Sarah P. Wessman, p.p.a. Henry Robert Wessman, v. Robert P. Gittens, Chairperson of the Boston School Committee, et al.

160 F.3d 790 (1998) Sarah P. WESSMANN, p.p.a. Henry Robert Wessmann, Plaintiff, Appellant, v. Robert P. GITTENS, Chairperson of the Boston School Committee, et al., Defendants, Appellees. No. 98-1657.

United States Court of Appeals, First Circuit. Heard July 30, 1998. Decided November 19, 1998.

791*791 Michael C. McLaughlin for appellant.

Chester Darling on brief for Citizens for the Preservation of Constitutional Rights, amicus curiae.

Frances S. Cohen, with whom Janet A. Viggiani, Hill & Barlow, Merita Hopkins, Corporation Counsel of the City of Boston, and Diane DiIanni, Special Assistant Corporation Counsel (Boston School Committee), were on brief, for appellees.

Elaine R. Jones, Theodore M. Shaw, Norman J. Chachkin, and Kimberly West-Faulcon, NAACP Legal Defense and Educational Fund, Ozell Hudson, Jr., Lawyers' Committee for Civil Rights Under Law of the Boston Bar Association, E. Macey Russell, Peabody & Arnold, Jonathan M. Albano, Denise J. Casper and Bingham Dana LLP on brief for Boston Branch, NAACP, and various individuals, amici curiae.

Before SELYA, BOUDIN and LIPEZ, Circuit Judges.

SELYA, Circuit Judge.

The City of Boston operates three renowned "examination schools," the most prestigious of which is Boston Latin School (BLS). The entrance points for admission to BLS occur principally at the seventh- and ninth-grade levels. In this litigation, plaintiff-appellant Henry Robert Wessmann, on 792*792 behalf of his minor child, Sarah P. Wessmann, challenges the constitutionality of BLS's admissions policy (the Policy). The district court rebuffed Wessmann's challenge. See Wessmann v. Boston Sch. Comm., 996 F.Supp. 120 (D.Mass.1998). On appeal, we must decide whether the Policy, which makes race a determining factor in the admission of a subset of each year's incoming classes, offends the Constitution's guarantee of equal protection. We conclude that it does.

I. BACKGROUND

We essay a brief historical reconnaissance to set the present dispute in perspective.

Over two decades ago, a federal district court adjudged the City of Boston (through its School Committee) to have violated the constitutional rights of African-American children by promoting and maintaining a dual public school system. See Morgan v. Hennigan, 379 F.Supp. 410, 480-81 (D.Mass. 1974) (Morgan I). Although the court found the school system as a whole guilty of de jure segregation, no specific evidence was produced to suggest that BLS's examination-based admissions policy discriminated against anyone or that those responsible for running BLS intended to segregate the races. See id. at 467-68. Nonetheless, BLS exhibited some of the symptoms of segregation: an anomalously low number of African-American students attended the school, see id. at 466 (tabulating statistics for examination schools), and the school had just changed its entrance testing methods pursuant to a consent decree settling charges that the earlier methods were themselves discriminatory, see id. at 467-68. These factors, combined with the City's inability to demonstrate that existing racial imbalances were not a result of discrimination, led the court to conclude that the City's examination schools (BLS included) were complicit in promoting and maintaining the dual system. See id. The presumption established by the Supreme Court in Keyes v. School Dist. No. 1, 413 U.S. 189, 210, 93 S.Ct. 2686, 37 L.Ed.2d 548 (1973), to the effect that a finding of intentional segregation in a "meaningful portion" of a school system suggests that other segregated schooling in the system is not accidental, played a pivotal role both in the district court's holding and in our ensuing affirmance. See Morgan v. Kerrigan, 509 F.2d 580, 594 (1st Cir.1974) (affirming Morgan I, 379 F.Supp. at 467).

The remedy adopted by the district court, among other things, obligated BLS to ensure that at least 35% of each entering class would be composed of African-American and Hispanic students. See Morgan v. Kerrigan, 401 F.Supp. 216, 258 (D.Mass.1975). Relying on the Keyes presumption, we affirmed this set-aside as part of a comprehensive plan to ameliorate pervasive and persistent constitutional infirmities throughout the Boston public schools. See Morgan v. Kerrigan, 530 F.2d 401, 425 (1st Cir.1976).

The Boston school system began gradually to mend its ways. By 1987, systemic progress permitted us to conclude that, for all practical purposes, the School Committee had achieved unitariness in the area of student assignments. See Morgan v. Nucci, 831 F.2d 313, 326 (1st Cir.1987). We based our conclusion not only on the distribution of students throughout the City's schools, but also on the good faith demonstrated by school administrators in conforming with the demands of meaningful change. See id. at 319-26. Because comparable improvement had not been accomplished in other areas, such as faculty and staff integration and the renovation of facilities, we instructed that federal court supervision of elements other than student assignment continue. See id. at 327-32. The district court thereupon relinquished control over student assignments, even while retaining active supervision over other aspects of the school system.

After 1987, the City's three examination schools — BLS, Boston Latin Academy, and the O'Bryant School — were no longer under a federal court mandate to maintain a 35% set-aside. Nevertheless, the School Committee remained committed to the policy until 1995, when a disappointed applicant challenged the setaside's constitutionality. The district court granted injunctive relief directing the complainant's admission to BLS. See McLaughlin v. Boston Sch. Comm., 938 F.Supp. 1001, 1018 (D.Mass.1996). The 793*793 School Committee then discontinued the 35% set-aside.

Concerned that the number of African-American and Hispanic students admitted to the examination schools might drop precipitously without a predetermined set-aside, school officials began researching alternative admissions policies in hopes of finding one that might prevent that result without offending the Constitution. The effort started in mid-1996 under the hegemony of Thomas Payzant, superintendent of the Boston public schools. Payzant commissioned Bain & Co. (Bain), a consulting firm, to review an array of admissions options ranging from lotteries to strict merit-selection plans and to report on how each option might affect the racial and ethnic composition of the examination schools' entering classes.

After Payzant informed the School Committee of Bain's preliminary findings, Robert P. Gittens, the School Committee chairman, appointed a task force to study the matter. The task force held meetings, hosted public hearings, and ultimately recommended the adoption of Bain's "Option N50." Bain's study showed that a major difference between Option N50 and some other possible alternatives (such as a strict merit-selection option) was that the former would minimize the diminution of black and Hispanic student admissions expected to result from abandonment of the 35% set-aside. Three members dissented from this recommendation. The School Committee nonetheless accepted Option N50, effective for the 1997-98 school year. Option N50 thereupon became the core of the Policy.

We recount the Policy's most salient features, leaving aside complexities not relevant to the case at hand. To gain admission to one of Boston's three examination schools, a student must take a standardized test. Based on a mathematical formula that purports to predict academic performance, school hierarchs combine each applicant's test score with his or her grade point average, derive a composite score, rank all applicants accordingly, and proceed to assign individuals to the applicant pool for the examination school(s) in which they have indicated an interest. To be eligible for admission to any of the examination schools, an applicant must be in the qualified applicant pool (QAP), a group composed of those who rank in the top 50% of the overall applicant pool for that particular school.

Half of the available seats for an examination school's entering class are allocated in strict accordance with composite score rank order. The other half are allocated on the basis of "flexible racial/ethnic guidelines" promulgated as part of the Policy. To apply these guidelines, school officials first determine the relative proportions of five different racial/ethnic categories — white, black, Hispanic, Asian, and Native American — in the remaining pool of qualified applicants (RQAP), that is, the QAP for the particular school minus those persons already admitted on the basis of composite score rank order alone. They then fill the open seats in rank order, but the number of students taken from each racial/ethnic category must match the proportion of that category in the RQAP. Because the racial/ethnic distribution of the second group of successful applicants must mirror that of the RQAP, a member of a designated racial/ethnic group may be passed over in favor of a lower-ranking applicant from another group if the seats allotted for the former's racial/ethnic group have been filled.

Sarah Wessmann encountered such a fate. BLS had 90 available seats for the 1997 ninth-grade entering class. Based on her composite score, Sarah ranked 91st (out of 705) in the QAP. To fill the first 45 seats, the school exhausted the top 47 persons on the list (two aspirants declined in order to accept invitations from another examination school). Had composite scores alone dictated the selection of the remainder of the ninth-grade entering class, Sarah would have been admitted. But the racial/ethnic composition of the RQAP was 27.83% black, 40.41% white, 19.21% Asian, 11.64% Hispanic, and 0.31% Native American. Consequently, the Policy required school officials to allocate the final 45 seats to 13 blacks, 18 whites, 9 Asians, and 5 Hispanics. As a result, black and Hispanic students whose composite score rankings ranged from 95th to 150th displaced Sarah 794*794 and ten other white students who had higher composite scores and ranks.

Acting to Sarah's behoof, her father sued a coterie of defendants (collectively, the School Committee), alleging that the Policy had defeated her candidacy and challenging its constitutionality. Following a 13-day bench trial, the district court held that the School Committee's interests in promoting a diverse student body and remedying vestiges of past discrimination were compelling, and that the means crafted by the School Committee to further these interests were not so expansive as to raise constitutional concerns. See Wessmann, 996 F.Supp. at 127-32. This appeal ensued.

II. ANALYSIS

We divide our analysis into four segments, beginning with the standards that govern our review, then addressing the general idea of "compelling governmental interests," and, finally, proceeding to consider seriatim the two justifications asserted by the School Committee in defense of the Policy.

A. Standards of Review.

The Supreme Court consistently employs sweeping language to identify the species of racial classifications that require strict scrutiny, see Adarand Constructors, Inc. v. Peña, 515 U.S. 200, 224, 115 S.Ct. 2097, 132 L.Ed.2d 158 (1995) (plurality op.) (concluding upon a review of the Court's precedents that government must "justify any racial classification subjecting [a] person to unequal treatment under the strictest judicial scrutiny"); Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 273, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) (plurality op.) (remarking that racial distinctions of "any sort" invite "the most exacting judicial examination") (citation and internal quotation marks omitted), and the Policy fits comfortably within this rubric. We conclude, therefore, that strict scrutiny is the proper standard for evaluating the Policy. Hence, the Policy must be both justified by a compelling governmental interest and narrowly tailored to serve that interest in order to stand.

The School Committee's rejoinder — that the Policy is not a quota — is a non sequitur. We agree that the Policy does not constitute a quota — at least not in the literal sense of an unchanging set-aside — but that fact gains the School Committee little ground. At a certain point in its application process — specifically, during the selection of the second half of each incoming class — the Policy relies on race and ethnicity, and nothing else, to select a subset of entrants. Thus, whether the Policy is truly a quota or whether it is best described otherwise is entirely irrelevant for the purpose of equal protection analysis. Attractive labeling cannot alter the fact that any program which induces schools to grant preferences based on race and ethnicity is constitutionally suspect. See Regents of Univ. of Cal. v. Bakke, 438 U.S. 265, 289, 98 S.Ct. 2733, 57 L.Ed.2d 750 (1978) (opinion of Powell, J.) (noting that regardless of whether the limitation at issue is described as "a quota or a goal," it is "a line drawn on the basis of race and ethnic status"); cf. Lutheran Church-Mo. Synod v. FCC, 141 F.3d 344, 354 (D.C.Cir.1998) (articulating similar sentiments anent employment preferences).

The School Committee also asserts an entitlement to more lenient review because the Policy neither benefits nor burdens any particular group. Under the flexible guidelines, the argument goes, the racial/ethnic distribution of the entering classes will change yearly, and thus, there is no real preference for any single group.

This assertion leads nowhere, for the manner in which the Policy functions is fundamentally at odds with the equal protection guarantee that citizens will be treated "as individuals, not as simply components of a racial, religious, sexual or national class." Miller v. Johnson, 515 U.S. 900, 911, 115 S.Ct. 2475, 132 L.Ed.2d 762 (1995) (citations and internal quotation marks omitted). Even though we may not know before the fact which individuals from which racial/ethnic groups will be affected, we do know that someone from some group will be benefitted and a different someone from a different group will be burdened. Because a court's obligation to review race-conscious programs 795*795 and policies cannot be made to depend "on the race of those burdened or benefitted by a particular classification," City of Richmond v. J.A. Croson Co., 488 U.S. 469, 494, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989) (plurality op.) (citations omitted), no more is exigible to bring strict scrutiny into play.[1]

A remaining issue under this heading concerns our review of the district court's findings and conclusions. We accord deferential review to specific findings of fact emanating from a bench trial. See Fed.R.Civ.P. 52(a). Here, however, because the issues advanced in this appeal — specifically, whether diversity and curing vestiges of past discrimination satisfy strict scrutiny — raise either questions of law or questions about how the law applies to discerned facts, our review is essentially plenary. See Veciños De Barrio Uno v. City of Holyoke, 72 F.3d 973, 978 (1st Cir.1995).

B. Compelling Interests: An Overview.

The question of precisely what interests government may legitimately invoke to justify race-based classifications is largely unsettled. Of course, we know that such state action is acceptable upon a showing, inter alia, that it is needed to undo the continuing legacy of an institution's past discrimination. See Miller, 515 U.S. at 920, 115 S.Ct. 2475. We also know that the Court has rejected the "role model" theory as a compelling interest. See Croson, 488 U.S. at 497-98, 109 S.Ct. 706. Beyond these examples, the case law offers relatively little guidance.

A few cases suggest (albeit in dictum) that remedying past discrimination is the only permissible justification for race-conscious action by the government. See, e.g., id. at 493, 109 S.Ct. 706 (stating that unless classifications based on race are "strictly reserved for remedial settings, they may in fact promote notions of racial inferiority and lead to a politics of racial hostility"). But in certain milieus, some courts have accepted race-based taxonomies that are not linked to remedying past discrimination, particularly in settings such as law enforcement and corrections. See Wittmer v. Peters, 87 F.3d 916, 919 (7th Cir.1996) (collecting cases);[2] see also Croson, 488 U.S. at 521, 109 S.Ct. 706 (Scalia, J., concurring) (stating that, "[a]t least where state or local action is at issue, only a social emergency rising to the level of imminent danger to life and limb" may justify race-conscious action).

In considering whether other governmental interests, beyond the need to heal the vestiges of past discrimination, may be sufficiently compelling to justify race-based initiatives, courts occasionally mention "diversity". At first blush, it appears that a negative consensus may be emerging on this point. The Wittmer court noted that the defendants "did not rely on generalities about racial balance or diversity" to justify their hiring program, suggesting that such an attempted justification would have lacked vitality. 87 F.3d at 920. Other courts have stated the conclusion more explicitly. See Lutheran Church-Mo. Synod, 141 F.3d at 354 (ruling out diversity as a compelling governmental interest in the employment context); Hopwood v. State of Texas, 78 F.3d 932, 948 (5th Cir.1996) (similar, in the educational context).

We think that any such consensus is more apparent than real. In the education context, Hopwood is the only appellate court to have rejected diversity as a compelling interest, and it did so only in the face of vigorous dissent from a substantial minority of the 796*796 active judges in the Fifth Circuit. See Hopwood v. State of Texas, 84 F.3d 720, 721 (5th Cir.1996) (Politz, C.J., with whom King, Wiener, Benavides, Stewart, Parker, and Dennis, JJ., joined, dissenting from denial of rehearing en banc). The question that divided the Fifth Circuit centered on the precedential value of Justice Powell's controlling opinion in Bakke. The panel in Hopwood pronounced that opinion dead. The dissenting judges countered that the reports of Bakke's demise were premature.

It may be that the Hopwood panel is correct and that, were the Court to address the question today, it would hold that diversity is not a sufficiently compelling interest to justify a race-based classification. It has not done so yet, however, and we are not prepared to make such a declaration in the absence of a clear signal that we should. See Agostini v. Felton, 521 U.S. 203, ____, 117 S.Ct. 1997, 2017, 138 L.Ed.2d 391 (1997). This seems especially prudent because the Court and various individual Justices from time to time have written approvingly of ethnic diversity in comparable settings, see, e.g., Wygant, 476 U.S. at 315, 106 S.Ct. 1842 (Stevens, J., dissenting); Washington v. Seattle Sch. Dist. No. 1, 458 U.S. 457, 472-73, 102 S.Ct. 3187, 73 L.Ed.2d 896 (1982), or have noted that the issue remains open, see Wygant, 476 U.S. at 286, 106 S.Ct. 1842 (O'Connor, J., concurring). But see Metro Broad., Inc. v. FCC, 497 U.S. 547, 614, 110 S.Ct. 2997, 111 L.Ed.2d 445 (1990) (O'Connor, J., dissenting) ("Like the vague assertion of societal discrimination, a claim of insufficiently diverse broadcasting viewpoints might be used to justify equally unconstrained racial preferences, linked to nothing other than proportional representation of various races.").

As matters turn out, we need not definitively resolve this conundrum today. Instead, we assume arguendo — but we do not decide — that Bakke remains good law and that some iterations of "diversity" might be sufficiently compelling, in specific circumstances, to justify race-conscious actions. It is against this chiaroscuro backdrop that we address the School Committee's asserted "diversity" justification for the Policy. Thereafter, we turn to its alternate justification: that the Policy is an appropriate means of remediating the vestiges of past discrimination.

C. Diversity.

The word "diversity," like any other abstract concept, does not admit of permanent, concrete definition. Its meaning depends not only on time and place, but also upon the person uttering it. See Towne v. Eisner, 245 U.S. 418, 425, 38 S.Ct. 158, 62 L.Ed. 372 (1918) (Holmes, J.) ("A word is not a crystal, transparent and unchanged, it is the skin of a living thought and may vary greatly in color and content according to the circumstances and the time in which it is used."); Hanover Ins. Co. v. United States, 880 F.2d 1503, 1504 (1st Cir.1989) (warning of the fallacy of believing that "a word is a word is a word"). It would be cause for consternation were a court, without more, free to accept a term as malleable as "diversity" in satisfaction of the compelling interest needed to justify governmentally-sponsored racial distinctions.

The School Committee demurs. Citing to Swann v. Charlotte-Mecklenburg Bd. of Educ., 402 U.S. 1, 16, 91 S.Ct. 1267, 28 L.Ed.2d 554 (1971) (stating that school authorities have "broad power to formulate and implement educational policy," including prescribing a specific percentage of minority students to attend each school "in order to prepare students to live in a pluralistic society") (dictum), it labors to persuade us that we would be warranted in deferring to its judgment because school officials necessarily enjoy substantial discretion in making education policy. We are not convinced.

