concurring in part and dissenting in part:
This “reverse discrimination” case presents some difficult questions arising from the application of an affirmative action plan. I conclude that the record before us is insufficient to make broad pronouncements of judicial policy. I would vacate and remand this case for two reasons. First, the record suggests that the district court improperly allocated the burdens of persuasion and production between *1315the parties. Second, the court did not make sufficiently detailed findings on the plan to hold it invalid.
I
The facts of this case, referred to in part by the majority, áre significant. Johnson, a 58 year-old white male, has lived in Santa Clara for many years. After working 17 years for a cement company, including 7 years as a dispatcher and 8 years as a supervisor, he quit rather than accept a transfer that would uproot his family. Instead, he took a job with the Agency. He worked the better part of 11 years in the position of Road Yard Clerk II, requisitioning road materials for operations, processing purchase orders and arranging material provisions in emergencies. Eventually, however, he concluded that the job offered him little chance for advancement. He requested a voluntary demotion to Road Maintenance Worker II, hoping that the Road Maintenance Division would provide better opportunities for promotion to a position such as road dispatcher.
Agency road dispatchers assign crews, equipment, and materials to road maintenance job sites. They must keep accurate records on the status and availability of workers, materials, and equipment at those sites and need a working knowledge of all those things as well as of the various roads maintained by Santa Clara county. Job candidates usually have four years of experience in either dispatch or road work, and can operate a radio telephone. In 1974, Johnson applied to be a road dispatcher but placed second and failed to get the job. The position re-opened in December 1979, by which time Johnson had gained more road work experience, including several months work “out of class” as a road dispatcher. He and eight others applied for the position. Seven of them scored over the required 70 on an examination given by a two-member oral board on April 24, 1980.
Johnson placed second, with a score of 75. Employees from the Agency’s Roads Division then conducted a second oral board for the seven qualifying applicants, and unanimously recommended Johnson for the job. Neither party denied that both the oral boards and the examination process were fair and in accord with Merit System Rules derived from the County charter. Section 7 of that charter prohibits sex discrimination and ensures promotions based on merit and equal opportunity.
Diane Joyce, the woman the Agency ultimately hired instead of Johnson, placed fourth on the examination, scoring 72.5, rounded up to a 73. She had worked as a road maintenance worker for five years and before that a bookkeeper, account clerk, and senior account clerk. She wanted to be a road dispatcher in 1974, but lacked the requisite experience for consideration. During the selection process in 1980, she contacted the County’s Women’s Coordinator, Lee, and told her about the job opening. Joyce reported that she had placed low on the eligibility list, however. Lee told Morton, the Agency’s Affirmative Action coordinator, about Joyce and Morton recommended Joyce’s promotion to Graebner, the Director of the Agency. Graebner then promoted Joyce. It was the only time he had ever directly made a promotion decision within the Roads Operation Division. Without Morton’s recommendation, one of Graebner’s subordinates would have made the road dispatcher appointment.
Graebner promoted Joyce under a voluntary, noncollectively bargained affirmative action plan dated December 18, 1978 (plan). The plan set no “quotas,” per se, in any job classification, but rather had long-range percentage goals aiming at some numerical parity with women available in the local labor market. It specified no past discriminatory Agency practices, identified no history of purposeful exclusion it sought to remedy, and did not follow Equal Employment Opportunity Commission (EEOC) guidelines, see 29 C.F.R. § 1608.1-.12 (1984). It simply stated that the relevant job classifications had traditionally underrepresented women, and recognized an extreme difficulty in increasing “significantly the representation of women in certain of *1316those technical and skilled-craft jobs.” When Graebner promoted Joyce under the plan, he did not carefully inspect the applications and related examination records of Joyce and Johnson, as the County’s Merit System Rules authorized him to do. He thought he simply retained discretion to pick from among the top candidates who had crossed a certain threshold of ability.
