Introduction
In Students for Fair Admissions, Inc. v. President & Fellows of Harvard College1 (SFFA), the Supreme Court invalidated the race-based affirmative action programs at Harvard University and the University of North Carolina (UNC) under the Equal Protection Clause of the Fourteenth Amendment.2 The SFFA majority held that the programs could not survive strict scrutiny: The universities’ interests in their affirmative action programs were not sufficiently compelling,3 and the schools “fail[ed] to articulate a meaningful connection between the means they employ and the goals they pursue.”4 Additionally, the Court determined that each affirmative action program used race as a negative and a stereotype, both of which are forbidden by the Equal Protection Clause.5
Despite its repeated and forceful rejection of the use of race in college admissions, the Court ended its decision by reassuring applicants, admissions officers, and lower courts that “nothing in this opinion should be construed as prohibiting universities from considering an applicant’s discussion of how race affected his or her life, be it through discrimination, inspiration, or otherwise.”6 The Court’s concluding caveat thus established what this Note dubs the “essay exception.”7
As the divisiveness of race-conscious policies and practices in America continues to escalate,8 the legal and moral urgency of deciphering SFFA’s essay exception grows. In all likelihood, the essay exception’s mechanics and reasoning will not only determine the survival of affirmative action programs in other contexts but will also serve as a touchstone for the evolving norms about the proper use of race in everyday decisionmaking.
This Note urges that the best reading of the essay exception is as an individuation requirement designed to curb the stereotype-reinforcing practice of race-based statistical discrimination in college admissions. On this account, even accurate race-based statistical discrimination is impermissible due to its violation of the racial equality axiom, which rejects the normative relevance of race absent social mediation. Properly understood, individuation thus allows assessors to evaluate race-based generalizations but prohibits assessors from supplying them.
This Note proceeds in four Parts. Part I provides a brief explanation of statistical discrimination and its primary alternative, individuation. Part II applies this model to the Court’s opinion in SFFA to show that the Court understood affirmative action programs as a form of race-based statistical discrimination. Part III argues that the Court’s individuation requirement is justified by the unique threat that race-based statistical discrimination poses to the fundamental principle of racial equality. Finally, Part IV examines what individuation requires in college admissions.
I. Statistical Discrimination
Though it often bears heavy moral implications, “[t]o discriminate is [merely] to distinguish.”9 Because individuals make distinctions for different reasons and in different ways, economists and philosophers separate discrimination into two conceptual categories: taste-based and statistical. This Part examines statistical discrimination in depth, first defining it in opposition to taste-based discrimination, and then surveying its consequences before turning to its alternative, individuation.
A. What is Statistical Discrimination?
“Statistical discrimination generally refers to the phenomenon of a decision-maker using observable characteristics of individuals as a proxy for unobservable, but outcome-relevant, characteristics.”10 The concept is best understood when juxtaposed with “taste-based” discrimination — the “disutility caused by contact with some individuals.”11 Taste-based discrimination results from a decisionmaker’s beliefs or preferences and can range from justified to repugnant. Person A may avoid spending time with smokers because the smell of cigarettes makes her sick; Person B may refuse to sit next to a man wearing a kippah on the subway because he dislikes Jews.12 Taste-based discrimination is “foundational” because “the basis for the discrimination lies at or near bedrock, and is not justified by reference to further or deeper goals.”13 Taste-based discrimination appears in “the lexicon of constitutional doctrine as ‘animus’”;14 laws or actions motivated by animosity toward a particular group face increased scrutiny and are often unconstitutional.15
Unlike taste-based discrimination, statistical discrimination “is instrumental” because the trait on which the discrimination is based is not itself of interest to the discriminator.16 That trait is significant “to the discriminator only because of what it indicates about the likelihood that the person who is the subject of the discrimination possesses another trait.”17 If the same characteristics on which bias is based, such as race, “operate[] as proxies for other less observable characteristics,”18 rational actors might discriminate based on those characteristics even absent personal animosity. Thus, what appears to be dislike for certain groups might actually be a rational response to limited information.19
A short example illustrates this phenomenon: Two employers, A and B, both decide to hire candidate Y, a man, instead of candidate X, a woman. Employer A strongly dislikes other women — regardless of demonstrated competence or personality fit, she prefers to have male employees. Employer B, however, has no preference for the gender of her employees. But she is concerned about the return on her investment in new-employee training and thus wants to hire the candidate who is less likely to take time off in the future. Employer B has read up on the issue and knows that, on average, women take more leave than men do.20 She therefore hires candidate Y. Employer A has engaged in taste-based discrimination, while B has engaged in statistical discrimination.
Statistical discrimination consists of three fundamental components:
The proxy: an observable characteristic that is statistically indicative of the target trait but is not itself outcome relevant.
