Certain former and present city workers (appellants) appeal from the district court's judgment that they are entitled to no relief under Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq., and 42 U.S.C. § 1981, for having lost their jobs by failing allegedly discriminatory civil service examinations. We affirm.

* Contreras, E. Gonzalez, Mock, and Zavala (accountants) were employed by the City of Los Angeles (City) as Senior Accountants in the Training and Job Development Division of the Mayor's office. Maria Gonzalez (Gonzalez) was employed in the same division as an Auditor. Each had served efficiently in his or her position for more than one and one-half years, having been originally hired on the basis of oral interviews at a time when the Mayor's office was not subject to the stringent hiring requirements of the City's civil service rules and regulations.

In early 1976, when the City transferred the function of the Training and Job Development Division to the newly created Community Development Department, the senior accountant and auditor positions previously assigned to the Mayor's office became subject to the City's civil service commission. Consequently, appellants were required to pass written examinations before assuming senior accountant and auditor status in the new department. The City Council, in an effort to avoid possible inequities to appellants and others employed in the Mayor's office, proposed an amendment to the city charter exempting the employees from the examination requirement. Unfortunately for appellants, that amendment was defeated by the electorate in November 1976.

Prior to the rejection of the amendment, accountants took and failed the senior accountant examination. Gonzalez passed the auditor examination but scored too low to be employed as an auditor in the new office. Although accountants thus became ineligible for their former senior accountant positions, each of them scored high enough to be employed at his former salary1 as regular civil service labor-market analysts. Gonzalez took a position with the State of California, but maintains her standing in this litigation by asserting a continuing interest in returning to her former auditor position.

Following voter rejection of the exempting amendment, appellants commenced this action in district court, claiming that the City's examinations unlawfully discriminated against Spanish-surnamed applicants. They requested that the district court enjoin the City from using the discriminatory examinations, order the City to develop examinations that accurately predict job performance, and order the City to retain appellants in their former positions until the new examinations are developed.

The three-step inquiry of Title VII actions is well-established, particularly in cases such as this where preemployment screening devices are used.2 At the outset, a plaintiff bears the burden of establishing a prima facie case that the employer's screening device selects employees in a significantly discriminatory pattern. Failure to meet this burden results in judgment for the employer. If the plaintiff succeeds in establishing a prima facie case of discrimination, the burden of proof then shifts to the employer to prove that the screening device is job related, i. e., that it actually measures skills, knowledge, or ability required for successful performance of the job sought by the applicant. Failure in this proof results in judgment for the plaintiff. If the employer succeeds in showing that the screening device does in fact measure job-related characteristics, the screening device does not violate Title VII unless the plaintiff then proves that the employer has available an alternative nondiscriminatory screening device that would effectively measure the capability of job applicants. See Dothard v. Rawlinson, 433 U.S. 321, 329, 97 S. Ct. 2720, 2726, 53 L. Ed. 2d 768 (1977); Albemarle Paper Co. v. Moody, 422 U.S. 405, 425, 95 S. Ct. 2362, 2375, 45 L. Ed. 2d 280 (1975); deLaurier v. San Diego Unified School District, 588 F.2d 674, 676 (9th Cir.1978).

In this case the district judge accepted evidence probative of all three stages of the Title VII inquiry. After doing so, he ruled that appellants had failed to establish a prima facie case of discriminatory impact upon Spanish-surnamed applicants, that the City had proven the senior accountant and auditor examinations to be job related, and that appellants had failed to prove the existence of a non-discriminatory alternative that would select qualified employees with the same degree of accuracy as the written examination. We must first determine whether the district court's finding that appellants failed to present a prima facie case of discrimination is clearly erroneous. Mayor of Philadelphia v. Educational Equality League, 415 U.S. 605, 621 n.20, 94 S. Ct. 1323, 1333-34 n.20, 39 L. Ed. 2d 630 (1974); White v. City of San Diego, 605 F.2d 455, 460 (9th Cir.1979). If it is not, we must affirm the district court's judgment for the defendants. Id. at 459-61.

In attempting to prove that the senior accountant examination had a significantly discriminatory impact on Spanish-surnamed applicants, accountants introduced the following statistical evidence: 5 of 17 or 29.4 percent of the Spanish-surnamed applicants taking the examination passed, while 22 of 40 or 55 percent of the whites who took the examination passed; also, the mean score for Spanish-surnamed applicants was 60.7, 61/2 points below the 67.2 mean score of white applicants. In addition to this statistical evidence, accountants introduced expert testimony that these statistics are reliable evidence of a discriminatory examination and that Spanish-surnamed individuals generally do worse than whites on written examinations.

The district judge, relying on evidence produced by the City, rejected the accountants' proof of a prima facie case for two reason. First, he placed little evidentiary weight on the examination results because the figures, although disparate, were not statistically significant when tested at a .05 level of significance. Second, he concluded that the accountants scored low because they failed to study seriously in preparation for the examination. We will discuss his reasons separately.

In considering the district court's reliance on lack of statistical significance, we observe that the Supreme Court has stated: " '(s)tatistical analyses have served and will continue to serve an important role' in cases in which the existence of discrimination is a disputed issue." International Bhd. of Teamsters v. United States, 431 U.S. 324, 339, 97 S. Ct. 1843, 1856, 52 L. Ed. 2d 396 (1977) ("Teamsters "), quoting Mayor of Philadelphia v. Educational Equality League, supra, 415 U.S. at 620, 94 S. Ct. at 1333. At the same time, however, the Court has cautioned that "(c)onsiderations such as small sample size may, of course, detract from the value of such evidence," Teamsters, supra, 431 U.S. at 340 n.20, 97 S. Ct. at 1857 n.20, and we have stated that the use of statistical evidence "is conditioned by the existence of proper supportive facts and the absence of variables which would undermine the reasonableness of the inference of discrimination which is drawn." 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) (footnote omitted). See also Morita v. Southern California Permanente Medical Group, 541 F.2d 217, 220 (9th Cir.1976), cert. denied, 429 U.S. 1050, 97 S. Ct. 761, 50 L. Ed. 2d 765 (1977) ("statistical evidence derived from an extremely small universe ... has little predictive value and must be disregarded"); Johnson v. Shreveport Garment Co., 422 F. Supp. 526, 539-40 (W.D.La.1976), aff'd, 577 F.2d 1132 (5th Cir.1978).

A total of 17 Spanish-surnamed applicants took the senior accountant examination. The City's expert witness testified that the statistics calculable from such a small sample are not significant at a .05 level of statistical significance,3 the level suggested by federal agency guidelines for the establishment of statistical proof. See e. g., 28 C.F.R. § 50.14, at § 12(b)(5) (1977). Thus, the expert testified, these statistics are not a reliable indicator of discriminatory effect. The unreliability of accountants' statistical base is illustrated by the fact that only three additional passing scores among Spanish-surnamed applicants would have produced a Spanish-surnamed passing rate sufficiently close to the white passing rate to preclude a finding of adverse impact under federal agency guidelines, even if one disregards the .05 level of significance requirement.4 Although accountants produced expert testimony suggesting that the statistical disparity of the examination results is reliable, we recognize that the district judge may accept some statistical inferences and reject others based upon his perception of the oral and documentary evidence placed before him. Mayor of Philadelphia v. Educational Equality League, supra, 415 U.S. at 621 n.20, 94 S. Ct. at 1333-34 n.20; White v. City of San Diego, supra, 605 F.2d at 460-61; Chance v. Board of Examiners, 458 F.2d 1167, 1173 (2d Cir.1972); United States v. Ironworkers Local 86, supra, 443 F.2d at 549. "It is for the District Court, in the first instance, to determine whether these statistics appear sufficiently probative of the ultimate fact in issue whether a given job qualification requirement has a disparate impact on some group protected by Title VII." Dothard v. Rawlinson, supra, 433 U.S. at 338, 97 S. Ct. at 2731 (Rehnquist, J., concurring).

As to the district judge's second reason, we begin with the premise that any statistical evidence produced by a plaintiff is subject to rebuttal by the employer. Id. at 331, 97 S.Ct. at 2727 (majority opinion); Teamsters, supra, 431 U.S. at 340, 97 S. Ct. at 1857. Employers "may endeavor to impeach the reliability of the statistical evidence, they may offer rebutting evidence, or they may disparage in arguments or in briefs the probative weight which the plaintiffs' evidence should be accorded." Dothard v. Rawlinson, supra, 433 U.S. at 338-39, 97 S. Ct. at 2731-32 (Rehnquist, J., concurring). In this case the City attacked accountants' statistical proof by introducing evidence that accountants failed to prepare seriously for the examination, and contended that such evidence further impeached the accuracy of the already dubious small sample size. Facts stipulated by the parties include the statement of a City official that accountants confessed to lack of preparation at a meeting held to air accountants' complaints about the examination. Although each accountant took the witness stand and expressly denied having made such an admission, the district judge did not believe their testimony and we must give "due regard ... to the opportunity of the trial court to judge of the credibility of the witnesses." Fed.R.Civ.P. 52(a). See deLaurier v. San Diego Unified School District, supra, 588 F.2d at 679. The same City official who revealed accountants' admission to their lack of study filed a report describing the meeting at which the admission occurred. That report, prepared several months before accountants filed a complaint in the district court, related plausible reasons for their failure to study: they believed not only that the voters would pass the charter amendment exempting them from the examination requirements, but also that they would receive a second opportunity to take the examination if the amendment failed.

