II. Coalition for TJ v. Fairfax County School Board

In Coalition for TJ, a three-judge panel of the Fourth Circuit rejected the claim that the decision by the Fairfax County School Board to change the admissions procedure at the highly competitive Thomas Jefferson High School (TJ) was impermissible race discrimination in violation of equal protection. Neither the old admissions procedure, nor the new procedure adopted in 2020, used race explicitly. Indeed, the admissions files even omitted the names of applicants to ensure that the process was race-blind, in the sense of being blinded to individual applicants’ racial statuses. Still, the complaint alleged that the change from one facially race-neutral policy to another was unconstitutional because the change was motivated by a desire to reduce the number of Asian American students admitted to TJ and in fact achieved that effect. On a motion for summary judgment, the lower court had held that the School Board’s decision to change its policy violated equal protection. The Fourth Circuit reversed and granted the School Board’s cross motion for summary judgment, finding no constitutional violation.

This case raises both the intent question and the effect question flagged above. First, the case raises the question of what intentions are constitutionally impermissible. In particular, given a fixed class size, is the intention to increase the enrollment of Black and Latino students (which the School Board acknowledged was part of its goal in changing the policy) the constitutional equivalent of an intention to “racially balance” the school or to decrease the enrollment of Asian American students? Second, even if an impermissible intent is present, how should courts assess whether the effect produced by the policy change establishes a constitutional violation? As we noted in the introduction, while the first question is extremely important, it has already attracted significant attention. The second question, by contrast, has so far largely been ignored.

But that soon may change. The Coalition for TJ [the Coalition] petitioned the Supreme Court for certiorari in the case, which was denied.Footnote ⁵ However, in a dissent from the denial of cert, Justice Alito focused exclusively on this second question, and noted his disagreement with the manner in which the Fourth Circuit assessed whether the change in policy had produced a discriminatory effect.Footnote ⁶ The Fourth Circuit had concluded that the Coalition could not demonstrate the new policy “purposefully discriminates against Asian American students” unless they could show “(1) that the policy exacts a disproportionate impact on a certain racial group, and (2) that such impact is traceable to an ‘invidious’ discriminatory intent.”Footnote ⁷ According to the Fourth Circuit, the correct way to analyze whether the policy produces this “disproportionate impact” (the effect prong) is to compare the percentage of Asian American students who were admitted under the new policy to the percentage who had applied to the school. And because Asian American applicants to TJ under the new policy were 48.59% of the applicant pool and 54.36% of the admitted class, the Fourth Circuit found that there was no disproportionate impact.Footnote ⁸ As a result, the case was without merit.

By contrast, the District Court, dissenting Circuit Judge Rushing, and Justice Alito all favored a different way of assessing whether the relevant “disproportionate impact” was present. Rather than assess how the admitted class compares to the applicant pool, these jurists favored comparing the percentage of Asian Americans in the admitted classes under the previous admissions policy with the percentage of Asian Americans admitted under the new policy. This “before/after” comparison, they claim, is what shows the relevant effect. And indeed, whereas in the five years prior to the policy change, Asian American students received 65–75% of the admissions offers, in the 2021 year (under the new policy), they received only 54%, which Judge Rushing describes as a 26% reduction in offers of admission.Footnote ⁹ According to this measuring stick, the policy has a clear “disproportionate impact.”

The narrow legal question at hand then is this: Which comparator is the right one to use? Should the effect prong of a claim of disparate treatment by proxy be assessed by comparing the percentage of admitted Asian American students to the applicant pool (or some other choice of reference population) as the Circuit Court held? Call this the population comparison or PC. Alternatively, should it be assessed by comparing the percentage of Asian American students admitted under the prior policy to the percentage admitted under the new policy? Call this the before/after comparison, or BAC. These two questions lead to broader moral questions: What is the normative rationale that underlies each of these standards? Which gives a better account of a distinctively discriminatory effect?

To date, the PC is the test used by most courts, though only a few have addressed this issue directly.Footnote ¹⁰ The main reason that courts adopt this approach lies in a flaw they identify with the BAC. In their view, the BAC makes the prior policy the normative default, the standard against which a claim of current discrimination is assessed. And if the prior policy is itself unjust, there seems to be something problematic about such an approach.Footnote ¹¹

We appreciate this problem with the BAC, and indeed we develop and expand on it below. But it is not dispositive of the issue, in our view, as there are powerful reasons to reject the PC as well. Suppose that members of a School Board are genuinely motivated by racial animus. The School Board thinks “we have too many Asians,” “they are taking over,” and accompanies those thoughts with all sorts of negative epithets about Asians that show genuine and deep animus toward Asian Americans. If the PC governs, the School Board would be free to adopt a facially neutral policy that is designed to reduce the Asian population at the school, and that does significantly reduce it, so long as the percentage of Asian American students at the school remains at or above their percentage in the relevant population. Both Justice Alito and the dissenting judge on the Fourth Circuit emphasize this noxious upshot of using the PC in their objections to the Fourth Circuit’s approach.

