The named plaintiffs in this class action suit are former employees at several salmon canneries in Alaska. They brought this action against their former employers, Wards Cove Packing Company, Inc. (“Wards”), Castle & Cooke, Inc. (“Castle”), and Columbia Wards Fisheries (“Columbia”), charging employment discrimination on the basis of race in violation of Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq., and the Civil Rights Act of 1866, 42 U.S.C. § 1981. The class is defined as all nonwhites who are now, will be, or have been at any time since March 20, 1971, employed at any one of five canneries. The individual canneries under scrutiny are Wards Cove and Red Salmon (operated by Wards), Bumble Bee (operated by Castle), and Ekuk and Alitak (run by Columbia).
The Title VII claims against Wards and Columbia were dismissed for lack of jurisdiction early in the proceedings. This court affirmed the dismissal as to Columbia, but reversed the decision as to Wards. Atonio v. Wards Cove Packing Co., Inc., 703 F.2d 329 (9th Cir.1983). On remand, and following a bifurcated liability trial, the district court held for defendants. The class appeals.
BACKGROUND
The five canneries are located in remote and widely separated areas of Alaska. They only operate for a short period each year, during the summer salmon runs. For the remainder of the year they lie vacant. *1123The salmon runs themselves are inherently unpredictable. Due to this fact, the number of canning lines to operate at each facility may vary from year to year, and in any given year a particular facility may not operate at all. Correspondingly, the num-. ber of employees needed to staff a particular cannery in any given year varies with the size of the salmon run. As early and as much as possible each winter, the companies attempt to gauge the size of the anticipated fish run for the upcoming season, and likewise the number of employees that will be needed. In making this assessment, management relies in part on forecasts provided by the Alaska Department of Fish and Game and the Fisheries Research Institute at the University of Washington. Despite these efforts, the actual fish run frequently varies to a considerable degree from the forecasts.
Each year the actual cannery operations begin in May or June, a few weeks before the anticipated fish run, with a period known as the preseason. Workers are brought in to assemble the canning equipment, repair the facilities from the winter damage, and generally prepare the entire cannery for the onset of the canning season. The district court found that many preseason job positions require a variety of skills and skill levels, and that there is too little time during the preseason to train unskilled workers for the skilled jobs.
Shortly before the fishing begins, the cannery workers arrive. Cannery workers, who comprise the bulk of the summer work force, are the individuals who staff the actual canning lines. The cannery worker position is unskilled. These workers remain at the cannery as long as the salmon run produces fish to be canned, and they are guaranteed payment for a minimum number of weeks if the run proves to be a short one. In turn, when the canning is completed, the cannery workers depart and the canneries are disassembled and winterized by postseason workers.
Salmon are extremely perishable and must be processed within a short time after being caught. Since the fish runs themselves are of short duration, cannery operations are often characterized by intense work and long hours. All the while, the Food and Drug Administration monitors the canning process closely, to ensure a safe consumer product. Basically, the canning process proceeds as follows. Independent fishermen catch the salmon and turn them over to company-owned boats called “tenders,” which transport the fish from the fishing grounds to the canneries. Once at the cannery, the fish are eviscerated, the eggs pulled, and they are cleaned. Then, operating at a rate of approximately four cans per second, the salmon are filled into cans. Next, the canned salmon are cooked under precise time/temperature requirements established by the FDA, and the cans are inspected to ensure that proper seals are maintained on the top, bottom and sides.
In addition to the cannery workers, each cannery staffs a variety of job classifications. Machinists and engineers are hired to maintain the smooth and continuous operation of the canning equipment. Quality control personnel conduct the FDA-required inspections and recordkeeping. Tenders are staffed with a crew necessary to operate the vessel. A variety of support personnel are employed to operate the entire cannery community, including, for example, cooks, carpenters, store-keepers, bookkeepers, beach gangs for dock yard labor and construction, etc. The nature of the industry is such that most of the jobs are seasonal and of short duration. The few employees that can be considered permanent or year-round consist of certain management and office personnel who staff the home offices in Seattle and Astoria, Oregon in the winter, and several machinists, carpenters and tendermen who maintain the winter shipyard in Seattle. The remainder of the employees needed for the summer canning season are hired beginning in the first few months of each year.
Due to the geographical realities, the companies must hire the necessary employees from various areas, primarily Alaska *1124and the Pacific Northwest. Nearly all employees are transported to and from the canneries by the companies each year, where they are housed and fed throughout the season. A few Alaska Native employees are able to reside in villages located near some canneries.
During their tenure, the appellants were primarily employed as cannery workers, the lowest-paid positions at the canneries. Appellants' discrimination complaints center on the fact that nearly all cannery worker positions are filled by nonwhites, while the higher-paying job classifications are predominantly white. This disparity, appellants allege, is due to hiring and promoting practices that allow intentional discrimination and produce a discriminatory impact as well. To illustrate these charges, appellants launched a wide-scale attack on the employee selection methods and the housing and messing practices used by the companies.