That accountants failed to prepare adequately for the examination is corroborated by their examination performance. Their average examination score, aside from being well below the established passing score, was 11.58 percent below the average score of the other 13 Spanish-surnamed applicants who took the test. Moreover, they scored well below the average Spanish-surnamed score in every examination category.

We conclude that the district court's finding that accountants failed to establish a prima facie case of discrimination was not clearly erroneous. In doing so we recognize the discriminatory implications of the disparate examination statistics and the expert testimony concerning their reliability. On the other hand, the City's expert testimony, the small statistical base, and the impeachment of the statistics by evidence of the accountants' failure to study, all convince us that the district court's conclusion was not an unreasonable interpretation of the evidence presented. "(W)here the evidence would support a conclusion either way but ... the trial court has decided it to weigh more heavily for the defendant(,) (s)uch a choice between two permissible views of the weight of evidence is not 'clearly erroneous.' " United States v. Yellow Cab. Co., 338 U.S. 338, 342, 70 S. Ct. 177, 179, 94 L. Ed. 150 (1949).

Gonzalez also relied upon statistical evidence in attempting to prove the discriminatory impact of the auditor examination: 6 of 23 or 26.1 percent of the Spanish-surnamed applicants passed the examination, while 23 of 36 or 63.9 percent of the white applicants passed. The mean score of Spanish-surnamed applicants, 57.3, was more that 10 points below the 67.4 mean score of white applicants. The district judge expressly found that this evidence indicated a statistical adverse impact on Spanish-surnamed auditor applicants. Although Gonzalez passed the auditor examination, she failed to secure her desired position because she did not score highly enough. She relies upon the disproportionate pass rate to contend that she lost her job through an unlawfully discriminatory examination.

The district judge ruled that Gonzalez did not establish a prima facie case of discriminatory impact. He did so by combining the distribution of scores on the auditor examination with the results of a separate, senior auditor examination, on which Spanish-surnamed applicants performed better than any other ethnic group. Because 75 of the 125 questions on the senior auditor examination also appeared on the 100-question auditor examination, the district judge concluded that the auditor examination could not have discriminated against Spanish-surnamed applicants. The district judge also relied upon the fact that Gonzalez was denied employment not solely on the basis of her written examination score, but also on the basis of an oral interview score.

The district judge erred in concluding that Gonzalez failed to establish a prima facie case of discriminatory impact. The City did not rebut the statistical evidence produced by Gonzalez. Moreover, the district judge's reason for concluding that Gonzalez failed in her proof the combined results of the senior auditor and auditor examinations did not validly refute the clear import of Gonzalez' evidence. The senior auditor results were taken from an extremely small sample size, since only 9 Spanish-surnamed applicants took the test. It was clear error for the district court to conclude that these statistically insignificant results of the senior auditor examination permitted disregard of the statistical results of the auditor examination, particularly when there was no evidence that the Spanish-surnamed senior auditor applicants performed well on the same questions that the Spanish-surnamed auditor applicants failed. Indeed, the City's expert witness testified that it is not sound statistical procedure to combine the results of two separate examinations.

Nor does the fact that Gonzalez was denied employment on the basis of a combined oral interview and written examination score support the district court's ruling that Gonzalez failed in her proof. Gonzalez' written examination score, although high enough to pass, was lower than the average passing score. Her oral interview score, on the other hand, was well above average. Thus, it was her written examination score that most substantially impaired her chances of obtaining the auditor position.

Statistical disparities alone may constitute prima facie proof of discrimination. New York Transit Authority v. Beazer, 440 U.S. 568, 584, 99 S. Ct. 1355, 1365, 59 L. Ed. 2d 587 (1979); Hazelwood School District v. United States, 433 U.S. 299, 307-08, 97 S. Ct. 2736, 2741-42, 53 L. Ed. 2d 768 (1977); League of United Latin American Citizens v. City of Santa Ana, 410 F. Supp. 873, 891 (C.D.Cal.1976). Although the prima facie case requirement is not automatically satisfied by statistical evidence of adverse impact, we hold that a prima facie case is established when such evidence of discriminatory impact is completely uncontroverted. Accordingly, we reverse the district court's ruling that Gonzalez failed to establish prima facie discrimination. We must therefore determine whether the City met its burden of proving that the auditor examination was job related.

A. The Employer's Burden of Proof Under Title VII

The allocation of proof in Title VII cases is well established. Once a plaintiff has met his or her burden of proving a prima facie case of discriminatory impact, the employer bears some burden of justifying the business practice in terms of business need. If this burden is met, the plaintiff then assumes the burden of proving the availability of an effective business alternative with less disparate racial impact.5 Despite this well-established procedure, courts differ on just what an employer must prove to discharge its burden. See Comment, The Business Necessity Defense to Disparate-Impact Liability Under Title VII, 46 U.Chi.L.Rev. 911, 912 (1979). Indeed, cases within this circuit might be read to differ on the weight of the employer's burden. Compare Craig v. County of Los Angeles, 626 F.2d 659 (9th Cir. 1980), and deLaurier v. San Diego Unified School District, 588 F.2d 674 (9th Cir. 1978), with Blake v. City of Los Angeles, 595 F.2d 1367 (9th Cir. 1979), and deLaurier v. San Diego Unified School District, supra, 588 F.2d at 685-92 (Hufstedler, J., dissenting). The question before us is: what must an employer show to meet its burden of proving that pre-employment test, having a disproportionate, adverse impact on a racial minority, are sufficiently justified by business need to survive a Title VII challenge?

In Craig v. County of Los Angeles, supra, we recently concluded that an employer must prove such tests to be "significantly job-related." 626 F.2d at 662. We stated that our standard for proof of job-relatedness was articulated by the Supreme Court:

"(D)iscriminatory tests are impermissible unless shown, by professionally accepted methods, to be 'predictive of or significantly correlated with important elements of work behavior which comprise or are relevant to the job or jobs for which candidates are being evaluated.' 29 C.F.R. § 1607.4(c)."

626 F.2d at 662, quoting Albemarle Paper Co. v. Moody, supra, 422 U.S. at 431, 95 S. Ct. at 2378. We went on to state that an employer "must demonstrate a significant relation between the challenged selection device or criteria and the important elements of the job or training program, not merely some 'rational basis' for the challenged practice. However, the employer need not establish a perfect positive correlation between the selection criteria and the important elements of work." 626 F.2d at 664 (citation and footnote omitted).

The employer's burden in Craig comports with that applied in deLaurier v. San Diego Unified School District, supra, 588 F.2d 674, a case decided almost two years earlier. In deLaurier we applied a standard "commonly referred to as the 'business necessity' or 'job relatedness' defense ...." Id. at 678. Although we stated that the employment practice in question must be shown to be " 'necessary to safe and efficient job performance,' " id. at 678, quoting Dothard v. Rawlinson, supra, 433 U.S. at 331 n. 14, 97 S. Ct. at 2728 n. 14, our application of this language to the facts demonstrated that the employer's burden is met with less than proof of absolute business necessity. The school district in deLaurier attempted to justify its mandatory maternity leave policy with evidence that teaching ability declines as the date of delivery approaches, and that the school district needs advance notice to procure long-term teacher substitutes. 588 F.2d at 678 n. 8. The school district did not thereby prove that its maternity leave policy, which required teachers to stop work at the start of their ninth month of pregnancy, was necessarily required for successful operation of the schools. See id. at 687-91 (Hufstedler, J., dissenting). Nonetheless, we upheld the maternity leave policy as a job-related employment practice. Id. at 678-79. We treated "job related" and "business necessity" as interchangeable terms, neither of which required proof that the challenged policy was absolutely necessary for operation of the business.

