Appellants are past and present members of the faculty of the University of Washington School of Nursing (the nursing faculty). One of them, Ruth Fine, has served as an associate administrator of the University’s hospital, as the director of nursing services, and as an associate professor. The intervenors, also past and present members of the nursing faculty, join in this appeal.

The nursing faculty filed suit in the district court alleging that the University engaged in discriminatory compensation practices in violation of 42 U.S.C. § 1983 , the Equal Pay Act, and Title VII. The district court sua sponte referred the case to a United States Magistrate sitting as a special master. After the nursing faculty presented its case, the special master issued a report recommending dismissal of the action pursuant to rule 41(b) of the Federal Rules of Civil Procedure. The district court granted the motion for an involuntary dismissal. We have jurisdiction under 28 U.S.C. § 1291 .

The nursing faculty argues that the district court erred in not reviewing the special master’s findings de novo, that we must consequently engage in de novo review, and that the district court erred in dismissing the action under rule 41(b). On its substantive claims, the nursing faculty contends that it demonstrated that the University violated section 1983 and the Equal Pay Act, that it made out a prima facie showing of discrimination prohibited by Title VII under both the disparate treatment *692 and disparate impact models, that the University cannot rely on a “competitive marketplace” defense, and that the nursing faculty is entitled to recover attorneys’ fees. We affirm.

I

For purposes of this appeal, we accept the facts agreed to by the parties, the evidence admitted during presentation of the nursing faculty’s case, and the district court’s factual findings. Because the case was dismissed under Federal Rule of Civil Procedure 41(b) after the nursing faculty presented its case, the University did not present its case.

Washington State created the University of Washington by statute. Wash.Rev.Code Ann. §§ 28B.20.010-.20.820 (1982 & Supp. 1983). The University consists of 16 separate schools, each under its own dean. Most schools are divided further into academic departments. The University functions under a very decentralized administrative scheme. Each school is responsible, subject to approval by the University’s president, for faculty appointments, entry level of new appointees and their salary, promotions, and salary increases for individual faculty members. Generally, the University’s president accepts departmental recommendations on these issues. The University’s budget office allocates money to each school after receiving a salary allocation for the University from the state legislature. The budget office normally designates a portion of the funds for across-the-board salary increases. The remainder is then divided among the schools, sometimes in varying percentages, for distribution at the schools’ discretion to their faculty.

In March 1972, members of the faculty of the School of Nursing filed a petition with Dr. Katz, Vice President for Academic Affairs and Provost of the University, alleging sex discrimination by the University. The University responded to the petition and provided certain salary data. Dr. Grayson, Vice President of the Health Sciences Center, and Dr. Katz met with representatives of the nursing faculty to discuss the petition. The University asserted that salary levels varied because each academic discipline commanded a salary based upon training, expertise, emphasis, subject matter, and the academic marketplace for that discipline. Thus, the University argued that it was inappropriate to compare the average salary in one discipline with the composite average salary paid University faculty members. Subsequently, the University undertook, and made available to the nursing faculty, a more sophisticated study of salaries in the School of Nursing. The study concluded that faculty salaries as a whole at the University of Washington lagged 9 percent behind the salaries paid by schools with which the University had traditionally compared its salaries. The study also concluded that average faculty salaries in the School of Nursing lagged 10.9 percent behind salaries paid by comparable schools of nursing, but that many other disciplines also lagged at least that much behind.

In September 1972, the University’s budget office allocated a 3 percent merit increase to each school. In response to the salary study findings, the budget office allocated an additional 2 percent increase to the School of Nursing in order to align the nursing faculty salaries with the average deficit in salaries at the University. Between 1973 and 1978, the University’s budget office gave an above average allocation to the School of Nursing in two budget years and an allocation equal to that given other schools in three budget years.

The nursing faculty remained dissatisfied. Throughout the summer and fall of 1972, members of the nursing faculty held meetings to discuss their petition and related issues. In November 1972, a group entitled the Women’s Salary Inequity Committee (the Committee) sought support from faculty members and sent a complaint letter to the Office for Civil Rights of the United States Department of Health, Education, and Welfare. The Committee also filed a complaint with the Washington *693 State Human Rights Commission. Thereafter, the Committee filed charges with the Equal Employment Opportunity Commission (EEOC) and the University’s Human Rights Commission.

In February 1974, the United States Department of Justice issued a right to sue letter to Spaulding as head of the Committee and the nursing faculty filed this action. Although this suit was originally filed as a class action against officials at the University of Washington, the University was later substituted as the sole defendant and the class claims were dropped. The intervenors, after objecting to the dismissal of the class claims, were granted permission to intervene and their claims were stayed pending resolution of the named plaintiffs’ claims.

In August 1977, after finding that he could not schedule the case for trial within 120 days after issue was joined, the district judge sua sponte referred the case to a United States Magistrate pursuant to 42 U.S.C. § 2000e-5(f)(5), rule 53 of the Federal Rules of Civil Procedure, and Local Magistrates’ Rule 21. The order of reference expressly stated that the magistrate was to sit as a special master, hear the case on the merits, and report recommended findings of fact, conclusions of law, and disposition. The order also stated that the special master’s report would be subject to review by the district court in accordance with rule 53(e) of the Federal Rules of Civil Procedure.

After the nursing faculty presented its case before the special master, the special master stated that he planned to grant the University’s rule 41(b) motion for involuntary dismissal. He concluded that the nursing faculty had failed to show that they performed substantially equal work compared to male faculty members in other departments and that the Equal Pay Act standard governed claims under Title VII and section 1983.

While the special master was preparing his findings and conclusions, we decided Gunther v. County of Washington, 623 F.2d 1303 (9th Cir.1979), aff'd, 452 U.S. 161 , 101 S.Ct. 2242 , 68 L.Ed.2d 751 (1981) (Gunther). We held that a plaintiff who fails to show that he performs substantially equal work is not precluded from suing under Title VII for relief from intentionally discriminatory compensation practices unless such practices are authorized under one of the four Equal Pay Act affirmative defenses. 623 F.2d at 1310-13 . The special master requested memoranda from the parties on the effect of our decision in Gunther on his 41(b) ruling. He then concluded that Gunther did not alter the outcome of the case and recommended that the district court dismiss the case.

The nursing faculty requested that the district court direct the special master to file a transcript of the proceedings before him with the district court. The district court denied this request, but certified the issue for interlocutory appeal under 28 U.S.C. § 1292 (b). We granted permission to appeal and held that the district court was required to provide a transcript. Spaulding v. University of Washington, 676 F.2d 1232, 1235 (9th Cir.1982) (Spaulding I). We expressly declined to decide whether the district court was required to review the special master’s findings under a de novo standard or a clearly erroneous standard. 676 F.2d at 1234 n. 1.

Prior to our decision in Spaulding I, the transcript was prepared and used by the district judge as he reviewed the special master’s factual findings under the clearly erroneous standard 1 and adopted both the findings of fact and conclusions of law with minor modifications. The district judge then ordered dismissal under rule 41(b).

II

The district judge did not decide whether he had jurisdiction over the nursing facul *694 ty’s claims, but went directly to the merits. This was incorrect.

Federal courts are courts of limited jurisdiction and may not resolve the merits of disputes if they lack jurisdiction. See 18 C. Wright, A. Miller & E. Cooper, Federal Practice & Procedure § 3522 (1975). Although the district court failed to address the issue of its jurisdiction, we may affirm its decision on any basis supported by the record. E.g., Jaffke v. Dunham, 352 U.S. 280, 281 , 77 S.Ct. 307, 308 , 1 L.Ed.2d 314 (1957) (per curiam).

