This is a sex discrimination case brought under Title VII of the. Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq. Congress made the provisions of Title VII applicable to state and local governmental entities such as the University of Washington by the Equal Employment Opportunity Act of 1972, Pub.L. No. 92-261, § 2 (1), 86 Stat. 103 , effective March 24, 1972. Heagney alleges that the University paid her an unfairly low *1159 salary because of her sex, and is seeking damages for the period subsequent to March 24, 1972. She seeks the difference between her estimate of a nondiscriminatory salary and her actual salary from March 24,1972 until March 15,1973, when she left the University. Heagney also claims that she should be awarded the full value of a nondiscriminatory salary for the period since her resignation because she was “constructively discharged” from her job. Finally, she requests reinstatement at the University.

The Equal Employment Opportunity Commission investigated her complaint and attempted to reach a settlement with the University. These efforts failed and the commission granted Heagney a “Notice of Right to Sue.” A magistrate tried the case in 1977, and issued his findings and conclusions on November 3,1977. In August 1978 the district judge adopted the magistrate’s findings and conclusions in their entirety, and entered judgment for the University. Heagney then brought this appeal.

Because we conclude that a relevant statistical report was erroneously excluded from evidence the case must be remanded to the district court for reconsideration on the issue of whether Heagney proved that she was underpaid because of her sex. The evidence supports the finding that Heagney was not constructively discharged, however, and we affirm that portion of the district court’s opinion.

I. FACTS

Joanne Heagney has a bachelor’s degree in chemistry from the University of Washington and has done some post-graduate work there. From 1962 until she resigned in 1973 she was employed at the Nuclear Physics Laboratory (NPL or the Lab), which is part of the University of Washington. Her primary duty at NPL was to design and fabricate targets used in nuclear physics experiments. Heagney initially held the job title of Radiochemist I; in 1965 she was promoted to Radiochemist II; and, finally, in November 1972 she was promoted to Materials Research Scientist.

Non-academic personnel at the University of Washington are categorized as either “classified” or “exempt.” Exempt employees are those filling jobs with unique or unstandardized requirements. Although Heagney held titled positions, she was an exempt employee. Salaries of classified employees are covered by a Washington state personnel law and are set and adjusted by an administrative agency. Salaries for exempt employees were, during the time considered here, more discretionary. Heagney’s salary was set by a supervisory committee of the NPL whose decisions were in turn approved by the College of Arts and Sciences and by the University’s personnel department. The University provided the NPL with certain salary guidelines.

Heagney relied principally upon statistical evidence to prove her case of sex discrimination. First she attempted to show that as a group women exempt employees at the University received disproportionately lower salaries than men. She then attempted to show that this pattern carried over into her individual case at the Lab.

As part of her showing that women exempt employees generally received lower salaries than males, Heagney introduced the results of a study undertaken by the Equal Employment Opportunity Commission (EEOC). This study shows that among all 717 exempt employees at the University in 1973, 70 percent of the males made over $12,000, but only 28 percent of the females made over this amount.

Several witnesses also testified about statistics gathered by the Office of Civil Rights of the Department of Health, Education and Welfare. These statistics compared the pay of male and female exempt employees based on 1972-1973 data. Although never introduced into evidence, the data apparently show that women exempt employees, on average, made less than their male counterparts with similar job titles and occupational codes. Witnesses from both the University and the Office of Civil *1160 Rights agreed that the salary data was deficient because job titles were so broad that they were not an accurate indication of the actual work content. For this reason the statistics did not provide a reliable basis for comparing work performed by females with work performed by males. The University’s Director of Personnel Services criticized the EEOC statistics on the same grounds.

To avert an enforcement action, in 1974 the Office of Civil Rights and the University signed a conciliation agreement concerning employment practices at the University. As one of three required steps concerning exempt employees, the University agreed to retain an outside consulting firm to undertake an in-depth analysis of the salaries paid to exempt employees, and to recommend changes. The University hired the consulting firm of Robert Hayes and Associates as a result of the agreement.

At trial, Heagney made several attempts to introduce the May 1975 Hayes final report into evidence. The University objected on the grounds that the Hayes study was based upon 1975 salary data and was therefore irrelevant to the question of whether there was salary discrimination during the 1972-73 period when Heagney was a University employee covered by Title VII. The magistrate sustained the University’s objection and kept the Hayes study out of evidence.