The Swann song upon which the School Committee relies cannot be wrested from the score. Cf. Gomillion v. Lightfoot, 364 U.S. 339, 343-44, 81 S.Ct. 125, 5 L.Ed.2d 110 (1960) (admonishing that, "[p]articularly in dealing with claims under broad provisions of the Constitution, which derive content by an interpretive process of inclusion and exclusion, it is imperative that generalizations, based on and qualified by the concrete situations that gave rise to them, must not be applied out of context in disregard of variant controlling facts"). Swann was decided 797*797 when dual educational systems were a reality and efforts to dismantle them were being frustrated by school officials who demonstrated little ardor for implementing the mandates of desegregation. Chary that the exigencies of the need for change might precipitate a rush to judgment, the Justices confirmed that federal courts must put the horse before the cart, that is, they must diagnose some constitutional malady before beginning to dispense remedies. See Swann, 402 U.S. at 16, 91 S.Ct. 1267. Thus, the Swann dictum, properly construed, recognizes that a low percentage of minority students in a particular school does not necessarily betoken unconstitutional conduct, but may result from innocent causes (say, the population distribution of a given district), and warns that, unless a skewed enrollment pattern is caused by unconstitutional student assignment practices, federal courts must defer to school officials' discretion and refrain from imposing remedies.

This well-accepted principle does not help the School Committee. The Swann Court had no occasion to consider the question, central to this appeal, of whether and to what extent the Constitution circumscribes school officials' discretion to formulate and implement an admissions policy that embraces a particular brand of pluralism. Cf. Bakke, 438 U.S. at 314, 98 S.Ct. 2733 (opinion of Powell, J.) ("Although a university must have wide discretion in making the sensitive judgments as to who should be admitted, constitutional limitations protecting individual rights may not be disregarded."). In the end, then, the School Committee's reference to Swann only begs the question: Swann reiterated that federal courts must grant remedies where there are constitutional violations, and the question here is whether the School Committee itself has violated the Constitution. It follows that, in order to persuade us that diversity may serve as a justification for the use of a particular racial classification, the School Committee must do more than ask us blindly to accept its judgment. It must give substance to the word.[3]

The School Committee endeavors to meet this challenge primarily by lauding benefits that it ascribes to diversity. Drawing on the testimony of various witnesses (school administrators, experts, and alumni), the Committee asserts that, because our society is racially and ethnically heterogeneous, future leaders must learn to converse with and persuade those who do not share their outlook or experience. This imperative becomes even more urgent because technology, now more than ever, forces heretofore estranged nations and cultures to communicate and cooperate. For these reasons, the School Committee exhorts us to find that diversity is essential to the modern learning experience.

Stated at this level of abstraction, few would gainsay the attractiveness of diversity. Encounters between students of varied backgrounds facilitate a vigorous exchange of ideas that not only nourishes the intellect, but also furthers mutual understanding and respect, thereby eroding prejudice and acting as a catalyst for social harmony. Indeed, Justice Powell's opinion in Bakke acknowledges that these very attributes may render an educational institution's interest in promoting diversity compelling. See id. In the last analysis, however, the School Committee's reliance on generalizations undercuts its construct. If one is to limit consideration to generalities, any proponent of any notion of diversity could recite a similar litany of virtues. Hence, an inquiring court cannot content itself with abstractions. Just as Justice Powell probed whether the racial classification at issue in Bakke in fact promoted the institution's stated goals, see id. at 315-19, 98 798*798 S.Ct. 2733, we must look beyond the School Committee's recital of the theoretical benefits of diversity and inquire whether the concrete workings of the Policy merit constitutional sanction. Only by such particularized attention can we ascertain whether the Policy bears any necessary relation to the noble ends it espouses. In short, the devil is in the details.

By its terms, the Policy focuses exclusively on racial and ethnic diversity. Its scope is narrowed further in that it takes into account only five groups — blacks, whites, Hispanics, Asians, and Native Americans — without recognizing that none is monolithic. No more is needed to demonstrate that the School Committee already has run afoul of the guidance provided by the principal authority on which it relies: "The diversity that furthers a compelling state interest encompasses a far broader array of qualifications and characteristics of which racial or ethnic origin is but a single though important element." Id. at 315, 98 S.Ct. 2733. A single-minded focus on ethnic diversity "hinder[s] rather than further[s] attainment of genuine diversity." Id. Nor is the Policy saved because the student assignments that it dictates are proportional to the composition of the RQAP. See id. (noting that the adoption of a "multitrack" program "with a prescribed number of seats set aside each for identifiable category of applicants" would not heal the admissions plan's constitutional infirmity).

When we articulated this concern at oral argument, the School Committee's able counsel responded that it is unnecessary for the Policy to consider other indicia of diversity because BLS historically has been diverse with respect to everything but race and ethnicity. For empirical confirmation of this assertion, the School Committee points to Bain's handiwork. Having analyzed various admissions options, Bain suggested that all the options would result in substantial gender, neighborhood, and socioeconomic diversity, but that, unless race and ethnicity were explicitly factored into the admissions calculus, attainment of racial and ethnic diversity might be jeopardized. This attempted confirmation does not pass constitutional muster.

If, as we are told, diversity has been attained in all areas other than race and ethnicity, then the School Committee's argument implodes. Statistics compiled for the last ten years show that under a strict merit-selection approach, black and Hispanic students together would comprise between 15% and 20% of each entering class, and minorities, in toto, would comprise a substantially greater percentage. Even on the assumption that the need for racial and ethnic diversity alone might sometimes constitute a compelling interest sufficient to warrant some type of corrective governmental action, it is perfectly clear that the need would have to be acute — much more acute than the relatively modest deviations that attend the instant case. In short, the School Committee's flexible racial/ethnic guidelines appear to be less a means of attaining diversity in any constitutionally relevant sense and more a means for racial balancing. The Policy's reliance on a scheme of proportional representation buttresses this appearance and indicates that the School Committee intended mainly to achieve a racial/ethnic "mix" that it considered desirable. Indeed, Bain's Option N50 was chosen and incorporated into the Policy because it held out the promise of increasing minority representation over the roughly 18% that Bain anticipated would result on a strict merit-selection basis.

The testimony at trial amply confirms this suspicion. Superintendent Payzant testified that a "fair representation of a cross-section of students" of the Boston public schools would constitute a proper "reference point" for defining a "diverse mix" of students. The "cross-section" to which he referred is comprised of the proportions of seventh- and ninth-grade black, Hispanic, white, and Asian students enrolled in Boston's public high schools. Another "reference point" mentioned by Payzant was the "proportional representation" embodied by the Policy, which, given his other testimony, is ultimately designed to move in the direction of the same racial/ethnic distribution.[4] Other school officials, 799*799 such as Dr. Elizabeth Reilinger and Dr. Edwin Melendez, testified to like effect, sometimes invoking the notion of "underrepresentation."

We do not question the School Committee's good intentions. The record depicts a body that is struggling valiantly to come to terms with intractable social and educational issues. Here, however, the potential for harmful consequences prevents us from succumbing to good intentions. The Policy is, at bottom, a mechanism for racial balancing — and placing our imprimatur on racial balancing risks setting a precedent that is both dangerous to our democratic ideals and almost always constitutionally forbidden. See Freeman v. Pitts, 503 U.S. 467, 494, 112 S.Ct. 1430, 118 L.Ed.2d 108 (1992); Croson, 488 U.S. at 507, 109 S.Ct. 706. Nor does the School Committee's reliance on alleviating underrepresentation advance its cause. Underrepresentation is merely racial balancing in disguise — another way of suggesting that there may be optimal proportions for the representation of races and ethnic groups in institutions. See Lutheran Church-Mo. Synod, 141 F.3d at 352.

It cannot be said that racial balancing is either a legitimate or necessary means of advancing the lofty principles recited in the Policy. The closest the School Committee comes to linking racial balancing to these ideals is by introducing the concept of "racial isolation." The idea is that unless there is a certain representation of any given racial or ethnic group in a particular institution, members of that racial or ethnic group will find it difficult, if not impossible, to express themselves. Thus, the School Committee says, some minimum number of black and Hispanic students — precisely how many, we do not know — is required to prevent racial isolation.

Fundamental problems beset this approach. In the first place, the "racial isolation" justification is extremely suspect because it assumes that students cannot function or express themselves unless they are surrounded by a sufficient number of persons of like race or ethnicity. Insofar as the Policy promotes groups over individuals, it is starkly at variance with Justice Powell's understanding of the proper manner in which a diverse student body may be gathered. See Bakke, 438 U.S. at 318, 98 S.Ct. 2733. Furthermore, if justified in terms of group identity, the Policy suggests that race or ethnic background determines how individuals think or behave — although the School Committee resists this conclusion by arguing that the greater the number of a particular group, the more others will realize that the group is not monolithic. Either way, the School Committee tells us that a minimum number of persons of a given race (or ethnic background) is essential to facilitate individual expression. This very position concedes that the Policy's racial/ethnic guidelines treat "individuals as the product of their race," a practice that the Court consistently has denounced as impermissible stereotyping.[5] Miller, 515 U.S. at 912, 115 S.Ct. 2475.

In the second place, the School Committee has failed to give us a plausible reason why we should believe that racial balancing of any type is necessary to promote the expression of ideas or any of the other ideals referenced in the Policy. We assume for argument's sake — albeit with considerable skepticism — that there may be circumstances under which a form of racial balancing could be justified by concerns for attaining the goals articulated by the Policy. To justify something so antithetical to our constitutional jurisprudence, however, a particularly strong showing of necessity would be required. The School Committee has provided absolutely no competent evidence that the proportional representation promoted by the Policy is in any way tied to the vigorous exchange of ideas, let alone that, in such respects, it differs significantly in consequence from, say, a strict merit-selection process. Nor has the School Committee concretely demonstrated that the differences in the percentages of students resulting 800*800 from the Policy and other, constitutionally acceptable alternatives are significant in any other way, such as students' capacity and willingness to learn. To the contrary, the School Committee relies only on broad generalizations by a few witnesses, which, in the absence of solid and compelling evidence, constitute no more than rank speculation. Given both the Constitution's general prohibition against racial balancing and the potential dangers of stereotyping, we cannot allow generalities emanating from the subjective judgments of local officials to dictate whether a particular percentage of a particular racial or ethnic group is sufficient or insufficient for individual students to avoid isolation and express ideas.

This brings us full circle. Although Justice Powell endorsed diversity as potentially comprising a compelling interest, he warned that a proper admissions policy would be such that if an applicant "loses out" to another candidate, he will "not have been foreclosed from all consideration for that seat simply because he was not the right color or had the wrong surname." Bakke, 438 U.S. at 318, 98 S.Ct. 2733. The Policy does precisely what Justice Powell deemed anathematic: at a certain point, it effectively forecloses some candidates from all consideration for a seat at an examination school simply because of the racial or ethnic category in which they fall. That happened to Sarah Wessmann. It violated the Equal Protection Clause. See People Who Care v. Rockford Bd. of Educ., 111 F.3d 528, 538 (7th Cir.1997) (concluding that preventing "children who are not beneficiaries of past discrimination" from becoming cheerleaders solely because of their race is "a barefaced denial of equal protection").

Again, let us be perfectly clear. We are aware that two of our sister courts of appeals have suggested that diversity may never constitute a compelling governmental interest sufficient to warrant race-based classifications. See Lutheran Church-Mo. Synod, 141 F.3d at 354; Hopwood, 78 F.3d at 948. For purposes of resolving this appeal, however, we need not speak definitively to that vexing question. Experience is "the life of the law," Justice Holmes commented, and more probably ought to be said before this chapter of constitutional inquiry is closed. We conclude today only that the School Committee's Policy does not meet the Bakke standard and, accordingly, that the concept of "diversity" implemented by BLS does not justify a race-based classification.

D. Vestiges of Past Discrimination.

The School Committee endeavors, in the alternative, to uphold the Policy as a means of redressing the vestiges of past discrimination. The court below accepted this explanation. See Wessmann, 996 F.Supp. at 131. We do not.

Governmental bodies have a significant interest in adopting programs and policies designed to eradicate the effects of past discrimination. See Miller, 515 U.S. at 920, 115 S.Ct. 2475; Mackin v. City of Boston, 969 F.2d 1273, 1275 (1st Cir.1992). Before embarking on such projects, however, government actors must be able to muster a "strong basis in evidence" showing that a current social ill in fact has been caused by such conduct. See Croson, 488 U.S. at 500, 109 S.Ct. 706. In giving meaning to the phrase "strong basis in evidence," we are guided primarily by the Court's particularized analysis in Croson and by the "body of appellate jurisprudence [that] has developed to provide that label with meaningful content." Engineering Contractors Ass'n of S. Fla., Inc. v. Metropolitan Dade County, 122 F.3d 895, 909 (11th Cir.1997), cert. denied, ___ U.S. ____, 118 S.Ct. 1186, 140 L.Ed.2d 317 (1998).

The threshold problem that we confront in this instance is that the School Committee disclaims the necessity for such evidence. Its disclaimer rests on the premise that a decree issued in the quarter-century-old desegregation litigation mandates local authorities to remedy any racial imbalance occurring in the school system and thereby obviates the need for an independent showing of causation. This premise lacks force.

The decree in question was entered in 1994 by Judge Garrity, pursuant to our instructions in Morgan v. Nucci, 831 F.2d 313 (1st Cir.1987). The particular provision to which the School Committee refers is entitled "Permanent 801*801 Injunction." It enjoins the School Committee "from discriminating on the basis of race in the operation of the public schools of the City of Boston and from creating, promoting or maintaining racial segregation in any school or other facility in the Boston public school system." Nothing in the plain language of this provision requires school officials to undertake any affirmative action, let alone to adopt a race-based classification (such as is contained in the Policy). Perhaps more important, the cited provision is not (as the School Committee would have it) a mandatory injunction. Rather, it operates as a negative injunction, forbidding the defendants from engaging in the acts that supported the original cause of action.[6] As long as school officials do not engage in discrimination against minorities — and there is no evidence that such conduct persists at BLS — they have not violated the injunction.

The School Committee's contention that racial imbalance, without more, mandates action also is discordant with established precepts of constitutional law. Once there is a finding of unitariness and the "affirmative duty to desegregate has been accomplished," school authorities are not expected to make "year-by-year adjustments of the racial composition of student bodies" absent a "showing that either the school authorities or some other agency of the State has deliberately attempted to fix or alter demographic patterns to affect the racial composition of the schools." Swann, 402 U.S. at 32, 91 S.Ct. 1267; see also Freeman, 503 U.S. at 494, 112 S.Ct. 1430 ("Once the racial imbalance due to the de jure violation has been remedied, the school district is under no duty to remedy imbalance that is caused by demographic factors.").

That ends this aspect of the matter. We concluded over ten years ago that Boston had restored the unitariness of student assignments, see Nucci, 831 F.2d at 319-26, and there is no contention here that any municipal actor has attempted intentionally to subvert the demographic composition of BLS (or any other school, for that matter). Under such circumstances, neither the Constitution nor the 1994 decree impose a duty on Boston's school officials to ensure the maintenance of certain percentages of any racial or ethnic group in any particular school.

Because the 1994 decree turns out to be a blind alley, the School Committee must identify a vestige of bygone discrimination and provide convincing evidence that ties this vestige to the de jure segregation of the benighted past. See Freeman, 503 U.S. at 494, 112 S.Ct. 1430. To meet this challenge, the School Committee cites an "achievement gap" between black and Hispanic students, on the one hand, and white and Asian students, on the other, and claims that this gap's roots can be traced to the discriminatory regime of the 1970s and before.

The scope of what social phenomena the law considers vestiges of past discrimination presents an open question. The presumptive vestiges are the well-known factors that the Supreme Court enumerated in Green v. County Sch. Bd., 391 U.S. 430, 435, 88 S.Ct. 1689, 20 L.Ed.2d 716 (1968) (mentioning student assignments, faculty, staff, facilities, transportation, and extra-curricular activities). Since Green, federal courts have recognized other permutations, including "quality of education." Freeman, 503 U.S. at 492, 112 S.Ct. 1430. What this means and how it is to be measured are difficult questions. Rather than entering that debate, we accept arguendo the School Committee's position that, in principle, a documented achievement gap may act as an indicator of a diminution in the quality of education. Even so, whether an achievement gap is a vestige of past discrimination depends on whether there is satisfactory evidence of a causal connection.

The court below short-circuited this inquiry. Citing Judge Garrity's 1994 order, the court reasoned that, once there has been a past judicial finding of institutional discrimination, 802*802 no more evidence is needed to justify a policy that employs racial classifications. See Wessmann, 996 F.Supp. at 131 (stating that the 1994 order is a "manifestation" of the reality of vestiges of past discrimination and that it alone provides a "compelling basis" for adoption of the Policy). The lower court was wrong.

There are times when a history of discrimination, in itself, may supply a powerful evidentiary predicate sufficient to justify some race-conscious action. See, e.g., Boston Police Superior Officers Fed'n v. City of Boston, 147 F.3d 13 (1st Cir.1998). Nevertheless, such holdings do not hinge on the mere existence of a past judicial finding, but, rather, on a variety of considerations, including what transpired since the time of that finding. In Boston Police, for example, our detailed inquiry revealed not only that discrimination had been a fact of the past, but that it persisted in the Boston Police Department, and that relatively little had been done to alleviate the situation at certain levels. See id. at 19-23. The record showed, for instance, that notwithstanding the existence of a consent decree, progress toward the integration of the police force had been "halting" and "modest." Id. at 23. The statistical evidence supporting this view pertained not to a distant past, but to present realities. See id. at 21, 23. In the final analysis, it was the combination of all this evidence, and the detailed showing that the effects of earlier discriminatory conduct continued to the present, that underpinned our conclusions. Withal, we took pains to warn against indiscriminate reliance on history alone lest it permit the adoption of remedial measures "ageless in their reach into the past, and timeless in their ability to affect the future." Id. at 20-21 (citation and internal quotation marks omitted).

