Meacham v. Knolls Atomic Power Laboratory , 381 F.3d 56 ( 2004 )

POOLER, Circuit Judge.

Plaintiffs are all former employees of defendant Knolls Atomic Power Laboratories (“KAPL”) who lost their jobs in the course of an involuntary reduction in force (“IRIF”). As all of the plaintiffs are over forty, they are protected under the Age Discrimination in Employment Act (“ADEA”). See 29 Ü.S.C. § 631(a). They sued KAPL, its president, John Freeh, and its parent company, Lockheed Martin Corporation, alleging age discrimination under both federal and state law. In particular, they claimed that KAPL designed and implemented its workforce reduction process to eliminate older employees and that, regardless of intent, the process had a discriminatory impact on ADEA-protect-ed employees. The jury rejected plaintiffs intentional discrimination claim but found that a facially neutral policy had a discriminatory impact on older employees and that KAPL could have accomplished its legitimate business goals by a method that was not discriminatory in its impact. The jury also found that defendants KAPL and Lockheed Martin acted willfully. Magistrate Judge David R. Homer, refused to set aside the verdict but ordered a new trial on damages for emotional suffering unless certain plaintiffs agreed to a reduction of those damages.

On appeal, KAPL attacks the verdict on several grounds. First, it urges that we reexamine and reject prior holdings of this court allowing an age discrimination claimant to proceed on a disparate impact theory. KAPL also contends that New York courts do not allow a disparate impact claim. Assuming the availability of a disparate impact claim to age discrimination plaintiffs, KAPL attacks the sufficiency of the evidence to support the disparate impact verdict because (a) the same practice — the' IRIF — cannot be the basis for both a disparate treatment and a disparate impact claim; (b) plaintiffs did not sufficiently specify the contested employment practice; (c) plaintiffs’ statistical evidence was inadequate and a portion of it was improperly admitted; (d) plaintiffs offered no evidence of an equally effective and no more costly alternative to the IRIF procedures; and (e) there was no competent evidence of willfulness. Finally, KAPL urges that portions of the damages award must be reduced or entirely set aside.

We affirm the verdict, holding that the arguments not waived by KAPL lack merit.

BACKGROUND

Relevant Facts

Our description of the background for this appeal is drawn from the trial testimony and exhibits as well as stipulated facts — viewed in light of the applicable standards of review, the limited number of issues on appeal, and questions concerning KAPL’s preservation of certain issues. Because the jury ruled in KAPL’s favor on the disparate treatment claim and plaintiffs question this verdict only in the alternative, we do not discuss facts relevant to it. Because the standard of review for denial of judgment as a matter of law (“JMOL”) requires us to credit the testimony that favors plaintiffs and does not allow us to reverse the jury’s verdict based on evidence the jury could have rejected, we do not set out, except in general terms, defendants’ refutation of plaintiffs’ witnesses’ testimony. See Tolbert v. Queens Coll., 242 F.3d 58, 70 (2d Cir.2001). And, because of the waiver issues, we address at some length the arguments defendants made in their motions for summary judgment and JMOL.

KAPL manages and operates the Knolls Atomic Power Laboratory, a research and development facility owned by the United States, under a contract with the Department of Energy (“DOE”). Its work is jointly funded by DOE and the United States Navy’s Nuclear Propulsion Program (collectively, “NR”), and consists of designing advanced nuclear propulsion systems, providing training to sailors to operate them, and overseeing maintenance, repair, refueling and decommissioning.

NR pays KAPL on a cost-plus basis for the work it performs. To control costs, NR, in consultation with KAPL, sets an annual staffing limit for the lab.¹ This limit is designed to adequately staff the work KAPL contracts to perform for NR. KAPL must stay within this ceiling.

For fiscal year 1996, NR set KAPL’s staffing level at 108 jobs below the previous year’s level. In addition, NR assigned KAPL new work that KAPL alleges required it to make thirty-five new hires. In combination, the staffing level reduction and the need for new hires required KAPL to eliminate 143 existing jobs. Thus, KAPL designed and implemented a Workforce Reduction Plan (“WRP”) providing the following methods to both pare the existing work force and acquire employees with unmet critical skills: (1) a voluntary separation plan (“VSP”) offering a $20,000 separation incentive for employees with at least twenty years of service; (2) transfers where possible to accommodate hiring needs; (3) retraining to allow existing employees to perform new skills; (4) an involuntary reduction in force; and (5) new hiring only when necessary for critical skills. KAPL consistently estimated that the average voluntary separation would cost it less than the average involuntary termination.

Prior to issuing the final workforce reduction plan, KAPL considered allowing all employees to participate in the VSP. KAPL’s human resources manager, Donald Burek, estimated that between 200 and 250 people would have elected to participate in such a plan. According to Freeh, KAPL rejected the idea of a VSP open to all employees except those with critical skills because, after speaking with other employers who had implemented such plans, he was concerned that denying the severance payment to persons with critical skills would cause a morale problem.

KAPL warned its employees that it could disapprove a VSP application on the basis of business needs. The company, in fact, denied at least one application because the employee had skills critical to the company.

By November 16, 1995, KAPL had completed the VSP process and allowed 107 employees to participate. The company then began to implement the IRIF component of its overall plan to reduce the workforce. Before the IRIF started, Burek made several different projections concerning its impact on the workforce by age group. The estimates of the participation by workers over forty in the IRIF ranged from 70% on the high end to 30% on the low end.

Initially, managers throughout KAPL determined which sections were facing an increased workload and which skills were excess. The particular sections, subsections, and units that would be impacted by the IRIF were determined by their staffing budgets. That is, if a section was over budget, it undertook an IRIF. If it was under budget, there was no IRIF even if workers in the section had skills that had been determined to be “excess.”

Over-budget managers were instructed to select employees for the IRIF by listing “all employees in [their] group[s] on [a] matrix”; ranking them between zero and ten for performance, flexibility, and criti-cality of their skills; and giving up to ten points for company service. Managers were to rate performance based on an average of the two most recent performance appraisals. The tests for making a flexibility determination were whether the employee’s “documented skills [could] be used in other assignments that [would] add value to current or future Lab work” and whether the employee was “retrainable for other Lab assignments.” Critical skills were those skills that were critical to continuing work in the Lab as a whole. In addition, KAPL directed managers to consider whether the “individuars skill [was] a key technical resource for the NR program” and whether “the skill [was] readily accessible within the Lab or generally available from the external market.” An employee would receive two points for service if she had between two and five years of service, four points for six to ten years of service, six points for eleven to fifteen years, eight points for sixteen to twenty years of service, and ten points for service above twenty years.

The guidelines further instructed the managers to identify for layoff the lowest ranked employees until its staffing was reduced to the required level. The managers were then to perform an adverse impact analysis to “determine whether [their] planned actions might have a disparate impact on a protected class of employees.” For minorities and females, the guidelines suggested using the EEOC’s four-fifths rule to determine whether an adverse impact existed. In explaining the rule, KAPL noted that “if the selection rate for a protected group is greater than 120% of the rate for the total population a serious discrepancy exists.” The instructions further provided: “A similar analysis should be performed to assure that we are not violating the Age Discrimination in Employment Act (‘ADEA’).”