Gonzalez refers us to Blake v. City of Los Angeles, supra, 595 F.2d 1367, a case decided six months after deLaurier. Blake might be interpreted as adopting a much more demanding employer burden of proof than that required in deLaurier. In Blake we stated that an employer must prove not only that his screening device is job related; he must further prove it " 'necessary to safe and efficient job performance ....' " Id. at 1376, quoting, again, the phrase from Dothard v. Rawlinson, supra, 433 U.S. at 332 n. 14, 97 S. Ct. 2728 n. 14. Although both deLaurier and Blake quoted this language, it could be argued persuasively that Blake applied it more nearly literally than did deLaurier. deLaurier essentially applied a job-relatedness standard. Blake defined "job-related," quite properly, as "the capacity of selection devices to measure traits that are important to successful job performance." 595 F.2d at 1377. Blake also stated, however, that " 'job-relatedness' is relevant only for the purpose of trying to prove that the characteristics which the various tests select are directly related to the business necessity." Id. This formulation might seem to suggest that business necessity is something over and above job relatedness that is, over and above what is "important to successful job performance." Id. On such a reading, Blake's use of the Dothard expression "necessary" would clearly mean something different from that expression's use in Craig and deLaurier. Thus, in order to determine the test to be applied to the auditor's examination given Gonzalez, we must first harmonize our cases.

Because it is the language from Dothard that is at the root of the issue before us, our effort to harmonize our cases must be guided by the Supreme Court's interpretation of that language. Before turning to the Supreme Court cases, we identify the issue in more detail.

The Craig test, by permitting job-related employment practices, views Title VII, as far as this case is concerned, as prohibiting only race-related employment criteria. The test maximizes employer freedom, restricting it only when employment decisions are made wholly or partially on the basis of race. It mandates employer color-blindness, but otherwise respects an employer's right to seek maximum employee productivity and efficiency. Thus, the Craig/deLaurier test tolerates a disparate impact on racial minorities so long as that impact is only an incidental product of criteria that genuinely predict or significantly correlate with successful job performance, and does not result from criteria that make race a factor in employment decisions.

The interpretation of Blake referred to above, on the other hand, views Title VII as much more restrictive of employer decision-making. That reading would not tolerate a disparate impact on racial minorities that results from job-related criteria. So understood, Blake would allow disparately impacting criteria only when forbidding them would seriously damage the business, that is, when they are "necessary" to operation of the business. Such a test would thus minimize an employer's freedom, permitting him to employ disproportionately-impacting criteria only if he can prove them necessary to the functioning of his enterprise. That such criteria are effective predictors of employee performance is insufficient under this view of Title VII. Thus, such a test would prohibit some pre-employment screening devices permitted by the Craig/deLaurier test: devices that actually predict employee performance, but that cannot be proven necessary to the operation of the business.

In determining whether Title VII requires application of the employer's burden of proof set forth in Craig and deLaurier or some other level of proof, we are not free to invoke our perception of ideal social policy. Rather, we must ascertain Congress' intent in enacting Title VII. Because the face of the statute does not clearly resolve the question before us, we look for guidance to the legislative history of Title VII, United States v. Culbert, 435 U.S. 371, 374 n.4, 98 S. Ct. 1112, 1114 n.4, 55 L. Ed. 2d 349 (1978); Train v. Colorado Pub. Interest Research Group, Inc., 426 U.S. 1, 10, 96 S. Ct. 1938, 1942, 48 L. Ed. 2d 434 (1976), and to interpretative pronouncements by the Supreme Court.

Congress' primary objective in passing Title VII was to eliminate race-related employment criteria. Senators Clark and Case, co-managers of Title VII on the Senate floor, issued an interpretative memorandum defining discrimination:

To discriminate is to make a distinction, to make a difference in treatment or favor, and those distinctions or differences in treatment or favor which are prohibited by (Title VII) are those which are based on any of the five forbidden criteria: race, color, religion, sex, and national origin. Any other criterion or qualification for employment is not affected by this title.

110 Cong.Rec. 7213 (1964). Critics of the legislation were concerned that Title VII would intolerably burden employers and would force them to abandon employee selection practices based on productivity and efficiency. See Griggs v. Duke Power Co., 401 U.S. 424 at 434 & n.10, 91 S. Ct. 849 at 855 & n.10, 28 L. Ed. 2d 158. However, "(p)roponents of Title VII sought throughout the debate to assure the critics that the Act would have no effect on job-related tests." Id. at 434, 91 S.Ct. at 855. Senator Case, in another interpretative memorandum, explained:

Whatever its merits as a socially desirable objective, title VII would not require, and no court could read title VII as requiring, an employer to lower or change the occupational qualifications he sets for his employees simply because proportionately fewer Negroes than whites are able to meet them....

Title VII says merely that a covered employer cannot refuse to hire someone simply because of his color; ... (I)t expressly protects the employer's right to insist that any prospective applicant, Negro or white, must meet the applicable job qualifications. Indeed, the very purpose of title VII is to promote hiring on the basis of job qualifications, rather than on the basis of race or color.

110 Cong.Rec. 7246-47 (1964). Senators Clark and Case assured the critics that "(a)n employer may set his qualifications as high as he likes ...," id. at 7213, and Senator Humphrey stated that "(t)he employer will outline the qualifications to be met for the job. The employer, not the Government will establish the standards." Id. at 13088.

By enacting Title VII, "(d)iscriminatory preference for any group, minority or majority, is precisely and only what Congress has proscribed." Griggs v. Duke Power Co., supra, 401 U.S. at 431, 91 S. Ct. at 853. The legislative history, of Title VII clearly reveals that Congress was concerned about preserving employer freedom, and that it acted to mandate employer color-blindness with as little intrusion into the free enterprise system as possible. See Developments in the Law Employment Discrimination and Title VII of the Civil Rights Act of 1964, 84 Harv.L.Rev. 1109, 1113-19 (1971); Comment, The Business Necessity Defense to Disparate-Impact Liability Under Title VII, 46 U.Chi.L.Rev. 911, 926-30 (1979); Note, Business Necessity under Title VII of the Civil Rights Act of 1964: A No-Alternative Approach, 84 Yale L.J. 98, 102-06 (1974). The question of preferential treatment has recently provided an occasion for the Supreme Court to state its understanding that Title VII "was not intended to 'diminish traditional management prerogatives.' " Texas Department of Community Affairs v. Burdine, --- U.S. ----, ----, 101 S. Ct. 1089, 1096, 67 L. Ed. 2d 207 (1981), quoting United Steelworkers of America v. Weber, 443 U.S. 193, 207, 99 S. Ct. 2721, 2729, 61 L. Ed. 2d 480 (1979).

Therefore, we conclude that the employer's burden of proof required by Craig and deLaurier is more consistent with Congress' Title VII intent than the employer's burden of proof required by the interpretation of Blake suggested above. Our review of Supreme Court case law reinforces this conclusion.

The Supreme Court first considered an employer's duty under Title VII in Griggs v. Duke Power Co., supra, 401 U.S. 424, 91 S.Ct. at 850, and concluded that "Congress has placed on the employer the burden of showing that any given requirement must have a manifest relationship to the employment in question." Id. at 432, 91 S.Ct. at 854. Griggs suggests that the Court perceived "business necessity" to be the same standard as "job-related," and viewed both as requiring only that an employer prove that his employment practices are legitimately related to job performance:

The Act proscribes not only overt discrimination but also practices that are fair in form, but discriminatory in operation. The touchstone is business necessity. If an employment practice which operates to exclude Negroes cannot be shown to be related to job performance, the practice is prohibited.

Id. at 431, 91 S. Ct. at 853.

Four years after the Griggs decision, the Supreme Court answered the very question considered by us today: "What must an employer show to establish that pre-employment tests racially discriminatory in effect, though not in intent, are sufficiently 'job related' to survive challenge under Title VII?" Albemarle Paper Co. v. Moody, supra, 422 U.S. at 408, 95 S. Ct. at 2367. The employer in Albemarle attempted to justify its use of aptitude tests that disproportionately excluded blacks by employing an expert to "validate" the tests in terms of job-relatedness. Various defects in the validation study convinced the Supreme Court that the employer had failed to satisfy its burden of proof. 422 U.S. at 431-36, 95 S.Ct. at 2378-80. Significantly, however, the Court did not interpret Title VII as requiring employer proof of a strong form of "business necessity." Indeed, the Court in Albemarle never used the term "business necessity." Rather, the Court "clarified" the "standard of proof for job relatedness," id. at 436, 95 S.Ct. at 2380, by articulating a standard that is "the same as that of the Griggs case that discriminatory tests are impermissible unless shown, by professionally acceptable methods, to be 'predictive of or significantly correlated with important elements of work behavior which comprise or are relevant to the job or jobs for which the candidates are being evaluated.' " Id. at 431, 95 S. Ct. at 2378, quoting 29 C.F.R. § 1607.4(c). It was this articulation of an employer's Title VII burden that we relied upon in Craig. 626 F.2d at 662.