Resolution of the debate about which comparator a court should use to assess a claim of disparate treatment by proxy depends on how one understands the nature of discriminatory harm. We should clarify, however, that this element should not be confused with the more generic legal requirement of standing, according to which the person or entity bringing the legal challenge must have suffered a harm. Any applicant to TJ who was not admitted has standing to challenge its admissions policy.Footnote ¹² Nor should it be confused with the requirement that the impermissible intent must have led the relevant actor to actually do something (adopt a policy, maintain a policy, etc.). The forbidden intention cannot simply be floating around in the mind of the actor without giving rise to an action in the world. After all, a plaintiff must have some policy to point to that she alleges was adopted or maintained with an impermissible intent. But these general requirements of standing and action differ from the more specific requirement that the challenged policy has a particular kind of effect, which both the BAC and PC standards point to. These tests do not function simply to ensure that a plaintiff has standing or that the actor whose action is challenged has taken some action. Rather, as we argue below, each standard puts forth a particular conception of discriminatory harm. That is, the BAC and PC tests each propose an answer to the question of what kind of “effect” in the world constitutes a distinctively discriminatory effect.

III. The puzzle of discriminatory harm and the flaws of both methods of assessing it

In this section, we will probe the conception of discriminatory harm that underlies each of the BAC and the PC. The idea of a discriminatory harm is novel to the particular kind of discrimination under study—disparate treatment by proxy—and the normative question at hand is what exactly it is. When does some outcome or effect amount to a distinctively discriminatorily harmful outcome or effect? As noted above, this question should be distinguished from the more generic question of whether adopting some policy effects any harm at all, broadly construed. In TJ, that more generic standing requirement may be satisfied so long as the admissions policy at issue denies someone entrance to the school. The question, though, is whether the harm of being denied entry into one’s school of choice is legally cognizable as a discriminatory harm. To answer this requires an account of discriminatory harm.

To better understand the nature of this inquiry into discriminatory harm, it might help to contrast it with other cases where harm is at issue. In many cases, it is obvious that a given state of affairs constitutes a harm and so we neglect to draw out explicitly the basis on which the state of affairs is a harmful one. For instance, when I break your arm, I effect a state of affairs in which your arm is now broken. We can immediately recognize this to be a harmful state of affairs, without working through some underlying theory of harm to explain why. This is not to say that there is no further explanation why having one’s arm broken constitutes a harm; there certainly is. But in cases like this, we can readily identify that the state is harmful without filling in a particular account on which it is. By contrast, in the cases at hand, where the effect of some policy is to simply admit some set of students and deny others, it is not only not immediately clear whether the resulting state of affairs constitutes a discriminatory harm, but it is not even clear how one should go about assessing this. Hence we have in TJ a debate between two different empirical tests, which emerge out of two different substantive conceptions of what makes some outcome a case of “discriminatory harm.” Each test characterizes the effect that a school admissions policy has on the demographics of its student population in a distinctive way, and thereby proposes that this way of conceptualizing the policy’s effect brings out what is potentially discriminatory about it. According to the BAC, determining whether some policy produces a discriminatory harm requires comparison of its resulting demographic profile with that of a previous policy. For the PC, the comparator should instead be the demographics of some catchment area. Scrutinizing the BAC and PC standards thus means asking why each standard’s approach to characterizing some effect—say, as a decrease, or as underrepresentation—should be what draws out what is distinctively morally significant about it (vis-à-vis discrimination). Unfortunately, the conception of discriminatory harm that underlies each standard is normatively flawed as we elaborate below, beginning with the BAC.

To understand the normative vision of discriminatory harm that underlies the BAC, a good place to begin is with Justice Alito’s dissent from the denial of certiorari in the TJ case. He writes:

If you think this example describes a discriminatory harm, as Justice Alito does, then perhaps the BAC captures something important. But what?

The answer isn’t simply that the policy was enacted with a legally problematic intention (however defined). Rather, there must also exist a distinctively discriminatory harm. The BAC takes the harm to inhere in the fact that there is a decrease in the number (or percentage) of Black players on the basketball team. Now, it is certainly true that the policy results in a decrease in the number of Black players on the team. But the question at hand is why the mere fact of a decrease should make it the case that the resulting state of affairs realizes a “discriminatory harm”? What we need is an argument for why this temporal comparison between the present composition of the team and the preceding composition grounds a claim of discriminatory harm. Otherwise, claiming that the policy has a discriminatory effect because it produces a decrease relative to the prior policy amounts to assuming rather than justifying that the BAC provides the right baseline for determining what constitutes discriminatory harm.

The above discussion thus imparts a methodological lesson. To avoid accidentally presupposing a substantive account of discriminatory harm in examining the effect that a policy has, we should try to characterize the outcome in as neutral terms as possible. So rather than describing the policy as “decreasing” or “under-representing” some group—terms which implicitly refer to a particular substantive choice of baseline—we should instead characterize the policy as resulting in X racial breakdown. Recasting the effect in this way helpfully brings out the ambiguity in how such states of affairs should be assessed: It is simply unclear whether a basketball team with 5 Black players out of a total 12 players really is one in which Black students are discriminatorily harmed.

Returning to the BAC in particular: notice that in setting forth a temporal or historical comparative account of harm, its animating normative picture of discrimination is one that is concerned with the change in a racial group’s representation over time.Footnote ¹⁴ In particular, in characterizing policy outcomes in terms of how they relate to past policy outcomes, the BAC both privileges the status quo and treats the status quo as a normative baseline for determining whether and when some state of affairs constitutes a discriminatory harm. Status quo entrenchment is a problematic feature of institutional design, and especially so in domains where concerns about discrimination arise. This aspect of the BAC would thus seem to pose serious risks and pitfalls for its account of discriminatory harm. To better see what these are, it will help to break up consideration of the BAC into two types of cases. First, there are cases in which the state of affairs before is the same as the state of affairs after the action at issue. Second, there are cases in which the state of affairs before differs from the state of affairs after the relevant action. We discuss each scenario below, with the aim of highlighting how the BAC standard operates to preserve the status quo and the key shortcomings that follow from this conception of discriminatory harm.

