ANDERSON, Circuit Judge:

This case involves alleged racial discrimination in the promotion of railroad journeymen/carmen to the position of foreman. Five black journeymen/carmen brought a Title VII employment discrimination action against their employer, Seaboard Coastline Railroad Company ("Seaboard"). The ultimate decision of the district court was that the plaintiffs had not established their disparate impact claim and that only one of the five plaintiffs had proved his disparate treatment claim. On appeal, the plaintiffs challenge the dismissal of the disparate impact claim, the dismissal of the disparate treatment claim as to the other four plaintiffs, and the scope of relief for the prevailing plaintiff. Seaboard cross-appeals the district court's finding of intentional discrimination as to the prevailing plaintiff. We affirm the decision of the district court.

I. BACKGROUND

A. Facts

Andrew Hill, Roosevelt Coleman, Joe Lock, Walter Jones and Freddie Lee were all employed by the Car Department of Seaboard's Tampa Division as journeymen/carmen ("carmen"). Each was passed over for the position of carman/foreman ("foreman"), a supervisory position normally filled from the ranks of the carmen. Promotions to foreman were made by the department's Master Mechanic, R.D. Brigman, who had assumed that position in February, 1980. The Division Superintendent had formal power to disapprove the Master Mechanic's recommendations but in fact did not exercise it. Brigman and all other supervisory personnel with input into the promotion process were white.

The procedure Brigman followed in making promotions consisted of two parts. First, he would examine the seniority roster of available carmen. Although the vacancies were not posted, all of the carmen were automatically considered when a vacancy occurred. Based on Brigman's personal observations and on the informal recommendations of the other supervisors, he would then narrow the list down to the six or eight most qualified candidates. Once the field of candidates was thus narrowed, Brigman made an in-depth comparison based on the personnel files of the employees.

The Master Mechanic considered a number of subjective and objective criteria. Brigman solicited supervisor comments regarding factors such as experience in repairs, familiarity with rules and regulations, intelligence and aptitude, verbal and written communication skills, punctual and regular work attendance, harmonious relationships with supervisors, trustworthiness, reliability, and prior supervisory experience. Brigman also relied on information from the carmen's personnel files relating to initial date of employment, grades on the car inspector exam, the number of personal injuries or accidents, attendance, past disciplinary problems and prior supervisory experience. Brigman did not testify as to the weight given the various factors considered.

During the relevant period, between January 1 and July 1980, at least five foreman positions became available. The carman pool at the time consisted of 267 white employees and 37 black employees. A total of five white employees were promoted, and at least one black employee was offered a promotion. Two whites, Reel and Hower, were promoted on May 1, 1980. One black, Williams, was selected for promotion on May 22, but refused it, because he would lose the ability to earn overtime. This was the first offer of promotion to foreman made to a black employee since at least 1957. A second black employee, Walters, may have been chosen for promotion in May, but the change did not become effective until August. Three more whites, Conrad, Benard and Clark, were promoted in June, although Benard was almost immediately demoted when the union brought his poor record to Brigman's attention. Another black employee, Coleman, was promoted to the foreman position after the EEOC charge had been filed. His temporary position was rescinded when funds were cut back.

B. Procedural History

The plaintiffs' Title VII claims proceeded to a non-jury trial on March 21 and 22, 1983. The employees asserted claims under both disparate treatment and disparate impact theories. Almost all of the evidence at trial consisted of the testimony of Master Mechanic Brigman. At the close of the plaintiffs' case, the district court dismissed the disparate impact claim, on the ground that subjective employment selection criteria cannot provide the predicate for a disparate impact case. The court concluded, however, that all of the plaintiffs had made out a prima facie case of disparate treatment, based on the framework set out in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802, 93 S.Ct. 1817, 1824, 36 L.Ed.2d 668 (1973), and that Seaboard's articulation of a nondiscriminatory reason for the decisions was not sufficient. Hill v. Seaboard Coast Line R. Co., 573 F. Supp. 1079 (M.D.Fla. 1983). However, in a later order on the relief to be granted, the court stated that the plaintiffs had not proved themselves qualified. This apparent contradiction was the subject of a remand, after the first appeal to this court. The district court was instructed on remand to "reconsider its findings on appellants' qualifications and its finding on intentional discrimination based on all the evidence. If the court finds that the appellants were not qualified, then the court must find that there was not intentional discrimination." Hill v. Seaboard Coast Line R. Co., 767 F.2d 771, 775 (11th Cir. 1985). The court of appeals also remanded the disparate impact claim, in light of intervening circuit precedent holding that the disparate impact theory may be used to challenge subjective criteria.