II
In Title VII cases, “the ultimate question” is “discrimination vel non." United States Postal Service Board of Governors v. Aikens, 460 U.S. 711, 714, 103 S.Ct. 1478, 1481, 75 L.Ed.2d 403 (1983) (Aikens). The existence of an affirmative action plan does not change this essential inquiry.
Title VII plaintiffs may challenge promotion decisions under either disparate treatment or disparate impact theories. Johnson asserts only a disparate treatment claim of the classic Title VII kind: the Agency intentionally and unjustifiably treated him differently only on the basis of his sex. “The ‘factual inquiry’ in a Title VII case is ‘[whether] the defendant intentionally discriminated against the plaintiff.’ ” Aikens, 460 U.S. at 715, 103 S.Ct. at 1482, quoting Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 253, 101 S.Ct. 1089, 1093, 67 L.Ed.2d 207 (1981). Of course, the fact that Johnson is a white male is irrelevant to his quantum of protection under Title VII. E.g., McDonald v. Santa Fe Trail Transportation Co., 427 U.S. 273, 280 & n. 8, 96 S.Ct. 2574, 2579 & n. 8, 49 L.Ed.2d 493 (1976). “[A] plaintiff’s ultimate burden of persuasion remains the same regardless of race [or sex]: he must prove that the defendant intentionally discriminated against him.” Lanphear v. Prokop, 703 F.2d 1311, 1315 (D.C.Cir.1983).
In McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973) (McDonnell Douglas), the Supreme Court first comprehensively addressed “the order and allocation of proof in a private, non-class action challenging employment discrimination.” Id. at 800, 93 S.Ct. at 1823. At the time, and later, the Court recognized that “[t]he facts necessarily will vary in Title VII cases, and the specification ... of the prima facie proof ... is not necessarily applicable in every respect to differing factual situations.” Id. at 802 n. 13, 93 S.Ct. at 1824 n. 13, quoted in Furn-co Construction Corp. v. Waters, 438 U.S. 567, 575-76, 98 S.Ct. 2943, 2948-49, 57 L.Ed.2d 957 (1978) (Furnco). The three stages of proof presentation and burden shifting under McDonnell Douglas entail no trifurcation of trial, but simply provide courts with “a sensible, orderly way to evaluate the evidence in light of common experience as it bears on the critical question of discrimination.” Furnco, 438 U.S. at 577, 98 S.Ct. at 2949; see generally B. Schlei & P. Grossman, Employment Discrimination Law 1321-22 (2d ed. 1983).
A Title VII plaintiff carries “the initial burden under the statute of establishing a prima facie case of racial discrimination.” McDonnell Douglas, 411 U.S. at 802, 93 S.Ct. at 1824. Johnson had to show that sex was the likely reason for the denial of his job opportunity. See Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 252-53, 101 S.Ct. 1089, 1093, 67 L.Ed.2d 207 (1981) (Burdine); Hagans v. Andrus, 651 F.2d 622, 625 (9th Cir.), cert. denied, 454 U.S. 859, 102 S.Ct. 313, 70 L.Ed.2d 157 (1981); White v. City of San Diego, 605 F.2d 455, 458 (9th Cir.1979). When this burden is met by the employee, it “creates a rebuttable ‘presumption that the employer unlawfully discriminated against’ him.” Aikens, 460 U.S. at 714, 103 S.Ct. at 1481, quoting Burdine, 450 U.S. at 255, 101 S.Ct. at 1094.