We first consider the district court’s jurisdiction over the section 1983 claim. The eleventh amendment to the Constitution bars suit in federal court by citizens against a state or its agency under section 1983 unless the state has waived its immunity. Quern v. Jordan, 440 U.S. 332, 338-45 , 99 S.Ct. 1139, 1143-47 , 59 L.Ed.2d 358 (1979); Alabama v. Pugh, 438 U.S. 781 , 98 S.Ct. 3057 , 57 L.Ed.2d 1114 (1978) (per curiam); Edelman v. Jordan, 415 U.S. 651, 662-63 , 94 S.Ct. 1347, 1355-56 , 39 L.Ed.2d 662 (1974); Bennett v. California, 406 F.2d 36, 39 (9th Cir.), cert. denied, 394 U.S. 966 , 89 S.Ct. 1320 , 22 L.Ed.2d 568 (1969); see Wolcher, Sovereign Immunity and the Supremacy Clause: Damages Against States in Their Own Courts for Constitutional Violations, 69 Calif.L.Rev. 189, 200-34 (1981).

The nursing faculty conceded in its amended complaint and in the pretrial order that the University is an agency of the State of Washington. Thus, we need not determine independently whether the University is a state agency. The nursing faculty does not assert that Washington waived its immunity. We therefore hold that the district court lacked jurisdiction over the nursing faculty’s section 1983 claim.

The district court did, however, have federal question jurisdiction over the Equal Pay Act claims. The University comes within the Act because it is both an employer and an establishment within the meaning of the statute. 29 U.S.C. §§ 203 (d), 206(d)(1). 2

Ill

The nursing faculty asserts that the district court erred in not reviewing the special master’s factual findings de novo. It bases this assertion on its interpretation of the applicable statutes, as well as on article III of the Constitution. Further, it argues that we must engage in de novo review of the facts because the district judge failed to do so. The nursing faculty also contends that the reference to the special master was improper because 42 U.S.C. § 2000e-5(f)(5) does not authorize referrals of Equal Pay Act claims. 3

We first consider whether the nursing faculty waived any of these objections. In Spaulding I, we held that the nursing faculty had not waived its right to have the district judge review a transcript of the proceedings before the special master. 676 F.2d at 1235 . The nursing faculty raised its objection to the use of the clearly erroneous standard of review at the same time that it raised its objection to the transcript provision. Thus, we must hold that the nursing faculty did not waive its right to contest the use of the clearly erroneous standard of review.

*695 Spaulding I, however, does not affect our determination of whether the nursing faculty waived its objection to referral of the Equal Pay Act claims. The nursing faculty waited until this appeal to raise the argument that Equal Pay Act claims cannot be referred to a special master under 42 U.S.C. § 2000e-5(f)(5). They are too late. Parties should object to a reference to a magistrate or a special master at the time the reference is made or within a reasonable time thereafter. See, e.g., Hayes v. Foodmaker, Inc., 634 F.2d 802, 803 (5th Cir.1981) (per curiam); Cruz v. Hauck, 515 F.2d 322, 331 (5th Cir.1975), cert. denied, 424 U.S. 917 , 96 S.Ct. 1118 , 47 L.Ed.2d 322 (1976); Diamond Door Co. v. Lane-Stanton Lumber Co., 505 F.2d 1199 , 1206 (9th Cir.1974); 5A J. Moore & J. Lucas, Moore’s Federal Practice 153.05[3] (2d ed. 1982); 9 C. Wright & A. Miller, Federal Practice & Procedure § 2606 (1971).

We therefore turn to the merits of the nursing faculty’s claim that the district court erred in not reviewing the special master’s factual findings de novo. The district judge referred sua sponte the nursing faculty’s claims to the special master under the authority of 42 U.S.C. § 2000e-5(f)(5) 4 and pursuant to rule 53 of the Federal Rules of Civil Procedure and rule 21 of the local magistrates’ rules. 5 The order of reference provided that the magistrate would sit as a special master and that the master’s report would be subject to review by the district court in accord with rule 53(e) of the Federal Rules of Civil Procedure. Rule 53(e) states that in a non-

jury action, the district court shall accept the master’s findings of fact unless clearly erroneous. Fed.R.Civ.P. 53(e)(2). After the master filed his report, the nursing faculty objected to the use of the clearly erroneous standard of review. The district judge rejected the nursing faculty’s claim that de novo review was required, relying upon White v. General Services Administration, 652 F.2d 913 (9th Cir.1981) (White).

In White , we held that a reference to a special master pursuant to the same local magistrates’ rule now before us was proper under 42 U.S.C. § 2000e-5(f)(5). We also stated that:

White, 652 F.2d at 915 (citations omitted). By incorporating all of rule 53 except for part (b), White teaches that 53(e)(2) is applicable. That subdivision states in part “the court shall accept the master’s findings of fact unless clearly erroneous.” Thus, we agree with other courts which have held expressly that the clearly erroneous stan *696 dard applies to review of a magistrate’s or special master’s factual findings in a Title VII ease. Cockrham v. South Central Bell Telephone Co., 695 F.2d 143, 145 (5th Cir.1983) (per curiam); Hayes v. Foodmaker, Inc., 634 F.2d at 802-03 (reference under 28 U.S.C. § 636 (b)(2)); Livas v. Teledyne Movible Offshore, Inc., 607 F.2d 118, 119 (5th Cir.1979) (per curiam); Harden v. Dayton Human Rehabilitation Center, 520 F.Supp. 769, 770-71 (S.D.Ohio 1981) (reference under 28 U.S.C. § 636 (b)(2)); see also Oliver v. Allison, 488 F.Supp. 885, 888-89 (D.D.C.1980) (mem.) (consensual reference to magistrate).

The nursing faculty contends that de novo review is required in order to alleviate an “irreconcilable tension” between cases referred under sections of the Federal Magistrates Act requiring de novo review and cases referred pursuant to rule 53, but they cite only cases in which reference was under 28 U.S.C. § 636 (b)(1) or (b)(3). Subsection 636(b)(1) expressly states that a district judge is required to make a de novo determination of those portions of a magistrate’s report to which a party objects. We require the same in references under subsection 636(b)(3). Coolidge v. The Schooner California, 637 F.2d 1321 , 1325-27 & 1326 n. 5 (9th Cir.), cert. denied, 451 U.S. 1020 , 101 S.Ct. 3011 , 69 L.Ed.2d 392 (1981) (consensual reference). But these cases differ from those referred under 28 U.S.C. § 636 (b)(2). Further, the nursing faculty’s argument that 28 U.S.C. § 636 (c), permitting a magistrate to hear and enter judgment in a civil matter upon consent of the parties, is implicit authority that de novo review is required unless the parties consent, lacks merit.

We hold that we should not undertake our own de novo review of the special master’s factual findings. See Fed.R. Civ.P. 52(a); accord Pullman-Standard v. Swint, 456 U.S. 273, 285-90 , 102 S.Ct. 1781, 1788-91 , 72 L.Ed.2d 66 (1982); cf. Rohde v. K.O. Steel Castings, Inc., 649 F.2d 317, 319 (5th Cir.1981); Coen v. Zick, 458 F.2d 326, 328 (9th Cir.1972) (findings of referee in bankruptcy which have been affirmed by district court not to be set aside unless clearly erroneous).

IV

On the merits of the Equal Pay Act claim, the district court held that the nursing faculty had failed to establish that its members performed work substantially equal to that performed by male faculty members in other parts of the University. The nursing faculty attempted to establish this claim by comparing the job of nursing faculty members to that of male faculty members in other schools of the University such as health services, social work, architecture, urban planning, environmental health, speech and hearing, rehabilitative medicine, and pharmacy practice. The nursing faculty does not assert that male and female members of the nursing faculty are treated differently. Indeed, one plaintiff and one intervenor are male members of the nursing faculty. The district court also held that Fine failed to establish that she performed work substantially equal to that performed by either of the two male associate hospital administrators she identified as comparators.

