WILLIAM FRENCH SMITH, ATTORNEY GENERAL, ET AL., PETITIONERS V. HENRY W. SEGAR, ET AL. No. 84-1260 In the Supreme Court of the United States October Term, 1984 The Solicitor General, on behalf of the Attorney General of the United States and the Administrator of the Drug Enforcement Administration (DEA), petitions for a writ of certiorari to review the judgment of the United States Court of Appeals for the District of Columbia Circuit in this case. Petition for a Writ of Certiorari to the United States Court of Appeals for the District of Columbia Circuit PARTIES TO THE PROCEEDING In addition to the parties named in the caption, Francis M. Mullen, Jr., the Administrator of the Drug Enforcement Administration, is a petitioner; Morris H. Davis is, with Henry Segar, a named respondent representing a class of similarly situated black special agents. A third party, Black Employees Against Discrmination In DEA, was dismissed as a plaintiff by stipulation when the plaintiff class was certified. TABLE OF CONTENTS Opinions below Jurisdiction Statute involved Statement Reasons for granting the petition Conclusion OPINIONS BELOW The opinion of the court of appeals (App., 1a-98a, infra) is reported at 738 F.2d 1249. The opinion of the district court (App., 99a-147a, infra) is reported at 508 F. Supp. 690. JURISDICTION The judgment of the court of appeals (App., 148a-149a, infra) was entered on June 22, 1984. Rehearing was denied on August 29, 1984 (App., 150a-154a, infra). On November 19, 1984, the Chief Justice extended the time for filing a petition for a writ of certiorari to and including January 26, 1985. The jurisdiction of this Court is invoked under 28 U.S.C. 1254(1). STATUTE INVOLVED Section 717(a) of Title VII of the Civil Rights Act of 1964, 42 U.S.C. 2000e-16(a), provides in pertinent part: All personnel actions affecting employees or applicants for employment * * * in executive agencies as defined in section 105 of title 5, * * * shall be made free from any discrimination based on race, color, religion, sex, or national origin. QUESTIONS PRESENTED 1. Whether an employer-defendant in a class-wide Title VII disparate treatment case who articulates a specific, reasonable employment practice to rebut plaintiffs' prima facie case of purposeful discrimination must prove the business necessity of that practice, even though plaintiffs have not alleged that the practice violates Title VII under a disparate impact theory. 2. Whether a class-wide claim of disparate treatment is established by statistical evidence that ignores published, objective civil service requirements for the positions in question. STATEMENT 1. A class of black special agents of DEA (including former agents and applicants for such positions) brought this action alleging violations of Title VII. /1/ They alleged a variety of discriminatory employment practices, including claims regarding initial grade level assignment, promotions, and salary determinations. /2/ App., 99a-100a, infra. At trial, respondents attempted to prove discrimination through the anecdotal testimony of ten witnesses and statistical evidence. Both courts below found the anecdotal testimony unpersuasive. Respondents' case therefore rested on the strength of their statistical evidence -- specifically, their multiple regression analysis. A multiple regression analysis is a statistical technique designed to estimate the effects of several independent variables on a given result. Properly used, such an analysis can estimate how much influence variables such as race, experience, and education had on perceived disparities in salary level. The independent variables included in respondents' study were: (1) education; (2) years of work experience and (3) race. App., 106a, infra. In 1975 and 1978, the average salary of black special agents was about $3,000 below the average salary of white special agents. Plaintiffs asserted that their regression analysis showed that race was probably a significant cause of part of that disparity. Ibid. Petitioners presented evidence that included expert testimony attacking "the methodological integrity and explanatory value of plaintiffs' statistics, alternative statistical analyses tending to show an absence of discrimination, testimonial evidence concerning DEA's equal employment opportunity programs, and cross-examination of plaintiffs' anecdotal accounts of individual discrimination" (App., 13a, infra). Most significantly, petitioners' expert witnesses testified that certain flaws in respondents' statist1cal analysis, including the omission of prior law enforcement or comparable experience as an independent variable, impaired the validity of that analysis and made it improper to conclude from it that race was a significant factor in the perceived salary disparity. /3/ Prior law enforcement experience is a critical factor under civil service requirements in determining whether a special agent will be placed initially at a GS-7 level or a GS-9 level; /4/ the differences in initial placement are a key factor in explaining the perceived racial disparity. Petitioners' experts also presented an extensive series of studies, known as cohort analyses. App., 13a-14a, infra. In these analyses, the experts divided the workforce into groups of agents based on the year the agents joined DEA and the GS grade level they received upon entry. Average salaries were then compared within each such group. Ibid. The cohort analyses showed that the average salaries of black special agents were no lower than those of their white cohorts. Ibid. /5/ 2. The district court ruled on the basis of the foregoing evidence that petitioners had violated Title VII with regard to salary disparities between black and white agents, grade at entry level, work assignments, supervisory evaluations, discipline, and promotions. /6/ The court rejected respondents' anecdotal testimony, ruling that "the allegations of specific discrimination by Black special agents (were) not corroborated by other evidence, and (were) substantially refuted by government witnesses. Defendants proved that the specific acts alleged did not occur because of the agent's race." App., 137a, infra. /7/ The court based its finding of discrimination on testimony regarding blacks' general "feelings and perceptions" of discrimination (App., 136a, infra), and