Among the practices challenged is the apparent lack of objective qualifications for many job classifications, and the use of subjective criteria in hiring and promoting. When filling most job positions, the respective hiring officers generally seek to hire the individuals who are, in the hiring officer’s opinion, the best for the job. Each different job classification naturally re-' quires the officer to consider the needs peculiar to that job. The district court found that the various job classifications at the cannery are not fungible, and that the most important qualifications for many of them, excluding cannery worker positions, are skill and/or experience. The court also found that the necessary skills are not readily acquirable during the season, primarily due to the time restrictions involved, and that cannery worker jobs do not provide training for other positions. Further, the district court found that preseason availability is a necessary qualification for many of the positions, but that it is never a requirement for cannery worker jobs.
The appellants also attacked the recruitment of employees for different jobs through separate channels. The great majority of cannery workers are hired from native villages in Alaska and through a primarily Filipino ILWU local in Seattle. Consequently, the cannery worker department is staffed almost entirely by these ethnic groups. Openings in other positions are not posted at the canneries, and the companies do not promote from within during the season. Instead, the companies fill other positions each year through applications received during the off-season at the mainland home offices, through re-hiring previous employees in those positions, and through word-of-mouth recruitment. Appellants also allege that nepotism is rampant in the canneries, with relatives of white company employees being given preference in hiring. Finally, appellants allege that nonwhites are segregated from whites in housing and messing, and that the bunkhouses and food provided for the nonwhites is far inferior to that provided for whites.
In holding for the defendant companies, the district court evaluated the mass of evidence introduced by both sides, including conflicting statistical data. The court analyzed all of appellants’ claims for intentional discrimination, concluding that the companies had successfully shown nondiscriminatory motivations. The court refused, despite appellants’ arguments to the contrary, to evaluate all of the claims under the disparate impact model of Title VII, relying on authority from this circuit. A few claims were subjected to disparate impact scrutiny, however, and the court again found for the defendants. Before this court, the appellants challenge these findings and raise a host of subsidiary issues.
ANALYSIS
I. Columbia Joint Venture Claims
Wards and Castle operate Columbia as a joint venture. Earlier in these proceedings, we affirmed the dismissal of Title VII claims against Columbia because they were not filed within the prescribed time limits and were, therefore, time-barred. See Atonio v. Wards Cove Packing Co., Inc., 703 F.2d 329, 332 (9th Cir.1983). Appellants now assert that dismissal of the *1125claims against the joint venture for procedural reasons does not affect the liability of the joint venturers as to those claims. Therefore, argue appellants, because they could have sued either or both of the joint venturers without suing the joint venture, the Title VII claims against Columbia can be asserted against Wards and Castle, both-of which were timely sued on separate discrimination charges.
We have no trouble agreeing that general common law agency principles, including joint and several liability, are applicable in Title VII cases. So too, however, are basic procedural and jurisdictional principles applicable. The controlling fact here, which appellants ignore, is that the Title VII claims against Columbia were not filed in time to grant jurisdiction. Nor were they ever filed against Wards or Castle in their capacity as joint venturers for Columbia. The claims were properly dismissed as untimely, and they simply no longer exist. Appellants cannot now evade the jurisdictional prerequisites by bringing these claims in through the back door.
II. Intentional Discrimination
The district court correctly recognized that while a plaintiff suing under section 1981 must always prove intentional discrimination, such is not always the case with Title VII. Of the two Title VII theories of liability, only disparate treatment requires a showing of intentional discrimination in order to establish a prima facie case. The alternate theory, disparate impact, requires no proof of intent and, logically enough, good intent is not a defense in impact cases.
Due to their inherent similarities, we can treat section 1981 and Title VII disparate treatment under one intentional discrimination analysis. The plaintiff in such a case has the initial burden of proving a prima facie case of intentional discrimination. If successful, the burden of production then shifts to the defendant to articulate some legitimate nondiscriminatory reason for the plaintiffs rejection. If the defendant carries this burden, the plaintiff can still prove by a preponderance of the evidence that the legitimate reasons offered were a pretext for discrimination. The plaintiff always has the burden of persuasion. Texas Dept. of Community Affairs v. Burdine, 450 U.S. 248, 252-256, 101 S.Ct. 1089, 1093-1095, 67 L.Ed.2d 207, 215-217 (1981); Kimbrough v. Sec. of U.S. Air Force, 764 F.2d 1279, 1283 (9th Cir.1985). In practical application, this allocation of burdens works to the effect that “after plaintiffs prima facie case and defendant’s ‘articulation,’ the trier of fact decides the question of discrimination based on the entire case.” Kimbrough, at 1283.