The Equal Pay Act prohibits an employer subject to its provisions from discriminating in the payment of wages between employees on the basis of sex. In order to make out a prima facie ease of an Equal Pay Act violation, the nursing faculty bears the burden of establishing that its members did not receive equal pay for equal work. Hein v. Oregon College of Education, 718 F.2d 910, 913 (9th Cir.1983) (Hein); Gunther, 623 F.2d at 1309 ; 6 see Corning Glass Works v. Brennan, 417 U.S. 188, 195 , 94 S.Ct. 2223, 2228 , 41 L.Ed.2d 1 (1974). If a prima facie case is established, the University may attempt to show that the payment of different wages is based on either a seniority system, a merit system, a system which measures *697 earnings by quantity or quality of production, or a factor other than sex. 29 U.S.C. § 206 (d)(l)(i)-(iv); Corning, 417 U.S. at 196 , 94 S.Ct. at 2229 ; Kouba v. Allstate Insurance Co., 691 F.2d 873, 875 (9th Cir. 1982); Gunther, 623 F.2d at 1308-09 .

The Act defines, at least in broad terms, what constitutes equal work by specifying that jobs are equal if their performance requires equal skill, effort, and responsibility and they are performed under similar working conditions. 29 U.S.C. § 206 (d)(1). A plaintiff may show that the jobs are substantially equal, not necessarily that they are identical. Gunther, 623 F.2d at 1309 . Actual job performance and content, rather than job descriptions, titles or classifications, is determinative. Id. Thus, each claim that jobs are substantially equal necessarily must be determined on a case-by-case basis. Hein, 718 F.2d at 913 ; Gunther, 623 F.2d at 1309 ; Usery v. Columbia University, 568 F.2d 953, 958 (2d Cir.1977). We disagree, therefore, with the University’s contention that jobs from different academic disciplines can never be substantially equal. The proper resolution of each case necessarily depends upon the actual content of the jobs. Gunther, 623 F.2d at 1309 . The district judge’s determinations on the issue of substantial equality are findings of fact. Hein, 718 F.2d at 913 ; Gunther, 623 F.2d at 1309 ; accord Horner v. Mary Institute, 613 F.2d 706, 713 (8th Cir.1980). We apply the clearly erroneous standard of review.

The nursing faculty argues that it need only demonstrate that the objective characteristics of the jobs are substantially equal. Then, according to the nursing faculty’s argument, the burden shifts to the University to show whether any subjective differences between the jobs, such as the subject taught, are legitimate and actual bases for the disparity. The cases relied upon by the nursing faculty are not cases under the Equal Pay Act, but, rather, are cases arising under Title VII and involve the allocation of burdens in a disparate treatment claim. See, e.g., Lynn v. Regents of the University of California, 656 F.2d 1337 (9th Cir.1981), cert. denied, 459 U.S. 823 , 103 S.Ct. 53 , 74 L.Ed.2d 59 (1982) (Lynn).

In attempting to meet its burden, the nursing faculty contends that nursing faculty members perform substantially equal work to that performed by specified comparator faculty members because both jobs require preparation and teaching of courses, research and publication, committee work, advising of students, and community service. The nursing faculty selected the departments relied upon for comparators because they are professional schools “with a degree mix similar to the school of nursing with practice related activities, where students are frequently put into clinical settings under faculty supervision.” The nursing faculty introduced statistics attempting to compare 66 individual faculty members from the selected comparator departments with members of the nursing faculty based on degrees held, experience, and merit. In addition, the nursing faculty points to evidence describing various faculty positions and showing that its members taught courses in comparator departments, engaged in research on interdisciplinary issues, and worked on committees with comparator faculty members. Despite the facial similarity in the nature of the faculty positions relied upon by the nursing faculty, the special master concluded that the nursing faculty had not shown substantial equality. The district judge adopted that finding. We conclude that the district judge’s finding on substantial equality is not clearly erroneous. See generally Horner v. Mary Institute, 613 F.2d at 714 (affirming finding that jobs of two physical education teachers were not substantially equal even though “superficially identical in that both involve teaching of physical education”).

The evidence supports the finding of the special master and the district judge on substantial equality. It indicates that the different departments in the University which plaintiffs used for comparators placed varying degrees of emphasis on research, training, and community service, and that nursing historically had been con *698 sidered a discipline distinct from those plaintiffs chose for comparison. Further, the statistical evidence submitted by the nursing faculty was deficient in several respects. It did not adequately account for prior job experience, rank, or multiple degrees, and, most important, it did not adequately evaluate the actual work performed by various faculty members. We agree with the district court’s observation that the nursing faculty’s statistical evidence “either ignores the central fact disputed in this lawsuit, which is whether or not the work done by plaintiffs is substantially equal to the work done by male faculty with whom they compare themselves, or it presumes equality.” The statistical evidence may demonstrate a pay disparity, but a difference in pay between jobs which women primarily hold and jobs which men primarily hold does not state a prima facie Equal Pay Act case if the jobs are not substantially equal. See Horner v. Mary Institute, 613 F.2d at 715 . The district court was not clearly erroneous in finding that the nursing faculty had not shown that the actual day-to-day responsibilities, skill, and effort required of the male comparators were substantially equal to that of the nursing faculty.

In reaching this decision, the district court adopted the special master’s findings concerning the differences in training and education in academic fields. The nursing faculty argues that a difference in training and educational background is irrelevant to whether jobs are substantially equal. In brief, the nursing faculty contends that “teaching is teaching.” The contention, although superficially appealing, lacks merit. As we stated previously, the determinative factor is actual job content. The Equal Pay Act specifies that equal skill is a component of equal work. 29 U.S.C. § 206 (d)(1). The administrative regulations interpreting the Equal Pay Act define skill as including “consideration of such factors as experience, training, education, and ability.” 29 C.F.R. § 800.125 (1983). Possession of skills irrelevant to the job requirements, however, cannot be considered in making a determination regarding substantial equality of skill. Id.; see Peltier v. City of Fargo, 533 F.2d 374, 377 (8th Cir.1976). Clearly, training in an academic field is necessary to a job as a university faculty member in that field. See Melanson v. Rantoul, 536 F.Supp. 271, 287 (D.R.I.1982) (refusing to compare plaintiff’s salary to average salary of male associate professors absent showing that each associate professor’s position requires equal skill). Some members of the nursing faculty had engaged in interdisciplinary teaching, research, or community activities, but the district judge did not err in adopting the master’s finding that these were not job required although some individuals brought their unique skills to the work. At best, those activities were only a part of their duties and, thus, do not convince us that the district judge was clearly erroneous on the substantial equality issue.

The evidence on Fine’s Equal Pay Act claim also shows that she did not prove substantially equal work. Fine was employed by the University of Washington Hospital from 1958 to 1976. She was director of nursing services from 1962 to 1976. From 1962 to 1968, she served as a clinical assistant professor in the School of Nursing. From 1968 to 1973, she was a part-time nontenured assistant professor in the School of Nursing and beginning in 1973, she was a part-time tenured associate professor. From 1969 to 1976, she also had the title of associate administrator. Since 1976, her exclusive position has been as an associate professor. She holds a master’s degree in nursing and has some post-graduate training in business administration. She argues that she performed substantially equal work to that performed by Lang and Stein, two male associate hospital administrators.

During Fine’s tenure at the hospital, she was responsible for supervising nursing personnel and for the respiratory therapy, central supply, medical records, and in-patient admitting departments. She testified that she reported directly to the hospital *699 administrator, that in her capacity as associate administrator she was responsible for department development, program changes, and planning, and that she was responsible for the nursing department budget. She also testified that she chaired a committee which planned a computer program system for University Hospital and Harborview Medical Center (Harborview), was in charge of disaster plans, and served on various other committees. Harborview was also associated with the University. In addition, she at times had administrative responsibility for the entire hospital at night, on weekends, and on an on-call basis.