On remand, Hill v. Seaboard Coast Line R. Co., 642 F. Supp. 319 (M.D.Fla. 1986), the district court found that only one of the plaintiffs, Coleman, was qualified, and had thus made out a prima facie case, and that at least one of the whites promoted, Benard, was less qualified. A later order found specifically that none of the other white employees promoted ahead of Coleman was less qualified than he. The court found Seaboard's proffered nondiscriminatory reason for the selection pretextual, holding that the company had discriminated against Coleman. On the disparate impact claim, the court found that none of the criteria, nor the system as a whole, was shown to have an unlawful impact on blacks.

On appeal, the parties assert the following claims. The plaintiffs contend that the district court erred (1) in finding that Hill, Lock, Jones and Lee failed to prove that they were qualified for the promotion; (2) in holding that Seaboard could have rebutted the prima facie cases of these plaintiffs; (3) in finding that Coleman was less qualified than all of the promoted whites except one; (4) in ruling that the plaintiffs had not made out a prima facie case under a disparate impact theory; and (5) by violating the law of the case on remand. Seaboard, on cross-appeal, argues that the district court's conclusion that Coleman was the victim of intentional discrimination was clearly erroneous.

II. DISPARATE TREATMENT CLAIMS

Appellants Hill, Lock, Jones and Lee argue that the district court erred in finding that they did not make out a prima facie case of discriminatory treatment. Because we conclude that the district court's finding that these appellants were not qualified for the foreman position was not clearly erroneous, we affirm the judgment against these appellants. Coleman argues that the district court erred in finding that he did not prove discrimination with respect to the promotions other than the June 23, 1980 promotion of Benard. On cross-appeal, Seaboard claims that the district court erred in finding that the company discriminated with respect to Coleman by promoting Benard. We affirm the district court both with respect to the finding of discrimination in the Benard promotion and with respect to the finding of no discrimination in the other promotions.

Two landmark Supreme Court cases delineated the method by which a plaintiff will normally persuade a factfinder that he or she has been the victim of intentional discriminatory treatment. Texas Dept. of Community Affairs v. Burdine, 450 U.S. 248, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981); McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). To make out a prima facie case under the McDonnell Douglas model, the plaintiff must show by a preponderance of the evidence that (1) he belongs to the protected group; (2) he applied and was qualified for a job for which the employer was seeking applicants; (3) he was rejected; and (4) after his rejection, the position remained open and the employer continued to seek applicants from persons of the plaintiff's qualifications. Texas Dept. of Community Affairs v. Burdine, 450 U.S. at 253, 101 S.Ct. at 1094 n. 6; Jones v. Gerwens, 874 F.2d 1534, 1539 (11th Cir. 1989).

If the plaintiff succeeds in making out a prima facie case, the burden shifts to the defendant to articulate a legitimate, non-discriminatory reason for its decision. Once the employer has met this burden of production, the plaintiff may still prevail on the ultimate issue of discriminatory intent by proving by a preponderance of the evidence that the defendant's justification is a pretext for discrimination. McDonnell Douglas Corp. v. Green, 411 U.S. 803-04, 93 S.Ct. at 1825; Smith v. Horner, 839 F.2d 1530, 1538 (11th Cir. 1988). We may reverse the district court's decision as to intentional discrimination only if its finding is clearly erroneous. Smith v. Horner, 839 F.2d at 1538.

With respect to appellants Hill, Lock, Jones and Lee, the district court found that they had not made out a prima facie case. Specifically, it found that each had failed to satisfy the second prong of the McDonnell Douglas framework: they were not qualified for the position of foreman because they did not possess the required intelligence, aptitude, and written and verbal communication skills. Brigman testified that this was one component of the minimum qualifications for the foreman position, and that none of the plaintiffs except Coleman had this qualification. Each of the plaintiffs testified that generally he was as or more qualified than those carmen actually promoted. The district court heard the testimony and concluded that the evidence offered by Brigman was more credible.

The plaintiffs contend that a district court may only consider objective criteria in evaluating whether a plaintiff has established a prima facie case under McDonnell Douglas. In other words, the plaintiffs argue that in determining whether a plaintiff is qualified for a position, the court must disregard any subjective qualifications required by the employer. If the subjective intelligence and communication requirement were eliminated, the plaintiffs could show that they were qualified for the position on the basis of their satisfaction of the objective criteria.