respondents' statistical evidence. Essential to the court's determination of liability was its rejection of the petitioners' contention that respondents' multiple regression analysis was not probative because it failed to account for prior criminal investigative or comparable, specialized work experience. Recognizing that a regression analysis "may be deemed insignificant" if relevant factors are excluded from the analysis, the court nonetheless placed the burden on the defendants to "rework plaintiffs' statistics incorporating the omitted factors or present other proof undermining plaintiffs' claims" (App., 140a, infra). Thus, the court rejected petitioners' criticisms of respondents' multiple regression analysis because the "criticisms were never empirically based" (App., 140a, infra). Instead, the court ruled that there was discrimination in initial placement of black and white special agents, relying upon findings that blacks were 16% less likely than whites to be placed initially in GS-9 positions, rather than GS-7 positions, /8/ without regard to the proportions in which black and white applicants were hired, or whether those who were hired were equally likely to meet the specialized experience requirements for GS-9 positions. /9/ 3. The court of appeals affirmed the district court's liability determination. The court's opinion establishes a "Framework for Analysis" (App., 18a, infra) providing that an employer automatically converts a disparate treatment case into a disparate impact case if it asserts, as a defense to a disparate treatment claim, a facially neutral employment practice (here, differentiation on the basis of prior job experience) as the reason for a perceived racial disparity (App., 18a-31a, infra). Acknowledging (App., 28a, infra) that this analysis conflicts with that of other circuits (citing Pouncy v. Prudential Ins. Co., 668 F.2d 795, 800 (5th Cir. 1982) and Rivera v. City of Wichita Falls, 665 F.2d 531, 539 (5th Cir. 1982)), the court nevertheless ruled that, "when an employer defends a disparate treatment challenge by claiming that a specific employment practice causes the observed disparity,and this defense sufficiently rebuts the plaintiffs' initial case of disparate treatment, the defendant should at this point face a burden of proving the business necessity of the practice" (App., 27a-28a, infra). Thus, under court's analysis, an employer automatically establishes a prima facie case of discrimination against itself, and takes on the burden of showing business necessity, when the employer shows that a facially neutral employment practice explains a perceived racial disparity, regardless of whether plaintiff has alleged that the facially neutral employment practice violates Title VII, or of the state of the rest of the record. /10/ The court of appeals then sustained the district court's finding of intentional discrimination on the basis of inferences of discrimination drawn from the multiple regression analysis presented by respondents. Although the court recognized (App., 8a, infra) that "(t)he choice of proper explanatory variables determines the validity of the regression analysis," like the district court it credited a multiple regression analysis that omitted the qualification of prior specialized experience for initial placement at the GS-9 level. The court gave three reasons for accepting that statistical analysis (App., 39a-41a, infra): (1) it asserted that the civil service definition of "specialized experience" is highly subjective; (2) it placed the burden of production of evidence regarding specialized experience on petitioners because it concluded that they had better access to the relevant information, and (3) it found no evidence in the record suggesting that blacks were less likely than whites to possess the requisite specialized experience. The court rejected petitioners' rebuttal evidence, which included expert testimony and extensive statistical evidence, stating (App., 26a, infra, footnote omitted) that "in the class action pattern or practice case, the strength of the evidence sufficient to meet (the) rebuttal burden will typically need to be much higher than the strength of the evidence sufficient to rebut an individual plaintiff's low-threshold McDonnell Douglas (Corp. v. Green, 411 U.S. 792 (1973)) showing". /11/ The court also ruled that it was not error for the district court to refuse to admit petitioners' alternative regression analysis showing that prior job experience, rather than race, explained the salary differentials at DEA. /12/ This evidence was offered at the remedial stage of the proceeding, and relied upon in moving for reconsideration (Mar. 1, 1982). The court of appeals ruled that it was not an abuse of discretion for the district court to refuse to reopen the liability stage to consider it. /13/ The court of appeals denied a petition for rehearing en banc with Judges Wilkey, Bork, and Scalia dissenting (App., 151a-152a, infra). /14/ REASONS FOR GRANTING THE PETITION Complex, class-wide Title VII litigation requires courts to engage in careful analysis and evaluation of the evidence in accord with the substantive principles of law underlying the statute. In this case, which turns solely on statistical evidence, /15/ the courts below departed from those principles in ways that not only caused them to reach the wrong result in this case, but also will have seriously harmful effects on the conduct of future Title VII litigation by the lower courts. Perhaps most significantly, the court of appeals ignored the method of proof for disparate treatment cases prescribed in Texas Department of Community Affairs v. Burdine, 450 U.S. 248 (1981), and McDonnell Douglas Corp. v. Green, 411 U.S. 792 (1973), and formulateed a conflicting analytical framework that automatically converts a disparate treatment case into a disparate impact case whenever an employer articulates a defense that a perceived statistical disparity is caused by the plaintiffs' failure to account for a specific reasonable employment practice in their statistical proof. The court of appeals' approach not only restricts the employer's defenses previously recognized by this Court; it also would make Title VII litigation that is already extremely complex even more complicated and difficult to resolve fairly. The