Melani v. Bd. of Higher Educ. of City of New York, 561 F. Supp. 769 (S.D.N.Y. 1983)

*770 *771 Vladeck, Elias, Vladeck & Engelhard, P.C., New York City, for plaintiffs; Judith P. Vladeck, Joseph J. Garcia, New York City, of counsel.

Frederick A.O. Schwarz, Jr., Corp. Counsel, New York City, for defendant; Norma Kerlin, Debra Jaret, New York City, of counsel.

GAGLIARDI, District Judge.

Plaintiffs commenced this action pursuant to Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e ("Title VII"), and 42 U.S.C. § 1983 against the Board of Trustees of the City University of New York ("Board").^[1] The action is based upon alleged sex discrimination in the Board's employment practices. The Board is a body corporate, existing under Articles 125 and 126 of the New York Education Law and governs and administers the City University of New York ("CUNY"), a public higher education institution consisting of nineteen *772 educational units.^[2] The court has certified a plaintiff class defined as all women now employed by the Board as members of the professional instructional staff at CUNY, or who at any time since October 1968 have been so employed or have sought such employment.^[3] Plaintiffs allege sex discrimination in virtually all facets of defendant's employment practices including hiring, promotion, salary, and fringe benefits.

Because defendant's alleged salary discrimination is the focal point of this controversy, a trial on that limited issue was held, without a jury, on June 17-19 and June 23-24, 1980. Since that time the court has refrained from issuing a decision on the salary issue and has held the remainder of the case in abeyance, due to pending settlement negotiations which appeared likely to resolve the entire case. Having been advised that settlement efforts have broken down completely, the court pursuant to Rule 52(a), Fed.R.Civ.P., now makes the following findings of fact and conclusions of law with regard to plaintiffs' Title VII salary discrimination claims.

New York Education Law § 6212 and the By-Laws of the Board ("By-Laws") § 6.1 set forth the positions designated as the CUNY professional instructional staff ("instructional staff"). These job titles include both the teaching positions of professor, associate professor, assistant professor, lecturer and instructor, as well as administrative and service positions such as registrar, college laboratory technician, and the higher education officer ("HEO") job titles.^[4]

The salary schedules for the instructional staff are set in accordance with New York Education Law § 6220, Article XII of the By-Laws, and the relevant collective bargaining agreements. The Board has ultimate responsibility for all appointments of faculty, and, with few exceptions, for appointments of non-faculty instructional staff. See By-Laws § 6.6. The Board sets salaries for instructional staff members not covered by the collective bargaining agreement, see By-Laws § 12.1, and approves any supplemental salary payments to an instructional staff member in addition to those set by the collective bargaining agreement. See By-Laws §§ 12.2 and 12.4.

Members of the instructional staff have been represented by several labor organizations. Since September 1, 1972, the Professional Staff Congress ("PSC/CUNY") has represented individuals in most of the instructional staff job titles. The Board has entered into collective bargaining agreements with PSC/CUNY covering the periods from September 1, 1973 to August 31, 1975; from September 1, 1975 to August 31, 1977; from September 1, 1977 to August 31, 1978; and from September 1, 1978 to August 31, 1980. These agreements set forth as many as ten salary levels within each job title and provide for automatic raises for each year of the agreement.

The plaintiffs in a Title VII action may establish a prima facie case of classbased discrimination by "demonstrating the existence of a discriminatory pattern and practice." International Brotherhood of Teamsters v. United States, 431 U.S. 324, 359, 97 S. Ct. 1843, 1866, 52 L. Ed. 2d 396 (1977) (quoting Franks v. Bowman Transportation Co., 424 U.S. 747, 772, 96 S. Ct. 1251, 1267, 47 L. Ed. 2d 444 (1976)).^[5] In order *773 to prove such a pattern or practice, plaintiffs must demonstrate more than "isolated or `accidental' or sporadic discriminatory acts." They must show that discrimination was "the regular rather than the unusual practice" of the defendant. Teamsters v. United States, supra, 431 U.S. at 336, 97 S. Ct. at 1855.

In determining whether proof of discriminatory intent is a necessary part of a Title VII plaintiff's prima facie showing, the court first must determine whether the claim is one of disparate treatment or of disparate impact. In a claim of disparate treatment, the plaintiff alleges that:

[t]he employer simply treats some people less favorably than others because of their race, color, religion, sex or national origin. Proof of discriminatory motive is critical, although it can in some instances be inferred from the mere fact of differences in treatment.

Teamsters v. United States, supra, 431 U.S. at 335 n. 15, 97 S. Ct. at 1854 n. 15. By contrast, in a disparate impact case the plaintiff alleges facially neutral employment practices that have a more severe impact on a particular group. Proof of discriminatory intent is not essential to such a case. Id.