In sum, whether past discrimination necessitates current action is a factsensitive inquiry, and courts must pay careful attention to competing explanations for current realities. See Freeman, 503 U.S. at 495-96, 112 S.Ct. 1430 (explaining that "though we cannot escape our history, neither must we overstate its consequences in fixing legal responsibilities"). The mere fact that an institution once was found to have practiced discrimination is insufficient, in and of itself, to satisfy a state actor's burden of producing the reliable evidence required to uphold race-based action.[7] See id. at 496, 112 S.Ct. 1430; Middleton v. City of Flint, 92 F.3d 396, 409 (6th Cir.1996).

Beyond history, the School Committee offers statistical and anecdotal evidence to satisfy its burden of demonstrating a strong evidentiary basis for the inauguration of remedial policies. The district court found the evidence favoring race-conscious remedies to be adequate, but the court's entire treatment of the subject comprises a lone paragraph composed of unrelievedly conclusory observations. See Wessmann, 996 F.Supp. at 131. In the absence of specific findings, we could remand. Given the time constraints applicable to the case, we opt instead to exercise plenary review, taking the statistical and anecdotal evidence in the manner suggested by the School Committee. See Veciños De Barrio Uno, 72 F.3d at 989 (observing that appellate courts ordinarily "should fill in blanks in the district court's account when the record and the circumstances permit this to be done without short-changing the parties").

The centerpiece of the School Committee's showing consists of statistical evidence addressed to a persistent achievement gap at the primary school level between white and Asian students, on the one hand, and black 803*803 and Hispanic students, on the other. One way to measure the achievement gap is in terms of relative performance on standardized tests. Over the years, whites and Asians have scored significantly higher, on average, than blacks and Hispanics. The School Committee theorizes that, because of this achievement gap, BLS receives fewer African-American and Hispanic applicants than otherwise might be the case, and even in comparison to this modest universe, an abnormally small number of black and Hispanic students qualify for admission. Accordingly, the Committee concludes that the statistics documenting the achievement gap, on their own, satisfy the "strong basis in evidence" requirement.

In mounting this argument, the School Committee relies heavily on a line of cases addressing affirmative action plans designed to remedy vestiges of past employment discrimination. See, e.g., Peightal v. Metropolitan Dade County, 26 F.3d 1545 (11th Cir. 1994); Stuart v. Roache, 951 F.2d 446 (1st Cir.1991). This reliance is mislaid. Fundamental differences distinguish the statistical inquiry involved in the employment discrimination context from the one proposed by the School Committee here. In employment discrimination cases, we know ex ante the locus of discrimination: it is the barrier to entry. Based on that premise, an appropriate statistical analysis compares the number of qualified minority applicants with those who gain entrance. The greater the disparity, the stronger the inference that discrimination is the cause of non-entry. See Croson, 488 U.S. at 501-02, 109 S.Ct. 706; Stuart, 951 F.2d at 451.

In this case, the "barrier to entry" comparable to those in the employment discrimination cases is BLS's requirement of an entrance examination and the resultant composite score — and no one (least of all, the School Committee) claims that the examination or any component thereof is discriminatory in operation or effect, or that it would be discriminatory if it were used as the sole criterion for admission. Such a claim was central to our conclusion in Stuart, 951 F.2d at 451, and it is totally absent here. What is more, any such claim would make precious little sense in the context of the School Committee's argument, for standardized achievement tests (a component of the entrance examination) are the primary objective measurement of the asserted achievement gap.

With the admissions process eliminated as an illegitimate barrier to entry, the achievement gap statistics, by themselves, must specifically point to other allegedly discriminatory conduct in order to suggest a causal link between those discriminatory acts and the achievement gap. Unlike the focused inquiry characteristic of the employment discrimination cases, however, the raw achievement gap statistics presented in this case do not by themselves isolate any particular locus of discrimination for measurement. Without such a focus, the achievement gap statistics cannot possibly be said to measure the causal effect of any particular phenomenon, whether it be discrimination or anything else. Cf. McCleskey v. Kemp, 481 U.S. 279, 294-95, 107 S.Ct. 1756, 95 L.Ed.2d 262 (1987) (contrasting legitimacy of inferences drawn from focused use of statistical methods in employment discrimination and venire-selection cases, on the one hand, with those drawn from application of general statistics to explain source of decisions in specific trials and sentencing proceedings). As such, the achievement gap statistics, by themselves, do not even eliminate the possibility that they are caused by what the Court terms "societal discrimination." Shaw v. Hunt, 517 U.S. 899, 909-10, 116 S.Ct. 1894, 135 L.Ed.2d 207 (1996); see also Coalition to Save Our Children v. State Bd. of Educ., 90 F.3d 752, 776-78 (3d Cir. 1996) (affirming findings that poor performance and achievement gap were caused by socioeconomic factors unrelated to past discrimination). To be sure, gross statistical disparities at times may suffice to satisfy a state actor's burden of production. See Croson, 488 U.S. at 501, 109 S.Ct. 706. But the achievement gap statistics adduced here fail to do so because it is unclear exactly what causative factors they measure.

The Croson Court relied on precisely this reasoning when it concluded, in the contractor context, that low minority membership in a local trade association "standing alone, cannot 804*804 establish a prima facie case of discrimination." Id. at 503, 109 S.Ct. 706. The Court reasoned that there could be "numerous explanations for this dearth of minority participation, including past societal discrimination." Id. Therefore, if such statistics are to be at all probative of discrimination, they must link cause and effect variables in a manner which would permit such an inference. See id. ("For low minority membership in these associations to be relevant, the city would have to link it to the number of local MBEs eligible for membership. If the statistical disparity between eligible MBEs and MBE membership were great enough, an inference of discriminatory exclusion could arise.").

We do not propose that the achievement gap bears no relation to some form of prior discrimination. We posit only that it is fallacious to maintain that an endless gaze at any set of raw numbers permits a court to arrive at a valid etiology of complex social phenomena. Even strong statistical correlation between variables does not automatically establish causation. See Tagatz v. Marquette Univ., 861 F.2d 1040, 1044 (7th Cir.1988); Ste. Marie v. Eastern R.R. Ass'n, 650 F.2d 395, 400 (2d Cir.1981) (Friendly, J.). On their own, the achievement gap statistics here do not even identify a variable with which we can begin to hypothesize the existence of a correlation.

The School Committee attempts to compensate for this shortcoming by pointing to certain alleged phenomena that it claims constitute substantial causes of the achievement gap. Chief among these is "low teacher expectations" vis-à-vis African-American and Hispanic students, a condition which the School Committee argues is an attitudinal remnant of the segregation era. To show the systemic nature of this alleged phenomenon, the School Committee leans heavily on the testimony of Dr. William Trent. Dr. Trent, a sociologist, identified teachers' low expectations of African-American and Hispanic students as a significant factor underlying the achievement gap in the Boston public schools. He based his conclusion on an analogy that he drew from studies he had performed in the Kansas City school system, including a "climate survey" of teacher attitudes and a multiple regression analysis designed to determine whether the low expectations reflected in teachers' answers to the questions posed in the climate survey might partially explain the achievement gap. Based on these materials, Dr. Trent had concluded that, in Kansas City, teacher "efficacy" — a term of art referring to a teacher's success in encouraging pupils to succeed — correlated with higher achievement test scores.

One difficulty with Dr. Trent's testimony is that it relies on evidence from one locality to establish the lingering effects of discrimination in another. Dr. Trent noted, for example, that data he examined from the Boston public schools revealed patterns "consistent with" his findings concerning the Kansas City schools. Croson, however, reaffirmed the Court's longstanding teaching that we must staunchly resist attempts to substitute speculation about correlation for evidence of causation. See Croson, 488 U.S. at 504, 109 S.Ct. 706 ("We have never approved the extrapolation of discrimination in one jurisdiction from the experience of another.") (citing Milliken v. Bradley, 418 U.S. 717, 746, 94 S.Ct. 3112, 41 L.Ed.2d 1069 (1974)). In this context, the Court emphasized that although government may adopt race-conscious remedies when there is evidence of lingering vestiges, it "must identify that discrimination, public or private, with some specificity" before it adopts the remedy. See id. at 504, 109 S.Ct. 706. At the very least, this would require solid evidence that Boston teachers have low expectations of minority students, and that these low expectations are related in a statistically significant way to the achievement gap.

Dr. Trent, however, never conducted a "climate survey" for the Boston school system. His conclusions for Boston were based only on a review of statistical data documenting the achievement gap (basically, the statistics regarding achievement test results and differing application and enrollment rates), statistics concerning teacher seniority, and anecdotal evidence about teacher attitudes supplied by school officials. When asked on cross-examination whether the data 805*805 that he relied on for his conclusions anent teacher attitudes were scientifically gathered, Dr. Trent responded in the negative. Dr. Trent thus freely conceded that the data he used was not of the quality necessary to satisfy the methodological rigors required by his discipline. Because Dr. Trent failed to follow his own prescribed scientific methodology for collecting data on the one issue central to his hypothesis about achievement gap causation, the trial court could not credit his conclusions.

An "opinion has a significance proportioned to the sources that sustain it." Petrogradsky Mejdunarodny Kommerchesky Bank v. National City Bank, 253 N.Y. 23, 25, 170 N.E. 479, 483 (1930) (Cardozo, J.). When scientists (including social scientists) testify in court, they must bring the same intellectual rigor to the task that is required of them in other professional settings. See Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 113 S.Ct. 2786, 125 L.Ed.2d 469 (1993); Braun v. Lorillard, Inc., 84 F.3d 230, 234 (7th Cir.1996); see also People Who Care, 111 F.3d at 537 (declaring, in reviewing admissibility of social science evidence purporting to quantify causes of achievement gap, that "the methods used by the expert to derive his opinion [must] satisfy the standards for scientific methodology that his profession would require of his out-of-court research"). The only excuse that Dr. Trent proffered for his failure to follow proper protocols was that a thorough study would have required more time than he had available. That is unacceptable. See Braun, 84 F.3d at 234. An expert witness can only deviate from accepted methods of scientific inquiry in ways that are consistent with the practices and usages of the scientific community.

The shortcomings in Dr. Trent's testimony largely relate to his failure to gather data systematically and point up the pitfalls that the School Committee invited by failing to validate the Policy in advance. Dr. Trent's charge was to trace the causal relationship, if any, between teacher attitudes and poor student performance. His failure to obtain reliable data disabled him from taking even the first step, for he could not validly establish whether Boston teachers' attitudes in fact were discriminatory, let alone show that they caused (or even significantly contributed to) the achievement gap. This first step is a cornerstone of the entire research project; in its absence, Dr. Trent could not legitimately eliminate other variables (including societal discrimination) that might explain the achievement gap in the Boston public schools.[8] See Croson, 488 U.S. at 503, 109 S.Ct. 706; People Who Care, 111 F.3d at 537. It follows inexorably that, with no methodological support, he could not produce a meaningful analysis of causation and, accordingly, his conclusions cannot bear the weight of the School Committee's thesis. See Mid-State Fertilizer Co. v. Exchange Nat'l Bank, 877 F.2d 1333, 1339 (7th Cir.1989) ("An expert who supplies nothing but a bottom line supplies nothing of value to the judicial process.").

Dr. Trent's reliance on anecdotal evidence fares no better. As a general matter, anecdotal evidence is problematic because it does not tend to show that a problem is pervasive. See Coral Const. Co. v. King County, 941 F.2d 910, 919 (9th Cir.1991) ("While anecdotal evidence may suffice to 806*806 prove individual claims of discrimination, rarely, if ever, can such evidence show a systematic pattern of discrimination necessary for the adoption of an affirmative action plan."). Thus, even though anecdotal evidence may prove powerful when proffered in conjunction with admissions or valid statistical evidence, anecdotal evidence alone can establish institutional discrimination only in the most exceptional circumstances. See Engineering Contractors Ass'n, 122 F.3d at 925-26 (collecting cases).

This case falls well within the general rule rather than the long-odds exception to it. The most specific testimony regarding low teacher expectations came from Deputy Superintendent Janice Jackson, whom the district court qualified as an expert witness. At one point in her career, Deputy Superintendent Jackson had received some training related to improving interactions between teachers and students (a program known as TEASA, an acronym for "teacher expectation and student achievement") by dispelling teachers' "unconscious biases." Without disparaging Ms. Jackson's credentials, we reject the contention that her observations, as presented at trial, provide any justification for a conclusion that a statistically meaningful number of Boston school teachers have low expectations of minority students.

Deputy Superintendent Jackson testified to two sets of observations. The first occurred before she joined the school system. She spent three months in classrooms as a "blind researcher," making certain observations for an unspecified research project.[9] In the process, she apparently visited six or seven schools (although she could recall only two by name). She testified that, during this time, she observed teachers treat minority and non-minority students differently. The differential treatment included calling on one set of pupils but not on another, disparate reprimands for the same behavior, and failure to push for "higher order thinking." Her second set of observations occurred once she became deputy superintendent. Here, her testimony is unrelievedly general. Although she visited numerous schools during this interval, her testimony is silent as to the purpose of these visits. It is equally silent as to crucial details, e.g., how many classrooms she visited or how many teachers she observed.

The short of it is that, to the extent that Ms. Jackson purposed to testify as an expert, she was obligated to present her data and her methodology with some specificity. In lieu of specifics, Ms. Jackson offered only her unsubstantiated recollection of past events. Croson reminds us that, in considering the legitimacy of race classifications, we must not blindly defer to the other branches' assurances that a particular condition exists. See Croson, 488 U.S. at 501-02, 109 S.Ct. 706; see also Hayes v. North State Law Enforcement Officers Ass'n, 10 F.3d 207, 214 (4th Cir.1993) (applying Croson and concluding that police chief's testimony did not constitute strong basis in evidence to justify race-conscious remedy even though the testimony was "based on his significant experience in the field of law enforcement"). Because Ms. Jackson was unable to quantify her observations in any manner whatsoever, the district court could not validly conclude that the number of "problem" teachers she observed was statistically significant.[10] Indeed, Dr. Trent was unable to utilize these observations to conduct any statistical analysis to provide a meaningful account of achievement gap causation. This inability illustrates one reason why anecdotal evidence, which includes testimony based on significant personal experience, rarely suffices to provide a strong basis in evidence.

The remaining evidence upon which the School Committee and the dissent relies — most notably the testimony of BLS Headmaster Michael Contompasis and Dr. Melendez — likewise exemplifies the shortcomings of anecdotal evidence. One cannot conclude 807*807 from the isolated instances that these witnesses recounted that low teacher expectations constitute a systemic problem in the Boston public schools or that they necessarily relate to the de jure segregation of the past.

Similarly unpersuasive are the School Committee's broad generalizations about socialization. Superintendent Payzant, for example, testified that he believed that teachers who started working in the era of dual school systems (who comprise approximately 28 % of current faculty) had been "socialized and shaped" by the thoughts and attitudes of the prior period. This may be so, but he gave no empirical evidence to confirm his conclusion. While the idea of "socialization" may be intellectually elegant, courts must insist on seeing concrete evidence. Without such substantiation, Payzant's testimony is no more compelling than the conclusory statements of the Richmond city councillor rejected as insufficient by the Supreme Court in Croson, 488 U.S. at 500-01, 109 S.Ct. 706.

To the extent that the School Committee notes other causal factors or indicia of discrimination, they, too, are insufficient either to show ongoing vestiges of system-wide discrimination or to justify a race-conscious remedy. For instance, pointing to the instability of leadership at the superintendent level, the School Committee suggests — so far as we can gather; the argument is neither developed nor clear — that it led to the absence of a standardized curriculum, which in turn has contributed to the achievement gap. If the argument is that the superintendent's office had a revolving door and thereby caused the achievement gap, then we see nothing that helps advance the contention that the achievement gap is a vestige of past discrimination. If the argument is that the superintendent's office has tacitly endorsed disparate curricula which in turn have caused the achievement gap, then the School Committee has provided us with absolutely no evidentiary basis to conclude that this is a consequence of discrimination; among other things, it has failed either to identify which curricula were discriminatory or to explain why they were so. It likewise has failed to identify the schools that adopted the discriminatory curricula in order to suggest a relationship between race and the existence of substandard school programs. Thus, it is impossible to tell, even on the assumption that differing curricula caused the achievement gap, whether this was a consequence of discriminatory conduct. In a nutshell, there is not a shred of evidence in the record supporting the contention that unstable leadership and the absence of uniform curriculum standards bore any relationship either to discrimination in the Boston schools or to the existence of the achievement gap.

We add an eschatocol of sorts. Even if the School Committee had proven the requisite connection between the Policy and the vestiges of past discrimination, the Policy could not endure. When authorized by the Constitution, race-conscious remedies not only must respond to a compelling governmental interest, but also must be narrowly tailored to rectify the specific harms in question. See Croson, 488 U.S. at 493, 109 S.Ct. 706 (explaining that any race-conscious means adopted to remedy past discrimination must be so narrowly tailored that there is "little or no possibility that the motive for the classification was illegitimate racial prejudice or stereotype"). Under this test, the Policy sweeps too broadly.

We limit our remarks in this regard to three points. First, since there is no discrimination at entry to BLS, we fail to see how the adoption of an admissions policy that espouses a brand of proportional representation is designed to ameliorate the harm that allegedly has occurred (a system-wide achievement gap at the primary school level). Indeed, when Deputy Superintendent Jackson was asked whether the flexible racial/ethnic guidelines in any way affect low teacher expectations, she responded that it was a "complicated question to answer" because, given the relatively recent implementation of the guidelines, there had "not even been an opportunity to check that." If a high-level school official cannot confirm that a race-conscious remedy will alleviate the purported major cause of a remnant of discrimination, we do not comprehend how that means can be narrowly tailored.

808*808 Second, the increased admission of black and Hispanic students cannot be viewed as partial compensation for injustices done at the primary school level. This is so because the victims of the achievement gap are public school students, and they are the ones who ought to be the focus of the remedy. The Policy does not focus in this direction, for many of the black and Hispanic students admitted under it come from private or parochial schools. Thus, the Policy is not sufficiently particularized toward curing the harm done to the class of actual victims. See Podberesky v. Kirwan, 38 F.3d 147, 158-59 (4th Cir.1994). After all, there is no reason to assume that granting a remedy to one member of a particular race or ethnic group comprises a condign remedy for harm done to another, especially when those who have been harmed are easily identifiable and still within the institution that allegedly suffers from the vestiges of past discrimination.