Lang holds a master’s degree in public health and served as an assistant hospital administrator at University Hospital and then as an associate hospital administrator at Harborview. He was responsible for numerous hospital departments, in charge of labor negotiation, program development, and had a major role in hospital planning and administration. The special master concluded that Lang was responsible for administering more departments than Fine and that those departments were more complex. In the absence of the hospital administrator, Lang assumed complete responsibility.

Stein holds a master’s degree in business administration and served as an associate hospital administrator from January 1974 to March 1978 at the University of Washington Hospital. He then served as acting hospital administrator for a time. As associate administrator, he was second in command. He had primary administrative responsibility for all major professional departments, represented the administration in labor negotiations, and coordinated hospital planning and budgeting.

The district judge was not clearly erroneous in finding that the jobs of Fine and Lang or Stein were not substantially equal.

V

The nursing faculty also claims sex-based wage discrimination by the University in violation of Title VII of the Civil Rights Act of 1964. Section 703(a) of Title VII makes it unlawful for an employer “to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment” or to “limit, segregate, or classify his employees or applicants” because of sex. 42 U.S.C. § 2000e-2(a).

The decision rendered at the end of the nursing faculty’s case indicates a determination by the district judge that no prima facie ease has been established. Evidence which is relevant to the issue of a prima facie ease may also be relevant to an employer’s defense of good faith. Thus, these are findings which could be applicable to either of those issues or to the issue of pretext. Essentially, however, the issue is one of whether a prima facie case was established. 7 This was our approach in Moore v. Hughes Helicopters, Inc., 708 F.2d 475 (9th Cir.1983) (Moore).

Our past cases have not been consistent on whether our review of the lower court’s ultimate finding regarding establishment of a prima facie case is de novo or under a clearly erroneous standard. See generally Moore, 708 F.2d at 480 ; Gay v. Waiters’ and Dairy Lunchmen’s Union, Local No. 30, 694 F.2d 531, 539-46 (9th Cir.1982). We need not decide which standard is correct because in this case, as in Gay and Moore , we reach the same conclusion under either one.

The nursing faculty asserts that it has established a prima facie Title VII case of sex-based wage discrimination under both the disparate treatment and the disparate impact models, although it cannot show substantial equality of jobs. The Supreme Court’s opinion in Gunther held that a Title VII cause of action exists outside the scope of the Equal Pay Act. 452 U.S. 161 , 101 S.Ct. 2242 , 68 L.Ed.2d 751 (1981). The Court affirmed our holding “that a plaintiff *700 is not precluded from establishing sex-based wage discrimination under some other theory [than substantial equality] compatible with Title VII.” Gunther, 623 F.2d at 1321 . We added that “[i]t is unnecessary to determine now what theories might be feasible.” Id. The Supreme Court also declined to decide “the precise contours of lawsuits challenging sex discrimination in compensation under Title VII,” Gunther, 452 U.S. at 181 , 101 S.Ct. at 2254 , and took pains to “emphasize at the outset the narrowness of the question” it faced, id. at 166 , 101 S.Ct. at 2246 , Accord Plemer v. Parsons-Gilbane, 713 F.2d 1127, 1131-34 (5th Cir.1983) (Gunther a special situation where plaintiffs presented clear, direct evidence of intentional discrimination).

A.

First, we examine the nursing faculty’s claim under the disparate treatment model of sex discrimination. Plaintiffs proceeding under this theory must prove by the preponderance of the evidence a prima facie case of discrimination. Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 252-53 , 101 S.Ct. 1089, 1093-94 , 67 L.Ed.2d 207 (1981) (Burdine ); McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 , 93 S.Ct. 1817, 1824 , 36 L.Ed.2d 668 (1973) (McDonnell Douglas). To do so, the nursing faculty must show “proof of actions taken by the employer from which we infer discriminatory animus because experience has proved that in the absence of any other explanation, it is more likely than not that those actions were bottomed on impermissible [i.e. sex-based] considerations.” Furnco Construction Co. v. Waters, 438 U.S. 567, 580 , 98 S.Ct. 2943, 2951 , 57 L.Ed.2d 957 (1978); White v. City of San Diego, 605 F.2d 455, 458 (9th Cir.1979). While the burden of establishing a prima facie case “is not onerous,” Burdine, 450 U.S. at 253 , 101 S.Ct. at 1094 , disparate treatment cases necessarily require direct or circumstantial proof of discriminatory motive, whereas no such proof of motive is required in.disparate impact cases. International Brotherhood of Teamsters v. United States, 431 U.S. 324 , 335 n. 15, 97 S.Ct. 1843 , 1854 n. 15, 52 L.Ed.2d 396 (1977) (Teamsters); Jackson v. Seaboard Coast Line R. Co., 678 F.2d 992, 1014 (11th Cir.1982); Heagney v. University of Washington, 642 F.2d 1157, 1163 (9th Cir.1981) (Heagney).

Of course, Title VII’s nature and purpose require that the McDonnell Douglas test be flexible. What must be shown to support an inference that the plaintiff was discriminated against depends on the facts of each case. Peters v. Lieuallen, 693 F.2d 966, 969 (9th Cir.1982); Hagans v. Andrus, 651 F.2d 622, 626 (9th Cir.1981). Thus, we do not specify the minimum factors required for plaintiffs to establish a prima facie Title VII case of sex-based wage discrimination.

Because the district court did not clearly err on the issue of substantial equality, the nursing faculty’s case must rest solely on evidence creating an inference that the wage disparity they point to was more likely than not the result of intentional sex discrimination. Under Gunther, it is certain that if the nursing faculty were to base its prima facie case only on a comparison of the work performed, it would have to show substantial equality. Gunther, 623 F.2d at 1321 . We stated that “because a comparable work standard cannot be substituted for an equal work standard, evidence of comparable work, although not necessarily irrelevant in proving discrimination under some alternative theory, will not alone be sufficient to establish a prima facie case.” Id. Thus, the nursing faculty has failed to establish a Title VII claim on an equal pay for equal work theory for the same reasons that we concluded it failed to establish a prima facie Equal Pay Act claim. See Piva v. Xerox Corp., 654 F.2d 591, 598 (9th Cir.1981). That payments are different is insufficient alone to establish a prima facie case.

We will not, therefore, infer intent merely from the existence of wage differences between jobs that are only similar. Gunther does not require this. The com *701 parability of the jobs, however, can be relevant to determining whether we can infer discriminatory animus. Gunther allows a “comparison of somewhat dissimilar jobs.” The plaintiff must still ultimately prove intent to discriminate to make out a case of disparate treatment. See Burdine, 450 U.S. at 253 , 101 S.Ct. at 1093 .

The nursing faculty suggests we interpret Gunther as providing for a “comparability plus” test — that is, as requiring only some degree of job comparability plus some combination of factors including direct and circumstantial evidence of discriminatory conduct and pay disparities. This would be, the nursing faculty argues, a sliding scale where the “plus” factors vary in inverse proportion to the degree of comparability shown. We reject the proposal and do not read Gunther as providing such a test. Gunther explicitly refused to adopt a precise formula for Title VII litigation. 452 U.S. at 181 ,101 S.Ct. at 2253. Such an unwieldy test might allow plaintiffs to bolster inadequate showings of comparability with a confusing potpourri of “plus factors,” plunging courts into standardless supervision of employer/employee relations.

The crux of the nursing faculty’s case, then, is something more within our competence — namely, the showing of an inference of discriminatory animus. The magistrate and the district court, after an extensive hearing, found no showing of discriminatory intent or motive and that whatever disparities in pay existed were not due to sex-based treatment. In arguing that the court was clearly erroneous, the nursing faculty submits a collection of: their witnesses’ testimony; evidence of an alleged “predisposition” toward discriminatory conduct by various University officials; and various statistics. We now consider those arguments.