Heagney has made the unadmitted Hayes report part of the record on appeal. The report indicates that the Hayes firm standardized exempt jobs by ranking each job according to a point scale using such factors as initial eligibility requirements, supervisory responsibility, creativity required and other indicia of skill, training and job importance. Based on 1975 data, the report describes a salary curve the firm prepared for the University that would insure that jobs with similar point rankings would receive comparable salaries. The report also includes a table showing how the salaries of exempt jobs in January 1975 actually compared with this curve. In relation to the salary curve, the University paid 39.2 percent of the women exempt employees below what the report established as a minimum salary. The comparable figure for male employees was 19.8 percent. The table shows that the University overpaid 14.5 percent of its exempt male employees, while it overpaid only 4.6 percent of its female employees. Thus approximately twice the percent of women were underpaid than men and more than three times the percent of men were overpaid than women. The study demonstrates a significant salary disparity between male and female exempt employees.

To show disparities in her own salary, Heagney introduced a survey prepared by Battelle Columbus Laboratories comparing salaries earned by scientists across the United States. The court found that Heagney’s salary was about 73 percent of the national mean for an equivalent job description in the Battelle survey. Three male employees at NPL had salaries below, but closer to, the national mean than Heagney’s salary for the type of work they did. Stowell had a salary that was 76 percent of the mean, Roth’s was 78 percent, and Cheney’s was 85 percent. In his oral opinion, the magistrate noted that the major problem he had with the use of this survey was classifying Heagney’s job properly so she could be matched with the appropriate category in the survey.

Finally, Heagney introduced statistics showing the starting salary, rates of salary increase, and education and experience of male and female exempt employees at NPL. Tabulations showing the salary history of Heagney, three male employees, and Kellenbarger, another female employee, are summarized in the margin. 1 These statis *1161 tics arguably indicate some inequity when Heagney’s salary is compared with male employees at least during the period prior to the date that the Equal Employment Opportunity Act became applicable. 2 .

At various times the Lab requested salary increases for Heagney and other exempt employees. For example, in 1968 Professor Vandenbosch wrote David Williams, the University’s personnel director:

If the evidence summarized above shows a discriminatory pattern prior to the effective date of the legislation extending Title

Heagney Kellenbarger (Female) Cheney (Male) Roth (Male) Stowell (Male)

Oct. 1962' $ 425 *

Jan. 1966 630 $ 620*

July 1966 661 651 $ 900*

June 1969 802 788 1021 $ 740*

Jan. 1970 882 867 1196 $ 845*

Mar. 1973 ** 1070 1004 1375 985 1048

*1162 VII proscriptions to state governmental entities, March 24, 1972, it is necessary to determine whether the pattern was carried forward after the effective date.

An examination of raises as a proportion of initial salary indicates that the male employees received increases similar to Heagney’s. Thus if the evidence shows an initial inequality between Heagney’s salary compared to the other exempt employees, the data indicate that Heagney did not appreciably narrow the gap through pay increases, although the gap did not widen. 4

The events which apparently precipitated the present suit began in June 1972, when the NPL gave Roth a 15.2 percent pay increase and Stowell 10 percent, but gave nothing to Heagney. She submitted a letter of resignation on September 1,1972, and stated as a condition for further employment that she be granted a 30 percent pay raise. In response Professor Weitkamp asked for a 10 percent raise for Heagney and 5 percent for the rest of the exempt staff. Heagney then contacted Human Rights officials at the University, and several months later contacted the EEOC and HEW. The University approved the 10 percent raise for Heagney effective in October 1972, and later approved a 5 percent raise for the other exempt employees retroactive to October.

The University has maintained that its promotion policies were designed to award especially meritorious service. As an explanation of why Heagney may not have been paid more, the magistrate found that during the “latter part” of her employment Heagney’s job enthusiasm slacked, she was not working a forty-hour week, and she may have had some conflict of interest between her job at NPL and a family-run business. Furthermore, her relationship with other personnel at the Lab had deteriorated.

In discussing Heagney’s apparent personnel problems the magistrate, in his oral decision, stated

He further found that many of Heagney’s other personnel problems had been “magnified” at trial to more importance than they had when they occurred.