“After a Title VII case is fully tried, we review the decision under the clearly erroneous standard applicable to factual determinations.”. Kimbrough, at 1281; Anderson v. Bessemer City, — U.S. -, -, 105 S.Ct. 1504, 1511, 84 L.Ed.2d 518, 528 (1985). The “ ‘district court must decide which party’s explanation of the employer’s motivation it believes.’ We will reverse that factual determination only if it is clearly erroneous ... and we will not ransack the record, searching for mistakes.” Casillas v. United States Navy, 735 F.2d 338, 342-343 (9th Cir.1984) (quoting United States Postal Service v. Aikens, 460 U.S. 711, 103 S.Ct. 1478, 75 L.Ed.2d 403 (1983)).
Before addressing appellants’ several allegations of error, it should be noted that appellants have relied heavily on this circuit’s decision in Domingo v. New England Fish Co., 727 F.2d 1429 (9th Cir.1984), modified, 742 F.2d 520 (1984). Domingo was originally brought as a companion case to the action at bar. Because of pre-trial delays in the instant action, including the earlier appeal, Domingo proceeded to trial and appeal much faster. The facts in Domingo are strikingly similar to the facts at bar, as are the claims of the plaintiffs. Both cases involve racial discrimination charges in the Alaska canning industry. Domingo resulted in a decision in favor of the class, and the present plaintiffs have cited that decision extensively. We do not, *1126however, feel compelled to blindly apply stare decisis. Although the similarities between the cases are striking, the differences between them are far more determinative.
In Domingo, we noted that the liability phase of the trial lasted one and one-half days, with the defendant company not challenging the plaintiffs’ statistics nor rebutting the plaintiffs’ prima facie case. 727 F.2d at 1433-1434. In essence, the defendants in Domingo did not defend against the allegations of discrimination. Conversely, the defendants at bar are different canning companies and they have defended themselves against these charges vigorously. The liability phase of the trial took twelve days, with the defense introducing witnesses and statistical evidence contrary to that of the plaintiffs. Taking these important factual differences into consideration, Domingo is entitled to no more or no less precedential value than the many other Ninth Circuit cases in this area of law.
A. Hiring, Promoting, Paying, Firing
When confronting the bevy of evidence below, the district court began its intentional discrimination inquiry by dividing the at-issue (non-cannery worker) jobs into two groups: skilled and unskilled. Both the plaintiffs and defendants had introduced labor-market statistics in an effort to bolster their contentions. For reasons discussed below, the district court rejected plaintiffs’ labor-market statistics, while crediting those of defendants. In addition, the plaintiffs introduced comparative statistics, which the court only credited in scrutinizing the unskilled jobs group.
Taking each group in turn, the court first found that the unskilled jobs were fungible, and, thus, comparative statistics were appropriate for use in establishing a prima facie case of discrimination. Since the comparative statistics showed a pattern of job segregation throughout the cannery work forces, the court found that the plaintiffs had put on a prima facie case with respect to the unskilled jobs. Nevertheless, for reasons discussed below, the court found that defendants had met their burden of production in showing motivation other than discriminatory animus, and that plaintiffs had failed in their ultimate burden of proving pretext.
Moving on to the skilled positions, the district court had more difficulty in finding a prima facie case of intentional discrimination, because the plaintiffs’ statistical evidence had been discredited. The court did find a “marginal” prima facie case, but only by way of combining all of plaintiffs’ evidence and claims of nepotism, individual instances of alleged discrimination, deterrence, failure to post openings, general lack of objective qualifications, lack of a formal promotion procedure, re-hiring past employees in their old jobs, and the discredited statistical evidence. The court found that none of these had significant probative value when considered alone. In conclusion, the court held that defendants had met their burden of production and that plaintiffs had failed to meet their ultimate burden of persuasion.
Appellants contend that the district court erred in not giving more credit to their evidence of nepotism. The district court noted that “[relatives of whites and particularly nonwhites appear in high incidence at the canneries. However, defendants have established that the relatives hired in at-issue jobs were highly qualified for the positions in which they were hired and were chosen because of their qualifications.” The court also found that plaintiffs’ statistics failed to recognize that a number of persons became related through marriage after starting work at the canneries, and that the testimony showed “that numerous white persons who ‘knew’ someone were not hired due to inexperience, and whites hired were paid no more than non-whites.” Therefore, the court concluded that there existed no “preference” for relatives at the canneries.1
*1127After carefully reviewing the record, we cannot say that the district court was clearly erroneous in making these findings. The Supreme Court has recently reiterated our role in reviewing these findings of fact. “If the district court’s account of the evidence is plausible in light of the record viewed in its entirety, the court of appeals may not reverse it even though convinced that had it been sitting as the trier of fact, it would have weighed the evidence differently. Where there are two permissible views of the evidence, the factfinder’s choice between them cannot be clearly erroneous.” Anderson v. Bessemer City, — U.S. at -, 105 S.Ct. at 1512, 84 L.Ed.2d at 528 (emphasis added). The fact finder’s account of the evidence, concluding that there were legitimate and nonpreferential reasons for the hires of friends and relatives, is entirely plausible in light of the whole record. Consequently, we will not disturb it.