At this juncture of the McDonnell Douglas analysis, the question arises whether the use of an affirmative action plan to rebut a charge of employer discrimination is properly part of the plaintiff’s prima facie case or whether it should be considered as a separate affirmative defense subsequent to the McDonnell Douglas analysis. If analyzed under the traditional McDonnell Douglas test, once a Title VII plaintiff establishes a prima facie case, *1317“[t]he burden then must shift to the employer to articulate some legitimate, nondiscriminatory reason” for the actions that inspired the complaint. McDonnell Douglas, 411 U.S. at 802, 93 S.Ct. at 1824. To rebut the presumption against him, therefore, the employer must “produc[e] evidence that the plaintiff was rejected, or someone else was preferred, for a legitimate, non-discriminatory reason.” Burdine, 450 U.S. at 255, 101 S.Ct. at 1094, quoted in Aikens, 460 U.S. at 714, 103 S.Ct. at 1481 (emphasis added). The only burden shifted to the employer is the burden of going forward with evidence that suggests a legitimate defense of his actions. Meeting this burden puts the trier of fact “in a position to decide the ultimate factual issue in the case.” Aikens, 460 U.S. at 715, 103 S.Ct. at 1482; see also Knutson v. Boeing Co., 655 F.2d 999, 1001 (9th Cir.1981) (employee retains the ultimate burden of showing intentional discrimination).
Under a McDonnell Douglas analysis, one legitimate reason an employer may offer to rebut the discrimination presumption, and fulfill his burden of going forward, is a bona fide affirmative action plan. See, e.g., Hunter v. St. Louis-San Francisco Railway Co., 639 F.2d 424, 426 (8th Cir.1981). The employer must articulate the reasons for rejection clearly, and do so through admissible evidence. See Burdine, 450 U.S. at 255 & n. 9, 257-58, 101 S.Ct. at 1095-96. When the employer has articulated a reason and rebutted the employee’s prima facie case, “the factual inquiry proceeds to a new level of specificity.” Id. at 255, 101 S.Ct. at 1094. The employee must have an “adequate ‘opportunity to demonstrate that the proffered reason was not the true reason for the employment decision,’ but rather a pretext.” Aikens, 460 U.S. at 716 n. 5, 103 S.Ct. at 1482 n. 5, quoting Burdine, 450 U.S. at 256, 101 S.Ct. at 1095; see McDonnell Douglas, 411 U.S. at 807, 93 S.Ct. at 1826. As long as the employee has this opportunity, the trial court must then decide whether, there has been impermissible discrimination.
In this case, the Agency offered an affirmative action plan and other evidence to rebut the presumption that its decision was sex-based. Mere presentation or articulation of any plan, of course, will not rebut the presumption. The employer must show sufficient facts suggesting a legal, bona fide affirmative action plan meeting the minima defined in United Steelworkers of America, AFL-CIO-CIC v. Weber, 443 U.S. 193, 99 S.Ct. 2721, 61 L.Ed.2d 480 (1979) (Weber), and the law of our own circuit applying Weber, La Riviere v. EEOC, 682 F.2d 1275 (9th Cir.1982). The plaintiff’s only Title VII burden of persuasion is to prove discrimination. He need not prove the illegality of an affirmative action plan merely alleged by his employer, or one for which his employer has not produced evidence to meet the prima facie requirements of a bona fide plan. The employer’s burden is to go forward with evidence of a legal plan, and it is in his best interest carefully to present the fullest evidence possible.
There is an alternative way to analyze an affirmative action plan: as a separate affirmative defense raised by the employer. This is not a distinction without a difference. The burden of proof under the McDonnell Douglas test never shifts from the plaintiff. The burden of proof to demonstrate the validity of a separate affirmative defense rests with the person asserting it: the employer in Title VII cases. It is logical to require the employer to prove the validity of its plan since it is the emT ployer’s plan that has allegedly caused a Title VII injury.
The statute already provides a similar type of analysis for one aspect of sex discrimination. Title VII provides a narrow exception that allows classifications or treatment based on sex, but only if sex is “a bona fide occupational qualification reasonably necessary” to an employer’s operation. 42 U.S.C. § 2000e-2(e)(1). That statutory exception takes the form of an affirmative defense for the employer: if the court disbelieves his proffered reasons, he may admit discrimination between employ*1318ees on the basis of sex but argue that the discrimination was justified. See generally K. Davidson, R. Ginsburg & H. Kay, Sex-Based Discrimination 667-95 (1974); Siro-ta, Sex Discrimination: Title VII and the Bona Fide Occupational Qualification, 55 Tex.L.Rev. 1025 (1977).
