The plaintiffs brought a class action lawsuit against the Douglas & Lomason Co. ("D & L" or "the Company") on July 5, 1985, alleging that D & L intentionally discriminated against blacks in its hiring, promotion, and termination practices. The district court, after a lengthy bench trial, entered judgment against the plaintiffs on their Title VII, 42 U.S.C. Sec. 2000e et seq., and 42 U.S.C. Sec. 1981 claims. We affirm the district court's decision.1

* D & L operates a plant in Cleveland, Bolivar County, Mississippi, that manufactures automobile parts. The plant employs approximately 657 people, 70% of whom are black. Applicants seeking entry-level positions with D & L need only meet certain minimal requirements, such as being at least eighteen years of age, physically able to perform the work, and able to read and write. The plaintiffs challenge certain hiring, promotion, and termination practices followed by D & L at the Cleveland plant after James Grizzard became plant manager in October 1982. The period between October 1982 and April 1986, when Grizzard ceased being manager at the Cleveland plant, was referred to at trial as "the Grizzard years."

* The plaintiffs first challenge the hiring practices adopted by Patty Haynes when she became personnel manager at the Cleveland plant in September 1981. Haynes's hiring practices, which D & L followed throughout the Grizzard years, substantially differed from those of her predecessor, Harold Keeton. Keeton allowed employees to take applications for other persons, allowed people to submit applications twenty-four hours a day, took applications whether or not D & L was hiring, and kept all applications in his "active" file for one year. He generally interviewed two to three applicants for every available position. In making hiring decisions, Keeton looked at the applicant's experience and attempted to hire those applicants with the most relevant work experience. Haynes, on the other hand, would not let employees take applications home, required all people who wanted to apply for employment to come to D & L's plant, accepted applications only when D & L was hiring, and kept applications for six months. She also accepted only one or two applications for every available position. Haynes testified that in making hiring decisions, she looked at the applicant's relevant work experience and would hire "the best person for the job."

During periods when D & L was hiring, Haynes permitted individuals who called the Company seeking information about possible employment to make appointments at which they could fill out an application. Haynes also accepted applications from individuals who came to the plant without appointments.2 If a person called or came to the plant while D & L was hiring but after the applicant pool reached what Haynes considered to be the optimal size, plant guards told the individual either that D & L was not hiring or was not taking applications.3 During periods that D & L was not hiring, Haynes or the guards informed individuals inquiring about employment either that the Company was not taking applications or was not hiring. Thus, it was possible for individuals to be told that D & L was not taking applications both when the Company was and was not hiring.

D & L also hired employees in what can best be called "spurts." Grizzard, after meeting with his department heads and supervisors, would determine whether D & L needed to hire additional workers. If D & L did need additional workers, Grizzard would so inform Haynes, who would begin taking applications. Haynes then would hire the number of employees sought by Grizzard. D & L rarely hired employees on a one-at-a-time or continuous basis and sometimes went months without hiring anyone.

D & L modified its application and hiring procedures in May 1985 by using the Mississippi State Employment Service ("MSES") to provide a pool of applicants from which it could hire. When D & L needed to hire employees, it would notify the MSES. The MSES then would refer both regular and on-the-job-training ("OJT") applicants to D & L.4 Haynes would interview the referred applicants and select new employees using separate job orders for the regular and OJT applicants hired. All hiring from the MSES referrals occurred on three separate spurts from May to September 1985. D & L has not hired any production workers since that time.B

The plaintiffs next contend that D & L engaged in a pattern or practice of discrimination by refusing to promote blacks to leader and foreman positions.5 D & L had no written criteria or guidelines for promotions, and Grizzard did not post notices concerning available promotion opportunities. Instead, department supervisors selected leaders and foremen for their departments with little or no input from Grizzard. Although D & L did not have written criteria for promotions, the supervisors consistently applied the following subjective criteria when selecting leaders and foreman: attitude, work record, relevant work experience, leadership abilities, willingness to help other employees, commitment to D & L, and seniority. Thus, D & L's promotion criteria are predominately of a subjective nature.

The plaintiffs also allege that D & L engaged in discriminatory practices by failing to provide blacks with temporary upgrades and permanent promotions to maintenance department positions. D & L employed a job-bidding process, as required by its collective bargaining agreement with the union representing D & L's employees, whereby employees had to submit bids for available "craft" positions. D & L, however, sometimes gave production employees temporary upgrades to craft jobs on a short-term basis.6 D & L did not allow employees to bid on temporary upgrade positions, and the supervisors of the departments into which an employee would be temporarily upgraded determined whom to upgrade. The chosen employee's supervisor, however, could block the upgrade if the employee was needed for production work.