The target: an unobservable, but outcome-relevant, characteristic correlated with the proxy.
The outcome: the ultimate goal of the discriminator.21
In the example above, gender is the proxy: It’s an easily observable trait that is statistically indicative of the target, likelihood to take leave. Taking leave is inversely related to the employer’s ultimate goal, productivity. Although gender does not cause lower productivity, taking leave does. Because the employer can’t directly observe a candidate’s likelihood to take leave, she uses a proxy, gender, to make a rational choice about whom to hire.
Additional examples abound — statistical discrimination is virtually everywhere. For example, age is a common proxy.22 Licensed commercial airlines may not employ pilots over the age of sixty-five.23 That’s because “age is a reliable even if imperfect indicator of poor vision (as well as impaired hearing and slower reflexes).”24 Excluding older people from being commercial pilots creates “a commercial airline pilot cohort that has better vision, better hearing, and faster reflexes.”25 Similarly, because state legislatures believe “that youth is a statistically sound predictor of irresponsible behavior,”26 states prohibit those under certain ages from driving,27 buying alcohol,28 and marrying.29
Importantly, however, the correlation between the proxy and target may be, but need not be, causal.30 Ice cream consumption is correlated with drownings and thus might be a useful proxy for drownings despite the fact that eating ice cream likely has no causal effect on one’s likelihood to drown.31 So long as a robust correlation exists, statistical discrimination is agnostic as to its source.
B. The Rationality and Consequences of Statistical Discrimination
Though statistical discrimination is often convenient, its appropriateness will depend on its rationality and consequences. At baseline, statistical discrimination is rational only when based on an accurate correlation between proxy and target.32 For example, statistical discrimination grounded in astrology or phrenology33 has no “sound empirical basis” and is therefore irrational.34 Accurate statistical discrimination requires a “positive correlation” between the proxy and target; the relationship must be “better than random” such that “the existence of the target is more likely given the indicator than it is without the indicator.”35
Even where the underlying correlation is accurate, however, “[t]here are strong and weak correlations.”36 And, even strong correlations are never perfect.37 Most people born and raised in Mongolia will possess native fluency in Mongolian.38 But: “It is possible to be born and educated in Mongolia and not be proficient in Mongolian, and even more possible to be born outside of Mongolia and have developed native fluency in the language.”39
Thus, because even accurate statistical discrimination relies on imperfect correlations, statistical decisionmaking will invariably produce type I and type II errors — false positives and false negatives, respectively.40 Imagine an employer who hires new employees based on shirt color. Shirt color does not affect job performance, but if shirt color were correlated with an outcome-relevant trait, like diligence, such that candidates in blue shirts were more likely to do diligent work, the employer’s decision to hire blue-shirters over red-shirters would be rational. Given a relatively robust correlation, it would also likely produce predominantly justified outcomes. The employer would end up with many qualified blue-shirters and would avoid hiring many less qualified red-shirters. Even so, he would also likely hire some less qualified blue-shirters and overlook some qualified red-shirters. In this scenario, the type I error the employer commits confers an unjustified benefit on the less qualified blue-shirters and the type II error he commits confers an unqualified disadvantage on the qualified red-shirters.41 Statistical discrimination thus involves a basic wrong in the sense that it uses an outcome-“irrelevant characteristic to select someone for disadvantage or benefit.”42
Moreover, statistical discrimination has spillover effects. One such effect is the reinforcement of self-fulfilling prophecies based on the correlation between the proxy and target traits, which can strengthen the underlying correlation itself.43 If, for example, Black Americans are objects of statistical discrimination in the labor market, they will have less incentive to invest in human capital, like education, because the expected return on that investment will, on average, be lower than the return for other groups.44
Finally, statistical discrimination may impede or advance goods such as equality, dignity, and democracy.45 Consequentialists and non-consequentialists alike must determine whether these goals are better served by “ignoring even statistically relevant differences”46 or by heeding them.47 Often, this analysis will depend on the alternative to statistical discrimination, which the next section takes up.
C. The Alternative to Statistical Discrimination: Informational Problems Require Informational Solutions
At its core, statistical discrimination is a rational response to an informational deficit. Unable to observe the outcome-relevant target trait, a decisionmaker relies on the observable proxy trait to reliably, though imperfectly, get at the target.48 Informational problems, however, have a straightforward alternative: Correct the informational deficit. “Individuation . . . looks at each individual separately, attempting to determine whether that individual possesses the attributes that are the object of interest.”49 In that way, it minimizes the errors and spillover effects of statistical discrimination.
