UNITED STATES of America, Plaintif–Appellee–Cross–Appellant, Janet Caldero, Celia I. Calderon, Martha Chellemi, Salih Chioke, Andrew Clement, Kristen D'Alessio, Laura Daniele, Charmaine Didonato, Dawn L. Ellis, Marcia P. Jarrett, Mary Kachadourian, Kathleen Luebkert, Adele A. McGreal, Margaret McMahon, Marianne Manousakis, Sandra D. Morton, Maureen Quinn, Harry Santana, Carl D. Smith, Kim Tatum, Frank Valdez, and Irene Wolkiewicz, Intervenors–Appellees–Cross–Appellants, Pedro Arroyo, Jose Casado, Celestino Fernandez, Kevin Lafaye, Steven Lopez, Anibal Maldonado, James Martinez, Wilbert McGraw, Silvia Ortega De Green, and Nicholas Pantelides, Intervenors–Appellees, v. John BRENNAN, James G. Ahearn, Scott Spring, and Dennis Mortensen, Intervenors–Appellants–Cross–Appellees,
New York City Department of Education; City of New York; Martha K. Hirst, Commissioner, New York City Department of City Administrative Services; New York City Department of Citywide Administrative Services, Defendants–Appellees. John Brennan, James Ahearn, Scott Spring, Dennis Mortensen, John Mitchell, and Eric Schauer, Plaintiffs–Appellants, v. Attorney General of the United States; Assistant Attorney General of the United States for Civil Rights; U.S. Department of Justice; New York City Department of Education; City of New York; New York City Department of Citywide Administrative Services; Martha K. Hirst, Commissioner, New York City Department of City Administrative Services, Defendants–Appellees, Janet Caldero, Celia I. Calderon, Martha Chellemi, Salih Chioke, Andrew Clement, Kristen D'Alessio, Laura Daniele, Charmaine Didonato, Dawn L. Ellis, Marcia P. Jarrett, Mary Kachadourian, Kathleen Luebkert, Adele A. McGreal, Margaret McMahon, Marianne Manousakis, Sandra D. Morton, Maureen Quinn, Harry Santana, Carl D. Smith, Kim Tatum, Frank Valdez, and Irene Wolkiewicz, Intervenors–Appellees,
Pedro Arroyo, Jose Casado, Celestino Fernandez, Kevin LaFaye, Steven Lopez, Anibal Maldonado, James Martinez, Wilbert McGraw, Silvia Ortega De Green, and Nicholas Pantelides, Intervenors–Appellees. Ruben Miranda, Plaintiff–Appellant, v. New York City Department of Education, Defendant–Appellee.
In 1996, the United States (the “Government”) sued the New York City Board of Education and related parties (the “City Defendants”) claiming a violation of Title VII's prohibition of disparate impact selection measures. The suit alleged that the City had, in hiring Custodians and Custodian Engineers (“CEs”) for its schools, (1) used, on three separate occasions, civil service examinations which discriminated against blacks and Hispanics, and (2) used recruiting practices which discriminated against blacks, Hispanics, Asians, and women. The parties entered into a settlement agreement in 1999 and asked the district court to enter it as a consent decree. The magistrate judge (Levy, M.J.)—who had jurisdiction by consent—approved the entire agreement, despite objections that primarily came from incumbent employees who were denied leave to intervene in the suit. The incumbent employees were unaffected by many of the agreement's provisions, but they objected to four paragraphs that provided permanent appointments and retroactive competitive seniority to 63 black, Hispanic, Asian, or female individuals, the “Offerees.”2 The City Defendants implemented the disputed parts of the settlement while the incumbent employees' appeal from their exclusion was pending. In 2001, this Court vacated and remanded, holding that the district court should have permitted the incumbent employees to intervene.
After the remand, the incumbent employees asked the district court to have the case sent to a district judge rather than the magistrate judge; this request was granted. They then brought two reverse-discrimination lawsuits against the City under § 703(a) of Title VII and the Equal Protection Clause (via 42 U.S.C. § 1983); these cases were consolidated with the original 1996 Government lawsuit. The new lawsuits sought equitable relief and damages. The incumbent employees also requested class certification. Meanwhile, the Government decided that it would defend the settlement agreement only in part. This prompted two groups of beneficiaries of the settlement to intervene to defend the retroactive seniority that the settlement had granted them. One group, the Arroyo Intervenors, consisted of ten Offerees who had taken and failed a challenged exam; the other group, the Caldero Intervenors, consisted of twenty-two Offerees who had not taken any challenged exam.
After years of contentious litigation, extensive discovery, and a set of three opinions spanning a total of approximately 150 pages, the district court (Block, J.) entered a final judgment. It held that some of the retroactive seniority provided by the settlement agreement violated Title VII, and that some of the retroactive seniority that did not violate Title VII violated the Equal Protection Clause. But, the court concluded that a significant remainder of the retroactive seniority was lawful. In particular, the court held (1) that the retroactive seniority of the test-failer Offerees3 did not violate Title VII or the Equal Protection Clause, except insofar as layoff seniority4 was granted to individuals who were not actual victims of discrimination; (2) that some of the test-failers were actual victims and others were not; (3) that the female non-test-failer Offerees' retroactive seniority did not violate Title VII or the Equal Protection Clause, except that their layoff seniority violated both Title VII and the Equal Protection Clause since—because there was no prima facie case of recruiting discrimination with respect to them—none of them were actual victims of such discrimination; and (4) that the minority male non-test-failers' retroactive seniority did not violate Title VII, but did violate the Equal Protection Clause. As a remedy the court entered a declaratory judgment, accompanied by schedules specifying the extent to which each individual's retroactive seniority was lawful. The class of incumbent employees was also certified, but their damage claims were rejected. Finally, the district court declined to enter any part of the disputed paragraphs of the settlement agreement into a consent decree.
On appeal, the district court's refusal to enter a consent decree is not challenged. The only disputes before us are the incumbent employees' lawsuits, which claim that the City Defendants' (voluntary) implementation of the settlement agreement violated § 703(a) of Title VII and 42 U.S.C. § 1983. The Title VII claim requires us to decide the applicability of Ricci v. DeStefano, ––– U.S. ––––, 129 S.Ct. 2658, 174 L.Ed.2d 490 (2009), which was decided after the district court issued its decision in this case.
The district court's Title VII analysis was based on the “affirmative action” framework of Johnson v. Transp. Agency, Santa Clara County, 480 U.S. 616, 107 S.Ct. 1442, 94 L.Ed.2d 615 (1987), and United Steel Workers of Am. v. Weber, 443 U.S. 193, 99 S.Ct. 2721, 61 L.Ed.2d 480 (1979). We hold that, contrary to the pre-Ricci law in this Circuit, Johnson and Weber do not apply to all race- or gender-conscious employer actions. In light of Ricci, the “manifest imbalance” and “no unnecessary trammeling” analysis of those cases extends, at most, to circumstances in which an employer has undertaken a race- or gender-conscious affirmative action plan designed to benefit all members of a racial or gender class in a forward-looking manner only. Where, as here, the employer instead provides individualized race- or gender-conscious benefits as a remedy for previous disparate impact, the employer must satisfy the requirements of Ricci, not Johnson and Weber, in order to avoid disparate-treatment liability. Under Ricci, the employer must show a strong basis in evidence that, at the time the race- or gender-conscious action was taken, the employer was faced with disparate-impact liability and that the race- or gender-conscious action was necessary to avoid or remedy that liability.
In addition to our central holding, we address several other issues. We discuss a few matters related to the identification of those individuals who, the City Defendants had a strong basis in evidence to believe, were victims of disparate impact and, therefore, as to whom the retroactive-seniority provided in the settlement agreement was a proper remedy. We also discuss what remedies the district court might properly consider, and we advise the district court that while it may be appropriate to strip certain settlement beneficiaries of the retroactive seniority they received from the settlement agreement, it is not correct to strip them of seniority stemming from their permanent appointments. Because it is possible that the case can be resolved on Title VII grounds, we decline at this time to address the important and difficult Equal Protection Clause questions that attach to the case. Finally, we affirm the district court's grant of class certification because, in so certifying, the district court did not abuse its discretion.
Accordingly, we AFFIRM the district court's judgment in part, VACATE it in part, and REMAND the case for further proceedings consistent with this opinion.
Factual and Procedural Background
I. The Parties
This appeal stems from three lawsuits involving five parties or groups of parties, who, from different directions, attack just about every aspect of the district court's judgment. The incumbent employees (the “Brennan Plaintiffs”)5 have appealed, and the Caldero Intervenors and the Government have cross-appealed. Since the City Defendants have withdrawn their cross-appeal and have not submitted briefs, the four parties before us are:
• The Brennan Plaintiffs, who seek to invalidate as much of the four challenged paragraphs of the settlement as possible.
• The Arroyo Intervenors, who seek to defend the settlement for Offerees who failed a challenged exam.
• The Caldero Intervenors, who seek to defend the settlement for Offerees who did not take any challenged exam.
• The Government, which wishes to defend some portions of the settlement, but not others.
II. General Factual Background
The focus of the Government's 1996 lawsuit was the allegedly discriminatory hiring of Custodians and Custodian Engineers by the City Defendants.
A. Custodians and Custodian Engineers
The Board of Education6 (the “Board”) employs Custodians and Custodian Engineers7 to take care of the approximately 1200 buildings in the City's school system. Custodian and CE positions are desirable supervisory jobs with good pay8 and civil-service protections. Each Custodian or CE is assigned to a school, and he or she supervises all the handymen and cleaners at that school. The Custodian or CE also has the responsibility for the upkeep, cleanliness, and safety of the assigned school. The main difference between Custodians and CEs is that CEs have more experience and a difficult-to-obtain stationary engineer's license. During the times at issue in this case, the Board employed over 900 Custodians and CEs.
Custodians and CEs are paid through a century-old, arcane, and idiosyncratic system known as the “indirect system.” See generally Beck v. Bd. of Educ. of City of New York, 268 A.D. 644, 646–47, 52 N.Y.S.2d 712 (N.Y.A.D.2d Dep't 1945). Under this system, the Custodian or CE is both a civil servant and an independent contractor. The Board allots each Custodian a lump sum of money based on the size and other characteristics of the assigned school. The Custodian uses some of this money to hire cleaners and helpers and to pay for supplies, and, up to a specified limit, he retains the rest for himself. The Board does not tell Custodians how to clean and maintain things or whom to hire; it only tells them whether the school is being cleaned and maintained satisfactorily. Custodians and CEs thus, purportedly, have an incentive to maintain their schools adequately but in a cost-effective manner. Id.
B. The Importance of Seniority
The seniority of a Custodian or CE is important in several ways, only three of which are relevant to the appeal before us: seniority improves a Custodian or CE's ability (1) to get transfers to more desirable schools; (2) to get Temporary Care Assignments (“TCAs”)9 ; and (3) to avoid being laid off. At least for transfers and layoffs, the effect of seniority is specified in a collective bargaining agreement between the Board and the union representing Custodians and CEs, Local 891 of the International Union of Operating Engineers (“Local 891”). Before discussing those matters, it is worth noting one way in which seniority is not at issue in this appeal: some seniority (which is termed “non-competitive” seniority) concerns a Custodian or CE's relationship with the Board in ways that do not affect the holder of that seniority's relationship to other Custodians or CEs. For example, the seniority that affects the amount of pay a Custodian or CE receives is, in this sense, “non-competitive.” See supra note 8.
Because Custodians and CEs earn higher salaries when they work in larger schools, it is desirable, when the opportunity arises, for them to transfer to larger schools. When there is a vacancy at a school, seniority plays a crucial role in determining who gets the transfer.10 Pursuant to the “rating and transfer plan” in the collective bargaining agreement for Custodians and CEs, each school in the system is assigned to a particular “seniority bracket” based on its square footage.11 CE brackets are associated with larger schools than the brackets for Custodians who have the same level of experience. In addition, some schools are available only to CEs, while some other schools are available to Custodians only if the Custodians have a refrigeration license.12
Every few months or so, as needed, the Board issues a Vacancy List to Custodians and CEs. The Vacancy List sets out the schools with a Custodian or CE vacancy, and any Custodian or CE who is eligible to transfer may bid for open schools and specify his or her order of preference.13 Custodians and CEs may bid for a school above their seniority bracket, so long as the bidder is not a Custodian bidding for a CE-only school, or a Custodian without a refrigeration license bidding for a school that requires such a license. Then the Board looks at who has applied for each vacant school. If only one eligible person bids for a school, then that person gets the school. The vast majority of vacant schools do not result in competition; either nobody or only one person bids for them. The most desirable vacancies, however, receive multiple bids. In such cases, if at least one bidder is in or above the school's seniority bracket, then any Custodian or CE who is below that seniority bracket will not get the school. In other words, a Custodian or CE can never get a school above his or her seniority bracket unless nobody in or above the required seniority bracket applies for the vacancy. If two or more candidates are in or above the required seniority bracket, then the transfer generally goes to the applicant with the highest performance rating from his or her current school's principal. However, if any other candidate's performance rating is within .25 points of the top candidate's performance rating, then seniority is used as a tie-breaker between them. The vacant school's principal can veto a Custodian or CE, but this does not appear to be a frequent occurrence. Indeed, nothing in the record indicates that it has ever happened. Once an applicant is determined by the Board to be the top applicant for a vacant school, the applicant cannot withdraw his or her bid; transfer to that school is mandatory. About 1% of the schools in the entire system get a new Custodian or CE each time the transfer process occurs.
2. Temporary Care Assignments
Seniority also affects Temporary Care Assignments—though only slightly. When a temporary school vacancy results from illness, vacation, or leave, the Board fills the vacancy through the TCA process. Unlike the transfer process, the TCA process is not found in the collective bargaining agreement. The record conflicts as to whether the TCA process is governed by any contract negotiated with Local 891.14 A Custodian or CE who receives a TCA will, for the duration of the TCA—usually two months—look after the TCA school in addition to his or her regularly assigned school. During that time, the salary of the Custodian or CE will be increased by 75%—or 100% in the unlikely event that the TCA lasts more than six months—even though he or she is not required to put in any additional time on the job. A Custodian or CE can reasonably expect to get a TCA once every two years.
The process for assigning TCAs is somewhat different from the transfer process, but seniority still plays a role. The record is, however, unclear as to how significant that role is. Once a Custodian or CE has at least one year of experience, he or she is added to the TCA waitlist for the district in which his or her regularly assigned school is located, or for adjacent districts.15 As TCAs arise, they are assigned to the Custodian or CE at the top of the list, so long as the TCA school does not require a license or special skill that the top person does not have. Upon finishing the TCA, the Custodian or CE is returned to the bottom of the list.
Because of the rotation system, seniority seems to play only a relatively minor role in the awarding of TCAs. The district court does, however, suggest a way in which seniority might make a difference. According to that court, there are three separate TCA lists in each district: one for CEs, one for Custodians with at least 15 years' seniority, and one for other Custodians. United States v. N.Y. City Bd. of Educ., 448 F.Supp.2d 397, 411–12 (E.D.N.Y.2006) [hereinafter NYC Board III ]. None of the parties' briefs mentions three lists, nor do these briefs cite any part of the record supporting this proposition. But, if the district court's statement is correct, then increasing a Custodian's seniority might increase the size of the 15–year TCA list and thereby work to the detriment of other Custodians on that list. This would correspondingly benefit other Custodians who are on the fewer–than–15–years TCA list by decreasing that list's length. Similarly, it is to the advantage of any particular Custodian or CE for there to be as few as possible other Custodians or CEs with at least one year of seniority in the same or adjacent districts. For then it would not take as long to get back to the top of the list after a TCA assignment.
Seniority directly determines the order in which Custodians or CEs will be laid off under N.Y. Civ. Serv. Law § 80(1). The statute and the CBA provide for a typical “last hired, first fired” system.
C. The Hiring Process
During the disputed period, the hiring process included four steps for Custodians and three steps for CEs. First, because a Custodian or CE is a civil servant within the “competitive class” as defined under New York law, an applicant wishing to become a Custodian or CE had to pass a civil service examination.16 The City Defendants administered three disputed exams: (1) Exam 5040, which was given in 1985 for the Custodian position; (2) Exam 8206, which was given in 1989 for the CE position;17 and (3) Exam 1074, which was given in 1993 for the Custodian position.
Second, an applicant had to submit “experience papers” explaining how the applicant satisfied the minimum qualifications for the position. The minimum qualifications—a few years of relevant experience for a Custodian, and a few more years plus the coveted high-pressure boiler license (also known as a “stationary engineer license”) for a CE—were stated in the notice for each of the aforementioned, disputed, exams. Once the exam was administered, the City Defendants reviewed the experience papers for passing applicants. A finding of insufficient experience could be administratively appealed. The exact proportion of appeals that overturned the City Defendants' original findings is unclear and disputed, but the proportion of successful appeals clearly was not insignificant.18
Third, for Exams 5040 and 1074 (the Custodian exams), but not for Exam 8206 (the CE exam), a practical test was required. The Exam 5040 practical was oral and it was given only to those applicants who had both passed the civil service exam and had had their experience papers accepted. Applicants were taken to the boiler room in a school in Brooklyn and asked questions about what they would do under various circumstances. For Exam 1074, this practical test was changed to a written format, so that the City Defendants could make sure that each applicant was asked the same questions. Also, unlike the Exam 5040 practical, the practical for Exam 1074 was administered—and individuals who failed it were eliminated—before the experience papers were reviewed. A small but non-trivial percentage of applicants failed the practical test.19
After these steps were completed, applicants who had passed the written exam, the “experience papers” stage, and the practical exam (if applicable), were placed on an “eligible list,” starting with the highest scorers on the written exam and ending with the lowest passing scorers. When the Board needed a Custodian or CE, it would follow the “Rule of Three.” Under this rule, it would call the top few people (usually three, but sometimes more if there were more than one vacancy) on the eligibility list for interviews. See N.Y. Civ. Serv. Law § 61(1). The vacancy or vacancies would then be filled by whichever ones of the interviewees were selected by the interviewers. Applicants who were interviewed three times without being hired would be removed from the eligibility list. The Board hired Custodians from the Exam 5040 eligibility list from Spring 1987 through Fall 1990. The Exam 8206 eligibility list was used to hire CEs from Spring 1991 through early 1994, and the Exam 1074 eligibility list was used for Custodians from early 1997 through early 2000.
D. Provisional Employees
In addition to the permanent Custodians and CEs, who had to go through the above described process and who had competitive seniority and civil-service protections, the City Defendants hired “provisional” Custodians and CEs. Provisional Custodians and CEs can be fired at any time, do not accrue competitive or non-competitive seniority, cannot bid for transfers to other schools, cannot obtain TCAs, and have to work wherever the Board puts them. It is therefore much better to be a permanent Custodian or CE than to be a provisional one. But provisional Custodians and CEs have the same responsibilities at the schools to which they are assigned as their permanent counterparts. And they are supposed to have the same experience and, in the case of provisional CEs, the boiler license as well. Provisional Custodians and CEs are hired when the Board needs Custodians and CEs but there is no appropriate eligibility list from which to hire permanent employees. See N.Y. Civ. Serv. Law § 65.
III. The Government's Investigation and Lawsuit
The U.S. Department of Justice began to investigate possible discrimination in the hiring of permanent Custodians and CEs in the early 1990s. A 1993 demographic survey revealed that more than 99% of the permanent Custodian and CE workforce was male, and 92% was white. In contrast, blacks constituted about 20% of the qualified labor pool for these positions, Hispanics made up about 19% and women about 8% of the pool. A 1996 demographic survey showed similar results. It also appeared (as was later confirmed by the investigation) that racial minorities and women were much more likely to be hired as provisional Custodians or CEs than as permanent Custodians or CEs, even though the qualifications for both were the same.20 These data aroused the Government's suspicion that something in the permanent hiring process was discriminatory.
Based on a statistical analysis of the results of the investigation, the Government sued the City Defendants on January 30, 1996. The Government originally made both pattern-and-practice claims (which require proof of intentional discrimination) and disparate-impact claims. But it ultimately pursued only the disparate-impact claims. Two sets of these were asserted.
In the first, the Government alleged that some of the tests employed brought about discriminatory results. Specifically, the claim was that Exams 5040, 8206 and 1074 had a disparate impact on blacks and Hispanics. This allegation was based on a report from two statisticians, who concluded that the statistical significance of the disparities in passage rates between white, black, and Hispanic takers of those three exams21 was overwhelming—it ranged from 2 to nearly 14 standard deviations. See NYC Board III, 448 F.Supp.2d at 407.
The second set of claims, the recruiting claims, alleged that the City Defendants' recruiting practices had a disparate impact on blacks, Hispanics, Asians, and women. In support of these claims, the Government produced a report by Dr. Orley Ashenfelter, a labor economist. The report said that the number of blacks, Hispanics, Asians, and women who took each of the three disputed exams was lower than the number that would be expected based on the representation of qualified individuals in the overall labor pool. Dr. Ashenfelter found that the differences were statistically significant at the 5% level, and for some groups the probability that the differences were the result of chance was much lower. Although Dr. Ashenfelter did not offer any opinion as to the cause for these disparities, the Government asserted that they resulted from limited advertising and word-of-mouth referrals that had a disparate impact on women and minorities.
2. IV. The Settlement
By 1999, extensive discovery had been conducted, but the Government had not moved for summary judgment and the case had not gone to trial. Before the Government had attempted to prove its case in court, the Government and the City Defendants entered into settlement negotiations, which concluded with the signing of a settlement agreement. Although much of the agreement did not give rise to any objections from third parties, four paragraphs, numbered 13–16, came under attack. Only those four paragraphs, and the consequences of their implementation, are at issue on this appeal.
A. Settlement Terms
Paragraph 13 provided that all “Offerees” who were serving as provisional Custodians or CEs as of the date the court approved the settlement agreement would be granted equivalent permanent positions. “Offeree” was defined earlier in the agreement to include two groups: (a) all black, Hispanic, Asian, or female individuals who were employed as permanent or provisional Custodians or CEs as of the date of approval of the agreement; and (b) all black, Hispanic, Asian, or female individuals who had taken one of the challenged exams and had been hired as provisional Custodians or CEs after the date of the agreement but before court approval of the agreement. The list of the 54 people falling into group (a) was attached as Appendix A to the agreement.
Paragraphs 14–16 gave the Offerees retroactive seniority in accordance with the following terms:
• For any Offeree listed in Appendix A of the agreement who had not taken a challenged exam, the retroactive seniority date was the date of his or her provisional hiring.
• For any Offeree listed in Appendix A of the agreement who had taken a challenged Custodian exam, the retroactive seniority date was the earlier of (i) his or her provisional hiring date, or (ii) the “Median Date” 22 for the challenged exam that he or she took, provided that if he or she had taken more than one exam then the earlier median date applied.