D & L considered employees in the general maintenance and rack maintenance departments to hold craft positions. As such, D & L required employees seeking those jobs to participate in the bidding process and maintenance department supervisors selected production employees for temporary upgrades to maintenance positions. The primary duties of general maintenance employees included pipe welding and repairing broken machinery. Welding, however, was the primary function of rack maintenance employees. Consequently, D & L considered welding experience to be more important than seniority for the purposes of temporary upgrades and permanent promotions to maintenance department jobs.

The plaintiffs finally contend that D & L discharged employees, using the Company's garnishment policy as a pretext, in retaliation for the employees filing charges with the Equal Employment Opportunity Commission ("EEOC"). D & L's shop rules provided that the Company would discharge any employee who received four or more garnishments from different creditors within a two-year period. Under this rule, only those garnishments not satisfied, withdrawn, or stayed within thirty days after D & L informed the employee of their existence counted against the four-garnishment limit. The two-year period began to run on the date D & L discussed the first garnishment with the employee and expired on the date that the thirty-day grace period for the fourth garnishment expired. Haynes, who was responsible for administering the garnishment policy, forgave all garnishments that D & L received before she became personnel manager.

* The plaintiffs contend that the district court erred by analyzing the hiring and promotion class claims under the disparate treatment model instead of under the disparate impact model.7 We examine discrimination claims using the disparate impact model when employment practices are "facially neutral in their treatment of different groups but ... in fact fall more harshly on one group than another and cannot be justified by business necessity." International Bhd. of Teamsters v. United States, 431 U.S. 324, 335 n. 15, 97 S. Ct. 1843, 1854 n. 15, 52 L. Ed. 2d 396 (1977). Although class action plaintiffs may challenge subjective or discretionary employment practices under the disparate impact model, Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 990, 108 S. Ct. 2777, 2786-87, 101 L. Ed. 2d 827 (1988), the use of subjective or discretionary decision-making does not itself create an inference of discriminatory conduct. Id. at 999, 108 S.Ct. at 2791 (plurality opinion). Instead,

"the plaintiff's burden in establishing a prima facie case [of discrimination] goes beyond the need to show that there are statistical disparities in the employer's work force. The plaintiff must begin by identifying the specific employment practice that is challenged.... 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 disparities."

Wards Cove Packaging Co. v. Atonio, 490 U.S. 642, 656, 109 S. Ct. 2115, 2124, 104 L. Ed. 2d 733 (1989) (quoting Watson, 487 U.S. at 994, 108 S.Ct. at 2788-89). Thus, "[t]he disparate impact model applies only when an employer has instituted a specific procedure, usually a selection criterion for employment, that can be shown to have a causal connection to a class based imbalance in the work force." Pouncy v. Prudential Ins. Co., 668 F.2d 795, 800 (5th Cir.1982); see also Wards Cove, 490 U.S. at 657-58, 109 S.Ct. at 2125 (noting that a plaintiff alleging a discriminatory impact claim must "specifically show[ ] that each challenged practice has a significantly disparate impact on employment opportunities for whites and nonwhites").8

The district court refused to analyze the plaintiffs' hiring and promotion claims using the disparate impact model because they did not identify a specific aspect of subjective decision-making by D & L that was shown to have any causal connection to the alleged class-based imbalance in D & L's general or supervisory work force. The plaintiffs pointed only to D & L's policy requiring individuals to fill out applications at the plant as a specific employment practice causing a class-based imbalance in the work force. This, however, does not justify analyzing the case under the disparate impact model because the plaintiffs did not demonstrate that the disparity they complain of was the result of the challenged policy. See Wards Cove, 490 U.S. at 657-58, 109 S.Ct. at 2125. Moreover, the plaintiffs identified no specific policy that allegedly caused a race-based imbalance in the number of persons who received promotions. Instead, the plaintiffs merely launched a wide-ranging attack on the cumulative effects of D & L's employment practices. The disparate impact model is not the appropriate vehicle from which to launch such an attack. Pouncy, 668 F.2d at 800. Consequently, the district court did not err in refusing to analyze the plaintiffs' claims using the disparate impact model.9 We thus review the plaintiffs' claims under the disparate treatment model.