On the other hand, the magistrate found that Heagney was “highly skilled.” She had received written commendations for her work. In 1971 Heagney had been invited to the Conference of Research Materials at Oak Ridge, Tennessee to give a paper describing her work. Professor Weitkamp noted that she had “achieved international recognition.” She was “one of perhaps one or two dozen people in the United States who have made a professional specialty of fabricating such targets.” When Heagney resigned Professor Bodansky of the NPL wrote her that her work had been “excellent” and they would welcome her back.

With regard to the statistical evidence the magistrate noted that none of it compared Heagney’s job with the same type of job performed by other male employees. According to the magistrate, Heagney’s job was unique. The district judge, in adopting the magistrate’s findings and conclusions, noted that the magistrate “consistently based his legal conclusions on a finding that *1163 the Plaintiff’s employment was unique and, as such, took her out from under the statistical evidence she advanced.” Therefore, because Heagney’s “proofs ... were based substantially, if not entirely, on statistical evidence” the judge did not believe that there was sufficient proof of sex discrimination. He agreed with the magistrate’s conclusion that there was no sex discrimination involved in establishing Heagney’s salary. The judge further found there was no evidence to support Heagney’s charge that she was constructively discharged.

II. PROOF OF SEX DISCRIMINATION

In International Brotherhood of Teamsters v. United States, 431 U.S. 324 , 335 n.15, 97 S.Ct. 1843 , 1854-1855 n.15, 52 L.Ed.2d 396 (1977) the Court discussed the difference between Title VII “impact” cases and “treatment” cases. In a disparate impact case, a plaintiff must show that some facially neutral employment practice has a substantially disproportionate impact upon a group protected by Title VII. Examples of such outwardly neutral employment practices include height and weight requirements, see Dothard v. Rawlinson, 433 U.S. 321 , 97 S.Ct. 2720 , 53 L.Ed.2d 786 (1977), or certain types of employment tests adversely affecting those of certain cultural backgrounds, see Griggs v. Duke Power Co., 401 U.S. 424 , 91 S.Ct. 849 , 28 L.Ed.2d 158 (1971). Once plaintiffs have shown the existence and impact of such a practice, which constitutes a prima facie case, the employer will be held liable unless the practice can be justified by “business necessity.” Dothard, 433 U.S. at 329 , 97 S.Ct. at 2727 ; Griggs, 401 U.S. at 430 , 91 S.Ct. at 853 .

A “treatment” case involves a situation where an employer treats an individual protected by Title VII differently simply because of the person’s minority status, religion or sex. In a treatment case, unlike an impact case, it is necessary to show an employer’s intent to discriminate, but intent can be inferred from circumstantial evidence. Teamsters, 431 U.S. at 335 n.15, 97 S.Ct. at 1854-1855 n.15. Once a plaintiff has introduced evidence from which it appears more likely than not that the defendant had an intent to discriminate, a prima facie “treatment” case is established. An employer will be liable unless it can “articulate” reasons which justify disparate treatment. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 , 93 S.Ct. 1817, 1824 , 36 L.Ed.2d 668 (1973); Golden v. Local 55 of the International Association of Firefighters, 633 F.2d 817 at 821 (9th Cir. 1980). If an employer does articulate legitimate, nondiscriminatory reasons for disparate treatment, the employee can then show that these reasons were a pretext to hide bias. McDonnell Douglas, 411 U.S. at 804-05 , 93 S.Ct. at 1825-1826 .

A plaintiff in a Title VII ease can plead both theories. Although the disparate impact theory is ordinarily asserted in class actions, an individual claimant may seek relief under such a theory as well. Wright v. National Archives & Records Service, 609 F.2d 702, 711-12 (4th Cir. 1979) (en banc). Heagney alleges that categorizing certain employees as “exempt” is a neutral employment practice which the statistics indicate has a disproportionate impact upon the salaries of female employees. She reasons that she is a member of the class of persons which has been disadvantaged by this practice, therefore the University should be liable.

