The city of Flint, Michigan, adopted a plan in 1985 requiring that 50% of all police officers who are promoted to the rank of sergeant be members of specified "minority" groups. This action under 42 U.S.C. §§ 1983 and 1981 was brought by white police officers who have been passed over at least once1 for promotion because the plan required that minorities be promoted ahead of them to maintain the designated numerical balance. The district court granted summary judgment to the city, and the plaintiffs timely appealed. We hold that the city has not met its heavy burden of proving that a compelling state interest necessitates its racial quota system, nor has it shown that the quota system is narrowly tailored to serve such an interest. We reverse the district court's decision and remand the matter for further proceedings consistent with this opinion.

* A. The 1973 Holliman v. Price Decision

For more than twenty years, federal courts in the Eastern District of Michigan have considered charges that the hiring and promotion policies and practices of the Flint police department have been tainted by discrimination. In Holliman v. Price, 1973 WL 280 (E.D.Mich. Jan. 3, 1973), minority plaintiffs challenged two elements of the police department's entry-level hiring policies. These requirements were that successful applicants (1) have sixty hours of college credit in the field of police administration or a directly related field, or have equivalent law-enforcement experience; and (2) satisfactorily complete a twenty-nine-item written test. The Holliman court, agreeing with the plaintiffs that the two challenged requirements were invalid, found that a validation study that the city had conducted "was totally insufficient by any professional standards" because it had been based on too small a sampling population. Id.at * 8.

Consequently, "in view of the well documented, historical underemployment of blacks in Flint's Department of Police," id. at * 10, the court granted plaintiffs' request for a preliminary injunction that compelled the city to proceed with plans to hire twenty new patrol officers. In addition, it ordered the city to reconsider previously rejected applicants for entry-level police positions, this time without assessing their qualifications based on the invalidated hiring criteria. Despite the relief granted, the court also stated that "[w]e decline at this time, however, to order any particular quota in [the] hiring of blacks vis-a-vis whites or other non-black minorities." Ibid. (emphasis added).

Four years later, in Alfaro v. Suber, Civ. Action No. 613 (E.D.Mich. May 18, 1977), minority applicants again challenged the police department's hiring policies and practices. The court found that, prior to 1973, "in the entire history of Flint, a total of only twenty-four Blacks had ever been hired" to serve as police officers. Slip op. at 5. On the other hand, 15 of 29 new hires (52%) during the 1972-73 year were minority members; 2 of 8 new hires the following year were minorities; and 56 of 105 (53%) newly hired officers in 1974-75 were minorities. Ibid.

In addition to considering the plaintiffs' suit over the police department's hiring policies and practices, the district court also ruled on their challenge to the department's promotion policies and practices. The court found that, in 1973, 229 applicants had taken the department's written examination for promotion to sergeant; 87% of the white applicants achieved the passing score of 70 or higher, and all 7 of the Black applicants passed. Moreover, the two racial groups' average scores had been virtually identical, though the court noted that "the minority sampling was too small to be predictive." Slip op. at 10.

In response to an earlier lawsuit in state court, there had been a stipulation and agreement between the city and the groups representing minority police officers in which the city had committed itself to promoting one minority member to sergeant for every four "non-minority" members that it promoted. It made its promotions by creating two separate promotion-eligible lists, one for minorities and one for non-minorities, each based on a composite score that aggregated the test results and a maximum of fifteen seniority points for up to fifteen years' service on the force. In addition, the city required that promotion-eligible candidates have either four years' prior experience as a Flint police officer, or three years' experience as an officer and one year of college. In 1975, a new promotion test was administered to 180 applicants, and 138 white officers passed. Of the 4 Black officers who took the test, 3 passed. All 3 of the Blacks who passed were promoted, pursuant to the stipulation.

In reviewing the record before it, the Alfaro court held that there had been a "gross disparity between employment levels of Black and White persons prior to 1973," providing "compelling evidence that the employment practices in the Police department have had a discriminatory impact." Slip op. at 15. However, the court found insufficient evidence to sustain plaintiffs' claims that the department's minority-recruitment efforts were discriminatory. Similarly, "[t]he 1973 and 1975 sergeants examination [sic] were not shown to have a statistically significant disparity." Id. at 17. The court concluded that, although the police department had clearly discriminated in its hiring policies and practices prior to 1973,

[s]ince that time, it is also apparent that the City has attempted to remedy many of the practices which caused the disparate impact and has since increased minority representation in the Police Department. Whether these gains have been the result of court action or pressures from federal funding agencies is not entirely certain.

Because of the general improvement in minority representation in the last three and one-half years, the Court does not believe that imposition of a quota for hiring is necessary. However, to assure that improvement continues, the Court will retain jurisdiction of the case for five years.

Id. at 20 (emphasis added). The court also prevented the police department from considering more than five years of seniority in applications for promotion. Id. at 18.

C. The 1983 Election and the 1984 HRC Hearings

In 1983, the voters of Flint elected a new city government. Mayor James Sharp ran on a platform that included pledges to implement "affirmative action plans" that would increase the city's minority hiring and promotion efforts.

In 1984, public hearings were conducted by the Flint Human Relations Commission ("HRC"). The HRC took testimony from several witnesses. Richard Dicks, a fourteen-year police veteran who had been an unsuccessful plaintiff in Alfaro and who had been named Flint's Deputy Administrator for Public Safety by Mayor Sharp, testified. Flint City Council President Melvin McCree reported the results of the City Council's 1981 and 1983 investigations into police hiring and promotion policies, and he conveyed a Council committee recommendation that the city adopt a plan for minorities to be 65% of new police officers. Furthermore, the city's new police chief, Miles White, who had been appointed by Mayor Sharp, and other political figures provided anecdotal testimony confirming contentions of past discrimination in police hiring and promotions. However, Chief White also stated at the hearing that the police department had undertaken an expensive effort in the 1960s to recruit at Black colleges across the nation; the effort "failed to produce significant numbers of recruits."