First consider a case in which there is no change in outcome. Since the BAC asks that courts compare how a group fared before some poorly motivated action occurred with how the group fares after that action, where there is no difference between how the group fared before and after the action at issue, there can be no violation. For this reason, the BAC enacts a status quo bias, in both a descriptive and normative sense.

In particular, where there is no change in the world, there can be no constitutional violation because the BAC requires that a group’s share of the admissions spots (or whatever else is at issue) diminish. As a result, this test cannot recognize poorly motivated decisions that yield no change. In particular, poorly motivated decisions to maintain policies are good examples of cases that are especially likely to evade scrutiny under this test, because such decisions are likely to yield no change in results. To illustrate, consider the following case.

Scrapped Plans

In declining to adopt the new policy on the grounds that it is likely to increase the admission of Black and Latino students, the school board in Scrapped Plans has, in our view, made a decision that is morally troubling. Moreover, and perhaps more importantly, Scrapped Plans seems as morally troubling as the decision to adopt a new policy for the same reasons. Because new policies are more likely to result in changes in who is admitted than decisions to maintain policies, the BAC is significantly more likely to find a violation in the second case than in the case of Scrapped Plans. Yet this asymmetry in how the BAC treats the two cases seems unjustified.

The second manner in which the BAC reinforces the status quo is more distinctively normative. When the BAC serves to test for some legal-normative violation, the “before” baseline establishes the normative benchmark to which all new policies are compared and in light of which they must be justified. This framing has two consequences. First, any future policy can never be considered on its own merits and must instead, always, be characterized as a departure from a previous policy and judged on those terms. Second, since the prevailing policy constitutes the yardstick of constitutional violation, the current policy cannot itself be scrutinized in the same way.Footnote ¹⁵ That is, whether the prevailing state of affairs is considered independently bad or good is immaterial to the BAC. Instead, prevailing policies earn their keep simply by having already been instated. In Scrapped Plans, for instance, the BAC would focus on the board’s decision to change its longstanding policy, while this policy itself receives no such scrutiny. So, the BAC in effect freezes the school’s policy at an arbitrary moment in time (the time of the court’s adoption of this standard, or perhaps the policy immediately preceding the policy that is being challenged) and proclaims that policy to be de facto justified. On this way of looking at it, the BAC does not just risk or embed a bias toward the status quo. Rather, to the extent that it effectively defines nondiscrimination by reference to the reasons behind and the effects of the prevailing policy—without ever bringing independent scrutiny to those reasons and effects themselves—it just is a status quo preserving standard.Footnote ¹⁶

Next consider the situation in which the outcome does change over time. When the outcome in the world changes after a poorly motivated decision (likely because a new policy is adopted), the BAC may also entrench the status quo. To illustrate, consider the following case:

Policy Change

The first thing to note about contexts like this one in which admissions are competitive and spots are limited is that any policy change is likely to change the school’s demographics. And any change in demographics results in some group’s percentage of the admitted students increasing and some group’s percentage of the admitted students decreasing. In other words, in virtually all cases of policy changes, there will be some group whose percentage of the school’s population has decreased in comparison to what it was before. The BAC will thus nearly always set the stage for a complaint by some group to any policy change. If decisions that maintain the status quo are likely not open to criticism and policy changes are likely to always risk criticism, the BAC seems to doubly entrench the status quo. And in a world marked by past and present injustice, to entrench the status quo just is to entrench many of these injustices.

Of course, a decrease in a group’s percentage among the admitted students does not, by itself, give rise to a legal claim of disparate treatment by proxy—even if we grant that the BAC presents the relevant standard for determining a discriminatory effect. Nonetheless, it is important to note that the BAC ensures that the effect element of the claim will be satisfied for some group in nearly every case (the only exception being the rare policy change that does not alter the demographics of the admitted student body in any way).

Next consider the PC. Whereas the BAC expresses a temporal conception of discriminatory harm, which favors stasis over time, the PC proposes that an effect be evaluated by reference to a particular normative ideal set by the demographics of some relevant population, say, the demographics of a school’s catchment area. On this view, a policy should select a population that roughly reflects the demographic composition of this reference population, and departure from such proportional representation constitutes a discriminatory harm for those groups that are underrepresented.

This measure of discriminatory harm, like that adopted by the BAC, requires a justification as it is not obvious why the fact that a group’s allocation of spots in a school (for example) is less than its representation in the catchment is a discriminatory harm. The fact that a particular pattern exists between these two measures is not meaningful on its own as either a harm or, specifically, a discriminatory harm. So, the first objection one might have to the PC is that it lacks a normative grounding as it does not explain why we should object to the failure of the school’s demographic profile to match the population from which it draws.

