No. 665, Docket 83-7767.United States Court of Appeals, Second Circuit.Argued January 25, 1984.
Decided July 3, 1984.
[EDITORS’ NOTE: THIS PAGE CONTAINS HEADNOTES. HEADNOTES ARE NOT AN OFFICIAL PRODUCT OF THE COURT, THEREFORE THEY ARE NOT DISPLAYED.]
Judith P. Vladeck, New York City (Joseph J; Garcia, Vladeck, Waldman, Elias Engelhard, New York City, of counsel), for plaintiffs-appellants.
Lillian Z. Cohen, Asst. Atty. Gen., New York City (Robert Abrams, Atty. Gen., Melvyn R. Leventhal, Stanley A. Camhi, Michael Tietz, Asst. Attys. Gen., New York City, of counsel), for defendants-appellees.
Appeal from the United States District Court for the Eastern District of New York.
Before MANSFIELD, PIERCE and WINTER, Circuit Judges.
WINTER, Circuit Judge:
 A certified class of current and former female employees of the State University of New York at Stony Brook (“Stony Brook” or “the University”) appeals from a judgment entered on August 8, 1983 by Judge Pratt. After a twelve-day bench trial, he held that Stony Brook had not engaged in a pattern and practice of sex discrimination. 587 F.Supp. 572 (E.D.N.Y. 1983).
 We affirm.
 This action was commenced under Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq. (“Title VII”) by thirty-four current and former female employees of Stony Brook. They alleged, inter alia, that Stony Brook had engaged in a system-wide pattern and practice of discrimination against teaching (“faculty”) and non-teaching professional (“NTP”) women with respect to virtually all conditions of employment including recruitment, hiring, placement at hire, promotion, tenure, and salary.
 By a memorandum and order dated June 22, 1977, the district court certified a class under Fed.R.Civ.P. 23(b)(2) “solely with respect to plaintiffs’ claim that at Stony
Brook there exists a system-wide pattern and practice of discrimination based on sex.” 587 F.Supp. 571, 574 (E.D.N.Y. 1983). The class certified consisted of:
all women who in the past may have been or in the future may be discriminated against on the basis of sex by defendants’ practices with respect to recruiting, hiring, termination, job assignment, promotion, compensation, and other terms, privileges, and benefits of employment and who either (1) have been employed by Stony Brook as a teaching or non-teaching professional at any time on or after February 11, 1974, or (2) may be so employed by Stony Brook in the future, or (3) once unsuccessfully applied for employment as a teaching or non-teaching professional with Stony Brook at any time on or after February 11, 1974, or (4) who may apply for such employment in the future. Id. at 575.
 After class certification, the parties and the district court agreed to a bifurcated trial in which the class issue of a pattern and practice of system-wide discrimination would be tried first, while trial of the claims of the individual plaintiffs would await final adjudication of the class claim.
 The system-wide, pattern and practice claim proceeded to trial after five years of discovery. The evidence consisted largely of testimony and documents concerning the structure and efficacy of Stony Brook’s affirmative action program, anecdotal testimony concerning individual cases of discrimination, and experts’ reports based largely on statistical data.
 Judge Pratt’s decision, 587 F.Supp. 572, familiarity with which is assumed, dismissed the complaint as to the class claims and directed that judgment be entered under Fed.R.Civ.P. 54(b). Conceding that “plaintiff’s statistics … show[ed] that Stony Brook’s work force [was] `sex-stratified’, in that women [were] distributed primarily in lower-level … teaching and administrative positions,” id. at 576, he nevertheless concluded that this stratification was not the result of Stony Brook’s use of gender tainted criteria or of facially neutral criteria that are not job-related and fall disproportionately upon women. Rather, he found that it was largely the result of historic and social conditions over which the defendants had no control.