Important testimony came from Dr. Marc Bendick, a labor-force analyst employed by the city, who reported that minorities comprised 46.3% of Flint's 1980 population and 45% of Flint's general labor force in 1984. After allowing for demographic, age, and education factors, for the tendency of census-takers to undercount minority members, for the city's tendency to employ a certain proportion of its entry-level officers from neighboring communities outside Flint where minority groups are not as populous, and for "the propensity of minorities to apply for civil service jobs at a somewhat higher rate than for jobs in the private business sector," Dr. Bendick determined, in a 1984 report, that minority persons should have comprised 39.6% of Flint's police officers in 1984; however, only 19.1% of the force was actually comprised of minorities at the time. Dr. M. Bendick, Minority Hiring and Promotion in the Flint Police Dept. 2, 7 (Apr. 1984). He noted that most Flint police officers had been hired in 1978 or earlier. Id. at 4.

Dr. Bendick also opined that 31.1% of Flint's police sergeants in 1984 should have been minorities, although only 10.8% actually were.2 However, in finding that minorities comprised 39.1% of the city's 1968 labor pool, 42.0% of the 1969 pool, and 44.5% of the 1970 pool, Bendick at 6, Dr. Bendick's statistics differed sharply from those of other statisticians. See, e.g., Statistical Abstract of the United States: 1986 16 (106th ed. 1985) (reporting that Blacks comprised 28.1% of Flint's 193,000 population in 1970). Ultimately, Dr. Bendick found that "[t]he appropriate end goal for proportionate minority representation at all ranks in the police department is the 45 percent which minorities constitute of the entry-level labor pool for the department." Bendick at 8.

2. The HRC Report and the Adoption of the "Flint Plan"

After the hearings, the HRC issued a report in August 1984, buttressing its findings of discrimination in the Flint police department by citing the district court decisions in Holliman and Alfaro. The HRC report included its "find[ing] that 45% is a reasonable end-goal for minority representation in each rank of the classified service of the police department," HRC Hearings Report ("HRC") at 5, and its further "find[ing] that implementation of a 1-to-1 promotional ratio until parity is achieved will not unnecessarily trammel the interests of non-minority officers." Ibid. The HRC acknowledged that the Flint police department had undertaken prior "initiatives ... including improvement in validation of exams; modifications of certification procedure; dissemination of information regarding employment opportunities to protected classes; pre-employment inquiries in compliance with state and federal laws; and in-service training and retraining programs for all personnel." HRC at 13. However, the HRC still emerged with a "find[ing] that the achievement of a 45% minority representation within all ranks in the Flint police department would satisfy the operational efficiencies that are associated with having a police department which more nearly reflects the population that it serves." HRC at 5.

A year after the HRC hearings, in July 1985, the Mayor presented to the Michigan Civil Rights Commission ("the Commission") "the City of Flint's Affirmative Action Plan" (hereinafter "the Flint Plan" or "the plan"), based on the HRC's findings. This plan covered the entire city workforce and declared an intent to "[s]earch for selection procedures that minimize adverse impact on minority [sic], females and handicappers." Plan at 21. The Commission approved the Flint Plan in January 1986, and the collective bargaining agreement between the city and the Flint Police Officers Association was amended to incorporate the plan. Promotions of entry-level police officers to the next rank of sergeant were to be made according to a 1:1 ratio, alternating between minority and non-minority candidates.

As of the early 1990s, the city's minority population is approximately 52%, and Blacks comprise 93% of all minorities in the city. Thus, Blacks constitute slightly less than 50% of the city's population.3 At the time that plaintiffs brought their suit, in May 1990, minorities comprised a higher percentage of the city's police sergeants than they did of entry-level officers. The city continued to administer its system of tests and seniority weighting to rank the non-minority applicants for sergeant, in order of their eligibility for the 50% of promotions to sergeant that were designated for them. At the same time, the city began ranking minority applicants separately, using the same criteria, for the 50% of promotions that were reserved for them.

Under this new arrangement, certain non-minority persons, like the plaintiffs, have been denied promotions to sergeant that they would otherwise have received in the absence of separate promotion lists. Therefore, they have not received the increased salaries and benefits that they would presumably have obtained if the Flint Plan had not modified the previous promotion system. Furthermore, they cannot begin accruing the seniority as sergeants that they would need in order to become eligible for future promotions. Plaintiffs therefore brought this action against the city for unlawful race discrimination under 42 U.S.C. §§ 1983 and 1981.

The district court considered plaintiffs' arguments that the HRC hearings, which the city used as a basis for implementing its promotion plan, were invalid because witnesses were not cross-examined and the Federal Rules of Evidence were not followed. The court correctly noted that legislative hearings are inherently political and contribute to the public weal in ways different from evidentiary proceedings conducted by judges. The court also observed that it was not familiar with any legislative body that conducts public hearings governed by court rules. Middleton v. Flint, 810 F. Supp. 874, 879 (E.D.Mich.1993).

Plaintiffs criticized the HRC findings on other grounds as well. Recognizing that the city may have discriminated in the past in hiring entry-level minority police officers, they argued that the police department's promotion policies and practices prior to the adoption of the Flint Plan were not discriminatory. Since sergeants can only be promoted from within the ranks of lower-level officers, and not directly from the greater municipal labor force, the plaintiffs maintained that a proper analysis of promotion policies should focus on the pace at which minority police officers are promoted. Plaintiffs' labor-force expert determined that minorities comprise a higher percentage of police sergeants than of lower-ranking police officers. The plaintiffs thus contended that promotions must have been awarded fairly. Therefore, plaintiffs claimed the city's decision to extend its affirmative action plan to promotions violates a long line of Supreme Court caselaw rejecting such plans in fields of endeavor not marked by past discrimination.

The district court, however, adopting Dr. Bendick's findings and statistical analysis, upheld the 1:1 ratio for promotions and granted the city's motion for summary judgment on the plaintiff's claims under 42 U.S.C. § 1983 and 1981. The plaintiffs brought this timely appeal from that decision.4

As a preliminary matter, we note that this appeal is from a plan that a municipality has voluntarily adopted on its own initiative. Thus, this plan differs from those that are crafted under the direction of a court order, resulting from a judicial factfinding process, as happened in United States v. Paradise, 480 U.S. 149, 107 S. Ct. 1053, 94 L. Ed. 2d 203 (1987) (plurality opinion). Nor was this plan presented to a court by the participating parties as part of a motion for a judicially approved consent decree, as was the case in Vanguards of Cleveland v. City of Cleveland, 753 F.2d 479 (6th Cir.1985), aff'd sub nom. Local 93, Int'l Ass'n of Firefighters v. City of Cleveland, 478 U.S. 501, 106 S. Ct. 3063, 92 L. Ed. 2d 405 (1986). Rather, this plan, much like the one adopted in City of Richmond v. J.A. Croson Co., 488 U.S. 469, 109 S. Ct. 706, 102 L. Ed. 2d 854 (1989), for awarding city contracts to minority business enterprises, has been voluntarily conceived in the course of a local government's political process.