A review board was to assess the manager’s selections “to assure adher[e]nce to downsizing principles as well as minimal impact on the business and employees.” Finally, KAPL’s general manager, John Freeh, and its chief counsel, Richard Cor-rea, were to review the final IRIF selections and the impact analyses.

The managers placed 245 of a total of approximately 2,063 exempt employees on the matrices and selected thirty-one for layoff. Thirty of the terminated employees were over forty.

After managers submitted the names of employees identified for layoff, KAPL performed an adverse impact analysis, but not the four-fifths analysis described in the Guidelines. Instead, Linda Geiszler, a human resources representative, compared the average age of the workforce before the IRIF with the average age of the workforce after the IRIF. Geiszler neither received instruction on how to conduct such an analysis from senior human resources personnel nor read any books or articles on how to do it. Although the review board considered the accuracy of managers’ ratings, it did not look at age discrimination issues.

Correa testified that although Freeh did not ask him to investigate why older workers were disproportionately represented in the layoffs, he did complete a legal review of the IRIF process. His paralegal checked the math on the matrices, and he spoke with human resource representatives and some, but not all, managers “to see if there was a legitimate, nondiserimi-natory reason for the selections.” When he did speak with managers, he questioned them only about the lowest four or five individuals on each matrix.

Although Correa was familiar with the ADEA, he explained that he did not look at who was placed on the matrices because “[w]hat was important was [that] the RIF decisions were properly made, they were legitimate decisions.” At one point, Cor-rea testified that “an impact analysis is only required for race and sex.” However, he later grudgingly acknowledged that “[r]ight now” disparate impact analysis was a viable age-discrimination theory in the Second Circuit. He further claimed that there was “no way to determine whether [Geiszler’s analysis was] appropriate or not, since there’s no guidance on it.” Correa concluded as a result of his review that the IRIF did not violate the ADEA.

Dr. Janice Madden, a qualified expert in the area of employment discrimination and an experienced statistician' who testified for the plaintiffs, analyzed the impact of several aspects of the IRIF on older employees. First, she compared the age composition of the pre-IRIF exempt workforce to the persons actually laid off. The pre-layoff exempt population consisted of approximately 40% workers under the age of forty and approximately 60% over forty. However, workers over forty constituted approximately ninety-eight percent of the laid-off workers. Madden estimated that the probability of this differential occurring by chance was approximately one in 348,000. Looking at the IRIF . section by section, Madden estimated the probability that thirty employees over forty would be laid off by chance as one in 6,639. Madden also testified that younger workers were less likely to get on the matrices in the first place. Of the 245 workers on the matrices, 179 were over forty. The odds of selecting thirty workers over forty from the 245 workers on the matrices were approximately one in 1,260.

Madden also testified that the matrices with the highest percentages of workers over forty resulted in the highest percentage of layoffs. However, even assuming that the matrices were drawn up properly, the odds of thirty older workers being laid off from these matrices were approximately one in seventy-three. Each of the disparities Madden found was statistically significant. She explained that a disparity is statistically significant if “the probability that something could happen by chance is less than 5 percent” and that each of the disparate outcomes she studied had a much lower than five percent chance of occurring randomly with the closest one being less than one percent.

Madden also considered each of the four criteria on the matrix and “found ... that the years of service score had no [ejffect. The performance score actually had very little [ejffect.... [TJhe two scores that were most responsible statistically for selecting who was RIFed among those individuals on the matrices [were] criticality and flexibility.” She also concluded that excess skills could not explain the age differential. Madden concluded that the procedures set up for review of the individual managers’ decisions “did not offer adequate protections to keep the prejudices of managers from influencing the outcome.”

Finally,' Madden criticized Geiszler’s analysis of adverse impact, stating that to do a proper adverse impact analysis, a statistician must compare the age composition of the pool from which the laid off employees were selected with the age composition of the group selected for layoff. Because KAPL’s exempt workforce exceeded 2,000 persons, Madden testified that “you literally could go from top to bottom of 31 people picking the oldest, you know, just blatantly and you wouldn’t affect the average age composition in any meaningful way.”

District Court Proceedings

In January 1997, appellants filed two separate age discrimination lawsuits in the United States District Court for the Northern District of New York. These two lawsuits, which made claims under the ADEA and New York’s Human Rights Law, were consolidated. Plaintiffs alleged both disparate treatment in the planning and implementation of the IRIF and disparate impact from the IRIF. Claiming that defendants’ violations of the ADEA were willful, plaintiffs demanded liquidated damages pursuant to 29 U.S.C. § 626(b). They also sought damages for emotional suffering under the Human Rights Law, front pay, back pay, and attorney’s fees.

On July 7, 1999, after discovery, defendants moved for summary judgment dismissing all of plaintiffs’ claims. As relevant to this appeal, they argued that the disparate impact claim should be dismissed because (1) a disparate impact claim is not cognizable under the ADEA, (2) plaintiffs failed to identify a facially neutral policy or practice; (3) plaintiffs did not establish that the design of the IRIF caused the disparate impact on older workers because plaintiffs’ expert failed to perform a regression analysis to rule out other causes; and (4) plaintiffs’ disparate impact claim was nothing “more than an ill-disguised disparate treatment claim” for which plaintiffs must prove discriminatory intent. To support their argument on the last point, KAPL cited Maresco v. Evans Chemetics, 964 F.2d 106, 115 (2d Cir.1992) for the proposition that a disparate impact claim must be dismissed where the alleged “facially neutral employment practice coalesces with the discharge which [the plaintiff] claims to have constituted disparate treatment.” The district court (Lawrence E. Kahn, Judge) denied defendants’ motion without opinion.

The parties consented to a jury trial before Magistrate Judge Homer, the liability portion of which began June 20, 2000. At the close of plaintiffs’ proof, defendants moved for judgment as a matter of law. With respect to the disparate impact claim, defense counsel argued that Dr. Madden’s testimony was defective because she failed te conduct a regression analysis to rule out factors other than age, including criticality, flexibility and performance. Because Madden purportedly did not perform a regression analysis, counsel argued that plaintiffs had not shown the policy caused the disproportionate number of layoffs of older employees.²

Plaintiffs’ counsel identified the challenged practice as both “the overall selection process of the people to be placed on the matrices and what’s on the matrices, the scoring of those people” and “each element within the overall practice.” Counsel also argued that the statistical disparity itself was sufficient to demonstrate causation. As alternative practices to the IRIF, plaintiffs’ counsel suggested (1) opening the VSP to the entire workforce except those with critical skills and (2) a hiring freeze.

In rebuttal, defense counsel criticized plaintiffs’ reliance on individual elements of the IRIF, arguing that this reliance was “in fact, a treatment case sort of masquerading as an impact case.” Defense counsel also argued that plaintiffs could not rely on the IRIF as a whole because they had not proven it was incapable of separation for analysis.