We now turn to the troublesome Supreme Court language that appears in Dothard v. Rawlinson, supra, 433 U.S. 321, 97 S.Ct. at 2723. Dothard invalidated a height and weight requirement for prison guards that disproportionately excluded women applicants and was not proven to be "job-related." 433 U.S. at 332, 97 S.Ct. at 2728. In doing so, the Court required employer proof identical to that required in its earlier cases: "the employer must meet 'the burden of showing that any given requirement (has) ... a manifest relationship to the employment in question.' " Id. at 329, 97 S.Ct. at 2727, quoting Griggs v. Duke Power Co., supra, 401 U.S. at 432, 91 S. Ct. at 854. Although the holding tracks prior Supreme Court cases, an unnecessary footnote contains the few words that cause our present difficulty: "a discriminatory employment practice must be shown to be necessary to safe and efficient job performance to survive a Title VII challenge." 433 U.S. at 332 n.14, 97 S. Ct. at 2728 n.14. This footnote formulation is belied by the broader standard applied in the Dothard text. See id. at 329, 331-32, 97 S. Ct. at 2726, 2727-28.

Since Dothard, the Court has indicated that the Griggs/Albemarle standard, rather than the Dothard footnote, controls Title VII inquiries. In New York Transit Authority v. Beazer, 440 U.S. 568, 99 S. Ct. 1355, 59 L. Ed. 2d 587 (1979), the plaintiffs challenged a Transit Authority (TA) refusal to hire narcotics users, specifically methadone users. The Court stated:

Respondents recognize, and the findings of the District Court establish, that TA's legitimate employment goals of safety and efficiency require that exclusion of all users of illegal narcotics, barbiturates, and amphetamines, and of a majority of all methadone users. The District Court also held that those goals require the exclusion of all methadone users from the 25% of its positions that are "safety sensitive." Finally, the District Court noted that those goals are significantly served by even if they do not require TA's rule as it applies to all methadone users including those who are seeking employment in nonsafety-sensitive positions. The record thus demonstrates that TA's rule bears a "manifest relationship to the employment in question." Griggs v. Duke Power Co., 401 U.S. 424, 432, 91 S. Ct. 849, 854, 28 L. Ed. 2d 158. See Albemarle Paper Co. v. Moody, 422 U.S. 405, 425, 95 S. Ct. 2362, 2375, 45 L. Ed. 2d 280.

Id. 440 U.S. at 587 n.31, 99 S. Ct. at 1366 n.31 (emphasis added) (citations omitted). Thus, the Court's most recent application of the employer's Title VII burden of proof not only follows the standards set forth in Griggs and Albemarle, but implicitly approves employment practices that significantly serve, but are neither required by nor necessary to, the employer's legitimate business interests.

We conclude that any employer burden of proof that could be suggested by the possible strong reading of Blake 's "business necessity" language would be inconsistent with Supreme Court case law and the Congressional intent underlying Title VII. Accordingly, we reject that interpretation of Blake. We interpret Blake in line with our precedents decided before (deLaurier ) and after (Craig ) that case. We hold that discriminatory tests are impermissible unless shown, by professionally accepted methods, to be predictive of or significantly correlated with important elements of work behavior that comprise or are relevant to the job or jobs for which candidates are being evaluated.6

The district court made two conclusions of law relevant to the validation issue:

4. The Defendants may demonstrate that a test is job related by any competent evidence. Washington v. Davis, 426 U.S. 229 (96 S. Ct. 2040, 48 L. Ed. 2d 597) (1976); see also, Blake v. City of Los Angeles, (435 F. Supp. 55 (C.D. Cal. 1977)).

6. There is no magic in any validating procedure, and the Defendants need only supply competent and relevant evidence upon the issue of the job-relatedness of their employment standards. Washington v. Davis, 426 U.S. 229 (1976) at 247 n. 13 (96 S. Ct. 2040 at 2051 n. 13, 48 L. Ed. 2d 597).

These conclusions would be error if they implied that the evidence need not show that the screening device was validated "by professionally acceptable methods" to be " 'predictive of or significantly correlated with important elements of work behavior which comprise or are relevant to the job or jobs for which candidates are being evaluated.' " Albemarle Paper Co. v. Moody, supra, 422 U.S. at 431, 95 S. Ct. at 2378, quoting 29 C.F.R. § 1607.4(c).

We doubt, however, that the district judge intended anything more than the appropriateness of accepting any competent and relevant evidence tending to establish the existence of a professionally acceptable validation. We so interpret the district judge's conclusion of law because footnote 13 of Washington v. Davis discusses three different standards adopted by the American Psychological Association, and itself cites Albemarle and 29 C.F.R. § 1607.7 The language of the sixth conclusion of law, however, was borrowed from the district court opinion in Blake v. City of Los Angeles, supra, 435 F.Supp. at 65, and there is also a "see" cite to that opinion in the fourth finding. Although the district court opinion in Blake itself quoted from footnote 13 of Washington v. Davis, the district judge in Blake focused on only part of that footnote and took too broad a view of permissible validation procedures. We therefore believe it is necessary to confirm our view that the district judge in the present case was making no similar error. We will do this by reviewing the evidence before the district court in the light of the "professionally accepted methods" validation standard of Albemarle, and the arguments pressed by Gonzalez.

To satisfy this requirement, examinations such as the one used by the City to hire auditors must be " 'validated' in terms of job performance." Washington v. Davis, 426 U.S. 229, at 247, 96 S. Ct. 2040, 2051, 48 L. Ed. 2d 597. Title VII requires no single method of examination validation, but only that the method chosen be professionally acceptable. See Albemarle Paper Co. v. Moody, supra, 422 U.S. at 431, 95 S. Ct. at 2378. To this end, the Equal Employment Opportunity Commission (EEOC) has issued guidelines defining minimum standards for professionally acceptable validation studies. See 29 C.F.R. § 1607.14 (1979). These standards are not mandatory, but they are "entitled to great deference," Griggs v. Duke Power Co., supra, 401 U.S. at 434, 91 S. Ct. at 855, and an employer who disregards them must articulate some cogent reason for doing so and generally bears a heavier than usual burden of proving job relatedness. United States v. City of Chicago, 549 F.2d 415, 430 (7th Cir. 1977); United States v. Georgia Power Co., 474 F.2d 906, 913 (5th Cir. 1973).

In Craig, we established a three-step procedure for validation of examinations used to select employees from among a group of applicants:

The employer must first specify the particular trait or characteristic which the selection device is being used to identify or measure. The employer must then determine that that particular trait or characteristic is an important element of work behavior. Finally, the employer must demonstrate by "professionally acceptable methods" that the selection device is "predictive of or significantly correlated" with the element of work behavior identified in the second step.

Craig v. County of Los Angeles, 626 F.2d 659, 662 (9th Cir. 1980), quoting Albemarle Paper Co. v. Moody, supra, 422 U.S. at 431, 95 S. Ct. at 2378. Applying the Craig procedure to the facts of this case, we conclude that the City successfully validated the auditor examination in terms of job-relatedness.

The City's validation of the auditor examination consisted of two phases: a job-analysis phase and an examination-review phase. In the job-analysis phase, a number of auditors and auditor supervisors employed in various civil service positions throughout the city were organized as a group of job experts for the purpose of determining what skill, knowledge, and ability was essential to the position of auditor. These job experts held four meetings. At the first meeting, they compiled a list of the tasks performed by City-employed auditors. At the second meeting, they determined what skills, knowledge, and ability are required to perform those tasks. At the third meeting, these skills, knowledge, and ability, known as job "elements," were ranked by the job experts on the basis of their importance to the job of auditor. Each job expert ranked the elements himself or herself, without the help of other experts, and the results were averaged to produce a final ranking. From the results of this final ranking, a City personnel analyst compiled a list of elements "critical" to the position of auditor. By definition, these critical elements were the various skills, knowledge, or ability sufficiently essential to the job of auditor to be tested on the auditor examination. At the fourth meeting, the job experts weighted the critical elements according to their relative importance to the auditor position. Again, the weighting was done on an individual basis, the final weighting being an average of all of the job experts' results. Standard deviation analysis was performed on this average, and on the average produced at the third meeting, to protect against one expert skewing the results by extreme rankings or weightings. The final product of these meetings, a compilation of elements critical to the position of auditor, weighted according to their relative importance, was used by the City's examining division to create the 100-question auditor examination.

The examination review phase of the City's validation study occurred after the applicants had taken the examination. In this phase, a new group of job experts was selected from among civil service auditors and supervisors employed by the City. These experts individually reviewed each question and decided if it tested one of the critical elements identified in the first phase. Only if five of the seven job experts agreed that a question tested a critical element was the question considered job related. As a result of this procedure, all but five questions were determined to be job related.