The nursing faculty contends that an inference of discriminatory intent follows from the University’s failure to cooperate in developing and providing information in response to the discrimination charge and its failure to appoint sufficiently experienced persons to act as women’s rights advocates. More than ample evidence shows, however, that the University did cooperate in good faith. It met with the nursing faculty, provided data as to salaries, degrees, and entry dates for nursing faculty, women at the University and representative males, developed studies on its own computers, and initiated new salary studies. The district court found that the University generally worked to improve the status of women employed there and one aspect of its efforts was the formation of an office for equal opportunity for women, organized directly under the University’s executive vice-president. 8

The nursing faculty contends that the all-male composition of the University’s three-person budget committee bolsters an inference of discriminatory animus. However, no evidence showed any discriminatory conduct on their part. In fact, the magistrate’s findings established that the School of Nursing salaries were not only kept competitive with other schools, but that the salary base there expanded at a rate faster than that of all but the School of Law.

The nursing faculty also argues that acts of the all-male budget committee forced the School of Nursing to rely heavily on the instructor rank, the lowest paid and lowest voting rank of the faculty. But, extensive evidence showed that the use of the in *702 structor rank was unrelated to sex and justified by legitimate non sex-related factors. The instructor rank enabled the school to hire inexperienced, non-doctorally prepared faculty. No showing was made that other disciplines hired persons of similar backgrounds at higher ranks, nor that the School of Nursing hired any doctorally prepared faculty at the instructor rank. Of the thirty-nine plaintiffs, the three who were doctorally prepared were hired at the level of assistant professor. Other testimony indicated, in fact, that other disciplines hired persons of the same background as teaching assistants, a non-faculty rank.

The nursing faculty next argues that its evidence showed that one of the University’s administrators, Dr. Grayston, Vice-President for Health Services, exhibited a “demeaning attitude” toward the nursing profession and the School of Nursing. This attitude, they argue, was merely part of a larger “predisposition to discriminate” that pervaded certain responsible administrators at the University and included “a clear lack of concern for the needs of female faculty members.” The extent to which the nursing faculty tries to construct a “pattern or practice” case is unclear from the briefs. However, such pigeon-holing of the nursing faculty’s theory is unnecessary because its evidence does not prove “more than the mere occurrence of isolated or ‘accidental’ or sporadic discriminatory acts,” nor does it establish that discriminatory treatment is “the [University’s] standard operating procedure — the regular rather than the unusual practice.” Teamsters, 431 U.S. at 336 , 97 S.Ct. at 1855 .

The nursing faculty correctly points out that we have said that the “existence of a discriminatory attitude ... tends to establish that it is more likely than not that the University’s decision was based on an impermissible criterion, and therefore tends to establish [the plaintiff’s] prima facie case.” Lynn, 656 F.2d at 1343 . That citation of Lynn does not, however, control these facts. Lynri reversed the district court for not adequately investigating the elements of Lynn’s prima facie treatment ease under McDonnell Douglas in a University’s tenure decision case. Id. at 1342-44. We said, “disdain for women’s issues, and a diminished opinion of those who concentrate on those issues,” could evidence an attitude which tended to establish Lynn’s prima facie case when presented with other evidence relevant to an inference of discriminatory motive. Id. at 1343. The district court there found that the tenure decision may have been influenced by such general disdain. Id. We emphasized in Lynn that “the view we express is a narrow one,” id. at 1343 n. 5. The importance of such a holding in the area of tenure decisions is obvious: where academic judgments should be made on the basis of the individual’s qualitative performance, an irrational disdain for women’s issues has no part in the deliberation. Nevertheless, in Lynn, we reaffirmed our sensitivity to “the need for courts to refrain from substituting their judgment for that of educators in areas affecting the content of curricula.” Id.

The nursing faculty, however, presents insufficient evidence of an endemic discriminatory attitude that would bolster a prima facie case. There is no evidence that the so-called academic disdain for women’s issues resulted in any wage discrimination. The testimony of former Dean Leininger that Grayston was “arrogant, authoritarian and autocratic toward female administrators” or that he was “ill-informed as to the role of academic nursing” or that he disagreed with some of the nursing faculty over the curriculum, without more, is not sufficient to overturn the factual findings entered below. Title VII does not allow plaintiffs to translate their subjective impressions into evidence of discriminatory motivation when those impressions stem from good faith disagreements. Such a rule would squelch any meaningful internal policy debate at the University.

We are similarly unconvinced that the University discriminatorily handled the Nursing School’s request for a Ph.D. program, because of a demeaning attitude to *703 ward nursing as an academic discipline. The program eventually was approved and the record demonstrates that the nursing faculty failed to prove that the Nursing School’s Ph.D. program was treated differently than any other proposals submitted by other schools and departments.

The nursing faculty also contends that the University instituted a series of equity adjustments to amend for “past discrimination” and that this practice is relevant to present discriminatory conduct by the University. The label placed on such payments, however, is not dispositive. The evidence showed these adjustments were part of a policy to pay each discipline’s faculty equitably — among themselves and as compared to faculty at other institutions. 9 Cf. Christensen v. State of Iowa, 563 F.2d 353, 357 (8th Cir.1977) (University’s good faith attempt to remove wage disparities through special evaluation system not faulted as a violation of Title VII).

Finally, the nursing faculty presented statistical evidence to support its prima facie case of disparate treatment. Generalized statistical evidence can be relevant to an adaptation of the McDonnell Douglas criteria for a prima facie case of disparate treatment. Lynn, 656 F.2d at 1342 (generalized statistics relevant to individual’s prima facie case because tenure decisions in academic context are highly subjective). See also O’Brien v. Sky Chefs, Inc., 670 F.2d 864 , 866 (9th Cir.1982) (class action).

The Supreme Court has said that, in treatment cases, “[pjroof of discriminatory motive is critical, although it can in some situations be inferred from the mere fact of differences in treatment.” Teamsters, 431 U.S. 324 , 335 n. 15, 97 S.Ct. 1843 , 1854 n. 15, 52 L.Ed.2d 396 . Statistics may be part of the circumstantial proof bolster-

ing such an inference. The statistics must clearly show the differences in treatment, thus their usefulness “depends on all of the surrounding facts and circumstances.” Id. at 340 , 97 S.Ct. at 1857 . For example, the probative force of the statistics can be reinforced with testimony of experiences of discriminatory treatment of individuals to bring “the cold numbers convincingly to life.” Id. at 339 , 97 S.Ct. at 1856 . Of course, if the plaintiff only presents the circumstantial evidence of statistics in lieu of all other proof of discriminatory intent, the case becomes a case of disparate impact.

Granting that statistics “when properly authenticated, constitute an accepted form of circumstantial evidence of discrimination,” EEOC v. Federal Reserve Bank of Richmond, 698 F.2d 633 , 645 (4th Cir. 1983), we are still sobered by warnings that statistical evidence has an “inherently slippery nature,” Wilkins v. University of Houston, 654 F.2d 388, 395 (5th Cir.1981), vacated and remanded, 459 U.S. 809 , 103 S.Ct. 34 , 74 L.Ed.2d 47 (1982), aff'd on remand, 695 F.2d 134 (5th Cir.1983), and “can be exaggerated, oversimplified, or distorted to create support for a position that is not otherwise supported by the evidence.” Note, Judicial Refinement of Statistical Evidence in Title VII Cases, 13 Conn.L.Rev. 515, 525 (1981). Thus, the weight of statistical proof relies implicitly on “the existence of proper supportive facts and the absence of variables which would undermine the reasonableness of the inference of discrimination.” White v. City of San Diego, 605 F.2d at 460 , quoting United States v. Ironworkers Local 86, 443 F.2d 544, 551 (9th Cir.), cert. denied, 404 U.S. 984 , 92 S.Ct. 447 , 30 L.Ed.2d 367 (1971).