Judge Homer denied defendants’ motion. Before the case was submitted to the jury, defendants renewed their Rule 50(b) motion. Again, defense counsel argued that plaintiffs’ case failed to the extent that they relied on the overall IRIF because plaintiffs did not show that the IRIF was incapable of separation for analysis. Counsel also argued once more that Dr. Madden did not establish causation. Finally, counsel contended that plaintiffs had not met their burden of showing an equally effective alternative to the IRIF. Judge Homer denied the motion.

On July 26, 2000, the jury returned a verdict in defendants’ favor on the disparate treatment counts. However, the jury also found that plaintiffs had “proven that a specific employment practice of the defendants ... although non-discriminatory on its face, had an adverse impact on the plaintiffs because of their ages”; that defendants failed to “articulate[ ] a business justification for selecting the plaintiffs for termination”; that plaintiffs had proven “that an alternative practice would have been equally effective in achieving the defendants’ legitimate employment goals as the method actually followed by the defendants” and that defendants acted willfully.

Prior to the beginning of the damages phase of the trial, eight of the plaintiffs settled with the defendants. The jury awarded the remaining plaintiffs damages ranging from approximately $69,000 to over $1.1 million.

After the damages phase, defendants renewed their motion for JMOL. On liability, they contested the sufficiency of the evidence on the grounds that (1) plaintiffs impermissibly relied on the overall selection process and (2) plaintiffs’ expert testimony was inadequate because she did not identify the correct population or rule out factors other than age as the cause of the disproportionate number of protected employees selected for layoff. Defendants also argued that there was- a legally insufficient basis for the jury’s findings that (1) defendants failed to articulate a business justification; (2) plaintiffs proved an alternative practice would have been equally effective in achieving the defendants’ legitimate employment goals; and (3) defendants acted willfully.

On damages, KAPL contended that there was no evidence to support the front pay award for seven plaintiffs or the back pay awards to ten plaintiffs. Defendants also attacked all the prospective mental anguish awards and sought reduction of awards for past suffering to between $5,000 and $30,000.

KAPL alternatively moved for a new trial pursuant to Federal Rule of Civil Procedure 59, contending that the verdict was against the weight of the evidence and that the court’s admission of Madden’s testimony — particularly her testimony on the regression analyses — was prejudicial error. KAPL also sought a new trial or remittitur on some of the damages issues.

In a published opinion, Judge Homer denied KAPL’s motion, with the sole exception of its request for remittitur on mental anguish damages. Meacham v. Knolls Atomic Power Lab., 185 F.Supp.2d 193, 245 (N.D.N.Y.2002). ’ The judge noted preliminarily that “[ajlthough courts of appeals are divided over the viability of the disparate impact theory under the ADEA, this theory of proof remains available in the Second Circuit.” Id. at 206. He then addressed KAPL’s various claims concerning the disparate impact verdict. He held that the plaintiffs’ prima facie case was sufficient, finding that (1) in this case, the entire IRIF could serve as a facially neutral employment policy because Madden’s testimony established that “no particular factor and no particular criterion caused the disparate impact on older employees”, id. at 208; (2) either the entire exempt work force or the exempt employees actually placed on the matrices properly could have been considered by the jury as the correct population for purposes of comparison to the plaintiffs; (3) Madden’s results demonstrated statistical significance, id. at 210-11; and (4) Madden’s testimony concerning her regression analyses was properly admitted because KAPL opened the door and, in any event, it was not prejudicial, id. at 211-12.

Judge Homer agreed with defendants that the jury erroneously found that KAPL had not offered a legitimate business justification. Id. at 213. However, the “erroneous finding [did] not merit judgment as a matter of law or a new trial” because the jury also found that plaintiffs offered proof of alternative practices that would be equally effective and no more costly than the IRIF. Id. at 213-14. The judge also held that the jury permissibly could have found that either a hiring freeze or an extension of the VSP would have permitted KAPL to stay within its staffing budget and retain critical skills at less cost than the IRIF. Id. at 215.

KAPL had argued that the willfulness verdict must be set aside both because it was inconsistent with the disparate treatment verdict and because it was unsupported by the evidence. Judge Homer rejected the inconsistency challenge based on the different mens rea standards applicable to disparate treatment and willfulness. He sustained the willfulness verdict based on (1) KAPL’s awareness that the ADEA prohibited employment practices that disparately impacted older employees; (2) defendants’ knowledge that implementation of the IRIF would result in a disproportionate impact on older employees and their failure to take “meaningful steps to avoid or mitigate that impact”; (3) KAPL’s use of a statistical measure for disparate impact that it knew “was incapable of identifying any disparate impact on older employees”; and (4) KAPL’s post-IRIF hiring of younger employees. Id. at 215-16.

Primarily on waiver grounds, Judge Homer rejected plaintiffs’ contention that disparate impact claims are not actionable under New York’s Human Rights Law. Id. at 216-17. Alternatively, he held that age discrimination disparate impact claims are actionable under the Human Rights Law. Id. at 217.

With respect to damages, the district court rejected all of KAPL’s front and back pay challenges but granted relief with respect to some plaintiffs’ mental anguish damages. Specifically, if the jury had awarded a plaintiff who had not offered evidence of treatment or of physical sequelae more than $125,000 for mental anguish, the court ordered a new trial unless the plaintiff agreed to accept a damages amount of $125,000. If a plaintiff who had offered proof of treatment or physical impact received more than $175,000 in emotional suffering damages, the court set a remittitur equal to that amount. Id. at 221-37.

KAPL filed a timely notice of appeal and plaintiffs a timely notice of cross-appeal. KAPL’s principal arguments on appeal are: (1) the ADEA does not support a cause of action based on disparate impact; (2) disparate impact is not cognizable under the Human Rights Law; (3) there is no evidence of disparate impact because (a) plaintiffs failed to isolate and identify a facially neutral employment process; (b) Madden did not demonstrate causation; (c) Madden’s post-discovery regression analysis was erroneously admitted and not probative of causation; and (d) plaintiffs did not establish an equally effective and no more costly alternative to the IRIF; (4) there is no evidence of willfulness; (5) the willfulness verdict is inconsistent with the disparate treatment verdict; (6) the emotional damages awards are excessive; and (7) there was no evidence to support the front pay awards to four plaintiffs. Plaintiffs seek to maintain their cross-appeal, which rests on two asserted evidentiary errors, only if defendants prevail on the principal appeal.

DISCUSSION

I. Principles implicated in our review.

This appeal requires us to apply general principles concerning waiver, stare decisis, and sufficiency of the evidence to support a jury verdict which we will briefly describe.