For the same reasons, we will not overturn the district court’s findings with respect to alleged individual instances of discrimination. A number of plaintiffs alleged that they were either overtly discriminated against in the hiring for at-issue positions, or that they were deterred from seeking at-issue positions because of the defendants’ alleged history of pervasive discrimination. Using the four-part test of McDonnell Douglas as a guideline,2 the district court did not give greater credit to the alleged instances because it found that the respective plaintiffs had not been hired for nondiscriminatory reasons. Primarily, the district court found that the individuals had made oral inquiries, which were not considered applications, or that the applications were untimely.3 The court also found that some applicants had been unavailable for preseason work and, therefore, unavailable for the positions they desired. There is ample evidence in the record to support the district court’s findings regarding these individual claims.
Nevertheless, appellants argue that the fact that the companies use separate hiring channels, word-of-mouth recruitment, and fail to announce vacancies should serve to excuse appellants from the necessity of establishing the timeliness of their applications and automatically elevate oral inquiries to the status of applications. We disagree. Appellants take this idea from a discussion of damages issues in Domingo. 727 F.2d at 1445. We find that discussion inapposite because, unlike the Domingo plaintiffs, the appellants have not first established wide-ranging discrimination. Appellants failed to convince the district court that they had been intentionally discriminated against, and they may not rely on Domingo in this manner to establish what they have failed to prove. We cannot find the district court clearly erroneous.
Appellants also contend that the district court erred in failing to credit their comparative statistics when analyzing the skilled positions. As previously indicated, the district court accorded these statistics, which compare the racial composition of the vari*1128ous job categories, little probative value because they did not reflect the number of employees possessing the requisite skills or those available for preseason work.
This court has recognized the importance of statistics as circumstantial evidence of discriminatory intent. In the same breath, however, the court often admonishes that statistics are “inherently slippery” and the weight given to them depends on “proper supportive facts and the absence of variables.” Spaulding v. University of Washington, 740 F.2d 686, 703 (9th Cir.), cert. denied, — U.S. -, 105 S.Ct. 511, 83 L.Ed.2d 401 (1984) (quotations omitted). The district court’s evaluation of conflicting statistics and determination of the probative weight they are to be accorded is a factual inquiry. Accordingly, we apply the clearly erroneous standard of review. Gay v. Waiters’ and Dairy Lunchmens Union, 694 F.2d 531, 550 (9th Cir.1982); see also Allen v. Prince George’s County, Md., 737 F.2d 1299, 1303 (4th Cir.1984).
Appellants fail to recognize the importance of minimizing variables to increase the reliability and significance of statistical evidence. In Domingo, we said that “[i]n many cases, it is necessary to consider the qualifications of the applicant pool because without that information, no inference of discrimination may be drawn; the lack of minority representation in the workforce might simply be due to a lack of qualified applicants.” 727 F.2d at 1436. Although we permitted the Domingo district court to credit comparative statistics in that case, it was because sufficient evidence of discriminatory treatment had already been presented, and the statistics were not necessary to raise an inference of discrimination. Id. We allowed them merely to demonstrate the consequences of the defendant’s already-proven discriminatory hiring practices.
The appellants at bar, however, have not previously presented sufficient evidence of discriminatory intent, and they desperately need these comparative statistics credited for that very purpose. This is a precise example of the type of putting the cart in front of the horse of which we were wary in Domingo. Appellants cannot use these general, unrefined statistics to meet their burden with respect to skilled positions. This ease clearly illustrates why courts and litigants must carefully examine proffered statistics to avoid the distortion of fact that they have the potential to produce.
The percentage of nonwhites employed in the Alaska salmon canning industry during the relevant time period was approximately 50 percent. Of these, approximately 88 percent were Alaska Natives or of Filipino descent. It is undisputed that the racial composition of cannery workers is predominantly nonwhite, and, therefore, those positions are primarily held by Filipinos and Alaska Natives. We know that this is because four of the canneries, Ekuk excluded, have a contract with Local 37 in Seattle to supply cannery workers, and we know further that Local 37 membership is predominantly Filipino. We also know that the Alaska Native cannery workers primarily come from sparsely populated areas immediately adjacent to four of the canneries.
Yet, the district court found that Filipinos constitute only about 1 percent of the population and labor force in the geographical region from which the canneries draw employees. Further, the district court found that Alaska Natives constitute only a small portion of the overall general population in the section of Alaska where canneries are located, which includes predominantly white city populations. From this comparison, it could easily be deduced that Alaska Natives and, more particularly, Filipinos are significantly overrepresented in the cannery worker jobs.
