Robert F. Byrnie v. Town of Cromwell, Board of Education, Body Corporate Cromwell Board of Education Body Corporate

*98 POOLER, Circuit Judge: Plaintiff Robert F. Byrnie appeals from a judgment of the United States District Court for the District of Connecticut, Gerard L. Goettel, Judge, dismissing his complaint alleging defendants Town of Cromwell Public Schools and Cromwell Board of Education (“Cromwell”) failed to hire him for a part time teaching position as an art teacher on the basis of his age, in violation of the Age Discrimination in Employment Act of 1967 (“ADEA”), 29 U.S.C. § 621 et seq., and on the basis of his gender, in violation of Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq. and Title IX of the Education Amendments of 1972, as amended, 20 U.S.C. § 1681 et seq. The district court dismissed Byrnie’s disparate treatment claims under both statutes on summary judgment on the grounds that Byrnie had failed to produce sufficient evidence to show that Cromwell’s nondis-eriminatory reason for hiring another candidate was pretext. The district court also dismissed Byrnie’s disparate impact claims under both statutes on summary judgment because the statistical evidence offered by Byrnie of gender and age discrimination was insufficiently linked to any identifiable hiring practice or policy. On appeal, Byr-nie contends summary judgment was inappropriate on his disparate treatment claims because unlawful discrimination could be inferred from a number of circumstances, including: the objective superiority of his paper credentials in comparison with the chosen candidate’s; the vague and conclusory nature of Cromwell’s business reason; and the destruction of documents made in connection with the hiring process. Byrnie also contends that summary judgment was inappropriate on his age-related disparate impact claim because he could allege a specific employment practice as the cause of statistical evidence indicating an age imbalance in Cromwell’s teaching hires. Summary judgment on his gender-related disparate impact claim was inappropriate, Byrnie argues, because he was prevented from alleging a cause for the statistical evidence showing gender imbalance among Cromwell’s teachers due to Cromwell’s destruction of documents. For the reasons below, we affirm the district court’s dismissal of Byrnie’s disparate impact claims but reverse with regard to Byrnie’s disparate treatment claims. BACKGROUND We presume familiarity with the district court’s thorough decision. See Byrnie v. Town of Cromwell Pub. Schs., 73 F.Supp.2d 204 (D.Conn.1999). In the spring of 1995, the Cromwell Board of Education decided to hire a part-time art teacher for Cromwell High School. Following the steps laid out in the Board of Education’s “Recommended Procedures for Hiring,” Cromwell advertised for a permanent, part-time art teacher in June, 1995. Applicants were required to submit a letter of application, a resume, transcripts, a letter of reference, and proof of Connecticut certification. Although the posting announcing the opening indicated that the only prerequisite for applicants was current Connecticut teaching certification, Cromwell’s job description for the position (which was not itself posted) specified, among other things, that the appropriate candidate have a degree in art education. Byrnie, who was 64 at the time, had a bachelor’s and a master’s degree in art education and had taught art at the high school level for 21}£ years. He had been substitute teaching at Cromwell High School for the five years prior to applying for the position and had been serving as the advisor to Cromwell High School’s literary magazine. In addition, Byrnie was certified to teach at the high school level in the areas of Art, History, Social Studies, and Adult Education. Forty-one people applied for the position. A Search Committee, consisting of Cromwell High School Principal Mark Nappi and two members of the Board of Education, pre-screened the applications, narrowing the pool down to 21 applicants. Of the original 41, three applicants were *99 male; two, including Byrnie, survived the pre-screening. At this point the Search Committee was enlarged to include a total of six teachers, administrators and board members. On July 3, 1995, four of the committee members met (two were unable to attend the meeting) to review and rate the 21 applications on a scale of 1-5 on ballot forms provided by Nappi. Interviews were granted to the five applicants receiving the highest cumulative scores. Esther Mancarella, the person eventually chosen for the position, received the highest cumulative score with eighteen points out of a possible twenty. Byrnie came in fourth with sixteen points. Of the five applicants chosen for interviews, only four were actually available so a lower scoring alternate was also chosen for an interview. All the candidates chosen for interviews, excluding Byrnie, were female — although the large majority of applicants had also been female. Two of the women were over age 40. At the same meeting, the committee agreed upon a set of interview questions to be asked at each interview. On July 12, 1995, five members of the Search Committee met to interview the candidates. The candidates were all asked the same questions in the same order by the same interviewer. At the conclusion of the interviews, the committee unanimously eliminated the alternative interviewee, and then each committee member wrote down his or her top three choices for the job in the order of preference. Based on those rankings, three of the four candidates were selected to proceed to the next round of interviews. Mancarella ranked first, having been selected by four committee members as their first choice and as the second choice of the fifth committee member. Byrnie ranked fourth, having only been chosen as the second choice of one committee member and the third choice of another member. Nevertheless, the committee unanimously decided to allow Byrnie to proceed to the next round of interviews in light of his service to Cromwell High School. The second round of interviews were conducted by Nappi and Cromwell High School’s Assistant Principal, Joseph Cas-sella, who each individually interviewed the candidates on July 18, 1995. Each interviewee was asked the same questions by the same interviewer in the same order. After the interviews, Cassella and Nappi conferred and agreed that Mancarella was the best candidate for the position. Nappi telephoned Mancarella, informing her that she was being recommended for the position and telling her to set up an appointment to meet with Superintendent James Gere. After the meeting, Gere submitted Mancarella’s name to the Board of Education which approved her appointment on August 8, 1995. Mancarella was a 42 year old with a bachelor’s degree in fine arts as well as some graduate course work. Although Mancarella had no experience teaching art at the high school level, she had four years experience teaching art at the middle school level and had been teaching art on a part-time basis