Plaintiffs in the instant case allege that defendant as a general practice pays less to female members of the instructional staff than it pays to similarly qualified male members. Accordingly, under the above definition, plaintiffs have raised a claim of disparate treatment. Plaintiffs' initial burden, therefore, is to show a pattern or practice of intentional salary discrimination against women.^[6]

In asserting their claim of disparate treatment, plaintiffs rely almost exclusively upon a series of statistical studies of the CUNY instructional staff conducted by Dr. Mark R. Killingsworth, a labor economist and professor of economics at Barnard College of Columbia University.^[7] The source for these studies was the CUNY Instructional Staff Profile ("ISP") data tapes which contained information regarding the instructional staff for each semester from fall 1972 to fall 1977, inclusive.

Plaintiffs' Study I is comprised of descriptive statistical compilations for two groups of the full-time, active CUNY instructional staff included in the ISP tapes. The first group consists of all such CUNY employees. The second includes only those employees hired after September 1972 ("post-72 hires").^[8] The various compilations within each group measure the average salary differential between men and women within sub-classes formed on the basis of a number of productivity-related factors: highest degree, age, years of CUNY service and years since the individual's highest degree. The survey of all staff on the ISP tapes indicates an overall average male-female salary differential of approximately $3,530. Sizeable salary differences between men and women persist when salaries are evaluated for staff comparable with regard to one or two of the characteristics mentioned *774 previously.^[9] Similar differences characterized the average earnings of men and women hired after September 1972.

Plaintiffs' Study II considers the combined effect on salary of the factors described in Study I along with a number of additional factors. To do so, plaintiffs' expert used the statistical technique known as multiple regression analysis. As with Study I, separate analyses were performed for all instructional staff on the ISP tapes and for the post-72 hires.

Multiple regression analysis permits the calculation of the effect of a variety of factors called independent variables on a particular result called the dependent variable. The use of the computer both reduces the time necessary to perform this analysis and makes it possible to include a large number of independent variables whose effect on the dependent variable may be thus determined. This technique has been recognized as useful for measuring the average salary differential, if any, between men and women or blacks and whites while controlling for or factoring out the effects of other productivity-related characteristics which may lead to differences in compensation. See Trout v. Hidalgo, 517 F. Supp. 873, 877-78 (D.D.C.1981); Greenspan v. Automobile Club of Michigan, 495 F. Supp. 1021, 1061 (E.D.Mich.1980); Fisher, Multiple Regression in Legal Proceedings, 80 Colum.L.Rev. 702, 721-25 (1980).

The regression analyses in plaintiffs' Study II included as many as 98 independent variables. These variables consist of various measurements of age, years of CUNY service, academic degrees, quality of academic degree,^[10] certificates and credentials (such as R.N., C.P.A. or Public Engineer), time elapsed since the completion of each degree, and time elapsed between the completion of successive degrees. An independent variable for sex was included in each of the analyses.

Plaintiffs thus sought to determine the average difference in salary for men and women while factoring out the effects of the independent variables studied. This analysis determined that the average salary differential between men and women with regard to all instructional staff on the ISP tapes was approximately $1800. With respect to post-72 hires, the differential was approximately $1,600. Plaintiffs also performed regression analyses for several subsets of the entire instructional staff membership and of the post-72 hires. These subsets included those individuals hired no more than one year after they received their highest degree, those hired within one year of receiving a Ph.D., those hired at age 30 or less, and those possessing a Ph.D. and hired at age thirty or less. In each case, plaintiffs found large average salary differentials between men and women after the effects of the other variables included in the studies were factored out.

The technique of regression analysis also provides a measure of the probability that the observed salary differential between men and women in each study resulted from random or chance factors. This measure is called the "P" value. Statisticians generally regard a measured differential as being statistically significant when the P value is .05 or less, that is, when the likelihood that the differential is the result of chance is less than 5 in 100. See D. Baldus & J. Cole, Statistical Proof of Discrimination *775 §§ 9.02 and 9.03 (1980). In virtually all of plaintiffs' studies, the P value indicates that the salary differential between men and women is statistically significant.^[11] Plaintiffs condensed these results in the following table:

                     Dollar Female Earnings Differential
                  Full-Time Active CUNY Employees, Fall 1977
      (Relative to Male Employees with Same Productivity Characteristics)
                                             Female Dollar Disadvantage
                                           (Chance Probability in Brackets)         
         Group                   All Persons                Persons Hired 10/72 or Later
          All                - $1,803. [p = *]              - $1,629. [p = .XXXXXXXXX]
                             (n = 5917, v = 98)             (n = 856, v = 70)
   Hired no more than
   1 year after receipt
   of highest degree         - $1,438. [p = *]              - $1,400. [p = .000031]
                             (n = 2822, v = 75)             (n = 370, v = 58)
   Hired no more than
   1 year after receipt
   of highest degree -
   highest degree was
   Ph.D                      - $ 987. [p = .000059]         - $1,164. [p = .0206]
                             (n = 750, v = 41)              (n = 135, v = 37)
   Hired at age 30
   or less                   - $1,289. [p = *]              - $ 896. [p = .0011]
                             (n = 2141, v = 70)             (n = 324, v = 50)
   Hired at age 30
   or less - highest
   degree was Ph.D           - $ 835. [p = .0011]           - $ 478. [p = .3149]
                             (n = 406, v = 39)              (n = 76, v = 35)
   Notes n = number of persons considered
         v = number of variables used
         * = infinitesimal