courts would have to decide, in the midst of class-wide disparate treatment litigation, whether any number of employment practices had a disparate impact not justified by business necessity, even though plaintiffs had not pleaded or presented evidence tending to establish a disparate impact claim. The added burden imposed on the government as an employer would be particularly onerous; it could be required to undergo the enormously expensive and time consuming validation procedures for many employment qualification standards prescribed by the Office of Personnel Management (OPM), even if those standards were not attacked as having a discriminatory impact not justified by business necessity. Otherwise, the government would risk a ruling that, through its defense against a disparate treatment charge, it had made out a prima facie disparate impact claim against itself and had not validated the qualification. There are several hundred federal occupational series, and each series has distinct qualification standards, normally containing different qualification requirements for each grade level. Thus, the rule enunciated by the court of appeals could seriously disrupt the federal personnel system. Moreover, the court of appeals' ruling undercuts one of the basic purposes of Title VII by eliminating the possibility of conciliation before suit; under the panel's approach, the employer need not be informed prior to suit what practices plaintiffs are challenging and why plaintiffs think those practices violate Title VII -- even though those practices are entirely explicit and were well known to the plaintiffs. Thus, conciliation of the disparate impact claims inserted into the lawsuit by the court's unprecedented rule would be impossible. The findings of the courts below that DEA had engaged in purposeful discrimination were based on statistical evidence lacking in probative value; that evidence failed to account for a key qualification -- prescribed by a published, measurable civil service qualification standard -- that controlled initial placement of individuals hired as DEA special agents. Given the absence of any other credible evidence of discrimination, the finding of liability under a disparate treatment theory was without support, and was contrary to principles of law established by this Court and by other courts of appeals. 1. The substantive principles of law underlying Title VII require that disparate treatment claims and disparate impact claims be analyzed differently. When class-wide claims of disparate treatment are in issue, "(p)roof of discriminatory motive is critical" (International Brotherhood of Teamsters v. United States, 431 U.S. 324, 335 n.15 (1977)). In contrast, the employer's motive is irrelevant in a disparate impact case (id. at 336). Once plaintiffs have shown that na identified employment practice has a disparate impact, and thereby established a prima facie case, the employer will be held liable unless the practice can be justified by business necessity. Dothard v. Rawlinson, 433 U.S. 321, 329 (1977); Griggs v. Duke Power Co., 401 U.S. 424, 430 (1971). The rule of the court of appeals automatically converting a disparate treatment case into a disparate impact case, and placing the burden on the employer to prove the business necessity of a facially-neutral employment practice -- without regard to either the pleadings or proof of the plaintiffs -- seriously distorts well-established Title VII analysis. The court of appeals' ruling conflicts with this Court's careful description of the method and order of proof to be followed in disparate treatment cases. In Burdine, this Court explained that the burden on the defendant is "to rebut the presumption of discrimination (raised by plaintiffs' prima facie case) by producing evidence that the plaintiff was rejected, or someone else was preferred, for a legitimate, nondiscriminatory reason" (450 U.S. at 254). This Court then held that if the employer presents "a legitimate reason for the action" the burden shifts back to the plaintiff to show that proffered reason was not the true reason for the employment decision, but was a pretext for discrimination. 450 U.S. at 255-256; see also McDonnell Douglas Corp. v. Green, supra. This method of proof nowhere contemplates a requirement that the employer show the business necessity of the legitimate nondiscriminatory reason for the employment decision. The imposition of such an additional requirement on the employer thus unfairly limits the defenses available to the employer under this Court's decisions. In determining whether a practice asserted by an employer as a defense to a disparate treatment charge is a legitimate, nondiscriminatory reason for the observed disparity, the court may properly evaluate the reasonableness of the practice. But the purpose of that inqury is simply to determine whether the employer had, in fact, intentionally discriminated against plaintiff -- i.e., whether the asserted practice was legitimate, or merely a pretext for discrimination. Under the court of appeals' ruling, however, the evaluation of the asserted employment practice is not limited to this inquiry. Instead, it wrongly changes the focus of the inquiry from whether purposeful discrimination occurred to whether there is a business necessity for the practice. Without the foundation of a disparate impact case -- in which plaintiffs identify, plead and present evidence that a particular employment practice has a disparate impact on a protected class -- there is no legal basis for requiring the employer to prove the business necessity of a facially neutral practice. /16/ The court of appeals incorrectly suggests that Burdine does not apply to a class action, on the theory that plaintiffs' prima facie showing in a class action will always be stronger than that presented in an individual case, so that "(t)he bare articulation of a nondiscriminatory explanation, while sufficient to rebut an individual plaintiff's low-threshold McDonnell Douglas showing, generally will not suffice as a rebuttal to a typical class-wide showing of pervasive discrimination." App., 25a, infra. But the affirmative case can be either very strong or barely sufficient to carry the burden, in either an individual suit or a class action. Indeed, since the court of appeals held that