• For Offerees not listed in Appendix A of the agreement—those Custodian Offerees in group (b), i.e., those hired as provisionals after the agreement but before court approval of it—the retroactive seniority date was the earliest provisional hire date for that Offeree's job title, as listed in an earlier stipulation regarding provisional hires. Thus, if the Offeree was a provisional Custodian, he or she would get the earliest provisional hire date for Custodians, which was February 28, 1992. If the Offeree was a provisional CE, he or she would get the earliest provisional hire date for CEs, which was April 13, 1990.
The agreement provided that the retroactive seniority dates would “apply for all purposes for which seniority is applied except any applicable probation requirement.” Accordingly, in addition to the TCA, transfer, and layoff benefits that are in dispute, the seniority dates provided various non-competitive benefits, which are not at issue here.
B. Court Approval
After the Government and the City Defendants agreed to the settlement, it went to Magistrate Judge Levy—to whose jurisdiction, as mentioned earlier, the parties had consented pursuant to 28 U.S.C. § 636(c)—for a fairness hearing, because the parties sought to have the agreement entered as a consent decree. See United States v. N.Y. City Bd. of Educ., 85 F.Supp.2d 130, 135 (E.D.N.Y.2000) [hereinafter NYC Board I ], vacated and remanded, Brennan v. N.Y.C. Bd. of Educ., 260 F.3d 123 (2d Cir.2001) [hereinafter NYC Board II ]. The district court gave notice and opportunity to object. Cf. 42 U.S.C. § 2000e–2(n) (preventing subsequent challenges to employment practices implementing a consent judgment if notice and opportunity to object have been given). There were over 300 objections, about half of which were form letters. NYC Board I, 85 F.Supp.2d at 134 & n. 3. Three objectors, who were white male incumbent permanent Custodians or CEs, moved to intervene: John Brennan, James G. Ahearn, and Kurt Brunkhorst. Id . at 134–35.
The Magistrate Judge determined that the Government had made out a prima facie case of disparate impact for both the testing and recruiting claims. Id. at 141–45. Next, he determined that the settlement was fair and reasonable. He rejected the objections of Brennan et al., along with other objections not relevant here. In particular, he noted that the settlement “avoided the need for a complex, expensive, and lengthy trial,” and that extensive discovery had already taken place. Id. at 146. He also said that the settlement was consistent with the objectives of Title VII; and that, although it was the Board and not Brennan, Ahearn, and Brunkhorst who bore the responsibility for the past discrimination the suit and agreement sought to correct, the effect of the Offerees' retroactive seniority on existing permanent Custodians and CEs was minimal. Id. at 146–51. Finally, the Magistrate Judge rejected the argument of Brennan et al. that they were entitled to intervene as of right pursuant to Federal Rule of Civil Procedure 24(a)(2). The court concluded that they did not have a protected interest in their seniority because (1) the remedies afforded the Offerees were “designed only to return employees to the positions they would have been in but for the alleged discrimination,” NYC Board I, 85 F.Supp.2d at 155, and (2) the possibility that the Offerees' retroactive seniority would affect the would-be intervenors was too “remote and speculative” to constitute a cognizable interest, id. at 156. Accordingly, the Magistrate Judge approved the settlement and denied the motion to intervene. Id. at 157.
V. The Settlement Is Implemented
Shortly after the Magistrate Judge's approval of the settlement, the City Defendants began implementing the agreement. They notified 63 Offerees—the 54 individuals listed in Appendix A to the agreement, plus nine other individuals—that they were entitled to relief under the agreement, if they agreed to release all discrimination claims against the City Defendants. 59 Offerees took the settlement, three resigned, and one declined the settlement. These 59 Offerees received permanent status, retroactive seniority, or both.
For transfer and layoff purposes, the retroactive seniority operated as one would expect. The Offerees' retroactive seniority was effective for both the seniority-bracket stage and the tie-breaker stage of the transfer process. And if layoffs were ever to occur, Offerees would be treated for purposes of the last-hired, first-fired rule as having been hired on their retroactive seniority dates. For TCA purposes, however, it is unclear what happened to Offerees who got retroactive seniority. As the district court pointed out, the record is contradictory as to (1) whether Offerees who were already permanent employees immediately went to the top of the list or stayed where they were, and (2) whether Offerees who were not permanent employees, or who had been permanent employees for less than one year at the time they accepted the settlement, were put at the top or the bottom of the list upon completing their probationary periods. See NYC Board III, 448 F.Supp.2d at 412 & n. 22. It is, however, clear that, as provided in the settlement agreement, the Offerees had to wait out a one-year probationary period before they were added to the TCA lists.
6. VI. The Second Circuit Vacates and Remands
Brennan, Ahearn and Brunkhorst appealed the Magistrate Judge's decision to this Court. In August 2001, we vacated and remanded with instructions for the district court to permit the Brennan Plaintiffs to intervene. NYC Board II, 260 F.3d at 133. We held that the Magistrate Judge had misapplied Federal Rule of Civil Procedure 24(a)(2). First, we rejected the Magistrate Judge's argument that the Brennan Plaintiffs' seniority rights were presumptively the result of illegal discrimination, because it “put the cart before the horse.” Id. at 129. Second, we held that an employment civil rights plaintiff or intervenor does not have to show a property interest in his job or its attributes in order to intervene; rather, “where a proposed intervenor's interests are otherwise unrepresented in an action, the standard for intervention is no more burdensome than the standing requirement.” Id. at 131. Third, we held that “because the exercise of seniority rights has a domino effect,” a Brennan Intervenor's “loss of a desirable transfer need not be directly to an Offeree ․ for the loss to be the result of the [settlement] Agreement.” Id. at 132. Therefore, since the Board did not adequately represent the Brennan Plaintiffs' interests, the district court should have allowed them to intervene. Id. at 132–33. We declined, however, to address the merits of the case:
Appellants also ask us to exercise discretionary jurisdiction and rule on the merits of the Agreement, rather than remand the case to the district court․ We think such a course would be ill-advised. Appellants have argued convincingly that they were denied the opportunity to develop a record that would have permitted a full and appropriate ruling on the fairness and constitutionality of the Agreement. Given the heavily factual nature of these issues, we believe that the best course is to remand the case to allow for a full development of the record.
Id. at 133 (citation omitted).
8. VII. Proceedings on Remand
After remand, this case lingered in the district court for about eight years, in the course of which several significant things happened: (1) the Brennan Plaintiffs filed two cases that were consolidated with the main case; (2) the Government began to attack parts of the settlement agreement; (3) the Caldero and Arroyo Intervenors entered the case in response to the Government' change of position; (4) additional discovery and hearings took place; and (5) the district court issued a set of decisions upholding most, but not all, of the settlement against the reverse-discrimination attacks brought by the Brennan Plaintiffs, but declining to approve the settlement as a consent decree.
A. The Brennan Plaintiffs Intervene and File a Related Complaint
In October 2001, on remand, the Brennan Plaintiffs,23 having been permitted to intervene, filed a Complaint in Intervention alleging primarily that the retroactive grants of seniority in paragraphs 13–16 of the settlement agreement violated Title VII and the Equal Protection Clause. The Government moved to dismiss these claims in November 2001. On February 28, 2002, before that motion was fully briefed, Magistrate Judge Levy approved the undisputed portions of the settlement agreement, with consent of the parties.
Because of certain arguments raised in the Government's motion to dismiss, and because the Brennan Plaintiffs wanted to add two new individuals with damage claims (John Mitchell and Eric Schauer), the Brennan Plaintiffs, not satisfied simply to intervene, filed an action on January 11, 2002 (“Brennan ”). In this suit, Mitchell and Schauer claimed that, in violation of Title VII and 42 U.S.C. § 1983, they had each lost a transfer to an Offeree because of the City Defendants' implementation of the settlement agreement. In addition to the damage claims, all the Brennan Plaintiffs sought declaratory and injunctive relief against the City Defendants, DOJ, and the Attorney General. The Brennan case was eventually consolidated with the Government's case in November 2003.
B. The Government Changes Its Position; Offeree Interventions Result
On April 8, 2002, the Government filed a memorandum in response to the Brennan Plaintiffs' February 2002 motion for a preliminary injunction. In this memorandum, the Government only partially opposed the motion. The Government opposed a preliminary injunction against the settlement as to Offerees who had taken a challenged exam, but not as to Offerees who had not. Two days later, the Government withdrew its previous counsel and substituted new counsel.
Next, the City Defendants moved for approval of the remaining paragraphs of the settlement agreement, and the Government opposed this motion as “premature.” In July 2002, a recruiting-claim beneficiary, Janet Caldero, learned that the Government was no longer defending the settlement as to Offerees such as she. She and others in her situation, having obtained counsel, moved to intervene in October 2002. The parties agreed to the 22 Caldero Intervenors' intervention in February 2003. Around this time, the Brennan Plaintiffs also invoked their rights as parties to object to the magistrate judge's 28 U.S.C. § 636(c) jurisdiction 2 to render a final decision, and, as a result, the case was returned to Judge Block.
Then, in September 2003, the Government changed its position again, triggering yet another intervention. In response to some interrogatories during discovery, the Government provided a chart detailing its current thinking about which Offerees were entitled to retroactive competitive seniority. This chart indicated for the first time that, in addition to the 32 Offerees who had not taken an exam, as to most of whom the Government had already indicated that it would not be defending the settlement, the Government would now also not be defending the settlement as to some (but not all) Offerees who had taken a challenged exam. Although the Government did not say that this latter group of Offerees was not entitled to any retroactive seniority, it did say that, for competitive seniority purposes, these individuals were entitled only to retroactive seniority dates later—usually by about two years—than those provided for by the settlement agreement. The chart also denominated the Offerees who had not taken an exam as “recruiting claimants,” and the others as “testing claimants.” That is, the Government separated the Offerees into two groups, based on the stage of the hiring process at which they allegedly suffered discrimination. Having learned of this document, some of the affected Offerees (the Arroyo Intervenors) moved to intervene. Their motion was granted in July 2004.24 See NYC Board III, 448 F.Supp.2d at 417.
VIII. The District Court's Opinions
After further discovery, the Brennan Plaintiffs and the Arroyo and Caldero Intervenors cross-moved for partial summary judgment. Additionally, the City Defendants moved to enter the settlement agreement as a consent decree, and the Brennan Plaintiffs moved for class certification. The district court issued its first of three opinions on September 11, 2006. See YC Board III, 448 F.Supp.2d 397. Some issues of fact remained, however. After a hearing on those facts, the district court issued its second opinion on April 20, 2007. See United States v. N.Y. City Bd. of Educ., 487 F.Supp.2d 220 (E.D.N.Y.2007) [hereinafter NYC Board IV ]. After another hearing dealing with still remaining fact questions, the district court issued its third opinion on May 28, 2008. See United States v. N.Y. City Bd. of Educ., 556 F.Supp.2d 202 (E.D.N.Y.2008) [hereinafter NYC Board V]. Final judgment was, at long last, entered on August 18, 2008.
A. The September 11, 2006 Opinion
In its 2006 opinion, the district court considered first whether the settlement awards violated Title VII, and then whether they violated the Equal Protection Clause. The court concluded:
• As to the 28 black or Hispanic Offerees before the district court in 2006 who had taken and failed a challenged exam:
• The lawfulness of one Offeree's retroactive seniority could not be determined because there was a genuine issue of material fact as to whether Exam 8206 had a disparate impact on Hispanics; a hearing was required on that issue.
• Seven Offerees were actual victims of discrimination and their entire retroactive seniority was lawful.
• There were factual disputes as to whether the remaining twenty Offerees were actual victims of discrimination, but their retroactive seniority for TCAs and transfers was lawful regardless of whether they were actual victims. Their retroactive layoff seniority, however, was lawful only if they were actual victims.25
• As to the 31 black, Hispanic, Asian, or female Offerees before the district court in 2006 who had not taken a challenged exam and/or were not among the groups on which those exams allegedly had a disparate impact:
• One Offeree was really a white male. His retroactive seniority was unlawful.
• The retroactive seniority of the other 11 male non-test-taker Offerees was unlawful unless they could show that they were actual victims of discrimination. According to the district court, giving them relief without such a showing would violate the Equal Protection Clause.
The 19 female non-test-taker Offerees could lawfully receive full retroactive seniority for transfer and TCA purposes, but not for layoffs.26
NYC Board III, 448 F.Supp.2d at 446–47.
The court's reasoning was as follows. First, it addressed a preliminary issue: whether the Brennan Plaintiffs were correct in their assertion that five supposedly Hispanic Offerees were not entitled to any relief because they were not in fact Hispanic. Relying on the EEOC's definition of “national-origin discrimination” in 29 C.F.R. § 1606. 1, the district court held that “an ancestral place of origin is sufficient to establish membership in a protected class.” NYC Board III, 448 F.Supp.2d at 422. Applying that definition, the district court held that one recruiting claimant, Ciro Dellaporte, was not Hispanic. Id. The other four—Kevin LaFaye, Steven Lopez, and the brothers Nicholas Pantelides and Anthony Pantelides, were Hispanic “because LaFaye's father and the Pantelides' mother were born in Puerto Rico, and Lopez's grandfather was born in Mexico.” Id.
Next, the district court sought to determine whether the retroactive seniority awards violated the Brennan Plaintiffs' rights under Title VII. The district court relied heavily on the “affirmative action” framework of Johnson, 480 U.S. 616, 107 S.Ct. 1442, 94 L.Ed.2d 615, and Weber, 443 U.S. 193, 99 S.Ct. 2721, 61 L.Ed.2d 480, consistent with which an employer defending an affirmative-action plan against a Title VII reverse-discrimination challenge needs to show only that (1) there is a “manifest imbalance” in a traditionally segregated job category and (2) the plan does not “unnecessarily trammel” the interests of adversely affected third parties. NYC Board III, 448 F.Supp.2d at 423–24. The district court rejected the Brennan Plaintiffs' argument that the Weber/Johnson framework did not apply. Id. at 428–31. According to the district court, “there is nothing in Title VII that vitiates an affirmative-action plan granting preferential seniority to non-victims of discrimination.” Id. at 429.
Applying the “affirmative action” framework, the district court found a manifest imbalance for both non-test-taker and test-failer Offerees. For the Offerees who had failed a test, the district court said that the Government had satisfied the “manifest imbalance” requirement by showing statistical evidence sufficient to make out a prima facie case of testing discrimination, id. at 425–26, except as to Hispanics who took Exam 8206; as to them a hearing was required, id. at 427. For Offerees who had not taken a test (and who therefore, the court reasoned, could only be victims of recruiting discrimination, if they were victims at all), the district court relied on Dr. Ashenfelter's analysis that found a statistical imbalance between the expected and actual numbers of black, Hispanic, Asian, and female takers of the three challenged exams. Id. at 427–28.
Having found manifest imbalances, the district court proceeded to the “unnecessary trammeling” stage of the inquiry. The district court found that the transfer and TCA effects of the Offerees' retroactive seniority did not unnecessarily trammel the rights of the Brennan Plaintiffs, because race, national origin, or gender would rarely be the deciding factor in whether a Brennan Plaintiff got a transfer, and even when one of these was the deciding factor the Brennan Plaintiff remained eligible for future transfers. Id. at 431. But the district court held that layoff seniority for non-victims of discrimination would unnecessarily trammel the Brennan Plaintiffs' rights. Id. at 431–34.
After resolving those Title VII issues, the district court examined the Equal Protection Clause of the Fourteenth Amendment, since the Brennan Plaintiffs had also attacked the settlement under 42 U.S.C. § 1983. For the minority male Offerees the district court applied strict scrutiny to the settlement grants of retroactive seniority. NYC Board III, 448 F.Supp.2d at 434–35. This meant the City Defendants had to show (1) a compelling interest for adopting the retroactive seniority, and (2) that the retroactive seniority was narrowly tailored to meet that interest. The district court said that, in this particular case, such a showing required the City Defendants to demonstrate a strong basis in evidence that the Board's own tests and/or recruiting practices were discriminatory. The court then held that the required showing was satisfied for Black and Hispanic Offerees who failed a test, because the Board had created the exams and the tests were adequately shown to be discriminatory against those racial groups. It was not, however, satisfied for non-test-taking male Offerees because although there was a significant disparity between the number of white test-takers and the number of black, Hispanic, or Asian test-takers, there was no sufficient evidence that this disparity was caused by the Board's own recruiting practices. Id. at 434–35. The district court rejected the Brennan Plaintiffs' claim that a compelling interest always requires a showing of intentional disparate treatment by the governmental entity asserting the interest, for such a requirement was, in the district court's judgment, inconsistent with Supreme Court precedent and the policy of encouraging settlement of Title VII cases. Id. at 436–38. The court also found that the transfer and TCA seniority remedy for the test-failer Offerees was narrowly tailored under the factors listed in United States v. Paradise, 480 U.S. 149, 171, 107 S.Ct. 1053, 94 L.Ed.2d 203 (1987). Id. at 438–40. Layoff seniority even for these Offerees, however, was not narrowly tailored. Id. at 440–41.
For female Offerees, the analysis was different. The district court, applying intermediate scrutiny, did not require any showing of governmental (Board) involvement in the discrimination the Board sought to remedy, so the City Defendants' interest in remedying gender discrimination could survive intermediate scrutiny even where their interest in remedying race discrimination might not survive strict scrutiny. Id. at 441–43. Having found that transfer and TCA seniority met the “narrowly tailored” standard during its strict scrutiny analysis, the district court easily determined that transfer and TCA seniority were “substantially related” for intermediate scrutiny purposes as well.27 Id. at 442–43.
The district court concluded its first opinion by addressing two other matters. First, the court refused to enter, as a consent judgment, the parts of the settlement agreement that had survived the Brennan Plaintiffs' Title VII and Equal Protection Clause attacks. Id. at 443–44. Second, the Brennan Plaintiffs had sought class certification with respect to their claims for declaratory and injunctive relief. Even though none of the other parties in either action had submitted responsive papers, the district court approved the class under Fed.R.Civ.P. 23. Id. at 444–46. The certified class consisted of “all custodial employees whose layoff-protection rights have been adversely affected by the grant of seniority benefits to beneficiaries who are non-victims of discrimination.” Id. at 446.
The district court then listed the factual issues that still required a hearing:
(a) which, if any, of the beneficiaries, in addition to Lloyd Bailey, Joseph Christie, Belfield Lashley, Gilbert Rivera, Peter Robertin, Felix Torres and Mayra Zephrini (Cintron), are actual victims of discrimination and received the relief to which they were entitled; (b) whether the results of Exam 8206 in respect to Luis Torres satisfy the evidentiary standards for establishing discrimination under Title VII and the Fourteenth Amendment; and (c) whether John Mitchell and Eric Schauer[, the plaintiffs seeking damages in the second consolidated action,] were denied transfers in favor of particular individuals who impermissibly received retroactivity [sic] seniority.
Id. at 447–48.
B. The April 20, 2007 Opinion
After holding hearings on the remaining factual issues, the district court issued a second opinion. See NYC Board IV, 487 F.Supp.2d 220. This opinion discussed the results of those hearings and responded to several motions for reconsideration. In it the court held that:
• Exam 8206 had been adequately shown to have had a disparate impact on Hispanics. Id. at 224–32.
• There was an issue of fact requiring a hearing as to whether the parties to the 1999 settlement agreement intended to distinguish between “testing claimants” and “recruiting claimants,” or whether instead “the intent of the Agreement was to give relief to blacks and Hispanics regardless of whether they failed a challenged exam.” Id. at 232–33.
• If the former were true, then the district court would not hold a hearing as to whether any recruiting-claim Offerees were actual victims of discrimination, because at summary judgment there was insufficient evidence to warrant a hearing on any claim that they were actual victims. Id. at 224 n. 3, 233–34.
• If the latter were true, the district court said, it would be necessary to reconsider the Equal Protection Clause “narrow tailoring” analysis for the male recruiting claimants. Id.
• As a clarification, the Equal Protection Clause prohibited not only contested retroactive seniority for the minority male recruiting claimants, but also seniority benefits based, under the settlement agreement, on these claimants' permanent appointment dates. Id. at 234–35.
• The Brennan Plaintiffs' class was expanded to include “all custodial employees whose seniority for purposes of transfers, TCAs and layoff protection has been adversely affected by the grant of seniority benefits to the Offerees.” Id. at 236. In other words, the class was no longer limited to individuals affected by the grant of seniority benefits to non-victims of discrimination, nor was it limited to individuals who lost layoff protection rather than transfers or TCAs.
C. The May 28, 2008 Opinion
The third opinion was issued in May 2008 after a hearing on the intent of the parties to the 1999 settlement agreement. See NYC Board V, 556 F.Supp.2d 202. The court addressed two issues: first, whether the parties in 1999 intended to categorize Offerees into “testing-claim beneficiaries” and “recruiting-claim beneficiaries”; second, who among the testing-claim beneficiaries was an actual victim of testing discrimination.
At the hearing, the district court heard the testimony of (1) Norma Cote, who had negotiated the settlement for the City Defendants, and (2) Katherine Baldwin, a Government lawyer who was not directly involved in the settlement negotiations but who reviewed and approved the agreement for DOJ policy compliance. Id . at 205–06. Cote testified that the Government had never separated the claimants into recruiting and testing beneficiaries, nor had it told the City Defendants that “the United States[ ] intended the settlement agreement to provide make-whole relief calibrated to each [O]fferee's individual injury.” Id. at 205. Indeed, she testified that she did not recall the Government “ever explain[ing] to [her] why they wanted these individuals to get retroactive seniority.” Id. at 205–06. Baldwin testified that DOJ's policy was “only [to] seek relief for identified victims of discrimination, [m]ake whole relief for identified victims of discrimination.” Id. at 206. Baldwin said that she and DOJ would not have approved the agreement if it had violated that policy. Id.
Based on this testimony, the district court “conclude[d] that had the [City Defendants] and the United States operated under the assumption—now embodied in the Court's holdings—that there was a sufficient evidentiary basis only for testing discrimination, blacks and Hispanics who had not taken one of the challenged exams would not have been included in the list of Offerees.” Id. at 207. The district court noted that the parties must have been aware in 1999 that (1) some recruiting claimants were being given retroactive seniority, because Asians and non-Hispanic, non-black women received that relief, and (2) “some blacks and Hispanics had not taken one of the challenged exams because that factor was relevant in determining whether relief would be based on the median hire date for a challenged exam or provisional hire date.” Id. at 206–07. But the district court said that Baldwin's testimony was “[m]ost compelling,” and that “[a]lthough the Court credits Cote's testimony that the policy was not communicated to the Board, ․ the United States simply would not have condoned any agreement that went beyond [make-whole] relief [to actual victims].” Id. at 207. The court also suggested that it was relying on “the general principle of contract law that ‘reformation may be available where the parties were under no mistake as to the words of the writing, but they supposed that the legal outcome would be different.’ “ Id. at 207 n. 4 (quoting 27 Williston on Contracts (4th ed.2003) § 70:128). For these reasons, the district court decided not to reconsider its holding that retroactive seniority for male recruiting claimants would not be “narrowly tailored” for Equal Protection Clause purposes. Id. at 207.