Plaintiffs also submitted a supplemental study ("Supplemental Report") focusing on sex discrimination in assignment of faculty to the ranks of professor, associate professor, assistant professor, instructor and lecturer. One of the studies in plaintiffs' Supplemental Report used the statistical technique of logit analysis. Logit analysis is conceptually similar to a regression analysis in which the dependent variable has discrete and not continuous values. See Greenspan v. Automobile Club of Michigan, supra, 495 F. Supp. at 1063. In plaintiffs' logit analysis, the dependent variable was rank and the independent variables consisted of productivity-related factors used in *776 plaintiffs' regression analyses. Plaintiffs thereby sought to measure for CUNY faculty hired since October 1972 the effect of sex on rank when the productivity-related independent variables were held constant. Plaintiffs compared the actual data on the number of female faculty members assigned to each rank with the expected number which would have been assigned to each rank if the women were treated as male faculty having the same productivity characteristics.^[12] These logit studies indicate that females, in comparison with males with similar characteristics, generally are disproportionately underrepresented in the upper ranks of the faculty and overrepresented in the lower ranks.

Defendant raises a number of criticisms of plaintiffs' studies and claims that plaintiffs therefore have failed to establish a prima facie case. Defendant argues as an initial matter that plaintiffs cannot rely exclusively on statistical evidence to demonstrate an employer's discriminatory motive, an essential element of a disparate treatment case. The Supreme Court has held, however, that when gross statistical disparities can be shown, they alone may constitute prima facie proof of an intentional pattern or practice of discrimination. See Hazelwood School District v. United States, 433 U.S. 299, 307-308, 97 S. Ct. 2736, 2741-2742, 53 L. Ed. 2d 768 (1977); Hudson v. International *777 Business Machines Corp., 620 F.2d 351, 355 (2d Cir.), cert. denied, 449 U.S. 1066, 101 S. Ct. 794, 66 L. Ed. 2d 611 (1980).

Defendant criticizes plaintiffs' descriptive presentation of salaries, Study I, as inadequate, in part because it considers male and female staff comparable with regard to no more than two characteristics. Defendant correctly recognizes that such a study is not highly probative because it fails to control for multiple influences on salary. See Wilkins v. University of Houston, 654 F.2d 388, 401-402 (5th Cir. 1981). For example, women may earn less than men having the same degree not as a result of discrimination but because those women have less experience. Study I, without more, would be insufficient to establish plaintiffs' prima facie case. Plaintiffs, however, rely primarily on their regression analyses which are not subject to this criticism.

Defendant also raises a series of objections to plaintiffs' regression analyses. Defendant contends that these studies are fatally flawed because they do not distinguish between instructional staff occupying faculty positions and those occupying administrative positions such as registrar, laboratory technician and the HEO job titles. Defendant claims that plaintiffs' studies, by aggregating these positions, reflect as sex-based salary differentials, differences in pay that actually result from differences in job responsibilities or titles.

Plaintiffs respond that their studies account for salary differences attributable to job title, and particularly to any distinctions between faculty and non-faculty positions through the independent variables of age, degree and CUNY experience which were included in their regression analyses. Dr. Killingsworth testified, for example, that of the non-teaching instructional staff, fewer than 5% had Ph.D. degrees, while approximately 36% of the teaching staff had obtained that degree. Similarly, while only 18% of the non-faculty staff had either an M.A. or Ph.D. or both, at least half of the faculty members had earned one of these degrees. In addition, virtually all the instructional staff members who do not have at least a B.A. degree are not faculty members. Dr. Killingsworth also testified that the faculty members on the average were several years older and had several more years of CUNY service than nonteaching members of the instructional staff. Dr. Killingsworth concluded: "If one includes variables of those kinds [degree, age, years of CUNY service] those variables reflect those differences that exist between the teaching faculty and the non-teaching faculty."

The court finds convincing Dr. Killingsworth's testimony that the independent variables in his Study II regression analyses adequately account for any salary differential attributable to the distinction between faculty and non-faculty positions. Although the correlation between the independent variables and the faculty/non-faculty dichotomy is not perfect, the data indicate that the variables included in the model sufficiently account for the distinction between faculty and non-faculty positions.^[13]

Defendant also criticizes the failure of plaintiffs' studies to distinguish among the different non-faculty job titles. This criticism is not persuasive. It is true that if employees hold positions of a widely disparate nature a regression analysis must control for the differences in job title by including either variables measuring additional qualifications specific to certain job titles or a variable for the job titles themselves. See Valentino v. United States Postal Service, 674 F.2d 56, 70-71 (D.C.Cir.1982). However, where the employees are similarly qualified, and hold similar jobs, their salaries may be appropriately compared through regression analysis without such variables. Id. Since an examination of the By-law description of the non-faculty job titles indicates that they are sufficiently *778 similar, plaintiffs' failure to include a variable for job title does not detract seriously from the probative value of the studies. See Trout v. Hidalgo, 517 F. Supp. 873, 883 (D.D.C.1981); see also Agarwal v. McKee & Co., 16 Employ.Prac.Dec. ¶ 8301 (N.D.Cal. 1977) (plaintiffs' studies established prima facie case despite their failure to include variable measuring job title, although inclusion of such a variable in defendants' regression studies constituted effective rebuttal).^[14]