at a treatment center for children identified as Socially Emotionally Maladjusted. In 1994, she had received a teaching award. Meanwhile, on July 24, 1995, Byrnie received a letter informing him that he had not been selected for the position. Convinced of his superior qualifications, Byr-nie contacted Gere and was told that he had been considered the least qualified of the four candidates considered for a second round of interviews and, in fact, he participated in the second round only as courtesy because of his service to the school system. At that time, Gere explained the hiring procedures followed by the search committee. Byrnie followed up on his concerns by meeting with Gere on July 27, 1995 and meeting with Nappi and Casella on August 8, 1995. At the August 8 meeting, Byrnie was denied a copy of the interview questions, but was permitted to examine them. In spite of Byrnie’s repeated requests, he was denied copies of the interview questions and Mancarella’s application materials until he filed a com *100 plaint under the Connecticut Freedom of Information Act (“FOIA”). Byrnie filed a complaint with the Connecticut Commission on Human Rights and Opportunities (“CCHRO”), and on January 8, 1996, Cromwell responded with a verified answer (“CCHRO Answer”). The CCHRO Answer was prepared by Cromwell’s attorneys based on a memorandum prepared by Nappi, which was in turn based on notes and records in Nappi’s possession at that time but were later lost or destroyed. In the CCHRO Answer, Cromwell explained that Byrnie had not been selected for the position because he performed poorly in the interviews and seemed unfamiliar with the teaching methods set forth in the Connecticut Competency Instrument (“CCI”), an assessment tool developed by the Connecticut State Department of Education in 1992. According to the CCHRO Answer, Byrnie was focused on his teaching portfolio and the history of art during the first interview. During his second interview, he exhibited a lack of familiarity with the CCI, and “Nap-pi had to explain the three major cluster areas set forth in the CCI.” Besides failing to “demonstrate that he understood the concepts set forth in the CCI,” Byrnie answered interview questions with vague, general, and unimaginative responses, according to the CCHRO Answer. By contrast, Mancarella displayed “enthusiasm,” poise, and an ability to work with a variety of different types of students, while her answers were “thought-out and coherent.” She also, according to the CCHRO Answer, “demonstrated knowledge of the CCI.... [Cromwell] bases its teacher evaluation system on the CCI.” On September 3, 1996, the CCHRO dismissed Byrnie’s complaint. On August 5, 1997, Byrnie filed a complaint with the district court alleging that Cromwell discriminated against him on the basis of his age, in violation of the Age Discrimination in Employment Act of 1967 (“ADEA”), 29 U.S.C. § 621 et seq., and on the basis of his gender, in violation of Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq., and Title IX of the Education Amendments of 1972, as amended, 20 U.S.C. § 1681 et seq. 1 Discovery produced testimony undermining the CCHRO Answer’s claim that familiarity with the CCI was one of the factors used to evaluate the candidates and suggested that neither Mancarella nor any of the other candidates had any significant familiarity with the CCI. In affidavits, two of Byrnie’s interviewers, Nappi and Cassella, stated that Byrnie was not hired because he did not demonstrate familiarity with the basic competencies of effective teaching. On October 25, 1999, the district court dismissed all of Byrnie’s claims in response to Cromwell’s motion for summary judgment. The district court held that Byrnie had made his prima facie case: Byrnie was qualified for the position he applied for, but the position was given to a 42 year old woman. See Byrnie, 73 F.Supp.2d at 210. The district court also found that Cromwell had offered a legitimate reason for failing to hire Byrnie: Mancarella was judged the better candidate based upon subjective criteria. See id. Byrnie’s case failed because there was not enough evidence to allow a jury to find Cromwell’s explanation to be pretextual and to allow an inference of discrimination. Byrnie’s paper credentials, while better than Mancarella’s, were not so overwhelmingly better as to allow an inference of discrimination based on that alone. See id. at 213. Cromwell’s subjective choice of Mancarella over Byrnie based on her superior interviewing performance was proper because the choice did not appear to be influenced by any discriminatory bias. See id. at 213-14. The selection process itself, the district court held, did not treat the applicants differently and thus could not *101 support an inference of discrimination. See id. at 218. And Byrnie’s statistical evidence showing that Cromwell’s teaching force had a lower percentage of males than Connecticut’s state-wide average could not, without a further showing of the cause of the statistical disparity, support a disparate treatment claim. See id. at 218. The statistical evidence also failed to support a disparate impact claim because Byrnie failed to allege an employment practice responsible for the under-representation of males and the elderly among Cromwell’s teaching hires. See id. at 221. Finally, Byrnie’s argument that the CCI-based interviewing questions caused a disparate impact on older job applicants was dismissed because he provided the court with no statistical evidence indicating such an effect. See id. at 222. Byrnie appealed the judgment on November 18, 1999. DISCUSSION We review a district court’s grant of summary judgment de novo. See Howley v. Town of Stratford, 217 F.3d 141, 151 (2d Cir.2000); Raine v. RKO Gen., Inc., 138 F.3d 90, 93 (2d Cir.1998). At summary judgment, a court is tasked with determining whether genuine disputes over material fact exist between the parties which should properly be submitted to a jury or whether, where no issues of material fact are found, the moving party is entitled to judgment as a matter of law. See Howley, 217 F.3d at 150-51; Kerzer v. Kingly Mfg., 156 F.3d 396, 400 (2d Cir.1998). Even where facts are disputed, in order to defeat summary judgment, the nonmoving party must offer enough evidence to enable a reasonable jury to return a verdict in its favor. See Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986); Scotto v. Almenas, 143 F.3d 105, 114 (2d Cir.1998). Thus the “non-moving party may not rely on conelusory allegations or unsubstantiated speculation.” Scotto, 143 F.3d at 114. Nevertheless, when examining the evidence, the court should resolve all ambiguities and draw all inferences in favor of the non-moving party. See id.; Kerzer, 156 F.3d at 400. Thus, “if there is any evidence in the record from any source from which a reasonable inference could be drawn in favor of the nonmoving party, summary judgment is improper.” Howley, 217 F.3d at 151. I. Byrnie’s Disparate Treatment Claim The ADEA makes it unlawful for an employer “to fail or refuse to hire ... any individual ... because of such individual’s age.” 29 U.S.C. § 623(a)(1). The protections of the ADEA reach individuals who are at least 40 years old. See 29 U.S.C. § 631(a). Similarly, Title VII makes it unlawful for an employer “to fail or refuse to hire ... any individual ... because of such individual’s race, color, religion, sex, or national origin.” 