The court then made determinations as to who among the 27 testing claimants were actual victims.28 The court said that, for each of these claimants, the question was “whether the retroactive seniority they received approximately corresponds to the seniority they would have received but for the discriminatory exams.” Id. at 209. Additionally, “the parties to the Agreement—i.e., the United States and the Board—bear the burden of proof.” Id. The court's determinations were as follows:
• Seven—Lloyd Bailey, Joseph Christie, Belfield Lashley, Gilbert Rivera, Peter Robertin, Felix Torres, and Mayla Zephrini (Citron)—were victims of discrimination who received appropriate make-whole relief, because the Brennan Plaintiffs had (supposedly) acknowledged this.29 Id. at 208 & n. 6.
• Ten—the Arroyo Intervenors—did not require a determination as to actual victim status because they had stipulated to layoff seniority dates.30 Id. at 208 & n. 7.
• Two, Ronald Johnson and Fidel Seara, did not require a determination because the former had retired and the latter had died. Id. at 208.
• For the remaining eight, the parties had stipulated to facts but were unable to agree on their legal effect, so the court had to decide whether they were actual victims who had received make-whole relief.
• Ricardo Cordero was an actual victim and received appropriate make-whole relief. He had failed Exam 5040, and his retroactive seniority date was appropriately the median hire date for that exam. Id. at 209.
• Vernon Marshall was an actual victim, but his retroactive seniority date required adjustment because, “for reasons that are not clear,” it was over two years earlier than the median date for the exam he failed. Id. at 209–10.
• Sean Rivera was not an actual victim. The settlement agreement had given him a retroactive seniority date of November 7, 1995. He scored an 80 on Exam 1074, which had a median hire date of October 27, 1997. This barely-passing score put Rivera close to the bottom of the Eligibility List for the exam, so that he had not been hired at the time of the 1999 settlement agreement and it was unlikely that he ever would be. Against the odds, he was hired from the list after all on February 4, 2000. The district court said that the Government' view that Rivera would have scored higher on a non-discriminatory exam, and would therefore have been hired earlier, was “purely speculative.” Id. at 210.
• The other five testing Offerees—Thomas Fields, Carla Lambert, Angel Pagan, Anthony Pantelides, and Luis Torres—were not actual victims either. The Brennan Plaintiffs, relying on a post hoc review of applicant qualifications performed by the City Defendants while they were still litigating against the Government, argued that these five would have failed the “experience papers” requirement even if they had passed the exams they took. The court said that although, in this post hoc review, unlike the actual hiring process, there was no opportunity for an administrative appeal of an “experience papers” rejection, the Government had failed, at the summary judgment stage, to meet its burden of providing “concrete evidence that the post hoc review was incorrect as to any of the five whose qualifications are challenged.” Id. at 210–11.
IX. The Miranda Lawsuit
In 2006, Ruben Miranda, a Hispanic male incumbent Custodian aligned with the Brennan Plaintiffs, filed yet another lawsuit alleging violations of Title VII and § 1983 by the City Defendants. Miranda alleged that he was denied two transfers in 2005, and that he would have gotten the transfers but for the retroactive seniority given to Offerees Janet Caldero and Marcia Jarrett, who got the transfers instead. He sought injunctive and declaratory relief, as well as damages. That case (“Miranda ”) was consolidated below, was dealt with in the district court's final judgment, and is part of this appeal.31
X. The Stay Applications
After numerous appeals and cross-appeals were filed, three parties sought stays or injunctions pending appeal. The Caldero Intervenors asked for a stay only as to the portion of the district court's declaratory judgment that stripped minority male recruiting-claim Offerees of seniority stemming from their permanent appointment dates. See United States v. N.Y. City Bd. of Educ., 620 F.Supp.2d 413, 415 (E.D.N.Y.2009) [hereinafter NYC Board VI] . The Brennan Intervenors sought to have the City Defendants enjoined from “providing any competitive seniority benefits to any beneficiary of the Settlement Agreement executed on February 11, 1999 (‘the Agreement’) to the extent that (1) the competitive seniority benefit resulted from the Agreement, (2) [the district court] found that those benefits were not a form of make-whole relief to actual victims of discrimination, and (3) [the district court] nonetheless found that the provision of those benefits did not violate the law.” Id. at 416 (modification omitted). The City Defendants sought a complete stay of the judgment. Id. The district court denied all three motions, stating that “neither the threat of irreparable harm nor the likelihood of success on appeal is sufficiently one-sided to justify disturbing what the Court believes to be the legally correct result.” Id. at 417.
I. Title VII Background
Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq., prohibits two types of employment discrimination on the basis of race, color, religion, sex, or national origin.
Disparate-treatment discrimination is prohibited by § 703(a) of the Act, 42 U.S.C. § 2000e–2(a). Such claims require that a plaintiff “establish that the defendant had a discriminatory intent or motive for taking a job-related action.” Ricci, 129 S.Ct. at 2672; see also 42 U.S.C. § 2000e–2(a) (“It shall be an unlawful employment practice for an employer ․ to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual's race, color, religion, sex, or national origin․”).
Disparate-impact discrimination is also barred by Title VII. Disparate-impact claims do not require a showing of discriminatory intent. Disparate impact occurs when an employer uses an employment practice that has a disproportionately adverse effect on protected groups. Ricci, 129 S.Ct. at 2672–73. An employer can rebut a prima facie showing of disparate impact by demonstrating that the employment practice is job-related, and the plaintiff, in turn, can rebut that showing by demonstrating that there is a less discriminatory alternative to the challenged practice. Id.
The Supreme Court, in Griggs v. Duke Power Co., 401 U.S. 424, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971), first interpreted § 703(a)(2) of the 1964 Act to prohibit disparate-impact discrimination. In response to some pro-defendant interpretations of disparate-impact doctrine made by the Supreme Court in Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989), Congress enacted the Civil Rights Act of 1991, Pub.L. No. 102–166, 105 Stat. 1071.32 The 1991 Act added a new section codifying the disparate-impact prohibition. See 42 U.S.C. § 2000e–2(k).
A separate section of Title VII, § 706(g), 42 U.S.C. § 2000e–5(g), addresses the equitable remedies a court may order after liability is proven. The statute provides courts with broad equitable powers:
If the court finds that the respondent has intentionally engaged in or is intentionally engaging in an unlawful employment practice charged in the complaint, the court may enjoin the respondent from engaging in such unlawful employment practice, and order such affirmative action as may be appropriate, which may include, but is not limited to, reinstatement or hiring of employees, with or without back pay (payable by the employer, employment agency, or labor organization, as the case may be, responsible for the unlawful employment practice), or any other equitable relief as the court deems appropriate․
Id. § 2000e–5(g)(1). But some types of individualized remedies are limited to actual victims of discrimination—that is, those individuals who would not have suffered the employer's adverse employment action in a nondiscriminatory world:
No order of the court shall require the admission or reinstatement of an individual as a member of a union, or the hiring, reinstatement, or promotion of an individual as an employee, or the payment to him of any back pay, if such individual was refused admission, suspended, or expelled, or was refused employment or advancement or was suspended or discharged for any reason other than discrimination on account of race, color, religion, sex, or national origin․ Id. § 2000e–5(g)(2)(A); see also Int'l Broth. of Teamsters v. United States, 431 U.S. 324, 367, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977).
As the Supreme Court and this Court have said, the § 703(a) “liability” phase and the § 706(g) “remedial” phase of a Title VII case are separate and must not be confused with one another. See id. at 360–61 (distinguishing “the initial, ‘liability’ stage” from “the second, ‘remedial’ stage”); Cates v. Trans World Airlines, Inc., 561 F.2d 1064, 1070, 1072 (2d Cir.1977). Section 706(g), being a remedial provision and not a liability provision, does not speak to the actions of private parties; only a court is limited by it. Local No. 93, Int'l Ass'n of Firefighters, AFL–CIO C.L.C. v. City of Cleveland, 478 U.S. 501, 521, 106 S.Ct. 3063, 92 L.Ed.2d 405 (1986). Private conduct is, instead, governed by the liability provisions of § 703.
II. Procedural Posture and Standard of Review
Before we discuss the numerous legal issues presented by this case, we pause to consider its procedural posture. The district court declined to enter the settlement agreement as a consent decree, and no party has appealed that decision. Accordingly, there is nothing left for us to decide with respect to the original 1996 NYC Board case brought by the Government. Indeed, the issue of settlement approval is now moot, for, as the district court explained, the settlement agreement expired by its terms on February 11, 2003:
The Court cannot enter those parts of Paragraphs 13–16 that have survived the Brennan intervenors' challenges as a consent judgment. In the first place, the Agreement was executed on February 11, 1999, and under Paragraph 11 was to “remain in force for a period of four years.” Also, the Agreement provided in Paragraph 47 that although the Court shall retain jurisdiction to ensure compliance, its jurisdiction “shall automatically terminate at the expiration of [the Agreement] as set forth in paragraph 1 1.” Magistrate Judge Levy, with the consent of all parties, entered all of its provisions save the challenged Paragraphs 13–16 as a consent judgment on February 28, 2002—well within that four year period—and the Court had jurisdiction to enforce it until February 11, 2003. The same enforcement limitation would have attached to Paragraphs 13–16 if they had also been reduced to judgment prior to the expiration period. In short, even if all of the provisions of Paragraphs 13–16 were deemed valid, the Court would nonetheless decline to reduce them to a judgment which it would be powerless to enforce.
NYC Board III, 448 F.Supp.2d at 443 (citation omitted).
Before us now is a pair of subsequent lawsuits brought by the Brennan Plaintiffs (Brennan and Miranda ) which had been consolidated with the original case below. Those lawsuits allege that the City Defendants' voluntary implementation, which took place around the year 2000, of the settlement agreement violated two statutes as to the Brennan Plaintiffs. The first is § 703(a) of Title VII, 42 U.S.C. § 2000e–2(a). The second is 42 U.S.C. § 1983, by virtue of an alleged violation of the Equal Protection Clause of the Fourteenth Amendment. As remedies for these alleged violations, the Brennan Plaintiffs seek declaratory and injunctive relief, and damages in an amount to be determined at trial.
The parties' briefs and the opinions below indicate confusion over the procedural posture of this case. We therefore emphasize here, as we will throughout our opinion, that this case is about whether the City Defendants are liable to the Brennan Plaintiffs for violations of § 703(a) or the Equal Protection Clause. We are not deciding what remedy the district court could have ordered if, instead of agreeing to settle the case, the Government had proven a disparate-impact violation by the City Defendants in NYC Board. We are also not deciding whether the district court's long ago decision not to approve the disputed portions of the settlement agreement and enter them as a consent decree was proper, in whole or in part. And we are not deciding whether, if the Government had directly attacked the seniority system itself as discriminatory, such an attack would have succeeded despite § 703(h) of Title VII, 42 U.S.C. § 2000e–2(h). None of those questions are before us.
The district court disposed of the vast majority of the relevant issues in Brennan and Miranda upon cross-motions for summary judgment. But, in what appears to be a confusion about the case's procedural posture, there were a few issues resolved by what the district court and the parties describe as “evidentiary hearings.” At these hearings, the district court heard testimony and took evidence. The court subsequently made findings of fact. It is not immediately clear to us why the district court held these hearings rather than a trial, and at oral argument the Government admitted that it didn't know either. Since the district court's 2006 opinion had already declined to enter the settlement agreement as a consent decree, the Brennan Plaintiffs' Title VII § 703(a) and 42 U.S.C. § 1983 claims were what remained. No party to these suits appears to have requested a jury trial, although the Brennan Plaintiffs' damage claims for intentional discrimination in violation of Title VII would certainly have given rise to the right to one, see 42 U.S.C. § 1981a(c). Ordinarily, under these circumstances, a district court would have held a trial to decide any factual issues not resolved at the summary judgment stage. Given what occurred at the evidentiary hearings, and given that the Brennan Plaintiffs do not appear to have objected to the lack of a jury trial, we think it best to treat the hearings as separate bench trials on separate issues. See Fed.R.Civ.P. 42(b) (“For convenience, to avoid prejudice, or to expedite and economize, the court may order a separate trial of one or more separate issues․”). And we review them as such.
This means that in this opinion, “[w]e review de novo the district court's partial grant of summary judgment, construing the evidence in the light most favorable to the non-moving party.” Weintraub v. Bd. of Educ. of City Sch. Dist. of City of N.Y., 593 F.3d 196, 200 (2d Cir.2010). And on this “appeal taken from a judgment entered following a bench trial, we review the district court's findings of fact for clear error, but we review de novo its conclusions of law and its resolution of mixed questions of fact and law.” APL Co. PTE Ltd. v. Blue Water Shipping U.S. Inc., 592 F.3d 108, 110 (2d Cir.2010).
III. Prima Facie Case and Defenses
We consider the Brennan Plaintiffs' Title VII claim first, leaving their constitutional claims to be considered later and only as necessary. See Ricci, 129 S.Ct. at 2672. In order to address the Brennan Plaintiffs' § 703(a) disparate-treatment claim in its present procedural posture, we must first determine which overall framework applies. Generally, § 703(a) claims are analyzed under the burden-shifting rules of McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802–04, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). Under McDonnell Douglas, a plaintiff must first make out a prima facie case, i.e., she “must demonstrate the following: (1) she was within the protected class; (2) she was qualified for the position; (3) she was subject to an adverse employment action; and (4) the adverse action occurred under circumstances giving rise to an inference of discrimination.” Leibowitz v. Cornell Univ., 584 F.3d 487, 498 (2d Cir.2009); accord Vivenzio v. City of Syracuse, 611 F.3d 98, 106 (2d Cir.2010); Aulicino v. N.Y. City Dep't of Homeless Servs., 580 F.3d 73, 80 (2d Cir.2009). Once the prima facie case has been shown, “the burden then must shift to the employer to articulate some legitimate, nondiscriminatory reason” for the adverse employment action. McDonnell Douglas, 411 U.S. at 802; accord Vivenzio, 611 F.3d at 106; Leibowitz, 584 F.3d at 498–99. The burden then shifts back to the plaintiff “to show that [the defendant's] stated reason for [the adverse employment action] was in fact pretext.” McDonnell Douglas, 411 U.S. at 804; accord Vivenzio, 611 F.3d at 106; Leibowitz, 584 F.3d at 499.
Because the Supreme Court's majority opinion in Ricci33 does not explicitly cite McDonnell Douglas or any other case in that line, one might reasonably ask whether some other framework ought to apply in reverse-discrimination cases of this sort. In our view, however, Ricci fits well within the McDonnell Douglas framework. As the Ricci Court explained, New Haven “rejected the test results solely because the higher scoring candidates were white. The question is ․ whether the City had a lawful justification for its race-based action.” Ricci, 129 S.Ct. at 2674. In other words, because the City's decision to reject the test results was explicitly based on a statistical racial disparity, it was beyond dispute that the plaintiffs had made out a prima facie case, so the burden shifted to the defendants to give a legitimate justification for the adverse employment action. The Court then held that, on summary judgment, the defendants had failed to provide a legitimate justification, id. at 2681; and so the “pretext” step of the inquiry was not reached. But if the Court had found a legitimate justification, it appears that a majority of the Justices would have proceeded to that step. Justice Alito's concurring opinion, joined by two other Justices, argued principally that “a jury could find that [New Haven]'s asserted justification was pretextual.” See id. at 2689 (Alito, J., concurring); see also id. at 2683–84 (noting that “[i]f an employer offers a facially legitimate reason for its decision but it turns out that this explanation was just a pretext for discrimination, the employer is again liable,” and citing St. Mary's Honor Ctr. v. Hicks, 509 U.S. 502, 506–07, 113 S.Ct. 2742, 125 L.Ed.2d 407 (1993), a case applying McDonnell Douglas ). The four dissenting Justices disagreed with Justice Alito on whether a jury could find a pretext, but they did not seem to disagree that the burden would shift back to the plaintiffs to show pretext. See id. at 2709 (Ginsburg, J., dissenting) (suggesting that New Haven might have been liable, even on the dissent's view of the law, if the city had been “seeking to exclude white firefighters from promotion”). We, therefore, conclude that Ricci does not impose a new § 703(a) summary-judgment burden-shifting framework, but instead constitutes, in this respect, a straightforward application of the first two steps of McDonnell Douglas.
The earlier Supreme Court cases dealing with § 703(a) challenges to affirmative action plans further support the view that McDonnell Douglas is the appropriate framework for this case. Like Ricci, these cases reconcile the disparate-treatment provisions of Title VII with the statute's other goals by recognizing a defense which employers may raise when they are sued for disparate treatment. The Supreme Court has explicitly stated that the “affirmative action” defense, too, is properly raised at the second step of the McDonnell Douglas framework. See Johnson v. Transp. Agency, Santa Clara County, 480 U.S. 616, 626, 107 S.Ct. 1442, 94 L.Ed.2d 615 (1987) (“This case also fits readily within the analytical framework set forth in McDonnell Douglas Corp. v. Green. Once a plaintiff establishes a prima facie case that race or sex has been taken into account in an employer's employment decision, the burden shifts to the employer to articulate a nondiscriminatory rationale for its decision. The existence of an affirmative action plan provides such a rationale.” (citation omitted)).
We therefore apply McDonnell Douglas to the Brennan Plaintiffs' § 703(a) claim. And the first question we must address is whether the Brennan Plaintiffs have satisfied the requirements for a prima facie case under McDonnell Douglas.34 We easily conclude that the Brennan Plaintiffs have made out a prima facie case. All but one of the individual plaintiffs (not counting unnamed members of the class certified by the district court) are white males.35 There is no dispute as to their qualifications. The City Defendants gave the Offerees retroactive seniority at the expense of the Brennan Plaintiffs.36 And the 1999 settlement agreement was explicitly race-and sex-based, thereby giving rise to the required inference of discrimination. Cf. Ricci, 129 S.Ct. at 2674 (“Whatever the City's ultimate aim—however well intentioned or benevolent it might have seemed—the City made its employment decision because of race.”).
In sum, because the City Defendants' actions were explicitly race- and sex-based—i.e., the grants of retroactive seniority were made because the three disputed exams and the recruiting process had resulted in racial and gender disparities among test-takers and test-passers—“[o]ur analysis begins with this premise: The [City Defendants'] actions would violate the disparate-treatment prohibition of Title VII absent some valid defense․ Without some other justification, this express, race-[and gender-]based decisionmaking violates Title VII's command that employers cannot take adverse employment actions because of an individual's race [or sex].” Ricci, 129 S.Ct. at 2673.37
Taken together, the briefs of the Caldero and Arroyo Intervenors and the Government raise two defenses on behalf of the City Defendants. The first defense, raised only on behalf of the Caldero and Arroyo Intervenors, is that the City Defendants' voluntary implementation of the settlement agreement was a valid affirmative action plan. The second, raised or suggested in somewhat different ways as to various specific Offerees by all three of these parties, is the “strong basis in evidence” defense recognized in Ricci. There is no dispute that to the extent that the City Defendants' actions are not justified by one of these defenses, the City Defendants violated the Brennan Plaintiffs' § 703(a) rights to be free from disparate treatment. We will consider each defense in turn.
IV. Affirmative Action
The Caldero and Arroyo Intervenors argue that the retroactive seniority awards were justified as part of an affirmative action plan, valid under Supreme Court precedent. They contend that, even if non-remedial, the seniority awards therefore do not violate § 703(a). Thus, according to these intervenors, even if, at the time the City Defendants implemented the settlement agreement, there was no reason to think that the individual recipients of retroactive seniority were victims of discrimination, such seniority would still be valid. The district court agreed in part, and held that, except for the retroactive awards of layoff seniority, the retroactive seniority awards constituted permissible affirmative action. We, instead, hold that the City Defendants' implementation of the settlement agreement was not affirmative action at all, let alone permissible affirmative action; and that it was, therefore, error for the district court to apply such an “affirmative action” defense to the Brennan Plaintiffs' claims.
A. Legal Background
The Supreme Court first recognized that a valid affirmative action plan constituted a defense to a § 703(a) reverse-discrimination lawsuit in United Steelworkers of America v. Weber, 443 U.S. 193, 99 S.Ct. 2721, 61 L.Ed.2d 480 (1979). In Weber, a union and an employer collectively bargained for an affirmative action plan under which 50% of the openings in an in-plant craft training program would be reserved for black employees, until the percentage of black craft workers in the plant reached a level commensurate with the percentage of blacks in the local labor force. A white production worker, who had not been selected for the training program, brought suit under § 703(a) of Title VII, arguing that he had been denied entry into the training program because of his race. The Supreme Court held that “Title VII's prohibition in § [ ] 703(a) ․ against racial discrimination does not condemn all private, voluntary, race-conscious affirmative action plans.” Id. at 208. The Court did not “define in detail the line of demarcation between permissible and impermissible affirmative action plans.” Id. But the Court noted that the plan was “structured to open employment opportunities for Negroes in occupations which have been traditionally closed to them,” and that it did not “unnecessarily trammel the interests of the white employees” because it was “a temporary measure” that did not “require the discharge of white workers and their replacement with new black hirees” or “create an absolute bar to the advancement of white employees.” Id. The Court therefore held that the affirmative action plan in Weber “f[ell] within the area of discretion left by Title VII to the private sector voluntarily to adopt affirmative action plans designed to eliminate conspicuous racial imbalance in traditionally segregated job categories.” Id. at 209.