Defendant's second major criticism of plaintiffs' regression analyses is that they omit certain critical independent variables measuring the faculty members' academic productivity while at CUNY and prior to their CUNY employment. Defendant claims that plaintiffs' studies attribute to sex discrimination salary differentials actually due to factors such as publications, total years of teaching experience, quality of teaching, committee work and community service. With regard to faculty productivity while at CUNY, it has been recognized that variables pertaining to an individual's accomplishments during his employment may incorporate the effects of discriminatory decisions rather than provide an independent measure of his productivity. See James v. Stockham Valves and Fittings Co., 559 F.2d 310 (5th Cir.1977), cert. denied, 434 U.S. 1034, 98 S. Ct. 767, 54 L. Ed. 2d 781 (1978); Meckleberg v. Montana Board of Regents of Higher Education, 13 Employ. Prac.Dec. ¶ 11,438 (D.Mont.1976); Finklestein, The Judicial Reception of Multiple Regression Studies in Sex and Race Discrimination Cases, 80 Colum.L.Rev. 737 (1980) ("Finklestein"). For example, as a result of discrimination, a woman may be assigned greater teaching responsibilities than a similarly situated man. Her ability to do research, her teaching performance and her time for administrative work all may suffer as a result of the initial discriminatory assignment. As defendant's expert concedes, a regression analysis seeking to determine the existence of sex-based salary discrimination should not include as independent variables productivity factors which may be affected significantly by the employer's actions. Plaintiffs consequently are not required to have included in their regression models variables measuring post-hire publications, teaching quality, and committee or community work.

With regard to the pre-hire productivity variables of publications and years of teaching experience outside of CUNY, plaintiffs argue that the variables already present in their regression studies such as age, highest degree, years between degree and quality of institution granting degree serve as "proxies" for the omitted factors. Moreover, plaintiffs' expert argues that the omitted variables could alter significantly the sex-based salary differential shown by his studies only if the omitted variables show a strong correlation to sex after all of the included variables are held constant.^[15] The court finds this argument convincing and finds that pre-hire years of teaching experience and pre-hire publications are closely related to and therefore adequately *779 accounted for by the variables for age, degrees and years between degrees.^[16]

Finally, defendant claims that plaintiffs' regression analyses are flawed by their failure to include a variable reflecting academic department and thereby to account for differing market conditions characterizing each department. Defendant contends that certain salary differences, for example, may be attributable to the fact that there is a greater demand for teachers of computer sciences than there is for French teachers, and that women tend to be concentrated in fields for which there is less demand. Defendant's expert, however, testified that he omitted any measurement of academic department from his own studies because inclusion of such factors would not yield a statistically significant improvement in his model due to the large number of departments. Title VII regression studies, moreover, need not account for every factor that conceivably might explain differences in salaries or promotions. See Valentino v. United States Postal Service, supra, 674 F.2d at 71 n. 24; Segar v. Civiletti, 508 F. Supp. 690, 696 n. 2 (D.D.C.1981); Finkelstein, supra, 80 Colum.L.Rev. at 744-45. Consequently, the omission of a variable for academic department does not negate the probative value of the studies.^[17]

Defendant's next series of objections concern the ISP tapes which provided all of the data for plaintiffs' studies. Defendant contends that errors and omissions in the ISP tapes render plaintiffs' studies meaningless.

Defendant argues first that the information contained on the ISP tapes was neither verified nor updated. The record indicates that plaintiffs checked the tape for internal inconsistencies and other obvious errors. In addition, plaintiffs made an initial effort to examine the CUNY personnel files and to obtain data therefrom. Plaintiffs' expert, however, determined that the information obtained from the personnel files reflected more omissions and internal errors than the ISP tapes. The only information source reasonably available to the plaintiffs for verification of the ISP tapes therefore was not useful for that purpose. The Second Circuit has noted that proof in discrimination cases is frequently based upon "data gathered by less than optimal means" and has refused to hold relatively minor limitations on plaintiffs' ability to gather data as fatal to the presentation of a prima facie case. See Guardians Association of the New York City Police Department, Inc. v. Civil Service Commission of the City of New York, 633 F.2d 232, 240 & n. 13 (2d Cir.1980), cert. granted on other grounds, 454 U.S. 1140, 102 S. Ct. 997, 71 L. Ed. 2d 291 (1982); Jones v. New York City Human Resources Administration, 528 F.2d 696 (2d Cir.), cert. denied, 429 U.S. 825, 97 S. Ct. 80, 50 L. Ed. 2d 88 (1976). Plaintiffs' failure to verify the ISP data consequently does not undermine the study based on those data.