42 U.S.C. § 2000e-2(a)(1). A plaintiff utilizes the same evidentiary framework for demonstrating either age discrimination or sex discrimination. See Hollander v. Am. Cyanamid Co., 172 F.3d 192, 198-99 (2d Cir.), cert. denied, 528 U.S. 965, 120 S.Ct. 399, 145 L.Ed.2d 311 (1999); Fisher v. Vassar College, 114 F.3d 1332, 1335 (2d Cir.1997) (en banc). A plaintiff alleging a violation of either statute must establish, by a preponderance of the evidence, a prima facie case consisting of four elements: (1) that plaintiff falls within the protected group, (2) that plaintiff applied for a position for which he was qualified, (3) that plaintiff was subject to an adverse employment decision and (4) that the adverse employment decision was made under circumstances giving rise to an inference of unlawful discrimination. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973); Fisher, 114 F.3d at 1335; Woroski v. Nashua Corp., 31 F.3d 105, 108 (2d Cir.1994). The burden upon the plaintiff to prove a prima facie case is minimal. See Carlton v. Mystic Transp., Inc., 202 *102 F.3d 129, 134 (2d Cir.), cert. denied, 530 U.S. 1261, 120 S.Ct. 2718, 147 L.Ed.2d 983 (2000); Chambers v. TRM Copy Ctrs. Corp., 43 F.3d 29, 37 (2d Cir.1994). Once a prima facie case has been established, the burden of production shifts to the employer who must defeat a rebuttable presumption of discrimination by articulating a legitimate, non-discriminatory reason for the employment decision. See Carlton, 202 F.3d at 134; Raskin v. Wyatt Co., 125 F.3d 55, 64 (2d Cir.1997). If the employer offers, via admissible evidence, a justification of its action which, if believed by a reasonable trier of fact, would allow a finding of no unlawful discrimination, then “ ‘the McDonnell Douglas framework — with its presumptions and burdens’ — disappear[s], and the sole remaining issue [is] ‘discrimination vel non.’ ” Reeves v. Sanderson Plumbing Prods., Inc., 530 U.S. 133, 120 S.Ct. 2097, 2106, 147 L.Ed.2d 105 (2000) (citation omitted). At summary judgment in an employment discrimination case, a court should examine the record as a whole, just as a jury would, to determine whether a jury could reasonably find an invidious discriminatory purpose on the part of an employer. See Howley, 217 F.3d at 151; Stern v. Trustees of Columbia Univ., 131 F.3d 305, 314 (2d Cir.1997); cf. Washington v. Davis, 426 U.S. 229, 242, 96 S.Ct. 2040, 48 L.Ed.2d 597 (1976) (holding that “an invidious discriminatory purpose may often be inferred from the totality of the relevant facts”). A court is to examine “the entire record to determine whether the plaintiff could satisfy his ‘ultimate burden of persuading the trier of fact that the defendant intentionally discriminated against the plaintiff.’ ” Schnabel v. Abramson, 232 F.3d 83, 90 (2d Cir.2000) (quoting Reeves , 120 S.Ct. at 2106). A motion for summary judgment may be defeated where “a plaintiffs prima facie case, combined with sufficient evidence to find that the employer’s asserted justification is false, may permit the trier of fact to conclude that the employer unlawfully discriminated.” Reeves, 120 S.Ct. at 2109. Taking the facts in the light most favorable to the non-moving party, Byrnie easily made out a prima facie case. He was a 64 year old male job applicant who, no one disputes, was eminently qualified for the art teacher position that Cromwell awarded to 42 year old female Mancarella, an applicant substantially younger than Byrnie. See Brennan v. Metro. Opera Ass’n, Inc., 192 F.3d 310, 317 (2d Cir.1999) (noting that “the fact that the replacement is substantially younger than the plaintiff is a more valuable indicator of age discrimination, than whether or not the replacement was over 40”). Cromwell, in turn, articulated a legitimate business reason for not hiring Byrnie: Mancarella performed better than Byrnie during the candidate interviews and thus seemed, based on subjective criteria, the better qualified candidate. Byrnie, in turn, cannot offer direct evidence of an improper discriminatory bias and thus must defeat summary judgment on the strength of his prima facie case combined with circumstantial evidence that Cromwell’s stated reason for failing to hire Byrnie is pretext. See Reeves, 120 S.Ct. at 2108 (“Proof that the defendant’s explanation is unworthy of credence is ... one form of circumstantial evidence that is probative of intentional discrimination, and it may be quite persuasive.”). A combination of factors, any of which judged on their own would be much less compelling, provide sufficient evidence to allow a reasonable jury to conclude that Cromwell’s explanation for failing to hire Byrnie was a pretext for impermissible discrimination. A. Comparative “Paper” Qualifications In terms of paper credentials, Byrnie was the better qualified candidate for the job. Byrnie had 21)£ years’ experience working as an art instructor at a number of different high schools. Further, Byrnie *103 had spent six years as a substitute teacher at Cromwell High School. Mancarella, by contrast, had four years’ experience teaching art at the middle school level, along with eight additional years’ part-time teaching experience as an art instructor. Byrnie had a bachelor’s and a master’s degree in Art Education, while Mancarella had a bachelor’s degree in Fine Arts and some graduate study course work. Both candidates presented strong recommendations. Finally, while both candidates had credentials sufficient for the position, technically Mancarella lacked the requisite education: the position called for a bachelors degree in Art Education, while Mancarella possessed a degree in Fine Arts. Courts have recognized that an employer’s disregard or misjudgment of a plaintiffs job qualifications may undermine the credibility of an employer’s stated justification for an employment decision. See Fischbach v. D.C. Dep’t of Corr., 86 F.3d 1180, 1183 (D.C.Cir.1996) (“Evidence indicating that an employer misjudged an employee’s performance or qualifications is, of course, relevant to the question whether its stated reason is a pretext masking prohibited discrimination.”); Tyler v. Re/Max Mountain States, Inc., 232 F.3d 808, 814 (10th Cir.2000) (quoting Fischbach); Alexander v. Fulton County, Ga., 207 F.3d 1303, 1340 (11th Cir.) (noting that “evidence showing an employer hired a less qualified applicant over the plaintiff may be probative of whether the employer’s proffered reason for not promoting plaintiff was pretextual”), reh’g and reh’g in banc denied, 218 F.3d 749 (11th Cir.2000). At the same time, “the court must respect the employer’s unfettered discretion to choose among qualified candidates.” Fischbach, 86 F.3d at 