In a disparate treatment class action, the plaintiffs "must prove discriminatory intent and demonstrate more than 'the mere occurrence of isolated or accidental or sporadic discriminatory acts.' " Carroll v. Sears, Roebuck & Co., 708 F.2d 183, 190 (5th Cir.1983) (quoting Teamsters, 431 U.S. at 336, 97 S.Ct. at 1855). The plaintiffs must establish "by a preponderance of the evidence that racial discrimination was the [employer's] standard operating procedure--the regular rather than the unusual practice." Id.

The plaintiffs may establish a prima facie case of disparate treatment "by the use of statistics if a 'gross' disparity in the treatment of workers based on race is shown." Id. However, "statistics are not irrefutable; they come in infinite variety and, like any other kind of evidence, they may be rebutted. In short, their usefulness depends on all of the surrounding facts and circumstances." Teamsters, 431 U.S. at 340, 97 S.Ct. at 1856-57. If statistical evidence is insufficient to establish discriminatory intent, the plaintiffs may bolster their case by introducing historical, individual, or circumstantial evidence. Bernard v. Gulf Oil Corp., 841 F.2d 547, 568 (5th Cir.1988). The employer then may rebut the plaintiffs' prima facie case "by introducing proof that plaintiffs' statistics are 'inaccurate or insignificant' or by providing a 'non-discriminatory explanation for the apparently discriminatory result.' " Id. (citations omitted).

"[T]he ultimate determination of the existence of intentional discrimination is a question of fact, reviewed on appeal under the clearly erroneous standard." Id. In determining whether a finding is clearly erroneous, we must give "due regard to the trial court to judge the credibility of the witnesses." Anderson v. Bessemer City, 470 U.S. 564, 573, 105 S. Ct. 1504, 1511, 84 L. Ed. 2d 518 (1985). "A finding is 'clearly erroneous' when although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been made." United States v. United States Gypsum Co., 333 U.S. 364, 395, 68 S. Ct. 525, 542, 92 L. Ed. 746 (1948).

The plaintiffs argue that the district court erred in finding that D & L did not discriminate against blacks by restricting access to application forms. They argue that D & L unlawfully restricted access when Haynes refused to give employment applications to a number of blacks while D & L was seeking to hire new employees.

In attempting to prove a prima facie case, the plaintiffs primarily relied on anecdotal evidence provided by class members. A number of individuals testified that they were unable to obtain applications from D & L. The district court found this anecdotal evidence unpersuasive because it was consistent with D & L's general application policies. This finding is not clearly erroneous. See part VIII.A infra. Additionally, the black individuals turned away at the gate by D & L guards never saw D & L give any applications to whites seeking employment, and no D & L employee ever directed derogatory or racially-oriented epitaphs at them.10 Moreover, around the same time that D & L did not give these black individuals applications, D & L provided applications to many other blacks. Finally, the record contains no evidence indicating that D & L's application procedures affected potential black applicants any differently than potential white applicants.11 Thus, the district court's finding that D & L did not discriminate against blacks who were seeking applications is not clearly erroneous. See Falcon v. General Tel. Co., 815 F.2d 317, 322 (5th Cir.1987) (finding that an employer's application procedures were not discriminatory where no evidence was introduced to prove that the procedures affected non-whites any differently than whites).

The plaintiffs contend that statistical and anecdotal evidence established that D & L discriminated against blacks when hiring production workers during the Grizzard years. In comparing the number of blacks hired to the number of blacks who applied for production jobs, however, the district court found no evidence of discriminatory intent.

The job of production worker--an unskilled, entry-level position--required only that applicants be at least 18 years old, be able to read and write, and be able to bend, lift, and stoop. During the Grizzard years, D & L hired 482 production workers--46.5% of those hired were black. Black employees constituted more than 70% of D & L's general work force, and more than 80% of the production work force, during the Grizzard years. In Bolivar County, where D & L's plant is located, 1980 census data indicates that the overall labor force was 51.8% black, with blacks constituting 70.1% of all operators, laborers, and fabricators.

* 1

The plaintiffs first argue that the district court erred in crediting D & L's statistical analysis of its hiring practices during the Grizzard years. The plaintiffs contend that because D & L's statistical analysis was fatally flawed, the district court should have relied on seven sets of statistical data presented by their expert.12 D & L, not surprisingly, finds fault with each of plaintiffs' seven standards and contends that the district court correctly relied upon the standard proposed by D & L's expert.