The analysis propounded by the plaintiffs has found some support in some of our sister circuits. See, e.g., Burrus v. United Telephone Co. of Kansas, 683 F.2d 339, 342 (10th Cir.), cert. denied, 459 U.S. 1071, 103 S.Ct. 491, 74 L.Ed.2d 633 (1982); Lynn v. Regents of the University of California, 656 F.2d 1337, 1344-45 (9th Cir. 1981), cert. denied, 459 U.S. 823, 103 S.Ct. 53, 74 L.Ed.2d 59 (1982). Under the analysis adopted by these cases, when both objective and subjective criteria make up the job qualifications, the plaintiff need only show that he meets the objective criteria to show that he is qualified; the employer may assert any failure to meet the subjective criteria as part of its rebuttal. Burrus v. United Telephone Co. of Kansas, 683 F.2d at 342. These courts have reasoned that requiring the plaintiff to show subjective qualifications poses dual difficulties: it may be difficult for the plaintiff to prove that he meets a subjective standard, and this standard may be difficult for the court to evaluate. Lynn v. Regents of the University of California, 656 F.2d at 1345.

Although this circuit has viewed the use of subjective factors in employment decisions with suspicion, it has not adopted the Lynn analysis. This circuit has often condemned such practices as using unpublicized, subjective criteria and hiring or promoting without giving notice of the position's availability, because they are ready conduits to race-based decisions. See, e.g., Stallworth v. Shuler, 777 F.2d 1431, 1434 (11th Cir. 1985); Roberts v. Gadsden Mem. Hosp., 835 F.2d 793, 798 (11th Cir.), modified, 850 F.2d 1549 (11th Cir. 1988). For example, in Crawford v. Western Electric Co., 614 F.2d 1300, 1315-16 (5th Cir. 1980), the court held that vague subjective advancement or hiring criteria were suspect, but did not go so far as to hold that they could never be considered in determining the plaintiff's qualifications. Accord Rowe v. General Motors Corp., 457 F.2d 348, 359 (5th Cir. 1972). We have not held that a plaintiff need meet only the employer's objective criteria in order to establish that prong of a prima facie case.

The context of Crawford, as well as Rowe, was more severe than the situation faced by the plaintiffs here: in those cases the employer depended on only subjective criteria in making its decision, not a mix of objective and subjective criteria. In that situation, as well as where the subjective standard is something akin to "likability" or "loyalty," there is a much greater justification for a court to depart from the McDonnell Douglas model. See Maddox v. Claytor, 764 F.2d 1539, 1546 (11th Cir. 1985) (reading Crawford as limited to a situation where the utter subjectivity of the system rendered the plaintiff's burden of showing qualification impossible). In contrast, under the circumstances of this case, it was within the district court's role and capabilities to assess whether the subjective employment qualification was bona fide and whether an employer's testimony that the plaintiff did not possess the qualification was credible. Considering the conclusion of the district court in light of the clearly erroneous standard and the analysis followed in this circuit, we cannot hold that the district court clearly erred in finding that the four appellants were not qualified.

Coleman argues that the district court applied the incorrect legal standard in evaluating his claim of discriminatory treatment. The district court required Coleman, as part of his prima facie case, to show that a less qualified white candidate was promoted in his place. With respect to the promotions of Reel, Hower, Conrad and Clark, the district court held that Coleman had not made out a prima facie case. The district court specifically concluded that none of these four white employees was less qualified than any of the blacks, including Coleman.

We agree with the appellants that the district court applied the incorrect standard. A plaintiff may establish a prima facie case even if he shows that the nonminority employee selected was equally as well qualified; he need not show, as part of his prima facie case, that his qualifications exceeded those of the nonminority employee. Perryman v. Johnson Products Co., 698 F.2d 1138, 1142 n. 7 (11th Cir. 1983) (plaintiff must show "that other employees with equal or less qualifications who are not members of the protected minority were promoted"). The purpose of the McDonnell Douglas test is to allow a plaintiff without direct evidence of discriminatory treatment to establish a rebuttable inference that the employer has discriminated. Such an inference is not justified where the plaintiff's qualifications are inferior to those of the nonminority employee favored by the job decision, but may be justified when the applicants' qualifications are equivalent.

Thus, the district court required the appellants to make a more difficult showing than mandated by law. However, a remand in this case for a redetermination of Coleman's ability to make out a prima facie case is not necessary. The district court implicitly found that the four white employees promoted over Coleman were more qualified. Therefore, remand would be futile: although Coleman might prevail in establishing a prima facie case, he could not overcome a legitimate business justification — i.e., choosing the most qualified candidates — specifically credited by the district court.