It is apparent, however, that the creation of jobs that are exempt from the Washington personnel law cannot be equated with such well-defined objective employment practices as personnel tests or minimal physical requirements. Classification of certain jobs as “exempt” only meant that the University had wider discretion to establish the employee salaries. Subjective employment decisions may result in discrimination, but the use of subjective criteria is not per se illegal. Ward v. Westland Plastics, Inc., No. 78-1666, slip op. at 4193 (9th Cir. Apr. 25, 1980). The gravamen of Heagney’s complaint is that the lack of well-defined employment criteria allowed a pattern or practice of discrimination to exist. We therefore conclude that “impact” analysis is inappropriate and that Heagney was required to prove disparate treatment.

*1164 The magistrate and trial judge were both of the opinion that generalized statistical evidence including all exempt employees was of little relevance to Heagney’s individual ease. Generalized statistics are, however, undoubtedly relevant in cases alleging disparate treatment. In Teamsters, the government brought an action against a union and trucking company alleging that the defendants had engaged in purposeful discrimination against Negroes and Spanish-surnamed individuals. The government sought a general injunction and individual relief for those minority individuals who had been denied line-driver jobs. In proving its case, the government relied heavily upon statistical evidence showing that the company employed few Negro or Spanishsurnamed individuals as line drivers, and that those minorities the company did employ were paid less. The Court concluded that such statistics could serve to support a prima facie case of disparate treatment. 431 U.S. at 337-342 , 97 S.Ct. at 1858.

After liability was established, the union and company argued at the remedial stage that the Court could not award individual relief to all persons who had sought line-driver jobs. In the defendants’ view, relief could only be granted where some evidence of discrimination had been introduced regarding a particular claimant. The Court disagreed. Instead, the Court concluded that once the government had shown that the defendants had engaged in discrimination, that conclusion supported an inference that every individual member of the class had been the victim of discrimination, even if no evidence had been introduced in support of a particular individual’s claim. 431 U.S. at 362-71 , 97 S.Ct. at 1868-1873. Thus every individual member of the class harmed received the benefit of the general statistical showing, and the burden shifted to the employer to rebut that inference by offering some legitimate non-discriminatory reason.

The logical conclusion from the Supreme Court’s opinion in Teamsters is that generalized statistics are relevant to an individual action under Title VII. If Heagney could demonstrate that the University regularly discriminated against its female exempt employees, such a showing, even if it were inadequate by itself to support a prima facie case, certainly would increase the probability that Heagney’s own case was also part of this pattern or practice. This would be true even if her job were unique. Several circuits, including our own, have relied on general statistical evidence to support an individual’s prima facie case of sex discrimination. See Davis v. Califano, 613 F.2d 957, 962-63 (D.C.Cir.1979); Sweeney v. Board of Trustees of Keene State College, 569 F.2d 169 , 177 (1st Cir.), vacated on other grounds, 439 U.S. 24 , 99 S.Ct. 295 , 58 L.Ed.2d 216 (1978); Kaplan v. International Alliance of Theatrical & Stage Employees, 525 F.2d 1354 , 1358 (9th Cir. 1975). 5

In Heagney’s case the EEOC and Office of Civil Rights statistics which she introduced into evidence do not adequately indicate the nature of the work performed, or whether females were denied promotions or pay raises they were entitled to receive. No meaningful comparison of female and male salaries is possible from these statistics. 6 See Hagans v. Andrus, No. 79-4424, *1165 (9th Cir. Feb. 5, 1981); Pack v. Energy Research & Development Administration, 566 F.2d 1111, 1113 (9th Cir. 1977). Without more, the judge’s ultimate conclusion that Heagney failed to prove sex discrimination is supported by the evidence.

The Hayes study, however, apparently adjusts for deficiencies in these earlier statistics by establishing a standardized basis for comparing job content with pay even though the job may be unique. The issues before this court, then, are whether the trial court erred by failing to admit the study, and if so, whether the error was harmless.

The magistrate excluded the Hayes study because it dealt with January 1975 data, not data from before March 15, 1973, when Heagney resigned. Under the circumstances, it is extremely unlikely that the approximately twenty-two month delay between Heagney’s resignation and the salary data collected by Hayes had any significant impact on the probative force of the study as to the existence of conditions when Heagney was an employee. For example, the fact that a piece of property was in a certain physical condition either sometime before or after a particular event will support an inference that the same condition existed at the time the event occurred. See Conner v. Union Pacific Railroad, 219 F.2d 799, 802 (9th Cir. 1955), Annot., 7 A.L.R.3d 1302 (1966). While conditions subject to human decisions may be prone to more variations, the same inference would apply in the absence of a contrary explanation.