Consider “driver monitoring” programs, which have become popular in recent years.50 Through plug-in devices or apps, insurance companies “keep tabs on how you drive, measuring things like speed, braking, cell phone use (to observe potential distracted driving), and time of day.”51 Insurers are thus able “to move away from using harmful socioeconomic factors to calculate insurance rates”52 because they no longer need to statistically discriminate. Before driver monitoring, observable factors like age, gender, education level, and even race were used as proxies for driving skill.53 But monitoring systems make that previously unobservable target trait directly observable. This allows insurers to conform insurance rates to risk more reliably and greatly reduces, or even eliminates, the type I and type II errors that resulted from statistical discrimination.54
Despite the obvious advantages of individuation, it also carries costs. “First, pure individuation is in the final analysis impossible.”55 This is because prediction of future behavior is always an analysis of past behavior. Even if all target traits were perfectly observable, outcomes would not be guaranteed; an employee whom the employer knows plans to take leave to raise children may decide to continue working and hire a nanny instead.56 Second, individuation is challenged by “the reality of the necessity of using proxies in a world of limited resources of time, personnel, and money.”57 That is, “[i]ndividuation requires individuators,”58 and individuators are often expensive. Finally, individuation results in outcome-relevant discrimination. Individuation rejects the command “to treat all citizens or all people the same even if they are not.”59 It means that bad drivers will pay more for car insurance and candidates who want to take leave to raise kids will have a harder time finding high-paying work. And though these costs may be significant, those of statistical discrimination may yet be higher.
II. Race-Based Statistical Discrimination in SFFA
With the foregoing primer on statistical discrimination in mind, this Part examines the Court’s opinion in SFFA to elucidate the Court’s model of affirmative action. Based on the Court’s four rationales and the essay exception, this Part argues that Court understood universities to be engaging in race-based statistical discrimination. On that model, the essay exception is an individuation requirement that curbs universities’ reliance on race-based stereotypes.
A. SFFA’s Four Rationales
The SFFA majority advanced four distinct rationales for invalidating the challenged affirmative action programs. First, the admissions programs failed strict scrutiny: The universities’ interest in using racial classifications was neither sufficiently compelling nor narrowly tailored.60 The Court explained that “[u]nlike discerning whether a prisoner will be injured or whether an employee should receive backpay”61 — situations in which the Court had upheld the limited use of racial classifications62 — “the question whether a particular mix of minority students produces ‘engaged and productive citizens,’ sufficiently ‘enhance[s] appreciation, respect, and empathy,’ or effectively ‘train[s] future leaders’ is standardless.”63 Moreover, Harvard’s and UNC’s “admissions programs fail[ed] to articulate a meaningful connection between the means they employ and the goals they pursue.”64 The schools used “imprecise”65 groupings that prioritized representation of “opaque racial categories” over meaningful diversity.66
Second, the admissions systems used race as a negative, a practice the Court’s Equal Protection jurisprudence forbids.67 Because “[c]ollege admissions are zero-sum[, a] benefit provided to some applicants but not to others necessarily advantages the former group at the expense of the latter.”68 If race was a plus for some, it was by definition a negative for others. And the Court saw this claim borne out in the data: “Harvard’s consideration of race has led to an 11.1% decrease in the number of Asian-Americans admitted to Harvard.”69
Third, Harvard’s and UNC’s admissions programs improperly relied on racial stereotypes, which is prohibited “throughout [the Court’s] Equal Protection Clause jurisprudence.”70 Emphasizing that it had “rejected the assumption that ‘members of the same racial group — regardless of their age, education, economic status, or the community in which they live — think alike,’”71 the Court held that the universities’ admissions practices did precisely that: “Harvard’s admissions process rests on the pernicious stereotype that ‘a black student can usually bring something that a white person cannot offer.’”72 The Court stressed that it had “time and again forcefully rejected the notion that government actors may intentionally allocate preference to those ‘who may have little in common with one another but the color of their skin.’”73
Because the word stereotype “is misleadingly ambiguous,”74 a brief detour is required to nail down exactly what the Court meant when it used the term. As Professor Frederick Schauer explained, stereotype might refer “to all statistical generalizations, . . . to all statistical generalizations about classes of people, . . . to those statistical generalizations that are inaccurate, . . . [or] to those statistical generalizations that, whether because of their inaccuracy or because of their other consequences, are worthy of condemnation.”75 The Court’s discussion of stereotype in SFFA suggests that it used the term in one of the first two, “non-moralized”76 senses. The Court introduced its prohibition on stereotyping by explaining that universities may not “operate their admissions programs on the ‘belief that minority students always (or even consistently) express some characteristic minority viewpoint on any issue.’”77 To the Court, the moral substance of the stereotype is immaterial, as is its accuracy. The stereotype that Asian American students are in the aggregate better than Caucasian American students at math is both positive and accurate.78 Yet, such reliance on such a stereotype is forbidden under SFFA’s prohibition on acting under the belief that “members of the same racial group — regardless of their age, education, economic status, or the community in which they live — think alike.”79
Finally, the Court considered the programs’ “lack [of] a ‘logical end point’” problematic.80 Although the universities offered various suggestions as to possible endpoints, the Court found none convincing. For example, the Court considered the schools’ “promis[es] to terminate their use of race only when some rough percentage of various racial groups is admitted,” to exacerbate and prolong the relevance of race rather than eliminate it.81 And though the schools suggested “that universities will no longer need to engage in race-based admissions when, in their absence, students nevertheless receive the educational benefits of diversity,” the Court deemed this standard so impossible to measure that a satisfactory end could never be determined.82
B. Affirmative Action as Statistical Discrimination
After offering its affirmative rationales for invalidating the challenged admissions programs and responding to the dissent, the majority closed its opinion with what this Note terms the essay exception:
[N]othing in this opinion should be construed as prohibiting universities from considering an applicant’s discussion of how race affected his or her life, be it through discrimination, inspiration, or otherwise. . . . A benefit to a student who overcame racial discrimination, for example, must be tied to that student’s courage and determination. Or a benefit to a student whose heritage or culture motivated him or her to assume a leadership role or attain a particular goal must be tied to that student’s unique ability to contribute to the university. In other words, the student must be treated based on his or her experiences as an individual — not on the basis of race.