Plaintiffs argue that KAPL waived several issues by failing to raise them in its motion for JMOL. “[A] party is not entitled to challenge on appeal the sufficiency of the evidence to support the jury’s verdict unless it has timely moved in the district court for judgment as a matter of law on that issue.” Kirsch v. Fleet St., Ltd., 148 F.3d 149, 164 (2d Cir.1998). To be timely, the motion must be made before the jury retires so that the non-moving party has an opportunity to remedy any defects in its proof. Id. The motion “must at least identify the specific element that the defendant contends is insufficiently supported.” Id. (quoting Galdieri-Ambrosini v. National Realty & Dev. Corp., 136 F.3d 276, 286 (2d Cir.1998)). We will reach the waived issue if to ignore it would result in manifest injustice. Id. We also may correct an error below despite the lack of a timely request in the district court if the issue is purely legal and no fact finding is required. Baker v. Dorfman, 239 F.3d 415, 420 (2d Cir.2000). However, we are much less likely to consider a waived claim “involving] a complex and novel issue of state law.” Pulvers v. First UNUM Life Ins. Co., 210 F.3d 89, 96 (2d Cir.2000).

Stare decisis, and in particular, our obligation to follow the rulings of prior panels, also plays a role in this case. We are compelled to . follow a decision of an earlier panel “unless it has been called into question by an intervening Supreme Court decision or by one of this Court sitting in banc,” United States v. Santiago, 268 F.3d 151, 154 (2d Cir.2001), or “unless and until its rationale is overruled, implicitly, or expressly, by the Supreme Court, or this court in banc.” In re Sokolowski, 205 F.3d 532, 534-35 (2d Cir.2000) (per cu-riam).

Where we reach the merits, we review the district court’s denial of JMOL de novo. Yurman Design, Inc. v. PAJ, Inc., 262 F.3d 101, 108 (2d Cir.2001). The district court, and thus this court, can grant the motion only if after viewing the evidence in the light most favorable to the non-moving party and drawing all reasonable inferences in favor of the non-moving party, it finds that there was insufficient evidence to support the verdict. Tolbert v. Queens Coll., 242 F.3d 58, 70 (2d Cir.2001). Neither the district court nor this court can set aside the jury’s credibility findings or find for the movant based on evidence the jury was entitled to discredit. Id.

II. Disparate impact under the ADEA.

KAPL contends, based on Hazen Paper Co. v. Biggins, 507 U.S. 604, 113 S.Ct. 1701, 123 L.Ed.2d 338 (1993) and decisions from other circuits³ that we should disavow our own precedent holding that policies having a disparate impact on older workers are actionable under the ADEA. According to plaintiffs, this argument was waived by KAPL’s failure to raise it before submission to the jury, and is barred by stare decisis. Alternatively they claim that the argument lacks merit.

At summary judgment, KAPL acknowledged that this circuit recognizes ADEA disparate impact claims, but, in light of the authority to the contrary, did not concede that an ADEA plaintiff could make a disparate impact claim. Defs.’ Mem. Supp. Summ. J. at 10-11. However, in moving for judgment as a matter of law on plaintiffs’ disparate impact claim, KAPL did not preserve this argument. Nor was the argument made in KAPL’s trial memorandum.

KAPL’s failure to move for JMOL based on the unavailability of an ADEA disparate impact claim constitutes a waiver. See, e.g., Kirsch, 148 F.3d at 169-70. We do not believe the attempted preservation of the issue at the summary judgment stage avoids the procedural default. Cf. Becker v. Poling Transp. Corp., 356 F.3d 381, 391 (2d Cir.2004) (stating that “the denial of a motion for summary judgment is moot in light of the fact the case has since been tried before a jury”). Nevertheless, the waiver rule is not an absolute one. Here, the district court considered whether a disparate impact claim was tenable under the ADEA notwithstanding KAPL’s failure to raise the issue. The district court’s decision allows us to proceed with review. See Stevens v. Department of Treasury, 500 U.S. 1, 8, 111 S.Ct. 1562, 114 L.Ed.2d 1 (1991) (considering the merits of an issue where — despite party’s failure to present it to the courts below— those courts decided it). Moreover, the waived issue is easily resolved under existing circuit precedent. Therefore, we will address it.

Stare decisis requires us to hold that the ADEA allows disparate impact claims. As defendants concede, we have repeatedly so held. See, e.g., Smith v. Xerox Corp., 196 F.3d. 358, 364 (2d Cir.1999); Criley v. Delta Air Lines, Inc., 119 F.3d 102, 105 (2d Cir.1997) (per curiam); Maresco, 964 F.2d at 115. We can reject this precedent only if a Supreme Court decision or a' decision from this circuit sitting en bane implicitly or explicitly overrules it. See Santiago, 268 F.3d at 154; Sokolowski, 205 F.3d at 534-35.

In Hazen Paper, a disparate treatment case, the Supreme Court held that an employer may lawfully discriminate, based on factors closely correlated with age. 507 U.S. at 611-13, 113 S.Ct. 1701. KAPL argues that this holding and dicta in Hazen Paper suggest that the Supreme Court ultimately will reject ADEA disparate impact claims. While this may be so, the Hazen Paper Court directly stated that it had not resolved the viability of ADEA disparate impact claims.⁴ 507 U.S. at 610, 113 S.Ct. 1701. The Supreme Court’s express statement contradicts any claim that Hazen Paper expressly or implicitly overruled our prior precedent. And, the Supreme Court’s dicta do not outweigh prior circuit authority. See Smith v. City of Des Moines, 99 F.3d 1466, 1469-70 (8th Cir.1996) (following prior case law allowing an ADEA claim for disparate impact despite its conclusion that dicta in Hazen Paper cast doubt on its viability); McLaughlin v. North Carolina Bd. of Elections, 65 F.3d 1215, 1225-26 (4th Cir.1995) (“Although subsequent Supreme Court opinions might be read to cast some doubt upon the constitutionality of the election scheme [we] upheld in [a previous case] ... we feel bound to treat [that] result ... as controlling unless and until the Supreme Court provides a much clearer signal to the contrary.”); Bronx Household of Faith v. Community Sch. Dist. No. 10, 127 F.3d 207, 218 (1997) (Cabranes, J., concurring in part and dissenting in part) (“we are not free to follow the Supreme Court’s strong suggestion in what is concededly dicta”). Of course, the decisions of other circuits do not expressly or implicitly overrule our prior cases. Therefore, we reject KAPL’s contention that disparate impact claims are not permissible under the ADEA.

III. Disparate impact under New York’s Human Rights law.

KAPL next argues that we should set aside the jury’s Human Rights Law verdict, which is the sole source of plaintiffs’ emotional suffering damages. In light of the state age discrimination law’s coverage of anyone over the age of eighteen, N.Y. Exec. L. § 296(3-a)(a), and the law in this circuit indicating that ADEA disparate impact plaintiffs must show that a challenged policy has an impact on the entire class, see Criley, 119 F.3d at 105, KAPL contends that a disparate impact claim is conceptually impossible. That is, no one employment policy can discriminate against everyone over the age of eighteen. At least one of New York’s four intermediate appellate courts has accepted this premise. See Bohlke v. General Elec. Co., 293 A.D.2d 198, 742 N.Y.S.2d 131, 132 (2002). However, the question has not been addressed by New York’s highest court. See Becker v. City of New York, 249 A.D.2d 96, 671 N.Y.S.2d 88, 90 (1998) (noting that “[n]either the United States Supreme Court nor the Court of Appeals has ruled conclusively that disparate impact, as opposed to disparate treatment, constitutes age discrimination”).