 Plaintiffs have appealed this judgment and have briefed their claims relating to faculty and NTP placement at hire, faculty tenure and promotion decisions, awards of tenure to faculty at hire, and faculty and NTP salaries. They have not briefed their claims relating to hiring and recruitment or comparable worth. We consider these latter claims abandoned on appeal.
 DISCUSSION 1. Claims of Legal Error
 We deal here with familiar legal principles. In order to prevail on their claim of a pattern and practice of discrimination, plaintiffs had to show by a preponderance of the evidence that Stony Brook’s “standard operating procedure — the regular rather than the unusual practice” is to discriminate on the basis of sex. International Brotherhood of Teamsters v. United States, 431 U.S. 324, 336, 97 S.Ct. 1843, 1855, 52 L.Ed.2d 396 (1977).
 Plaintiffs claim that Judge Pratt erred as a matter of law in two respects. Their first claim focuses upon his statement that “plaintiffs’ burden was to establish that after February 11, 1974, … with respect to its teaching and non-teaching professionals, Stony Brook has engaged in a pattern or practice of treating women less favorably than men, solely because they are women.” 587 F.Supp. at 577 (emphasis added). Plaintiffs argue that this statement demonstrates that the district court limited its consideration to the issue of discriminatory motive or treatment and did not consider whether the various criteria applied by Stony Brook in making employment decisions had an unlawful disparate impact. Griggs v. Duke Power Co., 401 U.S. 424, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971). After reviewing the decision of the district court as a whole, we must reject this argument. Judge Pratt’s findings of
fact refute not only the use of gender-tainted criteria but also the use of facially neutral criteria that are not job related but have a disparate impact upon women. Indeed, at several places in his opinion, he utilizes disparate impact analysis in concluding that prior experience, prior academic rank, and prior tenure are job related criteria. 587 F.Supp. at 585. In light of the explicit consideration of this claim, we perceive no significance in the isolated use of the word “solely.”
 The second claim of error of law is that Judge Pratt failed to apply to universities the legal obligations applicable to other employers under Title VII. Plaintiffs argue that he treated universities as a preferred, protected class of employer and point in particular to his statement that a university’s “treatment of minorities cannot be evaluated in the same manner as that of, say, a trucking company or factory that can hire many people with less differentiated qualifications.” 587 F.Supp. at 584.
 Although we agree that the legal obligations generally applicable to employers under Title VII are fully applicable to universities, we do not agree that Judge Pratt held universities to a less rigorous standard. His statement merely pointed out that generalized statistical data may be less persuasive evidence of discrimination where an employer hires “highly educated, specially qualified people” 587 F.Supp. at 584 on a decentralized basis than where the positions in question involve general skills and a central office does the hiring. His statements thus had to do with the logical strength of the inferences which may be drawn from certain kinds of statistics rather than with the legal duty imposed upon particular employers.
 Where uncoordinated and independent employment decisions are made by different persons, statistics as to hiring by the overall entity may be less significant in demonstrating bias than where a single office makes all employment decisions. Zahorik v. Cornell University, 729 F.2d 85, 92 (2d Cir. 1984). For example, one can draw stronger inferences about hiring criteria when persons are hired or not hired by a central office than when they are hired or not hired by different offices. The same is true when highly specialized jobs are in issue, as opposed to identical positions requiring only general skills. This is so, not because specialists are protected less by Title VII, but because the availability pool for certain specialized positions may be very small. As the size of the sample population declines, we can be less confident of the validity of statistical proof, because pure chance may produce a sample that deviates from population norms.
 We also recognize the possibility that the sex characteristics of a particular availability pool may differ from those of the general population because of prior discrimination by entities other than Stony Brook. However, Title VII imposes on Stony Brook only the duty to refrain from using non-job-related criteria that may perpetuate such prior discrimination. Griggs v. Duke Power Co., 401 U.S. at 430-31, 91 S.Ct. at 853. Therefore, where a statistical disparity between men and women may be explained by taking job-related criteria such as prior work experience or prior academic rank into account, evidence of the disparity does not prove unlawful discrimination.