Thus, if affirmative action plans are presented as affirmative defenses, the burdens of going forward and persuasion dictated by Supreme Court and Ninth Circuit precedent form a simple and logical framework for deciding these difficult issues. This approach seems to me to provide a better means of analyzing affirmative action plans.
The majority does not discuss this problem, but concludes without analysis that the employer need only show “some evidence” that the plan was in response to a conspicuous work force imbalance and that the plan is reasonably related to the plan’s remedial purposes. (At p. 1310 n. 2.) Then the plaintiff must prove the plan is invalid. I disagree with this analysis. In addition, all that is required by the majority’s holding is to vacate the judgment to allow the district judge to apply these new pronouncements on burdens of proof and persuasion. The subsequent broad pronouncements on whether the plan passes muster under Weber are premature.
The record does not indicate clearly whether the district court properly allocated the burdens of going forward and persuasion. Therefore, the record’ is sufficiently suspect that I would have remanded this case for the proper application of those burdens, and for consideration of the Agency plan as an affirmative defense.
Ill
The second reason that I write separately is because the district court made insufficient findings on the Agency’s affirmative action plan to justify its rejection of the plan.
Title VII embodies a broad legal principle of anti-discrimination. In Weber, the Supreme Court held that Title VII does not necessarily preclude voluntary, private affirmative action, and thus. carved out a very narrow exception to that principle: it rejects random discrimination against various minorities or whites or males, but allows planned discrimination under very limited circumstances. See 443 U.S. at 200, 99 S.Ct. at 2725. In La Riviere, we did not address the constitutional issue but applied Weber to a voluntary, public affirmative action plan. In each case the result is to sanction some discrimination not otherwise legally conscionable — because based on race or sex — when it aims to remedy purposeful past discrimination.
Courts, however, must be careful to “ensure that new forms of invidious discrimination are not approved in the guise of remedial affirmative action.” Setser v. Novack Investment Co., 657 F.2d 962, 968 (8th Cir.), cert. denied, 454 U.S. 1064, 102 S.Ct. 615, 70 L.Ed.2d 601 (1981). The bona fides of affirmative action plans must be evaluated carefully if the employer asserts he has treated plaintiff pursuant to one, because Title VII compels us to consider the rights of individuals, not groups. See, e.g., Connecticut v. Teal, 457 U.S. 440, 453-54, 102 S.Ct. 2525, 2534, 73 L.Ed.2d 130 (1982); City of Los Angeles Department of Water and Power v. Manhart, 435 U.S. 702, 708, 98 S.Ct. 1370, 1375, 55 L.Ed.2d 657 (1978).
Two overarching questions guide the analysis of plans: “(1) when is it permissible to initiate an affirmative action plan, and (2) what methods may be employed to prefer one race [or sex] over another.” Setser, 657 F.2d at 967 n. 4. Although Weber did not exclusively “define in detail the line of demarcation between permissible and impermissible affirmative action plans,” 443 U.S. at 208, 99 S.Ct. at 2729-30, it identified at least four factors which La Riviere adapted to public plans. The plan: (1) must aim at breaking down historic patterns of racial or social segregation and hierarchy; (2) may not unnecessarily trammel the interests or employment opportunities of other employees; (3) must not absolutely bar the advancement of other employees; and (4) must remain only a tempo*1319rary measure aimed at manifest racial or social imbalances. La Riviere, 682 F.2d at 1279-80, citing Weber, 433 U.S. at 208, 99 S.Ct. at 2729.
The majority evaluated the Agency’s plan in light of the four factor test and concluded that the plan was valid. Since a plan must pass muster under all four factors, a separate analysis of each factor is necessary.
A.
