Appellee Durwood Currier had worked for appellant Pratt & Whitney ("Pratt"), a division of United Technologies Corporation ("UTC"), for twenty-one years before he was terminated in a reduction-in-force ("RIF") in mid-2000. He was 61 years old and had advanced to positions of increasing responsibility during his tenure with the company, receiving commendations for the quality of his work and regular merit pay increases. Following his discharge, Currier filed suit under state and federal law alleging age discrimination.1 A jury found in his favor, awarding him $101,000 in back pay and $275,000 in compensatory damages. Although the district court viewed the case as "very close," it denied Pratt's motions for directed verdict, new trial or remittitur. Pratt now appeals, claiming, inter alia, that it was entitled to judgment as a matter of law and that the district court committed prejudicial error by allowing flawed statistical evidence to reach the jury. After careful review, we join the district court in concluding that, though the case was close, the verdict was supportable.

The relevant facts, as the jury could have found them, are as follows.

From 1987 to 1999, Currier worked as a business unit manager at Pratt's facility in North Berwick, Maine, a position in which he supervised 200 employees and developed the strategy for achieving the unit's goals in such areas as cost, safety and quality. In 1996, he received a glowing letter of recommendation that praised his leadership, his "maturity and business acumen," and his "going-forward potential." That same year, he took over a troubled business unit, guided it to an improved performance, and was rewarded with a three-week trip to Japan to tour similar facilities.

In 1998, Thomas Mayes became the new Operations Manager at the North Berwick facility, and in that position he supervised Currier and the then-six other business unit managers. Mayes, who was twenty years younger than Currier, conducted an evaluation of the business unit managers in 1998 and rated Currier more unfavorably than his performance seemed to warrant. For example, Currier's unit in 1998 achieved an approximately 68 percent reduction in defects, but he was given the same rating (3 on a scale of 1 to 5) for "quality" as two other, younger unit managers who experienced 73 percent and 68 percent increases in defects.2 In the "cost" category of the review, Currier was graded down because his unit's "cost per standard hour" was high; the evidence permitted the jury to conclude, however, that the higher cost was attributable to Currier's success in meeting the company's objective of reducing inventory by decreasing production lead time.3 Currier's perspective was that he was charged with the cost of the surplus labor resulting from that success, even though he assigned the extra workers to help other business units.

In early 1999, Currier was moved to a newly created manager's position, in which he was responsible for generating new business. Mayes' testimony suggested that, rather than a promotion, the transfer was a response to personnel issues and the increased costs in Currier's unit. Currier was given no job description and no goals, and his request for supervisory authority was denied. In early 2000, Mayes became General Manager of the North Berwick facility, and Stephen Pickett took over as Operations Manager. A few months later, UTC told Mayes that the salaried workforce in North Berwick needed to be reduced by about five percent. Mayes, Pickett and the head of Human Resources, Thomas Murphy, identified the job categories to be affected by the RIF, and they included the new business position held by Currier as one of the jobs to be eliminated. Rather than simply terminating Currier, however, management decided to evaluate him along with the six then-current business unit managers.

Under Pratt's standard layoff guidelines, managers in the job categories to be affected assess each employee in a targeted position using a numerical scoring system — the "Matrix" — that considers five criteria: "achieves results," "criticality of skills," "qualifications," "business orientation," and "interpersonal skills." The Human Resources department then reviews the scoring to ensure compliance with the company's guidelines and employment laws. The employee with the lowest score in a particular position is laid off.

In his evaluation of the business unit managers, Mayes gave Currier the lowest score, a 13 out of a possible 35 points. The others, all of whom were younger than Currier, received scores of 31, 28, 27, 27, 25 and 17. Currier criticizes the Matrix as a vague and "entirely subjective" list of criteria and asserts that his score "defie[d] the objective reality of [his] superb performance over more than 20 years." He points out that Mayes, who had been in South Berwick for only two years, did not consult with Currier's past supervisors or review his past performance evaluations.

To carry out the Matrix evaluation, Mayes utilized a form in which he circled a number from one to ten next to each of the five designated criteria. The evaluation process did not include a written explanation for Currier's low scores, but Mayes at trial gave three reasons for Currier's unfavorable outcome: his unit's financial performance target was off by more than any other unit's (based on cost per standard hour); his unit had "significant" labor relations problems in 1998 (stemming from an increased expectation that employees work overtime); and he did not perform well in the business development position. The jury also heard, however, that Currier received a merit pay increase in December 1999, nine months after starting the new job and six months before he was terminated.