 We do not suggest that statistical proof cannot be used in Title VII actions against universities. We mean only that where an employer hires specialists on a decentralized basis, litigants must either tailor the data to take that fact into account or analyze generalized statistics in light of the facts that independent, uncoordinated decisions are involved, the availability pool is a small sample, and the characteristics of the availability pool will vary from position to position.
 2. Claims of Factual Error
 We address here Judge Pratt’s findings as to Stony Brook’s affirmative action program, its relationship to plaintiffs’ allegations of a pattern or practice of sex discrimination
and the statistical data offered by plaintiffs in support of the factual claims presently before us. Again assuming familiarity with Judge Pratt’s extensive review of the evidence, we limit our discussion to the reasons why we regard his conclusions concerning the effectiveness of the affirmative action program and the limited usefulness of plaintiffs’ statistical data as not clearly erroneous.
 A. Stony Brook’s Affirmative Action Program
 After reviewing the University’s affirmative action program, Judge Pratt found as a fact that “Stony Brook ha[d] acted in good faith, even before Title VII required it to do so, to ensure that its policies and practices have operated without discrimination based on sex.” 587 F.Supp. at 579. The evidence in the record supports his conclusion.
 Stony Brook’s affirmative action program subjects its employment decisions to mandatory procedures and to review by affirmative action officials at Stony Brook. Every faculty or NTP hiring decision is individually reviewed by such officials as to recruitment, the pool of applicants interviewed, and the selection of the person finally hired. These procedures also provide review of claims of other forms of discrimination. Women at Stony Brook participate actively as members of the various equal opportunity commissions.
 The evidence indicates this program operates in good faith. On several occasions, the University reviewed and altered various employment policies in response to complaints of sex discrimination. In 1974, for example, a review of faculty salaries resulted in retroactive salary increases for certain individuals. In 1975, the University’s affirmative action commission concluded that women were underutilized given national labor pools, and prompt efforts were made to correct this.
 The sole evidence in the record, apart from the plaintiffs’ statistical studies, indicating flaws in Stony Brook’s affirmative action plan are occasional departures from its procedures where the hiring of senior tenured faculty with offers from competing schools was involved. We agree with Judge Pratt that these instances were isolated, that they involved circumstances in which the usual search procedures would have resulted in Stony Brook’s losing opportunities to hire highly desirable candidates and that they did not involve discrimination based on sex. They are, therefore, not probative either of bad faith in the execution of the affirmative action program or of a pattern and practice of discrimination.
 It should be emphasized that the finding that Stony Brook has an effective affirmative action program is of central importance to this litigation. The existence of a comprehensive affirmative action program is the antithesis of a pattern and practice of discrimination based on sex. Such a program is evidence of an intent to eliminate gender as an employment criteria and to root out subtle forms of discrimination. It thus directly controverts a claim that discrimination is the “standard operating procedure.” Teamsters, 431 U.S. at 336, 97 S.Ct. at 1855.
 We note, moreover, that affirmative action programs such as that at Stony Brook aid plaintiffs in obtaining evidence for use in Title VII litigation involving a pattern and practice of discrimination. First, they subject employment decisions to a process which openly scrutinizes them on an individuals basis with a view to detecting discrimination. Second, they allow individuals to bring claims, to present evidence in support thereof, and to force an explicit decision by the employer on those claims. Such procedures thus provide an individualized record of hiring and other employment decisions and greatly facilitate the discovery of particular instances of discrimination.
 The failure of the plaintiffs here either to mount a significant challenge to
the good faith of Stony Brook’s affirmative action program or to provide a meaningful number of individual instances of discrimination is thus a major weakness in their case. We by no means take the position that the probative impact of the existence of an affirmative action program may not be rebutted by statistical evidence alone. However, where a pattern and practice of discrimination is alleged, such evidence must be weighed in light of the failure to locate and identify a meaningful number of concrete examples of discrimination given the existence of procedures which would ordinarily cause such examples to come to light.