In this case, testimony by the University indicates that the Hayes study was undertaken because preliminary statistics, gathered in 1972-73 when Heagney was still a University employee, raised concerns about possible sex discrimination. In part, the purpose of the Hayes study was to confirm or deny any inferences raised by the statistics. Title VII became applicable to the University in March 1972. Equal Opportunity Act of 1972, Pub.L.No.92-261, § 2(1), 86 Stat. 103 . It seems likely that the University would have become more conscious of possible liability for sex discrimination rather than less so, especially because the University was then under investigation by the federal Office of Civil Rights. Although a witness from the Office of Civil Rights testified that the problems they were concerned about at the University continued past 1973, there is little likelihood that discrimination would increase during the post-1973 period. This foundation was adequate to demonstrate that the Hayes data was probative as to the existence of conditions prior to the time when the study was prepared. See Kinsey v. First Regional Securities, Inc., 557 F.2d 830, 839 (D.C.Cir.1977). Rule 401 of the Federal Rules of Evidence defines relevant evidence as that which has “any tendency to make the existence of any fact that is of consequence to the determination of the action more probable or less probable than it would be without the evidence.” The Hayes report had a tendency to make it more probable that Heagney was discriminated against and as such was admissible. See Fed.R.Evid. 402. We conclude that it was error to exclude the Hayes study.

Finally, the evidence in this case is too close to conclude that the error in failing to admit the Hayes study was harmless. The Hayes study indicates that more female than male exempt employees were underpaid. The evidence from Heagney’s individual case indicates that she may have been one of the women who were underpaid. This evidence included the salary comparison with male exempt employees, considering their educational attainments and experience, and at least three internal memoranda prepared by her supervisors, specifically noting that Heagney’s salary was not only low for a chemist, but low in relation to other University employees and exempt staff members.

*1166 The EEOC found that there was probable cause to believe that there was sex discrimination involved in establishing Heagney’s salary. The Office of Civil Rights also concluded that it appeared that sex discrimination was occurring. Against this background, the exclusion of the Hayes study was not harmless error and the case must be remanded to determine whether Heagney was underpaid because of her sex.

On remand the University may wish to submit evidence challenging the results of the Hayes study. The court must then determine whether the evidence, including the Hayes study, establishes a prima facie case of discrimination. If so, the burden will be on the University to offer some legitimate nondiscriminatory reason for the disparate treatment. 7

III. CONSTRUCTIVE DISCHARGE

Since the issue may arise on remand we find it necessary to discuss Heagney’s contention that she was constructively discharged.

If Heagney was constructively discharged she is entitled to recover back pay from the time of her resignation in 1973. In support of her argument Heagney relies upon Young v. Southwestern Savings & Loan Association, 509 F.2d 140, 144 (5th Cir. 1975), which held that forcing a plaintiff to attend a weekly religious ceremony created an intolerable working condition that forced the plaintiff to resign involuntarily.

The Fifth Circuit recently applied its Young decision to a factual situation quite similar to the instant case. In Bourque v. Powell Electrical Manufacturing Co., 617 F.2d 61 (5th Cir. 1980), a female plaintiff claimed that she was constructively discharged because she was not paid fairly. The court concluded that

617 F.2d at 65-66 . The court reasoned that, where possible, Title VII cases should be settled “within the context of existing employment relationships.” Id.

In his oral decision the magistrate found that Heagney, other than showing that she was possibly underpaid, had not introduced evidence showing circumstances from which it could be inferred that the University made her employment conditions difficult or intolerable. This conclusion is supported by the evidence. We hold that Heagney failed to prove that she was constructively discharged. On remand, if the court does find that Heagney was underpaid because of her sex, she may recover the difference between the salary she should have made absent discrimination and the salary she actually made at NPL for the period between March 24, 1972 and March 15, 1973.

The decision of the district court is AFFIRMED IN PART, REVERSED IN PART, and REMANDED for further proceedings consistent with this opinion.