On the other side of the coin, the district court found that the available general labor supply in the relevant geographic area was approximately 90 percent white. And, it is undisputed that the majority of at-issue jobs were held by whites.
With this background in mind, it is obvious that the institutional factors involved tend to distort the racial composition of the work force. Thus, when considering the *1129skilled positions, statistics which merely highlight the segregation of whites and nonwhites between the at-issue and cannery worker jobs, without more, cannot serve to raise an inference that the segregation is attributable to intentional discrimination against any particular race. They can, as Domingo pointed out, serve to demonstrate the consequences of discriminatory practices which have already been independently established.
When jobs are not fungible, as in this case, statistics must reflect the qualifications of the applicant pool in order to be probative and credible on the discrimination issue. The fact that the qualifications themselves are subjective does not obviate this requirement as a matter of law. In this case, the district court found that the qualifications most needed for the skilled positions were skill and/or experience in performing the respective jobs. Certainly, there is a degree of subjectivity present when an employer chooses the applicant that he or she feels is best qualified. But it is not necessary that plaintiffs’ statistics show that they were the best qualified. It is enough that they reflect the percentage of qualified nonwhites — in this case, those with some skill and/or experience in the desired jobs and who were available to begin work in the preseason. Whether or not such statistics have sufficiently reflected the minimum qualifications actually imposed by the employer so as to raise an inference of intentional discrimination is then a question of fact left for the fact finder. For these reasons, we do not hesitate to find that the district court did not clearly err in assigning appellants’ comparative statistics little probative value as to the skilled jobs.
The appellants further allege the district court erroneously held that the labor-market statistics offered by the defendant companies rebutted the appellants’ prima facie case of intentional discrimination. The district court, however, did not so hold. It found defendants’ labor-market statistics more probative than those of appellants because appellants’ statistics had counted re-hires of employees during successive seasons and at successive canneries within the same season. Important considerations apart from the statistics played a determinative part in the court’s conclusion that defendants had met their burden of production. As previously discussed, the court concluded from the evidence that all applicants were evaluated according to job-related criteria, albeit subjectively, and that oral inquiries and untimely applications served to eliminate hopeful employees, including some plaintiffs. Thus, appellants are incorrect in their basic assertion.
We further cannot find the district court clearly erroneous in its findings concerning job-related criteria. Appellants assert that the criteria were never imposed. The district court found otherwise. In so doing, the court took certain listed job qualifications verbatim from the defendants’ pre-trial order. These lists, however, merely supported the court’s conclusion that skill and/or experience were the general qualifications looked for in the hiring of employees for the specified jobs. After reviewing the record, we cannot conclude that the district court was clearly erroneous. See Anderson v. Bessemer City, — U.S. at -, 105 S.Ct. at 1510, 84 L.Ed.2d at 527.
Appellants also urge reversal on the ground that the district court’s findings failed to address the discriminatory nature of separate hiring channels and word-of-mouth recruitment. We decline to do so. Findings of fact are adequate if they are explicit enough on the ultimate issues to give this court a clear understanding of the basis of the decision and to enable us to determine the grounds on which the trial court reached its decision. Nicholson v. Board of Educ., etc., 682 F.2d 858, 866 (9th Cir.1982). See also Barber v. United States, 711 F.2d 128, 130-131 (9th Cir.1983); United States v. Alpine Land & Reservoir Co., 697 F.2d 851, 856 (9th Cir.) cert. denied, — U.S. -, 104 S.Ct. 193, 78 L.Ed.2d 170 (1983); De Medina v. Reinhardt, 686 F.2d 997, 1011-1012 (D.C.Cir.1982).
*1130The ultimate fact, that there existed no pattern or practice of discrimination in hiring, promoting, paying and firing, is supported by the numerous subsidiary findings of the district court. Throughout the findings, the court discusses the manner in which the canneries hire and promote employees. Included are findings about the fact that cannery workers are hired routinely through Local 37, but that skilled positions are filled through individual screening. It would have been convenient had the district court labelled certain findings as addressing “separate hiring channels” and “word-of-mouth” recruitment. It is inconsequential in the end, however, because it is abundantly clear from the district court’s opinion that these challenged practices were included in the ultimate finding. The court stated, “regardless of the manner in which a prospective employee came to the attention of the hiring personnel, the person was evaluated according to job related criteria.” Thereafter, in concluding the case, the court encompassed all of the claims when it said “defendants did not discriminate in the hiring, firing, promoting, or paying ...” The decision of the district court will not be disturbed.