The Court reaffirmed its Weber holding, and elaborated upon the requirements for a valid affirmative action plan, in Johnson v. Transportation Agency, Santa Clara County, 480 U.S. 616, 107 S.Ct. 1442, 94 L.Ed.2d 615 (1987). In Johnson the defendant employer voluntarily adopted an affirmative action plan for promotions of employees, and pursuant to this plan it promoted a female employee over the male plaintiff. The plan took sex into account as one of several factors, including the qualifications of applicants for promotions. The Supreme Court upheld this plan against the male plaintiff's § 703(a) attack because it satisfied two requirements. First, the plan was prompted by a “manifest imbalance that reflected underrepresentation of women in traditionally segregated job categories.” Id. at 631 (quotations omitted). “A manifest imbalance need not be such that it would support a prima facie case against the employer,” id. 13 at 632, although just how far below the prima facie level a “manifest imbalance” could go was left unclear. In any event, there was a “manifest imbalance” in Johnson, because, among the positions at issue, “none of the 238 positions was occupied by a woman,” and this fact was properly viewed as an imbalance even though “in some job categories women were not qualified in numbers comparable to their representation in the labor force.” Id. at 636. Second, the plan did not “unnecessarily trammel[ ] the rights of male employees or create[ ] an absolute bar to their advancement.” Id. at 637–38. This was so because the plan imposed “goals” but not “quotas,” the male plaintiff “had no absolute entitlement to the ․ position” and retained his current position and title along with the ability to apply for future promotions, and “the [defendant]'s Plan was intended to attain a balanced work force, not to maintain one.” Id. at 637–40.
B. Application of Johnson and Weber to the Settlement Agreement
In the instant case, the district court held, and the Caldero and Arroyo Intervenors urge, that, except for purposes of layoff seniority,38 the implementation of the settlement agreement satisfied Johnson and Weber because, they say, it was spurred by a manifest imbalance and it did not unnecessarily trammel the rights of the Brennan Plaintiffs. See NYC Board III, 448 F.Supp.2d at 423–34. In light of Ricci, we agree with the Brennan Plaintiffs that the district court's holding skipped a threshold step in the analysis. That is, to determine whether a voluntary, private, race- and sex-conscious employer action is eligible for the Johnson/Weber defense, courts must now ask whether the race- and sex-conscious action constitutes an affirmative action plan at all. In this case, we answer that question in the negative.39
The Supreme Court's recent decision in Ricci indicates that not all voluntary race- or gender-conscious employer action is properly analyzed under Weber and Johnson. In Ricci, the City of New Haven administered a promotional examination for its firefighters. The results yielded a significant racial disparity: according to the promotion rules and by virtue of the test scores, all ten individuals eligible for the then-existing lieutenant vacancies were white; and of the nine individuals eligible for the then-existing captain vacancies, two were Hispanic and seven were white. See 129 S.Ct. at 2666. Because of this racial disparity and, allegedly, because of fear of a disparate-impact lawsuit from black firefighters, the New Haven Civil Service Board did not certify the test results. Id. at 2667–71. A group of white firefighters then sued under § 703(a) for disparate treatment.
The Supreme Court, correctly, described New Haven's decision as a “race-based action.” Id. at 2674. But, even though the Court was addressing a § 703(a) reverse-discrimination suit attacking a voluntary, private “race-based action,” the majority opinion did not cite Weber or Johnson. Nor did the Court apply the “manifest imbalance” and “unnecessary trammeling” factors. Instead, the Court adopted a “strong basis in evidence” standard: “under Title VII, before an employer can engage in intentional discrimination for the asserted purpose of avoiding or remedying an unintentional disparate impact, the employer must have a strong basis in evidence to believe it will be subject to disparate-impact liability if it fails to take the race-conscious, discriminatory action.” Id. at 2677. Ricci thus makes clear that at least some race- or sex-conscious voluntary employer actions are not subject to the “affirmative action” analysis of Weber and Johnson.40 See id. at 2700 (Ginsburg, J., dissenting) (“This litigation does not involve affirmative action.”).
Before Ricci, this Court had held, relying on Weber, that “a showing of a prima facie case of employment discrimination through a statistical demonstration of disproportionate racial impact constitutes a sufficiently serious claim of discrimination to serve as a predicate for employer-initiated, voluntary race-conscious remedies.” Bushey v. N.Y. State Civil Serv. Comm'n, 733 F.2d 220, 228 (2d Cir.1984) (footnote omitted). Based on that reasoning, we upheld, against a § 703 reverse-discrimination challenge, a public employer's adjustment of promotional examination scores for prison employees, in order to avoid a disparate impact on minority officers. Id. at 227–28. After Ricci, however, the position taken in Bushey is no longer tenable. A prima facie case of disparate impact is not, post-Ricci, an adequate factual predicate for all types of race- or gender-conscious employer actions. At least some race- or gender-conscious actions, such as the rejection of test scores because of their racial distribution, are not affirmative action and therefore cannot be supported by a mere “manifest imbalance,” or even by a prima facie case of disparate impact.
2. Is the Implementation of the Settlement Agreement an Affirmative Action Plan?
Since, after Ricci, not all race- or sex-conscious voluntary employer actions constitute “affirmative action” that might be properly analyzed under Johnson and Weber, we must decide what is an affirmative action plan and what is not. We agree with the Brennan Plaintiffs that in order to be an affirmative action plan, an employer action must benefit all members of a protected class. Although the plan may call for individualized determinations, the plan itself cannot be individualized. Under this definition, the City Defendants' voluntary implementation of the settlement agreement is not an affirmative action plan. The City Defendants, and the intervenors who support them, therefore cannot rely on their affirmative action defense.
a. What Is an Affirmative Action Plan?
The Supreme Court has never said what, for purposes of the Weber/Johnson defense to § 703(a), is an affirmative action plan. But the Court has differentiated affirmative action from other forms of relief a court might order under § 706(g) of Title VII, 42 U.S.C. § 2000e–5(g). We think that distinction is applicable here.
Before delving into those cases, we pause to note that, in adopting for § 703(a) the definition of affirmative action found in the Supreme Court's § 706(g) cases, we do not mean to suggest that the authority of a court to order an employer to do something (as limited by § 706(g)) is coterminous with an employer's power to do that same thing voluntarily without violating § 703(a). The Supreme Court has expressly rejected that proposition, for the “suggestion that employers should be able to do no more voluntarily than courts can order as remedies ignores the fundamental difference between volitional private behavior and the exercise of coercion by the State.” Johnson, 480 U.S. at 630 n .8 (citation omitted). We rely on the § 706(g) cases only for their definition of what affirmative action is, and not for their discussion of when affirmative action is permissible.
Firefighters Local Union No. 1784 v. Stotts, 467 U.S. 561, 104 S.Ct. 2576, 81 L.Ed.2d 483 (1984) is the earliest § 706(g) case relevant to a definition of affirmative action. In Stotts, black firefighters in Memphis, Tennessee, brought a § 703(a) pattern-or-practice racial discrimination claim against the Memphis Fire Department. After settlement negotiations, a consent decree was entered and approved by the district court. Among other things, the consent decree provided a hiring goal for black firefighters. The next year, because of budget difficulties, Memphis decided to lay some employees off. Under a contract between the city and the Firefighters' Union, layoffs were to follow a “last hired, first fired” rule. The result of following that rule would have been to lay off a significant proportion of black firefighters, thereby undoing much of what the hiring goal had accomplished. Upon the black firefighters' request, the district court issued an injunction to prevent the black firefighters from being laid off. The Court of Appeals affirmed. In order to comply with this injunction, the city laid off or demoted several white firefighters instead, even though they had more seniority than the black firefighters protected by the injunction.
Of the several arguments made in favor of the Stotts district court's injunction, one in particular is relevant here. The Court of Appeals in Stotts reasoned that (1) if the black firefighters had shown a Title VII violation, the district court could, under § 706(g), have ordered the city not to lay off black firefighters, and (2) anything that a district court can order under § 706(g) when a Title VII violation has been proven, the court can also order to effectuate the purpose of a Title VII consent decree. Id. at 578. The Supreme Court disagreed with the first of these premises, and held that “the Court of Appeals imposed on the parties as an adjunct of settlement something that could not have been ordered had the case gone to trial and the plaintiffs proved that a pattern or practice of discrimination existed.” Id. at 579. A court could not have ordered such relief, the Court explained, because “a court can award competitive seniority only when the beneficiary of the award has actually been a victim of illegal discrimination.” Id. But there had been “no finding that any of the blacks protected from layoff had been a victim of discrimination and no award of competitive seniority to any of them” at the time of the consent decree. Id. Stotts did, however, expressly avoid deciding “[w]hether the City, a public employer, could have taken this course without violating the law,” if Memphis had acted “unilaterally,” that is, voluntarily and not under court order. Id. at 583.41
The next significant case is Local 28, Sheet Metal Workers' Int'l Ass'n v. EEOC, 478 U.S. 421, 106 S.Ct. 3019, 92 L.Ed.2d 344 (1986). In Local 28, the Government successfully sued a union with a long and egregious history of discriminating against blacks and Hispanics. The union was found to have engaged in a pattern or practice of discrimination in recruitment, selection, training, and admission to the union. At the § 706(g) remedy stage, the district court established a 29% nonwhite membership goal, ordered the union to adopt an affirmative action plan for reaching this goal, and appointed an administrator to oversee the plan. The plan required the union
to offer annual, nondiscriminatory journeyman and apprentice examinations, select members according to a white-non-white ratio to be negotiated by the parties, conduct extensive recruitment and publicity campaigns aimed at minorities, secure the administrator's consent before issuing temporary work permits, and maintain detailed membership records, including separate records for whites and non-whites.
Id. at 432–33. After two appeals to our court, which resulted in only two changes to the plan—a minor modification to the white-nonwhite ratio, and an extra year for the union to meet the goal—the union had only reached 10.8% nonwhite membership. Id. at 434. State and local authorities moved in the district court to hold the union in contempt.
The district court held the union in contempt, not simply because the union had not met the goal, but also because the union had failed to comply with the requirements of the underlying affirmative action plan, and had withheld information from the administrator and the court, thereby making monitoring the union's compliance difficult. Id. at 434–35. The district court issued another contempt order after another year of noncompliance. Id. at 435–36. The court at that time established a slightly different minority membership goal, and abolished the apprenticeship examination, which had been a vehicle for noncompliance. (The examination was replaced with selection by a three-member board, which would select white and minority apprentices at a one-to-one ratio.) A divided panel of our court affirmed in all respects relevant here, citing “the union's foot-dragging egregious noncompliance” with the plan. Id. at 438 (citing EEOC v. Local 638, Local 28 of Sheet Metal Workers' Int'l Ass'n, 753 F.2d 1172, 1183 (2d Cir.1985)).
At the Supreme Court, a four-Justice plurality upheld the affirmative action plan against the union's claim that § 706(g) permits a district court to award preferential relief only to the actual victims of unlawful discrimination. The plurality held that “ § 706(g) does not prohibit a court from ordering, in appropriate circumstances, affirmative race-conscious relief as a remedy for past discrimination․ [S]uch relief may be appropriate where an employer or a labor union has engaged in persistent or egregious discrimination, or where necessary to dissipate the lingering effects of pervasive discrimination.” Id. at 445 (plurality opinion). The plurality distinguished Stotts thus:
Stotts discussed the “policy” behind § 706(g) in order to supplement the holding that the District Court could not have interfered with the city's seniority system in fashioning a Title VII remedy. This “policy” was read to prohibit a court from awarding make-whole relief, such as competitive seniority, backpay, or promotion, to individuals who were denied employment opportunities for reasons unrelated to discrimination. The District Court's injunction was considered to be inconsistent with this “policy” because it was tantamount to an award of make-whole relief (in the form of competitive seniority) to individual black firefighters who had not shown that the proposed layoffs were motivated by racial discrimination. However, this limitation on individual make-whole relief does not affect a court's authority to order race-conscious affirmative action. The purpose of affirmative action is not to make identified victims whole, but rather to dismantle prior patterns of employment discrimination and to prevent discrimination in the future. Such relief is provided to the class as a whole rather than to individual members; no individual is entitled to relief, and beneficiaries need not show that they were themselves victims of discrimination. In this case, neither the membership goal nor the Fund order required petitioners to indenture or train particular individuals, and neither required them to admit to membership individuals who were refused admission for reasons unrelated to discrimination.
Id. at 474 (citation and footnotes omitted).42
The Local 28 distinction between affirmative action and make-whole relief, we think, makes equal sense in the § 703(a) context, for it harmonizes Ricci with Johnson and Weber. In Ricci, a § 703(a) case, the employer action that was challenged was the discarding of the results of a test that had already been administered. This action was individualized, for what it did, in essence, was to give promotion—or at least another chance at promotion—to the individual black firefighters who had taken the test, at the expense of those firefighters who would have been eligible for promotion if the test results had been certified. As the Supreme Court explained:
[W]e [do not] question an employer's affirmative efforts to ensure that all groups have a fair opportunity to apply for promotions and to participate in the process by which promotions will be made. But once that process has been established and employers have made clear their selection criteria, they may not then invalidate the test results, thus upsetting an employee's legitimate expectation not to be judged on the basis of race. Doing so, absent a strong basis in evidence of an impermissible disparate impact, ․ is antithetical to the notion of a workplace where individuals are guaranteed equal opportunity regardless of race.
129 S.Ct. at 2677. In other words, when an employer, acting ex ante, although in the light of past discrimination, establishes hiring or promotion procedures designed to promote equal opportunity and eradicate future discrimination, that may constitute an affirmative action plan. But where an employer, already having established its procedures in a certain way—such as through a seniority system—throws out the results of those procedures ex post because of the racial or gender composition of those results, that constitutes an individualized grant of employment benefits which must be individually justified, and not affirmative action.
The affirmative action plans in Johnson and Weber were quite different from such make-whole relief. The plan in Weber set out to achieve a better future racial balance among skilled craftworkers at Kaiser Steel's Gramercy plant, by requiring that 50% of production workers chosen for the skilled craftworker training program be black. See 443 U.S. at 199. This plan was adopted pursuant to an affirmative-action provision in the collective bargaining agreement recently negotiated between the United Steelworkers and Kaiser. Id. at 197–98. The plaintiff, Brian Weber, argued that the plan violated § 703(a) and (d) because he was denied entry into the training program, in favor of less senior black workers. But the Supreme Court treated the plan in Weber as affirmative action. That result is readily explained by the Local 28 distinction. Instead of, say, granting retroactive seniority to specific individual black production workers, or throwing out the results of previous selection processes, the plan benefited all members of the racially defined class in a forward-looking manner. And it was adopted in a newly negotiated collective bargaining agreement rather than unilaterally by the employer in derogation of an earlier agreement.
The plan in Johnson, likewise, did not grant individualized employment benefits to any specific women or racial minorities. The employer in Johnson, noting the underrepresentation of women in certain job classifications, decided to authorize the consideration of sex as one of several factors in deciding which of several qualified applicants to promote. 480 U.S. at 620–21. Although this plan ultimately resulted in the promotion of a woman over a man who was otherwise slightly better qualified, see id. at 623–24, the wheels were set in motion ex ante. The employer decided on a plan that benefited the entire class of women, and then it simply applied that plan in a particular instance. By way of contrast, the employer had not previously adopted a different, gender-neutral method of selecting employees for promotion and then opted to throw it out when confronted with the reality that a particular woman would not be promoted under that method.
The Arroyo and Caldero Intervenors contend, however, that the line between “affirmative action” governed by Weber and Johnson on the one hand, and race- or gender-conscious action taken “for the asserted purpose of avoiding or remedying an unintentional disparate impact,” Ricci, 129 S.Ct. at 2677, on the other, must be drawn quite differently. We find their attempts to circumscribe Ricci to be without merit.
The Arroyo Intervenors first suggest that Ricci essentially be limited to its facts. They say that Ricci applies only where an employer is motivated by “[f]ear of litigation alone,” id. at 2681, and that “Ricci's new legal standard has no application to a case like this one involving an employer's well-informed decision, after years of defending against employment discrimination claims, to enter into a settlement to redress ongoing and pervasive racial exclusion in a particular class of jobs.” Arroyo Reply Br. at 5. That argument confuses the Ricci strong-basis-in-evidence standard with the threshold question of whether Ricci applies. In Ricci, the Supreme Court said that an employer motivated by “fear of litigation alone” does not have an adequate strong basis in evidence. The Court did not suggest that the strong-basis-in-evidence standard applies only when an employer is motivated by “fear of litigation alone.” Indeed, it would make no sense to require a strong basis in evidence of only those employers who take race- or gender-conscious actions that are not “well-informed.”
Second, both the Arroyo and Caldero Intervenors assert that Ricci is limited to situations in which an employer, acting to avoid a disparate-impact Title VII violation that has not yet occurred, discards the results of a promotional exam that would otherwise result in the awarding of particular promotions to particular candidates. Arroyo Reply Br. at 9–12; Caldero Reply Br. at 64–66. That is, they seek to place at least three limitations on the scope of Ricci: (1) Ricci applies only when the employer seeks to avoid a current disparate-impact violation, while remedies for previous violations are subject to Weber and Johnson; (2) Ricci applies only when the employment benefit denied to the plaintiffs is hiring or promotion; and (3) Ricci applies only when, in the absence of the employer's race- or gender-conscious action, specific plaintiffs would have been entitled to specific positions. None of these narrow readings of Ricci seems to us valid.
Ricci is not limited to cases where the employer seeks to avoid a current violation. As the Ricci Court itself stated, its core holding applies whenever an employer takes race-conscious action “for the asserted purpose of avoiding or remedying an unintentional disparate impact.” 129 S.Ct. at 2677 (emphasis added). And limiting Ricci to the avoidance of current violations would make very little sense. It would be strange to make an employer be subject to Ricci when it promises a set of employment benefits to test-passers but then unilaterally decides not to give the benefits out, while at the same time allowing that employer to avail itself of the more easily satisfied Johnson/Weber standard when it gives out another set of benefits, such as the seniority rights that will lead to the aforementioned benefits. Ricci also is not solely about hiring or promotion. Nothing in the case suggests that when other past established employment benefits are involved employers can take individualized race- or gender-conscious action to remedy an asserted disparate impact without meeting the “strong basis in evidence” test. See id. at 2676 (“[W]e adopt the strong-basis-in-evidence standard as a matter of statutory construction to resolve any conflict between the disparate-treatment and disparate-impact provisions of Title VII.” (emphasis added)). For similar reasons, Ricci does not require that specific individuals be entitled to specific positions. When an employer lacks a strong basis in evidence that it would otherwise be liable for disparate impact, the employer should not “upset[ ] an employee's legitimate expectation not to be judged on the basis of race.” Id. at 2677. That expectation is upset by the race-conscious discarding of conditional entitlements to employment or employment benefits, just as much as it is when the entitlement is absolute.
We therefore hold that § 703(a), like § 706(g), draws a distinction between affirmative action plans, which are intended to provide ex ante benefits to all members of a racial or gender class, and make-whole relief, which is intended to provide ex post benefits to specified individuals who have suffered discrimination. And where this latter form of benefits is at issue, the employer may not invoke the “affirmative action” defense of Johnson and Weber.
b. The Employer Action in This Case
Having held that the Johnson/Weber framework does not apply to make-whole relief, we now apply this rule to the City Defendants' voluntary implementation of the disputed portions of the settlement agreement. In our view, the City Defendants' voluntary implementation of the settlement agreement falls on the Stotts side of the Local 28 distinction between affirmative action and make-whole relief. The City Defendants' implementation of the disputed portions of the settlement agreement granted competitive seniority to the Offerees. Competitive seniority is one of the forms of make-whole relief specifically enumerated in Local 28. Moreover, it was competitive seniority that was, in effect, at issue in Stotts, for when Memphis laid off white firefighters instead of black firefighters with less seniority, the city was treating the white firefighters as if they were the less senior employees.
And, perhaps even more importantly, the reason why grants of competitive seniority are generally not affirmative action, and therefore are generally limited to make-whole relief, is that retroactive seniority is by its nature individualized. The Offerees were a group of individuals defined by the settlement agreement to include provisional Custodians and CEs who were women or members of certain racial minority groups. Most of them were identified by name in an appendix to the agreement. It cannot be said that the City Defendants “provided [employment benefits] to the class as a whole rather than to individual members.” Local 28, 478 U.S. at 474. Instead, what the City Defendants did was “tantamount to an award of make-whole relief,” id., to individual provisional Custodians and CEs, arguably without adequate regard to whether there was any reason to think they might have been actual victims of discrimination, or to whether there was a strong basis in evidence that a disparate impact suit by them would succeed.
The Caldero Intervenors urge that the retroactive seniority awards in this case are affirmative action even though “they do not look like the most typical kind.” They say the awards are a reasonable, narrower substitute for a general race or gender preference for school transfers. Their view seems to be that if a broad, non-individualized racial preference is permissible provided that there is a manifest imbalance and the plan is narrowly tailored, then race- or gender-conscious relief for only some of the individuals who would have benefited from such a policy, had it been there in the past, must also be permissible. We do not agree. Just as, in the § 706(g) context, courts have the power to order general racial preferences (as in Local 28 ), but not individualized non-remedial relief (as in Stotts ), so, employers also cannot undertake individualized non-remedial relief and validate it by calling it “affirmative action.”
The cases cited by the Caldero Intervenors are not to the contrary. In Jana–Rock Constr., Inc. v. N.Y. State Dep't of Econ. Dev., 438 F.3d 195 (2d Cir.2006), an equal protection case, a construction company owned by an individual of Spanish descent claimed that New York's definition of “Hispanic,” which was used for purposes of selecting minority-owned businesses to benefit from affirmative action, violated the Equal Protection Clause because it included individuals of Latin American descent while excluding Spaniards. We rejected this challenge, holding that the Equal Protection Clause does not require a state using an affirmative action plan to “expand[ ]” the preferred class “to include other racial and ethnic groups that may have been discriminated against.” Id. at 207. Assuming arguendo that this equal protection reasoning applies in the § 703(a) context, Jana–Rock is distinguishable because it deals with the definition of a racial or ethnic beneficiary class. In our case, nobody is saying that the numerous other minority or female individuals who may have applied for Custodian or CE positions—or, for that matter, the minority or female incumbent Custodians and CEs like Miranda plaintiff Ruben Miranda—are not black, Hispanic, Asian, or female. They were excluded from the settlement agreement's benefits, not because they were deemed to be the “wrong” gender or race, but rather because they were not among the individuals selected for individualized relief. That there may be some flexibility in how a racial or gender class is defined does not undermine the requirement that affirmative action plans provide, on a prospective basis, equal benefits to all members of that beneficiary class.
The Caldero Intervenors also rely on Stuart v. Roache, 951 F.2d 446 (1 st Cir.1991) (Breyer, J.), in which the First Circuit upheld under the Equal Protection Clause a race-conscious promotion plan for Boston police officers. The promotion goals for black officers fell short of the projected number of such officers eligible for promotion: they amounted to 15.5% of all sergeants, while the eligible pool contained about 20% black officers. Id. at 454. But, like a differently defined racial or gender class, a smaller promotion goal does not turn affirmative action into individualized relief. A 15.5% goal and a 20% goal are both “provided to the class as a whole rather than to individual members .” Local 28, 478 U.S. at 474. The retroactive seniority awards at issue in our case are, instead, expressly “provided ․ to individual members.”