Currier contended that Mayes' decision-making was infected by age discrimination, and he relied on his record of proven success at Pratt to demonstrate that his supposed poor performance — as reflected in his low Matrix score — was a pretext to mask the real motive. To substantiate his claim that the motive was unlawful age bias, Currier presented the testimony of an expert statistician, Dr. Sat Gupta, who concluded that the RIF disproportionately affected older employees. Gupta reported that the average age of the five employees who were laid off was 53, while the average age of those retained was 45. At one point, he testified that the risk of layoff increased by about "30 thousand percent" with increased age.

The jurors were persuaded that discrimination occurred and rendered a verdict in Currier's favor. On appeal, Pratt contends that they were improperly and unfairly led astray by Gupta's flawed statistics. In particular, Pratt asserts that Gupta incorrectly based his analysis on the total number of salaried employees at the Berwick facility, 183, rather than on the 44 employees in the job categories that Mayes, Pickett and Murphy had designated as subject to the reduction-in-force. The company also complains that Gupta's analysis failed to consider whether any factors other than age and grade — such as the company's need for particular skill sets, salaries or longevity — accounted for the differing treatment among employees. Pratt maintains that, because Gupta's statistics were not drawn from the experience of "similarly situated" employees, his conclusions lacked any probative value and were thus both irrelevant and highly prejudicial. Pratt further asserts that, at a minimum, the court erred in refusing its request to instruct the jury on the limitations of statistical evidence.

Pratt also argues on appeal that it was entitled to judgment as a matter of law because Currier failed to meet his burden of proving age discrimination. Finally, the company challenges the jury's assessment of damages. We explain below why we find no reversible error on any of these issues.

We begin with Pratt's challenge to the statistical evidence, which it claims warrants a new trial, and then turn to its contention that the evidence presented at trial simply did not add up to a case of age discrimination. The damages issues are addressed briefly in the succeeding section.

A. Admissibility of the Statistical Evidence

Pratt made repeated efforts to exclude Gupta's testimony from the jury's consideration, filing multiple pre-trial motions asserting that his statistical analysis was "unreliable, methodologically unsound, incomplete, and irrelevant." The company continued to voice its objections at trial and ultimately was awarded a standing objection on the issue. Its primary complaint on appeal is that Gupta's calculations were based on the wrong set of employees; rather than the entire salaried workforce of 183 employees, Pratt maintains that Gupta's data should have been limited either to the six similarly situated business unit managers (plus Currier), or, at most, to the 44 employees who were actually susceptible to layoff. Gupta acknowledged that, if not all salaried employees were subject to layoff, his analysis would be distorted.

Assessing the admissibility of complex expert testimony before trial, when the court does not yet know the exact nature of the evidence to be presented, or even during trial, as the case develops witness by witness, is a challenge for any judge. A judge reasonably could resolve many admissibility questions either way, and rampant second-guessing by appeals courts would paralyze the judicial process. It is thus appropriate that a trial court's on-the-spot judgment, when made pursuant to the proper legal standard, be given broad deference. Accordingly, appellate courts will reverse a trial court's decision to admit or exclude expert opinion only for an abuse of discretion. Kumho Tire Co. v. Carmichael, 526 U.S. 137, 152, 119 S. Ct. 1167, 143 L. Ed. 2d 238 (1999); Gen. Elec. Co. v. Joiner, 522 U.S. 136, 143, 118 S. Ct. 512, 139 L. Ed. 2d 508 (1997); Hochen v. Bobst Group, Inc., 290 F.3d 446, 452 (1st Cir.2002).

Under well established Supreme Court case law, the trial judge serves as a "gatekeeper" for expert evidence, with the responsibility of "ensuring that an expert's testimony both rests on a reliable foundation and is relevant to the task at hand," Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 597, 113 S. Ct. 2786, 125 L. Ed. 2d 469 (1993); see also Hochen, 290 F.3d at 452; Fed.R.Evid. 702. Statistical analyses have been held admissible in disparate treatment discrimination cases "unless they are `so incomplete as to be inadmissible as irrelevant,'" McMillan v. Mass. Soc'y for the Prevention of Cruelty to Animals, 140 F.3d 288, 303 (1st Cir.1998) (quoting Bazemore v. Friday, 478 U.S. 385, 400 n. 10, 106 S. Ct. 3000, 92 L. Ed. 2d 315 (1986)).