B. Housing
The vast majority of cannery employees live at the canneries during the season in bunkhouses provided by the companies. The appellants claimed that nonwhites, particularly Filipinos, were segregated from whites and placed in inferior bunkhouses because of racial discrimination. The district court found that appellants had established prima facie case of intentional discrimination, but that the defendants’ evidence proved nondiscriminatory motivations which the appellants had failed to prove pretextual. Specifically, the court found that the employees were housed by time of arrival and by crew.
The record contains ample evidence to affirm the district court’s conclusion. While none of the cannery housing appears to have been luxurious, some bunkhouses were undoubtedly better than others. Testimony showed that workers who arrived at the canneries first, during the colder preseason period, were housed together in the best-insulated buildings. When the cannery workers eventually arrived, they were housed together in the remaining bunkhouses. This system enabled the companies to maintain only the amount of housing needed at any particular time. Furthermore, the workers were housed primarily by crew, thereby minimizing any inconvenience occasioned when different departments worked different shifts. While some mixing of crews did occur, the cannery workers were all housed together, regardless of race.4 Based on our review of the record, we do not find the district court clearly erroneous.
C. Messing
Cannery workers were also fed separately from the remainder of the work force. The appellants alleged that this was due to racial discrimination. The district court agreed that they had established a prima facie case of intentional discrimination, but that the defendants had met their burden of production and the appellants had not proved pretext. It is undisputed that the cannery worker mess halls served what is termed as a “traditional” oriental menu. The district court noted that the Local 37 contract provided for a separate culinary crew, and that Filipino and Asian persons dominated the membership in Local 37. Further, the court found that the quality and quantity of food served in the respective mess halls were the responsibility of the respective cooks, and that the older cannery workers preferred the traditional menu, to which the younger workers acceded. The court concluded that complaints about the food were attributable to personal taste, and that the segregated *1131messing arrangement was attributable to the union and not the conduct of defendants. There is support in the record for these findings, and we cannot find them clearly erroneous.
III. Disparate Impact
From the beginning, appellants have insisted that their claims also be analyzed under the disparate impact model of Title VII. Impact analysis exemplifies the broad prophylactic purpose of Title VII, which is designed to achieve equality of employment opportunities by removing artificial barriers that act as “built-in head winds” against the progress of minority groups. Connecticut v. Teal, 457 U.S. 440, 447-448, 102 S.Ct. 2525, 2530-2531, 73 L.Ed.2d 130, 137-138 (1982); Griggs v. Duke Power Co., 401 U.S. 424, 431-432, 91 S.Ct. 849, 853-854, 28 L.Ed.2d 158, 164-165 (1971). To make out a prima facie impact case, the plaintiff must show a facially neutral employment practice that has a ‘‘significantly discriminatory” impact upon a Title VII protected group. Connecticut v. Teal, 457 U.S. at 446, 102 S.Ct. at 2530, 73 L.Ed.2d at 137. It is not necessary to prove discriminatory intent. International Brotherhood of Teamsters v. United States, 431 U.S. 324, 335 n. 15, 97 S.Ct. 1843, 1854 n. 15, 52 L.Ed.2d 396, 415 n. 15 (1977); Moore v. Hughes Helicopters, Inc., 708 F.2d 475, 481 (9th Cir.1983). The burden of proof then shifts to the defendant to establish that the practice has “a manifest relationship to the employment in question,” Griggs, 401 U.S. at 432, 91 S.Ct. at 854, 28 L.Ed.2d at 165, or is justified by a business necessity, Moore, 708 F.2d at 481. “The employer may also rebut the employee’s prima facie case by showing the inaccuracy of the employee’s statistics.” Id. (citing Contreras v. City of Los Angeles, 656 F.2d 1267, 1273 (9th Cir.1981), cert. denied, 455 U.S. 1021, 102 S.Ct. 1719, 72 L.Ed.2d 140 (1982)). The plaintiff may still prevail by showing “that the employer was using the practice as a mere pretext for discrimination,” Connecticut v. Teal, 457 U.S. at 447, 102 S.Ct. at 2530, 73 L.Ed.2d at 137, “or that the employer’s purpose could be served by selection devices with less discriminatory impact,” Moore, 708 F.2d at 481 (citing Dothard v. Rawlinson, 433 U.S. 321, 329, 97 S.Ct. 2720, 2726, 53 L.Ed.2d 786, 797 (1977)).
The district court applied impact analysis to appellants’ claims of nepotism,5 but declined to do so for the balance of appellants’ discriminatory hiring and promoting claims. The court noted a conflict in this circuit concerning whether impact analysis is proper in situations where employees are challenging the subjective nature of employment practices. Caught in the bind, the district court wisely chose to follow the precedent authority by refusing to use impact analysis across the board. Appellants challenge this legal decision, and we must attempt to resolve the problem de novo.