KAPL concedes that it did not raise this issue until its post-verdict motion but argues that it would constitute manifest injustice to allow- the state law verdict to stand. Unlike KAPL’s claim that the ADEA does not extend to disparate impact claims, this argument cannot be answered by reference to our decisions. It is also not easily resolvable by construing state law or cases. We cannot easily predict whether New York’s Court of Appeals would apply Criley to its own statute as the federal and state statutes differ. The federal statute creates a class of older employees who are protected while the state statute prohibits discrimination on account of age against both older and younger workers. Compare 29 U.S.C. § 631(a) with N.Y. Exec. L. § 296(3-a)(a). The general principle that we construe the Human Rights law just as we do the ADEA, see Smith, 196 F.3d at 363 n. 1, has little application to these two quite different statutory provisions. Although Bohlke may foreshadow the New York Court of Appeals’ resolution of this issue, New York’s highest court' also could find that a policy having a disparate impact on all persons between, for example, fifty and sixty was actionable even though it had no impact on employees between eighteen and twenty-eight. Because the New York courts should resolve this issue, we decline to address it in the face of KAPL’s acknowledged waiver. See Pulvers, 210 F.3d at 96.

IV. Adequacy of the proof of disparate impact.

To prove disparate impact, plaintiffs must first identify the specific policy or policies responsible for the disparate impact. Smith, 196 F.3d at 367-68. They cannot rely on the company’s overall decision making process for a reduction ip force unless they “can show that the elements of the employer’s decision-making process are not capable of separation for analysis.” Id. An employer’s subjective decision-making process is a proper subject for a disparate impact analysis. Watson v. Fort Worth Bank and Trust, 487 U.S. 977, 990, 108 S.Ct. 2777, 101 L.Ed.2d 827 (1988).

After identifying the relevant policy, the plaintiffs must choose the correct population for statistical analysis. Smith, 196 F.3d at 368. In a reduction in force, this population will generally be the population subject to termination, here either KAPL’s entire exempt workforce or those exempt employees actually placed on the matrices. Id. Plaintiffs’ expert then must compare the retention rate of protected employees to that of the non-protected employees. Id.

If the comparison of retention rates demonstrates “a statistical disparity ... sufficiently substantial to raise an inference of causation,” defendant must “explain the business necessity of the challenged employment practice.” Id. at 365. After the defendants meet their burden, plaintiffs may prevail only “if [they] can show that the employer’s proffered explanation was merely a pretext for discrimination.” Id. They can do this by “showing] that another practice would achieve the same result at comparable cost without causing a disparate impact on the protected group.” Id.

KAPL contends that plaintiffs’ proof was deficient in several respects, the first of which is identification of the policy or practice causing the IRIF. They claim that plaintiffs (1) incorrectly focused on the entire IRIF rather than particular aspects of it and (2) impermissibly named the same practice — the design of the IRIF — as the focus of both their disparate impact and disparate treatment claims.

Plaintiffs asserts that KAPL’s second argument, which focuses on Maresco, has been waived. In Maresco, we said:

Both the disparate treatment and disparate impact theories can be invoked in a given case to establish ADEA liability, since they are simply alternative doctrinal premises for a statutory violation. We do not believe, however, that the disparate impact theory provides any significant analytical contribution in this case.

The facially neutral employment policy that Maresco invokes as the premise for disparate impact liability coalesces with the discharge which he claims to have constituted disparate treatment. Because evidence of the employer’s subjective intent to discriminate must be provided to support a claim of disparate treatment, allowing the disparate impact doctrine to be invoked as Maresco proposes would simply provide a means to circumvent the subjective intent requirement in any disparate treatment case. Furthermore, the eight discharges that occurred in connection with the divisional consolidation at issue in this case are unlikely to provide an adequate basis for the sort of statistical analysis frequently employed in disparate impact cases.

964 F.2d at 115 (internal citations and quotation marks omitted). In KAPL’s view, Maresco means that the same policy or practice can never be the focus of both a disparate impact and a disparate treatment claim. That is, an age discrimination plaintiff cannot argue alternatively that the defendants designed their policy with discriminatory intent and that, regardless of intent, the policy had a discriminatory impact. The preliminary question before us is whether KAPL sufficiently alerted the district court to its argument.

KAPL’s motion for summary judgment relied on Maresco for the proposition that plaintiffs had made only a disparate treatment claim. KAPL did not explicitly argue that plaintiffs can never maintain a disparate treatment and a disparate impact claim based on the same policy. See Defs.’ Mem. Supp. Summ. J. at 18-19. However, KAPL did describe Maresco as “dismissing a disparate impact claim where the plaintiffs ‘facially neutral practice ... coalesces with the discharge which he claims to have constituted disparate treatment.’ ” Id. at 19. (quoting Maresco, 964 F.2d at 115). KAPL’s references to Maresco at trial and before the jury retired were even more oblique. In its trial memorandum of law, KAPL did not cite Maresco and did not claim that disparate impact and disparate treatment claims could never focus on the same policy or practice. Instead, it argued that plaintiffs’ disparate impact claim was really a disparate treatment claim in disguise. In its initial Rule 50 motion, KAPL argued:

Number two, [plaintiffs’ counsel] pointed out something also that relates to the [Smith ] case. He spoke of the individual elements that also they are complaining about, sort of the individual flexibility, criticality, as applied. This impact case, I think is, in fact, a treatment case sort of masquerading as an impact case. They are complaining about managers subjectively applying their judgments to these decisions, which certainly can allow one to, if you’ve got animus, to inject it into the process. But that’s treatment, sir. That’s motivation. That’s not adverse impact. That’s not statistics. And that’s all I want to say.

Tr. at 2486.

We believe that the district court would have understood KAPL to have argued that because plaintiffs focused on individual judgments by individual managers, they did not make out a disparate impact claim. KAPL never argued clearly that one policy could not be the focus of both a disparate impact claim and a disparate treatment claim. Thus, we hold that KAPL waived its Maresco claim. Had KAPL made its argument clearly, even without citation to Maresco, we would consider the Maresco argument on the merits. However, whatever the outside boundaries of a party’s obligation to specify its objections, at a minimum, the party must have articulated the objection so that a reasonable judge would understand it. KAPL did not meet this standard.

We decline to exercise our discretionary jurisdiction to reach the Maresco claim.⁵

We turn now to plaintiffs’ purported failure to identify a policy other than the overall IRIF as having a disparate impact. As discussed previously, we held in Smith that a plaintiff cannot rely on his employer’s overall decision-making process for a reduction in force unless he “can show that the elements of the employer’s decision-making process are not capable of separation for analysis.” 196 F.3d at 368. Plaintiffs obligation to identify a specific employment practice causing a disparate impact stems from Watson. In that case, four justices of the Supreme Court said:

The plaintiff must begin by identifying the specific employment practice that is challenged. Although this has been relatively easy to do in challenges to standardized tests, it may sometimes be more difficult when subjective selection criteria are at issue. Especially in cases where an employer combines subjective criteria with the use of more standardized rules or tests, the plaintiff is in our view responsible for isolating and identifying the specific employment practices that are allegedly responsible for any observed statistical disparities.