Third, if palliating the effects of past discrimination is the ostensible justification for the Policy, then the Policy, on its face, has been crafted in puzzling ways. Suppose that in a particular year a group of Hispanic students does very well, such that they cluster between ranks 45 and 90, but that the Hispanic student population in the RQAP is sparse. Suppose further that whites and Asians form a significant majority of the RQAP. There is then a likelihood that, by reason of the Policy, a number of the Hispanic students — archetypal victims of discrimination — will be displaced by white and Asian students. Nor need we resort to hypotheticals to see such effects. At the O'Bryant School, the Policy's flexible racial/ethnic guidelines resulted in the rejection from the 1997 ninth-grade entering class of two Hispanic students in favor of a white student. Then, too, given the School Committee's position that Asian students have not been victims of discrimination, we are unable to comprehend the remedial purpose of admitting Asian students over higher-ranking white students, as happened in the case of Sarah Wessmann. This brings us back to the point of our beginning: in structure and operation, the Policy indicates that it was not devised to assuage past harms, but that it was simply a way of assuring racial/ethnic balance, howsoever defined, in each examination school class. See Croson, 488 U.S. at 506, 109 S.Ct. 706 ("The random inclusion of racial groups that, as a practical matter, may never have suffered from discrimination in the construction industry in Richmond suggests that perhaps the city's purpose was not in fact to remedy past discrimination."). The Constitution forbids such a focus.

III. CONCLUSION

We do not write on a pristine page. The Supreme Court's decisions in Croson and Adarand indicate quite plainly that a majority of the Justices are highly skeptical of racial preferences and believe that the Constitution imposes a heavy burden of justification on their use. Croson, in particular, leaves no doubt that only solid evidence will justify allowing race-conscious action; and the unsystematic personal observations of government officials will not do, even if the conclusions they offer sound plausible and are cloaked in the trappings of social science. See Hayes, 10 F.3d at 214 (noting the dangers of relying on "subjective evidence" to justify race-conscious policies); see also Wittmer, 87 F.3d at 919 (reminding us that "common sense undergirded the pernicious discrimination against blacks now universally regretted").

Our dissenting brother's valiant effort to read into Croson a broad discretion for government entities purporting to ameliorate past discrimination strikes us as wishful thinking. The Croson Court's own reference to the need for a "searching judicial inquiry," 488 U.S. at 493, 109 S.Ct. 706, and its rejection of Justice Marshall's position, see id. at 494-95, 109 S.Ct. 706; id at 535-36, 109 S.Ct. 706 (Marshall, J., dissenting), both suggest an attitude that is antipathetic to those who yearn for discretion. And unless and until the Justices reconfigure their present doctrine, it is the job of judges in courts such as this to respect the letter and spirit of the Supreme Court's pronouncements.

We need go no further.[11] While we appreciate the difficulty of the School Committee's 809*809 task and admire the values that it seeks to nourish, noble ends cannot justify the deployment of constitutionally impermissible means. Since Boston Latin School's admissions policy does not accord with the equal protection guarantees of the Fourteenth Amendment, we strike it down. The judgment of the district court must therefore be reversed.

We are mindful that Henry Wessmann asks not only that we declare the Policy unconstitutional — which we have done — but also that his daughter, Sarah, be admitted to BLS forthwith. The School Committee, which has vigorously defended the Policy, has tacitly conceded that, if its defense fails, Sarah should be allowed to enroll at BLS. The circumstances of this case are unusual, for the school year is under way and Sarah Wessmann — who already has spent elsewhere the first year and some months of what normally would be a four-year matriculation at BLS — does not have the luxury of time that a remand would entail. We therefore direct the district court to enter a judgment, in appropriate form, that, inter alia, commands Sarah's admission to BLS without delay.

Reversed.

BOUDIN, Circuit Judge, concurring.

Judge Selya's opinion demonstrates that the school committee plan under attack here does involve racial preference, whatever the complexity of the plan and subtlety in expression. Yet, this court concluded more than a decade ago that purposeful discrimination had ended so far as assignments were concerned and that the school committee was proceeding in good faith. See Morgan v. Nucci, 831 F.2d 313, 326 n. 19 (1st Cir.1987). To survive, the school committee's plan must serve a compelling state interest and be narrowly tailored to achieve that interest. Croson, 488 U.S. at 505-08, 109 S.Ct. 706; Wygant, 476 U.S. at 274, 106 S.Ct. 1842.

The foremost interest urged by the school committee is diversity, a ground that may or may not prove "compelling" to the Supreme Court. But even if diversity were an adequate ground, it has not been shown that this plan is necessary to achieve it. The record shows that minorities will be included in BLS in substantial numbers without the plan. Op. at 797-98. If some specific higher level is needed to achieve diversity of views and backgrounds, this has not been demonstrated in the record.

The alternative interest urged on this record is that the plan is necessary to remedy the residual effects of admitted past discrimination. The remnants on which the school committee relies are supposed teacher attitudes, specifically, a lower expectation of achievement by minorities in the Boston public schools. These attitudes are said to be linked backward in time to mind-sets developed during a regime of purposeful discrimination and forward to explain poorer performance by minority public-school students on the tests necessary for entry into the elite public schools. The remedy is the racial preference embodied in the plan.

In theory, low expectations could be caused by past discrimination and could have some effect on current performance. But here the quality of the supporting evidence is far from overwhelming. In this case, there is no study of Boston schools, only one for Kansas City; and the evidence as to Boston, while offered through experts, is largely based on general statements or anecdote. It is open to question whether such evidence can withstand Croson's requirement of a "searching judicial inquiry." Croson, 488 U.S. at 493, 109 S.Ct. 706.

In all events, the plan fails to meet the Supreme Court's narrow tailoring requirement. The plan is clearly overbroad when judged by the past-discrimination rationale; it provides preferences to minority groups that were never discriminated against by the Boston School authorities or affected by lowered expectations of public school teachers 810*810 (Asians and private school African-American applicants in particular). See Croson, 488 U.S. at 506, 109 S.Ct. 706; Wygant, 476 U.S. at 284 n. 13, 106 S.Ct. 1842. There is some reason to think that African-American private school applicants rather than public school students, may well be the principal beneficiary of the preferences created by the plan.

There is also no indication that the plan will do anything to alter the expectations of public-school teachers, which are claimed to be the source of the residual discrimination. One of the school committee's own experts, asked to say whether or how the plan would resolve this problem, was unable to supply an answer. Op. at 807. Another school committee expert, answering a direct question from the judge as to whether the plan's racial preference "attacked the problem of teacher attitude," essentially conceded that it did not. Tr. 6-126 to -127. These admissions suggest a further misfit between plan and remedy.

Finally, because the plan does not address the supposed cause of the problem, teacher attitudes formed in the ancien regime, the same arguments now urged to sustain the plan will be available for the indefinite future. Teachers retire slowly and themselves teach those who succeed them. The plan thus creates just the kind of "timeless" racial preferences of concern to the Supreme Court. Wygant, 476 U.S. at 276, 106 S.Ct. 1842; see Croson, 488 U.S. at 498, 505, 109 S.Ct. 706; cf. United States v. Paradise, 480 U.S. 149, 178, 107 S.Ct. 1053, 94 L.Ed.2d 203 (1987).

None of these defects of fit is surprising because, at bottom, the plan is not seriously suited to be a temporary measure to remedy low teacher expectations that are the supposed remnants of discrimination. It is instead a thoughtful effort to assist minorities historically disadvantaged while, at the same time, preserving the essentially competitive character of the schools in question. So viewed, there is no misfit between problem and remedy; the only misfit is with Croson's requirements for the use of racial preferences, requirements that only the Supreme Court can relax.

LIPEZ, Circuit Judge, dissenting.

Under the Equal Protection Clause of the Fourteenth Amendment, all racial or ethnic classifications by government actors are highly suspect and will be upheld only if they withstand strict judicial scrutiny. To meet the strict scrutiny standard, a challenged racial classification must serve a compelling governmental interest and must be narrowly tailored to achieve that goal. See Adarand Constructors, Inc. v. Peña, 515 U.S. 200, 224-25, 115 S.Ct. 2097, 132 L.Ed.2d 158 (1995). The Boston School Committee argues that the Boston Latin admissions program serves two compelling interests: promoting diversity in the public schools and remedying the vestiges of past discrimination. The majority rejects both arguments. Although I have reservations about the Committee's diversity argument on the facts of this case, I have none about its remedial argument. The district court properly found that the Boston School Committee had a strong basis in evidence for determining that the Boston Latin admissions program serves a compelling government interest in remedying the effects of prior discrimination, and that the program is narrowly tailored to achieve that goal.

I will explain more fully my disagreement with the majority by dividing my opinion into four parts: a summary of the Boston schools desegregation litigation that is particularly relevant to the development of the Boston Latin admissions program; a description of the proper legal framework for analyzing the evidence of the remedial interest; an analysis of the evidence establishing that interest; and an analysis of the program's narrow tailoring.

I.

Some Relevant History

Although the majority opinion provides an excellent background statement placing the present dispute in perspective, I wish to offer some additional background which more fully reveals the antecedents of the Boston Latin admissions program in the long history of court supervised desegregation of the Boston school system. That court-supervised desegregation 811*811 began in 1974, when the district court found that the Boston School Committee had engaged in "affirmative acts [which] intentionally created or maintained racial segregation." Morgan v. Hennigan, 379 F.Supp. 410, 427 (D.Mass.1974) (Morgan I). The Committee's acts included the manipulation of facilities utilization, new school construction, and redistricting to preserve distinctively white or minority districts. See id. at 425-41. The district court noted that the Committee was particularly successful in maintaining a segregated school system through the establishment and use of "feeder patterns" for students going from primary schools into the high schools. The primary schools were segregated in part because their districts were geographically drawn and based on residentially-segregated neighborhoods. As the high schools were far fewer in number than the primary schools, and each high school drew on several geographic primary school districts, the high schools might naturally have been more racially balanced than the primary schools. However, under the Committee's skewed feeder patterns, students were assigned from white "junior high" schools into predominantly white high schools beginning with the tenth grade, and from African-American "middle" schools into predominantly African-American high schools beginning with the ninth grade. See id. at 441-49. In the limited cases where students were allowed to transfer voluntarily into schools with vacant seats, the Committee manipulated the transfer program, alternately liberalizing it to allow whites to transfer out of African-American schools, or controlling it to keep African-Americans from transferring out of African-American schools. See id. at 449-56. The district court also found that the Committee took steps with regard to faculty and staff assignment systemwide, and admission to the selective citywide schools, in order to preserve the segregated status quo. See id. at 456-69.[12]

Relying on the presumptive logic set forth in Keyes v. School District No. 1, 413 U.S. 189, 93 S.Ct. 2686, 37 L.Ed.2d 548 (1973), the district court concluded that pervasive de jure segregation throughout the Boston public school system established a "prima facie case of unlawful segregative design" in the examination schools. Morgan I, 379 F.Supp. at 467, quoting Keyes, 413 U.S. at 208, 93 S.Ct. 2686. This presumption was not rebutted. See Morgan I, 379 F.Supp. at 467-68. Although little evidence was presented on the causes of the exam schools' imbalanced racial composition, the court attributed part of the disparity to the tying of admission to achievement (as opposed to aptitude) test scores.[13] The court also noted that the "advanced work classes" that prepared public school fifth and sixth graders for the examination schools were segregated. See id. at 433, 484 n. 16. On appeal, we summarized the situation as follows: "The examination schools are segregated because black children fare worse on the entrance examinations than whites. These children are products of the segregated elementary classes which constituted `tracks' to the examination schools and were more than 80% white.... Thus, the segregation of the lower schools had inevitable consequences for the examination ... schools...." Morgan v. Kerrigan, 509 F.2d 580, 594 n. 20 (1st Cir.1974) (affirming Morgan I).

In the subsequent remedial phase of the desegregation litigation, the court considered a proposal to use the SSAT, an aptitude rather than an achievement test, as part of the admissions process for the exam schools. This test would presumably reflect less of the disparity in elementary education among applicants resulting from the segregation of their schools. See Morgan v. Kerrigan, 401 F.Supp. 216, 243-44 (D.Mass.1975) (Morgan II). Although the district court also noted that there had been discussion about using "the median score as a floor for admissions," id. at 243, the district court declined to set an "arbitrary" examination-score minimum for the purpose of promoting desegregation of 812*812 the examination schools, imploring the parties to work towards solutions themselves. Id. at 244. The then-Headmaster of Boston Latin School stated in an affidavit that, in his practical experience, the median score was a good minimum standard for selecting students able to succeed at Boston Latin. Id. at 243. While the Headmaster "concede[d] that this standard [was] an assessment based on his experience, not on racial data or studies that would show that students scoring below that mark would be unable to learn and succeed within the Latin schools' program," id., this standard was accepted by the parties and has survived to this day as the conclusion that the top half of the applicant pool is "qualified" to succeed at Boston Latin.

As a remedy for its finding of discrimination by the Boston School Committee, the court ultimately mandated that "[a]t least 35% of each of the entering classes at Boston Latin School, Boston Latin Academy and Boston Technical High in September 1975 shall be composed of black and Hispanic students." Id. at 258. In addition, the court imposed the following requirements:

The School Department shall also institute and conduct programs (a) to make all students in the system aware of the admission requirements and type of instruction offered at the examination schools, and (b) to recruit black and Hispanic applicants to the examination schools in future years.

Any tutorial programs given to prepare students for entrance examinations shall be conducted on a desegregated basis, as shall advanced work classes (if they are to be continued). Any enrichment and remedial programs for students admitted to or enrolled in the examination schools shall be available and conducted on a desegregated basis. There shall be no tracking of students within the examination schools which results in racially segregated classes.

Id. at 258-59. In reviewing the district court ruling, we concluded that the 35% floor was not an impermissible racial quota because "overall racial composition [of the system was ordered] to be used as a starting point in designing a school desegregation plan. This approach is specifically approved in Swann [Swann v. Charlotte-Mecklenburg Board of Education, 402 U.S. 1, 91 S.Ct. 1267, 28 L.Ed.2d 554 (1971)]." Morgan v. Kerrigan, 530 F.2d 401, 423 (1st Cir.1976) (affirming Morgan II).

The issue of racial composition was considered especially important in the examination schools. For most other schools in the system, the district court set minority enrollment targets only within a broad statistical range. The racial/ethnic composition of any individual school was allowed a 25% deviation from the average racial/ethnic composition of the public school system as a whole. Such a wide margin of error was not necessary or appropriate for citywide schools[14] (such as the examination schools) for two reasons. First, the practical problems of implementing more complete desegregation in these schools were less significant than in the noncitywide schools, where additional busing was the main instrument for additional desegregation. Although busing posed a hardship for students who might otherwise have attended schools near their homes, most students attending citywide schools were already being bused to school. Second, minority enrollment at the citywide schools had to reflect more accurately the racial/ethnic composition of the system because the citywide schools were viewed as valuable vehicles for voluntary desegregation that could not have "[s]ignificant departures" from the overall composition and still attract students seeking an integrated learning environment. See 530 F.2d at 423. The target figure for overall African-American/Hispanic levels at the exam schools, 35%, was set slightly lower than the percentage of African-Americans and Hispanics in the system as a whole because there was a significant population of Asians at the exam schools already, bringing total levels of "black and other minority" enrollment closer to the 44% rough target set for other citywide schools. See Morgan II, 401 F.Supp. at 244.

In 1987, some twelve years after the initial court desegregation orders, we held that 813*813 court supervision of student assignments in the Boston schools should end in light of the Committee's good faith, the State Board's findings of compliance, and the general notion that the maximum practicable level of desegregation had been achieved (given the impact of factors exoteric to School Committee control, such as white flight). Morgan v. Nucci, 831 F.2d 313 (1st Cir.1987) (Nucci). In deference to the longstanding national tradition of local school autonomy, control over the student assignment process was returned to the Boston School Committee, with the express reservation that "`unitariness' ... in all aspects of the Boston schools has not yet been achieved." Id. at 318.[15] Finally, in 1994 the district court issued a "Final Judgment" in the Morgan litigation, and "permanently enjoined" the Committee "from creating, promoting or maintaining racial segregation in any school or other facility in the Boston Public school system." Morgan v. Burke, Civil Action 72-911G, (D.Mass. July 19, 1994), at 5 (Morgan IV).

The School Committee voluntarily continued the 35% African-American/Hispanic set-aside at the exam schools even after this policy ceased in 1987 (with Nucci) to be a court mandate. In 1996, however, the set-aside effectively ended with McLaughlin v. Boston School Committee, when the district court issued a preliminary injunction ordering the admission to Boston Latin of a student who challenged the set-aside's constitutionality. 938 F.Supp. 1001 (D.Mass.1996). In making that ruling, the court acknowledged that

the set aside had its origins in a court-ordered desegregation plan; that it had already played a crucial and successful role in desegregating the once virtually all-white examination schools; that the First Circuit had in no way called its validity into question in returning to the [Boston School Committee] control over student assignments; and that the appeals court had, if anything, strongly hinted that in-place corrective measures ought to be continued. Indeed, it may not be too much of a stretch to say that, without the set aside, there would not have been a finding of unitariness with respect to student assignments at all.

Id. at 1016. The court then warned that "abandonment of the 35% set aside at the present time without adopting other remedial measures would, within the next six years or sooner, convert BLS into an overwhelmingly white and Asian-American school with a black and Hispanic enrollment of about 15%." Id. at 1008.

Nonetheless, the district court found a likelihood of success for the disappointed white applicant Julia McLaughlin on the narrow tailoring aspect of her challenge to the admissions program. The district court cited the fact that "the set aside still does not have a built-in termination provision." Id. at 1016. It also noted that the Boston School Committee did "not appear to have explored the feasibility of less racially preferential plans for keeping BLS [the Boston Latin School] accessible to `qualified' students of all races and ethnicities, and not just to those students who, for a myriad of reasons, tend to excel on standardized examinations at the tender age of 12," even though "such plans may well be available for defendants' consideration."[16] Id. The court offered as one possible alternative to the rigid 35% set-aside a periodically 814*814 updated target for minority enrollment, with "percentages [tailored] to the relevant qualified applicant populations." In light of the preliminary injunction, the Committee replaced the set aside, and the district court's suggestion of a floating target was influential in molding the set-aside's immediate replacement, the current admissions program.[17]

II.