Griggs v. Duke Power, supra, was the first case to recognize that Title VII outlaws practices that are fair in form, but discriminatory in operation and impact. From this case grew the disparate impact model, challenging employment practices that are neutrally applied (thus making discriminatory intent difficult to prove), but that nevertheless operate to discriminate in effect. Examples of the type of objective, outwardly neutral employment practices clearly susceptible to impact scrutiny are pre-employment tests that adversely affect people of certain cultural backgrounds and pre-selection requirements such as height and weight restrictions. See Griggs, 401 U.S. 424, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971); Dothard, 433 U.S. 321, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977).
By and large, however, appellants have not challenged a specific facially neutral practice. Rather, appellants have mounted a broad-scale attack against the gamut of defendants’ subjective employment practices. We have firmly stated that subjective *1132practices are particularly susceptible to discriminatory abuse and should be closely scrutinized. Kimbrough, at 1284. At the same time, it is certain that subjective practices and decisions are not illegal per se. Id.; Heagney v. University of Washington, 642 F.2d 1157, 1163 (9th Cir.1981). The gray area of conflicting opinion is whether the close scrutiny of these practices can or should take the form of a disparate impact analysis.
In Heagney, upon which the district court relied, the plaintiff challenged the University’s power to classify certain jobs as “exempt” from state personnel laws, which, in turn, gave the school more discretion in setting salaries. We held that the crux of the complaint was an objection to the lack of well-defined criteria, which could not be equated with practices such as personnel tests or minimum physical requirements. Thus, impact analysis was inappropriate. 642 F.2d at 1163. O’Brien v. Sky Chefs, Inc., 670 F.2d 864 (9th Cir.1982), followed on the heels of Heagney, and once again held that the lack of well-defined criteria must be challenged under disparate treatment.
The conflict in this circuit arose with the decision in Wang v. Hoffman, 694 F.2d 1146 (9th Cir.1982), which challenged the manner in which the Army Corps of Engineers hired and promoted employees. Without cite to Heagney or O’Brien, oar court applied impact analysis to the plaintiff’s claim that the subjective selection process used by the Corps provided inadequate guidelines and could be manipulated in order to eliminate certain candidates. Although a language skills requirement— traditionally subject to impact analysis because it is objective and facially neutral— appeared central to the court’s concerns, the discussion reflects a deeper worry that the language requirement was added intentionally to disadvantage the plaintiff. 694
F.2d at 1149. Unfortunately, this speaks to intent, which is irrelevant in impact cases. Nevertheless, Wang seems to support the appellants’ arguments at bar. See also Peters v. Lieuallen, 746 F.2d 1390 (9th Cir.1984).
In subsequent cases we have recognized the conflict between Heagney and Wang, and expressed opinion without resolving the question. See Moore, 708 F.2d 475 (9th Cir.1983) (although unnecessary to decision, disparate treatment focus better suited to analysis of subjective decision making). Domingo, 727 F.2d 1429 (9th Cir.1984) (disparate treatment more appropriate approach because defendant’s practices were not facially neutral); Spaulding, 740 F.2d 686 (9th Cir.1984) (holding impact analysis only applicable to specific, facially neutral policies, rather than a full-scale challenge to an employer’s practices, of which lack of well-defined criteria is not facially neutral).
We choose to follow the Heagney line of authority because we believe it to be the correct view.6 Without question, employment discrimination is an evil which continues to plague our society and must be battled. Title VII was designed for that purpose, to make race (or sex, etc.) an irrelevant factor in hiring decisions. It must be remembered, however, that the disparate impact model was not explicitly provided for in the statute, but rather was first enunciated in Griggs as a mode of implementing the broad purposes of Title VII. While it has been argued that subsequent congressional actions have served to implicitly ratify the creation of disparate impact,7 it is no less clear that Congress was concerned about mandating colorblindness with as little intrusion into the free market system as possible. Courts have noted that it was deemed essential that employers remain free to set employment qualifications as they honestly saw *1133fit, so long as those qualifications were not based on race, color, religion, sex, or national origin. See Griggs, 401 U.S. at 436, 91 S.Ct. at 856, 28 L.Ed.2d at 167; Contreras, 656 F.2d at 1277-78.