Moving to the June 23, 1980, promotion, we hold that the district court did not err in finding that Seaboard discriminated by choosing Benard over Coleman. Seaboard argues that the district court applied the wrong legal standard, requiring Seaboard to meet a burden of persuasion with respect to its legitimate business reason for promoting Benard, as opposed to the lighter burden of production. Although we agree with Seaboard that such a requirement would be contrary to the law, we do not find that the trial court imposed such a requirement. Also, Seaboard argues that even if the court applied the correct standard, its finding that the business reason was pretext is clearly erroneous.

Seaboard apparently does not contest that Benard was less qualified for the position than Coleman. However, Brigman testified that the decision to promote Benard was made hurriedly, on the recommendation of a general foreman, when a sudden opening made the position available. The district court could find that this testimony met Seaboard's burden of production and rebutted the inference of discrimination raised by Coleman's prima facie case.

Once Seaboard articulated a nondiscriminatory reason for its decision, the district court's inquiry did not end. Coleman could prevail on his claim of discrimination if he established that the articulated reason was pretextual. Seaboard argues that there was no direct evidence that Brigman's decisions were racially motivated, and that, in fact, there was evidence that he treated blacks and whites equally. Also, Seaboard notes that as soon as Benard's deficiencies became apparent, he was removed from the foreman position. Seaboard's position is that the preponderance of the evidence demonstrated that the proffered business justification was the company's true motive.

However, the plaintiff's showing that the articulated business reason is pretextual may be made ". . . either directly by persuading the Court that a discriminatory reason more likely motivated the employer, or indirectly by showing that the employer's proffered explanation is unworthy of credence." Texas Dept. of Community Affairs v. Burdine, 450 U.S. at 256, 101 S.Ct. at 1095; Roberts v. Gadsden Mem. Hosp., 835 F.2d at 796. The district court relied on the latter category of proof. Brigman's explanation was inconsistent with his description of the promotion process in general and with the specific circumstances of earlier promotion decisions. Coleman had been considered as a qualified finalist in the three earlier promotions, so Brigman must have been personally familiar with his qualifications; nonetheless, Brigman actually chose Benard. In choosing Benard, Brigman forewent what he had described as the standard promotion procedure and, additionally, accepted a candidate with inferior qualifications. Likewise, Brigman testified that he had decided to promote a black carman, Walters, in late May, although the promotion did not come through until August. This carman was another qualified and available candidate passed over for a white employee upon Benard's promotion. As the factfinder, the district court must ultimately "decide which party's explanation of the employer's motivation it believes." United States Postal Service v. Aikens, 460 U.S. 711, 103 S.Ct. 1478, 1482, 75 L.Ed.2d 403 (1983). The evidence available to the district court undermined Seaboard's claim, indicating that it was "unworthy of credence." Cf. Smith v. Horner, 839 F.2d at 1539 (testimony as to business justification was "internally consistent and not contradicted by any evidence of record"); Clark v. Huntsville City Bd. of Educ., 717 F.2d 525, 528 (11th Cir. 1983) ("The inconsistency between the assertion that a qualified insider was rejected because an outsider was more qualified and a written policy to prefer qualified insiders regardless of the outsider's greater qualifications may cast doubt on the truth of defendant's reliance on the outsider's greater qualifications"). Thus, we cannot find that the district court was clearly erroneous in concluding that Seaboard's business justification for promoting Benard over Coleman was pretextual.

III. DISPARATE IMPACT THEORY

The plaintiffs argue that the district court erred in not finding that they had established discrimination under a disparate impact theory. In contrast to the disparate treatment theory, under the disparate impact theory the plaintiff need not show that the employer intended to discriminate. Instead, a plaintiff may establish a violation of Title VII by showing that a facially neutral employment practice has significant adverse effects on a protected group. Watson v. Fort Worth Bank Trust, ___ U.S. ___, 108 S.Ct. 2777, 2784, 101 L.Ed.2d 827 (1988); Stephen v. PGA Sheraton Resort, Ltd., 873 F.2d 276, 279 (11th Cir. 1989). The facially neutral employment practice may include objective job requirements, see, e.g., Griggs v. Duke Power Co., 401 U.S. 424, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971), or subjective or discretionary employment practices. Watson v. Fort Worth Bank Trust, ___ U.S. ___, 108 S.Ct. at 2787.