Many universities have for too long done just the opposite. And in doing so, they have concluded, wrongly, that the touchstone of an individual’s identity is not challenges bested, skills built, or lessons learned but the color of their skin. Our constitutional history does not tolerate that choice.83
Prior to SFFA, the decisions in Regents of the University of California v. Bakke84 and Grutter v. Bollinger85 had tightly circumscribed the permissible means and ends of affirmative action.86 Read against this background, the essay exception explains how the Court conceptualized Harvard’s and UNC’s affirmative action programs within these narrow confines. The essay exception explicitly recognized that race often “affect[s one’s] life, be it through discrimination, inspiration, or otherwise.”87 For example, the text acknowledged that overcoming racial discrimination can nurture “courage and determination” and that “heritage or culture” can motivate one “to assume a leadership role or attain a particular goal.”88 And, by highlighting qualities like courage, determination, and leadership, the essay exception endorsed those attributes, and similar ones, as legitimate criteria for college-admissions decisions. Thus, although the Court never said so explicitly, its inclusion of the essay exception and the examples therein acknowledged that race may be correlated with qualities that universities may lawfully prioritize.
The essay exception thus suggests that the Court conceived of affirmative action programs, as practiced by universities pre–SFFA, at least in part as a form of race-based statistical discrimination — though it certainly didn’t speak in those terms. On this model, race was an observable proxy for unobservable target traits such as courage, determination, and leadership. Universities believed that race was correlated with courage and determination.89 They discriminated based on race not because they believed race itself was a legitimate object of decisionmaking,90 but because they believed it to be indicative of another legitimate end.
Two of the Court’s affirmative rationales reaffirm this understanding. First, stripped of its economic dressing, race-based statistical discrimination is the use of accurate racial stereotypes. Here, the correlation between the proxy and the target is the stereotype that individuals from historically disadvantaged communities often overcome difficult situations that cause development of courage, determination, and leadership. The Court’s concern about the continued use of racial stereotypes in college admissions is thus a concern about the continued use of race-based statistical discrimination in college admissions. Second, the narrow tailoring element of strict scrutiny — the requirement that there be “a meaningful connection between the means they employ and the goals they pursue”91 — maps onto the strength of the correlation and the frequency and magnitude of errors in statistical discrimination. In constitutional law terms, a weak or inaccurate correlation will cause a “mismatch”92 between the proxy and the target trait.
Although the rule against race as a negative might be a recognition of the type II errors statistical discrimination causes, the Court’s discussion of the prohibition probably fits better with a taste-based, or foundational, model of discrimination. The Court’s rhetorical question illustrates this difference: “How else but ‘negative’ can race be described if, in its absence, members of some racial groups would be admitted in greater numbers than they otherwise would have been?”93 As the Court sees it, the problem here is not that colleges are misidentifying some Asian students as timid and some Black students as courageous94 but that colleges prefer to have Black students over Asian students. If the Court understood affirmative action exclusively as a form of statistical discrimination rather than taste-based discrimination, we would expect it to focus only on the former problem and not the latter. Moreover, one might also be skeptical of the characterization of affirmative action as statistical discrimination due to the Court declining to speak in those terms. After all, when the Court is concerned about the use of proxy traits and resulting error rates, it has said so pretty explicitly.95
Neither of these objections, however, undermines the conclusion that the Court understood affirmative action, in large part, as race-based statistical discrimination. The essay exception on the relationship between skin color and desirable experiences, such as “challenges bested, skills built, or lessons learned”96 — the relationship between the proxy and the target. Even if affirmative action programs operated in part through taste-based discrimination, the Court’s explanation of its reasoning suggests that it understood these programs to also have operated through race-based statistical discrimination.