Because we have already found that the nursing faculty’s statistics do not show *704 substantial equality of the jobs in question, we must evaluate the extent to which those statistics show comparability and then to what extent that comparability showing, along with other evidence of discriminatory animus, supports an inference of illegally discriminatory motive.

Statistics may not be any the less rigorous simply because they only go to comparability and not to substantial equality. This is especially true when we are asked not only to take them for what they mean, but also to draw an inference from them as to motive. The nursing faculty’s statistics are not general so much as they are strictly comparative: they compare salaries across jobs within the University that their experts have already chosen as comparable and make no comparisons with other universities or the labor market in general. The validity or usefulness of such comparative statistics therefore rests directly on their capacity to single out the factors that convert merely different treatment into unjustified discriminatory treatment. The more sophisticated the method of algebraic adjustment that is used, such as multivariate regression analysis, the more likely an illicit discriminatory factor can be ferreted out. See generally Baldus and Cole, Statistical Proof of Discrimination (Supp.1983); Fisher, Multiple Regression in Legal Proceedings, 80 Col.L.Rev. 702, 702, 705 (1980) (“Multiple regression analysis is a device for making precise and quantitative estimates of the effects of different factors on some variable of interest ... to extract a systematic signal from the noise presented by data”).

Intuitively, evidence of a pay disparity between jobs that are only comparable says very little about discrimination. We refuse to specify exactly what role such comparative statistics play, however, because the nursing faculty’s statistics are so rife with unreliability that they are a poor vehicle for such a purpose. The nursing faculty’s statistics were compiled by a paralegal services company that did not use a regression model but only, apparently, a simple matching technique. The selection of comparable faculty in other departments unrealistically assumed the equality of all master’s degrees, ignored job experience prior to University employment and ignored detailed analysis of day-to-day responsibilities. Finally, the nursing faculty’s statistics never compared female nursing wages to wages of female faculty in other departments. Without such a comparison, we have no meaningful way of determining just how much of the proposed wage differential was due to sex and how much was due to discipline. The statistics were generated from input that failed to control for exactly those differences between individuals that can legitimately lead to their being treated differently. Cf. Agarwal v. Arthur G. McKee and Co., 19 FEP Cases 503, 512 (N.D.Cal.1977), aff'd, 644 F.2d 803 (9th Cir.1981) (plaintiffs’ statistical study ignored similar essential factors); Valentino v. U.S. Postal Service, 511 F.Supp. 917, 957 (D.D.C.1981), aff'd, 674 F.2d 56 (D.C.Cir.1982); Vuyanich v. Republic National Bank of Dallas, 505 F.Supp. 224, 280 (N.D.Tex.1980); Keely v. Westinghouse Electric Corp., 404 F.Supp. 573 (E.D.Mo.1975). Further, the faultiness of the nursing faculty’s data base amplifies the effect of this methodological deficiency. The data base was built on information derived from the President’s record cards, which, ample testimony and exhibits showed, were inaccurate.

The nursing faculty responds that expert testimony reinforced its methodology. That argument, however, is wholly unconvincing. Dr. Lieberman, the nursing faculty’s expert, presented no specific support for the statistical methodology, was not an expert on the application of statistical studies to sex discrimination, and presented a conclusion that rested in part on the unreliable statistics. Leininger’s testimony that “there is a commonality in terms of social and helping process” and a “close interdigitation” between some of the disciplines based on joint conferences and seminars, with no other identification of specific duties and responsibilities, breathes no life into the nursing faculty’s statistical proof.

*705 B.

The nursing faculty also argues that it presented a prima facie case of discrimination under the disparate impact theory, first promulgated in Griggs v. Duke Power Co., 401 U.S. 424 , 91 S.Ct. 849 , 28 L.Ed.2d 158 (1971) (Griggs) (application of general intelligence test to minority group). Explaining the theory’s boundaries, the Court said that Title VII protects people from “not only overt discrimination, but also practices that are fair in form, but discriminatory in operation.” It does not mandate special employment treatment for “any person ... simply because he was formerly the subject of discrimination, or because he is a member of a minority group.” Griggs, 401 U.S. at 430-31 , 91 S.Ct. at 853 .

A showing of discriminatory animus is replaced under this theory by a showing of disproportionate impact, on a member of a group protected under Title VII, by an employer’s facially neutral practice. Heagney, 642 F.2d at 1163 . Where applicable, the elements of a prima facie case under the most general formulation of the disparate impact model are: (1) the occurrence of certain outwardly neutral employment practices, and (2) a significantly adverse or disproportionate impact on persons of a particular sex produced by the employer’s facially neutral acts or practices. Dothard v. Rawlinson, 433 U.S. 321, 329 , 97 S.Ct. 2720, 2726 , 53 L.Ed.2d 786 (1977); Moore, 708 F.2d at 481 ; Heagney, 642 F.2d at 1163 .

Under this model, the nursing faculty avoids proving discriminatory animus and benefits by being enabled “to shift a substantial burden to the employer upon establishing a prima facie case,” Gay v. Waiters’ and Dairy Lunchmen, 694 F.2d at 537 ; Contreras, 656 F.2d at 1271 . The benefit however is “weighed against the requirements of a prima facie disparate impact case, which are in some respects more exacting than those of a disparate treatment case.” Moore, 708 F.2d at 482 . Plaintiffs must prove not merely circumstances raising an inference of discriminatory impact; they must prove the discriminatory impact at issue. Moore, 708 F.2d at 482 (involving impact showing with respect to promotion policy); see also Johnson v. Uncle Ben’s, Inc., 657 F.2d 750 , 753 (5th Cir.1981), cert. denied, 459 U.S. 967 , 103 S.Ct. 293 , 74 L.Ed.2d 277 (1982).

Gunther, a disparate treatment case, offers no help in evaluating the nursing faculty’s impact claim. The nursing faculty has only the “comparable worth” theory upon which to rely. See, e.g., Power v. Barry County, 539 F.Supp. 721, 722 (W.D.Mich.1982) (defining comparable worth theory). Its case essentially becomes a claim that the nursing faculty received unequal pay for comparable work of equal value to its employer.

The nursing faculty’s impact case is simply stated: they have shown a disparate impact by showing a wage disparity between only comparable jobs and this disparate impact is caused by the facially neutral policy or practice of the University of setting wages according to market prices for jobs in the disciplines. We confront then the difficult question whether the disparate impact model is available to plaintiffs who, as the nursing faculty here, make a broad-ranging sex-based claim of wage discrimination, based on comparable worth. 10

*706 Doubtless because “an increasing patina of subtlety and complexity has obscured alleged claims of sex discrimination,” Bryant v. International Schools Services, Inc., 675 F.2d 562, 565 (3d Cir.1982) (challenge to policy giving unequal employment benefits to people hired to work in Iran), we must be sensitive to Title VII’s broad remedial purview, especially when employers use “artificial, arbitrary and unnecessary barriers to employment ... invidiously to discriminate on the basis of racial or other impermissible classification.” Griggs, 401 U.S. at 431 , 91 S.Ct. at 853 . However, the nursing faculty asks for an extension of Title VII that would plunge us into uncharted and treacherous areas.