This validation study satisfies our three-step procedure set forth in Craig v. County of Los Angeles, supra, 626 F.2d at 662. First, the initial meetings during the job-analysis phase specified the particular trait or characteristic which was to be measured by the examination. Second, the last two meetings during the job-analysis phase determined which characteristics or traits were important elements of the auditor position. Third, the examination-review phase demonstrated that the auditor examination was significantly correlated with those elements of work behavior identified in the job-analysis phase.

As mentioned earlier, a key requirement of this third step, a requirement essential to proof of job relatedness generally, is that the validation method be professionally acceptable. At trial, the City produced expert testimony that its validation procedures met professional standards. Gonzalez produced expert testimony that the procedures were professionally unacceptable. The district judge resolved this conflict of testimony in favor of the City's expert, not only by ruling that the examinations were job related, but also by resolving every testimonial dispute between these experts in favor of the City. We will not disturb such a credibility determination.8 See United States v. City of Chicago, supra, 549 F.2d at 429-30, 434.

Gonzalez contends that various methodological defects in the City's validation study demonstrated that it was not conducted in a professionally acceptable manner.9 These alleged defects were also the basis of her expert's opinion that the City's validation was unprofessional. Specifically, Gonzalez contends, and her expert testified, that the study did not protect against job-expert bias; did not ensure that the questions tested essential job attributes; did not attempt to eliminate questions that tested elements obtainable through a brief, on-the-job orientation; did not ensure that the level of question difficulty correlated with the level of job difficulty; and did not verify the level of the cutoff score. Methodological defects clearly may reduce the probative value of a validation study. Craig v. County of Los Angeles, supra, 626 F.2d at 664. Therefore, we must consider each of Gonzalez' alleged defects separately.

Gonzalez contends that the City did not protect against job-expert bias, and indeed risked such bias, by selecting experts from among its own employees and by informing the selected experts about the pending Title VII lawsuit. Although accountants attempted to introduce hearsay evidence of actual bias among job experts for the senior accountant examination, Gonzalez introduced no evidence of actual bias among job experts for the auditor examination. The existence of bias was left to inference from the fact that job experts were City employees who knew of the lawsuit. However, the Supreme Court has not found employee involvement in and knowledge of a lawsuit to be fatal defects in a validation study. See Albemarle Paper Co. v. Moody, supra, 422 U.S. at 433 n.32, 95 S. Ct. at 2379. Rather, the Court has cautioned that studies made by such employees "must be examined with great care." Id. The district judge saw no reason to suspect that City employees with knowledge of the lawsuit would be biased against Gonzalez or other Spanish-surnamed applicants. Their jobs were not threatened by the validation study, and there was no evidence that they were biased generally against Spanish-surnamed applicants. The entire validation process, which averaged the conclusions of many job experts and tested that average for unusual deviations, was carefully designed to neutralize the effect of any biases which may have existed. Thus, viewing the study with great care, we find that Gonzalez' first alleged defect does not undercut the study's validity.

Gonzalez next contends that the validation study's "most basic defect" occurred in the examination-review phase. Job experts were asked to determine whether each question tested a skill, knowledge, or ability included in one of the critical elements identified during the job analysis phase. Therefore, Gonzalez contends, the examination-review phase did not ensure that each question tested "essential knowledge, skills, or behaviors composing the job in question," 29 C.F.R. § 1607.5(a) (1977) (emphasis added), as required by EEOC guidelines. We disagree. The critical elements presented to the job experts in the examination-review phase had, in the job-analysis phase, been determined to be skills, knowledge or ability critical or essential to performance of the auditor function. Therefore, by determining that the trait tested by each question was included in a critical element, the experts in the examination-review phase determined that the question tested a trait essential to the position of auditor. Moreover, before a question would be validated as job-related, five of seven job experts in the examination-review phase had to agree that it tested a critical element.10

Gonzalez next alleges that the validation study was defective because it failed to identify questions that tested skills, knowledge, or ability obtainable through a brief, on-the-job orientation, as required by 29 C.F.R. § 1607.5(a) (1977). Again we disagree. When experts in the job-analysis phase identified critical elements of the auditor position, they selected only those traits that a "barely acceptable worker" at the "minimum level of proficiency" would need "on the first day on the job." Thus, the selected critical elements did not include skills, knowledge, or ability learned on the job. That the job experts of the examination-review phase were not expressly asked to determine whether the trait tested by each question was obtainable through a brief orientation, is irrelevant. By determining that each question tested traits included in a critical element, the job experts effectively determined that the skills, knowledge, or ability tested were those needed by minimally proficient employees at the commencement of their employment as auditors. Gonzalez does not refute this reasoning.

Gonzalez next contends that the examination-review phase failed to determine whether the difficulty level of the examination questions equaled the difficulty level of the auditor's position, as required by some cases. See, e. g., Kirkland v. Department of Correctional Services, 374 F. Supp. 1361, 1372 (S.D.N.Y.1974), aff'd in part and rev'd in part on other grounds, 520 F.2d 420 (2d Cir. 1975), cert. denied, 429 U.S. 823, 97 S. Ct. 73, 50 L. Ed. 2d 84 (1976). Even if we accepted this added requirement, Gonzalez would not be helped. As mentioned in the above discussion, those who identified the critical elements of the auditor position did so for the "barely acceptable worker" who functioned at a "minimum level of proficiency." Thus, the critical elements to which the questions were matched represented the requisite level of difficulty for success as an auditor. Moreover, the nature of the examination minimizes the importance of the level-of-difficulty equation. As the City's expert testified, the purpose of a civil service examination is not to identify all individuals who could function competently in the available job. Rather, the civil service examination, as required by the City charter, ranks the applicants so that only the best qualified are hired. The statistics of the auditor examination demonstrate the need for such ranking: 188 applicants took the examination in hopes of obtaining positions for which only 16 individuals had been hired in the previous two years combined.

Finally, Gonzalez contends that the examination passing score of 65 was arbitrarily set by the City charter, with no analysis of whether that score actually excluded some qualified applicants. We have two responses. First, Gonzalez passed the examination. Thus, a lower cutoff score would have had no effect on whether she was hired. Second, as discussed above, the nature of a civil service examination minimizes the significance of the cutoff score. Where an examination is designed to rank applicants so that only the top few may be hired (Gonzalez ranked 12th overall and she was not hired), the cutoff score is more a formality than a matter of consequence. Those who failed the examination would not have been hired even if they had passed by virtue of a lower cutoff score.

In summary, we conclude that the City's validation study satisfied our previously established three-step criteria for validations of pre-employment examinations. That the validation was professionally acceptable is demonstrated by expert testimony in the record and the failure of each of Gonzalez' challenges. The City established, by professionally acceptable standards, that the auditor examination was predictive of or significantly correlated with important elements of the auditor position. Thus, we must finally consider whether Gonzalez established the existence of a less discriminatory alternative.

"If an employer (meets) the burden of proving that its tests are 'job related,' it remains open to the complaining party to show that other tests or selection devices, without a similarly undesirable racial effect, would also serve the employer's legitimate interest in 'efficient and trustworthy workmanship.' " Albemarle Paper Co. v. Moody, supra, 422 U.S. at 425, 95 S. Ct. at 2375, quoting McDonnell-Douglas Corp. v. Green, 411 U.S. 792 at 801, 93 S. Ct. 1817, 1823, 36 L. Ed. 2d 668. Gonzalez contends that the auditor examination, even if job related, violates Title VII because the City could have effectively screened auditor applicants by use of less discriminatory oral interviews. In making this contention, Gonzalez cites extensively from Crockett v. Green, 388 F. Supp. 912, 919-21 (E.D.Wis.1975), aff'd, 534 F.2d 715 (7th Cir. 1976), a case which places upon the employer the burden of proving that no alternative screening device was available. The district court opinion in Crockett, which was rendered before the Supreme Court's decision in Albemarle Paper Co. v. Moody, supra, 422 U.S. 405, 95 S. Ct. 2362, 45 L. Ed. 2d 280, clearly misallocates the burden of proof. Once an employer has shown a selection device to be job related, it becomes the plaintiff's responsibility to prove that a less discriminatory alternative would satisfy the employer's hiring needs. Albemarle Paper Co. v. Moody, supra, 422 U.S. at 425, 95 S. Ct. at 2375; McDonnell-Douglas Corp. v. Green, supra, 411 U.S. at 804, 93 S. Ct. at 1825; deLaurier v. San Diego Unified School District, supra, 588 F.2d 674 at 676. With the burden of proof properly allocated in this case, we agree with the district court's conclusion that Gonzalez failed to prove the existence of less-discriminatory screening devices that would have satisfied the City's civil service hiring needs.