One could, of course, flesh out this account by providing such an explanation. In our view, the most plausible account of why a departure from proportional representation would constitute a discriminatory harm is the supposition that any such departures have their roots in prior discrimination. For example, the thought is that were it not for racial discrimination in the past, the racial breakdown of the student population at the magnet school would match the racial demographics of the catchment area. And thus any departure from this discrimination-free population baseline constitutes a discriminatory harm. If some thought along these lines underlies the PC (an idea we develop in Section V), the PC is thereby committed to something further: that a policy fares better in terms of discriminatory effect, the closer it gets towards this ideal state.

Let’s grant, for the sake of argument (for now), that mirroring the demographics of a given reference population is in fact ideal. Still, understood in such broad terms, the claim that descriptive improvements toward this ideal and normative improvements move in lockstep is unlikely. To return to Justice Alito’s hypothetical, it certainly does not seem as though adopting a policy that leads to an increase in the representation of white students on the basketball team would make for a better state of affairs than maintaining the current basketball try-out regime, simply because it is descriptively closer to the racial composition of basketball teams in an ideal world. But this is implicitly what the PC approach assumes. It assumes that because a given standard is the ideal state, those states of affairs that depart less from this state are preferable to those that depart from it more. In fact, this is the reason why the PC judges that in TJ, a policy change that reduces the percentage of Asian American students without bringing them below the population baseline is a change that does not make things worse (and perhaps even makes things better). But this conclusion strikes us as too quick. In the real world, in which prior injustice has occurred, it is not obvious that this is an improvement. In fact, by the lights of the PC, it would even be an improvement if the Asian student population decreased in favor of more white students, holding fixed Black and Latino representation, since that change would also bring the profile of admitted students more in line with the demographics of the applicant pool.

David Estlund calls this error the “fallacy of approximation.”Footnote ¹⁷ This fallacy rests on the thought that “if you have some ingredients of a good condition, but you will not have all of the ingredients, having more of them would be better.”Footnote ¹⁸ However, while this is sometimes the case, it is not always the case. Estlund illustrates the fallacy with the example of a person who is prescribed three pills for a medical condition but cannot get access to all three. If she can only take pills 1 and 2 but not pill 3, it is not clear that taking pills 1 and 2 is better (for her health, say) than taking only one pill or none at all. We would need to know more about the pills, how they work, and the interaction among them to make that assessment.

The “negative implication” of the fallacy of approximation is perhaps even more relevant for the scenario we are investigating, according to which “understanding some high standard is not useful for understanding value rankings where the standard is not fully met.”Footnote ¹⁹ If we bring that insight to bear in the case we are considering, we might say the following. The fact (supposing that it is a fact for the sake of argument) that in the ideal world, in which no prior discrimination had occurred, the racial breakdown of the population of admitted students would roughly mirror the racial breakdown of the geographic area the magnet school draws from does not in itself help us to discern which nonideal demographic breakdowns are better or worse than others. For this reason, even if we grant that the PC correctly expresses the ideal, it does not tell us whether any particular policy change is a normative improvement.

IV. Should the discriminatory effect element be abandoned?

The prior section highlighted the ways in which each of the conceptions of discriminatory harm present in the case law has significant problems. The BAC treats any departure from the present state of affairs as presumptively problematic and thereby enacts a form of status quo bias that is unjustified. The PC, by contrast, rejects stasis as a normative ideal in favor of an imagined ideal standard. But it too is flawed because it mistakenly assumes that states of affairs that are closer to this ideal are necessarily better than those that are further away. Given these significant problems with the two current approaches for assessing discriminatory harm, perhaps this element should be abandoned. If it were, the claim of disparate treatment by proxy would only require a showing of impermissible intent and no showing of discriminatory harm at all. (Though, as we noted earlier, plaintiffs would still need to have standing to bring such a claim.)

While this intent only version of the disparate treatment by proxy claim is surely possible and there is some language in the cases that supports it, we should note that all courts considering the claim say that a plaintiff must show both impermissible intention and some sort of impact.Footnote ²⁰ For example, in Coalition for TJ itself, the Circuit Court first assessed whether the requisite impact was present and, having found that it was not, noted that there was no need to go on to consider whether there was sufficient evidence of improper purpose (though it did so nonetheless).Footnote ²¹ If an effect were unnecessary, then failure to satisfy the effect prong would not suffice to invalidate the claim. The fact that the Fourth Circuit, and other courts, require some effect demonstrates that the intent only version of the claim does not fit existing case law.Footnote ²²

That said, the case law assessing claims of this form is continuing to evolve.Footnote ²³ Indeed, the United States Court of Appeals for the Second Circuit recently decided a case dealing with admissions policies at the New York City magnet schools that might be read to hold that a plaintiff need only show an intent to decrease the Asian population at the schools and the denial of admissions to some Asian students.Footnote ²⁴ In Chinese American Citizens Alliance of Greater New York v. Adams [CACAGNY],Footnote ²⁵ the posture of the case requires the court to assume that impermissible intent is present and so the sole issue on appeal was whether the plaintiffs have established the requisite “discriminatory effect.” The intention that is presumed is that the city intended to decrease the percentage (or number) of Asian students in the New York City magnet schools. Because the new policy led neither to a decrease in the Asian population at the schools (and thus did not satisfy the BAC), nor was the Asian population in the magnet schools lower than the percentage of Asian students in the school-age population eligible to apply to the magnet schools (and thus neither did it satisfy the PC), the trial court held that the plaintiffs had not stated an equal protection claim. On appeal, the Second Circuit reversed, finding that discriminatory effect should also be evaluated on an individual basis rather than only on the group or aggregate basis instantiated in either the BAC or PC. Using this approach, the court found that the requisite discriminatory effect was shown by the fact that some Asian applicants who would have been admitted under the prior policy were rejected under the new policy.