It is noteworthy, as well, that the Government claims it never intended a Stotts-like result when it entered the settlement agreement. The district court held a hearing to determine the intent of the settlement agreement, and at that hearing Norma Cote, the lawyer who had negotiated the agreement for the City Defendants, testified that she did not recall the Government “ever explain [ing] to [her] why they wanted these individuals to get retroactive seniority.” NYC Board V, 556 F.Supp.2d at 205–06. Katherine Baldwin, a DOJ supervisor who was not directly involved in the settlement negotiations but who approved the settlement after it was negotiated, testified that it was DOJ policy to seek only make-whole relief for actual victims, and that she would not have approved the settlement if it had contravened that policy. Id. at 206. The district court found, on the basis of this testimony, that the settlement agreement was intended to be limited to make-whole relief. Although the awards may have gone beyond make-whole relief, we are skeptical as to whether an employer can adopt an affirmative-action plan by accident. When an employer adopts an affirmative-action plan, it generally does so consciously, with the purpose “not to make identified victims whole, but rather to dismantle prior patterns of employment discrimination and to prevent discrimination in the future,” Local 28, 478 U.S. at 474. When applying strict or heightened scrutiny to race- or gender-based classifications in the Equal Protection Clause context, the Supreme Court has cautioned that “[t]he [actor's] justification must be genuine, not hypothesized or invented post hoc in response to litigation.” United States v. Virginia, 518 U.S. 515, 533, 116 S.Ct. 2264, 135 L.Ed.2d 735 (1996) (heightened scrutiny); accord Shaw v. Hunt, 517 U.S. 899, 908 n. 4, 116 S.Ct. 1894, 135 L.Ed.2d 207 (1996) (strict scrutiny). The same, we believe, is true when an employer raises the Weber/Johnson “affirmative action” defense to a reverse-discrimination suit under § 703(a).
The Caldero Intervenors, relying on the testimony recited in footnote 43, infra, and the text of the settlement agreement itself, assert that, notwithstanding the district court's finding to the contrary, the parties to the settlement agreement clearly intended to create an affirmative action plan in 1999. According to these Intervenors, it was manifest in 1999 that the settlement agreement would provide retroactive seniority to the specified Offerees without regard to whether there was any evidence that they were actual victims. Therefore, they say, the Government and the City Defendants must have intended to create an affirmative action plan. The Caldero Intervenors' premise is correct, but their conclusion does not follow.
The Caldero Intervenors correctly point out that the settlement agreement contained a specific definition of “Offeree,” and that this definition was not limited to actual victims. They also correctly point out that the Cote testimony, which the district court credited, indicates that the Government and the City Defendants did not perform any individualized investigation of the Offerees to determine the likelihood that they were victims of discrimination. Once it was determined that an individual satisfied the definition of “Offeree” in the settlement agreement, that individual was offered retroactive seniority.43 Yet all that this shows is that the Government and the City Defendants may have been less careful than they should have been. Although it was clear from the four corners of the settlement agreement and from the testimony that nobody was checking to make sure that the Offerees were actual victims or even likely victims, it was equally clear that the retroactive seniority awards would be individualized, and therefore, in light of Stotts, not affirmative action. Not every race- or gender-conscious employment benefit that goes beyond make-whole relief is affirmative action; some such benefits, like those at issue here, fall into the impermissible Stotts category instead.
Finally, the City Defendants' race- and gender-conscious actions are a poor fit for the wrongs they seek to redress. True affirmative action, like that undertaken voluntarily by the employers in Johnson and Weber and ordered by the court in Local 28, has the power to break down old patterns of discrimination and prevent future discrimination against minorities and women. But affirmative action also has its costs. Most significantly, it is race- or gender-conscious, and it puts non-minority and male individuals at a disadvantage. See Ass'n Against Discrimination in Employment, Inc. v. City of Bridgeport, 647 F.2d 256, 280–81 (2d Cir.1981) (“Balanced against the broad equitable power to remedy Title VII violations is a recognition that ‘the use of racial goals means, in practice, that certain nonminority persons will be kept out solely on account of their race or ethnic background’ and that this impinges on the basic principle ‘that individuals are to be judged as individuals, not as members of particular racial groups.’ “ (quoting EEOC v. Local 638, Local 28 of Sheet Metal Workers Int'l Ass'n, 532 F.2d 821, 827 (2d Cir.1976)) (modifications omitted)). Make-whole relief, likewise, is a powerful remedy for past wrongs. By putting the victims of discrimination where they would have been, but for the discrimination, make-whole relief not only undoes much of the harm caused to the victims themselves, but also provides examples so that others know that they, too, can overcome this country's history of discrimination in the workplace. See Franks v. Bowman Transp. Co., 424 U.S. 747, 763–68 & n. 28, 96 S.Ct. 1251, 47 L.Ed.2d 444 (1976). But make-whole relief is not free either. Someone has to pay for it—most often the non-minority and male employees, as well as the employer.
Because of the costs inherent in both affirmative action and make-whole relief, anyone attempting to provide either of these forms of relief—whether it be a court imposing a remedy under § 706(g) after a Title VII violation has been found, or the Government proposing a settlement to an employer, or an employer acting voluntarily—must be exceptionally careful to ensure that the employer's proposed action is properly tailored to achieve whichever of these two types of remedies for discrimination is sought. That is not to say that one cannot use affirmative action and make-whole relief at the same time; of course one can. See, e.g., Local 28, 478 U.S. at 473 n. 44 (noting that, along with imposing the affirmative action plan, the district court also gave backpay to specified individual victims of discrimination); cf. Ass'n Against Discrimination in Employment, 647 F.2d at 278 (noting that these “two categories [of relief] may overlap to some extent, although their intended functions differ”). But a court or employer planning to give out a race- or gender-conscious employment benefit, or a Title VII remedy, should always ask first: What is the purpose of what I am doing? (1) Am I trying to give make-whole relief to individual people who I think are victims of past discrimination, (2) am I trying to implement a non-individualized, class-wide affirmative action plan to dismantle prior patterns of discrimination and prevent future discrimination, or (3) am I trying to do both? Only after these questions have been answered can an appropriately tailored plan or plans be put in place. The Government and the City Defendants should have considered these questions before taking action. Instead, according to the testimony that the district court credited, the Government never explained why it wanted the Offerees to get retroactive seniority, and the City Defendants never asked.44
The necessity of avoiding Stotts-like “remedies,”—that is, non-remedial, individualized, race- or sex-conscious employment benefits—is not merely an abstract, doctrinal matter. It affects real people in tangible ways. In the instant case, Ruben Miranda, the plaintiff in Miranda, may be the clearest example of how the City Defendants' remedy does not fit. He is a Hispanic male. He studied for, took, and passed Exam 5040, which allegedly discriminated against Hispanics. As a reward for overcoming this hurdle, he was hired as a Custodian—only to be told, some years later, that other, newly appointed Custodians, a specified group of women, blacks, Hispanics, and Asians, would be put ahead of him in seniority. As a result, he could be laid off before these newcomers would be, and he would—and, as it turns out, did—lose school transfers to them. Thus, a Hispanic Custodian was hampered in attaining his career goals, including a transfer to a larger and higher-paying school, because of a settlement agreement that the Caldero and Arroyo Intervenors insist is an affirmative action plan for, among others, Hispanics.45
For all these reasons, while some or all of the retroactive seniority awards may be defensible on some other ground, they are not defensible as part of an affirmative action plan. The City Defendants therefore cannot use the Weber/Johnson defense to the Brennan Plaintiffs' § 703(a) prima facie case.
V. Strong Basis in Evidence
The Brennan Plaintiffs have made out a prima facie case of disparate treatment under § 703(a) of Title VII, and we have rejected the first defense relied on by the City Defendants, that of “affirmative action” under Johnson and Weber. We next consider the City Defendants' second defense: that the retroactive seniority awards to the Offerees were intended as make-whole relief to victims of the City Defendants' previous testing and recruiting practices that had a disparate racial and gender impact.
We hold that this defense—remedying disparate impact—is governed by Ricci. “[B]efore an employer can engage in intentional discrimination for the asserted purpose of avoiding or remedying an unintentional disparate impact, the employer must have a strong basis in evidence to believe it will be subject to disparate-impact liability if it fails to take the race-conscious, discriminatory action.” 129 S.Ct. at 2677. The City Defendants' action was race- and gender-conscious, and it was taken for the asserted purpose of remedying an unintentional disparate impact. It is therefore subject to Ricci.46
Our reading of Ricci is that a strong basis in evidence that an employer will be subject to disparate-impact liability requires that, at the time it takes the “race-conscious, discriminatory action,” the employer be faced with both (1) a prima facie case of disparate impact against itself (or perhaps a strong basis in evidence of a prima facie case), and (2) a strong basis in evidence either (a) that the employment practice having the disparate impact was neither job-related nor consistent with business necessity, or (b) that there was an equally valid, less discriminatory alternative, that the employer had refused to adopt, that would have served the employer's needs. As detailed in Part V.A, infra, we hold that, under Ricci, a “strong basis in evidence” of non-job-relatedness or of a less discriminatory alternative requires more than speculation, more than a few scattered statements in the record, and more than a mere fear of litigation, but less than the preponderance of the evidence that would be necessary for actual liability. This is what it means when courts say that the employer must have an objectively reasonable fear of disparate-impact liability.
The case before us also presents an additional problem not at issue in Ricci. Even if an employer has a strong basis in evidence that it faces liability for disparate impact, it may be unclear to whom the employer might face liability, or, if the disparate-impact case were brought by the Government, who might be entitled to court-ordered relief under § 706(g) and to what extent such individuals would receive relief. In the instant case, it was unclear at the time of the City Defendants' actions, and it remains unclear today, exactly who among the Offerees was a victim of any disparate-impact discrimination that might have occurred, and how much retroactive seniority was necessary to make such individuals whole. We hold that the strong-basis-in-evidence standard of Ricci applies to these questions too.
Thus, the employer must have a strong basis in evidence not only that it will actually be found liable for disparate impact, but also that, at the remedial stage following such a finding of liability, a court might well impose (pursuant to § 706(g)) a make-whole remedy equivalent to or broader than what the employer has done voluntarily. Put differently, when an employer, instead of litigating and having a court impose an equitable remedy, undertakes unilaterally the “process of recreating the past,” Int'l Broth. of Teamsters v. United States, 431 U.S. 324, 372, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977), the employer must have a strong basis in evidence that its race- or gender-conscious action, “as nearly as possible, ‘recreate[s] the conditions and relationships that would have been had there been no’ unlawful discrimination,” id. (quoting Franks v. Bowman Transp. Co., Inc., 424 U.S. 747, 769, 96 S.Ct. 1251, 47 L.Ed.2d 444 (1976)). As with the employer's fear of disparate-impact liability, the employer's reconstruction of the past must, that is, be objectively reasonable.
A. What Is a Strong Basis in Evidence?
It is not immediately obvious from Ricci what constitutes a “strong basis in evidence” in the Title VII disparate-impact context. The standard is new in this context, and the Supreme Court has not yet had occasion to describe its operations in detail, for in Ricci the court held that the standard had manifestly not been satisfied. See 129 S.Ct. at 2681 (“[T]here is no evidence—let alone the required strong basis in evidence—that the tests were flawed because they were not job-related or because other, equally valid and less discriminatory tests were available to the City.”). Our task, therefore, is to provide, as best we can, some guidance for the district court in applying the Ricci strong-basis-in-evidence test to the facts of this case.
1. Strong Basis in Evidence of Liability
Some of the general contours of the strong-basis-in-evidence standard are evident from Ricci. First, the standard is objective, not subjective, and it therefore focuses on the strength of the evidence of liability, not the strength of the employer's fear of litigation. See id. at 2675 (stating that “[a] mere good-faith fear of disparate-impact liability” is inadequate); id . at 2677 (referring to “an objective, strong basis in evidence”); id. at 2681 (“Fear of litigation alone cannot justify an employer's reliance on race․”).
Second, the strength of the evidence of disparate-impact liability is measured at the time the employer took the race- or gender-based action. In examining the evidence marshaled by New Haven, the Ricci Court considered only what the city knew at the time it made its decision, and not any further information that arose afterwards. See id. at 2678–81. Similarly, the Court generally referred to the absence of a strong basis in evidence in the past tense. E.g., id. at 2681 (“On the record before us, there is no genuine dispute that the City lacked a strong basis in evidence to believe it would face disparate-impact liability if it certified the examination results.”). The rationale underlying Ricci, moreover, confirms that the evidence is to be gauged at the time of the race- or sex-conscious employer action. The strong-basis-in-evidence standard is intended to “strike[ ] a ․ balance” between the Title VII provisions concerning disparate treatment and disparate impact, so that employers make the right decisions in the first place. Id. at 2675.
Thus, Ricci seeks to avoid both creating a legal framework under which “employers likely would hesitate before taking voluntary action for fear of later being proven wrong in the course of litigation and then held to account for disparate treatment,” and one under which employers would undertake “race[or gender-]based action at the slightest hint of disparate impact.” See id. at 2674–75. If evidence from after the employer's race- or gender-based decision were taken into account, there would be false negatives and false positives. Some employers might take inappropriate race- or gender-based actions in the hope or expectation that a strong basis in evidence would later emerge; other employers who actually do have a strong basis in evidence might refuse to take voluntary action for fear that later evidence would undermine that basis. Moreover, still other employers would be held liable for disparate treatment or for disparate impact even though based on the evidence of disparate impact then before them, they acted correctly at the time they made their decisions. All such results are inimical to Ricci.47
Third, either an actual prima facie case of disparate-impact liability is required, or a strong basis in evidence of a prima facie case is required. Ricci does not say which, as the City of New Haven was faced with an undisputed prima facie case of disparate impact. Id. at 2677–78. Nevertheless, because Ricci explicitly rejects the proposition that “an employer in fact must be in violation of the disparate-impact provision before it can use compliance as a defense in a disparate-treatment suit,” it seems likely that no more than a strong basis in evidence that a prima facie case exists would suffice.48 Id. at 2674. If an actual prima facie case were required, then an employer who used an employment practice that was obviously not job-related and who was faced with a strong basis in evidence, but not an actual prima facie case, of disparate impact, “likely would hesitate before taking voluntary action for fear of later being proven wrong in the course of litigation and then held to account for disparate treatment.” Id. But we decline to decide this issue here, leaving it to be determined by the district court in the first instance, if it proves to be necessary to deciding the case.
Fourth, because of the objective nature of the strong-basis-in-evidence test and its focus on the likelihood of actual liability, the test requires that the employer have a strong basis in evidence either (1) that its challenged employment procedures are not job-related, or (2) that there was a less discriminatory alternative procedure which the employer refused to adopt. These are, as the Ricci Court explained, the two conditions under which an employer can be liable for disparate impact after a plaintiff has shown a prima facie case of disparate impact. See Ricci, 129 S.Ct. at 2678.49
A somewhat more difficult question is just how strong the evidence of non-job-relatedness or a less discriminatory alternative must be. Two boundaries are clear from the Ricci Court's opinion. The evidence of liability certainly must be stronger than the evidence New Haven presented in Ricci. Summary judgment against the city could not there be avoided with what was described as “no evidence,” 129 S.Ct. at 2681, or, at most, “a few stray (and contradictory) statements in the record,” id. at 2680. At the other end, a strong basis in evidence that an employer will be liable for disparate impact must be less than what is required to prove a disparate-impact violation (and hence than what is needed to prove non-job-relatedness or the existence of a less discriminatory alternative). The strong-basis-in-evidence standard, Ricci says, “is not so restrictive that it allows employers to act only when there is a provable, actual violation.” Id. at 2676. The strong-basis-in-evidence test does not, therefore, require that there be a preponderance of the evidence of an actual disparate impact violation.
We think Ricci suggests that a strong basis in evidence is a balanced standard that falls somewhere in the middle between these upper and lower extremes. In borrowing the strong-basis-in-evidence standard from a line of Equal Protection cases, the Ricci Court stated that those cases “recognized the tension between eliminating segregation and discrimination on the one hand and doing away with all governmentally imposed discrimination based on race on the other .” Id. at 2675. Balancing those two goals requires “evidentiary support for the conclusion that remedial action is warranted,” id. (quoting Wygant v. Jackson Bd. of Educ., 476 U.S. 267, 277, 106 S.Ct. 1842, 90 L.Ed.2d 260 (1986) (plurality opinion)) (modification omitted), and not just “an amorphous claim that there has been past discrimination,” id. (quoting City of Richmond v. J.A. Croson Co., 488 U.S. 469, 499, 109 S.Ct. 706, 102 L.Ed.2d 854 (1989)).
Nevertheless, and notably, a “strong basis in evidence” for purposes of Ricci and Title VII is not necessarily the same as it is for Equal Protection Clause purposes. Id. at 2676. While in the equal protection context a strong basis in evidence has been described as “sufficient evidence to justify the conclusion that there has been prior discrimination,” Wygant, 476 U.S. at 277 (plurality opinion), Ricci, instead, explicitly does not require that an employer show that there has been a past disparate-impact violation of Title VII. Additionally, the Equal Protection Clause does not prohibit the government from taking actions which have an unintentional disparate impact, see Washington v. Davis, 426 U.S. 229, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976), and the type of evidence that supports a disparate-impact claim is different from that which would support a disparate-treatment claim.50 Finally, unlike the Equal Protection Clause, Title VII has repeatedly been construed so as not to undermine employers' ability to undertake “voluntary compliance,” which is “the preferred means of achieving the objectives of Title VII,” Ricci, 129 S.Ct. at 2674 (quoting Local No. 93, 478 U.S. at 515).
In sum, a strong basis in evidence of disparate-impact liability is an objectively reasonable basis to fear such liability. It is evaluated at the time an employer takes a race-conscious action. It relies on real evidence, not just subjective fear or speculation. Because it focuses on liability rather than mere litigation, it requires both objectively strong evidence of a prima facie case (or perhaps actual proof of a prima facie case ) of disparate impact, and objectively strong evidence of non-job-relatedness or a less discriminatory alternative.
2. Strong Basis in Evidence of Necessity
Even after an employer has shown a strong basis in evidence that it faces disparate-impact liability, the employer does not have carte blanche to take whatever race- or gender-conscious actions it pleases. Rather, “the employer must have a strong basis in evidence to believe it will be subject to disparate-impact liability if it fails to take the race-conscious [or gender-conscious], discriminatory action.” Id. at 2677 (emphasis added). That is to say, there must be a strong basis in evidence that the race- or gender-conscious action taken by the employer is necessary to avoid disparate-impact liability.
This necessity issue was not squarely presented in Ricci, for two reasons. First, there was no strong basis in evidence that New Haven would have faced disparate-impact liability, so it was unnecessary to determine what New Haven was permitted to do to remedy the disparate impact. Second, if New Haven had had a strong basis in evidence of disparate-impact liability, then there would have been little question that refusal to certify the test results was precisely what was necessary to avoid liability. There was, therefore, no need in Ricci to undertake the difficult process of determining who might have been a victim of discrimination and what sort of relief might have been required to make such individuals whole. Cf. Teamsters, 431 U.S. at 371–72. It seems easier to say what would be necessary to avoid a disparate-impact violation that is about to occur but has not yet happened, than it is to say what is necessary to remedy such a violation years after it took place.
We think it makes good sense to require that an employer's strong basis in evidence extend beyond the existence of disparate-impact liability, to the necessity of the employer's chosen race- or sex-conscious remedy for that disparate impact. As with the requirement of a strong basis in evidence of liability, a strong basis in evidence of necessity “gives effect to both the disparate-treatment and disparate-impact provisions, allowing violations of one in the name of compliance with the other only in certain, narrow circumstances.” Ricci, 129 S.Ct. at 2676. In doing so, requiring a strong basis in evidence of necessity avoids extreme positions that would undermine the careful Ricci balancing of disparate treatment and disparate impact under Title VII. If a showing of actual necessity—i.e., a showing that the race- or sex-conscious action's beneficiaries were actual victims who received make-whole relief—were mandated, employers would likely refuse to settle disparate-impact cases for fear of disparate-treatment liability. See Part V.B, infra; Ricci, 129 S.Ct. at 2674. But if anything less than a strong basis in evidence of necessity were required—for example, if “an employer's good-faith belief that its actions were necessary to comply with Title VII's disparate-impact provision” were “enough to justify race-conscious conduct,” then employers might give out race-conscious benefits “even where there is little if any evidence of disparate-impact discrimination” against the recipients of those benefits. See id. at 2674–75. We therefore hold that the strong-basis-in-evidence standard of Ricci applies not only to the question of disparate-impact liability, but also to the further question of whether the employer's race- or gender-conscious action is necessary to remedy that disparate impact. Here too, the employer's belief that its action is necessary to remedy disparate impact, i.e., that the beneficiaries of the action were victims of disparate impact and the action puts them roughly where they would have been in the absence of discrimination, must be objectively reasonable in the above defined sense.
B. The Government's “Actual Violation” Standard
The Government and the Brennan Plaintiffs ask us to go beyond the explicit requirements of Ricci and hold—either in all cases or, alternatively, in those cases where an employer's race- or gender-conscious action may implicate contractual seniority rights—that the employer can be liable even if its actions are supported by a strong basis in evidence that the actions are necessary to avoid or remedy a disparate-impact violation. According to them, a § 703(a) reverse-discrimination plaintiff can win by showing (a) that the employer's employment practices were not actually job-related and that there was not actually a less discriminatory alternative, or (b) that some or all recipients of race- or gender-conscious relief were not actual victims who received no more than make-whole relief. If the plaintiffs make this showing, they say, then the employer is liable to them for reverse discrimination. We disagree. The positions advocated by the Government and the Brennan Plaintiffs are irreconcilable with Ricci and rest on a confusion about the procedural posture of this case. Additionally, these parties' focus on contractual seniority rights is inappropriate because the remedy for an employer's unilateral and non-court-sanctioned breach of a collective bargaining agreement is to be sought (and obtained) by the union through the labor grievance and arbitration process, and not by means of a § 703(a) action brought by individual union members in federal court. As a result, the Ricci strong-basis-in-evidence standard is not altered in cases where the plaintiffs allege a breach of contract.
1. Ricci Does Not Require a Showing of Actual Liability or Actual Victims
The first, and most significant, problem with the arguments of the Government and the Brennan Plaintiffs is that they ask us to do something that Ricci explicitly tells us not to do:
Petitioners ․ suggest that an employer must be in violation of the disparate-impact provision before it can use compliance as a defense in a disparate-treatment suit․ [T]his is overly simplistic and too restrictive of Title VII's purpose. The rule petitioners offer would run counter to what we have recognized as Congress's intent that “voluntary compliance” be “the preferred means of achieving the objectives of Title VII.” Forbidding employers to act unless they know, with certainty, that a practice violates the disparate-impact provision would bring compliance efforts to a near standstill. Even in the limited situations when this restricted standard could be met, employers likely would hesitate before taking voluntary action for fear of later being proven wrong in the course of litigation and then held to account for disparate treatment.