The only evidence produced by Gonzalez to establish the existence of alternative auditor selection methods was expert testimony that Spanish-surnamed individuals generally do better in oral interviews than on written examinations and that oral examinations could be used to screen applicants. However, even if the district court accepted this testimony as establishing that oral interviews have a less disparate impact on minorities, it does not satisfy Gonzalez' burden of proving that oral interviews, as an alternative to written examinations, would satisfy the City's civil service hiring needs.11 That oral interviews were used in the past to hire auditors for the Mayor's office, a fact upon which Gonzalez relies, does not prove that such interviews would satisfy the merit hiring requirements of the new civil service division. Evidence in the record reveals that the oral interviews previously used were designed to meet the special needs of the Mayor's office, not those of a civil service classification. Moreover, that oral interviews were used to screen applicants in other job categories transferred out of the Mayor's office at the same time that the auditor position was transferred, another fact upon which Gonzalez relies, does not prove that such interviews would have satisfied the auditor-hiring needs of the new department. As the district court correctly concluded, Gonzalez did not prove that a less discriminatory alternative was available.

In summary, we conclude that accountants failed to establish a prima facie case of discriminatory impact by the senior accountant examination. Gonzalez, although successfully establishing a prima facie case of discriminatory impact by the auditor examination, failed to prove that a less discriminatory alternative was available to the City. Thus, the City's proof that its auditor examination was job related entitles it to judgment. Accordingly, we affirm the district court's decision that appellants are entitled to no relief under Title VII.

AFFIRMED.

TANG, Circuit Judge, specially concurring in part and dissenting in part.

* I write separately for two reasons. First, although I agree with the majority that the Title VII disparate impact business necessity defense requires only that the employer "demonstrate a significant relation between the challenged selection device or criteria and the important elements of the job or training program ...," ante, at 1277-1279, quoting Craig v. County of Los Angeles, 626 F.2d 659, 664 (9th Cir. 1980), I cannot sanction its assertion that we must "harmonize" Blake v. City of Los Angeles, 595 F.2d 1367 (9th Cir. 1979) cert. denied, 446 U.S. 928, 100 S. Ct. 1865, 64 L. Ed. 2d 281 (1980), with our other precedents before determining which test to apply to the auditor's examination. See ante, at 1276-1277. Until today, there was no confusion in this circuit concerning the appropriate test to apply in the business necessity defense. Second, I dissent from the majority's conclusion that the district court properly found the City's evidence sufficient to meet its burden on the business necessity defense. See ante, at 1282-1283. Because the lower court applied the wrong legal standard at the business necessity stage of the proof process, its findings cannot be reviewed under the clearly erroneous standard. Therefore, rather than deferring to the lower court's findings, I would remand the case for a reevaluation and an introduction of evidence in light of the proper legal standard.

The question raised in the majority opinion is whether we must first harmonize our ruling in Blake v. City of Los Angeles, 595 F.2d 1367 (9th Cir. 1979), with our other business necessity defense precedents before determining the proper test to be applied to the auditor's examination given Gonzalez. See ante, at ----. Having closely examined our treatment of the business necessity defense, both as a member of the Blake panel and, most recently, as a member of this panel, I have concluded that Blake fully conforms to the business necessity defense articulated in de Laurier and unmistakably rejects the inconsistent language in the de Laurier dissent. Because Blake is fully consonant with de Laurier, the present attempt to "harmonize" it with our other precedents not only wastes judicial resources but imparts needless confusion into an area of the law that had previously been more than adequately clear.

In de Laurier v. San Diego Unified School District, 588 F.2d 674, 678 (9th Cir. 1978), we decided that in order to rebut a prima facie case of employment discrimination under Title VII an employer must show " 'that the practice in question is necessary to safe and efficient job performance.' " Id., quoting Dothard v. Rawlinson, 433 U.S. 321, 331 n.14, 97 S. Ct. 2720, 2728 n.14, 53 L. Ed. 2d 768 (1977). In establishing this standard, we rejected the additional requirement, proposed in Judge Hufstedler's dissent, that the employer substantiate the nonexistence of less discriminatory alternatives. Compare de Laurier v. San Diego Unified School District, 588 F.2d at 681, with id. at 691 (Hufstedler, J., dissenting). The de Laurier majority, consistent with the Supreme Court's determinations in Dothard and Albemarle v. Moody, 422 U.S. 405, 425, 95 S. Ct. 2362, 2375, 45 L. Ed. 2d 280 (1975), placed the burden of proving less discriminatory alternatives squarely on the plaintiff. Consequently, once the defendant has rebutted the plaintiffs' prima facie case of discriminatory impact by demonstrating that the test device or rule is "necessary to safe and efficient job performance," the plaintiff shoulders the burden of establishing less discriminatory alternatives to the employment practice. de Laurier v. San Diego Unified School District, 588 F.2d at 681.

The majority's argument that Blake departs from the de Laurier standard can be reduced to the following proposition: Although the de Laurier business necessity defense requires that an employer demonstrate that the employment practice is " 'necessary to safe and efficient job performance ...' ", 595 F.2d at 1376, quoting Dothard v. Rawlinson, 433 U.S. at 332 n.14, 97 S. Ct. at 2728 n.14, both the de Laurier dissent and the Blake decision literally construe the word "necessary", thereby erroneously requiring an absolutely necessary relationship between the employment practice and the business need. See ante, at 1276-1277. In response, I maintain that an examination of Blake and de Laurier, together with a comparison of case law from other circuits, establishes that Blake, like de Laurier, requires only reasonable necessity. Moreover, Blake, unlike the de Laurier dissent, correctly assigned the burden of proving less discriminatory alternatives to the plaintiff, thereby underscoring its consistency with de Laurier. I shall address these points seriatim.

(1)

The Requirements that the Job Practice Be "Necessary"

The compatibility between the Blake and de Laurier definitions of the word "necessary" can be demonstrated by an analysis of those cases. In de Laurier, a school district's mandatory maternity leave policy was found " 'necessary to safe and efficient job performance'," 588 F.2d at 678, quoting Dothard v. Rawlinson, 433 U.S. at 331 n.14, 97 S. Ct. at 2728 n.14, even though the court below had not phrased the standard in those precise terms. Id. at 678-79. The dissent in de Laurier disagreed with the majority assessment. In its view, the "district failed to offer evidence in specific justification of the job-relatedness of its employment practice." Id. at 687. (Hufstedler, J., dissenting). As such, the distinction between the de Laurier majority and dissent on the definition of "necessary" merely reflects a disagreement over the district court's application of the business need standard to the facts of the case. The degree of necessity required was probably never at issue. Compare id. at 678 (practice must be necessary to safe and efficient job performance), with id. at 688 (Hufstedler, J., dissenting). The dissent did not disagree with the formulation of the standard; rather, it argued that on the facts presented, the defendant had not established the existence of a specific relationship between the employer practice and the business need. Id. at 688 (Hufstedler, J., dissenting). Neither the de Laurier majority nor the dissent required a demonstration of an "absolutely necessary" relationship between the employment practice and the business need.

Blake preserved the standard announced in de Laurier and expressly recognized that the term "necessary" meant nothing more than a "manifest relationship" between the job practice and business need. Blake, 595 F.2d at 1376, quoting Griggs v. Duke Power Co., 401 U.S. 424, 432, 91 S. Ct. 849, 854, 28 L. Ed. 2d 158 (1971). That Blake employed a "reasonably necessary" standard,1 rather than an "absolutely necessary" standard, is evident from its various descriptions of the test:

"(J)ob relatedness is relevant only for the purpose of trying to prove that the characteristics which the tests select are directly related to the business necessity."

Id. at 1377.

To justify the use of pre-employment selection devices as a business necessity, an employer must show that the tests are so closely job related that their use is "necessary to safe and efficient job performance." (Dothard, supra, 433 U.S. at 332 n.14, 97 S. Ct. at 2728).

Id. at 1378.

In Albemarle, the Supreme Court explained the legal standard for determining the job relatedness of selection devices: "discriminatory tests are impermissible unless shown by professionally acceptable methods to be 'predictive of or significantly correlated with important elements of work behavior which comprise or are relevant to the job or jobs for which candidates are being evaluated.' "

Id. (citation omitted).

Appellees failed to establish that the LAPD's selection devices were so closely job related that their use was justified by business necessity. At most, they showed that there was some rational relationship between their selection devices and certain elements of job performance by police.

Id. at 1379.