C. The Essay Exception as an Individuation Requirement
So far, the Court has made two seemingly counterintuitive assertions: First, affirmative action programs are a form of (at least plausibly accurate) statistical discrimination; race does develop courage, determination, leadership, and more through “discrimination, inspiration, or otherwise.”97 Second, even accurate race-based statistical discrimination — discrimination based on stereotypes that are true — is unlawful (and wrong). What mechanism could allow consideration of desirable attributes whose development is tied up with race and simultaneously excise reliance on racial stereotypes from the college admissions process? Individuation.
The essay exception does precisely that. It requires that applicants make explicit the connection between their race and their possession of desirable attributes, such as courage, determination, and leadership. Having made the target trait observable, colleges no longer have need or reason to use the statistical generalizations — racial stereotypes — on which they had previously relied.
III. What Justifies Individuation in SFFA?
The essay exception as an individuation policy is a straightforward alternative to affirmative action in the form of race-based statistical discrimination. But SFFA makes individuation a requirement, not merely an option. This holding relies on the conviction that race-based stereotypes are impermissible even when they are true — or, in other words, that race-based statistical discrimination is unlawful even when it is accurate. This Part examines the following three possible explanations for this position: First, individuation was already available; second, all statistical discrimination is unjust; and third, race-based statistical discrimination is particularly unjust. It concludes that the Court understands race-based statistical discrimination to be particularly unjust because it violates a foundational normative principle — the fundamental tenet of racial equality — that other forms of statistical discrimination do not.
A. Individuation Was Already Available
In the 1979 case, Orr v. Orr,98 the Supreme Court held that there was “no reason [for Alabama] to use sex as a proxy for need” in enforcing its alimony statute because “individualized hearings at which the parties’ relative financial circumstances are considered already occur.”99 As it did in Orr, the Court in SFFA could have held that there was no reason to use race as a proxy because individualized essays already exist.100 Yet, despite the ease of the analogy between Orr and SFFA, it cannot justify the individuation requirement. The Court didn’t indicate that preexisting essay opportunities influenced its decision, and more importantly, the opinion suggests that the individuation requirement is not a matter of convenience but one of justice.101
B. The Harm of Stereotyping
In justifying its imposition of an individuation requirement, the Court might also have relied on a conviction that all stereotyping in college admissions is wrong.102 This view holds some intuitive appeal. Compare an applicant’s two possible statements:
(1) Both my parents were college-educated.
(2) Both my parents were college-educated, so intellectual engagement was a family priority growing up. At the dinner table every night, we discussed what I’d learned in school and what was going on in the world.
An assessor who assumes, after reading the first statement, that the applicant is more intellectually engaged than an applicant whose parents did not attend college is likely to be correct.103 Even so, the assessor’s reliance on that assumption to award the applicant a benefit or penalty is uncomfortable, and the second statement feels like a more just basis for the assessor’s reliance than the first.
Universities could operate on this understanding. They could refuse to collect and make inferences from any type of categorical information, including gender, national origin, family income level, and parents’ education level. Instead, universities could provide opportunities for students to explain the relevance (if any) of these proxy traits to the outcome-relevant traits the university cares about.104 And indeed, at least some of the Court’s prior illegal-discrimination jurisprudence supports this theory: “[T]he Supreme Court has repeatedly held that otherwise-illegal discrimination cannot be justified based on statistical generalizations about groups, even if those generalizations are empirically supported.”105 For example, Craig v. Boren106 and Weinberger v. Wiesenfeld107 both invalidated laws premised on accurate gender-based generalizations due to the type I errors those generalizations caused.108
However, this conception is probably not what motivates SFFA’s essay exception for a few reasons. First, the antistereotyping rationale in SFFA is concerned specifically with racial stereotypes.109 In the Court’s view, “[t]he entire point of the Equal Protection Clause is that treating someone differently because of their skin color is not like treating them differently because they are from a city or from a suburb,” a distinction that may be used to stereotype.110 Second, “it’s not generally illegal for the government or private actors to rely on statistical generalizations.”111 And finally, the growing body of philosophical literature examining the coherence of a blanket rule against stereotyping based on its failure to treat people as individuals has come up empty.112
C. The Harm of Racial Stereotyping
Even if statistical discrimination isn’t always wrong, race-based statistical discrimination might be. The Court ended its opinion by condemning the principle it believed colleges had been acting on: “that the touchstone of an individual’s identity is not challenges bested, skills built, or lessons learned but the color of [one’s] skin.”113 This condemnation is rooted in a fundamental conviction, which this Note dubs the racial equality axiom: Absent social mediation, race does not cause normatively relevant traits and, therefore, is never itself normatively relevant.114 Skin color has no inherent relationship to intelligence, courage, leadership, or other talents. The amount of melanin in one’s skin is never alone a justified basis for differential treatment.115 Actions based on the premise that race does cause normatively relevant traits are wrong precisely because they violate these truths. Race-based statistical discrimination is thus wrong because it risks subjecting individuals to decisions made about them in violation of the racial equality axiom. The use of racial stereotypes always risks, and may cause, type I errors. The admission of a Black student to a university on the basis that she has more courage than her peers when she does not makes race itself normatively relevant.116 Adherence to the axiom of racial equality thus forbids reliance on race-based statistical discrimination.117
The ability of the racial equality axiom to explain three additional aspects of the Court’s equal protection jurisprudence bolsters its justificatory power. First, the racial equality axiom motivates the illegality of foundational, or taste-based, discrimination. Where discrimination is both taste- and race-based, race is itself normatively relevant. Race-based foundational discrimination is therefore irreconcilable with the racial equality axiom.