The Legal Framework for an Analysis of the Evidence

In rejecting the district court's conclusion that the School Committee's current admissions program for Boston Latin "appropriately addressed the vestiges of discrimination that linger in the Boston Public School system," Wessmann v. Boston School Committee, 996 F.Supp. 120, 131 (D.Mass.1998), the majority asserts that the School Committee failed to present satisfactory evidence of a causal connection between the achievement gap documented by the Committee and the prior de jure segregation of the Boston schools. I disagree with this conclusion. In my view, the majority judges the Committee's proof of causation unsatisfactory because the majority misperceives the Committee's evidentiary burden in defending its affirmative action program. That point requires some elaboration.

The law is clear that a public entity adopting an affirmative action program to remedy the lingering effects of past discrimination must have a "strong basis in evidence" for concluding that the lingering effects it identifies are causally linked to past discrimination. See City of Richmond v. J.A. Croson Co., 488 U.S. 469, 500, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989) (quoting Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 277, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) (plurality opinion)). The more elusive question is what factual predicate constitutes a "strong basis in evidence" justifying action by the public entity. From my reading of the relevant caselaw, I conclude that this "strong basis" requirement is met preliminarily if the public entity whose affirmative action program is challenged in court demonstrates that the entity adopted the program on the basis of evidence sufficient to establish a prima facie case of a causal link between past discrimination and the current outcomes addressed by the remedial program. If this prima facie case is not effectively rebutted by a reverse discrimination plaintiff, the public entity has met its burden of establishing a compelling remedial interest. I now turn to an analysis of the cases.

There are no educational context reverse-discrimination opinions from the Supreme Court which give us authoritative guidance on what constitutes a strong basis in the evidence sufficient to warrant remedial action. Contrary to the view of the majority, I think it is appropriate to look to analogous cases from the employment context. In that context, the Supreme Court has required that

a public employer ... must ensure that, before it embarks on an affirmative-action program, it has convincing evidence that remedial action is warranted. That is, it must have sufficient evidence to justify the conclusion that there has been prior discrimination.

Evidentiary support ... becomes crucial when the remedial program is challenged in court.... In such a case, the trial court must make a factual determination that the employer had a strong basis in evidence for its conclusion that remedial action was necessary.... [U]nless such a determination is made, an appellate court ... cannot determine whether the race-based action is justified as a remedy for prior discrimination.

Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 277-78, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) (Powell, J., plurality opinion) (emphasis added). In his earlier opinion in Bakke, Justice Powell (joined by no other Justices on this point) had stated that a "judicial, legislative or administrative" finding of prior discrimination was an absolute precondition 815*815 to establishing a remedial program. Regents of the Univ. of California v. Bakke, 438 U.S. 265, 307-09, 98 S.Ct. 2733, 57 L.Ed.2d 750 (1978).[18] Concerned that this requirement of a prior finding might be read into Justice Powell's opinion in Wygant, Justice O'Connor wrote in her concurring opinion in Wygant that "a contemporaneous or antecedent finding of past discrimination by a court or other competent body is not a constitutional prerequisite to a public employer's voluntary agreement to an affirmative action plan." Wygant, 476 U.S. at 289, 106 S.Ct. 1842 (O'Connor, J., concurring). Although it may seem excessively semantic to distinguish between the prerequisite of a court or other competent body's contemporaneous or antecedent finding of past discrimination to justify a voluntary affirmative action program, and the lesser prerequisite of a public entity's strong basis in evidence for concluding that such remedial action is necessary, there is actually much at stake in that distinction. As Justice O'Connor noted in her concurrence in Wygant:

... [P]ublic employers are trapped between the competing hazards of liability to minorities if affirmative action is not taken to remedy apparent employment discrimination and liability to nonminorities if affirmative action is taken. Where these employers, who are presumably fully aware both of their duty under federal law to respect the rights of all their employees and of their potential liability for failing to do so, act on the basis of information which gives them a sufficient basis for concluding that remedial action is necessary, a ... requirement [that they have a court's actual finding of their own past discrimination before they act] should not be necessary. Wygant, 476 U.S. at 291, 106 S.Ct. 1842 (O'Connor, J., concurring). This concern for the vulnerable position of a public entity with a remedial duty led Justice O'Connor to conclude that such an entity should not have to wait for its own liability to minorities to be proved conclusively in litigation before it could undertake remedial action.

These concerns were again reflected in Justice O'Connor's concurrence in Johnson v. Transportation Agency, 480 U.S. 616, 107 S.Ct. 1442, 94 L.Ed.2d 615 (1987). In Johnson, the Court upheld a voluntary affirmative action program instituted by a local transportation agency. The program allowed, inter alia, the agency to consider gender as a factor in the applicant's qualifications for employment. In her concurrence, Justice O'Connor noted that "[w]hile employers must have a firm basis for concluding that remedial action is necessary ... Wygant ... [does not] place[] a burden on employers to prove that they actually discriminated against women or minorities." Id. at 652, 107 S.Ct. 1442 (O'Connor, J., concurring). She explained that an employer would have a firm basis in the evidence, sufficient to justify remedial action, "if it can point to a statistical disparity sufficient to support a prima facie claim under Title VII by employee beneficiaries of the affirmative action plan of a pattern or practice claim of discrimination." Id. at 649, 107 S.Ct. 1442.

Justice O'Connor incorporated these standards from her Wygant and Johnson concurrences into an opinion for a majority of the Court only a few years later in City of Richmond v. J.A. Croson Co., 488 U.S. 469, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989). In Croson, the Court invalidated a minority subcontracting preference voluntarily instituted by 816*816 the city of Richmond because the city had not shown "any identified discrimination in the Richmond construction industry." Id. at 505, 109 S.Ct. 706. Justice O'Connor stated that the city had failed to justify its remedial goals with a "`strong basis in evidence.'" Id. at 500, 109 S.Ct. 706 (citation omitted). The city had produced "nothing approaching a prima facie case of a constitutional or statutory violation by anyone in the Richmond construction industry" sufficient to justify the city's use of subcontracting preferences to remedy its own role in perpetuating that discrimination. Id. (§ III-B, opinion of the court) (emphasis added). If the city had presented evidence supporting a prima facie claim under Title VII by the minority subcontractors, remedial action would have been warranted. In the absence of this evidence, the preference was an illegitimate racial classification.

In Stuart v. Roache, 951 F.2d 446 (1st Cir.1991), cert. denied, 504 U.S. 913, 112 S.Ct. 1948, 118 L.Ed.2d 553 (1992), we relied on Croson in discussing the burden of proof required of a public entity to justify a voluntary affirmative action plan giving minority police officers a preference in promotion to sergeant. We stated that "[a] majority of the Supreme Court in Croson used the words `strong basis' and `prima facie case' in [the context of voluntary race-conscious remedial action by a public entity.] Hence, that is the evidentiary standard that we use." 951 F.2d at 450 (citing Croson, 488 U.S. at 500, 109 S.Ct. 706, and Wygant, 476 U.S. at 292, 106 S.Ct. 1842). We noted in Stuart that the presence of statistical disparities creating a prima facie case "at least casts doubt on the fairness of the promotion process and requires further explanation" by the reverse-discrimination plaintiff.[19] 951 F.2d at 451. To elaborate on this last statement, we cited partially to this section of Justice O'Connor's concurrence in Wygant:

[A] public employer must have a firm basis for determining that affirmative action is warranted.... For example, demonstrable evidence of a disparity [between percentages of minorities on a staff and in a] relevant labor pool sufficient to support a prima facie Title VII pattern or practice claim by minority teachers would lend a compelling basis for a competent authority such as the School Board to conclude that ... a voluntary affirmative action program is appropriate....

To be sure, such a conclusion is not unassailable. If a voluntary affirmative action plan is subsequently challenged in court by nonminority employees, those employees must be given the opportunity to prove that the plan does not meet the constitutional standard this Court has articulated. However, as the plurality suggests, the institution of such a challenge does not automatically impose upon the public employer the burden of convincing the court of its liability for prior unlawful discrimination; nor does it mean that the court must make an actual finding of prior discrimination based on the employer's proof before the employer's affirmative action plan will be upheld.... In "reverse discrimination" suits, as in any other suit, it is the plaintiffs who must bear the burden of demonstrating that their rights have been violated.... [W]hen the Board introduces its statistical proof as evidence of its remedial purpose, thereby supplying the court with the means for determining that the Board had a firm basis for concluding that remedial action was appropriate, it is incumbent upon the nonminority teachers to prove their case; they continue to bear the ultimate burden of persuading the court that the Board's evidence did not support an inference of prior discrimination and thus a remedial purpose[.]

Wygant, 476 U.S. at 292-93, 106 S.Ct. 1842 (O'Connor, J., concurring).

Justice O'Connor's concurrence in Wygant, relied upon by us in Stuart, provides a clear statement of the strong basis in evidence 817*817 required of a public entity defending an affirmative action employment program in court: if minority beneficiaries of a voluntary program could, in the program's absence, have met the initial burden of production in a Title VII pattern-or-practice suit (a "prima facie case") on the basis of the evidence relied upon by the public entity, then the entity is justified in preemptively enacting a voluntary remedial affirmative action program. Put another way, the "strong basis in evidence" required of a government entity defending an affirmative action employment program in court is comparable to the evidentiary burden imposed on a minority plaintiff who makes a claim of a pattern or practice of employment discrimination under Title VII. Both must meet the standard for a prima facie case. In the reverse discrimination context, however, the public entity defending a challenge to its affirmative action employment program in court does so by presenting the evidence of a prima facie case of discrimination that a putative Title VII minority plaintiff might have presented in a potential lawsuit.

I must pause in this discussion of the relevant employment discrimination case law to make two points which are important to an understanding of the evidentiary burden of the School Committee. The School Committee asserts that evidence of differential success rates for African-American and Hispanic students applying to the examination schools from the advanced work classes in the Boston Public Schools, as compared to their white and Asian classmates, and evidence that African-American and Hispanic students apply to the examination schools at half the rate of white and Asian students, without more, constitute a "strong basis in evidence" sufficient to justify a remedial affirmative action program. In its amicus brief, the NAACP asserts that statistical evidence of the discriminatory effects of overreliance at the exam schools on composite score as a basis for admission, without more, provides a strong basis in evidence sufficient to justify a remedial affirmative action program. Both of these disparate impact contentions are wrong because they do not account for the centrality of proof of discriminatory animus in justifying a race-conscious remedial program. That centrality accounts for Justice O'Connor's reference to the strong basis in evidence provided for race-conscious remedial action by a prima facie Title VII pattern or practice claim. Unlike Title VII disparate impact claims, Title VII pattern or practice claims require evidence of discriminatory animus.[20] In addition to providing evidence of disparate impact as part of its strong basis in evidence for remedial relief, the reverse discrimination defendant must link that disparate impact to discriminatory animus, whether recent, or, as asserted in this case by the School Committee, rooted in a more distant history of discrimination.

In the Supreme Court employment cases I have discussed, the reverse discrimination defendant sought to establish that its affirmative action program was necessary to remedy harm caused by recent discrimination. The disadvantaged plaintiff challenged the existence of the discrimination itself, not whether that discrimination had actually harmed minorities. In response to that challenge to the existence of the discrimination itself, Justice O'Connor made the point in Wygant and Johnson that the public employer did not have the burden of convincing the court of its liability for prior unlawful discrimination, nor did the court have to make a finding of prior discrimination based on the 818*818 employer's proof to uphold the affirmative action program. Instead, the public employer only had to demonstrate that it had a strong basis in evidence for concluding that there was prior discrimination. Prima facie evidence of a Title VII pattern or practice claim that could be brought by minority plaintiffs would be such evidence.

In the instant case, where there is a long history of court findings of discriminatory acts by the School Committee, there is no dispute about the public entity's responsibility for prior discrimination. Instead, given the time lapse between those court findings of discrimination and the claim that this discrimination still disadvantages minorities, the issue in dispute here is whether the vestiges of that prior discrimination now affect minorities. This showing necessarily requires evidence of the existence of vestiges of the prior discrimination and of a present harm to minorities. It also requires evidence of causal connections between the past discrimination and the claimed vestiges, and between the vestiges and the present harm. Through these connections, the School Committee must show that the need for remedial action was a consequence of the effects of past discrimination. Nevertheless, Justice O'Connor's strong basis in evidence/prima facie analysis still defines the evidentiary burden of the School Committee in presenting its proof of the causal relationship between prior discrimination and the present effects on minorities.

The district court in the McLaughlin case recognized the applicability of Justice O'Connor's evidentiary framework:

[W]hile the party seeking to justify an affirmative action measure — here the defendant BSC [Boston School Committee] defending use of the 35% set aside — bears the initial burden of producing "a strong basis in evidence" in support of the measure, the "ultimate burden" of proof rests with the party challenging it to prove that it is unconstitutional. [citing Wygant] ... [T]he burden on the BSC is merely one of production: it must demonstrate that there is "a strong basis in evidence for [its] conclusion that remedial action was necessary." [citing Concrete Works of Colo. v. Denver, 36 F.3d 1513, 1521-22 (10th Cir. 1994)]. If it does so, the ultimate burden of proving that the evidence before the BSC did not reasonably support an inference of prior discrimination ... rests with the plaintiff.

McLaughlin, 938 F.Supp. at 1010. In the instant case, the Boston School Committee had the burden of demonstrating to the court that the evidence that impelled it to adopt the admissions program for Boston Latin met the prima facie standard set forth by Justice O'Connor in Wygant, Johnson and Croson and applied by us in Stuart: that is, the Committee had to show that the evidence before it would constitute a prima facie case of a constitutional or statutory violation if brought by a putative minority plaintiff. If this prima facie case was not effectively rebutted by Wessmann, who always retains the burden of persuasion, the School Committee prevails.

The majority finds the School Committee's reliance on cases addressing affirmative action plans designed to remedy vestiges of past employment discrimination inapt.

In this case, the "barrier to entry" comparable to those in the employment discrimination cases is BLS's requirement of an entrance examination and the resultant composite score — and no one (least of all, the School Committee) claims that the examination or any component thereof is discriminatory in operation or effect, or that it would be discriminatory if it were used as the sole criterion for admission. Such a claim was central to our conclusion in Stuart, 951 F.2d at 451, and it is totally absent here. What is more, such a claim would make precious little sense in the context of the School Committee's argument, for standardized achievement tests (a component of the entrance examination) are the primary measure of the asserted achievement gap.

I disagree with this analysis. Precisely as in the employment cases, there is an identifiable barrier to entry that could be challenged by minority applicants in the event the race-conscious aspects of the Boston Latin admissions program were elided. Amicus curiae NAACP makes this claim on appeal; it tried 819*819 doggedly to intervene below; and its position explains why the School Committee could not, in the context of litigation both current and anticipated, freely admit that composite score ranking may have had a discriminatory impact if used alone.[21] Despite this constraint, the School Committee did allude to the notion of discriminatory impact, as already noted, and there was evidence presented at trial that, for African-Americans and Hispanics, composite score was not reliably correlated with future performance at BLS. The Committee also took pains to indicate that its stipulation at trial that Wessmann would have been admitted had a straight-rank-order system been used was not a concession that such an admissions scheme was acceptable.

Although the admissions program challenged here was "voluntary," in the sense of not being impelled by the 1994 order or any other court proceeding or consent decree, the mere act of making a selection among students seeking admission to Boston Latin exposed the Committee to legal action by minority students. As Justice O'Connor characterized the analogous employment situation, "public employers are trapped between the competing hazards of liability to minorities if affirmative action is not taken to remedy apparent employment discrimination and liability to nonminorities if affirmative action is taken." Wygant, 476 U.S. at 291, 106 S.Ct. 1842 (O'Connor, J., concurring). The Wygant case dealt with racial preferences in employment, an arena where the shadow of Title VII suits by disappointed minority job seekers inevitably looms over a public employer using selection criteria with the potential to produce an unjustifiable "disparate impact" on minorities. The same considerations apply in the context of a selective public secondary school: the mere act of selection exposes the school system to challenge from minorities based on the disparate impact of the selection criteria used. Here, the relevant provision of federal law is Title VI of the Civil Rights Act of 1964, 42 U.S.C. § 2000d (West 1998), stating that "[n]o person in the United States shall, on the ground of race, color, or national origin, be excluded from participation in, be denied the benefits of, or be subjected to discrimination under any program or activity receiving Federal financial assistance."

There was evidence produced at trial that any exclusive focus on composite score in admissions had a disparate impact on African-Americans and Hispanics. Although the ability of minority plaintiffs to make colorable claims of Title VI violations would not be sufficient to justify a race-conscious affirmative action program (such claims would only be sufficient to force the Committee to find some alternative to composite score rank order admissions), the presence of the disparate impact underlying such claims, when causally related to the history of de jure segregation in the system, imposed on the School Committee a duty to ensure that it did not violate the Constitution by using selection criteria that perpetuated the effects of past governmental discrimination. As already noted, the district court in McLaughlin was skeptical of the legality of simply reverting to the use of composite score ranking, stating just two years ago that "abandonment of the ... set aside at the present time without adopting other remedial measures would, within the next six years or sooner, convert BLS into an overwhelmingly white and Asian-American school...." 938 F.Supp. at 1008 (emphasis added). The Committee chose a remedial measure for admission to the Boston Latin School in the wake of a long history of desegregation orders and under the threat of Title VI suits by disappointed minority applicants if no affirmative action were taken.

Just as the courts have always encouraged consensual resolutions to desegregation cases,[22] Congress' intent to encourage voluntary compliance with the requirements of 820*820 Title VI (and VII, for that matter) has always been a backdrop to the scheme of evidentiary burdens the federal courts have placed on litigants pursuant to that legislation. See Regents of the Univ. of California v. Bakke, 438 U.S. 265, 336, 98 S.Ct. 2733, 57 L.Ed.2d 750 (1978) (Brennan, White, Marshall, and Blackmun, JJ., concurring in part) (Title VI); Johnson v. Transportation Agency, 480 U.S. 616, 630 n. 8, 107 S.Ct. 1442, 94 L.Ed.2d 615 (1987) (Title VII). Similarly, in cases where there may be a duty to counteract the effects of past government discrimination, the Supreme Court has set evidentiary standards that facilitate a voluntary remedy. A government entity need not admit conclusive guilt for past discrimination's current effects before going forward with a remedial plan. Instead, it must satisfy the court that the evidence before it established a prima facie case of a causal link between past discrimination and the current outcomes addressed by the remedial program. If this prima facie case is not effectively rebutted by a reverse discrimination plaintiff, who always retains the burden of proving the illegality of the affirmative action program, the government has met its burden of establishing a compelling remedial interest under strict scrutiny analysis. With this legal framework in mind, I turn to an analysis of the evidence on the vestiges of discrimination.