the statistical evidence need not show gross disparities in order to establish a prima facie case, see App., 44a, infra, there is no reason to suppose that in a class suit the prima facie case based on statistical evidence will be more probative of discrimination than the prima facie case in an individual suit. Moreover, precisely the same statistical evidence can be used to establish a prima facie case in support of either an individual or a class claim. There is thus no justification for placing a greater burden on the employer to rebut a prima facie case in the latter situation. Although other courts of appeals have asserted that the Burdine analysis applies only to individual disparate treatment cases, /17/ the court below is alone in adopting a rule that drastically revises the Burdine analysis by utilizing the defendant's evidence to automatically convert a disparate treatment case into a disparate impact case. The court of appeals itself acknowledged that this ruling conflicts with the rule in other circuits (App., 28a, infra). See Pouncy v. Prudential Ins. Co., supra; Rivera v. City of Wichita Falls, supra; Pope v. City of Hickory, 679 F.2d 20, 22 (4th Cir. 1982); Mortensen v. Callaway, 672 F.2d 822, 824 (10th Cir. 1982). These circuits recognize that "(t)he discriminatory impact model of proof in an employment discrimination case is not * * * the appropriate vehicle from which to launch a wide ranging attack on the cumulative effect of a company's employment practices." Pouncy v. Prudential Ins. Co., 668 F.2d at 800. If the plaintiff wishes to use the disparate impact model, he must identify the particular practice he wishes to attack and present evidence showing that it has a disproportionate, adverse effect on him or the members of his class. See Pope v. City of Hickory, supra; Mortensen v. Callaway, supra. The court of appeals' ruling is unfair to employers, and in particular to the federal government as an employer, because it requires employers to defend against charges that are not contained in the pleadings and of which the employer may not have adequate notice. Many class-wide disparate treatment cases are in litigation for years, with numerous claims of discrimination in all kinds of employment practices. Such suits are already immensely complex and enormously expensive to defend. In the course of his defense the employer may attempt to rebut plaintiffs' statistical evidence by showing that the disparity is caused by a nondiscriminatory employment practice. If he succeeds in doing so, the employer has successfully defended against the charge of intentional discrimination against him, unless the plaintiff shows that the employment practice is a pretext for discrimination. There is no legal basis for making the employer go further, and show the business necessity of practice that plaintiffs have not identified or challenged as violating Title VII. See Pouncy v. Prudential Ins. Co., 668 F.2d at 800; Pope v. City of Hickory, 679 F.2d at 22; Mortensen v. Callaway, 672 F.2d at 824; Rivera v. City of Wichita Falls, supra; Heagney v. University of Washington, 642 F.2d 1157, 1163 (9th Cir. 1981). Moreover, the practical difficulties that this rule would cause are enormous. The validation study ordered by the district court in this case and the task of designing new personnel practices for DEA will take at least six years and cost approximately $400,000. /18/ Because virtually all employment qualification determinations in the federal civil service, whether for initial placement, promotion, or entry into training programs, are based on minimum qualification standards published by OPM, /19/ the implications of the court of appeals' ruling are apparent: the ruling could place at issue the validity of innumerable qualification standards, with the consequent staggering expense of validation of such standards, even in the absence of any challenge to those standards under a disparate impact theory. /20/ Moreover, litigation against any individual agency would threaten the employment practices of the entire government, because the court's ruling could require a federal agency to show the business necessity of a standard prescribed government-wide by the OPM. When objective employment qualifications are published, and thus readily available to plaintiffs, and plaintiffs have not charged or proved that any such practices violate Title VII under a disparate impact theory, there is no justification for imposing such burdens on the government as an employer. The court of appeals' ruling also undercuts an important goal of Title VII by virtually eliminating the possibility for conciliation between plaintiffs and employers. A carefully constructed administrative process is prescribed by Title VII, and initiation of that process has been made a prerequisite to bringing suit, so that -- to the extent possible -- differences can be worked out informally between employer and employee without resort to litigation, for "conciliation, rather than litigation, is a recognized goal of Title VII." Porter v. Adams, 639 F.2d 273, 277 (5th Cir. 1981). See 42 U.S.C. 2000e-16. The court of appeals' theory, allowing conversion of a disparate treatment case to a disparate impact case in mid-trial, circumvents the administrative process and the possibility of conciliation of the specific employment practices between employee and employer prior to trial. /21/ Moreover, the court's rule applies only to classwide cases, in which plaintiffs generally are represented by counsel, and usually ones well-versed in Title VII law. Thus, the requirement that counsel in such cases abide by the administrative exhaustion requirements will further the substantive purposes of Title VII and will not create a risk that individuals unlearned in the law will be deprived by a legal technicality of having their complaints heard. See Love v. Pullman Co., 404 U.S. 522 (1972). The court of appeals' ruling automatically converting a disparate treatment case into a disparate impact case in the midst of litigation, even if considered an alternative holding or dictum in this case, will have an immediate and crippling effect on orderly procedure and fair resolution of Title VII cases. Clearly, the court has prescribed a procedure to be used henceforth by lower courts in the conduct of Title VII