* In its Croson decision, the Court rejected a city's plan to set aside 30% of its construction contracts for "minority business enterprises" (MBEs). The Court noted that the city had not endeavored to show that minorities were actually facing discrimination in the local construction industry. Even though the city's population was 50% Black, and notwithstanding that only 0.67% of the city's prime construction contracts had been awarded to MBEs in recent years, the Court held that the city had to show that the numbers resulted from discrimination, not merely that such numbers existed:

While there is no doubt that the sorry history of both private and public discrimination in this country has contributed to a lack of opportunities for black entrepreneurs, this observation, standing alone, cannot justify a rigid racial quota in the awarding of public contracts....

....

... There is no doubt that "[w]here gross statistical disparities can be shown, they alone in a proper case may constitute prima facie proof of a pattern or practice of discrimination" under Title VII. But it is equally clear that "[w]hen special qualifications are required to fill particular jobs, comparisons to the general population (rather than to the smaller group of individuals who possess the necessary qualifications) may have little probative value."

488 U.S. at 499-501, 109 S. Ct. at 724-26 (quotations omitted). Furthermore, the Court saw no logic in some of the expansive features of the Richmond plan:

There is absolutely no evidence of past discrimination against [certain other minorities that had been included in the plan's 30% set-aside category, such as] Spanish- speaking, Oriental, Indian, Eskimo, or Aleut persons in any aspect of the Richmond construction industry.... It may well be that Richmond never has had an Aleut or Eskimo citizen. 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.

Id. at 506, 109 S.Ct. at 728.

Similarly, three years earlier, the Court struck down a school board's affirmative action plan that extended preferential protection from layoffs to certain minorities, holding that such racially based plans may not be imposed to the disadvantage of non-minority workers in government units when those agencies have not been shown to discriminate.

Societal discrimination, without more, is too amorphous a basis for imposing a racially classified remedy.... [A]s the basis for imposing discriminatory legal remedies that work against innocent people, societal discrimination is insufficient and over expansive. In the absence of particularized findings, a court could uphold remedies that are ageless in their reach into the past, and timeless in their ability to affect the future.

Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 276, 106 S. Ct. 1842, 1848, 90 L. Ed. 2d 260 (1986) (plurality opinion).

By contrast, where specific findings do evidence unequivocal and egregious manifestations of long-standing discrimination, the Court has upheld affirmative action plans and even extremely limited quota plans on a brief, finite basis. For example, under the extraordinary circumstances presented in Paradise, the Court upheld a one-time quota, requiring the promotion of eight minority state troopers to the rank of corporal, in order to remedy the unyielding record of bigotry in the Alabama Department of Public Safety ("the Alabama department"). The Court recounted the legal battles, court orders, and consent decrees that had been pursued for four decades in a futile effort to pry open the door for reasonable employment and promotion opportunities for minorities. It related a 1972 finding by the district court that the Alabama department had not merely discriminated against Blacks who sought entry-level positions, but had also not accepted a Black trooper at any rank during the prior thirty-seven years. Id. at 168, 107 S.Ct. at 1065. By 1978, of 232 Alabama state troopers at the rank of corporal or higher, there was still not one who was Black. Id. at 169, 107 S.Ct. at 1065. Despite court orders and two consent decrees, the Alabama department had not implemented "acceptable procedures for [the] advancement of black troopers into [their] structure." Id. at 163, 107 S.Ct. at 1062. To make matters worse, "for the purpose of frustrating or delaying full relief to [minorities, the Alabama department] artificially restricted the size of the trooper force and the number of new troopers hired." Id. at 156-57, 107 S.Ct. at 1059. In fact, to keep minority troopers at a minimum, the Alabama department, when compelled to hire minorities, selected individuals who were "other than the best qualified blacks from the eligibility rosters," id. at 157, 107 S.Ct. at 1059 (emphasis added), in order to stimulate higher rates of subsequent minority attrition. During the same time period that the people of Flint were electing Mayor Sharp, the district court, as quoted by the Supreme Court, summed up the very different situation in Alabama:

On February 10, 1984, less than two months from today, twelve years will have passed since this court condemned the racially discriminatory policies and practices of the Alabama Department of Public Safety. Nevertheless, the effects of these policies and practices remain pervasive and conspicuous at all ranks above the entry-level position. Of the 6 majors, there is still not one black. Of the 25 captains, there is still not one black. Of the 35 lieutenants, there is still not one black. Of the 65 sergeants, there is still not one black. Of the 66 corporals, only four are black. Thus, the department still operates an upper rank structure ... that totally exclude[s] black persons. Moreover, the department is still without acceptable procedures for advancement of black troopers into this structure, and it does not appear that any procedures will be in place within the near future. The preceding scenario is intolerable and must not continue.

Id. at 162-63, 107 S.Ct. at 1062. Under those extraordinary circumstances of "pervasive, systematic, and obstinate discriminatory conduct," id. at 190 n. 1, 107 S. Ct. at 1076 n. 1 (Stevens, J., concurring), the Court approved an extremely circumscribed 1:1 hiring quota, limited to requiring that eight minority members be promoted in tandem with eight non-minority members to the rank of state trooper corporal. Furthermore, even under those appalling circumstances, the divided Court stated that the quota would not continue to be valid if the Alabama department were to take other less drastic measures to mitigate its ongoing discriminatory practices. Thus, as the district court noted, "the Department had 'the prerogative to end the promotional quotas at any time, simply by developing acceptable promotion procedures.' " Id. at 164, 107 S.Ct. at 1063. When the Alabama department, in 1984, finally presented to the court an acceptable procedure for promotions to corporal, the court suspended the quota. Id. at 165, 107 S.Ct. at 1063. Thus, the quota was "temporary and extremely limited" and "was used only once at the rank of corporal and may not be utilized at all in the upper ranks." Id. at 182, 107 S.Ct. at 1072; see also id. at 188, 107 S.Ct. at 1075 (stressing that the quota had been used "only on one occasion") (Powell, J., concurring). While the Court observed that the district court's "firsthand experience with the parties" left it "best qualified to deal with the 'flinty, intractable realities of day-to-day implementation of constitutional commands,' " id. at 184, 107 S.Ct. at 1073 (quoting Swann v. Charlotte-Mecklenburg Bd. of Educ., 402 U.S. 1, 6, 91 S. Ct. 1267, 1271, 28 L. Ed. 2d 554 (1971)), our review clearly indicates that the fact pattern in Paradise does not prevail in this case.