Defendant's second argument with respect to the data base is that the ISP tapes *780 for the semesters from fall 1972 to fall 1974 were missing data needed for plaintiffs' regression analysis with regard to 20% of instructional staff members listed therein. There is no indication, however, that these omissions were other than random or that they imported any bias into plaintiffs' studies. In the absence of such evidence, these omissions will not defeat plaintiffs' prima facie presentation. See Vuyanich v. Republic National Bank, supra, 505 F. Supp. at 306.

Defendant's final objection to plaintiffs' data base is that the spring 1975 to fall 1977 ISP tapes omitted degree and other factors used as independent variables in plaintiffs' studies. Plaintiffs' expert concedes that these data were missing and that his regression studies consequently did not include data for individuals listed for the first time as members of the CUNY instructional staff on the spring 1975 through fall 1977 ISP tapes. These omissions, however, do not preclude reasonable inferences from being drawn with regard to the 1975-1977 hires.

First, plaintiffs' expert testified that the gross statistical differences in salary between men and women on the 1972-1974 ISP tapes also characterized those staff members appearing for the first time on the 1975-1977 tapes. Although these generalizations without more would not, for the reasons discussed at p. 776, supra, suffice to establish a prima facie case with regard to the 1975-1977 hires, they do serve as evidence of the persistence of the salary patterns characterizing the staff on the 1972-1974 tapes. The record, moreover, reflects that CUNY's procedures for determining an instructional staff member's salary level remained essentially the same throughout this entire period. Thus, there is a substantial basis in the record for the inference that the salary patterns reflected in plaintiffs' studies of the 1972-1974 data continued with respect to the 1975-1977 hires. See Hazelwood School District v. United States, 433 U.S. 299, 309 n. 15, 97 S. Ct. 2736, 2742 n. 15, 53 L. Ed. 2d 768 (1977) (proof of pre-Title VII discrimination may support inference of continuance of discrimination "particularly where relevant aspects of the decisionmaking process had undergone little change").^[18]

Finally, defendant contends that plaintiffs' studies are hopelessly flawed because they reflect the effect of employment decisions made prior to March 24, 1972, the date when Title VII became applicable to CUNY. Of course this argument has no bearing upon the validity of plaintiffs' analyses of post-72 hires which were designed explicitly to avoid this problem. With regard to the analyses of all instructional staff, it is true that a Title VII plaintiff must prove discriminatory conduct by the employer during the period when Title VII is applicable to that employer. See United Airlines v. Evans, 431 U.S. 553, 558, 97 S. Ct. 1885, 1889, 52 L. Ed. 2d 571 (1977). Where, as here, plaintiff has alleged post-Act discriminatory employment decisions, however, evidence of pre-Act discrimination may be relevant to prove that such pre-Act practices constituted a pattern of conduct which continued during the actionable period. See Hazelwood School District v. United States, supra, 433 U.S. at 309 n. 15, 97 S. Ct. at 2742 n. 15; Weise v. Syracuse University, 522 F.2d 397, 411 n. 24 (2d Cir.1975); Sobel v. Yeshiva University, 477 F. Supp. 1161, 1165 (S.D.N.Y.1979). Hence, the broader studies are relevant evidence that a pattern and practice of salary discrimination may have existed before the *781 applicable date of Title VII and continued after that time.^[19]

The probative force of plaintiffs' statistical evidence is not materially diminished by the objections raised by defendant. Plaintiffs have produced statistically significant evidence that women hired as CUNY instructional staff since 1972 receive substantially lower salaries than similarly qualified men. This evidence of post-Title VII discrimination is further buttressed by studies reflecting similar disparities when pre-Title VII salary decisions made according to the same procedures are also taken into account. The court finds that plaintiffs have established a prima facie case of sex-based salary discrimination.

In Teamsters v. United States, 431 U.S. 324, 360, 97 S. Ct. 1843, 1867, 52 L. Ed. 2d 396 (1977), the Supreme Court described as follows defendants' burden in a class-based Title VII action once plaintiffs have established a prima facie case:

The burden then shifts to the employer to defeat the prima facie showing of a pattern or practice by demonstrating that the [plaintiffs'] proof is either inaccurate or insignificant. An employer might show, for example, that the claimed discriminatory pattern is a product of pre-Act hiring rather than unlawful post-Act discrimination, or that during the period it is alleged to have pursued a discriminatory policy it made too few employment decisions to justify the inference that it had engaged in a regular practice of discrimination.

The Court added in a footnote that pattern and practice "suits have ... commonly involved proof of the expected result of a regularly followed discriminatory policy. In such cases, the employer's burden is to provide a nondiscriminatory explanation for the apparently discriminatory result." Id. at 360 n. 46, 97 S. Ct. at 1867 n. 46. Defendant's burden is simply to produce such an explanation and not to convince the trier of fact of its validity. The burden of persuasion remains with the plaintiff at all times. See Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 254, 101 S. Ct. 1089, 1094, 67 L. Ed. 2d 207 (1981); Sainte Marie v. Eastern Railroad Association, 650 F.2d 395, 399 (2d Cir.1981). Accordingly, defendant must introduce evidence showing that plaintiffs' studies are either "inaccurate or insignificant" or evidence suggesting a non-discriminatory interpretation of the outcome of those studies.^[20]

As discussed above, defendant has not demonstrated plaintiffs' evidence to be either inaccurate or insignificant. The court therefore must look to defendant's own evidence to determine whether it has articulated a non-discriminatory explanation of the salary disparities plaintiffs have shown.