cases -- a procedure binding on district courts in the leading circuit for the litigation of complex Title VII class claims against the government. /22/ See 1B Moore's Federal Practice Paragraph 0.402(2) (2d ed. 1984). The enormous expense and difficulties for employers, and particularly the federal government, as well as the damage to the orderly conduct of Title VII litigation, that will be caused by the rule announced by the court of appeals argues strongly for review by this Court at this time. 2. In finding intentional discrimination in this case, the court of appeals has once again ignored fundamental principles governing the evaluation of evidence in Title VII litigation, as it did in Aikens v. United States Postal Service, 665 F.2d 1057 (D.C. Cir. 1981), vacated, 460 U.S. 711 (1983), and Trout v. Lehman, 702 F.2d 1094 (D.C. Cir. 1983), summarily rev'd, No. 83-706 (Feb. 27, 1984). The court recognized that "(t)he trial was in large measure a duel of experts armed with sophisticated statistical means of proof," App., 7a-8a, infra, but it failed to evaluate that statistical evidence in light of applicable legal principles, and to require that there be a logical basis for inferring discrimination from it. Mayor of Philadelphia v. Educational Equity League, 415 U.S. 605, 620-621 (1974). We recognize that, in an appropriate case, intentional discrimination can be inferred from statistical evidence alone, if that evidence reaches proper proportions and takes into account relevant factors. See International Brotherhood of Teamsters v. United States, 431 U.S. at 339-340. But the statistical evidence presented here simply did not meet that standard. Relying on the uneven distribution of blacks and whites in certain grade levels, the court of appeals appears to have lost sight of the fact that the principal focus of Title VII is the protection of individuals from discrimination. See Connecticut v. Teal, 457 U.S. 440, 453 (1982); City of Los Angeles Department of Water & Power v. Manhart, 435 U.S. 702, 708 (1978); 110 Cong. Rec. 7213 (1964). Title VII "does not ultimately focus on ideal social distributions of persons of various races and both sexes. Instead it is concerned with combating culpable discrimination." Contreras v. City of Los Angeles, 656 F.2d 1267, 1275 n.5 (9th Cir. 1981). Properly used, statistical models and analyses can help a court determine whether such culpable conduct occurred. However, to perform that function, statistical evidence must meet minimum standards of reliability. For example, properly constructed statistical analysis would have aided the court here in determining whether the difference in initial placements resulted from discrimination or from legitimate, nonracial employment requirements. But to perform such a function, the statistical analysis had to account for the minimum measurable qualifications necessary for the higher initial placement. See e.g., De Medina v. Reinhardt, 682 F.2d 997 (D.C. Cir. 1982); Valentino v. United States Postal Service, 674 F.2d 56 (D.C. Cir. 1982); Eastland v. Tennessee Valley Authority, 704 F.2d 613, 624-625 (11th Cir. 1983). Conversely, a plaintiff's regression analysis that does not include such qualifications as variables lacks probative value. In the instant case, the salary disparity between blacks and whites results from a larger proportion of whites than blacks beginning at a GS-9 rather than a GS-7 grade level. Respondents' regression analysis included as independent variables nothing more than race, educational level, and the number of years of work experience. It wholly failed to differentiate between types of experience. But the published standards of OPM distinguished between the qualifications for placement at GS-7 and GS-9 levels with sufficient specificity to permit measurement. At least one year of criminal investigative or comparable experience was a basic requirement for placement as a DEA special agent at either GS-7 or GS-9 level. An applicant with two years of prior criminal investigative or comparable experience qualified for a GS-9 position, while those with less time spent in such specialized experience could not be assigned an entry grade level higher than GS-7. /23/ Because it failed to account for this second year of specialized experience as a minimum objective qualification for placement at the higher grade level, respondents' statistical analysis lacked probative value. /24/ It simply did not compare people with comparable qualifications. The recent decision of the Court of Appeals for the Fourth Circuit in Bazemore v. Friday, No. 82-1873 (Dec. 10, 1984), petition for rehearing and suggestion for rehearing en banc pending, provides a significantly different approach in evaluating multiple regression analysis than that applied by the court of appeals in the instant case. /25/ The Bazemore court found (slip op. 30) that a multiple regression analysis that failed to include nine independent variables lacked probative value where plaintiffs alleged that certain salary disparities were the result of racial discrimination: The purpose of multiple regression is to account for changes in a single variable through the examination of other variables which might be expected to influence that variable. In the present instance, that single variable is salary. An appropriate regression analysis of salary should therefore include all measureable variables thought to have an effect on salary level. * * * However, both experts omitted from their respective analysis variables which ought to be reasonably viewed as determinants of salary. As a result, the regression analyses presented here must be considered unacceptable as evidence of discrimination. What independent variables are reasonably required to be included wthin a multiple regression analysis before that analysis can be said to have significant probative value will depend on the facts of each case. In Bazemore, prior experience was one of the nine independent variables that the court believed should have been, but was not, included in the multiple regression analysis submitted in that case. The United States argued in Bazemore that the prior experience requirement need not be included in the multiple regression analysis because, inter alia, the hiring agency did not have any standards for determining the