Id. at 2674 (quoting Local No. 93, 478 U.S. at 515) (citations omitted). The Brennan Plaintiffs are asking us to give them the opportunity to prove, in the course of litigation, that the City Defendants were wrong when they deemed some Offerees likely victims of disparate-impact discrimination, and to hold the City Defendants liable for disparate treatment if any such error was made. But that is just what the Supreme Court has forbidden. The Brennan Plaintiffs are entitled to put those supporting the City Defendants to their proof that the City Defendants had a strong basis in evidence that the challenged testing and recruiting practices were not job-related or that there was a less discriminatory alternative to those practices, and that the City Defendants' actions were necessary to avoid the resulting disparate-impact liability as to particular Offerees—but that is all.51
True, the Brennan Plaintiffs and the Government would put the burden of proof on the reverse-discrimination plaintiffs, and not on the employer, to show the absence of an actual Title VII violation against actual victims. But that shift of the burden of proof does nothing to avoid the dangers on which Ricci focused. Employers, acting from an ex ante perspective, would still view themselves as “[f]orbidd[en] ․ to act unless they know, with certainty, that a practice violates the disparate-impact provision.” Id. In order for the strong-basis-in-evidence standard to work the way the Supreme Court seems to have intended it, an employer who does have a strong basis in evidence that a race- or gender-conscious action is necessary to avoid or remedy a disparate-impact violation must be able to rely on its strong basis in evidence when a reverse-discrimination challenge occurs. Otherwise, “employers likely would hesitate before taking voluntary action for fear of later being proven wrong in the course of litigation and then held to account for disparate treatment.” Id. If employers who had the required strong basis in evidence that specific Offerees were past victims of disparate impact could be found liable anyway when the plaintiff shows that there was no actual disparate-impact violation or that the race- or gender-conscious action benefited people who were not actual victims (or benefited actual victims beyond the extent of their victimhood), it would be of little comfort to the employer that the burden of proof rested with the other side.
The Brennan Plaintiffs and the Government also assert, in the alternative, that their “actual violation” approach applies only in those cases where the employer's voluntary action allegedly violates the contractual rights of those employees who are not the beneficiaries of the race- or gender-conscious action. But there is no basis for limiting Ricci in this way. Indeed, the Ricci Court stated, “we adopt the strong-basis-in-evidence standard as a matter of statutory construction to resolve any conflict between the disparate-treatment and disparate-impact provisions of Title VII .” Id. at 2676 (emphasis added). Ricci applies the strong-basis-in-evidence standard to all such conflicts, regardless of whether they involve contractual rights. Significantly, there were contractual rights in Ricci, and the Court never suggested that these would give rise to an exception to its careful balancing of disparate treatment and disparate impact. See id. at 2665 (describing the New Haven firefighters' CBA and its requirements concerning the promotional examinations, including a specific percentage weighting of the written and oral examinations); id. at 2679 (reviewing the job-relatedness of the percentage weighting under the strong-basis-in-evidence standard, not an actual job-relatedness standard).
2. The Consent–Decree, Settlement–Approval, and § 706(g) Cases Do Not Apply in the § 703(a) Context
The Government and the Brennan Plaintiffs cite, in support of their position, several cases addressing the consequences of a court's entering of a settlement agreement as a consent decree, a court's approval of a settlement, or a court's ordering of a remedy under § 706(g) of Title VII. If we were to import into § 703(a) the standards applicable in those cases, then the City Defendants could, indeed, be liable for disparate treatment even if they had demonstrated that they had the strong basis in evidence required by Ricci. But those cases, while they have varying degrees of relevance, do not control the case before us. In the instant case, the City Defendants unilaterally modified their employees' seniority dates while the magistrate judge's approval of the settlement was on appeal to this Court. No court had ordered the City Defendants to do this, nor had all appeals of the magistrate judge's decision been exhausted. And the district court's decision, subsequent to our appeal, not to enter as a consent decree the disputed paragraphs of the settlement agreement—which, by its terms, had by then expired—has not been appealed and would be moot in any event. See supra Part II. As a result, whatever standard applies in a case in which a court is ordering a remedy after a violation has been proven, or in which a court is deciding whether to approve a settlement or to enter it as a consent decree, is not germane to this case which does not present any of those situations. Here, the only question remaining before us is whether the employer's unilateral race- and gender-conscious action taken for the purported purpose of remedying disparate impact violates § 703(a). And that question is governed by Ricci.52
The Government first relies on United States v. City of Hialeah, 140 F.3d 968 (11th Cir.1998), and Kirkland v. N.Y. State Dep't of Correctional Servs., 711 F.2d 1117 (2d Cir.1983). Both of these cases deal with court-approved employer action, not unilateral employer action. In City of Hialeah, the Eleventh Circuit reviewed a district court's decision not to enter as a consent decree one part of a settlement agreement between the Government and the City of Hialeah. The disputed part of the consent decree purported to abrogate the contractual seniority rights of the city's unionized incumbent police officers and firefighters. The unions had not been invited to participate in the settlement discussions, and they did not give their consent. The Eleventh Circuit held that “[a] district court may not enter parts of a proposed consent decree that operate to diminish the legal rights of a party who objects to the decree on that basis.” Id. at 984; see also United States v. City of Miami, 664 F.2d 435, 436 (5th Cir.1981) (en banc) (Rubin, J., concurring); id. at 448 (Gee, J., concurring in part and dissenting in part).53 In the case before us, the district court did not enter the disputed portions of the settlement as a consent decree. The principle that a consent decree cannot dispose of the claims of a party that withholds its consent, therefore, has no application here.
Kirkland involved this Court's review of a district court's decision, in a Title VII disparate-impact case brought by private plaintiffs, to enter a class action settlement pursuant to Fed.R.Civ.P. 23(e). See Kirkland, 711 F.2d at 1121. The settlement agreement called for adjustments to the eligibility lists derived from the allegedly discriminatory civil service examination. We affirmed the district court's judgment, and we distinguished the Fifth Circuit's decision in City of Miami on the ground that the non-minority intervenors in Kirkland had no contractual rights that the settlement would alter. Id. at 1126–31. According to the Government, we must permit the Brennan Plaintiffs to show that the disputed tests and recruiting practices in the case before us are not job-related, because here, the Government says, the Brennan Plaintiffs' contractual rights will be impaired.
The problem with this contention and its reliance on City of Hialeah and Kirkland is that, unlike some consent decrees or court-approved class-action settlements, a court's holding of no more than that a private settlement agreement or an employer's unilateral action does not violate § 703(a) does nothing to impair the contractual rights of third parties. Consider a simple example: an employee is fired by her employer. She sues the employer on two theories: (1) the employer violated § 703(a) by firing her because she is female; and (2) the employer breached her employment contract. If a court rejects theory (1), the employee nevertheless remains free to pursue theory (2). Her contractual rights, and any other legal rights she may have against her employer, remain unimpaired. Moreover, it is well settled that no voluntary settlement—whether entered as a consent decree, approved under Rule 23(e), or agreed to in private—can dispose of the claims of a non-consenting third party. See Martin v. Wilks, 490 U.S. 755, 768, 109 S.Ct. 2180, 104 L.Ed.2d 835 (1989), superseded by statute on other grounds, Civil Rights Act of 1991, Pub.L. No. 102–166, 105 Stat. 1071 (“A voluntary settlement in the form of a consent decree between one group of employees and their employer cannot possibly “settle,” voluntarily or otherwise, the conflicting claims of another group of employees who do not join in the agreement.”); Local No. 93, 478 U.S. at 529 (“Of course, parties who choose to resolve litigation through settlement may not dispose of the claims of a third party, and a fortiori may not impose duties or obligations on a third party, without that party's agreement. A court's approval of a consent decree between some of the parties therefore cannot dispose of the valid claims of nonconsenting intervenors; if properly raised, these claims remain and may be litigated by the intervenor.”). The City Defendants voluntarily contracted with the Offerees and with the Government to give the Offerees retroactive seniority, but that contract cannot and does not “purport to resolve any claims the Union might have ․ as a matter of contract,” id. at 530.
The Government compounds its error at the victim-identification stage. In the Government's view, the district court should have applied the burden-shifting framework of International Brotherhood of Teamsters v. United States, 431 U.S. 324, 362–76, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977). In Teamsters, the Supreme Court held:
The proof of the pattern or practice supports an inference that any particular employment decision, during the period in which the discriminatory policy was in force, was made in pursuit of that policy. The Government need only show that an alleged individual discriminatee unsuccessfully applied for a job and therefore was a potential victim of the proved discrimination․ [T]he burden then rests on the employer to demonstrate that the individual applicant was denied an employment opportunity for lawful reasons.
Id. at 362 (footnote omitted). As for nonapplicants,
A nonapplicant must show that he was a potential victim of unlawful discrimination. Because he is necessarily claiming that he was deterred from applying for the job by the employer's discriminatory practices, his is the not always easy burden of proving that he would have applied for the job had it not been for those practices. When this burden is met, the nonapplicant is in a position analogous to that of an applicant and is entitled to the presumption discussed [above].
Id. at 367–68 (citation omitted).
Once again, the Government misperceives the procedural posture of the instant case. Teamsters applies when a court is ordering a remedy under § 706(g) of Title VII. But “the question of individual relief [under § 706(g) ] does not arise until it has been proved that the employer has followed an employment policy of unlawful discrimination.” Id. at 361. In Teamsters and other § 706(g) cases, a § 703(a) violation had been proven and the question was what remedy the court should order. The case before us in its present posture is not about judicially ordered remedies at all. It is about voluntary employer action and its relation to § 703(a). The only question we face is whether the City Defendants have violated § 703(a) by engaging in intentional reverse discrimination against the Brennan Plaintiffs. A § 706(g) case like Teamsters does not speak to that question. That is especially so because the Supreme Court has held that employers acting voluntarily can do more than a court could order them to do under § 706(g). See Johnson, 480 U.S. at 630 n. 8. But if we were to apply Teamsters here, we would be telling employers precisely that they cannot give individual relief to employees except under the circumstances in which a court could order such relief under § 706(g). Significantly, Ricci itself underscores the crucial distinction between voluntary employer action and court orders under § 706(g). As Ricci states, § 703(a) “is not so restrictive that it allows employers to act only when there is a provable, actual [disparate-impact] violation,” 129 S.Ct. at 2676, while an actual violation does have to be proven before a court can impose a remedy under § 706(g). See Teamsters, 431 U.S. at 361.
In distinguishing Teamsters from the case before us, we do not mean to suggest that its discussion of victims is not relevant to the issues before us; quite the contrary. Teamsters's analysis of how district courts applying § 706(g) should determine who is an actual victim has to be instructive for employers who are attempting unilaterally to remedy their own alleged disparate-impact violations. It inevitably informs what constitutes a strong basis in evidence that particular parties were actual victims. For, like a district court applying § 706(g), an employer “will have to make a substantial number of individual determinations in deciding which of the minority employees were actual victims of the company's discriminatory practices. After the victims have been identified, the [employer] must, as nearly as possible, recreate the conditions and relationships that would have been had there been no unlawful discrimination.” Id. at 371–72 (quotation marks omitted); see Part VI.C, infra. The employer's determinations of these matters must be supported by a strong basis in evidence under Ricci, or else the employer's race- or gender-conscious remedies will give rise to liability under § 703(a). But that does not mean that § 703(a) should be confused with § 706(g), or that the district court should use the Teamsters burden-shifting framework rather than Ricci to determine whether the employer is liable to the reverse-discrimination plaintiffs.54
3. The Brennan Plaintiffs Have Another Remedy for Any Breach of Contract by the City Defendants
The Brennan Plaintiffs, joined by the Government, further assert that, if we do not hold that the contractual seniority rights contained in the Collective Bargaining Agreement trigger an actual-violation-against-actual-victims standard, then “even though a consent decree cannot be entered, those whose contract rights are diminished have no remedy that can either prevent the unapproved settlement contract from being implemented or provide damages to those injured.” Brennan Br. at 54. If the Brennan Plaintiffs' assertion were true, it might well give us pause. But the problem with this argument is that “those whose contract rights are diminished” do have another remedy for any breach of contract. That remedy is the CBA's grievance and arbitration process. Instead of using that remedy, the Brennan Plaintiffs seek to bring a contract claim with the wrong plaintiffs in the wrong forum. Claims for the breach of a CBA should be brought by the union before an arbitrator as provided in the CBA, not by individual union members in federal court.
Whatever the outcome of this § 703(a) case, it is clear under long-established Supreme Court precedent that the Brennan Plaintiffs, through Local 891, are not barred from attacking the City Defendants' actions in a contract suit based on the collective bargaining agreement. In W.R. Grace & Co. v. Local Union 759, International Union of United Rubber, Cork, Linoleum & Plastic Workers of America, 461 U.S. 757, 103 S.Ct. 2177, 76 L.Ed.2d 298 (1983), for example, the employer, W.R. Grace, was under investigation by the EEOC for alleged discrimination against blacks and women. W.R. Grace saw its collective bargaining agreement with its employees' union expire in 1974, and there was a strike. During the strike, W.R. Grace hired strike replacements, including women who took jobs never previously held by women. When the strike was settled by the signing of a new collective bargaining agreement, the new agreement provided that the old employees would get their shift-preference seniority back. But W.R. Grace kept the strike-replacement women in their positions, ahead of men with greater seniority who were entitled to those positions under the new collective bargaining agreement. The men filed grievances pursuant to the CBA's procedures. While arbitration of those claims was pending, W.R. Grace entered into a conciliation agreement with the EEOC, which provided that the women in the bargaining unit would not only keep their jobs ahead of the male employees with more seniority; the company would also, in the event of a reduction in force, lay off male employees with more seniority if necessary to maintain the percentage of women in the bargaining unit.
W.R. Grace, acting under § 301 of the Labor Management Relations Act, 29 U.S.C. § 185, sought an injunction against the Union to bar the arbitration of any grievances that looked to relief in conflict with the terms of the conciliation agreement. The Union counterclaimed to compel arbitration, while the EEOC asked, among other things, for a declaratory judgment that the conciliation agreement prevailed over the CBA. While cross-motions for summary judgment were pending, W.R. Grace laid off some male employees in violation of their contractual seniority rights. The district court granted summary judgment to the EEOC and W.R. Grace, holding that the conciliation agreement was paramount. The Union appealed, and while that appeal was pending W.R. Grace laid off some more male employees pursuant to the conciliation agreement. Then the Court of Appeals reversed the district court's decision, and held that W.R. Grace had to arbitrate the grievances. An arbitrator then awarded backpay to some of the male employees, finding that the contract had been breached and that backpay was the appropriate remedy. W.R. Grace again sued under § 301, this time hoping to have the arbitration award set aside.
When this § 301 suit reached the Supreme Court, W.R. Grace and the United States argued that the arbitration award, and the collective bargaining agreement, could not be enforced because they were contrary to public policy. The Court disagreed:
Given the Company's desire to reduce its workforce, it is undeniable that the Company was faced with a dilemma: it could follow the conciliation agreement as mandated by the District Court and risk liability under the collective bargaining agreement, or it could follow the bargaining agreement and risk both a contempt citation and Title VII liability. The dilemma, however, was of the Company's own making. The company committed itself voluntarily to two conflicting contractual obligations. When the Union attempted to enforce its contractual rights, the Company sought a judicial declaration of its respective obligations under the contracts. During the course of this litigation, before the legal rights were finally determined, the Company again laid off employees and dishonored its contract with the Union. For these acts, the Company incurred liability for breach of contract. In effect, [the arbitrator] interpreted the collective bargaining agreement to allocate to the Company the losses caused by the Company's decision to follow the District Court order that proved to be erroneous.
Id. at 767 (footnotes omitted). Because it was the employer's fault that it had previously discriminated against women and then entered into two conflicting contracts, it was the employer that should bear the loss.
W.R. Grace also argued that enforcing the arbitration award was contrary to the policy of voluntary compliance with Title VII. Id. at 770–71 (citing Alexander v. Gardner–Denver Co., 415 U.S. 36, 44, 94 S.Ct. 1011, 39 L.Ed.2d 147 (1974)). The Supreme Court expressly rejected that argument too:
Enforcement of the [arbitration] award will not inappropriately affect this public policy. In this case, although the Company and the Commissions agreed to nullify the collective bargaining agreement's seniority provisions, the conciliation process did not include the Union. Absent a judicial determination, the Commission, not to mention the Company, cannot alter the collective bargaining agreement without the Union's consent. Permitting such a result would undermine the federal labor policy that parties to a collective bargaining agreement must have reasonable assurance that their contract will be honored. Although the ability to abrogate unilaterally the provisions of a collective bargaining agreement might encourage an employer to conciliate with the Commission, the employer's added incentive to conciliate would be paid for with the union's contractual rights.
Id. at 771 (citations omitted; emphasis added).
The City Defendants' voluntary implementation of the settlement agreement in our case, like the conciliation agreement in W.R. Grace, cannot preclude enforcement of existing contractual rights contained in a collective bargaining agreement. If any of the Brennan Plaintiffs were to file a grievance with Local 891, and if Local 891 were to pursue that claim to arbitration, a court could not disturb any arbitration award on public policy grounds or because of any conflict with the settlement agreement.
Grievance and arbitration is not simply an alternate remedy for any breach of the collective bargaining agreement; it is the proper remedy. The Supreme Court has rejected constructions of Title VII provisions that “would run counter to the national labor policy.” Am. Tobacco Co. v. Patterson, 456 U.S. 63, 77, 102 S.Ct. 1534, 71 L.Ed.2d 748 (1982). The Brennan Plaintiffs' position entails precisely such a construction. If the Brennan Plaintiffs wish to defend their contractual rights, the appropriate way to do so is not through a § 703(a) reverse-discrimination suit, but instead through the union grievance and arbitration process.55
It is, we emphasize, well established that collective-bargaining remedies are separate and independent from Title VII remedies. See Int'l Union of Elec., Radio & Mach. Workers, Local 790 v. Robbins & Myers, Inc., 429 U.S. 229, 236–37, 97 S.Ct. 441, 50 L.Ed.2d 427 (1976). “[T]he contractual rights under a collective-bargaining agreement and the statutory right provided by Congress under Title VII have legally independent origins and are equally available to the aggrieved employee.” Id. at 236 (quotation marks omitted). When an employee believes that she has suffered discrimination, she may proceed under the grievance procedure, under Title VII, or both, but the fact that the employee has one of these independent types of rights does not entitle her to anything with respect to the other. See id. (holding that the statutory period for filing a claim with the EEOC is not tolled by the pendency of grievance or arbitration procedures under a CBA). “The distinctly separate nature of these contractual and statutory rights is not vitiated merely because both were violated as a result of the same factual occurrence.” Alexander v. Gardner–Denver Co., 415 U.S. 36, 50, 94 S.Ct. 1011, 39 L.Ed.2d 147 (1974).56
The Brennan Plaintiffs are, of course, public employees, and the NLRA therefore does not apply to them. See Ford v. D.C. 37 Union Local 1549, 579 F.3d 187, 188 (2d Cir.2009) (per curiam). But the New York law governing the relationships between public employees and their unions follows the same principles as the NLRA. See, e.g ., Bd. of Educ., Commack Union Free Sch. Dist. v. Ambach, 70 N.Y.2d 501, 522 N.Y.S.2d 831, 517 N.E.2d 509, 512–13 (N.Y.1987) (quoting heavily from the U.S. Supreme Court's decision in Maddox, and holding that “[o]nly when a union has failed to represent an employee fairly—effectively depriving the employee of full use of the agreed procedures—does concern for employees' rights require that they be allowed to pursue grievances beyond the contractual mechanism”). Unless there is a breach of the duty of fair representation, “an individual union member normally lacks standing to enforce the terms of a collective bargaining agreement between the union and the employer.” Spano v. Kings Park Cent. Sch. Dist., 61 A.D.3d 666, 877 N.Y.S.2d 163, 167 (N.Y.App.Div.2009). It would be inappropriate for us to permit “individual employees,” contrary to these state-law policies borrowed from federal law, “to circumvent the grievance procedure in favor of other remedies.” Ambach, 522 N.Y.S.2d 831, 517 N.E.2d at 512.57
To summarize, in the case before us, we are neither asked to pass upon the propriety of entering the disputed portions of the 1999 settlement agreement into a consent decree, nor to enforce an agreement involving the incumbent employees' union. We are confronted only with a § 703(a) suit attacking a voluntary settlement between the Government and an employer. When the Government and an employer agree, in such a private settlement and without the consent of the union representing incumbent employees, to take away those employees' contractual rights for the benefit of alleged victims of discrimination, § 703(a) is not the correct way for the incumbent employees to seek vindication of their rights under the contract. The proper way is through the union grievance and arbitration process. As a result, the Brennan Plaintiffs' asserted contractual rights do not alter the generally applicable Ricci standard, and their § 703(a) claim is subject to the Ricci strong-basis-in-evidence defense just like any other such claim.58
VI. Application of the Strong–Basis–in–Evidence Standard
Under the strong-basis-in-evidence standard, the City Defendants need to show three things in order to avail themselves of the defense the Supreme Court recognized in Ricci: (1) that they were faced with a prima facie (or, potentially, a strong basis in evidence of a prima facie case) of disparate impact;59 (2) a strong basis in evidence that either the challenged employment practice was not job-related or there was a less discriminatory alternative to the practice; and (3) a strong basis in evidence that each Offeree was a victim of the challenged practice and received no more than make-whole relief. This showing needs to be made separately for each of the challenged employment practices—testing and recruiting.60
The district court—understandably, as it decided this case before Ricci—erred in its Title VII analysis in several respects. First, the district court applied the “affirmative action” framework of Johnson and Weber, and therefore required for Title VII purposes only a “manifest imbalance” instead of a prima facie case of disparate impact. Second, the district court failed to require of the City Defendants a strong basis in evidence—or, for that matter, any evidence at all—either of non-job-relatedness or a less discriminatory alternative. Third, the district court held that no showing concerning victimhood was required to justify the awards of transfer and TCA seniority, but that a showing of actual victimhood was required for layoff seniority. For all of these aspects of competitive seniority, the court should, instead, have required a strong basis in evidence of victimhood.
Accordingly, and as further described in the discussion below, we remand for the district court to apply this correct standard in the first instance.