. The following table summarizes the monthly salaries paid to Heagney, another female employee, and three male exempt employees. The lab apparently employed two other male exempt employees in supervisory jobs. The magistrate did not provide salary data for these two employees in his findings.

*1161 The percentages used in this section are rounded off to the nearest one-tenth of a percent.

Salary on date individual began working as an exempt employee at the NPL.

Salary on date Heagney resigned.

. When Cheney began working at the lab in 1966 he did not have a college degree but had 11 years of experience. At the time Cheney was hired, Heagney had four years of experience at the lab, and a bachelor’s degree in chemistry. Cheney was hired at a salary of $900 a month or almost 38 percent more than Heagney’s monthly salary at that time.

Roth had no experience other than previous part-time work at the lab when he was hired in 1969. He had a bachelor’s degree in physics. Heagney, by the time Roth was hired, had been working at the lab seven years. Despite Roth’s inexperience he was hired at a salary of $740 a month — $62 a month less than Heagney or a salary equal to about 92 percent of Heagney’s salary. ■

Stowell, after obtaining a degree in business administration, began full-time work at the lab as an electronics technician, a classified position. He had three years of full-time experience and one and a-half years of part-time experience when he was hired as an exempt employee. As an exempt employee, Stowell began at $845 a month. At this time Heagney had over seven years of experience and was making only $37 a month more than Stowell. Stowell thus started at almost 96 percent of Heagney’s salary.

Shirley Kellenbarger, the other female exempt employee at NPL, had a master’s degree in agronomy and 17 years (the EEOC report indicates that Kellenbarger had 12 years of experience. An internal memo prepared by the lab states that Kellenbarger had 9 years of experience) of work experience when she was hired in 1966. Her salary remained close to, but slightly below Heagney’s salary between 1966 and 1973.

. In 1969 Professors Vandenbosch and Weitkamp wrote the chairman of the Physics Department:

man of the Physics Department:

. Comparing the pay raises of each female employee to each male employee, beginning with the time the male employee was first employed at the NPL and ending when Heagney resigned, indicates the following. Cheney’s salary increased 52.8 percent between 1966 when he began working and 1973 when Heagney left, while Heagney’s salary increased 61.8 percent and Kellenbarger’s salary increased 54.2 percent during the same period. Roth and Heagney received comparable increases in the four years they were at the lab together. Roth received 33.1 percent in salary increases while Heagney received 33.4 percent. Kellenbarger’s salary increased 27.4 percent during this period. In the three years that both Stowell and Heagney were exempt employees Stowell’s salary increased 24 percent, Heagney’s 21.3 percent and Kellenbarger’s 15.8 percent.

. In McDonnell Douglas Corp. v. Green, 411 U.S. 792, 805 , 93 S.Ct. 1817, 1825-1826 , 36 L.Ed.2d 668 (1973), a case involving an individual claim of disparate treatment, the Court approved the use of generalized statistics, after the plaintiff had established a prima facie case, as a means of showing that an employer’s stated reasons for differing treatment were a pretext to hide bias. McDonnell Douglas did not limit the use of such statistical evidence, however, to this stage of the proceedings. As noted in Teamsters, 431 U.S. at 358 , 97 S.Ct. at 1866 , McDonnell Douglas “did not purport to create an inflexible formulation.” McDonnell Douglas further supports the conclusion that generalized statistics are relevant to proving an individual’s claim of discrimination.

. In Gunther v. County of Washington, 623 F.2d 1303, 1311 (9th Cir. 1979), cert. granted, - U.S. -, 101 S.Ct. 352 , 66 L.Ed.2d 213 (1980) we held that Title VII is broader than the Equal Pay Act, Pub.L. No 88-38, 77 Stat. 56 (1963) (codified at 29 U.S.C. § 206 (d)). Under Gunther , Heagney would not need to show that her work was the same as work performed by her male colleagues. The EEOC an OCR statistics *1165 do not provide any basis for comparison, however, and thus the probative value of the data is of little import in concluding that female employees were underpaid. The Hayes study, in contrast, does appear to provide some basis for making a meaningful comparison of male and female jobs. See Note, 54 St. John’s L.Rev. 738 (1980).

. McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 , 93 S.Ct. 1817, 1824 , 36 L.Ed.2d 668 (1973).