III.

The Evidence on the Vestiges of Discrimination

The district court found that "[t]he overwhelming evidence presented at trial confirmed the Boston School Committee's basis for concluding that remedial action was necessary...." 996 F.Supp. at 131. The majority says we should exercise plenary review of that decision, "taking the statistical and anecdotal evidence in the manner suggested by the School Committee." I agree with the majority's approach. I disagree with its conclusion.

1. The Achievement Gap

The evidence at trial revealed large gaps between African-Americans and Hispanics, on the one hand, and whites and Asians, on the other, in admissions to the exam schools, and in achievement and allocation of resources throughout the system. The most significant of these is the persistent, static gap in achievement between African-American and Hispanic students and white and Asian students. The gap is measured by achievement test scores of fifth and sixth graders on the Metropolitan Achievement Test and the Stanford 9. Expert witness analysis of the test results over a several year period showed a persistent and relatively unchanging gap in achievement in all subject matters which correlated with race: African-American and Hispanic students fared much worse on the tests than whites. The tests also revealed that, in general, Asians fared worse than whites on language skills achievement.

The evidence also demonstrated that African-Americans and Hispanics from the public schools apply to the examination schools at half the application rate of other students in the public schools. Even within the special advanced work classes which are designed to prepare Boston public elementary school students for the examination schools, African-Americans and Hispanics fared worse in the examination schools' admissions process than whites. Finally, there was evidence that African-American and Hispanic applicants to Boston Latin from private elementary schools do much better in the admissions process than their Boston public school counterparts. The existence of these "gaps" was undisputed at the trial.

2. The Relationship of the Achievement Gap to Prior Discrimination

The Committee presented evidence of a connection between low teacher expectations for minority students and low minority performance on achievement tests. The Committee presented further evidence that these low teacher expectations for minority students are prevalent in the Boston school system, and that this prevalence is a vestige of the long years of segregation in the Boston school system. Given these connections between student achievement, teacher expectations for students, and the impact of years 821*821 of segregation on these teacher expectations, the achievement gap itself is a current and lingering effect of discrimination. I will now summarize the evidence on these connections.

(a) The connection generally between teacher expectations and student performance

Both Dr. William Trent and Deputy Superintendent Janice Jackson testified that teacher expectations affected student performance. Dr. Trent[23] gave extensive testimony concerning the research he performed for the Kansas City School District in connection with the district's efforts to determine whether it had remedied the vestiges of prior segregation in compliance with a court order. In this study, Dr. Trent relied upon "climate surveys" — comprehensive questionnaires distributed to teachers, students and parents — to evaluate and score teacher efficacy. He explained that teacher efficacy measures "the sense that teachers perceived themselves or their colleagues as able to make a difference in their teaching activities, their perceptions that their schools operate with a sense of fairness, that they have a high regard for their students and student body, [and] that they perceive their school as open and accessible...." A higher efficacy score indicated that the teachers considered themselves able to make a difference in their teaching activities and that they held their students in high regard. The Kansas City study showed that there was a correlation between high teacher efficacy scores and higher student test performance, while low teacher efficacy scores were associated with low student test performance.

Deputy Superintendent Jackson also testified that teacher expectations affect student performance. Early in Jackson's career she was specifically trained in TESA (Teacher Expectations and Student Achievement) techniques. TESA's techniques were derived from studies demonstrating that interactions of teachers with students can convey whether the teachers believe that the students are capable of performing at a high level. Specifically, the techniques attempt to focus the teachers on their unconscious biases about low and high achievers, including how their racial attitudes influence their expectations of the students. For many years, Jackson instructed teachers on these techniques and conducted countless training sessions to demonstrate how teachers' interactions with students, even at an unconscious level, can have a significant impact on student performance.

Superintendent Thomas Payzant echoed Trent's and Jackson's statements: "My experience is that if you set high expectations for a student, they will try harder to reach the higher bar than if you set low expectations where they will tend to be content that they have met what they are expected to do and should be satisfied with their accomplishment."

(b) Low expectations for minority students in the Boston School System

Deputy Superintendent Jackson described her numerous opportunities to observe the interactions among students, teachers and administrators in the Boston public schools. In fact, Jackson testified that she observed between thirty and forty Boston public schools in her first year as Deputy Superintendent and at least another forty schools in the system her second year. During these classroom visits, she observed that teachers had different expectations for the African-American and Hispanic children versus the Asian and white children. She saw incidents of unjustified disciplinary action directed at minority students, and noted a frequent failure of teachers to call upon African-American and Hispanic students in class. She also saw recurring instances of teachers withholding praise for minorities and treating them 822*822 with condescending laxity when calling upon them in class.

Noting the substantial disparity in the application rates of African-American and Hispanic students compared with white and Asian students to Boston Latin School, Dr. Trent connected these low expectations for minority students described by Jackson to the minority students' low test performance and low admissions rate at the examination schools. Dr. Trent explained that "[t]o the extent teachers play a central role in encouraging and preparing students to apply for one of the more valued resources in the Boston Public Schools, [the difference in application rates] reflects a very different rate of success with encouraging or facilitating black and Hispanic students applying, in contrast to the rate at which white and Asian students do." He elaborated on this point:

Based, again, on the commentary reported and recorded by Deputy Jackson in her deposition, and my interviews, and on the pattern of applicant disparity, as well as looking at the data in terms of the generation of applicants, and there's chartered information that shows the same schools, for example, that generate substantial numbers of white and Asian applicants are not as successful with generating comparable numbers or rates of applicants for black and Hispanic students.

These suggest patterns of differential treatment or encouragement, or success, at least, with those black students, some of which is likely attributable to differential expectations, particularly within the commentary of the headmasters and the principals of the schools I had an opportunity to visit.

(c) Low expectations for minority students as a vestige of prior discrimination

Dr. Trent provided the primary evidence on the causal link between prior discrimination and low teacher expectations for minority students and the effect of these low expectations on the achievement gap. Dr. Trent tied these low expectations to the lingering effects of a segregated school system. He explained: "It appears that in the seniority ranks, as many as 28.4 percent of the teachers currently employed in the district have been with the district since prior to 1973, and about 47 percent have been with the district from 1980 and prior." These numbers indicated to Dr. Trent that many of the teachers' attitudes towards their students and expectations of them were shaped in a segregated school system. Dr. Trent explained the difficulty of changing such teacher attitudes following a desegregation order:

... [A]n organizational theory in research suggests that organizational change and organizational stability, particularly the climate of attitudes and dispositions within organizations, is a critical feature, and that generally recruitment hinges on socializing new members into the existing culture of the organization. So to the extent that there is this kind of faculty seniority and tenure in the district, we could be seeing the persistence of particular attitudes that predate or at least were shaped and developed during the period of intense contestations regarding the desegregation of the schools [in Boston]. And to the extent that there is little change, effort, professional development, and other sorts of efforts that would address those prevailing attitudes, you could have a persistence of those attitudes and the socialization of new individuals into those existing cultures.

School desegregation research has shown the importance of changing the composition of the makeup of the schools, the racial composition of faculty, and the importance of changing leadership, as well as providing professional development experiences in order to facilitate greater success with the desegregation effort.

Based on his interviews with school personnel, Dr. Trent testified that there was no evidence that the school system had been successful in counteracting the diminished expectations for minority students.

As noted by the majority, Dr. Trent did not conduct a specific study of the Boston school system, as he had done for the Kansas City School District, to evaluate whether teachers had different expectations for minority students and whether those low expectations were attitudinal remnants of the segregation 823*823 era. In lieu of such a study, Trent relied on the seniority statistics discussed above as well as statistics relating to the different student application rates to the exam schools, indicating a substantial disparity between the application rates of African-Americans and Hispanics compared to whites and Asians. He also relied on statistics that show that African-American and Hispanic students attending the Boston public schools receive invitations to attend Boston Latin at a much lower rate than do African-Americans and Hispanics attending private schools, whereas the invitation rates for white and Asian private school and white and Asian public school students were comparable. Finally, he relied on interviews with headmasters, principals and other personnel, as well as an interview with Deputy Superintendent Janice Jackson, who, as noted, had special training in observing teacher interactions with students, and had made a personal survey of over seventy schools in the Boston school system over the course of two years. Based on this information, he offered his expert opinion that low teacher expectations for minorities, shaped in the segregation era, are a cause of the current achievement gap, and thus the achievement gap itself is a vestige of discrimination.

Superintendent Payzant also testified to vestiges of prior discrimination in the school system. He first explained that, prior to 1973, there were

specific practices in the school district that resulted in students being assigned to schools that often led to particular races being represented in some schools and not others, practices that resulted in resources being unequally distributed so that there were disparities among the schools, disparities among the programs that were offered and the access and opportunity that students had to them....

....

And I cite that because I think the institutional history of the school district prior to 1973 was shaped by those practices. And as I testified a moment ago, there is a significant percentage of teachers and other educators in the Boston Public Schools who really were socialized and shaped by the expectation that the institution had in the pre-1973 period.

He further explained the significance of this "socialization" in the pre-1973 period: "[W]hat was going on [pre-1973] was part of what people saw as their normal working conditions and they learned to function in those settings and, whether consciously or unconsciously, accepted the practices that were occurring in many instances." He noted that

there was a disparity in terms of allocation of resources, access to programs, variation in quality from school to school, ... [and] the result was that some students were getting a higher quality education than others and often that was defined by race, and that was part of the background with respect to expectations that were set for what students could or could not do.

Robert Gittens, Vice-Chair of the Boston School Committee, also testified that the effects of segregation and prior discrimination permeated the school system despite its efforts to remedy the problem. He explained that he was often in contact with African-American parents who complained of being alienated from the school system. They also complained that their children were not being motivated or challenged academically. He noted, however, that he did not hear these complaints from white parents. He further stated: "We have schools today that are overwhelmingly black and Hispanic[.] I believe that there is a pervasive sense that there are kids in the system who cannot, will not ... succeed." He elaborated:

I think that one of the things that has happened is we have a teaching force that has been in the system for a very long period of time. And I think that there are large numbers of teachers who, and I've heard this from headmasters and principals and others, that there are teachers who say, "I've been teaching this way for 20 or 25 years. Why should I change the way I teach?"

3. A prima facie case of causation

The majority's criticisms of the Committee's evidentiary case are twofold: 1) the Committee did not establish the presence of 824*824 differential teacher exceptions for minority students and 2) the Committee failed to establish any causal connection between prior discrimination and the achievement gap itself; that is, it failed to establish that prior discrimination affected low teacher expectations for minority students and further failed to establish that these low expectations were a cause of the achievement gap. I respond to these criticisms in turn.

The majority asserts that Dr. Trent's failure to conduct a survey of the type conducted in Kansas City disabled him from validly establishing that teachers had different expectations for minority students. Specifically, the majority criticizes Dr. Trent's reliance on "anecdotal" evidence about teacher attitudes supplied by school officials rather than the broad survey of teachers in Kansas City. Used pejoratively, the word "anecdotal" describes accounts of isolated instances few in number. Used descriptively, the word describes observational testimony that could embrace many instances of a phenomenon. We should be wary of dismissing as "anecdotal" the extensive observational accounts of experienced school administrators testifying about the prevalence of different teacher expectations in their school systems, particularly when, as in the case of Deputy Superintendent Jackson, the administrator is trained to make such observations.

Jackson had observed a large variety of classrooms before her work in the Boston schools. As a result, her observations of over seventy schools in the Boston school system over a two year period were made within a broader frame of reference than those of the individual teachers, students, and parents responding to the surveys relied on by Dr. Trent in his Kansas City survey. Jackson's extensive training in making such observations would allow her to notice nuanced behavior that survey respondents would not have discerned. While the risk that the personal bias of the observer will infect the results is greater when relying on data filtered through a few observers, as opposed to data reported on paper by a large number of survey respondents, Jackson was trained to make her observations in a professional, objective manner. Observations of wide scope, reported by parties with training in methods of objective observation, do permit generalization about "pervasive," "systematic" problems.

The majority opinion and the concurrence contend that there is a similarity between the evidence relied upon by the School Committee and the evidence found inadequate by the Supreme Court in Croson. I find this analogy unconvincing. In Croson, there was no evidence of local, industry-specific discrimination beyond the statements of one council-person to the effect that he was "familiar with the practices in the construction industry in this area," and that in his experience "the general conduct of ... the industry ... is one in which race discrimination and exclusion on the basis of race is widespread." Id. at 480, 109 S.Ct. 706. Such a vague generalization led the Court to conclude that "none of the evidence produced by the city points to any identified discrimination in the Richmond construction industry." Id. at 506, 109 S.Ct. 706. In this case, the testimony of Trent, Jackson and others on the achievement gap, low teacher expectations for minorities, and the causal connection to prior discrimination in the Boston public schools was based on extensive observations by skilled professionals of the conduct and performance of students and teachers in the Boston schools, as well as statistical evidence specific to those schools.

The majority also asserts that "Dr. Trent thus freely conceded that the data he used [on teacher expectations] was not of the quality necessary to satisfy the methodological rigors required by his discipline." I do not find this concession in Dr. Trent's testimony. In fact, during cross-examination, in response to a question concerning whether it was reasonable for him to rely on Deputy Superintendent Jackson's assessments of teacher expectations in the Boston school system, Dr. Trent responded: "In many instances in the social sciences, qualitative researchers work with information that is reported by the respondents. That is not unusual." Although a survey of teachers and other personnel in the Boston school system would have provided a more substantial basis for Dr. Trent's opinion on the 825*825 existence of low teacher expectations for minorities, the absence of such empirical data does not justify the majority's rejection of his testimony.

I also disagree with the majority that Trent's testimony relied on "evidence from one locality to establish the lingering effects of discrimination in another" in the manner criticized in Croson. Dr. Trent's expertise in identifying patterns of vestigal effects of discrimination necessarily was acquired through his prior studies of other school systems, including Kansas City. He did not attempt to use national statistics or statistics from other localities to infer the existence of similar local conditions, as was done by the city council in Croson. See Croson, 488 U.S. at 504, 109 S.Ct. 706.

The Committee's evidence on the connection between prior discrimination and low teacher expectations, and low teacher expectations and the achievement gap, was undeniably a mix of statistical and anecdotal evidence. Although the majority finds this mix unacceptable, courts have often held that statistical evidence documenting a disparate impact or a pattern or practice of disparate treatment can combine with anecdotal evidence of acts of discrimination to establish a prima facie case of discrimination.

In International Brotherhood of Teamsters v. United States, 431 U.S. 324, 334-38, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977), statistical evidence on outcome (namely, the low number of minority truck drivers compared to the relevant qualified pool) was supported by individual testimony on prior discrimination (over forty acts of explicit discrimination against minority aspirants). These testimonials "brought the cold numbers convincingly to life." Id. at 339, 97 S.Ct. 1843. In Coral Construction Co. v. King County, 941 F.2d 910, 919 (9th Cir.1991), cert. denied, 502 U.S. 1033, 112 S.Ct. 875, 116 L.Ed.2d 780 (1992), the Ninth Circuit warned of the dangers of relying overly on either statistics or anecdotal accounts, individually, to establish a widespread pattern of discrimination. Nonetheless, the court found that "the combination of convincing anecdotal and statistical evidence is potent." 941 F.2d at 919. In EEOC v. O & G Spring and Wire Forms Specialty Co., 38 F.3d 872, 878 (7th Cir.1994), cert. denied, 513 U.S. 1198, 115 S.Ct. 1270, 131 L.Ed.2d 148 (1995), statistical evidence established only that the employer in question had no African-American employees; the anecdotal evidence, limited to a sampling of four disappointed applicants, was relevant to demonstrating discriminatory conduct. The plaintiffs successfully established a prima facie case of a pattern or practice of disparate treatment. In Ensley Branch, NAACP v. Seibels, 31 F.3d 1548, 1565 (11th Cir.1994), the Eleventh Circuit stated that anecdotal evidence could be "used to document discrimination, especially if buttressed by relevant statistical evidence."

In the instant case, experienced school administrators and officials, such as Superintendent Payzant and Vice-Chair Gittens, offered their judgment, on the basis of extensive day to day experience with the Boston school system, that there were links between prior discrimination in the system, low teacher expectations, and the achievement gap. Dr. Trent offered his expert opinion on these links without challenge to the admissibility of his testimony. He testified that his conclusions were based on a reasonable methodology in his profession. He evaluated statistics documenting student performance and teacher histories in the Boston school system. He studied the observations of a well-trained administrator in the Boston school system describing teacher performances in the classroom, the impact of these performances on students, and faulty attempts to alter teacher attitudes. He knew the history of segregation in the Boston school system. Seeing statistics and patterns in Boston that he had observed in other school systems where he had found a link between student achievement gaps and prior discrimination, he testified to the probability of such a link in the Boston school system. He had an adequate basis for making that judgment.

The majority also finds fault with Dr. Trent's testimony, and the Committee's case generally, because of the failure to control for "competing explanations for current realities." In the majority's view, the Committee's evidence had to account for other variables 826*826 that might explain all or part of the achievement gap in terms of societal discrimination. This insistence reflects a misconception of the School Committee's evidentiary burden, and would elevate the Committee's evidentiary burden far beyond the prima facie standard contemplated in the Title VII cases. The Committee did not have to present proof that would permit the court to make an independent finding of causation. The Committee had to satisfy the court that the Committee had before it a strong basis in evidence for its judgment that the achievement gap is linked to past discrimination in the Boston school system. As indicated by the discussion of the Supreme Court and First Circuit precedents in Part II, that strong basis is provided by a prima facie case of causation. In voluntarily undertaking remedial measures, the School Committee did not have to establish an airtight case for its own liability to minority students. By setting the bar of proof for the School Committee unrealistically high, the majority has ignored precedents that impose only a prima facie burden on the School Committee.