A. Prima Facie Case
1. Testing Discrimination
Before the district court, it was undisputed that the City Defendants were faced with a prima facie case of disparate-impact testing discrimination for all three of the challenged exams, except that the Brennan Plaintiffs questioned whether there was a prima facie case that Exam 8206 had a disparate impact on Hispanics. See NYC Board III, 448 F.Supp.2d at 425–27. After an “evidentiary hearing,” the district court found that such a prima facie case had been made out. See NYC Board IV, 487 F.Supp.2d at 224–32. The Brennan Plaintiffs have not appealed that determination. Accordingly, there is no reason for the district court to revisit it on remand.
2. Recruiting Discrimination
The district court held that the City Defendants were not faced with a prima facie case of disparate-impact recruiting discrimination, reasoning that although the Government had shown a disparity between the expected and actual numbers of female and minority test-takers, there was no “evidence connecting the ․ disparity to the Board's recruiting practices, a causal link not established by the statistical evidence presented in support of the recruiting claim.” NYC Board IV, 487 F.Supp.2d at 234; see also NYC Board III, 448 F.Supp.2d at 446 n. 57. The question of whether the district court erred in this respect is complicated. The Government, for example, wishes to argue to that court that it did. Because the court below reached its decision on this issue before Ricci, and because Ricci raises a question regarding the proper standard to apply, we think it appropriate to remand this issue for further consideration in the light of the still-to-be-determined Ricci standard. We note in particular that even if there was not an actual prima facie case of disparate impact recruiting liability, Ricci might—but does not necessarily—permit a showing that the City Defendants had a strong basis in evidence of such a prima facie case. See supra Part V.A.1. Whether such a strong basis in evidence would suffice remains an open question. But the district court need not reach or decide this question if it finds that the City Defendants lacked even a strong basis in evidence of a prima facie case of recruiting discrimination.
At the same time, we reject the Brennan Plaintiffs' contention that the City Defendants' allegedly discriminatory recruiting practices—word-of-mouth recruiting, and limited advertising—are not prohibited by Title VII even if they cause a disparate impact. Like the other courts that have addressed similar arguments, we are unpersuaded. As the Sixth Circuit has stated, the “assertion that disparate impact analysis is inapplicable to ․ recruiting practices is plainly incorrect. The very purpose of Title VII's disparate impact theory is to eradicate barriers which discriminate on the basis of race, gender, religion, and other protected classifications.” United States v. City of Warren, 138 F.3d 1083, 1094 (6th Cir.1998) (quotations, modifications, and footnote omitted); see also Thomas v. Wash. Cnty. Sch. Bd., 915 F.2d 922, 924–26 (4th Cir.1990) (a school board violated Title VII through a combination of nepotism, word-of-mouth recruiting, and “the general practice of posting notice of vacancies only in the schools,” where minorities were unlikely to see them); United States v. Ga. Power Co., 474 F.2d 906, 925 (5th Cir.1973) (“Word-of-mouth hiring and interviewing for recruitment only at particular scholastic institutions are practices that are neutral on their face. However, under the facts of the instant case, each operates as a ‘built-in-headwind’ to blacks․”). In addition, this court, many years ago, concluded that a prima facie case of disparate impact existed based on “subjective word-of-mouth hiring methods.” See Grant v. Bethlehem Steel Corp., 635 F.2d 1007, 1016 (2d Cir.1980).
The Brennan Plaintiffs do not deny that recruitment practices qualify as employment practices. Their argument is, essentially, that § 703(a)(2), which makes it illegal for an employer “to limit, segregate, or classify his employees or applicants for employment in any way which would deprive or tend to deprive any individual of employment opportunities,” does not refer to recruiting, because people who don't even apply for a job because of recruiting discrimination are not “applicants for employment.” But, as the Government says, the Supreme Court has all but held that “applicants” includes potential applicants in the § 703(a)(2) disparate-impact context. See Wards Cove, 490 U.S. at 651 & n. 7 (suggesting that a prima facie case of disparate impact would be made out “if it were found that [a] dearth of qualified nonwhite applicants was due to practices on [an employer's] part which—expressly or implicitly—deterred minority group members from applying for ․ positions.”); see also Dothard v. Rawlinson, 433 U.S. 321, 330, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977).61 Accordingly, we make explicit what may have only been implicit in our decision in Grant and hold that potential applicants for employment are “applicants for employment” for § 703(a)(2) purposes.62
B. Job–Related and Less Discriminatory Alternative
The next component of the Ricci strong-basis-in-evidence standard requires that the employer show, for each disputed employment practice, a strong basis in evidence either that the practice was not job-related or that there was a less discriminatory alternative to that practice. See Ricci, 129 S.Ct. at 2678. The district court considered evidence concerning job-relatedness and less discriminatory alternatives neither for the challenged exams, nor for the recruiting practices. See NYC Board III, 448 F.Supp.2d at 426 n. 37. On remand, the district court should determine whether the City Defendants had a strong basis in evidence alternative.63 See supra Part V.A.1. Whether the district court needs to make such a determination with respect to the two challenged recruiting practices depends on whether the recruiting defense survives the initial, prima facie case/strong-basis-in-evidence of prima facie case, step of the Ricci inquiry.
C. Necessity and Make–Whole Relief
Next, the district court must determine whether the City Defendants had a strong basis in evidence that their race- and gender-conscious actions were necessary to avoid disparateimpact liability. See Ricci, 129 S.Ct. at 2677; supra Part V.A.2. As we have explained, that means the City Defendants must have had, at the time they took their race- and gender-conscious actions, a strong basis in evidence that each Offeree was an actual victim of discrimination who received no more than make-whole relief, i.e., a strong basis in evidence that, in such a disparate impact suit, he or she would have been entitled to receive, by a court order pursuant to § 706(g) of Title VII, the race- or sex-conscious relief that he or she received from the City Defendants through the voluntary settlement agreement.
The district court's decision as to job-relatedness and less discriminatory alternatives will affect this step in several important ways. First, if there was no strong basis in evidence that any of the challenged testing and recruiting practices were either not job-related or that there was a less discriminatory alternative to them, then the district court need not make any determinations as to particular Offerees.64 The Brennan Plaintiffs would then have successfully shown that implementation of the settlement agreement as a whole made the City Defendants liable for reverse discrimination, under § 703(a) of Title VII. That is so because an employer cannot have a strong basis in evidence that anyone is a victim of discrimination if there is no strong basis in evidence that discrimination has occurred.
Second, and for the same reason, if the City Defendants can show a strong basis in evidence of non-job-relatedness or of a less discriminatory alternative as to one or more of the challenged employment practices, but not as to others, then the City Defendants have violated § 703(a) to the extent they have given relief to Offerees for whom there was no strong basis in evidence that such Offerees were victims of the practice or practices for which there was a strong basis in evidence of non-job-relatedness or of a less discriminatory alternative. For example, if the City Defendants had a strong basis in evidence that their tests were not job-related, but they lacked a strong basis in evidence that their recruiting practices were neither job-related nor that there existed a less discriminatory alternative to these recruiting practices (or could not make out a prima facie/ strong-basis-in-evidence of a prima facie showing of recruiting discrimination), then the City Defendants would, for each Offeree, have to show a strong basis in evidence that that Offeree was a victim of testing discrimination. Essentially, the City Defendants would be required to show a strong basis in evidence that the Offeree took the test, or that the Offeree could successfully have “undertake [n] [his or her] difficult task of proving that [he or she] should be treated as [an] applicant[ ] and therefore [is] presumptively entitled to relief accordingly” under § 706(g). See Teamsters, 431 U.S. at 364.65
Third, if the district court does find a strong basis in evidence of disparate-impact liability, the appropriate extent of relief the City Defendants could have given each Offeree without violating § 703(a) turns on which theory—non-job-relatedness or a less discriminatory alternative—supports the district court's determination. If the Government successfully shows on remand, as it hopes to, that there was a strong basis in evidence that the provisional hiring process was a less discriminatory alternative to a particular hiring practice, then it follows that any Offeree, as to whom there is a strong basis in evidence that the Offeree was a victim of that particular discriminatory practice, properly received retroactive seniority to his or her provisional hiring date. For if an Offeree was hired under the provisional hiring process and that process was a less discriminatory alternative, then there is a strong basis in evidence that that Offeree would have received a permanent appointment if the City Defendants had used the less discriminatory alternative instead.
Alternatively, if there was a strong basis in evidence that an employment practice that had a disparate impact discriminated only because it was not job-related, it becomes much more difficult for the district court to determine how much retroactive seniority the City Defendants could properly give each Offeree without violating § 703(a). Suppose, for example, that the challenged tests were discriminatory; it is nigh-impossible to tell exactly when (if at all) an Offeree would have been hired if the test had not been discriminatory. In that hypothetical world, some Offerees might have passed with flying colors and would then have been the first to be hired from the eligibility lists; others might have found themselves in the middle of the eligibility lists; and still others might still have failed, or might have passed with such a low score that they would have been placed near the bottom of the eligibility lists, where they probably would never have been hired. And, those who passed a test might have been rejected at the interview stage (which is not here being challenged as discriminatory), once or twice, thus having their employment delayed; or three times, resulting in removal from the eligibility list. Finally, they might have failed the “experience papers” stage, and then they might or might not have succeeded in having that decision overturned on administrative appeal. To say the least, “[t]he task remaining for the District Court on remand will not be a simple one.” Teamsters, 431 U.S. at 371. “Th[e] process of recreating the past will necessarily involve a degree of approximation and imprecision.” Id. at 372.
Recreating the past is difficult in the § 706(g) context addressed in Teamsters. It is even more difficult when an employer undertakes voluntary remedial action that is challenged under Ricci and § 703(a). In the § 706(g) context, a court necessarily has already found an actual violation of Title VII. See id. at 361–62. All the court has then to do is to determine who is an actual victim of that discrimination, and “ ‘recreate the conditions and relationships that would have been had there been no’ unlawful discrimination,” id. at 372 (quoting Franks, U.S. at 769). That process is not easy, because there may be more victims of discrimination than there are vacancies that were discriminatorily refused to them, and their interests must be balanced against “the legitimate expectations of other employees innocent of any wrongdoing.” Id. But when an employer undertakes voluntary action to remedy an unintentional disparate impact, the employer has not proven, and does not need to prove, that it has actually violated Title VII. Ricci, 129 S.Ct. at 2674. Instead, the employer must have a strong basis in evidence that it would be liable in a disparate-impact suit unless it took the race- or gender-conscious action. See id. at 2675–76. Once an employer has a strong basis in evidence that it faces disparate-impact liability, the employer must also undertake a recreation of the past that is supported by a strong basis in evidence. Where, as here, incumbent employees subsequently bring a § 703(a) challenge to the employer's race- or gender-conscious action, the court must determine whether there really was a strong basis in evidence to support the employer's recreation of the past.66 That is a very difficult project indeed. It is one that only a district court can undertake, and whose difficulty should serve to caution any appellate court attempting to review the district court's conclusions.
The district court, deciding this case before Ricci, asked whether each Offeree was an actual victim, instead of whether the City Defendants had a strong basis in evidence that each Offeree was an actual victim. As a result, we must vacate the judgment and remand so that the district court can apply the correct standard. In addition, several of the district court's findings require comment.
First, as discussed above, the district court did not make any individualized determinations as to alleged victims of recruiting discrimination, because it concluded that there was no prima facie case of recruiting discrimination. Because we have remanded on the question of whether, under Ricci, there was a prima facie case/strong-basis-in-evidence of a prima facie case of recruiting discrimination, individualized determinations pertaining to recruiting-discrimination victimhood will be required to the extent that the district court finds (a) that the recruiting defense survives Ricci's prima facie case/strong-basis-in-evidence of prima facie case requirement; and (b) that there was a strong basis in evidence that the challenged recruiting practices either were not job-related or had a less discriminatory alternative.
Second, the district court twice said that the Brennan Plaintiffs “acknowledge that seven of the [Offerees] were victims of discrimination under the testing claims and received appropriate make-whole relief: Lloyd Bailey, Joseph Christie, Belfield Lashley, Gilbert Rivera, Peter Robertin, Felix Torres and Mayra Zephrini (Cintron).” NYC Board III, 448 F.Supp.2d at 419; see also NYC Board V, 556 F.Supp.2d at 208. The Brennan Plaintiffs state, and the Government agrees, that no such concession was made. The district court did not provide a citation to the record to support its assertion, either in 2006 or in 2008. On remand, the district court should either explain the basis for its assertion, deem the matter forfeited, or make individualized strong-basis determinations for each of these individuals.67
Third, the district court did not address whether the ten Arroyo Intervenors were victims of testing discrimination. The district court had held, as to them and other alleged victims of testing discrimination, that retroactive seniority was limited to actual victims for layoff purposes only. The parties then resolved the Arroyo Intervenors' layoff seniority dates by stipulation. That stipulation, however, provides that “[s]hould a higher court conclude that ‘actual victim’ status is relevant to the lawfulness of any of the seniority benefits that these beneficiaries received under the Agreement for the purpose of obtaining transfers or temporary care assignments, each shall have the right to a hearing on ‘actual victim’ status in [the district court] on remand.” Additionally, “[i]n the event of any remand by a higher court for a purpose that is related to the lawfulness of the seniority benefits received by the [Arroyo Intervenors], these stipulated dates shall no longer apply.” Because we have concluded that the retroactive transfer and TCA seniority given to the Arroyo Intervenors, like that given to the other Offerees, violates § 703(a) of Title VII to the extent it is not supported by a strong basis in evidence that each individual was an actual victim of discrimination, the stipulation no longer applies on remand.68 As a result, subject of course to any possible new stipulation, the district court, on remand, will have to address whether the City Defendants had a strong basis in evidence that the Arroyo Intervenors were actual victims of discrimination who received no more than make-whole relief when the City Defendants implemented the settlement agreement.
Fourth, the district court did make actual-victim status determinations for the remaining eight Offerees: Ricardo Cordero, Thomas Fields, Carla Lambert, Vernon Marshall, Angel Pagan, Anthony Pantelides, Sean Rivera, and Luis Torres. The district court found that (1) Cordero and Marshall, each of whom failed an exam, were actual victims and the retroactive seniority they received was lawful, NYC Board V, 556 F.Supp.2d at 209–10; (2) Sean Rivera, who was, against the odds, hired after years of delay because his barely passing score put him at the very bottom of the eligibility list where it was unlikely he would ever be hired, was not an actual victim because his “hypothetical performance [on a nondiscriminatory exam] is purely speculative,” as is “what such a score would have yielded in terms of his hire date,” id. at 210; and (3) the other five Offerees would not have been hired even under a nondiscriminatory exam because they were not qualified, id . at 210–11. We vacate this portion of the judgment because the district court should have asked whether the City Defendants had a strong basis in evidence that these individuals were victims, not whether the individuals actually were victims.69
We note, moreover, that some of the district court's findings in this portion of its judgment are problematic in two other ways. There is a tension between the finding that Rivera's hypothetical performance on a nondiscriminatory exam was speculative, while Cordero's and Marshall's were not. Anyone who received a lower score than he or she would have received on a nondiscriminatory test is a likely victim of discrimination, whether the lower score resulted in outright failure or, instead, delay in being hired. See, e.g., Guardians Ass'n v. Civil Serv. Comm'n, 633 F.2d 232, 248 n. 30 (2d Cir.1980) (“There is no reason why the acceptance of a belated offer of employment should be deemed a waiver of the right to seek redress for a discriminatory delay.”).
In addition, the district court erred in a somewhat more subtle way in allocating the burden of proof. The district court cited Johnson v. California, 543 U.S. 499, 505, 125 S.Ct. 1141, 160 L.Ed.2d 949 (2005), an Equal Protection case, for the proposition that “the parties to the [Settlement] Agreement—i.e., the United States and the Board—bear the burden of proof” as to the actual-victim status of each Offeree. NYC Board V, 556 F.Supp.2d at 209. Although under Ricci the § 703(a) reverse-discrimination defendant bears the burden at trial of showing a strong basis in evidence that its race-or gender-conscious actions were necessary to avoid disparate-impact liability, see 129 S.Ct. at 2664 (“[R]ace-based action like the City's in this case is impermissible under Title VII unless the employer can demonstrate a strong basis in evidence that, had it not taken the action, it would have been liable under the disparate-impact statute.”), it is also true, as the Government points out, that if the Government had pursued its disparate-impact claims to judgment and the district court had imposed a remedy under § 706(g), then the case governing the burden of proof would have been Teamsters, not Johnson v. California. Under Teamsters, as this Court has interpreted it,
[e]ach class member must show that he or she was among those adversely affected by the challenged policy or practice. If this showing is made, the class member is entitled to individual relief unless the employer in turn can establish by a preponderance of the evidence that a legitimate non-discriminatory reason existed for the particular adverse action.
Robinson, 267 F.3d at 161–62.70 Viewing Teamsters through the strong-basis-in-evidence lens of Ricci, an employer, before it takes a race-or gender-conscious action to avoid or remedy a disparate-impact violation, must have a strong basis in evidence that (1) each beneficiary of the race- or gender-conscious action could “show that he or she was among those adversely affected” by the disparate-impact practice, and (2) the employer would be unable to “establish by preponderance of the evidence that a legitimate non-discriminatory reason existed for the particular adverse action.”71
Fifth, the district court's 2006 opinion addressed the Brennan Plaintiffs' claim that five Offerees “are not members of a protected class because each lacks the requisite cultural or linguistic ties to qualify as Hispanic.” NYC Board III, 448 F.Supp.2d at 422. The district court, relying on the EEOC's definition of national-origin discrimination, see 29 C.F.R. § 1606.1,72 held that one Offeree, Ciro Dellaporte, was not Hispanic because he was “plainly ․ of Italian ancestry.” NYC Board III, 448 F.Supp.2d at 422. (The City Defendants, who had erroneously listed him as Hispanic in their records, did not thereafter defend his appointment.) The other four, the district court held, were Hispanic: Kevin LaFaye's father and the mother of Nicholas and Anthony Pantelides were born in Puerto Rico, and Steven Lopez's grandfather was born in Mexico. Id. The Brennan Plaintiffs appeal the district court's decision, but only as to the Pantelides brothers. We agree with the district court that the City Defendants properly treated the Pantelides brothers as Hispanic, in light of their ancestors' national origin. See Espinoza v. Farah Mfg. Co., 414 U.S. 86, 88, 94 S.Ct. 334, 38 L.Ed.2d 287 (1973) (“The term ‘national origin’ on its face refers to the country where a person was born, or, more broadly, the country from which his or her ancestors came.”). The district court therefore need not revisit this issue when it applies the strong-basis-in-evidence standard on remand.
With these considerations in mind, we remand for the district court to take up the difficult task of determining whether the “hypothetical nondiscriminatory past” that the City Defendants attempted to recreate, when they implemented the settlement agreement, was supported by a strong basis in evidence.73
VII. Equal Protection
The parties also ask us to review the district court's decisions under the Equal Protection Clause. We think, however, that the better course is to follow one “of the cardinal rules governing the federal courts: ․ never to anticipate a question of constitutional law in advance of the necessity of deciding it.” Brockett v. Spokane Arcades, Inc., 472 U.S. 491, 501, 105 S.Ct. 2794, 86 L.Ed.2d 394 (1985) (quotation marks omitted). That is especially so here because the case presents difficult and far-reaching issues concerning the application of the Equal Protection Clause to remedies for past disparate impact. We decline to decide those issues in this case until the district court has addressed the Title VII issues under the proper standard—i.e., Ricci rather than Johnson and Weber. Only if this court, upon reviewing a district court decision that has applied the correct Title VII standard, determines that some part of the City Defendants' voluntary implementation of the disputed paragraphs of the settlement agreement did not violate § 703(a) of Title VII, would it be necessary or prudent for us to address the constitutional issues.
The Supreme Court has cautioned that, although “[a] fundamental and longstanding principle of judicial restraint requires that courts avoid reaching constitutional questions in advance of the necessity of deciding them,” in some cases it is “inadvisable to vacate and remand” for the court below to address a statutory issue where “considerations of judicial economy” strongly favor addressing the constitutional issue on the merits. Lyng v. Nw. Indian Cemetery Protective Ass'n, 485 U.S. 439, 445–47, 108 S.Ct. 1319, 99 L.Ed.2d 534 (1988). The principle of judicial restraint would not be “vindicated by sending [a] case on what would almost certainly be a brief round trip to the court[ ] below.” Id. at 447. Here, however, the trip to the district court will likely be—perhaps to the able district judge's chagrin—anything but brief. Because the district court did not permit the Brennan Plaintiffs to make any showing as to job-relatedness of, or a less discriminatory alternative to, the challenged employment practices, a factual record will have to be developed. Then the district court will have to decide whether that factual record shows that the City Defendants had a strong basis in evidence. To the extent that the district court finds a strong basis in evidence that the City Defendants faced disparate-impact liability, the court will have to determine whether the City Defendants also had a strong basis in evidence that giving each Offeree retroactive seniority was necessary to remedy any disparate-impact Title VII violation. And then, the district court will have to impose a remedy for whatever reverse-discrimination § 703(a) violations it finds. See Part IX, infra.
Even more important, there is at least some possibility that the constitutional issues will not come back to us.74 If the district court finds that the disputed portions of the settlement agreement violated § 703(a) in their entirety, i.e., the City Defendants failed to satisfy the strong-basis-in-evidence standard of Ricci with respect to any of their employment practices, then neither we nor the district court will need to address whether the City Defendants also violated the Equal Protection Clause. In cases like this one, where our addressing a constitutional issue could dispose of the case, but there is at least one non-trivial non-constitutional issue which could also dispose of the case after a remand to the district court, the better practice is to remand for the district court to address the latter first. See, e.g., Westchester Day Sch. v. Vill. of Mamaroneck, 386 F.3d 183, 191 (2d Cir.2004) (“Prudence counsels against reaching out to establish a far-reaching constitutional rule when there are many other bases upon which this case may ultimately be decided.”).
Moreover, the prudential principle of avoiding unnecessary constitutional adjudication “has been strictly followed where, as in the present case, difficult or far-reaching constitutional issues are raised.” Fed. Election Comm'n v. Cent. Long Island Tax Reform Immediately Comm., 616 F.2d 45, 52 (2d Cir.1980) (en banc) (per curiam). There are at least two such issues in the case before us. First, the parties call upon us to decide a question the Supreme Court explicitly left open in Ricci: whether “meeting the [Title VII] strong-basis-in-evidence standard would satisfy the Equal Protection Clause in a future case.” 129 S.Ct. at 2676. Second, the Brennan Plaintiffs argue here, as they did below, that no showing of previous unintentional disparate impact can ever satisfy the Equal Protection Clause. Rather, they say, only remedying pervasive and egregious intentional disparate treatment is a “compelling interest” for purposes of strict scrutiny75 (or an “important interest” for purposes of heightened scrutiny).76 See Ricci, 129 S.Ct. at 2682 (Scalia, J., concurring). These issues are manifestly “difficult” and “far-reaching.” See Richard Primus, The Future of Disparate Impact, 108 Mich. L.Rev. 1341 (2010); Richard Primus, Equal Protection and Disparate Impact: Round Three, 117 Harv. L.Rev. 493 (2003).