In this appeal, the Flint city government cites the HRC hearings and the early 1970s court battles, along with statistical comparisons of the composition of the city's police sergeants to the general Flint labor force, to argue that the city's 1:1 promotion plan is as justified as the one upheld in Paradise. By contrast, the plaintiffs point to a clear turning point in 1973 that was recognized by the district courts in both Holliman and Alfaro. After 1973, the Flint police department's hiring policies and practices changed dramatically, albeit under the pressure of litigation. Therefore, the Holliman and Alfaro courts refused to impose quotas, or even to establish formal goals and timetables, that would interfere with the improved hiring and promotion process being developed within the police department.

Plaintiffs also point to the police department's higher percentage of minority police sergeants than of minority entry-level officers as evidence that the promotion policies were not marred by discrimination. It is therefore the plaintiffs' contention that their city, like the Richmond of Croson, has implemented an "affirmative action plan" that can no longer be justified, if indeed it ever could have been.

In considering race-based quota5 plans, or even less drastic variations of "affirmative action,"6 the courts employ a "strict scrutiny" analysis. Wygant, 476 U.S. at 273-74, 106 S.Ct. at 1846-47; United Black Firefighters Ass'n v. City of Akron, 976 F.2d 999, 1006 (6th Cir.1992). "The Court has recognized that the level of scrutiny does not change merely because the challenged classification operates against a group that historically has not been subject to governmental discrimination." Wygant, 476 U.S. at 273, 106 S.Ct. at 1846. Under that standard, Flint must prove that it had a compelling state interest when it enacted its plan, and that the plan is narrowly tailored to further that compelling state interest. Croson, 488 U.S. at 505-07, 109 S.Ct. at 728-29; Long v. City of Saginaw, 911 F.2d 1192, 1195 (6th Cir.1990). This position was strongly reemphasized by the Supreme Court in Adarand Constructors, Inc. v. Pena, --- U.S. ----, ----, 115 S. Ct. 2097, 2113, 132 L. Ed. 2d 158 (1995). "[W]e hold today that all racial classifications, imposed by whatever ... governmental actor, must be analyzed by a reviewing court under strict scrutiny."

In conducting a strict scrutiny review, we first consider whether the Flint Plan is justified by a "compelling state interest." Thus, we inquire whether the city "had a strong basis in evidence for its conclusion that remedial action was necessary." Wygant, 476 U.S. at 277, 106 S.Ct. at 1849 (emphasis added); Vogel v. City of Cincinnati, 959 F.2d 594, 601 (6th Cir.), cert. denied, 506 U.S. 827, 113 S. Ct. 86, 121 L. Ed. 2d 49 (1992); Long, 911 F.2d at 1196. "[T]he purpose of strict scrutiny is to 'smoke out' illegitimate uses of race by assuring that the legislative body is pursuing a goal important enough to warrant use of a highly suspect tool." Croson, 488 U.S. at 493, 109 S.Ct. at 721 (emphasis added). In considering the city's claim that its plan is compelled by a history of past discrimination, the district court was presented with three kinds of evidence: prior judicial findings, anecdotal evidence, and competing statistical analyses that sought to compare minority hiring and promotions with purportedly relevant demographic pools. We review, in turn, each of these three evidentiary presentations.

* As the record before us makes clear, the federal courts have been invited by litigants for more than twenty years to order hiring and promotion quotas in the Flint police department, and the courts have refused to do so. Those courts were presented with the pre-1973 history, just as the HRC was during its public hearings. Similarly, those courts were offered statistical analyses. They found that discrimination existed before 1973, but they also found that the police department was taking steps to rectify past wrongs. Since the time of Holliman and Alfaro, minority representation on the police force continued to increase, in some measure because of the litigation that brought the police department's practices under the watchful eyes of the courts. Minority promotions to sergeant increased at an even sharper pace, again in an atmosphere marked by the awareness that the courts were watching, but without the imposition of quotas by the court. Notwithstanding their findings of past discrimination in the years before 1973, those pre- Croson courts consistently found a quota on hiring or promotion to be an unnecessarily drastic remedy, in light of the circumstances before them.

In the 1970s, as part of a state court consent decree, the Flint Police Department instituted a temporary 1:4 quota system that was directly aimed at its sergeant-promotion program. Flint Police Patrolmens [sic] Ass'n v. City of Flint, No. 74-33234-CL (Genesee County Cir. Ct. Dec. 2, 1974) (unpublished). This system was a temporary plan that expired by its own terms when fifteen percent of the city's sergeants were minority members or, in any event, after five years. Today, that minority percentage has more than doubled. Therefore, the prior judicial findings that repeatedly and explicitly rejected invitations to impose hiring and promotion quotas weigh heavily against the present Flint Plan.

The district judge in this case observed that the prior court decisions in the 1970s "did not consider ... the type of statistical analysis for a disparate impact analysis that [Flint] put forward in 1984 or in the case at bar, in which discrimination at the entry level is taken into account in determining the need for affirmative action at the promotion level." Middleton, 810 F. Supp. at 880. Nevertheless, the district court concluded that the prior cases "do not constitute sufficient evidence to indicate the need for a 1:1 quota in the hiring of police sergeants." Ibid. The factual record before us sustains the district court's finding.

Next, the district court observed that "the anecdotal testimony [before the HRC] serves to show a culture on the police force that is discriminatory," see id. at 881, and it regarded that material to be "considerable and worthy of weight." Id. at 880. Nevertheless, the court once more concluded that such evidence could not, on its own, provide the necessary legal support to justify the quota. Id. at 881.