particular kinds of prior experience that warrant an increased starting salary. Bazemore was concerned not with prior experience standards for placement at different levels, but with salary differences among people performing at the same job level. In contrast, the court of appeals in the instant case credited a regression analysis that failed to include explicit prior experience qualifications that distinguish between persons eligible for placement at a GS-9 level and those who could be placed no higher than a GS-7 level. The prior experience requirements in this case are sufficiently described in the OPM Handbook X-118 to be measurable and, thus, do not suffer from the vagaries that characterize the prior experience requirements at issue in Bazemore. We believe that the Bazemore court went too far in requiring that the multiple regression analysis in that case include prior experience as an independent variable; our position is not that the plaintiffs' multiple regression analysis must include any qualification that the employer at some point in the litigation asserts as necessary for the position at issue. Rather, a rule of reasonableness must apply. In the instant case, the prior experience requirement for placement as GS-9 was published by the government, thoroughly explained in Handbook X-118, and a factor that any reasonable analysis should have considered in determining why a disproportionate number of whites were initially placed at the higher job level in DEA. The failure of a regression analysis to include such a factor destroys its significance. The court of appeals gave three reasons for crediting respondents' regression analysis despite its failure to account for this minimum experience requirement. None is substantial. First, the court of appeals' objection that the experience requirement provided in the relevant standard was too subjective to be included within a regression analysis is without foundation. The OPM Handbook X-118 specifically describes the type of experience that would satisfy the specialized experience requirement for initial placement of special agents -- criminal investigative -- and gives a number of examples of acceptable experience: investigative experience as a member of a military intelligence or criminal investigative component of any of the branches of the Armed Services or in any of the various intelligence or investigative branches of the public service, investigation of complex casualty or insurance claims, investigation of criminal cases for practicing attorneys, and experience as a uniformed law officer where the principal duties are criminal investigations. Handbook X-118 also specifies various types of experience related to investigative work, but that would not qualify for the experience requirement: experience in performing the normal duties of policemen, guards, watchmen or private detectives assigned principally to the protection of life or property, experience as a probation or parole officer not involving pre-sentence investigations or the like, investigations of minor insurance damage claims, clerical work incidental to the adjudication of claims, and experience as a payroll clerk, cashier, or inspector whose primary duty was to examine plants and materials. Clearly, the standards set forth in the Handbook refute any suggestion that the specialized experience requirement is not sufficiently objective to be included within a multiple regression analsysis. /26/ The court's second reason for excusing respondents' failure to include the exerpience requirement, i.e., that petitioners had better access to the necessary information, is also erroneous. All the necessary information was contained in personnel folders that were equally available to respondents. Respondents chose to use only the information contained in Department of Justice computerized personnel records, called JUNIPER tapes. These tapes are not used in the hiring or promotion process and do not record data critical to grade assignment under the X-118 standards. Respondents' failure to go beyond the information contained in the JUNIPER tapes and include the information necessary to determine whether an applicant was qualified for initial placement at the GS-9 level caused their regression analysis to be fatally deficient. The court finally asserted that "even if we accept arguendo that a second year of specialized experience -- or, for that matter, of criminal investigative experience -- is a minimum objective qualification, DEA's objection would still fall far short * * * (because a) part from the unsubstantiated declamations of DEA's appellate attorneys, nothing in the record so much as hints that black special agents are less likely than white special agents to possess this qualification" (App., 40a-41a, infra). But the question at issue was solely whether respondents' regression analysis sufficiently reflected the actual requirements for the jobs at issue to constitute reliable evidence. Thus, in requiring petitioners to present evidence concerning the impact of the specialized experience requirement on blacks, the court again placed an improper burden on defendants. It is incumbent on plaintiffs to produce statistics that reflect the minimum objective qualifications for the positions at issue. See Davis v. Califano, 613 F.2d 957, 964 (D.C. Cir. 1979); Valentino v. United States Postal Service, 674 F.2d 56, 70 (D.C. Cir. 1982); Eastland v. Tennessee Valley Authority, 704 F.2d 613, 623-625 (11th Cir. 1983). Because of its repudiation of that principle, the decision below warrants this Court's review. CONCLUSION The petition for a writ of certiorari should be granted. Respectfully submitted. REX E. LEE Solicitor General RICHARD K. WILLARD Acting Assistant Attorney General ROBERT E. KOPP JOHN C. HOYLE Attorneys JANUARY 1985 /1/ Special agents are criminal investigators. Their duties include: surveillance of suspected narcotics dealers, transactions of "buys" of illegal drugs as evidence for prosecutions, development of cases for prosecution by United States Attorneys, and, in some instances, supervision of the work of other special agents. App., 101a, 103a, infra. /2/ Claims of discrimination in recruitment and hiring were settled prior to trial. App., 100a, infra. In addition to claims of disparate treatment, respondents also