We hold that on facts such as these, where plaintiffs’ sex-discrimination claim is a wide-ranging claim of wage disparity between only comparable jobs, the law does not go so far as to allow a prima facie case to be constructed by showing disparate impact. Accord Power v. Barry County, 539 F.Supp. at 726 (rejecting comparable worth theory in suit by female members at sheriff’s office: “a review of the legislative history of Title VII ...” led the court there to conclude that Gunther’s “recognition of intentional discrimination may well signal the outer limit of legal theories cognizable under Title VIL”). This conclusion is supported by considerations of precedent, prudence and judicial competence, see generally Note, Sex-Based Wage Discrimination Under The Title VII Disparate Impact Doctrine, 34 Stan.L. Rev. 1083 (1982) (surveying cases and presenting arguments why courts should deny disparate impact claims based on corn-parable worth theories); Nelson, Opton, and Wilson, Wage Discrimination and the “Comparable Worth” Theory in Perspective, 13 U.Mich.J.L.Ref. 233 (1980) (explaining problems of comparable worth theory as a legal vehicle), and accords with the law and goals of Title VII. We so hold without making any broad statement as to the general availability of the impact model in other broad based sex-wage cases.

Other courts have compellingly rejected the comparable worth theory in sex-based wage discrimination claims under an impact theory. Those cases involve claims by the employer, as here, that it was paying market wages. For instance, in Lemons v. City and County of Denver, 620 F.2d 228 (10th Cir.), cert. denied, 449 U.S. 888 , 101 S.Ct. 244 , 66 L.Ed.2d 114 (1980), several city-employed nurses sued in a class action to require the City to pay them equally with comparable jobs in the community. 11 Denver set its pay by the market wages of nurses in the community. The court held that Title VII did not require employers “to ignore the market in setting wage rates for genuinely different work classifications.” Id. at 229, quoting Christensen v. State of Iowa, 563 F.2d 353 (8th Cir.1977) (Christensen). The court also said that existing authority did not require the city and county “to reassess the worth of services in each position in relation to all others.” Id. at 229. When the Supreme Court stressed that Gunther was not a case involving comparisons of the work in question, the Court distinguished Lemons precisely as a case which did. 452 U.S. at 166 n. 7, 101 S.Ct. at 2246 n. 7. The case before us is on all fours with Lemons .

*707 Christensen involved a similar comparable worth claim by female clerical employees against the University of Northern Iowa. The University referred to wages in the local labor market to set wages. The plaintiffs argued they were underpaid visa-vis comparable males who were doing jobs of equal value to the University, and that reliance on the local wage market simply perpetuated their depressed wages. Without reaching the Bennett Amendment issue eventually addressed by the Supreme Court in Gunther, the court held that the plaintiffs failed to establish a prima facie case that the wage disparity was caused by sex discrimination. 563 F.2d at 355 . It stressed that “[t]he federal policy embodied in Title VII is that individuals should be entitled to equal opportunities” in the job market, not equal results. Id. at 356 .

We agree with Lemons and Christensen and join those courts in refusing to accept a construction of Title VII allowing the nursing faculty to establish a prima facie violation of that Act “whenever employees of different sexes receive disparate compensation for work of differing skills that may, subjectively, be of equal value to the employer, but does not command an equal price in the labor market.” Christensen, 563 F.2d at 356 .

The case before us simply does not fit into the disparate impact model. The model was developed as a form of pretext analysis to handle specific employment practices not obviously job-related, such as: employers’ intelligence tests which adversely affect minority persons, e.g., Griggs; height and weight or other requirements affecting those of a certain sex, e.g., Dothard v. Rawlinson, 433 U.S. 321 , 97 S.Ct. 2720 , 53 L.Ed.2d 786 (1977); Gerdom v. Continental Airlines, 692 F.2d 602 (9th Cir.1982) (en banc); Harriss v. Pan American World Airways, Inc., 649 F.2d 670 (9th Cir.1980) (policy requiring commencement of leave upon pregnancy), or policies which exclude applicants based on arrest records, e.g., Gregory v. Litton Systems, Inc., 472 F.2d 631 (9th Cir.1972). See also Lerner, Employment Discrimination: Adverse Impact, Validity and Equality, 1979 Sup.Ct.Rev. 17 (discussing Griggs and progeny). As the court in Pouncy v. Prudential Insurance Co. of America, 668 F.2d 795, 800 (5th Cir.1982) (Pouncy), made clear:

The nursing faculty unconvincingly cites cases for the proposition that “the disparate impact analysis has been applied to wage discrimination cases.” They do not involve wide-ranging allegations challenging general wage policies but rather challenges to specific employer practices, namely, fringe benefits policies, with respect to which employers exercise judgment. The rules by which an employer determines the availability of fringe benefits can be evaluated in terms of their job-relatedness. It has been such “selection procedures to which the disparate impact model has traditionally applied,” Pouncy, 668 F.2d at 801 , and not the mere payment of market wages.

In Arizona Governing Committee for Tax Deferred Annuity and Deferred Compensation Plans v. Norris, — U.S.-, 103 S.Ct. 3492 , 77 L.Ed.2d 1236 (1983), aff'g in part, rev’g in part, 671 F.2d 330 (9th Cir.1982), for instance, the employer’s policy of providing retirement benefits based on sex-segregated annuity tables was challenged. The Court held that discrimination proscribed by Title VII occurs where an employer’s policy is to offer female employees a benefit plan requiring them to make the same contributions as male employees, yet receive lower benefits. Id. at 3499, 3501. Los Angeles Department of Water and Power v. Manhart, 435 U.S. 702 , 98 S.Ct. 1370 , 55 L.Ed.2d 657 (1978), also involved sex-segregation under a benefits plan.

Our opinion in Wambheim v. J.C. Penney Co., 705 F.2d 1492 (9th Cir.1983), also involved a specific employer policy, and is not contrary to our holding here. Wam *708 bheim, a class action, narrowly held that “disparate impact analysis is appropriate in this § 703(a)(1) case” where Wambheim and her class challenged Penney’s “head-of-household” rule under its medical and dental insurance coverage, and that plaintiffs successfully established a prima facie case. 705 F.2d at 1494 . That case we found “unusual” because it involved an allegation of violations under section 703(a)(1): discrimination with respect to “compensation, terms, conditions, or privileges of employment,” whereas the disparate impact theory had been developed in section 703(a)(2) cases. However Wambheim is inapposite here because it also specifically dealt with a particular employer policy rather than a full-scale assault on the employer’s salary practices.

We cannot manageably apply the impact model when the kernel of the plaintiff’s theory is comparable worth. The problem is compounded in this case. When the disparate impact model is removed from the cases involving challenges to clearly delineated neutral policies of employers, it becomes so vague as to be inapplicable. See Pouncy, 668 F.2d at 801 . The nursing faculty claims to have pinpointed a facially neutral policy at the University having the discriminatory impact they assert. That policy is the University’s relying on the market to set their wages. We find that they have failed to do so, and emphasize that such a practice is not the sort of “policy” at which disparate impact analysis is aimed.

Relying on competitive market prices does not qualify as a facially neutral policy or practice for the purposes of the disparate impact analysis that was first articulated in Griggs . For Title VII purposes, simply labelling an employer’s action a “policy or practice” is not sufficient. What matters is the substance of the employer’s acts and whether those neutral acts are a non-job-related pretext to shield an invidious judgment.

Every employer constrained by market forces must consider market values in setting his labor costs. Naturally, market prices are inherently job-related, although the market may embody social judgments as to the worth of some jobs. Employers relying on the market are, to that extent, “price-takers.” They deal with the market as a given, and do not meaningfully have a “policy” about it in the relevant Title VII sense. Fringe policies, which are discretionary, are altogether another matter. Additionally, allowing plaintiffs to establish reliance on the market as a facially neutral policy for Title VII purposes would subject employers to liability for pay disparities with respect to which they have not, in any meaningful sense, made an independent business judgment. As we have previously said, “Title VII does not ultimately focus on ideal social distributions of persons of various races and both sexes. Instead it is concerned with combatting culpable discrimination. In disparate impact cases, culpable discrimination takes the form of business decisions that have a discriminatory impact and are not justified by their job-relatedness.” Contreras v. City of Los Angeles, 656 F.2d 1267 , 1275 n. 5 (9th Cir.1981) (use of allegedly discriminatory auditor’s examinations challenged).