"While anecdotal evidence may suffice to prove individual claims of discrimination, rarely, if ever, can such evidence show a systemic pattern of discrimination." O'Donnell Constr. Co. v. District of Columbia, 963 F.2d 420, 427 (D.C.Cir.1992) (emphasis added) (quoting Coral Constr. Co. v. King County, 941 F.2d 910, 919 (9th Cir.1991)). "Anecdotal evidence is most useful as a supplement to strong statistical evidence...." O'Donnell, 963 F.2d at 427 (emphasis added). The record before us on appeal supports the district court's finding that the anecdotal evidence did not justify the city's decision to implement a compulsory ratio for promoting police officers of different groups to sergeant.

In upholding the Flint Plan, the district court based its justification on the statistical evidence developed by Dr. Bendick, the city's labor-force analyst. The district court noted that the Paradise Court expressed concern over whether a department's promotion procedures can be considered fair if entry-level hiring at that department is marred by discriminatory practices:

[Defendants] argue that no remedial relief is justified in the promotion context because the intentional discrimination in hiring was without effect in the upper ranks, and because the Department's promotional procedure was not discriminatory. There is no merit in either premise.

Discrimination at the entry level necessarily precluded blacks from competing for promotions, and resulted in a departmental hierarchy dominated exclusively by nonminorities.

480 U.S. at 168, 107 S. Ct. at 1065 (emphasis added). When Flint introduced its plan in 1984, the percentage of police sergeants who were minority-group members was 10.8%. Bendick at 7. According to the city's expert, basing his numbers primarily on the percentage of minority members in the general workforce of Flint,7 the percentage of minority sergeants should have been 31.1% at that time. In the decade since 1984, as the percentage of minorities in the general labor pool increased, so did Bendick's estimation of the proper representation of minority sergeants on the police force. Consequently, in a 1992 update, he ultimately determined that, in light of a general Flint labor pool whose minority composition had grown to 45% by 1984, the percentage of minority sergeants in the Flint police force should have reached 41.5% by 1987. Dr. M. Bendick, Representation of Minorities 8 (Feb. 1992). Although the minority proportion of sergeants reached 36% in 1992, the discrepancy at the time that the city adopted its plan was more pronounced, with minorities comprising 10.8% of the city's sergeants rather than the 31.1% that Dr. Bendick believed should have been in place in 1984. These numbers, when combined with the anecdotal accounts of continuing racism in the Flint police department, persuaded the district court that there was a "sufficient basis in evidence" to find that Flint had a compelling state interest to adopt its plan.

We hold that, as a matter of law, the district court erred by adopting, without deeper analysis, Dr. Bendick's raw statistics concerning the city's general labor pool as the relevant population against which to gauge the City of Flint's police promotion policies and practices. Furthermore, we hold that a disparity between the percentage of a protected class employed in a particular workforce or occupation and the raw percentage of class members in a regional labor pool, standing alone, cannot be "a strong basis in evidence" sufficient to justify hiring or promotion quotas.8 That is, it is permissible to remedy discrimination. It is not permissible to remedy disparity, without more.

In Croson, the Court confronted statistics showing that "minority businesses received 0.67% of prime contracts from the city [of Richmond] while minorities constituted 50% of the city's population." 488 U.S. at 499, 109 S.Ct. at 725. The Court acknowledged:

There is no doubt that "[w]here gross statistical disparities can be shown, they alone in a proper case may constitute prima facie proof of a pattern or practice of discrimination" under Title VII. But it is equally clear that "[w]hen special qualifications are required to fill particular jobs, comparisons to the general population (rather than to the smaller group of individuals who possess the necessary qualifications) may have little probative value."

Id. at 501, 109 S.Ct. at 726 (emphasis added) (quoting Hazelwood Sch. Dist. v. United States, 433 U.S. 299, 307-08 & n. 13, 97 S. Ct. 2736, 2741-42 & n. 13, 53 L. Ed. 2d 768). Therefore, the Croson Court concluded, "where special qualifications are necessary, the relevant statistical pool for purposes of demonstrating discriminatory exclusion must be the number of minorities qualified to undertake the particular task." Croson, 488 U.S. at 501-02, 109 S.Ct. at 726 (emphasis added). Although presented with a gross numerical disparity between the proportion of minorities living in Richmond and the proportion of minority business that had received city construction contracts, the Croson Court held that:

standing alone this evidence is not probative of any discrimination in the local construction industry. There are numerous explanations for this dearth of minority participation, including past societal discrimination in education and economic opportunities as well as both black and white career and entrepreneurial choices. Blacks may be disproportionately attracted to industries other than construction.

Id. at 503, 109 S.Ct. at 726-27 (citing 1986 Presidential report that "black-owned businesses are more than proportionately represented in the transportation industry, but considerably less than proportionately represented in the wholesale trade, manufacturing, and finance industries"). The Court decried the " 'completely unrealistic' assumption that minorities will choose a particular trade in lockstep proportion to their representation in the local population." Id. at 507, 109 S.Ct. at 729.

By way of comparison, we note abundant evidence that persons whose origin is the Indian subcontinent are very heavily represented in the lodging and motel business. Edwin McDowell, Hospitality Is Their Business, N.Y. Times, Mar. 21, 1996, at D1 (46% of America's economy hotels and 26% of total lodgings owned by Indian-Americans); Greg Johnson, Column One, L.A. Times, Apr. 1, 1996, at A-1 (12,500 lodging properties owned by Indian-Americans surnamed Patel). This by a population that represents less than .5% of the United States population. 1996 World Almanac 386. Is this a sign of discrimination, whether in favor of or against this particular group? Hardly. Yet, any such "disproportion" necessarily must create a corresponding disproportion for Indian-Americans in other lines of endeavor. Thus, as implied by the Supreme Court's holdings in Croson and Adarand, a plan is constitutionally unsupportable when it is driven, at bottom, by nothing but the bare assumption that representation of various groups in specific employment categories should be derived from general proportions, even with whatever "adjustments" the experts, litigants, or courts may think appropriate.