Pratt argues that the court failed to perform its gatekeeping duties in this case by ignoring the lack of foundation for Gupta's statistics. It complains that the court presumed incorrectly that the list of 183 employees that Pratt had supplied to Currier ("Exhibit 22") in compliance with the Older Workers Benefit Protection Act, 29 U.S.C. § 626(f),4 defined the group subject to termination. And, as a result of that presumption, Pratt maintains that the court allowed into evidence irrelevant and prejudicial statistics.

Pratt's initial difficulty in demonstrating an abuse of discretion is that the district court faced a moving target, both in the nature of the company's objections and in the state of the evidence. Not until its post-trial motion for judgment as a matter of law did Pratt explicitly make the argument it now offers about Gupta's analysis, i.e., that his conclusions are invalid because his calculations used the full universe of salaried employees rather than the 44 who were subject to layoff. In its pre-trial motions, Pratt emphasized that the data on which Gupta relied — Exhibit 22's list of 183 employees — was flawed because it did not account for variables other than age and was not confined to similarly situated employees, in that there was no breakdown in terms of job, skills, or length of time served.

Pratt made similar arguments when it moved for directed verdict both at the close of plaintiff's case and at the close of all evidence. In its first motion, Pratt also pointed to Gupta's admission that his statistical sample — and thus his analysis — would be wrong if any of the other layoffs that were part of the RIF were attributable to age neutral factors. Pratt did not assert at that time, however, that the RIF process began with a restricted pool of eligible employees and that Gupta's analysis was entirely without foundation on that basis.5

We suspect that the argument did not surface earlier because the original size of the at-risk pool of employees appeared to remain an open question through the end of the trial. Although the district court observed that the evidence unequivocally established that "ultimately only 44 out of the 183 salaried employees were at risk for the reduction in force," the key word is "ultimately." The five job categories chosen to absorb the RIF were picked by Mayes, Pickett and Murphy, and the record contains no evidence that would foreclose the jury from reasoning that the original winnowing down of the 183 to 44 was not "age neutral." As plant manager, Mayes, the individual whom Currier accuses of bias, was the person in charge of effectuating the RIF, and the jury could have found that he played the leading role in selecting the targeted groups. In brief, while the particular layoff decisions focused on only 44 employees, the jury reasonably could have viewed the layoff process to have begun with Mayes' scrutiny of the 183.6 And, in any event, the factual development of the case belies any abuse of discretion in this respect in the court's decision to admit the statistical evidence.

Pratt's related argument that Gupta's analysis was meaningless because it failed to take into account the wide differences in circumstances among the terminated employees was a point made clearly and repeatedly by Pratt through cross-examination of Gupta and in the testimony of the company's own expert witness. We see no abuse of discretion in the district court's decision to view this weakness in Gupta's analysis as a matter of weight rather than admissibility and thus properly a subject of argument and jury judgment. As the Supreme Court has observed, "[v]igorous cross-examination, presentation of contrary evidence, and careful instruction on the burden of proof are the traditional and appropriate means of attacking shaky but admissible evidence," Daubert, 509 U.S. at 596, 113 S. Ct. 2786; see also McMillan, 140 F.3d at 303 ("[I]f [the expert's] analysis omitted what defendants argue are important variables, or was deficient in other respects ... it was up to defendants to exploit and discredit the analysis during cross examination.").

Pratt took the opportunity to "exploit and discredit" Gupta's conclusions through other evidence as well. Mayes, the sole individual accused of acting with discriminatory motive, completed Matrix evaluations only for the business unit managers and thus had no direct involvement in the other layoff decisions; as the district court noted, Mayes' limited role diminished the probative value of Gupta's collective analysis of the five terminations. Similarly, Pratt elicited evidence that one of the other employees laid off, who also was 61 years old at the time, had a history of performance issues — a fact that again challenged the validity of Gupta's statistical conclusion of age bias.