Second, the racial equality axiom explains SFFA’s skepticism about racial diversity as a state interest.118 Justice Powell’s opinion in Bakke had recognized racial diversity as a legitimate interest: Just as “[a] farm boy from Idaho can bring something to Harvard College that a Bostonian cannot offer[,] . . . a black student can usually bring something that a white person cannot offer.”119 But if the Black student can contribute something different from his white peers, it is not because his skin color caused him to have different thoughts, talents, or experiences. If he has different thoughts, talents, or experiences, they are due to the social mediation of discrimination, celebration, motivation, or more. An assumption that a student’s skin color provides meaningful information about his contribution to diversity is offensive in a categorically different way than a similar assumption based on national origin or income, for which there are no analogous equality axioms.120 When social mediation alone supplies a characteristic’s moral salience, distinctions drawn on the characteristic itself offend the fact of fundamental equality expressed in the racial equality axiom.121
Finally, the racial equality principle helps explain the lessened scrutiny applied to sex-based distinctions in constitutional law.122 Sex, like race, exists absent social mediation. However, unlike race, sex does cause normatively relevant traits even without social mediation. For example, “the large differences in testosterone levels and behavior between men and women”123 are not coincidences. There is thus no sex equality axiom equivalent to the racial equality axiom. The type I errors that sex-based statistical discrimination risks, and can produce, may invalidate the practice when the root cause of difference is social mediation alone, but they do not violate a principle as foundational as the racial equality axiom.
One might object to the invocation of the racial equality axiom by urging that it does not capture the concept of race as used by proponents of race-based affirmative action. The most basic definition of race is “minimal race,” or the “real or imagined bodily differences (as of skin color and hair texture), marking their bearers as . . . sharing real or imagined ancestors, who . . . have a real or imagined common geographical origin.”124 But race can also refer to “racialized group” status, which:
adds to minimal race the idea that . . . people differentiated by [minimal race] are represented as differentially possessing normatively significant character traits and talents for reasons intrinsic to them, in ways that are used to justify practices of categorical inequality (segregation, stigmatization, discrimination), and that . . . such representations help cause such practices.125
Racialization “recognizes that ideas about” the “connection[s] between race and normatively significant character traits” motivate and (purport to) justify ideologies that rationalize racial “stigmatization, segregation, and discrimination.”126 To many advocates, affirmative action, then, is a form of:
[d]iscrimination on the basis of unjust racialization, . . . [that] does not take racialization as a proxy for some other relevant characteristic. Rather, racialized status itself — being a member of a racially segregated and stigmatized group (rather than being a specific victim of discrimination or its effects) — is the relevant feature of interest.127
Understanding race as racialization, however, doesn’t help. First, racialization depends on minimal race. One is only “a member of a racially segregated and stigmatized group” by virtue of possessing the “real or imagined bodily differences” that indicate a shared ancestry or “geographical origin.”128 “Discrimination on the basis of unjust racialization” thus also violates the racial equality axiom because it makes minimal race normatively relevant.129 More importantly, however, the schools did not operate their affirmative action programs with racialization in mind. As the Court pointed out, it borders on preposterous to think that the six broad-brush racial categories used by Harvard and UNC captured anything meaningful about racialization.130 Based solely on their own practices, the universities’ affirmative action programs ran afoul of the racial equality axiom.