Where plaintiffs use statistical evidence to challenge an employer's hiring practices, that evidence, to be probative of discriminatory intent, must compare the relevant portion of the employer's work force with the qualified population in the relevant labor market. EEOC v. Olson's Dairy Queens, Inc., 989 F.2d 165, 168 (5th Cir.1993). The dispute between the plaintiffs and D & L revolves around how to properly define the "relevant labor market." D & L contends that the labor market should be defined using an "applicant flow" analysis--i.e., arguing that D & L's available labor force is best determined by examining the applications of those persons who actually sought employment with the Company. The plaintiffs, on the other hand, contend that an applicant flow analysis should not be used here because some of the applications submitted during the Grizzard years were not available for examination.13 Thus, the plaintiffs argue that any one of their seven different methods of measuring the relevant labor market was superior to D & L's applicant flow analysis.14

Actual applicant flow figures are the preferred method by which to measure an employer's hiring practices and performance. Olson's Dairy Queens, 989 F.2d at 168; Phillips v. Joint Legislative Comm., 637 F.2d 1014, 1025 (5th Cir.1981), cert. denied, 456 U.S. 960, 102 S. Ct. 2035, 72 L. Ed. 2d 483 (1982); Hester v. Southern Ry. Co., 497 F.2d 1374, 1379 (5th Cir.1974); see also Mister v. Illinois Cent. Gulf R.R. Co., 832 F.2d 1427, 1435 (7th Cir.1987) ("Statistical analysis of the actual applicants has the advantage of self-selection: the study examines how the employer actually treated people who wanted the job. Applicant studies are preferable as a rule because Title VII governs the treatment of applicants."). "Other statistical measures are necessarily imperfect in differing ways and varying degrees. The best the court can do is to accept what figures are available; allow for imperfections, skewing factors, and margins of error; and then take the figures for what they are worth. Sometimes this is much, sometimes little." Phillips, 637 F.2d at 1025. Thus, when applicant flow figures are flawed or otherwise unavailable, we have allowed plaintiffs to use other statistical measures to establish a prima facie case of discrimination. See, e.g., id. (noting that actual applicant flow figures were unavailable because the employer did not identify applicants by race); Robinson v. Union Carbide Corp., 538 F.2d 652, 657-58 (5th Cir.1976) (upholding the district court's refusal to employ an applicant flow analysis), modified in part on other grounds, 544 F.2d 1258 (5th Cir.1977).

D & L's statistical expert, Dr. Joan Haworth, testified that, according to an applicant flow analysis, D & L did not discriminate in hiring black applicants. Haworth found that from April 1984 to January 1985, blacks constituted 46.7% of all applicants. She then compared this figure to the percentage of applicants hired by D & L during the Grizzard years who were black--45.5%--and determined that the difference was not statistically significant.15 Thus, Dr. Haworth concluded, D & L did not engage in discriminatory hiring practices.

The plaintiffs challenge Dr. Haworth's analysis on several grounds. They first assert that D & L did not keep all applications from the April to October 16 period, thus skewing the analysis so that the number of black applicants from that period was understated. As evidence of D & L's selective retention of applications, the plaintiffs point to the large difference between the number of applications submitted by blacks during the April to October period and the October 17 to January period.16 The district court found, however, that one explanation for the large increase in applications submitted by blacks during the latter period was that Oliver Robinson, a black D & L employee, had begun a campaign to encourage blacks in Bolivar County to apply for employment at D & L and file discrimination charges against D & L with the EEOC.17 This finding is not clearly erroneous.

As additional evidence that D & L failed to retain applications, the plaintiffs contend that the number of applications retained by D & L indicates that Haynes accepted approximately eight and one-half applications for every seven positions filled during the pre-suit period, but accepted about ten applications for every seven positions filled during the post-suit period. The plaintiffs contend that the difference demonstrates that D & L did not retain all applications from the April to October period. Haynes, however, testified that she believed she kept all the applications from the April to October period, although she ultimately was unsure whether she had discarded some applications. Moreover, Haworth testified that while she believed she had all the relevant applications, she nevertheless had a "reasonably large amount" of the applications for the purposes of her analysis. The district court resolved this dispute by finding that D & L did retain all applications pertaining to the relevant period, and we cannot say that this resolution was clear error. See Trevino v. Holly Sugar Corp., 811 F.2d 896, 901 (5th Cir.1987) (finding that the district court did not err in accepting an employer's statistical analysis because, while it was flawed, the plaintiffs' analysis suffered from more serious deficiencies); Nash v. City of Houston Civic Ctr., 800 F.2d 491, 497 (5th Cir.1986) (noting that where there are competing yet permissible views of the evidence, we must affirm the district court's finding).