The cases on which Pratt relies to assert reversible error in the admission of the evidence are largely distinguishable. For example, in LeBlanc v. Great American Insurance Co., 6 F.3d 836, 847-849 (1st Cir.1993), the court's holding was that the statistical evidence was insufficient as a matter of law to support a jury finding of discrimination; Currier does not argue that the statistics here prove age discrimination, but offers them to bolster his claim that other facts pointed to age bias. See infra at 256. In two other cases, the challenge was to the accuracy or context of the expert's conclusions. In Irvine v. Murad Skin Research Labs., Inc., 194 F.3d 313, 320-21 (1st Cir.1999), the court concluded that the expert lacked "adequate factual data to support [his] conclusions," rendering his computations related to damages unreliable. In Hillstrom v. Best Western TLC Hotel, 354 F.3d 27, 32 (1st Cir.2003), statistics that showed a high rate of termination for older employees lacked probative value because there was no evidence on the universe of employees supervised by the manager at issue.7

Here, the information on which the statistical analysis was based was presented, and there is no claim that the statistics were an inaccurate representation of what the expert analyzed. Accuracy, of course, is not the whole story. As we have noted, various factors blunted the significance of Gupta's conclusions and, indeed, we think his analysis skittered near the line of inadmissibility. The jury was not, however, uninformed. Challenges to the probative value of Gupta's analysis were amply brought to the jury's attention. In these circumstances, we find no abuse of discretion, and thus no reversible error, in the district court's decision to admit Gupta's statistics and allow the jury to assess their significance.8

Pratt argues that it should have been granted judgment as a matter of law because Currier failed to present sufficient evidence to meet his burden of proving age discrimination. We review the district court's denial of Pratt's motion under Fed.R.Civ.P. 50 de novo, Zachar v. Lee, 363 F.3d 70, 73 (1st Cir.2004), with our review "weighted toward preservation of the jury verdict," Rodowicz v. Mass. Mut. Life Ins. Co., 279 F.3d 36, 41 (1st Cir.2002). "We must affirm unless the evidence was `so strongly and overwhelmingly' inconsistent with the verdicts that no reasonable jury could have returned them." Walton v. Nalco Chem. Co., 272 F.3d 13, 23 (1st Cir.2001) (citation omitted).

Currier's initial burden was to establish a prima facie case as required under the familiar McDonnell Douglas framework, which is applied when the plaintiff lacks direct evidence of discrimination. See, e.g., Hillstrom, 354 F.3d at 30; Cruz-Ramos v. Puerto Rico Sun Oil Co., 202 F.3d 381, 384 (1st Cir.2000) (describing the burden-shifting model developed in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802-04, 93 S. Ct. 1817, 36 L. Ed. 2d 668 (1973)); Brennan v. GTE Gov't Sys. Corp., 150 F.3d 21, 26 (lst Cir.1998). In the context of a RIF, he must show that he was at least 40 years old; that his job performance met his employer's reasonable expectations; that he experienced an adverse employment action; and that younger persons were retained in the same position or that the employer otherwise did not treat age neutrally. Cruz-Ramos, 202 F.3d at 384; Brennan, 150 F.3d at 26. This burden is "not onerous," Cruz-Ramos, 202 F.3d at 384, and the requirements were easily met here.9

It is equally plain that Pratt adequately met its responsive burden to articulate a legitimate, nondiscriminatory reason for its action. See Cruz-Ramos, 202 F.3d at 384. The company pointed to Currier's low score in the Matrix review and identified three areas of concern: his unit's cost overruns; his unproductive performance in the new business position; and labor relations problems in his unit relating to overtime. The burden thus returned to Currier to prove age discrimination. Id. As we review his effort, "we inquire whether the evidence as a whole would permit a reasonable factfinder to conclude that the proffered reason was pretextual and the true reason was an age-based animus." Brennan, 150 F.3d at 26.

Like the judgment to admit the statistical evidence, it is a close call whether the evidence presented in this case adds up to age discrimination. Currier acknowledges that at least two of Pratt's asserted reasons for his low ranking had a basis in fact. He admitted at trial that he was not effective in the new business position, and that "I have myself to blame for it." He does not dispute that his unit missed its cost target in 1998. On the labor relations issues, he did not directly refute Mayes' testimony regarding employee discontent about excessive overtime but noted that "[o]vertime is built into the system" and that his unit was "no worse, no better than anybody else." Currier further acknowledged that, before his termination, he had never felt that he was treated differently based on his age by Mayes or any other Pratt supervisor.