The nursing faculty asserts it has shown three other facially neutral policies as a foundation for impact analysis here. The nursing faculty argues that the University had a facially neutral policy of making all salary increases in percentages of present salaries. The evidence, however, shows that the disciplines at the University were not in a lock-step for the purpose of percentage pay increases, and that part of the appropriated money from the state legislature was sometimes allocated in varying percentages. On some occasions, the nursing faculty received greater percentage increases.

The nursing faculty also contends the University had the practice of requiring nursing faculty to start as “instructors” and progress through a four-rank promotional system. However, ample evidence showed that members of the nursing faculty were not required to begin as instructors, and the entry rank depended on the *709 person’s prior teaching experience, published research or outstanding professional accomplishments.

The final facially neutral policy alleged is the University’s “discretionary budgeting policies based on subjective considerations.” We fail to see how this qualifies as a facially neutral policy, even if it were proved. Ordinarily, the lack of well-defined criteria as facilitating wage discrimination is a claim better presented under the disparate treatment model. See Heagney, 642 F.2d at 1163 (treatment model more appropriate where gravamen of plaintiff’s claim goes to the use of subjective criteria); Pouncy, 668 F.2d at 801 ; but cf. Hung Ping Wang v. Hoffman, 694 F.2d 1146, 1148 (9th Cir.1982) (lack of objective criteria and a job promotion disparity may be basis for impact claim).

VI

We also affirm as to Ruth Fine’s claim under Title VII. As we held above, she failed to prove substantially equal work. Thus, she must make out a case of intentional discrimination with other evidence and the comparability of her job with Lang and Stein. There is no evidence before us, however, which creates an inference of intentionally discriminatory treatment suffered at the hands of the University. She also presented no evidence of disparate impact.

VII

The nursing faculty next asserts that Reg Williams, the male faculty member, has a valid claim for discrimination because he received a salary “infected” by the discrimination the female faculty members suffered. However, even where the female faculty members are able to frame a cognizable Title VII or Equal Pay Act claim, it does not allow the male employees “to bootstrap their job grievances ... into an employment discrimination claim rooted in federal law.” Ruffin v. County of Los Angeles, 607 F.2d 1276, 1281 (9th Cir.), cert. denied, 445 U.S. 951 , 100 S.Ct. 1600 , 63 L.Ed.2d 786 (1979) (sex-based wage discrimination suit brought by male officers dismissed on summary judgment). There we held “that the County’s past employment practices as to its female deputy-sheriffs, however ill-conceived, did not serve as a foundation for this suit brought under Title VII and the Equal Pay Act.” Id. at 1281. Williams makes no claim that he received a lower wage because of his sex.

VIII

In light of the above, there is no reason to disturb the district court’s denial of attorneys’ fees.

AFFIRMED.

. The parties agree that the district judge applied the clearly erroneous standard in his review. The record is unclear. In Spaulding I, we observed that the nursing faculty “obtained a de novo review by the district court.” 676 F.2d at 1234 . For purposes of this appeal, we will accept the issue as framed by the parties.

. The parties did not raise before the district court nor on appeal whether National League of Cities v. Usery, 426 U.S. 833 , 96 S.Ct. 2465 , 49 L.Ed.2d 245 (1976), and its progeny prevents the equal pay provision of section 206(d)(1) from being enforced against a state or state agency. See, e.g., Pearce v. Wichita County, 590 F.2d 128 (5th Cir.1979); Marshall v. City of Sheboygan, 577 F.2d 1 (7th Cir.1978); see also Ruffin v. County of Los Angeles, 607 F.2d 1276 , 1282 n. 3 (9th Cir.1979) (declining to decide issue), cert. denied, 445 U.S. 951 , 100 S.Ct. 1600 , 63 L.Ed.2d 786 (1980). Because the parties neither raise the issue nor contend that it affects our jurisdiction, we do not reach the question.

. The nursing faculty also contends that section 2000e-5(f)(5) does not authorize referrals of section 1983 claims. We need not address this contention because the district court lacked jurisdiction over the section 1983 claim. See part II.

. Section 2000e-5(f)(5) provides that:

. Rule 21 of the Western District of Washington Magistrates’ Rules provided that:

The rule is now Magistrates’ Rule 5.

. Our decision in Gunther v. County of Washington, 623 F.2d 1303 (9th Cir.1979), aff’d, 452 U.S. 161 , 101 S.Ct. 2242 , 68 L.Ed.2d 751 (1981), is also reported at 602 F.2d 882 .

. Because we rule on the adequacy of the prima facie case, we do not address whether the University may rely on the competitive marketplace as a defense to a successful prima facie case under Title VII.

. Part of the nursing faculty’s evidence was not admitted as substantive proof, but rather to show that the University was on notice that various groups felt women’s issues needed more attention. These were a 1970 report by the Women’s Commission of the Associated Students of the University; documents written by the Coordinator of the Women’s Caucus, Washington Library Association; a letter on behalf of the University’s home economics department; and Reports by the Faculty Senate Subcommittee on Academic Women and the Task Force on Personnel Policy and Practice.

The magistrate found explicitly, however, that the University worked to improve the status of women during the time relevant to this lawsuit and that finding was approved and adopted by the district court. We cannot say that the finding of the district court is clearly erroneous.

. The nursing faculty presents contested evidence that a less discriminatory alternative compensation system exists at another university. This is not relevant to a prima facie case of disparate treatment. Once a plaintiff establishes a prima facie case and the defendant has responded with legitimate, nondiscriminatory reasons for his actions, the plaintiff then has a chance to show the proffered reasons are merely pretextual. Burdine, 450 U.S. 248, 254-56 , 101 S.Ct. 1089, 1094-95 , 67 L.Ed.2d 207 . It is at this last stage that such evidence could be relevant.

. The controversial "comparable worth” theory essentially holds that men, women and minorities should be paid equally for jobs that are of comparable value to the employer. "At its broadest, ‘comparable worth’ encompasses any claim for sex-based wage discrimination ... in which the plaintiff does not perform work ‘substantially equal’ to that of a higher paid employee of the opposite sex.” Schlei and Grossman, Employment Discrimination Law 475 (2d ed. 1983). This value to the employer, some argue, can be objectively measured. The market does not accurately portray that value because of defects in the market caused by pervasive discrimination. Compare Blumrosen, Wage Discrimination, Job Segregation and Title VII of the Civil Rights Act of 1964, 12 U.Mich.J.L.Ref. 397 (1979) (same) with Nelson, Opton and Wilson, Wage Discrimination and the “Comparable Worth" Theory in Perspective, 13 U.Mich.J.L.Ref. 233 (1980) (no social science technique exists to show that residual wage differentials are, in fact, sex-based). Comparable worth theorists have also attempted to argue that discriminatory practices in the labor market are ultimately *706 inefficient because not job-related. This, of course, can be said only based on the belief that the market is defective. See Note, Equal Pay, Comparable Work and Job Evaluation, 90 Yale LJ. 657, 672 (1981). But cf. Beller, The Economics of Enforcement of An Antidiscrimination Law: Title VII of the Civil Rights Act of 1964, 21 J.L. and Econ. 359 (1978) (empirical analysis suggests that laws cannot eliminate both sex and wage discrimination because conflicting provisions simultaneously increase and decrease relative employment of the protected group, with a cancelling-out effect).

. The nurses in Lemons did not want to be paid as nurses in the community were paid because, they argued, the local market price for nurses was depressed because of past social discrimination. The nursing faculty's arguments here also rely on their historically depressed wages being due to discrimination in an imperfect market. Courts are not competent to engage in a sweeping revision of market wage rates. Cf. Christensen v. State of Iowa, 563 F.2d 353, 356 (8th Cir.1977) (Title VII not intended to abrogate the laws of supply and demand).