Presumably, therefore, the disparate impact model was created to challenge those specific, facially-neutral practices that result in a discriminatory impact and that by their nature make intentional discrimination difficult or impossible to prove. Were the facial-neutrality threshold to disappear or be ignored, the distinction between disparate impact and disparate treatment would diminish and intent would become a largely discarded element. Rather than being an irrelevant factor as envisioned, race (or sex, etc.) could then become an overriding factor in employment decisions. Employers with work forces disproportionate to the minority representation in the labor force could then face the choice of either hiring by quota or defending their selection procedures against Title VII attack. We do not find such a result has been mandated by Congress or through Supreme Court interpretation of Title VII. Therefore, practices and policies such as a lack of well-defined criteria, subjective decision making, hiring from different sources or channels, word-of-mouth recruitment, and segregated housing and messing, which are not facially neutral, lend themselves far better to scrutiny for intentional discrimination. Consequently, we hold that disparate impact analysis was correctly withheld by the district court when considering these claims.8
Appellants’ nepotism allegations, which we have previously held proper for impact analysis, Bonilla v. Oakland Scavenger Co., 697 F.2d 1297 (9th Cir.1982), cert. denied, — U.S. -, 104 S.Ct. 3533, 82 L.Ed.2d 838 (1984), were properly so considered by the district court. As previously discussed, the court found that no pattern or practice of nepotism existed because there was no preference for relatives. We do not hold those findings to be clearly erroneous. In addition, we find appellants’ remaining allegations of error, concerning re-hire preferences and termination of Alaska natives, to be without merit.
Accordingly, for the reasons set out in this opinion, the district court is
AFFIRMED.
. The district court's analysis of appellants' nepotism claims applies equally under both the disparate impact and intentional discrimination inquiries. Disparate impact is discussed infra.
. It was set out in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802, 93 S.Ct. 1817, 1824, 36 L.Ed.2d 668, 677 (1973), that a plaintiff can establish an inference of discrimination by meeting four criteria. First, that the individual belongs to a Title VII protected class. Second, that he or she applied and was qualified for an open position. Third, that he or she was rejected. Finally, that after the rejection, the employer continued to seek applicants. See also Diaz v. American Telephone & Telegraph, 752 F.2d 1356, 1359 (9th Cir.1985).
Appellants accuse the district court of misapplying the test, by using it too stringently. They are incorrect, apparently due to a misreading of the court’s opinion. Contrary to the accusation, the court clearly said that the McDonnell Douglas elements are not an inflexible formulation, but rather provide some guidance.
. Applications could be untimely if made too early or too late. Testimony showed that some plaintiffs had orally inquired during one season about positions for the next season a year away, and such inquiries were not considered an application unless followed up by a written application to the home office during the winter. Conversely, because the companies generally received far more applications than there were job vacancies, an application was untimely if received after the opening was filled. The district court found that the defendants did not treat whites and nonwhites differently in these respects.
. The only exception to this appears to have been the few female cannery workers, who were housed apart from male cannery workers. White female cannery workers were housed with nonwhite female cannery workers, just as white and nonwhite male cannery workers were housed together.
. The district court also used impact analysis to test an English language requirement and in an alternate holding in favor of defendants on the housing and messing claims. The language requirement finding is not challenged on this appeal, and for reasons discussed infra we will not review the court's impact discussion regarding housing and messing claims.
. Moreover, we agree with the district court. Principled institutional decision-making requires that we adhere to Heagney, the first in line. We believe the other panels acted improperly in ignoring Heagney. It is the law of this circuit by which we are bound until overruled by appropriate en banc proceedings.
. See Helfand & Pemberton, The Continuing Vitality of Title VII Disparate Impact Analysis, 36 Mercer 939, 944-954 (1985).
. In addition to the conflict within this court, the circuit courts of appeals are split on the applicability of disparate impact analysis to subjective employee selection practices. Those which have applied impact analysis are the Fifth, Sixth, Tenth, Eleventh and D.C. Circuits. See Rowe v. General Motors Corp., 457 F.2d 348 (5th Cir.1972); Rowe v. Cleveland Pneumatic Co., Numerical Control, Inc., 690 F.2d 88 (6th Cir.1982); Lasso v. Woodmen of World Life Ins. Co., Inc., 741 F.2d 1241 (10th Cir.1984); Williams v. Colorado Springs School Dist., 641 F.2d 835 (10th Cir.1981); Griffin v. Carlin, 755 F.2d 1516 (11th Cir.1985); Segar v. Smith, 738 F.2d 1249 (D.C.Cir.1984). Those circuits which have said they will only apply impact analysis to specified objective employee selection practices include the Fourth, Fifth, Eighth, and Tenth Circuits. See EEOC v. Federal Reserve Bank, 698 F.2d 633 (4th Cir.1983); Pope v. City of Hickory, 679 F.2d 20 (4th Cir.1982); Vuyanich v. Republic Nat'l Bank, 723 F.2d 1195 (5th Cir.1984); Carroll v. Sears, Roebuck & Co., 708 F.2d 183 (5th Cir.1983); Carpenter v. Stephen F. Austin State Univ., 706 F.2d 608 (5th Cir.1983); Pouncy v. Prudential Ins. Co., 668 F.2d 795 (5th Cir.1982); Talley v. United States Postal Service, 720 F.2d 505 (8th Cir.1983); Harris v. Ford Motor Co., 651 F.2d 609 (8th Cir.1981); Mortensen v. Callaway, 672 F.2d 822 (10th Cir.1982).