487 U.S. at 994, 108 S.Ct. 2777 (O’Connor, Justice,-writing for herself, the Chief Justice, and Justices White and Scalia).

One year later, in Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 109 5.Ct. 2115, 104 L.Ed.2d 733 (1989), the Court considered plaintiffs’ claim that several hiring practices including nepotism, separate hiring channels, rehire preferences, and subjective decision making resulted in a racially imbalanced workforce. Id. The Wards Cove court rejected plaintiffs’ global approach, requiring instead that they demonstrate that “specific elements of the [employer’s] hiring process have a significantly disparate impact on nonwhites.” Id. at 658, 109 S.Ct. 2115. Thus, although subjective decision making is subject to a disparate impact analysis, it must be shown that it is the subjective elements of the decision-making process that caused the disparate impact.

With these principles in mind, we examine in the light most favorable to plaintiffs, their statistical proof. At the outset, we note that KAPL’s IRIF was broken down and analyzed by Dr. Madden as a four-step process, and that the structure of the IRIF lent itself to that analysis; plaintiffs were thus required to “isolate [ ] and identify [ ] ... specific employment practices” allegedly responsible for the IRIF’s lopsided outcome. Watson, 487 U.S. at 994, 108 S.Ct. 2777. Plaintiffs met this burden. A reasonable juror could have viewed Dr. Madden’s testimony as establishing that (1) the procedures used to select employees to be placed on the matrices selected older employees at a disproportionate rate; (2) older employees were selected from the matrices at a rate disproportionate to their numbers on the matrices; (3) of the criteria listed on the matrices, managers’ judgments on flexibility and criticality had the greatest impact on the persons selected for layoff; (4) one basic flaw in the IRIF process was the degree of subjective decision making it allowed to individual managers to set flexibility and criticality scores; and (5) another was the failure to audit for that tendency. Thus, the jury could have found that the degree of subjective decision making allowed in the IRIF created the disparity.

But sustaining the jury’s causation finding is insufficient to preserve the verdict against KAPL because, as the district court ruled, the lab offered a facially legitimate business justification for the IRIF and its constituent parts: “to reduce its workforce while still retaining employees with skills critical to the performance of KAPL’s functions.” Meacham, 185 F.Supp.2d at 213.⁶ Unchallenged, KAPL’s justification would preclude a finding of disparate impact. However, plaintiffs prevail nonetheless if they demonstrate that KAPL’s justifications are a pretext for discrimination by challenging their veracity or by showing “that another practice would achieve the same result at a comparable cost without having a disparate impact on the protected group.” Smith, 196 F.3d at 365.

The district court ruled that the jury could have found that the availability of two suitable alternatives—a hiring freeze and a selective expansion of the VSP— exposed KAPL’s justification as pretextual. Meacham, 185 F.Supp.2d at 214-15. We affirm this result on different grounds, notwithstanding that Judge Homer’s ruling was erroneous. A hiring freeze or an extended VSP might have been cost effective, but they are not alternatives to the specific employment practices plaintiffs identified as discriminatory: unaudited and heavy reliance on subjective assessments of “criticality” and “flexibility.” Rather, they are replacements for the IRIF itself, which was (and is) a personnel decision justified by KAPL’s business objectives. See, e.g., Atonio v. Wards Cove Packing Co., Inc., 10 F.3d 1485, 1502-03 (9th Cir.1993).

Faced with the need to cut its workforce and simultaneously to retain employees with critical skills, KAPL undertook a multi-step workforce reduction plan that included a VSP, transfers, retraining, a limitation on new hiring to candidates with critical skills, and—finally— the IRIF. Employers are free to decide that layoffs are necessary; and “criticality” and “flexibility” may be appropriate criteria to use in making a termination decision. But if a particular criterion is subjective (as “flexibility” and “criticality” are), and if (as here) evidence shows that (i) the subjectivity disproportionately impacted older employees; (ii) the employer observed that the disproportion was gross and obvious; and (iii) the employer did nothing to audit or validate the results, then an employer may be liable for discrimination if equally effective alternatives to the challenged features of the employment practice are available.

At least one suitable alternative is clear from the record: KAPL could have designed an IRIF with more safeguards against subjectivity, in particular, tests for criticality and flexibility that are less vulnerable to managerial bias. For instance, KAPL’s in-house counsel, Richard Correa, spoke to certain managers about certain scores, but his review was far from systematic and appears to have consisted primarily of having a paralegal check the managers’ math. A sample of the factors considered in scoring flexibility shows that the criterion was imprecise at best.⁷ This evidence, combined with Dr. Madden’s testimony and KAPL’s inadequate response when the criticality and flexibility criteria led to extremely skewed results (see the discussion of willfulness below) could easily have led the jury to conclude that KAPL could have made simple adjustments to the criticality and flexibility criteria that would have led to a nondiscriminatory distribution of layoffs.⁸

We turn now to KAPL’s claim that because Dr. Madden did not rule out other causes, plaintiffs failed to establish that KAPL’s facially neutral IRIF policy caused the disparate impact on older workers. This argument misconstrues the role statistics play in a disparate impact analysis. Statistical disparity alone — assuming it is substantial as it was here — raises an inference of causation. Smith, 196 F.3d at 365. The Second Circuit cases that KAPL cites for its argument to the contrary are all disparate treatment cases, not disparate impact cases, and (in part because defendants failed to preserve the Maresco issue) are not relevant to this disparate impact case. See Appellants’ Br. at 47 (citing Bickerstaff v. Vassar College, 196 F.3d 435, 445-50 (2d Cir.1999); Hollander v. American Cyanamid Co., 172 F.3d 192, 198 (2d Cir.1999); Raskin v. Wyatt Co., 125 F.3d 55, 60, 68 (2d Cir.1997)).

KAPL also argues that Madden’s testimony concerning regression analyses was erroneously admitted and that it prejudiced KAPL. Madden admitted at her deposition that she had not performed any regression analyses. Armed with this admission, defense counsel sought to undermine her direct testimony by asking whether she had performed any regression analyses. She answered that she had.⁹ On redirect, plaintiffs’ counsel asked Madden to describe the analyses she had performed. Over KAPL’s objections, the court allowed the testimony but provided that KAPL could recall Madden at a later time and at plaintiffs’ expense. Madden then testified that she had found that older workers’ performance skills declined disproportionately to those of younger workers between evaluations conducted in 1991 and evaluations conducted in 1995. She also testified that flexibility and criticality scores were much better predictors of who would be laid off than were performance and years of service scores.