In this case, there is only one population group that is eligible for promotion to the rank of Flint police sergeant. That relevant statistical pool is the population of previously hired Flint police officers. Plaintiffs argue that 30% of those officers in 1992 were minorities, while 36% of the sergeants were minorities at that time, seemingly indicating that the police department's promotion policies and practices have been free of discrimination against minorities. However, the city maintains that the entry-level police population has itself been tainted because of past discrimination against minorities seeking to join. Thus, if Dr. Bendick's supposition is correct, there should have been more entry-level minority officers, and consequently more promoted sergeants.

We have carefully reviewed Dr. Bendick's report and deposition testimony, and, while some of his statistics may be described as somewhat useful, we find his interpretive statistics, projections, and analyses to be misleading because they are based on conclusory assumptions that have no basis in the law. See supra at 399. While he adjusts his numbers carefully to allow for age and education differentials, for inefficiencies in census-taking, for lower minority presence in the populations outside Flint that have supplied some of the city's officers, and for an assumed "propensity of minorities to apply for civil service jobs at a higher rate than for jobs in the private sector," he assumes without any substantiation whatsoever that minority men and women as a group are exactly as inclined as are non-minority men and women as a group--neither more so nor less so--to seek employment in the police department, and that they are exactly as qualified under non-discriminatory conditions. Thus, he assumes that their 45% presence in the 1984 general labor force should have resulted in their 39.6% presence as police officers and 31.1% as sergeants. Bendick at 7.

This court has observed in the past that statistical comparisons with the general workforce of a city are not reliable indicators of discrimination because such a population sample is too general to constitute a relevant statistical pool. See Vogel, 959 F.2d at 600; Long, 911 F.2d at 1202. Furthermore, in Long, when we were presented a record for review that included statistics showing a limited minority presence in a Michigan city's police force, we noted that "this court should be ever mindful that there could have been and generally were numerous explanations for disparities between the percentage of minority and nonminority police officers on the Saginaw police force, many of them unrelated to discrimination of any kind." Id. at 1198-99. We also noted that the City of Saginaw had experienced a significant exodus of white residents during the prior decade. The population shift resulted in minorities comprising a higher percentage of the city's general labor force after 1980 than they had constituted in previous years, when many police officers had been hired. Id. at 1200.9 In addition, "the city ... failed to consider the high noncoercive turnover (voluntary terminations, retirements, illnesses, etc.) rate among minorities (27.5% of the total turnover rate within the Police Department) as compared to the lesser rate among nonminorities that would, within the context of th[e] case, obviously result in significant statistical distortions...." Ibid.

Just as "[t]here is no iron law of human behavior that every racial or ethnic group will perform equally well on nonbiased examinations in all fields of human endeavor," Billish v. City of Chicago, 989 F.2d 890, 896 (7th Cir.) (en banc) (quoting United States v. City of Chicago, 870 F.2d 1256, 1261 (7th Cir.1989)), cert. denied, 510 U.S. 908, 114 S. Ct. 290, 126 L. Ed. 2d 240 (1993), "it is common for different groups to rely on different mobility ladders." Thomas Sowell, Preferential Policies: An International Perspective 132 (1990) (quoting Cynthia H. Enloe, Police, Military and Ethnicity: Foundations of State Power 143 (1980)). If the Black population of Flint has inclined to rely on police service as an important mobility ladder, then it could well be that even 45% minority representation would be inadequate to withstand subsequent claims by individual Black applicants that they have faced discriminatory barriers to employment or promotion within the department. On the other hand, as seemed to be the case with the construction-related entrepreneurship discussed in Croson, it may be that the police department of Flint is not the mobility ladder of choice for the city's minorities. Indeed, during the course of his deposition in this case, Thomas Bugbee, who had been City Personnel Director under Mayor Sharp and others, revealed that, for many years during the 1970s and 1980s, "we had a lot of white males coming down [applying for police department positions] who had baccalaureate degrees out of Ferris and Michigan State. And we were not getting college degreed blacks, for example, coming down. Generally, you know, they were recruited heavily for other kinds of positions." Deposition at 36.

Therefore, standing alone without a more serious inquiry along the lines that we have discussed, Dr. Bendick's assumption that the Black population of a race-neutral city police department will virtually mirror the Black population of the general workforce at large is merely conclusory.

Furthermore, as this court has noted previously, evidence of past discrimination that is remote in time will not support a claim of compelling governmental interest when other evidence is adduced to show that the governmental body has taken serious steps in subsequent years to reverse the effects of past discrimination and to implement appropriate new standards. Thus, in Brunet v. City of Columbus, 1 F.3d 390 (6th Cir.1993), we held that strong evidence proffered in 1989 that a city fire department had discriminated prior to 1975 "is too remote to support a compelling governmental interest to justify the affirmative action plan," especially in light of evidence that the city had subsequently taken steps to improve its recruitment efforts. Id. at 409. Similarly, in our case, as the HRC's 1984 report grudgingly acknowledged, the Flint police department had undertaken "initiatives ... including improvement in validation of exams; modifications of certification procedure; dissemination of information regarding employment opportunities to protected classes; pre-employment inquiries in compliance with state and federal laws; and in-service training and retraining programs for all personnel." HRC at 13.

Even if the city had presented statistics evidencing a compelling state interest in imposing its plan, the plan would need to pass a second measure of strict scrutiny to ensure that it is narrowly tailored to achieve that compelling state interest. In Paradise, the Supreme Court identified several factors to consider "[i]n determining whether race-conscious remedies are appropriate [i.e., narrowly tailored]." 480 U.S. at 171, 107 S.Ct. at 1066. These include: (1) the necessity for the relief; (2) the efficacy of alternative remedies; (3) the flexibility and duration of the relief, including the availability of waiver provisions; (4) the relationship of the numerical goals to the relevant labor market; and (5) the impact of the relief on the rights of third parties. Ibid.

This opinion is not intended to deny that racial discrimination and disadvantage exists at many levels of our society and in many ways. Our colleague, Judge Keith, has eloquently stated:

I know this to be true, because as a black man in these United States, I face racism daily. There is not a day that passes when I am not reminded of the color of my skin.... Every aspect of American life--even the judicial system--is tainted by racism.

Jim Fitzgerald, Wrong Questions, Right Answers, Detroit Free Press, Dec. 28, 1990, at 8-F.

Judge Keith has been further quoted, in a speech by former United States District Judge Robert M. Duncan, as stating, "Every day of my life I am in some fashion personally troubled by racial matters in America...." Judge Duncan went on to say, of himself, that despite "42 years of Bar membership, [I] have yet to feel one day of true equity in America." Cincinnati Bar Ass'n Rep. at 5 (Sept. 1994).