The jury, however, also heard testimony suggesting that Currier's low Matrix score did not reflect his capabilities. Just a few years earlier, he had been asked to take over a troubled unit and was rewarded for his efforts in turning it around. With that background, it reasonably may have seemed implausible to the jury that, in a fair rating, Currier would rank at the bottom of a group that included one person who had only a couple of months experience as a unit manager — and, indeed, that his score would be half or less than half of five of his six peers. In addition, the jury could have credited Currier's explanation that his more recent cost overruns did not reflect poor performance but were a temporary byproduct of his successful strategy to reduce product lead time. As the district court observed, the testimony about the Matrix also supported an inference of pretext:

The five Matrix categories ("achieves results," "criticality of skills," "qualifications," "business orientation," and "interpersonal skills") were entirely subjective and the jury could well have been dissatisfied with Mayes' vague explanations as to why Currier received low scores in some of the Matrix categories. With regard to the "qualifications" category, Mayes testified that "based on [his] observations, the other business units, based on performance as shown, their ability to use those skills and their perception to manage business were better." ... With regard to the "criticality of skills" category, Mayes testified that he had not seen Currier apply his skills to the "changing business environment."

In assessing Currier's skills, Mayes did not review Currier's past performance reviews, which included the praise for his "going-forward potential." On the evidence offered, the jury reasonably could have concluded that Mayes gave an inadequate explanation for Currier's decline from the excellent evaluations he was given just a few years earlier.

While this evidence amply supports a jury conclusion of pretext in Currier's ranking, the more tenuous finding is that age discrimination furnished the motive. Currier did not point to even passing comments about age by Mayes or other Pratt supervisors, and he testified that the first time he felt age bias was when he was terminated. In Reeves v. Sanderson Plumbing Products, Inc., 530 U.S. 133, 147, 120 S. Ct. 2097, 147 L. Ed. 2d 105 (2000), however, the Supreme Court held that, "[i]n appropriate circumstances," the plaintiff's prima facie case together with sufficient evidence of pretext could be enough to support an inference of discrimination. The circumstances here seem to us appropriate. The prima facie case and — to use the district court's adjective — the "unpersuasive" explanations for Currier's low ratings loom large in the absence of any other explanation. There is no suggestion that Mayes disliked Currier, or that they were in any way incompatible.

We need not rely on the Reeves holding, however, because Currier offered more to support the jury's determination. When two unit manager jobs opened up in the months before the RIF, Currier was passed over for the positions even though he had communicated his desire to move back from the new business slot. Both of the employees selected were significantly younger than Currier (20 and 15 years), and one had no prior experience as a business unit manager. If the jurors viewed the company's performance justification for Currier's termination as pretext — as apparently they did — they likewise could have rejected Mayes' explanation that performance concerns motivated this earlier decision-making. Similarly, the jury may have seen evidence of age bias in the contrast between Mayes' 1998 performance evaluation of Currier and his evaluations of other unit managers. See supra at 248-49. In addition, although the probative force of Gupta's testimony was limited for reasons we have explained, the jury properly may have given it some weight, at least in the absence of any conflicting evidence. Even after Gupta acknowledged that adding other variables to his analysis might have led to a different outcome, he maintained that the result he achieved would retain significance:

I think the evidence was so strong that perhaps I would still say that age was a dominant player. Again, there is some speculation how much impact other variables have but it does not dilute such a tremendous impact.

In sum, after careful consideration, we agree with the district court that "[t]he case was not strong but ... there was sufficient evidence for the jury to infer that Currier's age was the real reason that Mayes selected him to be terminated."

C. New Trial or Remittitur Based on Excessive Damages

Pratt argues that the evidence failed to support either the jury's award of $101,580 in back pay or its compensatory damages award of $275,000, and it claims that the district court abused its discretion by denying the company's motion for a new trial or remittitur. See Davignon v. Clemmey, 322 F.3d 1, 11 (1st Cir.2003). Pratt acknowledges that a party seeking remittitur "bears a heavy burden of showing that an award is `grossly excessive, inordinate, shocking to the conscience of the court, or so high that it would be a denial of justice to permit it to stand,'" Koster v. TWA, 181 F.3d 24, 34 (1st Cir.1999) (citation omitted).

We agree with the district court that the damages award here does not meet that standard, and its reasoning accords with our own. See District Court Order on Post-Trial Motions, at 20-22. It thus suffices for us to say that, with respect to back pay, the jury could have found that appellant continued to try to find some type of work even after applying for Social Security benefits, and that, as to compensatory harm, the amount awarded was not grossly disproportionate to the non-pecuniary losses to which appellant testified. In short, there was no abuse of discretion in the district court's denial of Pratt's motion.

For the foregoing reasons, the judgment of the district court is affirmed in its entirety.