The first question is whether the plan breaks down historic patterns of discrimination. The majority concludes that the Agency’s plan was remedial, relying solely on the imbalance in the work force. Although I agree that statistics are useful and probative on this issue, a statistical imbalance should not suffice to satisfy this prong of Weber. An employer should use an affirmative action plan not only for identifying and making whole victims of discrimination, but also to identify and change the discriminatory organizational practices and conditions that initially produce victims. See United States Commission on Civil Rights, Affirmative Action in the 1980s: Dismantling the Process of Discrimination 41 (1981) (Commission Report); cf. Firefighters Local Union No. 1784 v. Stotts, — U.S. —, 104 S.Ct. 2576, 2588, 81 L.Ed.2d 483 (1984) (courts can override seniority systems only for each member of minority plaintiff class who was an actual victim of the discriminatory practice and on whom the practice had an individual impact); see generally 29 C.F.R. § 1608.4 (1984). Thus, plans must arise in a remedial context: they must respond to a deep historical problem, rather than to random, insensitive acts of individuals over the years. See, e.g., Weber, 443 U.S. at 198 n. 1, 99 S.Ct. at 2725 n. 1 (citing deliberate racial exclusion as a legitimizing precondition for affirmative action plans in the construction industry, among others).
A judicial inquiry demands no official, administrative or judicial determination of discrimination. See Local Union No. 35 v. City of Hartford, 625 F.2d 416, 422 (2d Cir.1980), cert. denied, 453 U.S. 913, 101 S.Ct. 3148, 69 L.Ed.2d 997 (1981). Nevertheless, the employer must be able to point to past or present discriminatory patterns and practices that created the traditional segregation of the job categories in question. Merely “arguable” violations are not enough. Cf. Weber, 443 U.S. at 211, 99 S.Ct. at 2731 (Blackmun, J., concurring). Past discrimination implies some discriminatory cause for the present underrepresentation in the employer’s job category and allows the court to go beyond the effects of socially created, accidental differences. Remedies affect causes; as a remedy, a plan must bear a relationship to the causes of discrimination so that it can change the discriminatory organizational practices that initially created victims. See, e.g., Commission Report, supra, at 41.
In this regard, statistics that show a conspicuous work force imbalance help greatly: they function as quantitative clues that “raise questions rather than settle them” and pinpoint areas requiring “further investigations into the factors that produce the statistical profile.” Commission Report, supra, at 36. These statistical studies, of course, should adjust for demographically relevant variables in their comparisons between the work force and local labor pool. See, e.g., United States v. City of Miami, 614 F.2d 1322, 1352-53 & n. 5 (5th Cir.1980) (Gee, J., dissenting), modified on other grounds, 664 F.2d 435 (1981) (en banc). Because plans are remedial, however, an employer cannot rely on statistics alone as a short-cut around the critical need for causal evaluation and analysis. See Commission Report, supra, at 36. Therefore, I consider the majority’s approach too superficial for the analysis required under the first prong of Weber, and disagree with it.
B.
The second inquiry is whether the plan unnecessarily trammels the interests of non-minority employees. A bare affirmation that the more reasonably the plan relates to its remedial purpose in specifics, *1320see, e.g., Ende v. Board of Regents of Northern Illinois University, 565 F.Supp. 501, 507 & n. 3, 509 (N.D.Ill.1983), the less chance it will so invade “the interests of majority workers that it violates Title VII,” Parker v. Baltimore & Ohio Railroad Co., 652 F.2d 1012, 1016 (D.C.Cir.1981), is insufficient. We must scrutinize carefully the reasonableness of the plan’s measures to prevent the majority employees from having their “rights ... bartered away by the employer in order to buy his private peace.” City of Miami, 614 F.2d at 1352 n. 3 (Gee, J., dissenting). At the most obvious level, therefore, a plan may not simply “require the discharge of white workers and their replacement with new black hirees.” Weber, 443 U.S. at 208, 99 S.Ct. at 2730; accord La Riviere, 682 F.2d at 1279. More generally, the plan may not unnecessarily close off channels of employment to non-minorities. See Parker, 652 F.2d at 1014. A court’s serious evaluation of a plan’s measures can settle such questions as whether it in reality uses a “quota” rather than a “goal,” and can establish the reasonableness of the plan with respect to non-minority employees. The record does not demonstrate such an inquiry. Therefore, I conclude that the majority is not in a position to pass on this issue.