However, even the enduring and reprehensible fact of racism does not justify every type and degree of government action based upon race. Rather, the Supreme Court has laid down, in Adarand, Croson, and the other cases set forth above, particular standards as to when and to what extent race may be used by government as a factor to disadvantage some Americans at the expense of others. In resolving the case before us, we are bound to apply those standards.

In this case, the district court proceeded to analyze the Flint Plan under each of the factors identified in Paradise. First, it found that the plan provided necessary relief, especially in light of the history of discrimination on the police force, as shown through the statistical and anecdotal evidence. Second, the court found that the alternative strategies that had been suggested by the plaintiffs were inadequate and that, in the absence of a quota, under-representation would continue until an entire generation of police sergeants retired. Third, the court found the plan flexible enough to satisfy strict scrutiny. In the court's words:

There has been no suggestion that the minorities appointed to sergeant are not qualified. This plan is not of indefinite duration. The Paradise Court also upheld a plan with a long duration. Judge Gadola, in Detroit Police Officers Ass'n v. Young, 765 F. Supp. 393, 396 (E.D.Mich.1991) upheld a 1:1 Plan with a sixteen year duration as reasonable. This Plan will probably expire this year or next year, eight to nine years after inception.

810 F.Supp. at 882 (citation and footnote omitted).

Fourth, the court, adopting the city's definition of the relevant labor market, found that the plan's 41.5% goal set a proper target, and that Paradise "provides authority for the statistical analysis that removes the effect of prior discrimination to determine the relevant qualified pool."

Finally, the district court held that the plan was compatible with the Supreme Court's views on minimizing the adverse impact on third parties:

The Supreme Court has said that unlike the discharge of workers, the granting of promotions is a less painful way to effect affirmative action. Thus, in the Paradise case, the Supreme Court upheld a rigid hiring quota on a police force; whereas, the Court struck down race-based layoffs in Wygant, 476 U.S. at 282-83, 106 S.Ct. at 1851 ("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."). Thus, the case at bar may be distinguished from Long v. City of Saginaw, 911 F.2d 1192 (6th Cir.1990), in which the Sixth Circuit struck down Saginaw's decision to lay off police officers to conform to affirmative action goals.

In reviewing the district court's decision on this appeal, we have already questioned whether the city proved that relief was necessary at any time during the duration of the Flint Plan. Thus, we hold that the first element of the Court's five-step inquiry into "narrow tailoring," the "necessity element," was not met. Neither was the second element, which gauges the efficacy of less drastic, alternate possibilities. For two decades, the district courts have rejected invitations to impose hiring and promotion quotas on Flint's police department, and the evidence of discrimination in the 1970s was more compelling than it is in the 1990s. Nevertheless, the courts, supported by all available statistics, found that the city had taken significant alternative steps since 1973 to ameliorate past wrongs. Even in the egregious case of Paradise, where a quota was upheld, the quota merely required the Alabama department to promote eight Black troopers to corporal in tandem with eight such white promotions. 480 U.S. at 172, 107 S.Ct. at 1067. Even then, the quota was "ephemeral," and it would "evaporate[ ]" as soon as the Alabama department implemented reasonable alternative promotion procedures without an adverse impact. Id. at 178, 107 S.Ct. at 1070. As the Paradise Court observed:

Thus, the [district] court imposed a 50% promotional quota in the upper ranks, but only if there were qualified black candidates, if the rank were less than 25% black, and if the Department had not developed and implemented a promotion plan without adverse impact for the relevant rank.... The court highlighted the temporary nature and flexible design of the relief ordered....

Id. at 163-64, 107 S.Ct. at 1062.

By contrast, the Flint Plan has lasted for nearly nine years, perhaps more,11 continuing even after the percentage of minority sergeants in the force was higher than the percentage of minority entry-level officers. The nine-year plan was imposed even though the city was already being successfully prodded, in less drastic, alternate ways, in the years after 1973: prodded partly by the courts, partly by the city's desire to win federal grants, partly by the electorate, partly by the positive role played by pioneering minority officers who worked for change from within, partly by a sincere desire to hire the best officers, and partly by the consciences of those faced with calls to ameliorate the injustices of the past. Complete reform may not have come overnight, but the message of both Croson and Paradise is that the complete implementation of such reform is not a prerequisite to avoiding a quota remedy.

Third, in terms of its flexibility and duration, the Flint Plan was not sufficiently flexible to meet the requirements of strict scrutiny. In judging "flexibility," courts often consider whether such a plan provides a waiver in a case when sufficient numbers of qualified minorities are not available for hire or promotion. See, e.g., Local 93, Int'l Ass'n of Firefighters v. City of Cleveland, 478 U.S. 501, 512, 106 S. Ct. 3063, 3070, 92 L. Ed. 2d 405 (1986); Vogel, 959 F.2d at 599. On the surface, the Flint Plan seems to "pay lip service" to this concern, stating that "[t]hese goals are flexible targets; they are not quotas or rigid standards. Furthermore, these goals shall not result in unqualified persons being appointed to positions." However, these words are merely conclusory assertions. The operational terms of the plan are different. In Flint, for a police officer to qualify for promotion to sergeant, an applicant must take an examination and emerge, in conjunction with other factors, with a minimum score of 70. That grade is stated as the dividing line that distinguishes the "qualified" candidate from the "unqualified" candidate. However, Flint has, in fact, created a free-floating passing standard, at one time awarding as many as ten bonus points to all test-takers, thereby altering otherwise failing grades as low as 60 into "qualifying" scores. On one list before us in the record, it appears that four of ten "passing" scores of "qualified" minority candidates for sergeant were boosted into the 70s by the 10-point bonus. This inference was confirmed by counsel at oral argument. We also note that an upward alteration of only an additional five points would have turned another six "unqualified" candidates into "qualified" ones. This "flexibility" with grading destroys the city's claim that its plan flexibly waives the 1:1 requirement so that it "shall not result in unqualified persons being appointed to positions."

