and math both suggest that universities will attempt to undermine the effect of affirmative action bans by changing their admissions formulas to capture race by proxy. While the math of how to do this is technical, the intuition is simple. For example, suppose that the race of each applicant could be perfectly predicted using an extensive set of other observable information. Then universities could perfectly replicate their racial preferences without explicitly using race by instead giving preference to the characteristics that predict that an individual is from a minority race. More generally, universities can engage in racial discrimination by proxy by giving reduced (or negative) weight to any characteristic or attribute that is predictive (controlling for all other observed variables) of a student being White or Asian and increased (or positive) weight to any characteristic that is predictive of a student being Black or Hispanic.

A crude version of this is already emerging with a push towards colleges becoming standardized-test-optional (e.g. Columbia) or removing consideration of such tests entirely (the UC system). Apparently, whilever explicit racial preferences are allowed, these tests are an important tool for identifying smart students of each race, but if affirmative action is banned these tests are suddenly less useful to those who wish to engage in racial balancing. (A more subtle option is to downweight math SAT scores relative to reading/writing scores, where the Asian performance advantage is smaller). Another notable proxy ripe for exploitation is location of residence or schooling. Applicants from schools or localities with high minority shares can be given preference over those from White and Asian majority areas. Historical developments since the banning of affirmative action in California provide a likely template for how universities will attempt to undermine a nationwide ban on affirmative action. Despite racial preferences being illegal, UC San Diego curiously has an explicit goal of reaching a full-time ‘Latinx’ undergraduate share of 25%. In the 2022 admission cycle, despite a stable racial composition of applicants, it managed to engineer a one-third increase in the Hispanic admit share relative to 2021, while the White share declined by almost one-fifth. Or consider UC Berkeley, which following the appointment of a new undergraduate admissions director in 2019, implemented new “equity-centered” admissions procedures. The result, for the class matriculating in 2020? An almost 50% increase in Hispanic and Black admits, chalked up in part to greater minority outreach, despite zero net increase in minority applicants.

If the Supreme Court does ban racial preferences, it seems inevitable that courts will soon have to speak on whether discrimination by racial proxy is similarly illegal. It is difficult to see why it would not be. Even if a characteristic is, strictly speaking, race neutral, the moment it is used in an admissions scheme with the purpose of affecting the racial composition of admits, the admission scheme is race-conscious and racially discriminatory.  “I am not discriminating against you because you are Black, I am discriminating against you because you live amongst too many Black people”, said no-one in good faith ever. Progressives in the US rarely hesitate before describing historical ‘redlining’ by the Federal Housing Authority as racist and discriminatory, and the converse should hold here.

Fortunately, there is a simple conceptual test of whether use of a variable is motivated by its ability to proxy for race. If a university truly inherently cares about some attribute, it should already have been taking it into account (with appropriate weight) long prior to affirmative action being banned. Sudden new or dramatically changed interest for some characteristic once racial preferences are banned is strongly suggestive that the variable is being used for the purpose of affecting racial composition. Harvard has already publicly stated what it claims will happen to the racial composition of its student body if Grutter is overturned. When someone tells us what will happen if they don’t cheat, perhaps we should take them seriously.

While the operations of such race-conscious admissions might currently be popular on elite university campuses, representative opinion polls and the results of ballot questions in blue states including Washington and California consistently show that an overwhelming majority of the American public opposes the use of race as a factor in college admissions. It would seem that Americans agree with Chief Justice Roberts’ observation that “it is a sordid business, this divvying us up by race.” And irrespective of one’s personal view, the text of Title VI of the Civil Rights Act is crystal clear. Perhaps soon the day shall come when on college campuses at least, that “sordid” business is no longer allowed.