IV. What Does Individuation Require in College Admissions?
If adherence to the racial equality axiom justifies individuation, what does individuation actually require? Of course, many, if not most, instances of individuation will be uncontroversial. Imagine a Native American student who writes about her own advocacy to include more Native American history in her high school’s curriculum.131 She explains that she grew up hearing stories about her ancestors’ experiences in early America. Learning about their suffering and bravery inspired her to advocate for that same history to be shared with her peers. This student has made explicit how her race contributed to her development of valuable skills like advocacy and leadership.
Trickier cases arise when the development of the target trait is itself tied up with racial stereotypes. This Part thus focuses on what individuation requires when students invoke racial stereotypes in their essays and argues that, in practice, treating racial stereotypes appropriately requires drawing a bright line between supplying those stereotypes and evaluating them.
A. Two Essays
In their effort to untangle the Court’s opinion in SFFA, Professors Benjamin Eidelson and Deborah Hellman propose the following as a hypothetical essay that would undoubtedly be permissible after SFFA:
I was the only Black student in most of my AP classes in high school. Each time I raised my hand, I felt a rush of anxiety: was I about to confirm all of the worst stereotypes about people of my ethnicity? But as hard as this experience was, I’ll always be grateful for the courage and determination that it helped to engender in me.132
In contrast, the Court cited approvingly to a real essay written by a Black student, Andrew Brennen, as part of his application to UNC.133 It:
described the stereotyping that [Brennen] experienced as a Black man — such as expectations that his interests should be limited to “rap music” or “the hood” — and explained that he “do[es] what [he] do[es] because people do not expect it from [him], [and] because others who look like [him] are not able to do it.”134
On the surface, these two essays are similar; they both discuss the impact that widely held racial stereotypes can have on individuals and describe how the applicants developed positive qualities from feeling those stereotypes applied to them. In that sense, they are both minimally individuated — they specify the nature of the relationship between the proxy and the target traits. But on closer inspection, the essays differ in an important respect: Brennen’s identifies the specific racial stereotypes that applied to him while the hypothetical applicant does not. Brennen’s essay is thus fully individuated in a way the hypothetical essay is not.
As Eidelson and Hellman acknowledge, in the hypothetical essay, “Black” could be replaced with any other race or ethnicity — such as “Irish” — without sacrificing intelligibility.135 A student who was the only Irish person in his AP classes, for example, might worry about perpetuating stereotypes of the Irish as “hot-tempered” or “devoutly religious and conservative.”136 The fact that the hypothetical essay remains coherent when the proxy is swapped reveals that it fails the individuation requirement. The essay does not specify how this applicant’s race led to the development of courage and determination; it merely explains how the stereotypes faced by an applicant of any race — when assumed or supplied by the readers — might inspire courage and determination. That is, the conclusion that the student developed courage and determination is only reasonable if the reader supplies the racialized inference herself; there is no other way she could know the content of the “worst stereotypes” about Black or Irish people the hypothetical student worries she is “confirm[ing]” with her questions or comments.137
The real essay, however, makes explicit the connection between this specific proxy, Blackness, and the chosen target, the motivation and ability to defy expectations. Replacing “Black” in this essay with “Irish” would render the essay incomprehensible — Irish people are not expected to have interests limited to “rap music” and “the hood.” An Irish student who claimed to possess tenacity due to overcoming those expectations would be met, rightly, with incredulity. Certainly, an Irish applicant could write a deeply moving essay about developing tenacity by overcoming ethnic stereotypes; but that applicant would have to explain how his specific proxy trait, Irishness, led to the development of the target trait at issue, just as Brennen did.
Just as the proxy race could be replaced in the hypothetical essay, so too could the racial stereotypes referenced. Based on the essay’s reference to AP classes, it likely intends to refer to a stereotype about the intelligence of Black individuals. Unfortunately, many pernicious stereotypes about Black people in America still exist today.138 The essay could plausibly refer to assumptions about Black individuals’ higher propensity for criminality, drug use, or low socioeconomic status.139 Living in fear of confirming those stereotypes might also engender courage and determination. Brennen’s essay, on the other hand, specifies which stereotype about people of his race motivated him to defy expectations: “that his interests should be limited to ‘rap music’ or ‘the hood.’”140
B. Supplying Versus Evaluating Stereotypes
Though the distinction between Brennen’s essay and the hypothetical one might seem trivial, it directly addresses the Court’s main concern: the continued use of racial stereotypes in college admissions systems. As Eidelson and Hellman point out, all communication, even through college admissions essays, is “fundamentally cooperative.”141 Thus, Eidelson and Hellman contend, at some level, an applicant “will be counting on the reader to fill in various details based on the common ground that the two share as inhabitants of the same social world.”142 But that’s not exactly right. Although effective communication indeed depends on common knowledge or trust, it need not depend on inference. In other words, there’s a big difference between filling in the details and evaluating the credibility of a statement.
Consider, for example, two claims about the skills high school students gain from playing team sports:
(1) High school students can gain many important skills from playing team sports like soccer.