Defendant's evidence with regard to its employment practices consisted of statistical analyses of instructional staff salaries *782 and of the testimony of a number of university officials. The statistical studies were conducted by Dr. Edgar F. Borgatta, Professor of Sociology and Psychology at the Graduate Center of CUNY. With one exception, defendant's studies were based upon a sample drawn from the fall 1978 ISP tapes and included the following instructional staff members: all men and women in the rank of instructor; all women in the ranks of assistant professor, associate professor, and professor; and male faculty selected at random to provide a number of assistant professors, associate professors and professors approximately equal to the number of females in each of those ranks. Hence, the sample included neither lecturers nor any non-faculty staff. Defendant sent questionnaires to the individuals in the sample to obtain information beyond that available in the ISP tapes or the personnel files.

Defendant then divided the sample into those individuals hired before 1972 and those hired between 1972 and 1978. With regard to each of the pre-72 hires, defendant calculated the total increase in salary for the period between 1971 to 1978. Defendant used these salary increases rather than the annual salary data for the pre-72 hires in order to factor out the effect of the initial salary decisions which were made before Title VII became applicable to CUNY. The mean salary increase for the period from 1971 to 1978 for female pre-72 hires was approximately $200 more than that of male pre-72 hires, but defendant's expert concluded that this difference was statistically insignificant. Defendant then sought to examine by means of regression analysis the effect, if any, of gender upon the increase in salary from 1971 to 1978 for the pre-72 hires when productivity-related independent variables were held constant. The independent variables in these regression analyses were years of CUNY service, highest degree at time of hire, pre-hire college teaching experience, pre-hire publications and years since highest degree. In these analyses, defendant found no statistically significant differences between the 1971-78 salary increases received by male pre-72 hires and those received by females in that group.

Defendant then conducted a series of studies concerning those members of its sample who were hired between 1972 and 1978. With regard to these individuals, defendant studied annual salary rather than salary increases, since all salary decisions for such individuals were made after Title VII became applicable to CUNY. Defendant, through regression analysis, measured the difference in 1980 salary between men and women when years of CUNY service, degree at hire, pre-hire publications and pre-hire teaching experience are held constant. The average salary difference thus determined was characterized by defendant's expert as statistically insignificant.

Finally, defendant conducted one study that was not based upon the above-described sample from the ISP tapes. This study compared the average salary increase from 1971 to 1978 for men and women hired before 1972 who were lecturers or non-faculty members. The study did not, however, attempt to factor out the effect of any productivity related variables.

Defendant's statistical presentation sheds no light on defendant's salary policies with regard to lecturers and non-faculty instructional staff members. Defendant did not perform regression analysis with regard to these employees and therefore offered no evidence as to the salary differences between similarly qualified men and women in those ranks. As discussed at p. 777, supra, courts generally have held gross averages such as those calculated by defendant for this group to be without significant probative value. Defendant therefore failed to present any statistical data sufficient to provide a non-discriminatory explanation for plaintiffs' prima facie case of salary discrimination with regard to lecturers and non-faculty staff.

The court also finds that defendant's statistical presentation with regard to the other faculty members is fatally flawed. Defendant's *783 expert himself described the studies as "faulted" because of the manner in which the sample was chosen. By choosing a sample in which the proportion of men in the four faculty ranks was approximately equal to the proportion of women faculty in those ranks, defendant's expert conceded that he was controlling for rank.^[21] It follows that defendant's study could serve to demonstrate, at most, the absence of any salary discrimination between men and women in the same rank. As all parties acknowledge, instructional staff salary is highly correlated with rank. Thus, if men have been disproportionately assigned to higher ranks and women to lower ranks, as plaintiffs' logit analysis strongly suggests, there will be a disparity between male and female salaries due to a disproportionate and possibly discriminatory assignment of men and women to ranks. Defendant's study, by considering only equal numbers of men and women in each faculty rank, will mask any such disparity in salaries.

In instances in which assignment to rank is controlled by the employer and arguably is tainted by discrimination, courts have discredited studies which purported to measure salary discrimination by holding rank constant. See James v. Stockholm Valves, supra, 559 F.2d at 332; Mecklenberg v. Montana Board of Regents, supra, 13 Employ.Prac.Dec. at 6496; cf. Presseisen v. Swarthmore College, supra, 442 F. Supp. at 619 (requiring regression analysis to include rank level in view of proof that assignment to rank is not discriminatory); see also Finklestein, supra, 80 Colum.L.Rev. at 740-42, 748-49. The court therefore finds that defendant's studies provide no useful evidence as to whether women faculty are systematically paid less than men.