The plaintiffs next argue that even if the applicant flow analysis was the best standard by which to measure the labor pool available to D & L, the district court erred in accepting Dr. Haworth's analysis because it was based on samples too small to allow for a meaningful statistical analysis. When statistics are based on small sample sizes, the plaintiffs contend, the presence or absence of statistical significance is not a reliable indicator of discrimination. Whether or not a sample is too small to yield statistically meaningful results, however, must be determined on a case-by-case basis. Watson, 487 U.S. at 995 n. 3, 108 S. Ct. at 2789 n. 3; Rendon v. AT & T Technologies, 883 F.2d 388, 397 (5th Cir.1989). The district court found Dr. Haworth's analysis to be more credible and more persuasive than the seven analyses offered by the plaintiffs' expert. We find that the district court did not err in crediting D & L's statistical analysis, which demonstrated that D & L did not engage in discriminatory hiring practices, over the analyses offered by the plaintiffs.182

The plaintiffs further argue that the district court erred in accepting Haworth's analysis of D & L's hiring practices during the period in which the Company exclusively relied upon MSES to provide acceptable applicants. In relying on Dr. Haworth's analysis, the district court found that her job-order-by-job-order analysis was more persuasive because it more closely reflected the actual process through which D & L hired employees. The plaintiffs' analyses, tout au contraire, grouped the six different job orders into one set of statistics.19

D & L placed job orders with MSES three times during the Grizzard years. On each occasion, D & L sought two types of referral--one order requested regular referrals and one order requested OJT referrals. Dr. Haworth compared the percentages of blacks and whites hired by D & L for each separate job order and found that in only one job order did the difference between the actual and expected number of blacks hired exceed two standard deviations.20 The district court found, in light of the record as a whole, that this single statistical disparity did not establish a prima facie case of discriminatory intent. We find that the district court's conclusions regarding the statistical analyses of D & L's hiring practices are not clearly erroneous. See Lewis v. NLRB, 750 F.2d 1266, 1276 n. 17 (5th Cir.1985) (noting that a single statistically significant disparity "by no means commands an inference of discrimination").

The plaintiffs attempted to buttress their statistical evidence with anecdotal evidence regarding class members whom D & L allegedly did not hire because of their race. While Title VII plaintiffs may use evidence regarding individual instances of discrimination to bolster their class claim, Bernard, 841 F.2d at 568, the district court found that D & L did not discriminate in refusing to hire the individuals at issue. This finding is not clearly erroneous. See part VIII.B infra. Accordingly, the plaintiffs' anecdotal evidence, combined with their statistical evidence, did not establish a prima facie case of disparate treatment.

The plaintiffs introduced statistical, historical, and anecdotal evidence in an attempt to prove that D & L unlawfully discriminated against blacks when promoting employees to leader and foreman positions. The plaintiffs' statistical evidence regarding the promotion claim compared the number of blacks promoted to leader and foreman positions during the pre-suit Grizzard years with the percentage of blacks present in D & L's work force; the plaintiffs also sought to compare the percentage of promotions received by blacks during the pre-suit Grizzard years with the respective percentages for the pre-Grizzard and post-suit periods. D & L, on the other hand, introduced statistics comparing the number of blacks holding leader and foreman positions to the number of blacks holding supervisory positions in Bolivar County. The district court ultimately determined that D & L's statistical analysis was the better one and therefore credited it. Moreover, the district found that the historical, anecdotal, and other circumstantial evidence submitted by the plaintiffs did not establish that D & L engaged in a pattern or practice of discrimination. The plaintiffs, of course, challenge these determinations.

* A court may infer that an employer engaged in racial discrimination when promoting workers if statistics, when comparing the number of non-whites and whites promoted, demonstrate a gross statistical disparity. Lewis, 750 F.2d at 1271. Statistical evidence, however, must be meaningful in light of all the surrounding facts and circumstances. Pouncy, 668 F.2d at 1275. "In establishing an inference of discrimination from statistical evidence, the 'required comparison [is] to a qualified pool of employees presumptively eligible for promotion.' " Lewis, 750 F.2d at 1275 (quoting Pouncy, 668 F.2d at 803).