The Blake court did not demand an "absolutely necessary" relationship; rather it required a close or direct relationship. Blake did not, as the majority implies, ignore the Supreme Court's Albemarle decision; it expressly relied upon Albemarle in formulating the job relatedness standard. See id. at 1378, citing Albemarle v. Moody, supra, 422 U.S. at 431, 95 S. Ct. at 2378. Although Blake assertedly relied upon dictum in Dothard, ante, at 1279-1282, the language cited in Blake is probably a holding.2 At any rate, because de Laurier expressly adopted the same formulation in construing and applying the business necessity defense, the majority's criticism is patently underinclusive. See de Laurier, 588 F.2d at 678, quoting Dothard v. Rawlinson, 433 U.S. at 331 n.14, 97 S. Ct. at 2728 n.14.

In establishing the perceived distinction between the Blake and de Laurier standards, the majority points to a specific passage from Blake and argues than an "inordinate focus on (it) ... might seem to suggest that business necessity is something over and above job relatedness." Ante, at 1276-1277, quoting Blake, 595 F.2d at 1377. The Blake assertion is entirely consistent with the de Laurier test: "(J)ob relatedness is relevant only for the purpose of trying to prove that the characteristics which the various tests select are directly related to the business necessity."3 Blake, 595 F.2d at 1377 (emphasis added). An "inordinate focus" on the sentence as written in Blake would yield nothing that is inconsistent with our prior or subsequent precedents.

Moreover, the application of the business necessity defense in Blake fully comported with de Laurier. Blake offered two holdings on the business necessity defense issue and neither even remotely suggests that an absolute necessity is required. First, the court held that, when an employment practice creates a disparate impact, a defendant's good faith is immaterial to the business necessity defense. Id. at 1376-77. See Griggs v. Duke Power Co., 401 U.S. 424, 432, 91 S. Ct. 849, 854, 28 L. Ed. 2d 158 (1971). Second, Blake found that the trial court applied an incorrect business necessity standard. Because the district court had based the award of summary judgment upon defense affidavits that " 'for the most part, merely point(ed) out what (was) obvious and need(ed) little proof,' " id. at 1378, we concluded that the trial court had exacted only a rational relationship between the height requirement and police duties, a standard far below that required by the Albemarle decision. Blake, 595 F.2d at 1378. See Albemarle, 422 U.S. at 431, 95 S. Ct. at 2378.

Blake 's retention of the de Laurier business necessity defense and its rejection of the absolute necessity standard is perhaps best illustrated by comparing the language in Blake with the wording used in cases that do require an absolutely necessary relationship between the job practice and the business need. When courts require strict necessity, specific, unmistakable language is employed: "Necessity connotes an irresistible demand .... (A practice) must not only directly foster safety and efficiency ..., but also must be essential to those goals." United States v. Bethlehem Steel Corporation, 446 F.2d 652, 662 (2nd Cir. 1971) (emphasis added). See, e. g., Kirby v. Colony Furniture Co., Inc., 613 F.2d 696, 705 n.6 (8th Cir. 1980); Green v. Missouri Pacific Railroad Company, 523 F.2d 1290, 1295-98 (8th Cir. 1975); Rock v. Norfolk & W. Ry. Co., 473 F.2d 1344, 1349 (4th Cir.), cert. denied 412 U.S. 933, 93 S. Ct. 2754, 37 L. Ed. 2d 161 (1973); United States v. St. Louis-San Francisco Ry. Co., 464 F.2d 301, 308 (8th Cir.) (en banc ), cert. denied, 409 U.S. 1107, 93 S. Ct. 900, 34 L. Ed. 2d 687 (1972); United States v. Jacksonville Terminal Co., 451 F.2d 418, 451 (5th Cir. 1971), cert. denied, 406 U.S. 906, 92 S. Ct. 1607, 31 L. Ed. 2d 815 (1972). Similar language is simply not present in Blake. On the contrary, the court in Blake specifically required that the practice be "directly related to the business necessity." Blake, 595 F.2d at 1377. By demanding that the practice do more than "directly foster" the business need, the circuit cases cited above expressly reject the "directly related" standard established in Blake. Hence, Blake 's language is fundamentally irreconcilable with decisions that require an absolutely necessary relationship.

(2)

Allocation of Burden of Proof Regarding Less Discriminatory

It might be argued that, if Blake did require absolute necessity between the employment practice and business need, it would subtly shift the plaintiff's burden of proving the availability of less discriminatory alternatives into the defendant's business necessity defense because no less discriminatory alternatives could exist if the practice were absolutely necessary to the business need. It could be argued further that, by implicitly requiring proof of the absence of less discriminatory alternatives, the employer is burdened with an unduly onerous task: proof of a negative.4 Indeed, these concerns probably constitute the unarticulated predicate to the majority's perception of intracircuit discord.

The court in Robinson v. Lorillard Corporation, 444 F.2d 791, 798 n.7 (4th Cir. 1971), made these same arguments yet the question underlying the assertions whether proof of an absolutely necessary relationship between the job practice and business need subsumes the proof of alternatives issue was neither explored nor proven. Although there may be an immediate intuitive appeal in the Robinson view, the more probable interpretation of the language in Dothard suggests that the Court's definition of "necessary" excludes any consideration of less discriminatory alternatives in the employer's burden of proof.

If we assume that the words "necessary to safe and efficient job performance" refer to the means/end relationship between the job practice and business need (an assumption the majority and I seem to agree upon, see ante, at 1274-1277), the Dothard Court appears to have required simply that the means the job practice efficiently maximize the end the business need. Since the end in issue at this stage of the proof process, i. e., the employer's business purpose, is economic and not necessarily coextensive with interests sought to be protected by less discriminatory alternatives, see ante, at 1277-1279 (legitimacy of employer goals of efficiency and productivity), it is entirely possible to find an optimal or "necessary" advancement of business need, thereby fulfilling the employer's burden, and a corresponding substantial infringement upon employee rights. See Albemarle Paper Co. v. Moody, 422 U.S. at 431-32, 95 S.Ct. at 2378, (test may significantly predict work behavior and still be discriminatory). Consequently, the distinction between an absolutely necessary relationship and a reasonably necessary relationship, rather than indicating the degree to which a less discriminatory alternatives analysis is implicated in the business necessity defense, is exclusively concerned with the measure of efficiency between the job practice and business need. The existence or nonexistence of less discriminatory alternatives is neither definitionally nor operationally material to an analysis that solely concerns itself with the substantiality of this means/end relationship.

Although an examination of less discriminatory alternatives could overlap with proof offered at the business necessity stage, the overlay would in no sense be an inevitable product of requiring a "necessary" relationship between the job practice and business need. Indeed, the Court suggested as much in Dothard when it asserted that the employer's business need, strong prison guards, "could be achieved by (alternatively) adopting and validating a test for applicants that measures strength directly. Such a test, fairly administered, would fully satisfy the standards of Title VII ...." Dothard v. Rawlinson, 433 U.S. at 332, 97 S. Ct. at 2728 (footnote omitted) (emphasis added). The Court offered this alternative test as evidence of the relative inefficiency of testing employed by the institution, not as evidence of the existence of less discriminatory alternatives. This latter concern is parenthetically indicated by the requirement that the strength test be "fairly administered." Proof of fair administration would entail an examination of less discriminatory alternatives to the optimally efficient business need test and, therefore, constitute an element of the plaintiff's burden.

Several cases seem to support the propositions that: (1) by focusing solely upon the degree of efficiency between job practice and business need, the "necessary" language in Dothard excludes consideration of less discriminatory alternatives as part of the employer's proof; and (2) any discussion of alternatives during judicial review of the employer's burden of proof centers on more efficient alternatives, and not, necessarily, on less discriminatory alternatives. See Albemarle Paper Co. v. Moody, 422 U.S. at 432, 95 S.Ct. at 2378 (study's failure to analyze particular skills needed in job precluded a finding of significant correlation); Craig v. County of Los Angeles, 626 F.2d at 668 (county's test failed to consider alternative fact that, by virtue of cultural rapport, Mexican-American police officer would be more efficient in dealing with Mexican-American constituency); Blake, 595 F.2d at 1376 (fact that police department functioned successfully without previous regulation indicated that the rule was not necessary to efficient operation).

Of course, the more obvious evidence that the Court in Dothard did not intend that the word "necessary" subsume consideration of less discriminatory alternatives is the likelihood that the Court would not, within a period of only two years, implicitly contradict the allocation scheme announced in Albemarle. Therefore, the predicate to the possible criticism of Blake discussed here that the word "necessary" effects a subtle shift in burdens must be regarded with some caution.

Even if we work under the assumption that an "absolutely necessary" standard shifts the burden of proving alternatives,5 however, it is clear that Blake does not employ the subtle technique attributed to it by the majority. First, as is demonstrated in the preceding section Blake does not require absolute necessity. Second, as shown below, Blake implicitly acknowledged the danger of a sub silentio shift in burdens by expressly segregating the "alternatives" discussion from its business necessity analysis. Finally, Blake is directly contrary to the de Laurier dissent and entirely consistent with the de Laurier majority on the issue of proving alternatives.