Similarly, the CUNY administrators who testified regarding university salary practices offered no explanation of the salary discrepancy between men and women shown in plaintiffs' studies. Although these administrators and defendant's expert speculated that women devote more time to child-rearing, have fewer publications, and are concentrated in academic fields for which the market demand is not great, they did not testify that such generalizations characterize the women on the CUNY instructional staff or explain the salary disparity measured by plaintiffs. Such speculations are insufficient to rebut plaintiffs' prima facie showing of sex-based salary discrimination.

Defendant therefore has failed to demonstrate that plaintiffs' statistical evidence is either "inaccurate or insignificant" or to provide "a nondiscriminatory explanation for the apparently discriminatory result." Teamsters v. United States, supra, 431 U.S. at 360 & n. 46, 97 S. Ct. at 1867 & n. 46. Accordingly, defendant has failed to rebut plaintiffs' prima facie case.

The court finds that defendant has discriminated against plaintiffs in the payment of salaries in violation of Title VII. The parties are given 30 days in which to submit proposed orders with regard to proceedings to determine the amount of plaintiffs' damages on the claim of salary discrimination.

So Ordered.

[1] The Board of Trustees of the City of New York has been substituted as defendant for its predecessor entity, the Board of Higher Education of the City of New York.

[2] CUNY is composed of eight senior colleges, nine community colleges, a graduate division, and the Mount Sinai Medical School. The employment practices of the medical school are not at issue in this action.

[3] The instructional staff does not include clerical or custodial workers.

[4] The HEO job titles include individuals having various administrative functions such as personnel management or campus planning.

[5] Although Teamsters v. United States, supra, was a "pattern or practice" suit brought by the government pursuant to § 707(a) of the Civil Rights Act of 1964, 42 U.S.C. § 2000e-6(a), the standard set forth therein is equally applicable to private class-based actions alleging widespread discrimination. See Presseisen v. Swarthmore College, 442 F. Supp. 593, 598-99 (E.D.Pa.1977), aff'd 582 F.2d 1275 (3d Cir. 1978).

[6] Claims of sex-based wage discrimination under Title VII, unlike such claims brought under the Equal Pay Act, 29 U.S.C. § 206(d), are not limited to claims of unequal pay for equal work. Equally cognizable under Title VII are claims that wages are "depressed because of intentional sex discrimination." County of Washington v. Gunther, 452 U.S. 161, 166, 101 S. Ct. 2242, 2246, 68 L. Ed. 2d 751 (1981).

[7] The only evidence introduced by plaintiffs in addition to Dr. Killingsworth's studies and testimony was Dr. Marilyn Gittel's description of her work on the Chancellor's Advisory Committee on the Status of Women at the City University of New York and a copy of the report produced by that committee in 1971. Plaintiffs did not call any class members as witnesses to testify as to their own experience regarding salary at CUNY.

[8] Plaintiffs divided the post-72 hires from the other instructional staff members in order to isolate those salary decisions made after Title VII became applicable to defendant on March 24, 1972.

[9] Male employees whose highest degree is a B.A. receive, on the average, about $20,180 annually, while the average salary of female B.A.'s is $17,665. Average male salary also exceeds average female salary in the five different age groups, in the nine different years-of-CUNY-service groups and in the nine years-since-highest-degree groups.

[10] Two sets of "quality of degree" variables were used with regard to the degree or degrees held by the instructional staff member at time of hire. The first set of variables indicate whether the institution awarding an M.A. or Ph.D. was included in K. Roose and C. Anderson, A Rating of Graduate Programs, American Council on Education (1973). The second set indicate the quality level accorded the institution awarding the B.A., M.A. or Ph.D. by Martin Trow et al. in Technical Report: Carnegie Commission Survey of Higher Education, Carnegie Commission on Higher Education 97-125 (1972).

[11] As Dr. Killingsworth stated in his report:

The regression measure of the net relation between sex and salary is an average measure (net of the influences of the other factors considered in the analysis) of the systematic relation between sex and salary. As with any other average, there is, of course, some variation around this figure. Accordingly, the "chance probability" ... is a kind of index of the probability that the true average could be zero, given the regression measure of the average, the observed variation in the data, and the number of persons included in the analysis. In this sense, the "chance probability" is somewhat analogous to the "margin of error" used in evaluating the results of public opinion polls and in ascertaining whether any differences in responses as between different choices presented in such polls might be attributable to mere chance factors rather than to a true difference in responses.

[12] Plaintiffs condensed their results in the following table:

                    Summary of Implications of Logit Results
                  Assignment to Faculty Rank at Hire  Females
                (Persons Hired as Faculty Since October 1972 Only)
                                                       If Treated as Male with
Faculty Rank                      Actual                  Same Characteristics  
  to Which
Assigned at Hire        Number            Percent      Number            Percent
Full Professor             8                 1.37       19.66               3.37
Associate
Professor                 22                 3.77       31.09               5.32
Assistant
Professor                154                26.37      159.27              27.27
Instructor               340                58.22      299.46              51.28
Lecturer                  60                10.27       74.52              12.76
     Total               584               100.00      584.00             100.00
                  Summary of Implications of Logit Results
                 Assignment to Faculty Rank at Hire  Males
             (Persons Hired as Faculty Since October 1972 Only)
                                                       If Treated as Female with
Faculty Rank                      Actual                  Same Characteristics  
  to Which
Assigned at Hire        Number            Percent      Number              Percent
Full Professor            43                5.82        18.44                 2.50
Associate
Professor                 78               10.55        62.90                 8.51
Assistant
Professor                284               38.43       289.55                39.18
Instructor               267               36.13       311.33                42.13
Lecturer                  67                9.07        56.78                 7.68
      Total              739              100.00       739.00               100.00