* The plaintiffs introduced statistical evidence comparing the percentage of leader and foreman promotions given to blacks during the Grizzard years--23%--to the percentage of black employees in D & L's production work force--80%. According to the plaintiffs, these statistics indicated a pattern of discrimination in the promotion of D & L employees to leader and foreman positions. The plaintiffs contend that this comparison is appropriate because D & L selected leaders and foremen from within the organization. The district court rejected the plaintiffs' approach as being overbroad because not everyone in D & L's production work force was qualified for promotion. The district court further found that because D & L had hired individuals from outside D & L's work force to be foremen and supervisors, the plaintiffs' approach was too narrow because it excluded individuals outside D & L's production work force who were qualified to be leaders and foremen.21 We agree that the plaintiffs made the forbidden assumption that all D & L employees were equally qualified for promotion. See Lewis, 750 F.2d at 1275 (rejecting the plaintiffs' analysis because they brought into their comparison "employees who are not to be considered for promotion"); Pouncy, 668 F.2d at 803 (refusing to infer that an employer discriminated against black employees in awarding promotions simply because the plaintiffs demonstrated that the percentage of black employees promoted was "far less" than the percentage in black employees in the employer's work force). Consequently, the district court did not err in finding the plaintiffs' statistical evidence for leader and foreman promotions unpersuasive.

The plaintiffs also introduced statistical evidence showing that blacks constituted 64% of persons promoted to leader and foreman positions during the pre-Grizzard and post-suit periods, but only 23% of those promoted during the pre-suit Grizzard period. The plaintiffs also note, however, that during the pre-Grizzard period, blacks apparently received favorable treatment with regard to promotions: whites worked on average for 71.8 months before promotion while blacks worked on average only 43 months. See also note 29 infra. Because of the bias in favor of blacks during the pre-Grizzard period, the district court found it unhelpful to compare statistics from that period with statistics from the post-Grizzard periods. Cf. Lewis, 750 F.2d at 1275 n. 14 (noting that an employer's affirmative action program could skew statistical analyses in disparate treatment cases). Because D & L was not required to continue the preferential treatment given black employees during the pre-Grizzard period, see United Steelworkers v. Weber, 443 U.S. 193, 205-06, 99 S. Ct. 2721, 2728, 61 L. Ed. 2d 480 (1979), the district court thus did not err in disregarding the plaintiffs' statistical evidence regarding the pre-Grizzard period.

In finding no evidence of discriminatory conduct, the district court accepted D & L's statistical evidence, which compared the percentage of black supervisors at D & L to the percentage of black supervisors county-wide.22 We believe that the district court erred in comparing the total number of black supervisors at D & L to the number of black supervisors county-wide. An employer cannot escape liability under Title VII by demonstrating that, at the bottom line, minorities are represented in supervisory capacities at non-discriminatory levels. See Wards Cove, 490 U.S. at 655-57, 109 S.Ct. at 2124; Connecticut v. Teal, 457 U.S. 440, 450, 102 S. Ct. 2525, 2532, 73 L. Ed. 2d 130 (1982). Thus, the relevant comparison is between the number of blacks promoted during the pre-suit Grizzard years and the number of blacks that statistically should have been promoted in the absence of any discrimination.23

The plaintiffs have demonstrated that blacks received 23% of all promotions during the pre-suit Grizzard years. At the same time, blacks constituted 19.1% of the supervisors in production-related occupations in Bolivar County.24 The district court, because blacks were represented in D & L's work force in a ratio higher than statistically predicted,25 inflated the 19.1% figure to 25%--i.e., blacks statistically should have received 25% of all promotions to leader and foreman positions. Because the number of blacks who received promotions is not statistically different from the number of blacks expected to receive promotions,26 the district court did not err in failing to find an inference of disparate treatment with regard to D & L's promotion practices. See Terrell v. University of Tex. Sys. Police, 792 F.2d 1360, 1362 n. 3 (5th Cir.1986) ("When the judgment of the district court is correct, it may be affirmed on appeal for reasons other than those given or relied on below."), cert. denied, 479 U.S. 1064, 107 S. Ct. 948, 93 L. Ed. 2d 997 (1987).