Moreover, as the record before us shows, the Flint Plan was imposed for a duration of nearly nine years and remained in force at least until the strict percentage of 41.5% was reached. Although the district court stated, 810 F. Supp. at 882, that the Paradise Court upheld a plan with "a long duration," we can find nothing in the Paradise opinion that justifies the district court's undocumented statement. Indeed, the Paradise Court specifically noted that the plan was "temporary," 480 U.S. at 163, 107 S. Ct. at 1062; it was "ephemeral," and it "evaporated" at the ranks of corporal and sergeant upon "implementation of promotion procedures without an adverse impact." Id. at 178, 107 S.Ct. at 1070. Thus, it had been implemented "only on one occasion, when [the district court] ordered the promotion of eight blacks and eight whites to the rank of corporal in February 1984," as a "one-time occurrence." Id. at 188, 107 S. Ct. at 1075 (Powell, J., concurring).

Similarly, in Local 93, the Court affirmed this court's opinion upholding a consent decree that created promotional goals. As the parties emphasized in their application to the district court, and as the district judge reemphasized when he granted the request, the parties shortened the life of the proposed decree from nine years to four years. Thus, the court found that "the amended proposal is more reasonable and less burdensome than the nine-year plan that had been proposed originally." 478 U.S. at 512, 106 S.Ct. at 3070. By contrast, Flint's plan was not flexible; it lasted about a decade, perhaps more, and it has compromised the integrity of its qualification process by boosting previously failing grades of 60-69 into the passing range.

Fourth, as we have noted above, the record does not reflect any relationship between the numerical goal of 41.5% and the relevant labor market. Rather, the city's statistical expert provided a conclusory and somewhat misleading numerical analysis that does not justify the constitutional violation here.

Another way of looking at the third and fourth factors together considers whether "the degree of remedial preference is tied to the effects of past disadvantages or discrimination." Peightal v. Metropolitan Dade County, 26 F.3d 1545, 1558 (11th Cir.1994) (citation omitted). It seems obvious that a plan's tailoring is less "narrow" if it results in a very large degree of preference for minority group members (and corresponding disadvantage for non-minority group members) than if the degree of preference is smaller. This is certainly the situation if the degree of preference chosen is not in some way related to evidence bearing on the degree of discrimination to be remedied. See Shurberg Broadcasting of Hartford, Inc. v. FCC, 876 F.2d 902, 912-16 (Silberman, J., concurring).

In a situation such as Paradise, where the exclusion had been total, it was far more plausible to enforce a draconian remedy, as the information necessary to any type of narrow tailoring was absent. In our case, we have ample information. We know that Black police officers had been hired at least as far back as the 1950s (5 on the force in 1957, 5 hired between 1952 and 1961); that at the 1973 and 1975 sergeant exams, those Blacks that applied passed the exam and were hired at roughly the same rate as whites; that over 10% of sergeants were Black by 1984; and that throughout the period between 1985 and 1991, Black representation at the sergeant level was roughly equal to or greater than such representation at the lowest officer level. Under these circumstances, it is impossible to find that a plan justifying a virtually unlimited degree of preference is narrowly tailored.

In order for the city to procure the necessary numbers of Black sergeants to meet the quota, the record shows that the degree of preference had to be very great. Based on the 1987 exam, twenty-one white officers were passed over to reach the first of the "preferred" Black officers, and an additional twenty-nine had to be passed over before the process was complete. The differences in scores were considerable. Among white officers, the spread passed over was roughly from the 90th percentile to the 50th percentile of the test takers. In contrast, some Black applicants below the 50th percentile were taken. If race were not a factor at all, few observers would say that taking candidates from the bottom half of any ranking was equivalent to taking candidates from the top tenth, whether the job in question was electrician, baseball player, or symphony musician.

Further, we note that quotas and preferences are easily transformed from one into the other. Certainly, where the ranking criteria are already known, the correspondence is exact. In our case, if it were deemed objectionable to admit that there was a 1:1 quota, exactly the same result could have been reached by adding 20 points to the score of each minority applicant at the 1987 exam. A pre-existing commitment to a fixed amount of preference (as with a veterans' preference) has the result, in any given case, of determining exactly the proportion of the favored group that will be selected.

Even when the degree of preference is established in advance, knowledge of the general characteristics of the selection criteria and the applicant pool may well allow a close approximation, for each round of selection, of the desired quota outcomes. Thus, a 15-point preference might be known with confidence to produce approximately a 25% quota, a 20-point preference a 50% quota, and so on. Even though these figures will be more approximate and contingent than when the degree of preference is assigned after the fact, the general result and correlation will be the same. Justice Brennan noted this point in his opinion in Regents of the University of California v. Bakke, 438 U.S. 265, 378, 98 S. Ct. 2733, 2793, 57 L. Ed. 2d 750 (1978) (Brennan, J., concurring in part and dissenting in part).

Turning now to the fifth factor, the Flint Plan has drastically affected the rights of certain third parties. By denying certain otherwise eligible white candidates their earned opportunity to be promoted, an opportunity that has been denied for possibly as long as nine years, the plan denied to the plaintiffs salary increases, increased pensions, and related benefits, as well as the opportunity to accrue seniority to gain subsequent promotions to such higher ranks as lieutenant. While the Supreme Court has held that the denial of promotion opportunities is not as grievous an infringement of personal rights as are other sanctions like demotions, layoffs, or outright discharges, see, e.g., Wygant, 476 U.S. at 282-83, 106 S.Ct. at 1851, the denial of these opportunities for nearly a decade does constitute a very significant, and legally cognizable, adverse impact on the rights of third parties.

In sum, based on the five factors by which the courts have evaluated "affirmative action plans" for "narrow tailoring," we hold that the Flint Plan was not "narrowly tailored," as required to make its racial discrimination constitutional.

Because the record of prior judicial findings, anecdotal testimony, and general labor-force statistics does not provide "a strong basis in evidence" that the Flint Plan serves a "compelling state interest" and because, in any event, it was not "narrowly tailored" to achieve the claimed interest, we REVERSE the district court's grant of summary judgment to the City of Flint, and hold that the Flint Plan's 1:1 quota for promotions is unconstitutional. We REMAND the matter for further proceedings consistent with this opinion, including determining whether any of the plaintiffs were actually injured by the application of the plan, and a determination of whether the city is continuing to practice a de facto quota-driven scheme under the guise of hiring and promoting from one unified list.