We review the district court’s evidentiary rulings for abuse of discretion, see Wolak v. Spucci, 217 F.3d 157, 161 (2d Cir.2000), and we reverse only if refusal to do so is “inconsistent with substantial justice,” Fed.R.Civ.P. 61. Thus, although Federal Rule of Civil Procedure 26(a)(2) requires disclosure and supplementation as needed of expert reports, any non-prejudicial error the district court may have made does not require reversal. The district court gave KAPL a chance to remediate any prejudice by recalling Madden at plaintiffs’ expense. Although KAPL chose not to recall Madden,' it did offer extensive testimony from its own expert to rebut Madden’s conclusions. Therefore, plaintiffs gained no unfair advantage from their partial failure to disclose.

Having addressed KAPL’s objections to the jury’s disparate impact verdict, we affirm it.

V. The willfulness verdict.

An employee is entitled to liquidated damages equal to his back pay for willful violations of the ADEA. 29 U.S.C. §§ 216(b), 626(b). KAPL challenges the jury’s willfulness verdict arguing that (1) it is inconsistent with the jury’s finding that KAPL did not intentionally discriminate and (2) in light of KAPL’s careful IRIF plan, there was no evidence to support the verdict.

There is no inconsistency in the verdicts. Disparate treatment liability requires that the defendant employer intend to discriminate on the basis of age, Cronin v. Aetna Life Ins. Co. 46 F.3d 196, 204 (2d Cir.1995), while an ADEA violation is willful if the employer “knew or showed reckless disregard for the matter of whether its conduct was prohibited by the statute,” Hazen Paper, 507 U.S. at 614, 113 S.Ct. 1701. Given these different standards, a jury could reasonably find both that KAPL did not intentionally discriminate on the basis of age and that it showed reckless disregard for the impact of its IRIF policy on older employees.

We also hold that the jury had sufficient proof of reckless disregard to find willfulness. KAPL’s counsel knew that this circuit recognizes a claim for disparate impact under the ADEA. In addition, the IRIF guidelines instructed managers to test for age discrimination, by use of the 4/5 test. Geiszler, however, did not use the 4/5 test. . Instead, she compared the average age of KAPL’s 2000-plus employees before and after the IRIF and found no significant difference. Not only did Madden testify to this method’s gross inadequacy, but it is also inconsistent with the methodology we have approved. See Smith, 196 F.3d at 365-66 (approving as methods for judging disparate impact, the four-fifths test and the statistical approach used by Madden).

KAPL knew that its IRIF had affected a disproportionate number of older employees. A reasonable jury could have found that KAPL management’s failure to challenge Geiszler’s patently inadequate approach evinced a desire not to know that the overwhelming — and overwhelmingly disparate — impact of the IRIF was on older workers, and that KAPL’s failure to modify its procedures in the face, of this disparity reflected a reckless disregard for whether it had violated the ADEA by implementing a policy with such an impact.

Seen in the light most favorable to plaintiffs, the proof demonstrated that KAPL’s procedures for guarding against disparate impact age discrimination were inadequate. Some managers who selected employees for termination testified that they received no training on avoiding age discrimination in the IRIF. The review board also received no training on age discrimination and did not monitor for age discrimination. And, as we have outlined, Cor-rea’s review was quite limited. The jury reasonably could have found that this proof, too, supported a conclusion that KAPL managers acted in reckless disregard of their obligations under the ADEA.

Because there was sufficient proof of reckless disregard to enable a jury to find that KAPL acted willfully and there is no inconsistency between the disparate treatment and willfulness verdicts, we affirm the jury’s verdict on willfulness.

VI. Damages.

A. Mental anguish.

Where plaintiffs offered no proof other than testimony establishing shock, nightmares, sleeplessness, humiliation, and other subjective distress, Judge Homer limited their mental anguish damages to $125,000. The judge limited damages where either physical sequelae or professional treatment was established to $175,000. In some instances, these caps significantly reduced the amounts the jury had awarded. KAPL contends, however, that all awards should have been reduced to between $5,000 and $30,000. In KAPL’s view, the damages must be reduced to this level because (1) the district court failed to address the damages in light of KAPL’s good faith and (2) plaintiffs’ claims do not rise above garden variety emotional distress claims for which New York courts traditionally uphold awards of no more than $30,000.

In determining whether the mental anguish damages, which stem exclusively from plaintiffs’ state law claim, should have been further reduced, we must use New York standards. Gasperini v. Center for Humanities, 518 U.S. 415, 431, 116 S.Ct. 2211, 135 L.Ed.2d 659 (1996). New York law provides that jury verdicts may be set aside and new trials ordered where the jury’s award “deviates materially from what would be reasonable compensation.” N.Y. C.P.L.R. § 5501(c). Mental anguish damages must also be “reasonably related to the discriminatory conduct” of the defendant. In re New York City Transit Auth., 78 N.Y.2d 207, 573 N.Y.S.2d 49, 54, 577 N.E.2d 40 (1991). The reviewing court looks to “the duration of a complainant’s condition, its severity or consequences, any physical manifestations, and any medical treatment.” Id. at 55. Finally, the reviewing court must determine how the award compares with others awarded for similar injuries and whether it is supported by evidence before the jury. Id.

New York cases vary widely in the amount of damages awarded for mental anguish. Many do reduce awards to $30,000 or below. See, e.g., In re Buffalo Athletic Club, 249 A.D.2d 986, 672 N.Y.S.2d 210, 211 (1998); In re Manhattan and Bronx Surface Transit Operating Auth., 220 A.D.2d 668, 632 N.Y.S.2d 642, 644 (1995); In re New York State Office of Mental Retardation and Developmental Disabilities, 183 A.D.2d 943, 583 N.Y.S.2d 580, 582 (1992); In re City of Fulton, 221 A.D.2d 971, 633 N.Y.S.2d 914, 915 (1995). However, other cases uphold awards of more than $100,000 without discussion of protracted suffering, truly egregious conduct, or medical treatment. See, e.g., Rio Mar Rest. v. NYSDHR, 270 A.D.2d 47, 704 N.Y.S.2d 230, 231 (2000); In re Allender, 233 A.D.2d 153, 649 N.Y.S.2d 144, 145 (1996); Boutique Indus., Inc. v. NYSDHR, 228 A.D.2d 171, 643 N.Y.S.2d 986, 986 (1996). For truly egregious conduct with severe and verified results on a complainant’s mental and physical health, courts have upheld awards far in excess of the amounts upheld here. See, e.g., In re New York City Transit Auth., 181 A.D.2d 891, 581 N.Y.S.2d 426, 428-29 (1992). In addition, the passage of time since the cited cases were decided could reasonably support higher verdicts. When confronted with the range of mental anguish verdicts approved under New York law, we do not find the verdicts in this case to deviate substantially from verdicts awarded under similar circumstances.