challenged a number of aspects of DEA's personnel practices such as initial grade assignments, work assignments, supervisory evaluations and discipline and promotion decisions, as having a disparate impact on black agents (App., 20a, infra). However, respondents failed to isolate an employment selection criterion, such as an aptitude or intelligence test or an educational requirement, that would properly be subject to a disparate impact analysis. See Griggs v. Duke Power Co., 401 U.S. 424 (1971); Pouncy v. Prudential Ins. Co., 668 F.2d 795, 799-800 (5th Cir. 1982). Consequently, neither the district court nor the court of appeals applied a disparate impact analysis to respondents' claims, except to the extent discussed infra. /3/ They also pointed out that the respondents' multiple regression analysis had a very low R2 value (10 Tr. 1853-1855, 1859). The R2 value measures the difference in the dependent variable that is likely to be attributable to the independent variables: where the independent variables totally control the dependent variable, the R2 value is one; where they are entirely unrelated, the value is zero. Finkelstein, Regression Models in Administrative Proceedings, 86 Harv. L. Rev. 1442, 1448-1449 (1973); Vuyanich v. Republic Nat'l Bank, 505 F. Supp. 224, 273 (N.D. Tex. 1980), vacated on other grounds, 723 F.2d 1195 (5th Cir. 1984). The R2 values here showed that all the independent variables in respondents' multiple regression analysis taken together explain only 21% to 37% of the variations in salary of special agents hired after the Act's effective date (App., 52a-53a n. 27, 107a, 109a, infra). Similarly, these variables explained less than 20% of the difference in rate of hiring at different grade levels between blacks and whites (PX 7, Tables 8, 8a). See page 7, infra. /4/ While at least one year of criminal investigative or comparable experience is a basic prerequisite for service as a special agent, an applicant with two years of prior criminal investigative or comparable specialized experience would qualify for a GS-9 position upon entering DEA (App., 101a-102a, infra). Applicants with less than this length of such experience could not be assigned an entry level higher than GS-7 (ibid.). /5/ Petitioners' evidence also showed that DEA's Administrator at the time of the trial, Peter Bensinger, had an impressive record on equal employment opportunity matters, including responsibility for appointment of numerous minority persons to high level posts in the State of Illinois before he assumed his position at DEA. As Administrator of DEA, Mr. Bensinger continued his commitment to equal employment opportunity policies, establishing advisory committees among DEA's black, Hispanic, female, and Asian American employees, as well as an overall EEO Advisory Committee that met with him quarterly. He estimated that he spent 10% to 15% of his time as Administrator on EEO matters. He made efforts to reduce the processing time for individual complaints and installed a monthly EEO tracking system. The court of appeals noted that "the record contains significant uncontradicted evidence of DEA's institutional good faith in implementing equal employment opportunity programs" (App., 78a, infra). /6/ The district court ruled that respondents had failed to prove discrimination in Schedule A appointments, traning, and harassment and reprisal. App., 146a, infra. /7/ With regard to the testimony of the three witnesses called by respondents who claimed discrimination in initial grade assignment, the district court found that one "did not have Civil Service status at the time he applied to DEA and thus could not be equated with" two white agents who were placed at a higher grade than he (App., 119a, infra); another did not present any evidence regarding the qualifications of individuals hired at a higher grade level than he (App., 120a, infra); and the third was inadevertently placed at a GS-7 level due to an administrative error, but the mistake was corrected and he was placed at a GS-9 level one day after his initial incorrect placement (App., 122a, infra). The court concluded that petitioners "successfully rebutted" respondents' anecdotal testimony regarding initial placement "at every turn" (App., 142a, infra). /8/ App., 112a, infra. /9/ In a subsequent opinion and order, the district court ordered extensive relief against petitioners, including classwide backpay relief, promotional goals and timetables, and frontpay specifically linked to those goals and timetables. The court of appeals substantially overturned these orders and remanded for reconsideration (see note 13, infra). We accordingly do not seek review of any of the relief ordered. /10/ Judge Edwards filed a concurring opinion in which he stated (App., 97a, infra) that "(s)o long as the defendant has adequate notice of the need to prove business necessity, it is irrelevant that this duty is prompted by the defendant's, rather than the plaintiff's, proof." /11/ The court noted (App., 56a-57a, infra) that the district court erred in ruling that "(a)fter the establishment of a prima facie case * * * the burden of persuasion shifts to the employer," but excused that error on the ground that after this Court's decision in Texas Department of Community Affairs v. Burdine, 450 U.S. 248 (1981), the district court responded to a motion for clarification by stating that it had allocated the burdens in conformity with Burdine. /12/ Although the court of appeals stated (App., 58a, infra) that this regression analysis is not part of the record, it was filed as an affidavit attached to a memorandum of points and authorities concerning remedies on November 2, 1981; the district court simply chose to rely on plaintiffs' regression analysis and refused to reconsider its liability finding (Memo. Op. of Feb. 17, 1982, at 3). /13/ The court of appeals vacated the imposition of strict promotion goals and timetables in the district court's remedial order, as well as the class-wide frontpay tied to these goals and timetables, and remanded the case for further remedial proceedings by the district court. App., 81a-82a, infra. The court also affirmed the district court's use of a class-wide backpay scheme, but vacated the particular calculation of the backpay formula. Ibid. Four days after it issued its