 B. Faculty and NTP Placement at Hire
 In support of their claim of discrimination in the placement of faculty at hire, plaintiffs offered a statistical study that evaluated such placement on a university-wide basis as though Stony Brook hired male and female academics in groups and then assigned them to various positions throughout the university. This is a major flaw since, as Judge Pratt found, the University hires academics on an individual basis to fill specific teaching positions with particularized job requirements as they become open. The studies also did not consider prior academic rank or work experience in comparing the placement of male and female academics. Judge Pratt correctly held that prior rank and experience themselves may reflect academic merit and that universities may consider them as one of the many factors that govern faculty appointment decisions.
 Plaintiffs’ statistical evidence of discrimination in the initial placement of female NTP’s purported to show that women were disproportionately hired into lower-ranking NTP positions. These studies were also flawed since they did not take availability pools for each position into account and thus assumed that all NTP’s at Stony Brook were available to be hired into each NTP job at every level. Since there are a wide variety of discrete NTP positions at Stony Brook and some of those positions require considerable specialization, plaintiffs’ assumption was inconsistent with the record and their study was not probative of discrimination. Moreover, the defendants’ data demonstrated that more than the statistically expected number of women had been hired in each NTP rank when availability pools were taken into account.
 C. Faculty Promotion and Awards of Tenure at Hire
 The plaintiffs also submitted statistics purporting to show that women were concentrated in the lower academic ranks, received fewer second promotions than men, were rarely promoted to tenure and received a disproportionately low number of awards of tenure at hire. However, the studies relating to promotion and tenure again did not take into account the relevant pools of qualified applicants. They assumed, for example, that all faculty members were available for promotion to tenure each year, although newly-hired junior faculty, which included the preponderant number of women faculty, are not realistically eligible for promotion. In addition, defendant’s statistical studies showed that, when eligibility was taken into account, Stony Brook tenured similar proportions of male and female faculty.
 Plaintiffs’ statistical evidence on the issue of tenure at hire also did not take into account prior academic rank or prior tenure.
However, rank achieved elsewhere is relevant to an applicant’s qualification for tenure at Stony Brook, and thus may, along with other criteria, be taken into account in determining whether to award tenure to a newly-hired faculty member. Defendants’ study showed that, when prior rank is taken into account, no difference in the initial placement of men and women exists.
 D. Faculty and NTP Salary
 Plaintiffs’ statistical studies in support of their claims of salary discrimination against women NTP’s purported to show that men were paid more than women within each NTP rank. Judge Pratt concluded, however, that plaintiffs’ studies were both “unrealistic and illogical” because they took no account of either prior work experience or existing duties at Stony Brook. 587 F.Supp. at 603. He concluded that plaintiffs’ regression analyses could not realistically predict salaries or establish proof of salary discrimination among NTP’s unless such variables are taken into account. We agree.
 NTP ranks are established on a statewide basis without considering Stony Brook’s particular employment needs. At Stony Brook, each NTP rank encompasses a wide variety of discrete jobs which require different qualifications and experience. They also vary widely as to responsibilities and are not linked in a promotion chain. The NTP rank itself merely establishes outside parameters for salary and does not reflect the tasks or responsibilities of a particular job except in a highly general fashion. Existing duties and prior work experience are thus crucial variables that must be taken into account in comparing relative salaries of NTP’s. Since plaintiffs’ data failed in this regard, it is not probative of discrimination.
 Finally, the record clearly shows that NTP rank and salary are fixed for each position before persons are interviewed or hired to fill the slot. In view of that and the fact that NTP jobs are separate rather than links in a promotion chain, plaintiffs’ salary claim with regard to NTP’s seems merely to reargue their placement at hire claim.