B. Front Pay.

Finally, KAPL claims that there was no evidence to support the jury’s front pay awards for plaintiffs Deborah Bush, who received $70,625.15 for ten years; Bruce Palmatier, who received $13,211.05 for nine years; David Townsend, who received $25,121.50 for twelve years; and Carl Woodman, who received $42,344.21 for twelve and a half years. KAPL contends that the awards are not justified because plaintiffs have not demonstrated that they continued to make efforts to obtain comparable employment. KAPL also contends that the awards project too far out in the future.

Front pay is available under the ADEA “where the factfinder can reasonably predict that the plaintiff has no reasonable prospect of obtaining comparable employment.” Whittlesey v. Union Carbide Corp., 742 F.2d 724, 729 (2d Cir.1984). The employer ordinarily has the burden of demonstrating that suitable employment was available. Greenway v. Buffalo Hilton Hotel, 143 F.3d 47, 53 (2d Cir.1998). However, if the employee failed to take reasonable steps to mitigate his damages, the employer need not prove that suitable employment existed and the plaintiff is not entitled to front pay. Id. at 54-55. The first issue presented is whether plaintiffs failed to make reasonable efforts to mitigate their damages, thus relieving defendants of the burden to demonstrate the existence of suitable employment.

Each of the plaintiffs obtained employment. Six months after Bush’s termination, she found a job as an administrative assistant at a salary lower than she had received at KAPL. She left that job for another job that paid less but gave her better prospects of promotion. By the time of trial, her new job paid her approximately $3,000 less per year than she had been paid at KAPL.

Woodman, who had worked as a specialist in non-destructive testing, was rehired at KAPL as a janitor. Woodman continued to look for jobs in his field. He found a job as a non-destructive tester with another company but continued to look for other opportunities. In early summer 1998, the new company fired him. In November 1998, KAPL rehired him as a nondestructive test inspector, a non-exempt job. He continued to look for a specialist job at KAPL.

Palmatier, who worked as an exempt electrical specialist at the time of the layoffs, returned to KAPL in May 1996 as a non-exempt hourly electrician. With overtime, his wages slightly exceeded the salary he received at the time of layoff. He did not look for non-union work because layoffs are by seniority in the non-exempt ranks and he believed he would have better protection as a member of the union.

In April or May 1996, Townsend, a former specialist, obtained a job as a health physics technician at a KAPL subcontractor. From that position, he became employed as a janitor at KAPL. Later, he was promoted to the position of radioactive waste processor. Townsend also applied for specialist jobs but was turned down. In February 1998, he took a job at KAPL as a radiation technician and has continued to seek promotion within his unit. However, because of his prior experience, Townsend is now wary of applying for exempt positions.

The district court did not clearly err in finding that plaintiffs made reasonable efforts to mitigate damages and that defendants failed to prove that other, more comparable employment existed. Each plaintiff diligently sought employment and reasonably explained why he or she was unable to or chose not to find more remunerative employment.

The amount of time for which front pay will be awarded is committed to the district court’s discretion. See, e.g., Reed v. A.W. Lawrence & Co., Inc., 95 F.3d 1170, 1182 (2d Cir.1996). KAPL wholly fails to demonstrate why the district court abused its discretion in choosing time periods between nine and twelve and a half years as the front pay period for plaintiffs who already had taken substantial steps to mitigate their damages. Therefore, we affirm the front pay awards.

VII. Cross-appeal.

As noted previously, plaintiffs request affirmative relief on their cross-appeal only if the defendants prevail on their appeal. Because we have affirmed the judgment of the -district court, we have no need to address the issues raised on the cross-appeal.

CONCLUSION

For the reasons discussed, we affirm the judgment of the district court.

1 . .“Staffing level” is not an exact term. The agreement between NR and KAPL instead provided for a fixed number of "man-years.” If the same workers remained at the same jobs for the entire year, the staffing level would be identical to the number of man-years. However, because workers do not remain at the same jobs throughout the year and because the IRIF took some time to implement, KAPL had to eliminate more employees than it would have eliminated if it had simply been required to reduce its workforce to a certain level by a certain date. We nevertheless use the more easily comprehensible terms, “staffing levels” and "number of employees” because the distinctions between those terms and "man-years” do not affect our analysis.

2 . This argument is the only claim defendants made concerning the disparate impact theory. However, they also argued with respect to pattern and practice intentional discrimination that plaintiffs failed to submit acceptable proof of an alternative practice that would have met defendants' business goals and thus shown pretext.

3 . See Smith v. City of Jackson, Miss., 351 F.3d 183, 187 (5th. Cir.2003), cert. granted,-U.S. -, 124 S.Ct. 1724, 158 L.Ed.2d 398 (2004); Adams v. Florida Power Corp., 255 F.3d 1322, 1326 (11th Cir.2001); Mullin v. Raytheon Co., 164 F.3d 696, 697 (1st Cir.1999); Ellis v. United Airlines, Inc., 73 F.3d 999, 1001 (10th Cir.1996); EEOC v. Francis W. Parker Sch., 41 F.3d 1073, 1076-78 (7th Cir.1994). But see Arnett v. California Pub. Emp. Ret. Sys., 179 F.3d 690, 696-97 (9th Cir.1999), vacated on other grounds, 528 U.S. 1111, 120 S.Ct. 930, 145 L.Ed.2d 807 (2000); Smith v. City of Des Moines, 99 F.3d 1466, 1470 (8th Cir.1996).

4 . The Court is likely to resolve this issue next term. On March 29, 2004, it granted certio-rari in Smith v. City of Jackson, 351 F.3d 183 (5th Cir.2003), which held that the disparate impact theory of recovery is not available in age discrimination cases. — U.S.-, 124 S.Ct. 1724, 158 L.Ed.2d 398 (2004).

5 . To reach Maresco, we would need to search a large and complex record to determine whether plaintiffs' disparate impact and disparate treatment claims actually do "coalesce.” Defendants' failure to preserve means that we would have to do that without the benefit of the district court's insights.

6 . Although it denied KAPL’s post-verdict motions, the district court noted that the jury erroneously found that KAPL failed to offer a legitimate business justification for terminating plaintiffs' employment through the IRIF. The error was deemed harmless due to the jury's finding with respect to equally effective alternatives. Meacham, 185 F.Supp.2d at 213-14.

7 . According to one set of instructions:

Flexibility: The individual’s flexibility to do other work within KAPL was considered. We considered principally ability to do other work within MDO, but kept in mind ability to do other work around KAPL. Highest ratings went to those with demonstrated previous experience in other assignments. For those without as much broad experience as those getting top values, higher ratings were provided to those who had demonstrated flexibility in different types of work within their assignments. • For the population below this, points were provided for demonstrated operative behavior in taking on or being asked to take on work. Lowest ratings went to those who had demonstrated resistance to being flexible to new tasks, or who, despite a flexible attitude were limited in their skills.

8 . The alternative of reworking the challenged criteria is all but self-evident from the record, particularly ' since company counsel could ■ have asked for a second (perhaps outside) opinion when he received the IRIF’s startlingly skewed results. We are satisfied that the alternative was adequately presented to support the result here.

9 . She performed these analyses between the deposition and her trial testimony.