1183; see also Simms v. Oklahoma ex rel. Dep’t of Mental Health & Substance Abuse Servs., 165 F.3d 1321, 1330 (10th Cir.) (“Our role is to prevent unlawful hiring practices, not to act as a ‘super personnel department’ that second guesses employers’ business judgments.”) (citations omitted), cert. denied, 528 U.S. 815, 120 S.Ct. 53, 145 L.Ed.2d 46 (1999). When a plaintiff seeks to prevent summary judgment on the strength of a discrepancy in qualifications ignored by an employer, that discrepancy must bear the entire burden of allowing a reasonable trier of fact to not only conclude the employer’s explanation was pretextual, but that the pretext served to mask unlawful discrimination. In effect, the plaintiffs credentials would have to be so superior to the credentials of the person selected for the job that “no reasonable person, in the exercise of impartial judgment, could have chosen the candidate selected over the plaintiff for the job in question.” Deines v. Tex. Dep’t of Protective & Regulatory Servs., 164 F.3d 277, 280-81 (5th Cir.1999); see also Fischbach, 86 F.3d at 1183 (“Title VII liability cannot rest solely upon a judge’s determination that an employer misjudged the relative qualifications of admittedly qualified candidates.”). Byrnie’s credentials on their own could not meet this weighty burden. While he was well-qualified for the art teacher position, we cannot say that Mancarella was unqualified or that Cromwell was unreasonable in electing to offer her the position in light of a comparison of her paper credentials with Byrnie’s. Nevertheless, just because the discrepancy between Byrnie’s and Mancarella’s qualifications does not on its own have the strength to create a material issue of fact, that does not mean the discrepancy is stripped of all probative value. Indeed, although Cromwell claims the decision to hire Mancarella rested finally upon her superior performance in the interviews, it is worth noting that even prior to the interviews — and thus based upon paper credentials alone — she was ranked, along with two other applicants, higher than Byrnie. After an initial screening, the pool of 41 applicants was reduced to 21 who were then subjected to a second screening. During the screening meeting the applications were rated on a scale of 1- *104 5 by each member of the Search Committee, and those applicants who received the highest number of points were.invited to interview. In reviewing the applications, the Search Committee looked at whether the applications were complete, the presentation of the cover letters, and the applicants’ education, experience and achievement. 2 Based on paper credentials alone, Mancarella scored highest while Byrnie came in fourth. While no doubt other factors could also be important, Nap-pi could not suggest what other factors might account for determining Mancarella was a more suitable candidate on the strength of her paper credentials. The credibility of the Search Committee is not helped by the fact that it needed to relax the educational requirements of the position in order for Mancarella to survive these initial screenings, let alone be selected as the most deserving of an interview. 3 Further, the Search Committee had to ignore the requirement that an application be complete in Mancarella’s case because her application was missing transcripts from a number of the colleges she had attended. To be sure, this alone does not support a finding of pretext; lawful reasons can easily be found for the low ranking of Byrnie’s application in comparison to his less experienced competitor — for example, his self-presentation in the application cover letter may have been off-putting. See Byrnie, 73 F.Supp.2d at 216 (suggesting a variety of lawful reasons that could explain the ranking of Byrnie’s application). Similarly, the failure to screen out Mancarella’s application based on its being incomplete and her lacking the appropriate qualifications is not enough on its own to call into question the good faith of the hiring process because these are not significant irregularities. See Weinstock v. Columbia Univ., 224 F.3d 33, 45 (2d Cir.2000) (noting that procedural irregularities may “ ‘raise a question as to the good faith of the process where the departure may reasonably affect the decision.’ ”) (quoting Stern, 131 F.3d at 313). Nevertheless, a reasonable trier of fact is entitled to find that it does bear on the credibility of the employer which must finally be evaluated from the perspective of the entire record. B. Subjective Hiring Criteria Cromwell only used the applicant’s paper credentials to determine who would be selected for interviews. Mancarella was chosen as the most qualified candidate on the strength of her interview performances. The two men who interviewed Byrnie in his last interview both affirmed that “[h]is responses did not demonstrate that he was the most qualified candidate for the position. In my judgment, his responses did not demonstrate familiarity with the basic competencies necessary for effective teaching.” As the district court properly noted, “[tjhere is nothing unlawful about an employer’s basing its hiring decision on subjective criteria, such as the impression an individual makes during an interview.” Byrnie, 73 F.Supp.2d at 213. At the same time, we have also cautioned that “an employer may not use wholly subjective and unarticulated standards to judge employee performance for purposes of promotion.” Knight v. Nassau County Civil Serv. Comm’n, 649 F.2d 157, 161 (2d Cir.1981). This is because “[ajny defendant can respond to a [discrimination charge] with a *105 claim of some subjective preference or prerogative and, if such assertions are accepted, prevail in virtually every case.” Robinson v. 12 Lofts Realty, Inc., 610 F.2d 1032, 1040 (2d Cir.1979) (internal quotation marks omitted). Accordingly, an “employer’s explanation of its reasons must be clear and specific” in order to “afford the employee a full and fair opportunity to demonstrate pretext.” Meiri v. Dacon, 759 F.2d 989, 996-97 (2d Cir.1985). Where an employer’s explanation, offered in clear and specific terms, “is reasonably attributable to an honest even though partially subjective evaluation of ... qualifications, no inference of discrimination can be drawn.” Lieberman v. Gant, 630 F.2d 60, 67 (2d Cir.1980). Although Byrnie contends otherwise, Cromwell’s explanation of its hiring decision based on the candidates’ interview performances is not so vague or conclusory as to disallow Byrnie the possibility of demonstrating pretext. Quite to the contrary: the justification for not hiring Byrnie, on its face, raises credibility problems. A subjective evaluation, besides being clear and specific, must also be honest. See id. That is to say that “[w]hile the business judgment rule protects the sincere employer against second-guessing of the reasonableness of its judgments, it does not protect the employer against attacks on its credibility.” Chapman v. AI Transp., 229 F.3d 1012, 1048 (11th Cir.2000) (Birch, J., concurring in part