[13] Defendant also objected to the inclusion of lecturers in the same study as other faculty members because a lecturer, under section 11.26 of the By-laws, is not required to pursue academic research. The nature of the lecturer position, however, is not so different from that of the other faculty ranks as to require that it be omitted from plaintiffs' studies.

[14] Under Dr. Killingsworth's reasoning, described at p. 12, supra, wage differences attributable to the most distinctive of the non-faculty job titles would be accounted for by variables already included in the regression, e.g. the differences for the "college physician" position are accounted for by the degree variables in the regression analysis.

[15] As Dr. Killingsworth stated in his study:

Thus, suppose, hypothetically, that female faculty have less [total] work experience, on average, than male faculty. A priori, one might suppose that introduction of an explicit "years of work experience" factor would reduce the "sex effect" derived from our regression results. However, if most of the sex difference in work experience were associated with such sex differences in such factors as age, years of CUNY service, length of time taken between successive degrees completed and the like all of which are explicitly included in our studies then introduction of the "years of work experience" factor would have little or no impact on the measured "sex effect." In other words, the key question is whether there is a sex difference in the particular "omitted variable" of interest that remains after all the included variables actually used in the study have been taken into consideration.

[16] Defendant contends that such variables are particularly inappropriate proxies for work experience for women, when women more frequently than men interrupt their professional work for child-rearing or other family related matters. Such a proposition, if true for the members of plaintiff class, may affect the utility of the proxy variables in plaintiffs' study. Such sex-based stereotypes, however, may not be invoked to defeat plaintiffs' prima facie case, but rather, if supported by evidence relevant to the class at issue, appropriately may form a part of defendant's rebuttal. See Trout v. Hidalgo, supra, 517 F. Supp. at 881; Vuyanich v. Republic National Bank, 505 F. Supp. 224, 306-307 (N.D.Tex.1980), aff'd on reconsideration, 521 F. Supp. 656 (N.D.Tex.1981), appeal docketed No. 81-1357 (5th Cir. Aug. 10, 1981).

[17] The Fifth Circuit in Wilkins v. University of Houston, supra, 654 F.2d at 402, found the failure of the plaintiffs' studies to consider department or college fatal to a prima facie case of sex-based salary discrimination by a university. Plaintiffs in that case, however, used only gross statistical comparisons and not regression analyses accounting for factors such as experience or degrees that might be compensated more highly. In Presseisen v. Swarthmore College, supra, 442 F. Supp. at 615, inclusion of a variable measuring college division was recognized as an effective part of defendant's rebuttal, but was not required as part of plaintiffs' prima facie showing.

[18] Penalizing plaintiffs with regard to the omissions in the 1975-77 ISP tapes seems particularly inappropriate here because defendant in the spring of 1975, despite this pending litigation, without the consent of the CUNY Chancellor and without notice to plaintiffs, simply stopped gathering the relevant instructional staff data. Cf. Valentino v. United States Postal Service, supra, 674 F.2d at 73 n. 31 (where employer has destroyed records, court may draw adverse inference). Defendant contends that the 1975-77 data was available to plaintiffs in the individual CUNY personnel files. As stated above, however, the information gained from those files was too incomplete and fraught with errors to be of use.

[19] In its post-trial memorandum, defendant also argues that the computer print-outs for plaintiffs' studies of all instructional staff members reflect 52.8% "missing cases" and thus an analysis of only 47.2% of that group. Plaintiffs' print-outs, defendant claims, reflect a similar rate of "missing cases" for the studies of post-72 employees. Defendant, however, did not cross examine plaintiffs' expert with regard to the meaning of the term "missing cases" in those documents. Neither did defendant's own expert testify as to the significance of that term or of those figures on plaintiffs' print-outs. In the absence of such testimony, the court will not attempt to guess the meaning of those figures and instead will rely on plaintiffs' expert's testimony with regard to his data base.

[20] The Bennett Amendment, 42 U.S.C. § 2000e-2(h), makes applicable to a Title VII claim of sex-based salary discrimination the affirmative defenses to a case brought pursuant to the Equal Pay Act, 29 U.S.C. § 206(d) (1). See County of Washington v. Gunther, supra, 452 U.S. at 171, 101 S. Ct. at 2249. A Title VII defendant, however, has the burden of pleading and proving those affirmative defenses. See Kouba v. Allstate Insurance Co., 691 F.2d 873 (9th Cir.1982). Defendant here has not raised any of these defenses.

[21] Defendant's expert did not dispute that there existed a substantial disparity between the percentage of males in each faculty rank and the percentage of males in each rank in defendant's sample.