Contrary to the majority, I believe that Boston Police Superior Officers Federation v. City of Boston, 147 F.3d 13 (1st Cir.1998), where we upheld an affirmative action program, supports the conclusion that the School Committee satisfied its evidentiary burden to justify the remedial measure. In Boston Police, we first relied upon "the BPD's history of racial discrimination [that] is well-documented in the decisions of this court." Id. at 20. We described the twenty years of litigation attempting to remedy the department's discrimination against African-Americans. We relied upon court findings made seven years prior to the litigation at issue and specifically noted that "we do not think this evidence is too temporally remote to justify the conclusion that the BPD's past racial discrimination has manifest effects in the present status of black officers." Id. We considered the statistical evidence that demonstrated a continuing disparity between African-American and white police officers being promoted to lieutenant positions. We then concluded:

This tortuous history, combined with the persistent effects of discriminatory practices at the entry and sergeant levels, sufficiently links the BPD's past discrimination and the statistical disparities contemporaneous with Ruiz's promotion. Given the BPD's halting and, at times, quite modest progress in remedying its earlier discrimination, we are reluctant to infer that the vestiges of that discrimination had substantially disappeared when the BPD promoted Ruiz. The evidence warranted the district court's conclusion that the 1996 "statistical disparity combines with judicial findings of past entry-level discrimination by the BPD to imply convincingly that historical discrimination has affected the promotion of minority sergeants to the rank of lieutenant, and that the lingering effects of that discrimination were present in 1995 when the BPD promoted Ruiz."

Id. at 22-23 (citations omitted). Implicit in our decision in Boston Police was the common sense proposition that a long history of discrimination in a social institution affects for a considerable period of time the attitudes and behavior of people who have worked in that system. Although such history, standing alone, cannot provide the strong basis in evidence for a public entity's adoption of an affirmative action program, it can contribute to the evidentiary basis underlying the judgment that there is a causal relationship between a current outcome, such as an achievement gap, and past discrimination. The Committee appropriately invokes that history to support, in part, its admission program for Boston Latin.

I am concerned that the majority's evidentiary requirements in this case will force school systems contemplating affirmative action programs designed to address the effects of past discrimination to establish, through the collection of quantifiable social science data, that past discrimination is the sole or primary cause of variable achievement. Given the Supreme Court's emphasis on the prima facie justification for such an initiative, I conclude that "substantial factor" causation meets the causal burden of production 827*827 for a prima facie case.[24] In his dissent[25] in Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989), Justice Stevens noted that under existing law the notion of a prima facie case under Title VII contemplates a similar standard. As he put it, the asserted causal link between the acts of an employer and the harm to a Title VII plaintiff "must have substance" but "need not constitute the sole or primary cause of the harm." Id. at 672-73, 109 S.Ct. 2115 (Stevens, J., dissenting). This standard reflects the general tort standard embodied in the Restatement; See Restatement (Second) of Torts §§ 430-433 (1965). As Justice Stevens also noted, it is consistent with the views of a majority of Justices in Price Waterhouse v. Hopkins, 490 U.S. 228, 109 S.Ct. 1775, 104 L.Ed.2d 268 (1989).[26] It is also consistent with Congress' response to that decision in the 1991 Civil Rights Act.[27]

The evidence presented by the School Committee established a prima facie case that differential teacher expectations, grounded in the long history of segregation in the Boston school system, were a substantial causal factor in the undeniable achievement gap found in the Boston school system. In the face of this evidence, Wessmann had the burden of challenging the Committee's prima facie case by disproving the alleged causal linkages between prior discrimination, teacher expectations, and the achievement gap.

4. Plaintiff's evidence challenging the causal link

Wessmann conceded the existence of the achievement gap but attempted to rebut the causal explanation advanced by the Committee by asserting a "neutral explanation" for the disparities (that is, that they were due to socioeconomic conditions attributable, at most, to societal discrimination). Wessmann's only witness on these matters, Professor Stephan Thernstrom, hypothesized that three variables — poverty rates, levels of education, and family structures — might be the cause of the achievement gap in the Boston system. He based his hypothesis on national statistical evidence that demonstrates powerful relationships between low achievement and poverty rates, levels of education and family structure. Professor Thernstrom suggested that, if a controlled study had been done to isolate the causes of the achievement gap, the study would have revealed that any causal connection between the achievement gap and teacher expectations was insignificant. He further rejected the Committee's argument that low teacher expectations are connected to prior discrimination on the basis that the Committee did not conduct a study on this issue.

During cross-examination, Professor Thernstrom acknowledged that he had no data particular to Boston on any alternative causes of the achievement gap, thus confirming his reliance on the national statistics criticized by the Supreme Court in Croson. He also acknowledged that, in his recent book evaluating the causes of a national achievement gap between students,[28] he had written that neither poverty rates, levels of education, nor family structure could account for the national achievement gap between African-American and white students. In fact, in a chapter titled "Low Expectations, 828*828 Low Performance," he stated that "ask little of children in the way of academic achievement and little is what you tend to get." In that chapter, he specifically cited actions of the Boston School Committee in 1990, and stated that the Boston public schools "were failing to do their job and most of all failing African-American pupils."

As the district court noted in its opinion, the one expert witness to testify on Wessmann's behalf conceded that, ultimately, alternative theories of causation could not fully explain the achievement gap between white and African-American students. Hearing this evidence, the district court rejected Thernstrom's direct testimony and accepted only his cross-examination testimony, along with other evidence presented by the Committee, in reaching its conclusion that the Boston Latin admissions program "appropriately addressed the vestiges of discrimination that linger in the Boston Public School system." 996 F.Supp. at 131.

By asserting that the district court erred in crediting the extensive observational testimony of experienced, well-trained school administrators, and by requiring quantifiable data to establish a causal link between past discrimination and present outcomes, the majority would reduce strict scrutiny to a standard that is indeed "fatal in fact." See Adarand, 515 U.S. 200, 237, 115 S.Ct. 2097, 132 L.Ed.2d 158 (1995). In my view, the district court properly concluded that the School Committee had a strong basis in evidence for adoption of the Boston Latin admissions program, thereby meeting its evidentiary burden, and that the plaintiff failed to carry her burden of persuading the court that this affirmative action program was unconstitutional.

IV.

Narrow Tailoring

To survive strict scrutiny, the School Committee's admissions program must serve a compelling interest in remedying past discrimination and must also be narrowly tailored to serve that goal. "When race-based action is necessary to further a compelling interest, such action is within constitutional constraints if it satisfies the `narrow tailoring' test this Court has set out in previous cases." Adarand, 515 U.S. at 237, 115 S.Ct. 2097. United States v. Paradise, 480 U.S. 149, 107 S.Ct. 1053, 94 L.Ed.2d 203 (1987), is the leading Supreme Court case on the meaning of narrow tailoring. It emphasizes the following factors: "the necessity for the relief and the efficacy of alternative remedies; the flexibility and duration of the relief, including the availability of waiver provisions; the relationship of the numerical goals to the relevant labor market; and the impact of the relief on the rights of third parties."[29] Id. at 171, 107 S.Ct. 1053. Fundamentally, narrow tailoring analysis asks whether a program is "overinclusive" or "underinclusive" to serve the purposes of the specific compelling interest on which the program is based. See generally id. at 190 n. 1, 107 S.Ct. 1053 (Stevens, J., concurring); Church of the Lukumi Babalu Aye, Inc. v. City of Hialeah, 508 U.S. 520, 546, 113 S.Ct. 2217, 124 L.Ed.2d 472 (1993).

The following language from Croson offers some insight into the rationale underlying these factors:

[The narrow tailoring prong of strict scrutiny analysis] ensures that the means chosen "fit" [the] compelling goal so closely that there is little or no possibility that the motive for the [racial] classification was illegitimate racial prejudice or stereotype.

Classifications based on race carry a danger of stigmatic harm. Unless they are strictly reserved for remedial settings, they may in fact promote notions of racial inferiority and lead to a politics of racial hostility.

Croson, 488 U.S. at 493, 109 S.Ct. 706. On this account, the "narrow tailoring" and "compelling interest" requirements serve the same purpose: ensuring that the program in question is not simply motivated by a desire to favor one race. "Narrow tailoring" serves the additional purpose of minimizing inadvertent harms that may result from an affirmative action program, on the basis of a "costbenefit" 829*829 analysis.[30] Given these narrow tailoring considerations, I am satisfied that the School Committee has produced a race-conscious admissions program for Boston Latin which is narrowly tailored to address a compelling remedial goal.

1. Flexibility and Duration of the Relief

The admissions program's racial/ethnic composition standards automatically vary each year with the composition of the Qualified Applicant Pool. The record indicates that there has been some case-by-case flexibility in the past: a well-qualified Native American admittee who would not have been admitted under a mechanical application of the program's guidelines (there was too low a percentage of Native Americans in the qualified pool that year to have a spot allocated to that racial/ethnic group) was nonetheless admitted after a glance at the straight compositescore ranking indicated this student had placed above the lowest-ranked admittee.

The limited duration requirement has generally been met by "built-in mechanisms" to shrink the scope and limit the duration of remedial programs. See Boston Police, 147 F.3d at 23, 24-25 (quoting Mackin v. City of Boston, 969 F.2d 1273, 1278 (1st Cir.1992)). A race-conscious program's "temporary" nature ensures that it "will not last longer than the discriminatory effects it is designed to eliminate." Fullilove v. Klutznick, 448 U.S. 448, 513, 100 S.Ct. 2758, 65 L.Ed.2d 902 (1980) (Powell, J., concurring). The policy at issue here is explicitly subject to review. The superintendent must present findings and recommended modifications to the Committee every three years. The district court found that the review provision "ensures that the Policy will not outlive the examination schools' compelling need for it." 996 F.Supp. at 132. While it is not "self-terminating" in an involuntary, mechanical sense, nothing about this race preference implies permanence in the sense of the Supreme Court's general warning against "remedies that are ageless in their reach into the past, and timeless in their ability to affect the future." Wygant, 476 U.S. at 276, 106 S.Ct. 1842 (plurality); see also Croson, 488 U.S. at 498, 109 S.Ct. 706.

2. Relationship of the Numerical Goals to the Composition of the Qualified Pool

This element of narrow tailoring assures that the beneficiaries of any program will be qualified, thereby minimizing the cost of the preference to society while assuring that the favored applicants are not unjustly enriched. Cases approving preferences in the promotions of police officers cite the "relationship of the [program's] numerical goals to the relevant labor market," Paradise, 480 U.S. at 171, 107 S.Ct. 1053, or the fact that the preference is directed accurately towards "the smaller group of individuals who possess the necessary qualifications...." Boston Police, 147 F.3d at 21 (quoting Croson, 488 U.S. at 501, 109 S.Ct. 706). This "qualification" factor does not make the transition from the employment to the educational context gracefully: education, after all, is directed at shaping individuals in a prospective manner. Students ranking in the lower half of the applicant pool by composite score had succeeded at Boston Latin in the past (Headmaster Contompassis recalled one in the sixty-fourth percentile from the top who graduated near the top of his class). Nonetheless, the Committee's admissions program has clearly been structured to meet this qualification requirement. No one is admitted, on account of the "flexible racial/ethnic guidelines," from outside the qualified applicant pool (or even from anywhere near the median score point).[31] In fact, for the 1997-1998 school year, all students admitted were within the top 10.6% (by composite score rank) of the applicant pool.

830*830 3. Impact of the Relief on the Rights of Third Parties

In Paradise, the Court asked if the promotion preference given in that case was an "absolute bar" to advancement. See Paradise, 480 U.S. at 171, 182-83, 107 S.Ct. 1053 (quoting Local 28 of Sheet Metal Workers v. EEOC, 478 U.S. 421, 481, 106 S.Ct. 3019, 92 L.Ed.2d 344 (1986)). Without minimizing the disappointment Sarah Wessmann feels at her exclusion from Boston Latin, the answer to the analogous inquiry in this case is clearly "no." Forty-five spots were open to Wessmann on the basis of composite score alone, and several times as many were available when Wessmann attempted to enter at the seventh grade level (even accounting for the 35% set-aside which existed during that year).

The numerous employment promotion-preference cases also ask whether the preference imposes a layoff-like burden, i.e., whether it defeats legitimate entrenched expectations of the disfavored. Layoffs raise more severe narrow tailoring concerns than a hiring preference:

In cases involving valid hiring goals, the burden to be borne by innocent individuals is diffused to a considerable extent among society generally. Though hiring goals may burden some innocent individuals, they simply do not impose the same kind of injury that layoffs impose. Denial of a future employment opportunity is not as intrusive as loss of an existing job.

Wygant, 476 U.S. at 282-83, 106 S.Ct. 1842. Rather than being removed from a superior school because of a racial preference, Wessmann was denied the opportunity to move from a good school to a better school. There is no constitutional right to attend a school of one's choice. See Johnson v. Bd. of Educ., 604 F.2d 504, 515 (7th Cir.1979); United States v. Perry Cty. Bd. of Educ., 567 F.2d 277, 279 (5th Cir.1978). It is true that Wessmann, unlike the non-promoted police officers in Stuart or Boston Police, will not have another chance to enter Boston Latin. However, her situation resembles denial of a promotion with retention of current job status: she remains at the Latin Academy which, while not the equal of the Latin School in terms of overall student performance, is nonetheless "challenging" Wessmann, according to the affidavit of her Headmaster, the trial testimony of her father, and other evidence in the record.

4. Necessity of Relief to Achieve the Compelling Interest, and the Efficacy of Alternative Remedies.

(a) Necessity of relief to achieve the compelling interest

"To evaluate [a determination of necessity] we must examine the purposes [a preference program] was intended to serve." Paradise, 480 U.S. at 171, 107 S.Ct. 1053. Here, that goal is to overcome expeditiously the effects of past discrimination in the Boston schools. Urgency of relief is a legitimate consideration in finding necessity. See Paradise, 480 U.S. at 173-77, 107 S.Ct. 1053. If admissions had been based on straight rank order for the 1997-1998 school year, the percentage of those invited to the seventh grade who were African-American would have dropped from 13% to 6%; for Hispanics the drop would have been from 5% to 3%. For the ninth grade, the dropoff would have been 21% to 14% for African-Americans and 10% to 7% for Hispanics. Given the low levels of African-American and Hispanic students who would have been admitted to Boston Latin on a straight-from-the-top admissions methodology (in comparison to either the composition of the general population, of the entire public school system, or of the qualified applicant pool), a preferential program was required to remedy the lingering effects of de jure segregation in the system.

Given that one of the standards for evaluating narrow tailoring is over- or under-inclusiveness to the purpose, one might ask whether this program is "under inclusive" to address the remedial needs facing the School Committee. The ultimate impact in numbers is small — eleven of ninety invitees to the ninth grade, and thirty-seven of 440 to the seventh, were admitted who would not have been if straight rank order admissions had been used. Overall, the preference affects outcome for about 9% of all admittees. The resulting representation of African-Americans and Hispanics in the student body of 831*831 Boston Latin does not even approach proportionality with the composition of the Boston public school system as a whole, which is 74% African-American and Hispanic.

Wessmann asserts that underinclusivity is evidence that the program is only intended to address an interest in diversity. Where affirmative action deals with selective programs like promotions or competitive schools, however, the impact of the preference may have to be small because of the narrow tailoring requirement that beneficiaries be qualified. Any preference can only respond to remedial needs in proportion to the success of minorities in raising themselves into the qualified pool. Thus, the preference in Stuart, for example, gave "only limited advantage" to minority applicants, increasing their numbers "gradually over time." 951 F.2d at 454. Under the circumstances, there is no grounds for holding that the program here is not narrowly tailored towards the remedial goal because it pursues that goal deliberately.

(b) Efficacy of alternative remedies

The Committee adopted race neutral alternatives to work towards a race-blind policy in the future. These alternatives include more extensive schooling (extra classes during the school year and summer school), enhanced school programs (the advanced work classes), and admissions test preparatory courses in the public schools feeding into the exam system. However, Professor Edwin Melendez, a School Committee member, described the alternatives as "not a panacea to deal with the issue of unequal access overnight," but rather as "definitely a long-term strategy." The alternatives did not address adequately the problem of urgency.

The School Committee also considered alternative admissions plans detailed by an outside consulting firm. Consideration of these proposals was delegated to a special ad hoc Task Force, composed of a broad array of concerned individuals and experts including two members of the School Committee, Professor Charles Ogletree of Harvard Law School, several individuals with experience in formulating affirmative action programs, and numerous public figures with personal connections to Boston Latin from various ethnic communities within the city. The Task Force held eight committee meetings and five neighborhood hearings, all open to the public. The Task Force members discussed and evaluated the consultant's initial proposals and then worked closely with the consultants until a plan emerged that was satisfactory to the remedial needs at issue here. This plan was presented by the Task Force chairs to the School Committee, which also heard the views of the dissenting members of the Task Force. Further public hearings were held to consider the final Task Force report before it was adopted by the Committee. There is no indication in the record that any less race conscious program was ever proposed to the Committee (before or during this litigation) which could have effected immediate remedial relief while simultaneously maximizing the quality of the student body.[32]

Relying on Podberesky v. Kirwan, 38 F.3d 147, 160-61 (4th Cir.1994), Wessmann argues that the Committee must show that a race-neutral policy was first tried, not just considered, and that it failed to accomplish the goal in question. In that case, however, Podberesky himself submitted the proposed race neutral plan which piqued the interest of the Fourth Circuit. His plan had obviously not been considered by the university. In this case, several race neutral plans presented by the consulting firm to the Committee were rejected after consideration as inadequate to the purpose of immediate remediation. Moreover, there is no Supreme Court authority for the proposition that implementation of a racially-preferential plan must be preceded by a failed race-neutral attempt to accomplish the same goals.

The Majority's Narrow Tailoring Critique

The majority cites three narrow tailoring deficiencies in the Committee's admissions program: (1) African-American and Hispanic applicants from private schools were by definition not hurt by the inadequacies of the public schools, and thus should not be allowed to benefit from any remedial program; 832*832 (2) basing admissions on the composition of the remaining qualified applicant pool will not necessarily benefit African-Americans and Hispanics, despite the fact that the program is predicated on remedying past discrimination against those groups; furthermore, the program benefits racial groups not victimized by discrimination in the Boston schools; and (3) the program does not directly address the problem of lower teacher expectations. The subtext of these arguments is a suggestion that the program is justifiable only on racial diversity grounds, and that it is tailored only towards a form of racial balancing.