and dissenting in part); see also Stern, 131 F.3d at 313 (“While we do not second-guess an employer’s hiring standards, the reasons for its employment decision, including its alleged reliance on such standards, are subject to scrutiny under Title VII ...”). For example, the Eighth Circuit found, in a case similar to this one, that a school’s decision to hire one teacher over another based on their interview performances raised a “genuine controversy at this summary judgment stage as to whether age was not a factor in the application of such subjective criteria or the formation of such subjective impressions.” Widoe v. Dist. # 111 Otoe County Sch., 147 F.3d 726, 730 (8th Cir.1998). In that case, the school’s claim that the plaintiff was not hired, in part, because her recommendations were outdated and her references were hard to find, was undermined by the fact that plaintiff had already been teaching at the school, had been observed by a member of the hiring committee and had, in fact, received favorable letters of recommendation from committee members. See id. at 729-30. Within this context — that is, from the perspective of the entire record — the court found there was a “genuine issue of fact” concerning “whether defendant’s proffered reasons are pre-textual.” Id. at 731. In this case, although Cromwell is entitled to use subjective criteria in choosing whom to hire, Byrnie is also entitled to challenge the credibility of the decision’s rationale. Based on Byrnie’s 21}é years teaching experience, it does not seem difficult to take issue with the credibility of Cromwell’s assertion that Byrnie lacks “familiarity with the basic competencies necessary for effective teaching.” This assertion is particularly hard to swallow given that Byrnie had been a substitute teacher at Cromwell for five years and was often asked to take over classes for extended periods when other teachers were on leave. It strains credulity to believe that a teacher unfamiliar with the competencies necessary for effective teaching would be relied upon for so long. This is particularly the case given that one of Cromwell’s own teachers wrote a glowing recommendation for Byrnie stating that Byrnie was the English Department’s first choice for long-term substitute teaching assignments. The credibility of Cromwell’s justification for not hiring Byrnie is further weakened when compared to the justification initially offered by Cromwell to the CCHRO in 1996. Cromwell’s CCHRO Answer explains that the successful job candidate needed to demonstrate an educational philosophy and teaching methodology that *106 was consistent with the teaching methods set out in the Connecticut Competency Instrument. Cromwell then noted that Byrnie was unfamiliar with the basic teaching methods as set forth in the CCI, and to emphasize the point, explained that “Nappi had to explain the three major cluster areas set forth in the CCI.” Byr-nie’s answers were “not in line with the CCI,” while Mancarella, by contrast, “demonstrated knowledge of the CCI.” Later, however, during his deposition, Nappi conceded that most, if not all, candidates “were unfamiliar with the terminology of the CCI in and of itself.” Mancarella, herself, Nappi testified, was only somewhat familiar with the CCI, and, although the CCHRO Answer failed to mention it, she, like Byrnie, also had the three major cluster areas of the CCI explained to her. Thus, the CCHRO Answer seems misleading in suggesting that Byrnie’s unfamiliarity with the CCI resulted in him comparing unfavorably with Mancarella since it is not clear Mancarella — or anyone — possessed significant familiarity with the CCI. At deposition, Nappi reconciled the fact that most of the candidates lacked familiarity with the CCI with the CCHRO Answer’s assertion that Mancarella “demonstrated knowledge of the CCI” by explaining that the CCI was “just a compilation of good teaching practices,” and Mancarella “was familiar with the basic concepts of effective teaching strategies.” The two answers are not necessarily inconsistent. By dropping reference to the CCI in the affidavits of Nappi and Casella, Cromwell extricates itself from the misleading nature of the CCHRO Answer, although the fact that the answer attempts to portray as important to the outcome of the hiring process Byrnie’s unfamiliarity with a document the successful candidate had, at best, only limited familiarity with herself is pertinent information for the trier of fact. Unfortunately for Cromwell, downplaying the importance of the CCI results in the claim that a person with over twenty years teaching experience lacks familiarity with the basics of teaching in and of themselves. A juror could believe that the latter statement was the result of an attempt to salvage an earlier explanation that was collapsing during civil discovery — and the apparent implausibility of the new explanation was a price preferable to advancing a new, potentially inconsistent explanation of the hiring decision. See EEOC v. Ethan Allen, Inc., 44 F.3d 116, 120 (2d Cir.1994) (noting that a juror can reasonably view an employer’s changing explanations as “pre-textual, developed over time to counter the evidence suggesting age discrimination”). Cromwell relies as well on Byrnie’s poor interviewing skills as the reason for choosing Mancarella for the position and calls attention to Byrnie’s own testimony about his interview answers as proof. Because we are not a super-personnel department, it does not matter whether we find Byr-nie’s interview responses appropriate. An employer is entitled to arrive at a subjective evaluation of a candidate’s suitability for a position. The issue here is whether such a poor interview performance could make credible Cromwell’s asserted justifications for its hiring decision. Byrnie’s interview performance speaks to the question whether he or Mancarella was more qualified for the job — one of the justifications Nappi and Cassella advanced for not hiring Byrnie. It is not probative, however, of their basic claim that Byrnie lacked “familiarity with the basic competencies necessary for effective teaching.” Further, when the latter claim is juxtaposed with the former, one may reasonably question “whether age [or gender] was not a factor in the application of ... subjective criteria or the formation of ... subjective impressions” concerning Byrnie’s interview performance. Widoe, 147 F.3d at 730. Cromwell’s dual explanations for the hiring decision, and the lack of evidence (and even countervailing evidence) of one of their justifications might raise a material question about Cromwell’s real reason for *107 not hiring Byrnie. We need not decide whether this mixed bag of evidence is by itself sufficient to defeat summary judgment, however, because we conclude that Cromwell’s destruction of evidence, in combination -with the evidence undermining Cromwell’s justification is adequate to defeat summary judgment. C. Spoliation Byrnie seeks an adverse evidentiary inference with regard to documents which Cromwell destroyed. The district