Here’s how the court describes its holding: “when a discriminatory intent in enacting a facially neutral law or policy is shown, the requisite proof of discriminatory effect is not limited to an aggregate disparate harm, but rather can be satisfied by proof of an adverse harm to one or more individuals based on their race that has resulted from that discriminatory law or policy.”Footnote ²⁶ But to the extent that this is offered as a conception of discriminatory harm and not merely as a reiteration of the standing requirement, this statement is puzzling and, ultimately, circular. While an Asian student plaintiff has indeed been denied admission, what makes this effect “based on their race” and thus a discriminatory effect rather than merely the sort of harm that is necessary for standing? After all, the policy at issue does not employ race. Perhaps the claim rests on the fact that the student is denied admissions and is Asian. But there are many students of different races who were denied admissions under the new policy (and perhaps would have been admitted under the old policy). Why exactly have these other students not also stated a claim for a violation of equal protection? After all, the policy under which they were denied admission was motivated by an impermissible intention and they suffered harm. So what exactly makes the harm “discriminatory” in the case of an Asian student but not “discriminatory” in the case of a student of another race? Granting that the facially neutral policy was adopted with the invidious intention of harming Asians, how does the content of that intention make a difference to the nature of the harm that is suffered by different individuals on account of the same policy? Setting aside matters of blame or culpability, it would appear that the harm of being denied admission is the same, regardless of who it is that is being denied admission. The court in CACAGNY provides no answer for why it is that only Asian students suffer the discriminatory harm. For this reason, this case would seem to suggest that the discriminatory effect element is unnecessary.

But when we dig deeper into the reasoning of this case, we see that the court implicitly relies on the BAC baseline after all. Notice that the court in CACAGNY does not treat all Asian students who were denied admissions under the new policy as having suffered a harm “based on their race.” Rather, it is only those who would have been admitted under the prior policy who qualify as having suffered the requisite discriminatory harm.Footnote ²⁷ This restriction on which Asian students may bear proper witness to the policy’s discriminatory effect shows that the court in CACAGNY implicitly relies on a version of the BAC in assessing what constitutes discriminatory harm, albeit in an individualized rather than an aggregate population-wide form. Perhaps the court does so because it realizes that the BAC fills in a conception of discriminatory harm that is otherwise missing in its analysis.

As this discussion illustrates, existing case law maintains the importance of showing discriminatory harm—though the question of what such a harm is has remained persistently ambiguous. In addition, while an intent-only version of the claim of disparate treatment by proxy remains a theoretical possibility, we are not yet ready to abandon the necessity of showing discriminatory harm. In part, this judgment rests on the very real values on which it rests, as we explain in the next Section.

V. Genuine discriminatory harm as captured by both the BAC and the PC

While both the BAC and the PC articulate flawed conceptions of discriminatory harm, as we explored in Section III, there is a kernel of truth that each captures as well. In this Section, we explore what that kernel is. In the next Section, we will explain the tension between these two visions and how that tension presents a very real dilemma for how best to understand the concept of discriminatory harm.

In Section III, we explained that the BAC privileges the status quo. There, we focused on the normatively troubling features of status quo bias: in particular, its tendency towards the entrenchment of existing injustices. But be that as it may, this does not mean that status quo bias is wholly unjustified, as the stability of the institutions that govern access to important opportunities indeed has moral value. In particular, philosophers have long noted the pervasive significance of reliance in our lives. Reliance lies at the heart of rational agency itself: reliance, for example, that that which typically serves as a means to an end, will this time, when I take it, also serve as a means to my desired end. This in turn means that reliance has special normative import in a liberal society. The exercises of agency characteristic of a liberal society require individuals to have more than just the mere latitude to author and pursue their own lives. Agent deliberation, action, and planning must meet up with a broader causal environment that is sufficiently stable and predictable in order for individuals to be able to successfully author their lives. As a general matter, then, the normative significance of reliance draws from the normative significance of supporting agency itself.

Thus, from the perspective of safeguarding agents’ abilities to plan their lives in rational pursuit of key institutional goods, there is some disvalue in making changes to policies because it weakens reliance—even if the change would be more rational, fairer, or even all things considered an improvement. Of course, even while this interest in stability is real, it must be weighed against countervailing interests in modifying and improving the processes by which important opportunities are allocated. This is especially pressing when current institutional arrangements are unfair or work to compound prior injustice. So, there is a real tension here that we will return to and further develop in Section VI.