opinion, the court of appeals issued a supplemental order stating that "(i)n considering the issues of appropriate remedies on remand, the District Court should * * * consider the Supreme Court's opinion in (Firefighters Local Union No. 1784 v.) Stotts, to the extent applicable" (App., 82a, 154a, infra). /14/ Judge Starr did not participate. /15/ Both courts below rejected respondents' anecdotal evidence (page 6, supra; App., 42a-45a, infra). /16/ In the instance case, the court of appeals did not even consider whether there was evidence in the record establishing that any employment practice had a significant disparate impact. It merely assumed that if the employer raises an employment practice as an explanation for a perceived racial disparity, the employment practice must be taken to have a sufficiently disparate impact to constitute a prima facie case under a disparate impact theory. This assumption is highly questionable; in the instant case there is no evidence that the specialized experience requirement had a sufficiently severe impact on the initial placement of special agents to satisfy the requirements for a prima facie case under the disparate impact theory. Blacks were only 16% less likely than whites to be initially placed at a GS-9 rather than a GS-7 position, and for those hired after the effective date of Title VII, the difference was only 12%. App., 12a, infra. The federal employee selection guidelines provide that "(a) selection rate for any race * * * which is less than four-fifths * * * (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact, while a greater than four-fifths rate will generally not be (so) regarded" (29 C.F.R. 1607.4 D; 28 C.F.R. 50.14, Sec. 4D (emphasis added)). /17/ Craik v. Minnesota State University Board, 731 F.2d 465, 468-471 (8th Cir. 1984); Payne v. Travenol Laboratories, Inc., 673 F.2d 798, 818 (5th Cir.), cert. denied, 459 U.S. 1038 (1982). But see Gerdom v. Continental Airlines, Inc., 692 F.2d 602, 608-609 (9th Cir.), cert. dismissed, 460 U.S. 1074 (1982); Croker v. Boeing Co., 662 F.2d 975, 990-991 (3d Cir. 1981). Craik and Payne suggest that International Brotherhood of Teamsters v. United States, 431 U.S. at 360-362, establishes the relevant pattern of proof. But the description in Teamsters of the burden imposed on the employer after the plaintiffs establish a prima facie case of disparate treatment -- to show that plaintiffs' proof is "inaccurate or insignificant * * * (or) to provide a nondiscriminatory explanation for the apparently discriminatory result" (431 U.S. at 360-361 & n.46) -- does not differ significantly from Burdine's description of the employer's burden. /18/ See Joint Stipulation of July 31, 1981; Stipulation of Apr. 25, 1983; Notice of Filing of Jan. 7, 1985; Supplemental Notice of Filing of Jan. 15, 1985. /19/ The standard involved here is Series GS-1811, Criminal Investigator, contained in the OPM's Qualification Standards for Positions Under the General Schedules (1968), known as Handbook X-118. /20/ The problem is exacerbated by the fact that the standards vary greatly in the extent to which they require the exercise of expertise in their application to individual employee or applicant qualifications. Standards that leave much discretion to expert personnel judgments are simply not susceptible to validation, in the same way that test scores, or height or weight limitations are, and in some instances may not be susceptible to validation at all. /21/ Conciliation is likely to be an effective way to resolve disparate impact claims in federal employment, because the use of selection procedures with a disparate impact is inconsistent with federal employment policies. The federal government has published guidelines providing that "(t)he use of any selection procedure which has an adverse impact on the hiring, promotion, or other employment * * * opportunities of members of any race * * * will be considered to be discriminatory and inconsistent with these guidelines, unless the procedure has been validated in accordance with these guidelines, or the provisions of section 6 (excusing validation when it is not "technically feasible") * * * are satisfied." 28 C.F.R. 50.14, Sec. 3. /22/ This ruling is another example of what has been perceived as a disturbing practice of this court of appeals of announcing important rulings as alternative holdings or dicta, thereby allegedly attempting to evade accountability. See Scalia, Vermont Yankee: The APA, the D.C. Circuit, and the Supreme Court, 1978 Sup. Ct. Rev. 345, 372. /23/ The standard contained a number of examples of the types of experience that qualify as comparable. See pages 25-26, infra. Respondents presented no evidence that the standard was applied in a discriminatory way. See note 26, infra. /24/ The lack of probative value of respondents' regression analysis is confirmed by its very low R2 value: all of the independent variables included in that analysis explained no more than 20% of the differences in initial placement between agents. See note 3, supra. /25/ The government has filed a petition for rehearing with suggestion of rehearing en banc in Bazemore, arguing that the court of appeals erred in failing to find that the continuing effects of pre-Act discrimination in compensation violate Title VII, in failing to find that a performance rating system violated Title VII under a disparate impact theory, and in failing to find that some blacks were entitled to relief even under the approach employed by the court of appeals panel. Private plaintiffs have also filed a rehearing petition and have raised, inter alia, the issue of who has the burden of including the asserted independent variables in a regression analysis. /26/ There is absolutely no suggestion that the "comparable experience" standard was applied in a discriminatory way. The evidence of the way the standard had been applied to the class members and their white cohorts was available to respondents in the personnel folders, and they could have challenged any discriminatory application of that standard in their anecdotal evidence. Since they presumably presented their most persuasive anecdotal evidence, the fact that both courts below concluded that it showed no discriminatory treatment strongly suggests that the standard was applied in a non-disriminatory manner.