court agreed with Byrnie that the missing application materials of the other candidates, besides Byrnie and Mancarella, for the art teacher position had potential relevance, but noted that such relevance would be “minimal.” Byrnie, 73 F.Supp.2d at 207 n. 8. We agree with the district court that were the materials of the other applicants the only thing missing, the potential significance of the missing applications could not amount to much given that once the interview candidates were selected, the application materials no longer played a role in the hiring process. However, as Cromwell concedes, the application materials were not the only materials destroyed. The “written ballot form that Mr. Nappi provided” to the screening committee members to enable them to rank the 21 applications to determine which applicants would receive an interview was missing, along with the tally sheet that compiled their votes. Also missing were the forms on which, as the CCHRO Answer explains, “[e]ach Committee member independently wrote down his or her top three choices for the job opening, in order of preference,” after the first round of interviews, along with the tally sheet adding up the votes. 4 In addition, any notes made by the interviewers in the first and second rounds of interviews — and Nappi testified that it was likely that such notes were made — were destroyed. Finally, the notes Nappi relied upon in drafting the summary of the hiring process that the CCHRO Answer was based upon are missing. We have defined spoliation as “the destruction or significant alteration of evidence, or the failure to preserve property for another’s use as evidence in pending or reasonably foreseeable litigation.” West v. Goodyear Tire & Rubber Co., 167 F.3d 776, 779 (2d Cir.1999). The spoliation of evidence germane “to proof of an issue at trial can support an inference that the evidence would have been unfavorable to the party responsible for its destruction.” Kronisch v. United States, 150 F.3d 112, 126 (2d Cir.1998). This sanction serves a threefold purpose of (1) deterring parties from destroying evidence; (2) placing the risk of an erroneous evaluation of the content of the destroyed evidence on the party responsible for its destruction; and (3) restoring the party harmed by the loss of evidence helpful to its case to where the party would have been in the absence of spoliation. See West, 167 F.3d at 779. In borderline cases, an inference of spoliation, in combination with “some (not insubstantial) evidence” for the plaintiffs cause of action, can allow the plaintiff to survive summary judgment. Kronisch, 150 F.3d at 128. Where one seeks an adverse inference regarding the content of destroyed evidence, one must first show that “the party having control over the evidence ... had an obligation to preserve it at the time it was destroyed.” Kronisch, 150 F.3d at 126. Such an obligation usually arises when a “party has notice that the evidence is relevant to litigation ... but also on occasion in other circumstances, as for example when a party should have known that the evidence may be relevant to future litigation.” Id. The law in this circuit is not clear on what state of mind a *108 party must have when destroying it. In Reilly v. Natwest Markets Group Inc., we noted that at times we have required a party to have intentionally destroyed evidence; at other times we have required action in bad faith; and at still other times we have allowed an adverse inference based on gross negligence. See Reilly v. Natwest Mkts. Group Inc., 181 F.3d 253, 267 (2d Cir.1999), cert. denied, 528 U.S. 1119, 120 S.Ct. 940, 145 L.Ed.2d 818 (2000). In light of this, we concluded a case by case approach was appropriate. See id. Finally, a court must determine “whether there is any likelihood that the destroyed evidence would have been of the nature alleged by the party affected by its destruction.” Kronisch, 150 F.3d at 127. The burden falls on the “prejudiced party” to produce “some evidence suggesting that a document or documents relevant to substantiating his claim would have been included among the destroyed files.” Id. at 128. The district court dealt with this issue by noting that there was no evidence indicating when the missing documents were destroyed or with what state of mind. See Byrnie, 73 F.Supp.2d at 208 n. 8. Nevertheless, Cromwell had to have had notice of the prospect of potential litigation at the time the notes that formed the basis for Cromwell’s CCHRO Answer were destroyed since it is only logical to anticipate that a complaint lodged with the CCHRO might be pursued at a later date in a court room. The CCHRO Answer, further, promises to make available to the CCHRO “the name or any identifiable information about any [applicants]” if needed — suggesting that applicant information existed at the time the CCHRO Answer was prepared. And, arguably, Cromwell had notice of potential litigation even earlier: in October, 1995, Cromwell received a FOIA request seeking materials related to the hiring process, and on August 8, 1995, in a meeting with Nappi and Cassella, Byrnie had expressed concerns about the hiring process and asked for copies of the interview questions. Gere testified that “shortly after the process was completed it became very apparent to us that Mr. Byrnie had concerns,” and explained that when the FOIA request was received, he was assured by Nappi that “information” had been “retained.” See Kronisch, 150 F.3d at 127 (documents destroyed years before suit brought could reasonably be found to have been destroyed in anticipation of litigation where fear of potential future litigation plausibly motivated the spoliation). Although no one could say when all the documents related to the 1995 art teacher job search were destroyed, Gere testified at his deposition that destruction of personnel records sometime after an employee search was concluded was a routine process. However, in the end, it does not matter when the documents were destroyed since even if the documents were destroyed days after the search ended and before anyone had wind of Byrnie’s “concerns,” Cromwell was still required by federal regulations implementing Title VII and the Americans with Disabilities Act to retain all records pertaining to employment decisions for a period of two years. See 29 C.F.R. § 1602.40. 