The plaintiffs contend that because D & L based its promotion decisions on subjective criteria, D & L could have failed to promote blacks for discriminatory reasons. The district court rejected the plaintiffs' argument. A promotional system that is based upon subjective criteria is not "discriminatory per se." Page v. U.S. Indus., Inc., 726 F.2d 1038, 1046 (5th Cir.1984); see also Watson, 487 U.S. at 990, 108 S.Ct. at 2786 ("[A]n employer's policy of leaving promotion decisions to the unchecked discretion of lower level supervisors should itself raise no inference of discriminatory conduct."). However, "promotion systems utilizing subjective evaluations by all white supervisors" can be evidence of discrimination in disparate treatment cases. Payne v. Travenol Laboratories, Inc., 673 F.2d 798, 827 (5th Cir.1982); Lewis, 750 F.2d at 1276.

This case is distinguishable from our prior holdings, however, because the D & L supervisors who evaluated and promoted employees were both black and white. As the district court found,

[w]hile the subjectivity of the criteria of attitude and commitment could result in racial disparity in the potential for promotion, it is clear from the record evidence that the individuals recommended or considered for promotion included black individuals recommended by white foremen and supervisors, white individuals recommended by black foremen and supervisors, and individuals recommended by both.

Record Excerpts at 90; cf. St. Mary's Honor Ctr. v. Hicks, --- U.S. ----, ----, 113 S. Ct. 2742, 2751, 125 L. Ed. 2d 407 (1993) (noting that evidence in a hiring case tending to negate a finding of discrimination includes the fact that the hiring officer is of the same minority group as the plaintiff). The district court's finding is supported by the evidence.27 Accordingly, the district court did not err in refusing to find that the subjective component of D & L's promotion policy constituted evidence of discriminatory intent.28 See Bernard, 41 F.2d at 547 (holding that an employer's unwritten and subjective standards for promotion were not discriminatorily applied).

The plaintiffs introduced evidence of the comparative seniority of white and black employees prior to promotion to support their argument that D & L was discriminating against blacks by promoting less senior whites.29 The district court rejected this evidence because it ignored that more senior black employees turned down several promotions later given to less senior white employees.30

The average length of time between promotions may be relevant in proving discrimination by an employer. See Pouncy, 668 F.2d at 804. However, we agree with the district court that the statistics submitted by the plaintiffs are fatally flawed because they overlook the black employees who turned down promotions later offered to white employees. See Trevino, 811 F.2d at 902 (finding that the plaintiffs' statistics demonstrating a "significant disparity between Hispanics and non-Hispanics for the total elapsed time from the date of hire" to promotion was "not compellingly persuasive evidence" of discrimination). For example, Robert LaPresto, a white employee, was promoted to leader after being employed by D & L for approximately four months. However, the position was first offered to four black employees who rejected it. D & L promoted James Haney, another white employee, to leader nine months after hiring him, but only after two black employees first declined the offer of promotion. Six black employees rejected a promotion given to white employee Randy Daniels, whom D & L promoted to leader less than two months after his employment began.31 We also note that during the post-suit period, during which blacks received 64% of the promotions to leader and foreman and about which the plaintiffs do not complain, blacks worked on average 154.9 months before promotion while whites worked only 51 months. This comparison demonstrates that statistics regarding the relative seniority of workers promoted by D & L cannot support a finding of disparate treatment. Cf. Pouncy, 668 F.2d at 803 (finding that a discrepancy between the mean salaries of black and white employees could be explained by "any number of nondiscriminatory factors," including different skill levels, previous training, and experience). Consequently, like the district court, we refuse to infer from this evidence that D & L's promotion practices unlawfully disfavored black employees.

The district court considered evidence put forward by the plaintiffs with regard to alleged individual instances of discrimination and concluded that D & L gave legitimate reasons why it did not promote most of the individuals at issue. The district court also found that D & L did not promote the other individuals simply because they were overlooked or did not come to the attention of management.32 See part VIII.C infra. Thus, the anecdotal evidence did not, as the plaintiffs assert, support a finding that D & L engaged in a pattern or practice of discrimination. See Texas Dept. of Community Affairs v. Burdine, 450 U.S. 248, 259, 101 S. Ct. 1089, 1097, 67 L. Ed. 2d 207 (1981) ("The fact that a court may think that the employer misjudged the qualifications of the applicants does not in itself expose him to Title VII liability...."); Odom v. Frank, 3 F.3d 839, 849-50 (5th Cir.1993) (noting that the perpetuation of a "good old boy" network does not amount to racial discrimination). Consequently, the district court did not clearly err in concluding that the plaintiffs' anecdotal evidence did not support their claim of disparate treatment. See Lewis, 750 F.2d at 1276-77 (finding that anecdotal evidence supported a finding only of isolated instances of discrimination).