Still, in some cases, reliance interests are not only of general value to all in a liberal society but failing to ensure them can even present a distinctively discriminatory harm. Consider the facts in Coalition for TJ itself. At issue in this case is the policy’s effect on Asian representation within the new class of admitted students. It is notable that in our world, Asian students have historically been (and in many cases, still are) affected by discrimination in schools. For example, within the context of competitive school admissions, subjective assessments of leadership skills or personality have been applied to Asian students, in the past and in myriad settings, in a discriminatory manner. Unconscious reliance on familiar stereotypes of Asian students, e.g., passive or dull, affects Asian students’ abilities to perform well on these dimensions. Suppose, now, that discriminatory harms like these in part explain Asian students’ investment in admissions criteria such as standardized test scores and accordingly their relatively better performance on these measures compared with those where they are unfairly disadvantaged. If so, then the policy change away from test-based assessments would disproportionately harm Asian students in considerable part because of their past experiences with discrimination. This would seem to make the effect of the policy change on Asians discriminatory indeed. Furthermore, the “overrepresentation” of Asian students relative to their population share in a context in which test-based assessments predominate (like TJ admissions before the change) is importantly not an overrepresentation due to a set of prior unfair advantages that the group had.

To support our view that, to the extent that BAC captures genuine discriminatory harm, it does so when and because it protects the reliance interests of individuals who have navigated injustice, consider a modified version of the TJ case, which we call TJ*.

TJ and TJ* would be judged similarly under the BAC standard, yet there seems to be a normatively important difference between them. Most notably, in TJ*, it is plausible to think that the status quo of overrepresentation of white students has come about because of unfair advantages these students and their families had access to. For this reason, a decrease in their representation may be a step in the right direction, because it lessens their overrepresentation as compared to what would have occurred absent prior racial discrimination (an intuition that the PC seems to capture).Footnote ²⁸ The real TJ case, by contrast, wherein it is Asian students who are “overrepresented,” feels different because it seems that prior and current discrimination against Asian students could plausibly explain why Asian students do especially well on standardized tests and are thus overrepresented among the accepted students under the old test-based policy (an intuition that the PC failed to register, as we explained earlier). This contrast between the real TJ case and TJ* suggests that the BAC’s presumption in favor of the status quo only tracks a genuinely discriminatory effect when the group that is disadvantaged by some policy is disadvantaged today in part because of discrimination in the past.Footnote ²⁹

This intuition draws further support from the fact that the groups challenging policy changes under a disparate treatment by proxy theory do so on behalf of plaintiffs, like the plaintiffs in Coalition for TJ, who have themselves suffered prior discrimination. In fact, the hypothetical Justice Alito constructs to illustrate the genuine discriminatory harm that BAC registers, and PC misses, offers an example with a remarkably similar structure. In particular, it seems to us that Black students’ representation on school basketball teams is not something that is entirely independent of their marginalization in other areas of school. Rather, there is very plausibly a link between the two. Perhaps opportunities to play sports were comparatively more open to Black young people than were opportunities to excel in other ways, or perhaps common stereotypes about the abilities and deficits of Black students worked to steer Black students into basketball rather than science, for example. If this were the case and the present-day representation of Black players on the basketball team could in part be traced back to past discrimination practices against Black students, then it does seem troubling to kick Black players off the team today in the name of nondiscrimination. After all, the reason Black students as a group are even in this position to be disadvantaged by the policy change in the first place is because of past discrimination.

Justice Alito is right to note that there is something indeed discriminatory about a policy that reduces a group’s representation in the latter case, since the reason why the policy has that effect on the group is likely because of a history of discrimination against this group. When “overrepresentation” is the result of a group’s navigating past discrimination (as might be the case with Asian students in the actual TJ case and Black students in Justice Alito’s basketball example), policies that disadvantage that group on that basis today are naturally thought of as having an effect today that is a legacy of this past discrimination and so is itself discriminatory. This is the conception of discriminatory harm that the BAC appropriately captures.

At this point, we should make clear that we do not here mean to be positing that these causal stories about navigating past discrimination are in fact what explains the “overrepresentation” of Asian students at TJ or the hypothetical “overrepresentation” of Black students on the basketball team in Justice Alito’s example. Ultimately, validating whether there is such a link is a matter of historical and social-scientific rather than normative and theoretical inquiry. Nor do we mean to suggest that concerns to safeguard reliance and attend to how groups have navigated past injustice are what in fact motivate various actors to defend the BAC. Rather, our aim has been to put forth plausible explanations for the divergence in our moral intuitions about these cases—the real TJ and Justice Alito’s basketball hypothetical, on the one hand, and TJ*, on the other—and connect them to an account of why the BAC in different cases does and does not seem to latch onto something significant vis-à-vis discrimination. These explanations center on the after-effects of past discrimination on how groups navigate the social world today. The thought is that when past discrimination against a particular social group explains why it is disadvantaged by a given policy change today, these effects on the group fit within the normative category of “discrimination.”

Thus, while normative discussions of the effects of preserving the status quo tend to focus on the preservation of unfair advantage, there are also circumstances under which there are good reasons to support a presumption in favor of such stability. Cases like Justice Alito’s basketball hypothetical bring out a side of the prevailing state of affairs with which we have more positive associations: It also reflects how groups have managed to navigate past discrimination and the forms of agency that they have developed in response to injustice. From this vantage, the status quo contains sites of opportunity even for those social groups who are and historically have been disadvantaged. Insofar as policy changes threaten to alter disadvantaged groups’ access to these opportunities, it seems to us that they should indeed face a burden.