5 Several courts have held that destruction of evidence in violation of a reg *109 ulation that requires its retention can give rise to an inference of spoliation. See Latimore v. Citibank Fed. Sav. Bank, 151 F.3d 712, 716 (7th Cir.1998) (“The violation of a record[-]retention regulation creates a presumption that the missing record contained evidence adverse to the violator.”); Favors v. Fisher, 13 F.3d 1235, 1239 (8th Cir.1994) (because employer violated record retention regulation, plaintiff “was entitled to the benefit of a presumption that the destroyed documents would have bolstered her case”); Hicks v. Gates Rubber Co., 833 F.2d 1406, 1419 (10th Cir.1987) (same); see also Steffen Nolte, The Spoliation Tort: An Approach to Underlying Principles, 26 St. Mary’s L.J. 351, 368-69 (1995) (collecting cases announcing a tort cause of action for spoliation based on violation of record-retention regulations). We agree that, under some circumstances, such a regulation can create the requisite obligation to retain records, even if litigation involving the records is not reasonably foreseeable. For such a duty to attach, however, the party seeking the inference must be a member of the general class of persons that the regulatory agency sought to protect in promulgating the rule. Cf. W. Page Keeton, et al., Prosser and Keaton on Torts § 36, at 224-25 (5th ed. 1984) (describing similar requirement for finding a statutorily created duty in tort). For example, violation of a rule that records be retained for securities disclosure purposes would not create a duty to preserve covered records for use in a subsequent employment discrimination suit. On the other hand, where, as here, a party has violated an EEOC record-retention regulation, a violation of that regulation can amount to a breach of duty necessary to justify a spoliation inference in an employment discrimination action. Cf. Favors, 13 F.3d at 1239 (Title VII rules, Title VII action); Hicks, 833 F.2d at 1419 (same). In such a case we can be confident that the responsible agency had in mind persons in the plaintiffs position and accordingly that our finding a duty to preserve documents in such circumstances will advance the goals of the rule. Although a regulation may supply the duty to preserve records, a party seeking to benefit from an inference of spoliation must still make out the other usual elements of a spoliation claim. The party must demonstrate first that the records were destroyed with a culpable state of mind (i.e. where, for example, the records were destroyed knowingly, even if without intent to violate the regulation, or negligently). Second, a party must show that the destroyed records were relevant to the party’s claim or defense. We are satisfied that Byrnie has established both. Cromwell admitted its policy to destroy such records soon after the hiring process had been completed and that the records in this case were so destroyed. At no point has Cromwell asserted that their destruction was merely accidental. This is evidence of intentional destruction sufficient to show a culpable state of mind on Cromwell’s part. Given that we have held that bad faith — an intent to obstruct the opposing party’s case — need not be shown to justify an inference of spoliation, see Reilly, 181 F.3d at 267, intentional destruction of documents in the face of a duty to retain those documents is adequate for present purposes. There is also ample evidence that the documents were relevant to Byrnie’s case. While the district court was correct in finding the missing applications to be minimally relevant, the missing notes the interviewers- made during the first and second rounds of interviews have more relevance. Those notes would clarify what aspects of Byrnie’s interview performances were reflected in the poor subjective evaluation he received from the Search Committee and whether that evaluation adhered to permissible criteria. Further, the documents Nappi relied upon in preparing the memo that formed the basis of Cromwell’s *110 CCHRO Answer would be relevant as well, given that the answer has been undermined by the deposition testimony of Nap-pi and others. Keeping in mind that “holding the prejudiced party to too strict a standard of proof regarding the likely contents of the destroyed evidence would subvert the prophylactic and punitive purposes of the adverse inference, and would allow parties who have intentionally destroyed evidence to profit from that destruction,” we think there is a “likelihood that the destroyed evidence would have been of the nature alleged by the party affected by its destruction.” Kronisch, 150 F.3d at 127-28. Again, the fact that Cromwell had to retreat from the specifics of its initial explanation for failing to hire Byrnie towards a more general, but less plausible, claim that Byrnie simply lacked familiarity with the basic competencies needed for good teaching makes it possible to infer that these explanations are pretext covering over the real explanation that would be disclosed by the destroyed documents. Cromwell argues that even if Byrnie is entitled to an adverse inference, such an inference must be limited to giving the greatest weight possible to other existing evidence favorable to the plaintiff. However, the rule offered by Cromwell rewards those most thorough in the art of document shredding since as the existing evidence of unlawful behavior dwindles, the deterrent force of the threat of an adverse inference fades as well. Cromwell relies on Felice v. Long Island R.R. Co., where we noted that, in a case where both parties failed to call relevant witnesses, the “jury should not be encouraged to base its verdict on what it speculates the absent witnesses would have testified to, in the absence of some direct evidence.” 426 F.2d 192, 195 n. 2 (2d Cir.1970). Because we are dealing with destroyed evidence, we adhere to Kronisch, where we held that a plaintiff had “produced enough circumstantial evidence to support the inference that the destroyed [evidence] may have contained documents supporting (or potentially proving) his claim, and that the possibility that a jury would choose to draw such an inference, combined with plaintiffs circumstantial evidence, is enough to entitle plaintiff to a jury trial.” Kronisch, 150 F.3d at 128. As Kronisch makes clear, a party seeking an adverse inference may rely on circumstantial evidence to suggest the contents of destroyed evidence. It then becomes a matter for the jury to decide, based on the strength of the evidence presented, whether the documents likely had such content. Thus, based on the record before us at summary judgment, enough circumstantial evidence exists to permit a reasonable trier of fact to conclude that the destroyed documents would show unlawful discrimination. Whether a reasonable trier of fact actually will draw such ah inference is a matter left to trial. The record before us presents a close case, and, as we have recognized elsewhere, “at the margin, where the innocent party has produced some (not insubstantial) evidence in support of his claim, the intentional destruction of relevant evidence by the opposing party may push a claim that might not otherwise survive summary judgment over the line.” Id. We find that substantial evidence supports Byrnie’s claim. Byrnie had clearly superior paper credentials while Mancarella only survived the initial application screenings due to procedural irregularities — yet her application was ranked higher than Byrnie’s. Further, Cromwell modified its justification for the hiring decision in the period between its CCHRO Answer and the present litigation. One of its current justifications for not hiring Byrnie — that he was unfamiliar with the basic competencies of effective