(2) High school students can develop a keen sense of curiosity from playing team sports like soccer.
A typical reader is likely to infer from the first statement that the writer is referencing skills such as communication, collaboration, and discipline. In this scenario, the reader supplies the critical information called for by the statement, namely which skills high school soccer players might develop. On the other hand, a reader of the second statement is tasked only with evaluating whether it is reasonable to claim that students can develop curiosity through playing soccer.
Each example college admissions essay discussing race is analogous to one of these scenarios. The hypothetical essay requires the reader to make a judgment call about what “the worst stereotypes about people of [his] ethnicity” are.143 In part, that judgment call will rely on the typical process of decoding language.144 But it also necessitates that the reader select particular stereotypes about Black individuals to apply to the applicant. On the other hand, Brennen’s essay asks the reader only to confirm that liking rap music and living in the hood are common and culturally significant expectations about Black Americans today. Unlike Brennen’s essay, the hypothetical essay thus requires the reader to do what SFFA forbids: make race-based generalizations in violation of the racial equality axiom.
The error costs of supplying racial stereotypes are significantly higher than those of evaluating racial stereotypes. An assessor who commits a type I error while reading the hypothetical essay imposes a racial stereotype on the writer that he did not invoke and that may not even apply to him. She thus violates the racial equality axiom described above.145 On the other hand, the assessor who commits a type I error in reading Brennen’s essay approves the writer’s assertion of a falsity. She does not, however, impose a race-based generalization on the author. Of course, Brennen has himself invoked a racial stereotype. But one’s violation of the racial equality axiom against oneself does not “demean[] the dignity and worth of a person”146 in the same way that another’s violation does. And on the Court’s explanation, it makes no difference whether the stereotype to which the applicant is subjected is a positive or negative one. The wrong is in the “allocat[ion of] preference to those ‘who may have little in common with one another but the color of their skin.’”147
Of course, evaluation requires the reader to either rely on her prior knowledge or trust the speaker’s assertion. However, an assessor’s reliance on her background understanding of the world and social life to evaluate a statement does not risk affirmation of racial stereotypes in the same way that supplying those racial stereotypes does. Supplying racial stereotypes requires the assessor to exercise independent judgment about which racial generalizations to elevate and which to ignore. The assessor thus actively participates in furthering the relevance of particular racial stereotypes even if she does not endorse them. The type I error risk that this choice creates violates the racial equality axiom. On the other hand, evaluating the existence and weight of a racial stereotype requires the reader to assess only whether the social fact presented to her is accurate. In doing so, she does not actively promote or disregard any race-based generalizations and does not run afoul of the racial equality axiom.
Two objections might be raised. First, admissions officers shouldn’t be required to artificially blind themselves to the social realities of modern America.148 Moreover, even if they wanted to, “[a]dmissions officers could not actually constitute themselves as blank slates when it came to race, with all of the naïveté (and . . . gullibility) that would entail.”149 But the essay exception does not require admissions officers to ignore race. They can and must use their background knowledge to objectively evaluate whether the story a candidate tells about himself and why he would make a valuable addition to the university is a reasonable one. The assessor may not, however, apply her own notions of what it means to be a particular race to a candidate who has not voluntarily adopted that description.
Second, an objector might worry that, as described above, the individuation requirement unreasonably burdens applicants, either obliging those who wish to discuss race to provide a sufficiently detailed history of racism in the United States150 or “disabl[ing] them from writing good essays qua essays.”151 But, to satisfy the individuation requirement as conceived of here, a Black applicant need not provide “the history and current reality of anti-Black racism in the United States”;152 he need only provide sufficient contextual information such that the assessor can reasonably assess the accuracy of the story the applicant tells. Although this requirement could detract from the quality of applicants’ writing, Brennen’s essay and others like it should help allay these concerns.
Conclusion
A statistical discriminator uses accurate stereotypes to dole out advantages and disadvantages. But when those stereotypes are about race, statistical discrimination’s inevitable errors violate the principle that race is never normatively relevant absent social mediation. Social mediation itself — the mistreatment or inspiration that inspires courage and leadership — rather than race, is the proper object of consideration.
On the Court’s reasoning, affirmative action as practiced by universities was illegal because it was a form of race-based statistical discrimination. As an individuation mechanism, the essay exception excises college’s illegitimate reliance on racial stereotypes. It gives applicants the opportunity to explain the connection between race and normatively relevant traits and allows assessors to evaluate, but not supply, racial stereotypes offered by the applicant. Understanding SFFA through the lens of statistical discrimination not only crystalizes the moral truth of fundamental racial equality but also provides practical guidance to admissions offices, applicants, lower court judges, and the general public on how to abide by SFFA’s principles going forward.