Quite apart from its content, the organization of the Blake opinion indicates that the court followed the precedent of de Laurier and Albemarle. The discussions of the business necessity defense and the plaintiff's proof of less discriminatory alternatives are plainly demarcated by separate subtitles. Blake, 595 F.2d at 1375, 1383. By structurally dissociating these two burdens, Blake implies that, under a reasonable necessity standard, the plaintiff's burden is not subtly shifted into the business necessity defense. Cases that require absolute necessity stand in direct contrast.6

The comparison between Blake and the de Laurier dissent is most important on the issue of proof of less discriminatory alternatives: Blake directly conflicts with the allocation advocated in the de Laurier dissent. In discussing the issue of alternatives, the de Laurier dissent, unlike Blake, drastically blurred and probably ignored the distinction between the business necessity defense and the plaintiff's burden of showing less discriminatory alternatives:

Even if the district court's findings were unavailable, the business necessity conclusion cannot be sustained because the district court completely failed to consider the existence of less burdensome alternatives, such as individualized decision-making, that would serve the legitimate interests of the school district as well as the mandatory leave policy. The evidence presented by both parties, and the court's own findings of fact, strongly suggest the availability of such alternatives, and Dothard requires that they be considered in establishing the business necessity defense.

de Laurier, 588 F.2d at 691 (Hufstedler, J., dissenting) (emphasis added). Although the de Laurier dissent is wrong in its construction of Dothard, our subsequent reliance on Dothard in Blake was fortified by the correct interpretation and an express citation to Albemarle : "Even if an employer meets his burden of demonstrating business necessity, Title VII plaintiffs may prevail if they show that alternative selection devices are available that would serve the employer's legitimate interests without discriminatory effects." (Albemarle, supra, 422 U.S. at 425, 95 S.Ct. at 2375). Blake, 595 F.2d at 1367. Thus, Blake unmistakably rejects the de Laurier dissent's analysis on the proof of alternatives issue. Contrary to the de Laurier dissent, Blake assigns to the plaintiff the burden of demonstrating less burdensome alternatives, thereby correctly applying the de Laurier majority standard.

An analysis of Blake and a comparison between it and the de Laurier majority indicates that the same degree of necessity is demanded in both cases reasonable necessity. Moreover, those circuits that have adopted a strict or absolute necessity approach to the business necessity defense expressly indicate the use of the more exacting standard by employing words such as "irresistible" and "essential" in expressing the degree of necessity required and discard the "directly foster" standard as too lenient. Blake, on the other hand, adopts a "directly related to" standard that is substantively indistinguishable from the "directly foster" test. Thus, Blake eschews the more demanding tests of other circuits and, consistent with de Laurier, adopts an intermediate standard requiring more than a rational relationship but less than an absolutely necessary connection. This standard fully comports with our decisions, including that announced today by the majority. See ante, at 1275-1277 (requiring more than a rational relationship).

An examination of the Blake decision also contradicts any claim that it either sanctions or encourages a sub silentio shift in the burden of showing the availability of less burdensome alternatives. By requiring less than absolute necessity and by separating the discussion and application of the business necessity defense from the analysis of less burdensome alternatives, Blake precludes any subtle shift in burden of proof, even if such a shift were possible. Moreover, Blake clearly disavows the de Laurier dissent's allocation of proof on the alternatives issue. Blake, therefore, correctly observes the de Laurier majority's precedent on the issue of proving less discriminatory alternatives.

There is no intracircuit discord over the business necessity defense. The Craig decision, a purported foil to the Blake analysis, see ante, at 1276-1277, directly cites the Blake formulation of the business necessity defense, 626 F.2d at 662 (citing with approval the page at which Blake quotes the contested Dothard language), and characterizes the job relatedness standard as "rigorous" immediately after praising the "instructive" benefits of Blake. 626 F.2d at 666. Compare Blake, 595 F.2d at 1377 (business necessity defense is narrow), with Craig, 626 F.2d at 656 (job relatedness standards are rigorous). Moreover, in Harriss v. Pan American World Airways, Inc., 649 F.2d 670 at 675 (9th Cir. as corrected Dec. 30, 1980), we again affirmed Blake's formulation of the business necessity defense and quoted Dothard and Blake for the proposition that "a discriminatory employment practice must be shown to be necessary to safe and efficient job performance." Harriss, slip op. at 675. Indeed, Harriss relied on Blake in order to hold that the defendant's evidence was "sufficiently compelling." Id. at 675. The majority opinion today, not Craig, is the first decision to suggest that there might be a conflict in the development of this circuit's business necessity standards. As a matter of precedent, therefore, Craig has validated Blake's status in this circuit.7

The majority concedes that the district court committed legal error by requiring only that the City supply "competent and relevant evidence" on the issue of the job relatedness of their employment standards, ante, at 1279-1282, yet it concludes that despite this error the lower court's determination that the City met its burden of establishing a business necessity must be upheld on appeal. I cannot share this view. Accordingly, I respectfully dissent from this holding; rather than affirm the district court I would remand the case with instructions to reevaluate the evidence under the proper legal standard.

A district court's findings of subsidiary fact in Title VII disparate impact cases must normally be reviewed by an appellate court under the clearly erroneous standard. See Harriss v. Pan American World Airways, Inc., 637 F.2d 1297 (9th Cir. 1980). We would, therefore, usually accord great deference to the lower court's factual findings on the issue of business necessity defense and refrain from substituting our judgment for that of the district court.

When a district court's factual findings are induced by an erroneous conception of the law, however, we are not free to indulge in a deferential review. On the contrary, we must discard the clearly erroneous standard and independently review the record because it has long been acknowledged that a lower court's legal misconception can directly infect its findings. See United States v. Singer Mfg. Co., 374 U.S. 174, 194 n.9, 83 S. Ct. 1773, 1784 n.9, 10 L. Ed. 2d 823 (1963); United States v. Gypsum Co., 333 U.S. 364, 394, 68 S. Ct. 525, 541, 92 L. Ed. 746 (1948); Parson v. Kaiser Aluminum & Chemical Corp., 575 F.2d 1374, 1382 (5th Cir. 1978), cert. denied, 441 U.S. 968, 99 S. Ct. 2417, 60 L. Ed. 2d 1073 (1979); James v. Stockham Vales & Fittings Co., 559 F.2d 310, 314 n.1 (5th Cir. 1977), cert. denied, 434 U.S. 1034, 98 S. Ct. 767, 54 L. Ed. 2d 781 (1978); Traveler's Indemnity Co. v. United States, 543 F.2d 71, 73 (9th Cir. 1976); Ritter v. Morton, 513 F.2d 942, 949 (9th Cir.) cert. denied, 423 U.S. 947, 96 S. Ct. 362, 46 L. Ed. 2d 281 (1975)8; People of State of California State Lands Comm'n v. Simon, 504 F.2d 430, 440 & n.33 (Em.App.), cert. denied, 419 U.S. 1021, 95 S. Ct. 496, 42 L. Ed. 2d 294 (1974). As a consequence, we must assume that the district court's findings in this case were infected by its erroneous understanding of the business necessity defense and independently review those findings.

The majority opinion not only fails to acknowledge this principle but blatantly states that we can deferentially review the lower court's findings:

At trial, the City produced expert testimony that its validation procedures met professional standards. Gonzales produced expert testimony that the procedures were professionally unacceptable. The district judge resolved this conflict of testimony in favor of the City's expert, not only by ruling that the examinations were job related, but also by resolving every testimonial dispute between these experts in favor of the City. We will not disturb such a credibility determination.

Ante, at 1282.

This approach is an obvious application of the clearly erroneous standard and cannot be justified under the circumstances here presented. The district judge's conclusion that the examinations were job related is entirely without conclusive force because it rendered that determination under the incorrect assumption that job relatedness could be established and the defendant's burden met with the introduction of evidence that was merely "relevant" or "competent". Similarly, its determination that the City's expert was more credible cannot be insulated by a sympathetic appellate review because that conclusion was directly engendered by its erroneous assumption that the City could satisfy its burden of proof on the business necessity defense with the introduction "relevant" or "competent" evidence.

Although the normal course in such a case would be an appellate de novo review of the district court's findings of fact, in view of the complexity of the evidence received and the possibility that the party's presentation of evidence was inhibited by the lower court's legal error, such an examination would more properly proceed at the district court level. I would, therefore, remand the case for an introduction and evaluation of evidence and a redetermination on the issue whether the City met its burden of proving job relatedness and validation.