teaching — further weakens Cromwell’s credibility because of its implausibility and Cromwell’s failure to adduce any evidence in support of the assertion. In the final analysis, while the foregoing additional evidence might not have been sufficient in itself to defeat summary judg *111 ment, it does when coupled with the allowable inference of spoliation. We therefore conclude that a jury could reasonably find that the documents Cromwell destroyed — • most prominently, the notes relied upon by Nappi in preparing the account of the hiring process that formed the basis of the CCHRO Answer and any notes made during the candidate interviews — would have filled out the picture created by the evidence Cromwell actually did produce in an unflattering manner by showing unlawfully discriminatory reasons informed the decision not to hire Byrnie. II. Byrnie’s Other Claims Because we find that summary judgment against Byrnie on his disparate treatment claims is inappropriate, we do not address Byrnie’s contention that, based on his affidavits and depositions, genuine issues of fact exist that the district court overlooked. Similarly, we do not reach the question whether statistical data presented by Byrnie showing that Cromwell tended to hire a lower percentage of male teachers than were hired on average statewide in Connecticut defeated summary judgment on his disparate treatment claim. Byrnie also objects to the district court’s dismissal of his disparate impact claim. A plaintiff makes out a prima facie case for a disparate impact claim by showing “that a facially neutral employment policy or practice has a significant disparate impact.” Brown v. Coach Stores, Inc., 163 F.3d 706, 712 (2d Cir.1998). Simply gesturing towards the hiring process as a whole will not satisfy the requirement that the plaintiff identify a “specific employment practice” that is the cause of the statistical disparities. Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 656, 109 S.Ct. 2115, 104 L.Ed.2d 733 (1989) (quoting Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 994, 108 S.Ct. 2777, 101 L.Ed.2d 827 (1988) (O’Connor, J.) (plurality)). Byrnie contends that Cromwell’s reliance upon the CCI as a guide for evaluating candidates has resulted in a hiring process that disproportionately favors younger candidates, as evidenced by the fact that the average age of teachers hired by Cromwell is 35. 6 However, as the district court pointed out, since the CCI was not developed until 1992, all teachers who received their teaching certificate prior to 1992 would apparently be adversely affected by unfamiliarity with the CCI’s terminology — thereby impacting an age group extending well beyond the group protected by the ADEA. See Byrnie, 73 F.Supp.2d at 222. Accordingly, the district court appropriately dismissed Byrnie’s claim of age discrimination via disparate impact. Similarly, because Byrnie fails to suggest what specific employment practice accounts for a statistically low number of males being hired at Cromwell, summary judgment was properly granted to Cromwell on that claim as well. CONCLUSION Viewing the record as a whole, we find enough evidence to allow Byrnie’s disparate treatment claims to survive summary judgment. The superiority of Byrnie’s paper credentials in comparison to the chosen candidate’s credentials — along with the fact that the chosen candidate was not even technically qualified for the job — when combined with the weak plausibility of the employer’s explanations for its hiring decision — particularly in light of the misleading nature of the employer’s initial explanation offered to the CCHRO during its investigation of Byrnie’s dis *112 crimination claims — provides sufficient circumstantial evidence to make reasonable a jury’s inference that Cromwell’s spoliation of documents related to the hiring process destroyed evidence that would allow a finding of unlawful discrimination. Nevertheless, the district court correctly found that Byrnie’s statistical evidence, because unconnected to any specific employment practice or policy, cannot support a disparate impact claim. Accordingly, we affirm the district court’s grant of summary judgment on Byrnie’s disparate impact claims, but reverse the grant of summary judgment on Byrnie’s disparate treatment claims. Footnotes [1] Byrnie also asserted claims under comparable provisions oí Connecticut's Fair Employment Practices Act, Conn. Gen.Stat. Ann. § 46a-60(a)(l). Those claims were dismissed, along with the federal claims, at summary judgment. [2] These, at least, were the review criteria according to Cromwell’s answers to Bj'rnie’s discovery interrogatories. At deposition testimony, however, Nappi stated that "[tjhere was no set criteria or qualifications" for reviewing applications and that he could not remember what guided his review if not a set criteria. [3] Cromwell notes that its "Recommended Procedures for Hiring” charges the building level administrator with setting the qualifications for any given job. Be that as it may, the selected qualifications must, according to the same document, be reviewed by the Board of Education. The reviewed and approved list of job qualifications for an art teacher requires the successful applicant to possess a degree in art education. [4] We know the rankings of the applicants chosen by the screening committee for interviews because they were included in the CCHRO Answer. Similarly, the CCHRO Answer contains a tally of the votes each candidate received from the Search Committee after the first round of interviews. [5] 29 C.F.R. § 1602.40 states, in pertinent part, that "[a]ny personnel or employment record made or kept by a ... school (including ... application forms submitted by applicants and other records having to do with hiring ... ) shall be preserved by such ... school ... for a period of two years from the date of the making of the record or the personnel action involved, whichever occurs later.” The regulation implements 42 U.S.C. § 2000e-8(c) which requires "[e]veiy employer ... subject to this subchapter” to "(1) make and keep such records relevant to the determinations of whether unlawful employment practices have been or are being committed, [and] (2) preserve such records for such periods ... as the [EEOC] shall prescribe by regulation or order....” We believe that the voting tally sheets and interview notes destroyed by Cromwell fall within § 1602.40’s category of "records having to do with hiring.” [6] Bymie's age discrimination disparate impact claim has been finessed a bit from the version offered to the district court where he alleged under-representation of teachers over age 50. Otherwise his age discrimination disparate impact claim would simply fail because he has not alleged, as he must, that the employment practice affects the entire protected group. See Criley v. Delta Air Lines, Inc., 119 F.3d 102, 105 (2d Cir.1997) (per curiam) (noting that ADEA disparate impact claim "must allege a disparate impact on the entire protected group, i.e., workers aged 40 and over”).