C.
The third issue to be addressed is whether the plan absolutely bars the advancement of non-minority employees. This inquiry is closely related to the one above, and the answer will depend on several factors, including the number and nature of the applicants, the number of contested jobs available, and employee expectations. The affirmative action program in Weber did not deprive white employees of opportunities for advancement, but created new opportunities by establishing a training program for both blacks and whites. Weber, 443 U.S. at 198, 99 S.Ct. at 2724. Thus, although “black employees received] training in preference to senior white employees,” id. at 199, 99 S.Ct. at 2725, the expansion of opportunities previously limited by seniority prevented any problem. See also La Riviere, 682 F.2d at 1280. Under Weber, the safest plans minimize non-minority employee displacement and do not concentrate the costs of affirmative action “upon a relatively small, ascertainable group of non-minority persons.” EEOC v. Local 638, 532 F.2d 821, 828 (2d Cir.1976); cf. Gewirtz, Remedies and Resistance, 92 Yale L.J. 585, 605 (1983) (effective remedies must operate within third-party cost constraints). The district court did not analyze this issue sufficiently, and I would have remanded this case for a more detailed consideration.
D.
The final question is whether the plan remains only a temporary measure. The majority concluded that the plan was temporary because it did not state expressly that it would be permanent, and purported to attain rather than to maintain a sexual balance. Thus, the majority concludes that Weber endorses all plans which do not admit to being permanent, but which end when their goals are met. This interpretation of Weber is too broad, and reads the temporary requirement out of the test by making it a hurdle over which any artfully drawn plan can jump. The saving grace of the plans in Weber and La Riviere was their expressly temporary duration. See Weber, 443 U.S. at 208-09, 99 S.Ct. at 2729-30; La Riviere, 682 F.2d at 1280. Moreover, a broad reading of Weber is unnecessary in this case. The disputed facts in the record, particularly Graebner’s testimony, does not justify the majority’s conclusion that the plan was temporary. An appellate court is not in a position to make this factual determination. The district court should have the opportunity to weigh this evidence properly, and I would remand this case for further factual hearings.
IV
In conclusion, the record demonstrates that the district court evaluated the Agency's 1978 plan, although it did consider *1321newer Agency and county-wide plans. The relevant plan, of course, is the one the Agency applied to the allegedly adverse employment decision. I would have remanded this case for the district court to reconsider the plan as an affirmative defense of the Agency under the proper allocation of burdens discussed above. The district court then should have reviewed the aspects of the appropriate plan under a detailed Weber analysis to create an appropriate record for review. See Parker, 652 F.2d at 1020; cf. Pullman-Standard v. Swint, 456 U.S. 273, 291, 102 S.Ct. 1781, 1791, 72 L.Ed.2d 66 (1982) (when “a district court has failed to make a finding because of an erroneous view of the law, ... there should be a remand for further proceedings to permit the trial court to make the missing findings”).
I concur in the result reached by the majority opinion because I believe that the district court erred by misallocating the burdens of persuasion and proof, and a judgment based on such a record should not stand. I cannot join this opinion, however, because I disagree with the majority’s • allocation of the burdens of persuasion and production, and because I believe that the district court should have the first opportunity to correct its own errors and conduct the intensive factual analysis necessary to resolve these issues. In order for the majority to uphold the plan, it has taken several unnecessary doctrinal steps that weaken the Supreme Court’s test in Weber as interpreted by La Riviere. From these pronouncements, I dissent.