 Plaintiffs make their strongest showing in the area of faculty salaries. Both parties focused heavily, at trial and on this appeal, on the statistical evidence relating to salaries. After carefully reviewing the record, we conclude that Judge Pratt’s finding that plaintiffs failed to prove a pattern or practice of discrimination with respect to faculty salaries is not clearly erroneous.
 We begin by observing that the determination of faculty salaries at Stony Brook is decentralized. Initial salaries normally are determined by a department chairman based on a combination of factors including education, qualifications, experience, the salary scale in the appointee’s discipline at Stony Brook, the rank of appointment, and the salaries of peers at other educational institutions in the same field. The department chairman recommends an initial salary, but then must obtain final approval from a university official and, when the starting salary exceeds $24,000, from the Chancellor of the State University system. Raises occur through across-the-board percentage increases, usually as a result of collective bargaining, and merit increases occur in the discretion of the department chairman subject to approval by SUNY Central. Raises also occur upon promotion.
 Plaintiffs’ primary evidence of gender-based discrimination in faculty salaries was a statistical report prepared by their expert, Dr. Killingsworth. Based on a multiple linear regression analysis of personnel data supplied by defendants, the report concluded that women employed as faculty at Stony Brook during the 1976-77 academic year were underpaid between $1706 and $2000 as compared to similarly qualified men.
 Judge Pratt found that plaintiffs’ report did not prove salary discrimination. He relied in part on a report prepared by defendant’s expert, Dr. Meier, which criticized Dr. Killingsworth’s study. The design of the competing studies largely reflects the parties’ different theories of this case. As Judge Pratt noted, much of the difference between plaintiffs’ and defendants’ studies stems from Dr. Killingsworth’s omission from his regression analysis of any variable reflecting prior experience. While this is consistent with the contention that recognition of these criteria merely perpetuates prior discrimination that occurred elsewhere, Judge Pratt rightly rejected that contention on the grounds that prior experience is job related.
 There were also other differences between Dr. Killingsworth’s and Dr. Meier’s studies. Dr. Meier’s study separately compared faculty in each of Stony Brook’s academic divisions. Dr. Killingsworth’s study, on the other hand, aggregated faculty into broader groups by fields of degree and used inconsistent aggregations. Dr. Killingsworth’s study also included the salaries of visiting faculty, lecturers, artists and performers, and other categories that Judge Pratt found not comparable to regular faculty. 587 F.Supp. at 602.
 Even defendants’ study, however, found a smaller but statistically significant female salary disadvantage as of 1976-77. Dr. Meier nevertheless argued that this figure did not prove salary discrimination. First, he pointed out that his study found no statistically significant salary disadvantage for women hired after 1972, a finding confirmed by Dr. Killingsworth’s study. Second, Dr. Meier concluded that the rate of salary increase for pre-1972 hires in every year after 1972 was either the same for men and women, or slightly favored women.
 Thus, the statistical evidence arguably shows that defendants discriminated with respect to salary before 1972 and that a residual effect of that discrimination continues to exist. However, Dr. Meier also concluded that the salary difference resulting from inclusion of the pre-1972 group disappears when unique categories of employees such as faculty members at the defunct Education Department, artists, performers and the Nobel Laureate are excluded from the study. When these employees are excluded and results for the pre-1972 and post-1972 hires are averaged, women faced no statistically significant salary disadvantage as of the 1976-77 academic year.
 Moreover, women faculty presently at Stony Brook who were hired before 1972 are a very small group, and easily identified individually. Nevertheless no direct evidence of discrimination as to them other than the statistical study was produced. In the context of Stony Brook’s comprehensive affirmative action program and Dr. Meier’s explanation of the disparity in salary revealed in his study, the failure to produce such direct evidence is significant. Judge Pratt’s conclusion that the statistical evidence of possible salary disparities is too insubstantial to prove a pattern and practice of discrimination is not, therefore, clearly erroneous.