In order to facilitate the logical and efficient presentation of evidence in these often complex cases, the courts have developed a framework for the shifting allocation of burdens of proof and production. Stephen v. PGA Sheraton Resort, Ltd., 873 F.2d at 279. First, the plaintiff has the burden of demonstrating that the defendant employed a facially neutral employment practice that had a significant discriminatory effect. Id. See also Connecticut v. Teal, 457 U.S. 440, 102 S.Ct. 2525, 2530, 73 L.Ed.2d 130 (1982). The plaintiff must demonstrate that it is the application of a specific employment practice that has caused the disparity. Wards Cove Packing Co. v. Atonio, ___ U.S. ___, ___, 109 S.Ct. 2115, 2125, 104 L.Ed.2d 733 (1989). See also Watson v. Fort Worth Bank Trust, ___ U.S. at ___, 108 S.Ct. at 2788-89 (plurality) ("Especially in cases where an employer combines subjective criteria with the use of more rigid standardized rules or tests, the plaintiff is in our view responsible for isolating and identifying the specific employment practices that are allegedly responsible for any observed statistical disparties").

After the plaintiff has made out a prima facie case, the employer may rebut the inference of discrimination by articulating a nondiscriminatory business justification for the employment practice. At this stage, the employer has the burden of producing evidence of the business justification; the ultimate burden of persuasion remains on the plaintiff. Wards Cove Packing Co. v. Atonio, ___ U.S. at ___, 109 S.Ct. at 2126.

Finally, if the employer meets the burden of production as to a sufficient business justification, the plaintiff may nonetheless establish discrimination by demonstrating that the proffered employment justification was insubstantial and pretextual, or by demonstrating pretext because other selection devices, without similarly undesirable racial effect, would equally serve the employer's interests. Wards Cove Packing Co. v. Atonio, ___ U.S. at ___, 109 S.Ct. at 2126; Stephen v. PGA Sheraton Resorts, Inc., 873 F.2d at 279.

The district court specifically found that neither the promotion system as a whole nor any of its particular criteria had a discriminatory impact. The decision of the district court was a finding of fact, which we will not overturn unless it is clearly erroneous. Eastland v. Tennessee Valley Authority, 704 F.2d 613, 620 (11th Cir. 1983) (findings of fact reviewable for clear error, including the ultimate fact of the existence of discrimination); see also Williams v. City of Dothan, 818 F.2d 755, 766 (11th Cir. 1987) (Hill, J., dissenting) (discriminatory effect of housing project, in non-Title VII case). The appellants have not met their burden on appeal, i.e., to demonstrate to this court that the district court clearly erred.

The evidence of disparate impact offered by the plaintiffs is sparse. The evidence proffered by the plaintiffs is that as of a fixed date preceding the filing of the EEOC charge, there was only one black foreman out of 25 positions. Thus, while there were approximately twelve percent black carmen, only four percent of the foremen were black. This assertion, while undeniably true, is inadequate proof of disparate impact. First, we note that, while it is not dispositive, there clearly was no disparate impact in the promotions made within the six-month period preceding the filing of the EEOC charge. Moving beyond the six-month period, the plaintiffs have not argued to us, nor did they argue to the district court, over what period of time the whites were promoted. To the extent that they were promoted before the current promotion procedure was adopted, their promotion may not be probative of disparate impact in the operation of the current system. Nor have the plaintiffs argued that the same number of blacks relative to whites obtained in the applicant pool during the period when this promotion procedure was applied. The plaintiffs' bare assertion that their figures demonstrate disparate impact cannot, without more, suffice. See generally Powers v. Alabama Dept. of Educ., 854 F.2d 1285, 1297 (11th Cir. 1988), cert. denied, ___ U.S. ___, 109 S.Ct. 3158, 104 L.Ed.2d 1021 (1989) ("statistical comparisons offered in support of a disparate impact claim should be drawn as narrowly as possible to avoid the effects of factors other than discriminatory practices"). Thus, the argument made to the district court and to this court was insufficient to show the significant adverse impact required to establish a prima facie case.

IV. CONCLUSION

The district court did not err in finding that Hill, Lock, Jones and Lee failed to make out a prima facie case of disparate treatment. While the court applied an incorrect legal standard to the claim of the fifth plaintiff, Coleman, the correct standard could not change the outcome. We affirm the district court's finding of intentional discrimination with respect to Benard's promotion over Coleman. The district court did not err in dismissing the disparate impact claim, because of insufficient proof of adverse impact.

AFFIRMED.