We, therefore, decline at this time to address the parties' contentions under the Equal Protection Clause.
VIII. Class Certification
The Caldero Intervenors argue that the district court should not have certified the Brennan class. They contend that the district court abused its discretion in certifying, and violated due process, by failing to give any party other than the Brennan Plaintiffs an opportunity to brief the issue of class certification. We are not persuaded.
We review class-certification decisions for abuse of discretion. E.g., Robinson, 267 F.3d at 162. Here, when the district court originally certified the class, it acknowledged that the Brennan Plaintiffs were the only party who had had a chance to submit a brief on the issue of class certification. NYC Board III, 448 F.Supp.2d at 444. We agree that it might have been better had the district court waited for full briefing of the issue instead of deciding it in 2006. But the Caldero Intervenors failed to make any Rule 23 argument in their opening brief to us. Their reply brief suggests some Rule 23 issues they would have raised below if they had had an opportunity. To the extent these purport to be Rule 23 arguments against certification, they are not in the opening brief and are therefore forfeited. See, e.g., Poupore v. Astrue, 566 F.3d 303, 306 (2d Cir.2009).77 Although the opening brief does note that “a Title VII class action, like any other class action, may only be certified if the trial court is satisfied, after a rigorous analysis, that the prerequisites of Rule 23(a) have been satisfied,” Gen. Tel. Co. of Sw. v. Falcon, 457 U.S. 147, 161, 102 S.Ct. 2364, 72 L.Ed.2d 740 (1982), there is no argument that the actual analysis in the district court's opinion was not rigorous. Nor can we say, in the absence of specific arguments, that the lack of briefs by the Caldero Intervenors rendered the court's certification decision not “rigorous.”
The Caldero Intervenors' due process claim is also unavailing, for it is too conclusory to avoid forfeiture on appeal. Their opening brief states only that there was a due process violation and cites Nelson v. Adams USA, Inc., 529 U.S. 460, 465, 120 S.Ct. 1579, 146 L.Ed.2d 530 (2000). That is not sufficient to raise the issue on appeal. See Norton v. Sam's Club, 145 F.3d 114, 117 (2d Cir.1998) (“[W]e have concluded that merely ․ stating an issue without advancing an argument ․ [does] not suffice.”).78
If a violation of § 703(a) or the Equal Protection Clause is found, the district court will have to decide upon a remedy. Here, the Brennan Plaintiffs attack the remedy imposed by the court below as too narrow, while the Government and the Caldero Intervenors attack it as too broad.
The district court, having found violations of both Title VII and the Equal Protection Clause, did not award any damages to the Brennan Plaintiffs, nor did it issue an injunction. Instead, the court issued the following declaratory judgment:
1. As a remedy for testing discrimination, the awards of retroactive seniority to black and Hispanic Offerees for purposes of layoff protection violate[ ] Title VII and the Fourteenth Amendment, except insofar as the awards provide make-whole relief to actual victims of testing discrimination.
2. As a remedy for recruiting discrimination, the awards of retroactive seniority to female Offerees for purposes of layoff protection violate[ ] Title VII.
3. As a remedy for recruiting discrimination, the awards of seniority—both retroactive and based on permanent-appointment date—to black, Hispanic and Asian male Offerrees [sic] for purposes of transfers, TCAs and layoff protection violate[ ] the Fourteenth Amendment.
4. Ciro Dellaporte is not a member of a protected class and, therefore, not entitled to any relief.
5. Except as stated above, the relief provided by the Agreement comports with Title VII and the Fourteenth Amendment.
Attached to the judgment were two schedules listing the “proper” competitive seniority dates for the Offerees.
We do not decide the proper scope of the remedy here. We review a Title VII remedial order issued under § 706(g) for abuse of discretion. “Our function is not to exercise our own discretion, but to determine, in light of the purposes of the Act, whether the district court judge has abused his.” Ass'n Against Discrimination in Employment, Inc. v. City of Bridgeport, 647 F.2d 256, 279 (2d Cir.1981). Similarly, “federal courts have leeway to fashion appropriate relief, and appellate tribunals have accorded district courts broad discretion to frame equitable remedies for constitutional violations so long as the relief granted is commensurate with the constitutional infraction.” Cooper v. U.S. Postal Serv., 577 F.3d 479, 496 (2d Cir.2009) (quotation marks and brackets omitted). And, since the scope of the City Defendants' liability is yet to be determined, it would be premature for us to say exactly what the scope of the district court's equitable discretion might be.79 As for the Brennan Plaintiffs' damage claims, those too depend on the district court's determination of the scope of liability. The damage claims are premised on specific Brennan Plaintiffs' having lost transfers to specific Offerees. That is why the district court dismissed those claims; it had held that “[t]he transfers at issue ․ comported with Title VII and the Fourteenth Amendment.” Before any damage award would be appropriate, the district court must first determine, under the standards we have set forth in this opinion, whether the City Defendants violated the law when they gave the transfers to those particular Offerees.80
It will, of course, be impossible for the district court to fashion a remedy that makes everyone happy, much less whole. To the extent that the City Defendants' employment practices were discriminatory, there were too many victims to count. A few, the Offerees, got permanent appointments and retroactive seniority. Others, like Ruben Miranda, did not. And there may well be an untold number of non-Offeree individuals who failed the challenged tests or were prevented from learning of the Custodian and CE openings because of recruiting discrimination. Had the Government successfully pursued its case to judgment instead of settling, this would probably have been one of those cases which would have “require[d] class-wide, rather than individualized, assessments of monetary relief,” because “identification of the individuals entitled to relief would drag the court into a quagmire of hypothetical judgments and result in mere guesswork.” Robinson, 267 F.3d at 161 n. 6 (quotation marks omitted). Instead, the Government and the City Defendants entered into a settlement that attempted to identify victims of discrimination anyway—without consulting Local 891. To the extent that this settlement violated Title VII or the Equal Protection Clause, the Brennan Plaintiffs were victims of reverse discrimination, as were the other members of the Brennan class, of whom there are hundreds.81 At this point, nobody can really know who would have been employed as a Custodian or CE, and at what level of seniority, in a hypothetical world in which there had been no unlawful discrimination.
There can be no doubt that the Brennan Plaintiffs, and many of the members of the certified Brennan class, were harmed by the permanent appointments of the Offerees and by the seniority stemming from those appointments. And to the extent the settlement benefited some Offerees unlawfully, redress, in the form of adjustments in seniority, appears justified. But, once the district court has reduced each “unlawful” Offeree's seniority to his or her permanent-appointment date, the value of further modifications to seniority dates begins to dissipate quickly. Such additional modifications place a heavy cost upon innocent Offerees, while conferring possibly speculative benefits upon diffuse victims of reverse discrimination.82
In any event, we need not and do not decide exactly what remedy the district court should impose (or to what extent the district court should find the liability on which any remedy would necessarily be premised). It is for the district court to decide what, if any, is the scope of the City Defendants' liability, and then to exercise appropriate equitable discretion in imposing a remedy. In doing so, the district court should explain why it exercised its discretion in the way that it did, so that a reviewing court can determine whether that discretion has been abused.83
Because the district court in its Title VII analysis reached results inconsistent with the Supreme Court's subsequent decision in Ricci, its judgment must be vacated and remanded—with two exceptions. First, we affirm the district court's grant of class certification. Second, paragraph 4 of the district court's declaratory judgment, which states, “Ciro Dellaporte is not a member of a protected class and, therefore, not entitled to any relief,” has not been appealed and therefore must stand.84 The judgment of the district court is therefore AFFIRMED in part and VACATED in part, and the case is REMANDED for further proceedings consistent with this opinion.
I join the majority in concluding (1) that the district court judgment can be affirmed insofar as it (a) granted class certification to the Brennan Intervenors, and (b) declared Ciro Dellaporte not a member of a protected class entitled to relief under the challenged settlement, but (2) that the judgment must be vacated in all other respects and remanded for further consideration in light of Ricci v. DeStefano, ––– U.S. ––––, 129 S.Ct. 2658, 174 L.Ed.2d 490 (2009). In reaching this conclusion, I agree with the majority that Ricci is not limited, as the Caldero and Arroyo Intervenors urge, to its particular facts. I also agree that the challenged settlement cannot be characterized as an affirmative action plan, so that we need not consider these intervenors' argument that Ricci does not apply to such plans. With due respect, however, I cannot join in the majority opinion because I think its extended discussion of Title VII jurisprudence generally, and the scope of the Ricci rule in particular, is not required to our decision to remand and yields an abundance of dicta that could confuse future consideration of judgments actually based on Ricci. As the Supreme Court has cautioned, “however helpful it might be for us to adjudge every pertinent statutory and constitutional issue” that could arise in the application of a law or decision, “we cannot properly reach out and decide matters not before us.” Granfinanciera, S.A. v. Nordberg, 492 U.S. 33, 64 n. 19, 109 S.Ct. 2782, 106 L.Ed.2d 26 (1989); see United States v. Tomasi, 313 F.3d 653, 660 (2d Cir.2002) (Sotomayor, J., concurring in the judgment) (“While clarity in the law is always to be desired, judges should not indulge themselves by reaching out to decide issues not squarely before them in order to accomplish this result.”); see also PDK Labs., Inc. v. United States D.E.A., 362 F.3d 786, 799 (D.C.Cir.2004) (Roberts, J., concurring in part and concurring in the judgment) (identifying as “cardinal principle of judicial restraint” that “if it is not necessary to decide more, it is necessary not to decide more”).
A. Ricci Requires No Gloss from This Court To Permit Application by the District Court on Remand
In Ricci, a case originating in this circuit, the Supreme Court addressed the same question raised here: “whether the purpose to avoid disparate-impact liability excuses what otherwise would be prohibited disparate-treatment discrimination.” Ricci v. DeStefano, 129 S.Ct. at 2674. Acknowledging that “statutes and principles” seemed to “point in different directions,” the Court set itself the “task ․ to provide guidance to employers and courts for situations when these two prohibitions could be in conflict absent a rule to reconcile them.” Id. Toward that end, the Court pronounced the following rule:
[B]efore an employer can engage in intentional discrimination for the asserted purpose of avoiding or remedying an unintentional disparate impact, the employer must have a strong basis in evidence to believe it will be subject to disparate-impact liability if it fails to take the race-conscious, discriminatory action.
Id. at 2677. The Court explained that the requisite strong basis had to be “objective,” id., and could not be satisfied by “a few stray (and contradictory) statements in the record,” id. at 2680. Further, it applied the strong-basis-in-evidence requirement to all three prongs of disparate-treatment analysis set forth in 42 U.S.C. § 2000e–2(k)(1)(A), see id. at 2677–79 (concluding that (1) “City was faced with a prima facie case of disparate-impact liability,” but (2) the evidence raised “no genuine dispute that the [discarded] examinations were job-related,” and (3) respondents “lacked a strong basis in evidence of an equally valid, less discriminatory testing alternative that the City, by certifying the examination results, would necessarily have refused to adopt”). The Supreme Court apparently did not think that district courts required any further guidance to begin applying the stated rule: “Our holding today clarifies how Title VII applies to resolve competing expectations under the disparate-treatment and disparate-impact provisions.” Id. at 2681. Thus, I would simply remand this case for further consideration in light of Ricci without attempting to anticipate or resolve questions that are not now before us and that may never arise in this case. Case-by-case review of judgments actually based on Ricci will afford sufficient opportunity to discuss application of the strong-basis-in-evidence rule in particular circumstances.
B. Concerns Raised by The Majority's Discussion
The majority's efforts to define the parameters of the Ricci rule in advance of application by the district court raise a number of concerns in my mind. The following are merely illustrative.
First, in order to engage in a detailed discussion of Title VII law generally and Ricci in particular in light of the evidence in this case, the majority finds itself obliged at the outset to decide the standard applicable to review of the factual record. While on an appeal from an award of summary judgment, we usually review the facts in the light most favorable to the non-moving party, see, e.g., Wilson v. C.I.A., 586 F.3d 171, 183 (2d Cir.2009), the majority elects to employ the “clear error” standard applicable after trial, see, e.g., Skoros v. City of New York, 437 F.3d 1, 12 (2d Cir.2006). This choice is grounded more in record confusion, however, than in law. As the majority explains, while the district court disposed of “the vast majority of the relevant issues ․ upon cross-motions for summary judgment ․, in what appears to be a confusion about the case's procedural posture,” some issues were resolved after “ ‘evidentiary hearings.’ “ Ante at . Unable to discern “why the district court held these hearings, rather than a trial,”1 the majority simply decides “it best to treat the hearings as separate bench trials on separate issues,” and to review the findings for “clear error.” Id. at [45–46]. This course of action has consequences for the ensuing discussion as the difference between review in the light most favorable to the non-movant and clear error review can be significant: the former standard favors the loser; the latter favors the winner. There is no need to resolve this review standard to reach a remand decision in this case. Indeed, there is good reason not to do so. On remand, the factual record may change as the parties seek better to address Ricci's requirements. Moreover, the district court can then clarify what, if any, “factfinding” informs a new final judgment.
Second, the majority concludes that the Brennan Intervenors' claim of disparate-treatment discrimination is properly reviewed according to the three-step analytical framework outlined in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). The extensive discussion of McDonnell Douglas that accompanies this determination is not only unnecessary to our decision to vacate and remand, but it also risks confusing the question of discriminatory intent with the question of whether even actions taken with such intent can be excused by a purpose to avoid disparate-impact liability. McDonnell Douglas analysis was devised to answer the first question but, as a number of the judges of this court have recognized, it is not well-suited to answer the second. See Ricci v. DeStefano, 530 F.3d 88, 99–100 (2d Cir.2008) (Cabranes, J., with Jacobs, C.J., Raggi, Wesley, Hall, and Livingston, JJ.) (dissenting from denial of rehearing en banc ) (suggesting that issue should be resolved by mixed-motive analysis); accord id. at 89 (Calabresi, J.) (concurring in denial of rehearing en banc ) (agreeing that mixed-motive analysis should have been used but for parties' failure to present argument to district court or original appellate panel).
Significantly, in Ricci, the Supreme Court neither mentioned nor used McDonnell Douglas analysis in holding that plaintiffs were entitled to summary judgment on their claim of discriminatory treatment. See Ricci v. DeStefano, 129 S.Ct. at 2681. Rather, after observing that the City defended against the treatment claim by professing a purpose to avoid disparate-impact liability, see id. at 2673, the Court proceeded to consider the three factors relevant to a disparate-impact claim, see 42 U.S.C. § 2000e–2(k)(1)(A), and concluded that defendants lacked a strong basis in evidence to think that they would have been subjected to such liability had they failed to take the challenged race-conscious, discriminatory actions. See Ricci v. DeStefano, 129 S.Ct. at 2677–81.
No different approach is needed in this case, where defendants do not dispute that their challenged settlement actions were animated by intentional considerations of race, ethnicity, and gender. Their defense, as in Ricci, is that “the purpose to avoid disparate-impact liability excuses what otherwise would be prohibited disparate-treatment discrimination.” Id. at 2674. A decision to remand for reconsideration in light of Ricci does not require us to shoehorn this defense into the second step of McDonnell Douglas analysis. Indeed, the fit is awkward. At that step, a defendant need only articulate a nondiscriminatory purpose for his actions, not demonstrate that intentional discrimination is supported by a valid defense with a strong basis in evidence, as Ricci requires. See Holcomb v. Iona Coll., 521 F.3d 130, 141 (2d Cir.2008) (explaining that “[i]t is not our task, at the second stage of the McDonnell Douglas framework, ․ to determine whether the [defendant's] explanation of its action is convincing” bur rather only to “ask whether defendant has introduced evidence that, ‘taken as true, would permit the conclusion that there was a nondiscriminatory reason’ “ (quoting St. Mary's Honor Ctr. v. Hicks, 509 U.S. 502, 509, 113 S.Ct. 2742, 125 L.Ed.2d 407 (1993) (emphasis in Hicks )).
Nor does our decision to remand require us now to decide whether, despite a strong basis in evidence, a defendant's claimed purpose to avoid disparate-impact liability might be attacked as pretextual. See Ricci v. DeStefano, 129 S.Ct. at 2683 (Alito, J., concurring). Much less need we decide that such a pretext attack occurs at step two of McDonnell Douglas analysis, as the majority indicates.2
Because defendants' challenged settlement conduct was plainly animated by race, ethnicity, and gender considerations, here, as in Ricci, any “analysis begins,” not with McDonnell Douglas, but “with this premise: [defendants'] actions would violate the disparate-treatment prohibition of Title VII absent some valid defense.” Id. at 2673 (emphasis added). Thus, in ordering remand, we can leave it to the able district judge to decide in the first instance whether defendants have the strong basis in evidence necessary to pursue a defense of disparate-impact liability. See id. at 2677.3
Third, the majority engages in an extensive discussion of affirmative-action precedent and strongly suggests that such plans would continue to be judged by reference to standards derived from Johnson v. Transportation Agency, Santa Clara County, 480 U.S. 616, 107 S.Ct. 1442, 94 L.Ed.2d 615 (1987), and United Steelworkers of America v. Weber, 443 U.S. 193, 99 S.Ct. 2721, 61 L.Ed.2d 480 (1979), rather than Ricci. The matter is by no means clear. As the majority recognizes, the Supreme Court in Ricci did not so condition its ruling, signaling that “its core holding applies whenever an employer takes race conscious action ‘for the asserted purpose of avoiding or remedying an unintentional disparate impact.’ “ Ante at  (quoting Ricci v. DeStefano, 129 S.Ct. at 2677 (emphasis added by majority)). Because this case does not involve an affirmative-action plan, there is no need to “harmonize [ ]” Ricci with affirmative-action precedent to order remand, much less to do so in a way that could be construed to cabin Ricci to cases challenging make-whole individualized relief. Id. at . To be sure, the majority's earlier statement that “[i]n light of Ricci,” Johnson/Weber analysis “extends, at most, to circumstances in which an employer has undertaken a race-or gender-conscious affirmative action plan designed to benefit all members of a racial or gender class in a forward-looking manner only,” id. at  (emphasis in original), appears properly to leave for another day any controlling resolution of Ricci's application to affirmative-action challenges. But that is all the more reason for this panel to exercise restraint and to order remand without anticipating cases not now before us.
Fourth, the majority attempts to provide the district court with detailed guidance as to how to apply the Ricci strong-basis-in-evidence test to the record facts. This too is unnecessary to our decision to remand and puts the cart before the horse. However appropriate it may be for an appellate panel to provide instruction when a district court has misapplied Supreme Court precedent, such instruction is premature when a remand is ordered for application of a new Supreme Court decision. That is particularly so here, where, as I noted earlier, the Supreme Court signaled in Ricci that its own opinion was intended to provide the necessary “guidance” by “clarif[ying] how Title VII applies to resolve competing expectations under the disparate-treatment and disparate-impact provisions” of Title VII. Ricci v. DeStefano, 129 S.Ct. at 2674, 2681. In these circumstances, we should let Ricci speak for itself on remand without added gloss from this panel.
In particular, I question the majority's attempt to “hold” in the absence of any judgment applying Ricci, what is and is not sufficient to satisfy the strong-basis-in-evidence requirement. Ante at [76–77]. Holdings consist of determinations necessary or pivotal to a decision; judicial statements that are unnecessary to the decision in the case are mere dicta. See United States v. Rubin, 609 F.2d 51, 69 n. 2 (2d Cir.1979) (Friendly, J., concurring) (“A judge's power to bind is limited to the issue that is before him; he cannot transmute dictum into decision by waving a wand and uttering the word ‘hold.’ ”); see also Black's Law Dictionary 800 (defining “holding”), 1177 (defining “obiter dictum ”) (9th ed.2009). Our decision to remand this case for application of Ricci's strong-basis-in-evidence standard does not require us now to pronounce principles of sufficiency for identifying when that standard is met.
I am, moreover, dubious of the majority's pronouncement that, “under Ricci, a ‘strong basis in evidence’ “ requires “less than the preponderance of the evidence that would be necessary for actual liability.” Ante at [76–77]. To be sure, in Ricci, the Supreme Court stated that the strong-basis-in-evidence requirement is “not so restrictive that it allows employers to act only when there is a provable, actual violation.” Ricci v. DeStefano, 129 S.Ct. at 2674. But there is a difference between the preponderance finding of a violation made by a jury after trial and a court's identification of sufficient evidence to permit a preponderance finding at trial, a difference that I fear may be lost in the majority's broad “holding .” A party who cannot produce sufficient record evidence to support a possible preponderance finding by a jury on an issue on which the party bears the burden is going to wind up with summary judgment awarded against him. At some point, this court may have to consider whether a “strong basis in evidence” must be sufficient to survive a summary judgment motion. After all, on such a motion, evidence is viewed in the light most favorable to the non-moving party, which means that even some weak cases can survive summary judgment. We need not decide that question, however, to order remand in this case. Therefore, we also need not “hold” that Ricci can be satisfied on less evidence than would permit a preponderance finding at trial. Indeed, on remand, the district court may find support for the disparate-impact liability defense so compelling or deficient as not to require a precise definition of the standard of proof. That was, after all, the circumstance in Ricci.
Fifth, the majority also discusses in some detail various challenges that the Brennan Intervenors might raise to defendants' disparate-impact-liability defense. Again, this discussion is unnecessary to our decision to remand. Further, it is premature now to rule that certain challenges could not defeat the defense in this case, or to imply that certain evidence is inadmissible. These matters are best decided in the first instance by the district court. Certain challenges may be resolved on remand, whether as a matter of law or fact, in ways that will narrow or even eliminate the need for further appellate review.
As indicated at the outset, these concerns are illustrative and not exhaustive. Nevertheless, they suffice to explain why I join in the Court's decision to affirm in part and to vacate and remand in part without also joining in the majority opinion. Rather, I would simply conclude (1) that the district court judgment can be affirmed insofar as it (a) granted class certification to the Brennan Intervenors, and (b) declared Ciro Dellaporte not a member of a protected class entitled to relief under the challenged settlement, but (2) that the judgment must be vacated in all other respects and remanded for further consideration in light of Ricci.
CALABRESI, Circuit Judge:
Judge RAGGI, concurs in the judgment of the Court and files a separate opinion.