1. The inclusion of minority applicants from private schools

Our Constitution has always protected the right of parents to choose a school by moving to another district, by opting into a suburban busing program like METCO,[33] or by sending their children to private school. It therefore seems incongruous to punish the initiative of those who removed their children from an educational environment tainted by vestiges of discrimination. Moreover, there is no reason to assume that leaving the public school system was not in itself a burden on these families. If there were hardships involved for families that fled the public schools (tuition, transportation, adjusting in general to the transition), those hardships could be fairly linked to the past discrimination that made the public schools an unacceptable environment for their children in the first place.

2. Potential disfavoring of groups subject to past discrimination; separate categorization of whites, Asians and Native Americans

The majority notes that Hispanics, "archetypical victims of discrimination," could potentially be disfavored by the program under a hypothetical[34] scenario whereby their representation in the remaining qualified applicant pool was concentrated near the top of the composite score rankings. That potential for Hispanics (or, for that matter, African-Americans) to be disfavored at Boston Latin suggests some limitations in the Committee's remedial program. Those limitations do not discredit the program. If the facts in one particular year were so anomalous that the program disfavored either African-Americans or Hispanics, the case-by-case flexibility documented above could be invoked to remedy the anomaly.

The majority raises a troubling point when it questions the School Committee's separate categorization of Asians under the admissions program's racial/ethnic guidelines.[35] Perhaps it would have been preferable to group Asians together with whites and any other identifiable groups against whom there was no asserted history of discrimination. However, in the context of the long history of de jure discrimination against African-Americans and Hispanics in the Boston school system, this separate categorization of Asians does not undermine the fundamental remedial purpose of the admissions program. Moreover, the record suggests the possibility of a remedial interest with regard to Asians. There is evidence of a significant language-skills achievement gap for Asian students in the public schools. If there is future litigation about this aspect of the program, more achievement gap evidence relating to Asians might be produced.

3. Not directly addressing lower teacher expectations

The majority opinion and the concurrence both find a narrow tailoring flaw because the 833*833 School Committee's admissions program does not directly address the problem of lower teacher expectations for African-American and Hispanic students, identified by the School Committee as a vestige of discrimination and a substantial causal factor of the achievement gap. The admissions program is designed to remedy the impact of lower teacher expectations, and not the expectations themselves — and is perfectly acceptable as such. Although the School Committee has a responsibility to eliminate the vestiges of discrimination from the system over time, the Committee also has a responsibility to address the harm those vestiges currently impose on African-American and Hispanic students.

If the School Committee was not simultaneously addressing the underlying teacher expectations problem, there might be a narrow tailoring concern related to the duration of the remedial program. The School Committee is, in fact, addressing this underlying problem. It has been making substantial efforts towards instituting a standardized curriculum in the primary schools. A standard curriculum sets forth what is expected from every student, independently of what any individual teacher might otherwise expect from any individual student. Under the Focus on Children reform initiative, adopted in August 1996, citywide curricular standards are to be implemented within five years. The School Committee is also making more direct efforts to address the problem of teacher bias. For several years it has retained the services of the Efficacy Institute, an educational firm which had previously undertaken large-scale teacher reeducation in the New York City public schools.

V.

Conclusion

The majority characterizes my dissent as "wishful thinking" about the meaning of Wygant, Croson and other Supreme Court precedents in this difficult area of the law. Not surprisingly, I disagree. I believe that I am faithful to those precedents and, unlike the majority, apply them accurately to the evidence presented to the district court.

The majority goes awry because it reads Wygant's requirement of a "strong basis in evidence"[36] for an affirmative action program and Croson's reference to a "searching judicial inquiry"[37] into the justification for an affirmative action program as demands for evidence grounded in quantifiable social science data rather than human judgments. There is no such demand in Wygant, Croson or any other Supreme Court precedent. Numbers are not the only source of the requisite degree of certainty about low teacher expectations for minorities and causation. In this case, the extensive observations of experienced administrators in the Boston public schools, supplemented by the testimony of a highly qualified expert who recognized in the Boston public schools a phenomenon he had studied extensively elsewhere, were as probative as the statistical surveys and regression analyses demanded by the majority.

The majority also goes awry because it uses Croson's reference to a "searching judicial inquiry" as the basis for disregarding two critical points made by Justice O'Connor in Wygant:

(1) The strong basis in evidence required of a public entity defending an affirmative action program in court is provided by evidence sufficient to support a prima facie case of discrimination against the favored minority.

(2) "In `reverse discrimination' suits, as in any other suit, it is the plaintiffs who must 834*834 bear the burden of demonstrating that their rights have been violated."[38]

We applied this evidentiary framework in Stuart,[39] thereby signaling to the district court its applicability in this affirmative action case. The district court followed that teaching, and we should as well.

For all of the reasons stated herein, I would affirm the judgment of the district court.

[1] To the extent that the School Committee suggests that burdening different groups equally makes classifications based on race constitutionally acceptable, we need respond only with a citation to Loving v. Virginia, 388 U.S. 1, 8, 87 S.Ct. 1817, 18 L.Ed.2d 1010 (1967).

[2] Wittmer exemplifies the caution that this area of constitutional jurisprudence demands. In that case, the court accepted the preferential hiring of black boot camp officers as a compelling interest based on the notion that such officers were "needed because black inmates are believed unlikely to play the correctional game of brutal drill sergeant and brutalized recruit unless there are some blacks in authority in the camp." Wittmer, 87 F.3d at 920. The Wittmer court's conclusions rested on uncontroverted expert social science evidence regarding prison administration, and the court carefully limited its holding to the record before it, stating that if future academic research should show that its factual assumptions about boot camp discipline were unwarranted, race-conscious action would be impermissible. See id. at 920-21.

[3] Drawing on Supreme Court precedent that notes the vital function of public education in our society, see, e.g., Ambach v. Norwick, 441 U.S. 68, 76-77, 99 S.Ct. 1589, 60 L.Ed.2d 49 (1979); Brown v. Board of Educ., 347 U.S. 483, 493, 74 S.Ct. 686, 98 L.Ed. 873 (1954), the School Committee suggests that BLS's status as a secondary school, as opposed to a university, materially alters the decisional calculus and warrants judicial deference to school officials' determinations anent the racial and ethnic composition of the student body. Once again, the School Committee's suggestion rests on out-of-context dicta and is entirely unpersuasive. More important, the School Committee's citation to Brown is self-defeating, for the Brown Court made it abundantly clear that constitutional principles cannot take a back seat to the discretion of local school officials in respect to matters such as the racial composition of student bodies.

[4] Over the past decade, the relevant proportions have been more or less as follows: 48% black, 25% Hispanic, 8% Asian, and 17% white.

[5] To be sure, the School Committee attempts to vindicate its focus on groups by enumerating the administrative burdens that would accompany an individualized admissions process. But administrative convenience is not a sufficient justification for promoting racial distinctions. See Croson, 488 U.S. at 508, 109 S.Ct. 706.

[6] Were this decree intended to function as a mandatory injunction, it likely would be unenforceable for failure sufficiently to describe what is expected of the party enjoined. See Fed. R.Civ.P. 65(d); see also Board of Educ. v. Dowell, 498 U.S. 237, 246, 111 S.Ct. 630, 112 L.Ed.2d 715 (1991); David B. v. McDonald, 116 F.3d 1146, 1148 (7th Cir.1997), cert. denied, ___ U.S. ___, 118 S.Ct. 692, 139 L.Ed.2d 638 (1998).

[7] The situation at BLS illustrates why no such knee-jerk inference can be drawn. In 1987, we ruled that the Boston public schools had achieved unitariness with respect to student assignments. See Nucci, 831 F.2d at 326. Thereafter, though not obligated to do so, BLS continued its policy of setting aside 35% of the seats in its incoming classes for black and Hispanic pupils and jettisoned the set-aside only after a court challenge. Nothing in the record suggests that BLS has in the past ten years been discriminating against minorities. Under such circumstances, a conclusion that a previous finding of discrimination in and of itself establishes that current disparities are due to that discrimination would be little more than an ipse dixit. See Freeman, 503 U.S. at 496, 112 S.Ct. 1430 (warning that "[t]he causal link between current conditions and the prior violation is even more attenuated if the school district has demonstrated its good faith").

[8] Far from being a burden-shifting mechanism, as our dissenting brother suggests, the requirement of considering various salient causal factors is part and parcel of a party's duty to limn a plausible causal relationship between particular independent and dependent variables. See, e.g., Coward v. ADT Security Sys., Inc., 140 F.3d 271, 274 (D.C.Cir.1998) (noting this requirement in the context of analyzing a prima facie showing of wage discrimination). While a litigant need not include every conceivable independent variable in such an analysis, see, e.g., Bazemore v. Friday, 478 U.S. 385, 400, 106 S.Ct. 3000, 92 L.Ed.2d 315 (1986) (opinion of Brennan, J.), the major independent variables must be considered, see Coward, 140 F.3d at 274. This is necessary not only to render the statistical study probative, but also to save it from being dismissed as irrelevant. See Bazemore, 478 U.S. at 400 n. 10, 106 S.Ct. 3000; see also Koger v. Reno, 98 F.3d 631, 637 (D.C.Cir.1996) (stating that "Bazemore [does not] require acceptance of regressions from which clearly major variables have been omitted"). Here, we need not concern ourselves with whether the School Committee's showing satisfies the Bazemore standard, for there is no valid statistical study of causation to begin with.

[9] Although Ms. Jackson was not supposed to know the nature of the project, she did testify that she was not asked to make TEASA-type observations.

[10] We note in passing that the testimony of school officials with respect to the pervasiveness of this phenomenon was far from unequivocal. Although Deputy Superintendent Jackson offered one conclusion, statements by BLS Headmaster Contompasis and Dr. Melendez leave a considerably different impression.

[11] As a prevailing party, the plaintiff is presumptively entitled to fees and expenses under 42 U.S.C. § 1988 (1994). Should the parties not be able to agree on a disposition of this aspect of the matter, we direct the plaintiff to file in the district court, within 30 days following the issuance of our mandate, an application for such fees and expenses (including those incident to proceedings in this court). The defendants shall have 30 days thereafter in which to file an opposition. The district court shall then proceed to hear and determine the application.

[12] Segregation of Hispanics was not described in Morgan I but became a concern of the court by the remedial phase litigation, Morgan v. Kerrigan, 401 F.Supp. 216, 242 (D.Mass.1975).

[13] This aspect of the admissions regime had been the subject of charges filed with the Massachusetts Commission Against Discrimination prior to Morgan I, and had recently been changed for a three year trial period pursuant to a settlement.

[14] These schools included, in addition to the examination schools, certain vocational schools in the system which also drew students from varied geographical locations within the city.

[15] Faculty and staff desegregation, along with the failure to develop a plan to address poor upkeep of facilities, were the primary continuing concerns of the court, not student assignments. See id. at 327-32.

[16] This statement of the court seems to challenge the conclusion that test scores can identify a pool of qualified applicants in a nondiscriminatory way just as strongly as it questions the notion of racial preferences. Compare the same court's skepticism about standardized testing in 1975:

Nor is it clear that the SSAT is the most valid or even a generally valid method of identifying black and white students who can benefit from a Latin school curriculum. It is attractive because it is available, and apparently has value as a predictor of academic success. It is generally accepted, however, that blacks fare worse on this type of examination. There have been suggestions, but little evidence, that the SSAT itself is a culturally biased test.

Morgan II, 401 F.Supp. at 243-44. In fact, admission to the exam schools has never been based solely on test scores: the "composite" scores used to distinguish applicants have always used some combination of standardized test scores with grade point averages.

[17] The influence of the McLaughlin court's suggestions was acknowledged by the parties in this case. In the trial before the district court in this case, there was this exchange: "THE COURT: ... [Judge Garrity] suggested the very plan that we're dealing with now. MR. McLAUGHLIN: He did."

[18] Justice Powell's complete statement follows:

We have never approved a classification that aids persons perceived as members of relatively victimized groups at the expense of other innocent individuals in the absence of judicial, legislative, or administrative findings of constitutional or statutory violations.... After such findings have been made, the governmental interest in preferring members of the injured groups at the expense of others is substantial, since the legal rights of the victims must be vindicated. In such a case, the extent of the injury and the consequent remedy will have been judicially, legislatively, or administratively defined. Also, the remedial action usually remains subject to continuing oversight to assure that it will work the least harm possible to other innocent persons competing for the benefit. Without such findings of constitutional or statutory violations, it cannot be said that the government has any greater interest in helping one individual than in refraining from harming another. Thus, the government has no compelling justification for inflicting such harm.

Bakke, 438 U.S. at 307-09, 98 S.Ct. 2733 (opinion of Powell, J.) (citations and footnote omitted).

[19] In the Stuart case, the complaint leading to the consent decree alleged two loci of discrimination: in addition to past discrimination that lowered the numbers of African-American officers eligible to take the examinations for promotion to sergeant, the exams themselves were not a "fair non-discriminatory device for screening applicants." 951 F.2d at 451. Differential pass-rates for African-Americans and whites seem to have been the primary support for the contention that the exams were discriminatory.

[20] In every instance in which the phrase has been used by the Supreme Court, a "pattern or practice" claim under Title VII refers to a pattern or practice of disparate treatment, rather than disparate impact. See also Maurice E.R. Munroe, The EEOC: Pattern and Practice Imperfect, 13 Yale L. & Pol'y Rev. 219, 227 n. 48 (1995) ("on a strict statutory construction, `pattern or practice' discrimination refers only to disparate treatment discrimination.")

"In disparate treatment cases, the plaintiff is required to prove that the defendant had a discriminatory intent or motive." Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 986, 108 S.Ct. 2777, 101 L.Ed.2d 827 (1988). See also International Bhd. of Teamsters v. United States, 431 U.S. 324, 335 n. 15, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977) ("Proof of discriminatory motive is critical, although it can in some situations be inferred from the mere fact of differences in treatment."); 42 U.S.C. § 2000e-6(a) (defining a "pattern or practice" as "intended to deny" the full exercise of equal employment rights (emphasis added)).

[21] The NAACP's amicus brief states that its motion to intervene was motivated by the belief that "it was contrary to the [School Committee's] interests to present evidence of [the School Committee's] potential liability to African-American and Hispanic students for discriminatory admissions practices."

[22] See Vaughns v. Bd. of Educ. of Prince George's County, 18 F.Supp.2d 569, 578 (D.Md.1998) (collecting cases).

[23] Dr. Trent is a quantitative sociologist who has researched and written extensively on the effects of desegregation on students, while they are in school and after their graduation. He has focused on whether achievement gaps between African-American and white students relate to the lingering effects of prior segregation, and particularly whether there is a connection between low teacher expectations for minority students (as a continuing effect of prior discrimination) and the performance of those students.

[24] See the Restatement (Second) of Torts § 431(a) (1965) for one common statement of the legal standard for substantial factor causation.

[25] The Justices were not debating the appropriateness of substantial factor causation in a prima facie case.

[26] The opinion by a plurality of four Justices stated that if an impermissible consideration (race, gender, etc.) "played a motivating part" in an employment decision, a prima facie case of causation was made sufficient to shift the burden of proof onto the employer, Price Waterhouse, 490 U.S. at 244, 250, 258, 109 S.Ct. 1775; Justice O'Connor's concurrence would have shifted the burden only if the employer gave the impermissible consideration "substantial weight," 490 U.S. at 261-62, 109 S.Ct. 1775.

[27] If an impermissible consideration is shown to have been a "motivating factor" in an employment decision, the causation element of a prima facie case has been produced. See 42 U.S.C. § 2000e-2(m) (West 1998)(codifying § 107 of the 1991 Civil Rights Act).

[28] Stephan Thernstrom and Abigail M. Thernstrom, America in Black and White: One Nation, Indivisible (1997).

[29] Paradise involved a court-ordered remedial preference in police promotions.

[30] See Jed Rubenfeld, Affirmative Action, 107 Yale L.J. 427, 437-443 (1997). Such inadvertent harms include (among others) injuring third parties, unjustly granting a windfall to unqualified beneficiaries, foreclosing consideration of race-neutral alternatives, and furthering notions of dependency.

[31] Regarding the legitimacy of the current plan's central thesis that the top half of the applicant pool is qualified to succeed at Latin School, I note that this thesis has endured a succession of admissions regimes and, conceptually, predates the Morgan litigation. The use of the composite ranking is also a longstanding tradition in the admissions regime for Boston Latin.

[32] Cf. Paradise, 480 U.S. at 177 n. 28, 107 S.Ct. 1053 (indicating presence of no suggested alternatives in record).

[33] The METCO program allows Boston-resident minorities to opt into the suburban public schools, providing transportation for them as well.

[34] Hispanics were disfavored in practice at O'Bryant School. However, admissions to that school are not at issue in this case.

[35] One Asian student whose composite score ranking was below Wessmann's was admitted under the guidelines. In other situations, the admission program's racial/ethnic guidelines have worked to disfavor Asians. For the 1997-1998 school year, the percentage of Asian ninth grade admittees rose from 23% of the total under straight rank ordering to 24% under the program; for the seventh grade, Asian admittees dropped from 24% under straight rank ordering to 23% under the program.

[36] Wygant, 476 U.S. at 277, 106 S.Ct. 1842 (Powell, J., plurality opinion).

[37] 488 U.S. at 493, 109 S.Ct. 706 (plurality). The phrase "searching judicial inquiry" comes from section III-A of Justice O'Connor's Croson opinion, in which she was joined only by the Chief Justice and two other Justices. That section of her opinion addresses the argument that "benign" uses of race (primarily affirmative action plans) should be subject to a lower standard of review than "strict scrutiny," as had been proposed frequently by dissenting members of the Court. The phrase is not offered as a commentary on the quantum of evidence required to justify a "remedial" program. See 488 U.S. at 493-98, 109 S.Ct. 706.

[38] Wygant, 476 U.S. at 292, 106 S.Ct. 1842 (O'Connor, J., concurring).

[39] 951 F.2d at 450-53.