In sum, the BAC could rest on two broad kinds of harms that track genuine concerns of discrimination. First, in a society with a history of injustice, there will be certain social resources or avenues to opportunity that disadvantaged groups have come to rely on in responding to past discrimination. In these cases, policy changes which disproportionately disadvantage the group by altering access to these goods trace back to this past discrimination because the reason they became especially important to the group is the unjust ways that other avenues toward opportunity have been shut off to them. In these cases, the harm today is quite literally an effect of past discrimination and so seems to itself be a discriminatory effect. Second, seeing the status quo as embodying not just injustice but also adaptive responses to injustice and successful navigation of unjust social conditions allows us to see how policy changes may further disadvantage already disadvantaged groups. It does so by undercutting both their access to important social goods as well as their agential powers in acting and planning around key social institutions. If the moral wrong of discrimination is tied to the entrenchment of disadvantage, then these effects would straightforwardly constitute discriminatory harms.

Having unearthed what we see as the genuine moral intuition that underlies the BAC, we turn now to the PC, and, in similar fashion, attempt to recover the kernel of moral value on which it rests. Rather than compare the state of affairs after the policy change to the state of affairs before the policy change—as the BAC did—the PC compares the state of affairs after a policy change to a particular normative ideal, one in which each group is represented in rough proportion to its percentage of the relevant population. By focusing on that comparison, we can see what normative vision undergirds the PC.

In our view, the most promising normative rationale for adopting the PC is that it instantiates a view of what the world would have been like had discrimination not marred the past. Absent such discrimination, the PC seems to assume, admission to competitive high schools like TJ would not correlate with race. Rather, each racial group would be likely to be represented among the student body in proportion to their representation in the applicant pool or in the catchment area. Absent prior discrimination, kids from different racial groups would not differ from one another in terms of aptitude, ability, or interest. And so, the racial composition of the student body of an elite magnet school would likely mirror the racial composition of the geographic area from which the students come. The reason to measure a policy change with reference to the state of affairs that would have occurred absent discrimination is so that a court may assess whether a disputed policy change, even when animated by a problematic motivation, nonetheless moves us closer or further away from this idealized state of affairs. When the change moves us closer to this state, even if poorly motivated, no harm (moral or constitutional) occurs. Or so the thought might go.

It is important to note that this account of PC’s normative rationale does not rest on the view that proportional representation as such is what fairness requires. Nor does it depend on what “really” would have happened in such an imagined counterfactual world. Rather, it rests on a sort of legal fiction about what the world would have been like absent discrimination. Hence, an objection claiming that even absent discrimination, there would be differences among racial and ethnic groups along multiple dimensions, perhaps due to cultural differences among different groups, erroneously takes the counterfactual claim in the vein of an empirical claim. The point instead is that in the absence of knowledge about what would have been the case in such a counterfactual world—whether, for instance, cultural differences would or would not create such differences—the law presumes that since people from different racial groups do not differ in ability or other meaningful ways, any observed differences must result from unjust social practices.Footnote ³⁰ By calling this presumption a “legal fiction,” we do not mean to suggest it is false. Rather, its status as legal fiction rests on the fact that while its underlying truth is unknowable or unprovable, the law adopts it as a lodestar for normative and expressive reasons.Footnote ³¹

To be sure, the PC is only a partial idealization of what the world would be like absent racial discrimination because the composition of the catchment area of the school itself would likely be very different from what it currently is absent prior race discrimination. The PC is not animated by a view about what the world as a whole would be like absent racial discrimination. Rather, it instantiates an epistemic commitment that absent prior race discrimination, there would be no differences in talents and abilities between racial groups and, for that reason, the population of admitted students would roughly match the racial composition of the catchment area.Footnote ³²

The kernel of normative value that underlies the PC is a commitment to a vision of racial equality in which racial differences make no difference in the aggregate to the abilities, opportunities, or achievement of students. It is a vision of a world freed from the after-effects of racial discrimination and thus, in a substantial way, freed from the significance of race in terms of station or status. This vision is borne out in a racially egalitarian distribution of key benefits and burdens, with the thought being that part of having achieved a racially just society is to have overcome the racially patterned outcomes that characterize our current social order.

VI. An unavoidable tension

We have seen in the preceding two sections how each of the BAC and PC standards expresses a view of the kind of effect that constitutes a distinctively discriminatory harm. In considering deviation from the status quo as the hallmark of discriminatory harm, the BAC risks carrying forward a history of group disadvantage and marginalization. But it also potentially pays heed to the efforts of individuals and groups at navigating that discrimination, and it suggests that there should be a presumption against policy changes which undermine those efforts. The BAC’s conception of discriminatory harm suggests, reasonably in our view, that actions that disadvantage a group in a way that traces back to the group’s encounters with discrimination in the past are themselves discriminatory. Furthermore, in attending to how groups adapt to existing social conditions, the BAC takes note of cases wherein policy changes would deepen the distributive harms that a disadvantaged group suffers or present a distinctively burdensome obstacle to a disadvantaged group’s ability to rationally pursue life plans. These too raise genuine worries of discrimination. The PC, on the other hand, sets forth a view of discriminatory harm that compares the effect of a given policy to a normative standard that is drawn from a society in which, without group stratification, it is imagined that the demographic profile of recipients of scarce competitive goods would reflect the demographics of some broader population. If this is the pattern of outcomes that would obtain under these idealized nondiscriminatory conditions, then any departures from this pattern, even departures that have been longstanding, are marks of a discriminatory state of affairs. To the extent that the PC standard sets forth a baseline from which discriminatory harm may be assessed that does not normalize past and present patterns of discrimination, it too offers a worthy perspective on what nondiscrimination requires.

The problem that arises in TJ and in conditions of injustice more broadly is that these two conceptions of discriminatory harm pull in opposite directions. For even when key social institutions are marked by injustice, some groups that are socially disadvantaged nonetheless rely on them and moreover may have come to rely on them precisely because of past discrimination. In these cases, the BAC is sensitive to the normative interest in safeguarding this reliance. The problem with adopting this standard of discriminatory harm wholesale is that there is also something rather perverse in requiring—for the sake of nondiscrimination no less—that an unjust status quo be preserved in order to ensure stability and reliance in this state of affairs. By contrast, the PC standard sets its sights exclusively on some normatively idealized state of affairs. So long as the status quo is unjust, the PC dictates that it must simply be overcome en route to this ideal.

On this way of seeing things, the BAC and PC conceptions of discriminatory harm mark a substantive normative conflict. And contrary to popular portrayals of this conflict, it is not best understood as one between different social groups—say Black and Latino students “versus” Asian students—that are differentially advantaged and disadvantaged by some state of affairs. Rather, we want to suggest that what is at stake here, and in many circumstances marked by injustice, are two different normative goods: that of affirming the forms of agency that have cropped up even in a world marked by unjust obstacles and that of making the world anew in light of past and present injustice. It is important to see that these goods need not track different social groups. The same disadvantaged group might have members who are variously empowered within and constrained by an unjust social system; in fact, the same individual might find themselves in one moment able to successfully maneuver conditions of injustice and in another moment, saddled by those very conditions. It is a fraught fact that the social structure that forms the terrain on which people create their lives is often both an unjust impediment and a condition on which agency is built. This is the profound conflict that gives rise to the opposing baselines against which discriminatory harm is assessed proposed by the BAC and PC: the choice between supporting the very real life plans forged in our actual conditions of past and present injustice and remaking the unjust social structure to be as it should be.

To be sure, there may be circumstances where the two normative concerns do not point in opposing directions and may instead be reconciled with one another. It is possible that in some cases, the best way to support the life plans of disadvantaged social groups now just is to remake unjust social structures; and vice versa, the best way to remake unjust social structures is to start from the real life plans of social groups today and ask what changes to the social structure would facilitate their development and realization. One might even say that this is how we should construct and pursue a robust program of nondiscrimination: by taking our cues from the real-life plans and forms of agency of marginalized social groups today. In pointing out the tension between building anew and supporting the lives built upon existing structures, we do not mean to take issue with this general line of thought. Rather, our point is simply that in an unjust world, many of the lives and plans that people live and make are grounded in those very structures that are part and parcel of the broader unjust social conditions. And this creates a conflict when considering how we should treat those structures, for it suggests that they should neither be straightforwardly dismantled nor continually supported because people have developed lives and plans that rely on them. Selective schools and the prevailing “meritocratic” standards for entrance are an especially illustrative example of this tension. Such schools are, at the same time, bottlenecks and ladders to opportunity. Barring broader interventions into the social structure, to change how they function in the first respect is to change how they function in the second. Changes in this context will therefore always be double-edged: a more perfect admissions process (granting such a thing) means that those who had organized their life around the previous process will be left in the lurch.

The practical question at hand is thus, which of the two standards best appreciates the sum total of the relevant normative considerations in our particular circumstances. That this conflict is an inescapable feature of life in an unjust world leads to a perhaps fruitful comparison between the divergence in how the BAC and PC standards treat the unjust status quo and two opposing orientations to theorizing about ethical matters in conditions of injustice. The BAC standard’s apparent sensitivity to how social life develops in conditions of injustice—in particular, how conditions of injustice differentially shape the opportunities and choices available to different groups and, in turn, how those groups have come to adapt to and rely on social institutions in different ways—seems to reflect an attentiveness to facts on the ground as they actually stand in our society. By contrast, the PC standard expresses a view of what constitutes a discriminatory effect that is taken from a vision of how goods in an idealized nondiscriminatory society would be distributed. It fixes its gaze on the pattern of outcomes that would obtain under some ideal set of circumstances and does not modify its prescription for the circumstances of our real world.

On this way of seeing things, the opposing BAC and PC standards embody opposing judgments as to the extent to which normative theorizing should incorporate facts about our actual nonideal circumstances in its judgments of what constitutes a discriminatory harm.Footnote ³³ That discrimination is a normative phenomenon borne out of an unjust world would seem to give an upper hand to the standard that does not idealize away those very features that give rise to discrimination as a social problem. So, that would seem to suggest a point against the PC. However, the conclusion that this orientation thus favors the BAC would be too quick. The BAC’s status quo baseline does not capture all of the morally relevant on-the-ground facts of our actual context. A standard that expresses such a strong presumption in favor of stability and takes policy change alone to indicate a potentially discriminatory effect does not reflect this rich set of considerations and treats as negligible the substantial moral costs of maintaining the status quo. Our nonideal circumstances also give us strong reasons to destabilize rather than stabilize our institutions. In the case of TJ, these reasons flow from, among other considerations, the entrenched disadvantage of the Black and Latino communities in the region, their longstanding exclusion from accessing key public goods in the area, and the effects that changing the distribution of the best educational resources could have on mitigating this disadvantage. The normative significance of these facts is lost in the BAC, which, in privileging the status quo ante as it in fact was, rules out all considerations about what ought to be.