The plaintiffs further contend that the district court erred in rejecting their attempt to buttress their statistical evidence of discrimination with evidence regarding D & L's history of discrimination. The plaintiffs argued that D & L's supervisors both used racially derogatory language in speaking to black employees and physically abused them, without any repercussions from upper management. After reviewing the anecdotal evidence presented by the plaintiffs, the district court concluded that D & L at no time encouraged or condoned any mistreatment of black employees by supervisors. Specifically, the district court found that

[t]he sum total of the evidence with regard to these allegations of what constitutes a claim of hostile environment in the form of disparate treatment of black employees amounts, at best, to proof of isolated acts of the use of vulgar or racially derogatory language by no more than two or three company officials and a single instance of physical contact between a company official and a single black employee, all occurring over a period of three or more years. What is also abundantly clear is that when reported to the appropriate company officials, every complained of incident was immediately followed by affirmative company action in the form of apologies to the employees involved, counseling of the managers involved, and/or discipline of those management officials. Moreover, the plaintiffs' allegations relate to the time period after the arrival of the Plant Manager James Grizzard, with there being not a scintilla of evidence that Grizzard participated in or condoned any of the complained of acts, with the exception of possibly two instance of the use of vulgar language, both of which, under the undisputed circumstances, the court finds to have been justified, and one, about which the union filed a grievance, for which Mr. Grizzard apologized.33

Record Excerpts at 129-30. The district court further found that some of the individuals who alleged that D & L officials used racially derogatory language were not credible.

It is within the province of the district court to resolve credibility disputes. Anderson, 470 U.S. at 564, 105 S.Ct. at 1504. Moreover, the "mere utterance of an ethnic or racial epithet which engenders offensive feelings in an employee" is not indicative of a pattern or practice of racial discrimination in violation of Title VII. Rogers v. EEOC, 454 F.2d 234, 238 (5th Cir.1971), cert. denied, 406 U.S. 957, 92 S. Ct. 2058, 32 L. Ed. 2d 343 (1972). After undertaking an independent review of the record, we conclude that the district court did not err in finding that the evidence demonstrated, at best, only isolated instances of racially motivated acts insufficient to establish class-wide disparate treatment.34 See Bernard, 841 F.2d at 569 (holding that the anecdotal testimony of the plaintiffs' witnesses did not establish class-wide disparate treatment).

The class plaintiffs finally argue that the district court erred in finding that D & L did not discriminate against blacks when giving temporary upgrade assignments to general maintenance and rack maintenance positions.35 They contend that the statistical evidence demonstrated a gross disparity in the treatment of white and black employees, thus establishing a prima facie case of discrimination.36 It does not appear that the plaintiffs presented any expert testimony explaining the meaning or statistical significance of this evidence. Consequently, our review of the evidence has been unduly hampered by the plaintiffs' failure to establish whether their statistics were meaningful or significant in light of the particular facts of this case. See Teamsters, 431 U.S. at 340, 97 S.Ct. at 1857 (noting that the "usefulness [of statistics] depends on all of the surrounding facts and circumstances"). Nevertheless, after closely reviewing the plaintiffs' evidence, we conclude that the plaintiffs failed to establish a prima facie case of discrimination.

Eleven employees, three of whom were black, received temporary upgrades to general maintenance jobs during the Grizzard pre-suit period. During this period, two white employees--William Myers and James Partridge--received 68% of the total temporary-upgrade hours. By comparison, two black employees--James Triplett and Alfred Kemp, Sr.--received approximately 88% of all temporary-upgrade hours during the pre-Grizzard period; in the post-suit period, one white employee--Robert Haynes, Jr.--and one black employee--Arthur Perry--received approximately 94% of all temporary-upgrade hours. The plaintiffs introduced no evidence demonstrating that other individuals desired or were qualified for temporary-upgrade positions.37 Consequently, we have nothing with which to compare the plaintiffs' statistics. As a result, the statistics are meaningless.38

Other evidence supports the district court's determination that D & L did not discriminate in giving employees temporary upgrade assignments. For example, the percentage of blacks in permanent craft positions during the Grizzard years ranged from 66.7% to 73.7%. During the immediately preceding years, blacks comprised 57.1% to 63% of the employees in craft positions.39 Moreover, D & L during the Grizzard years awarded seven out of the eight permanent promotions to craft positions to black employees. Moreover, Alfred Kemp, one of the black employees who received no temporary upgrades during the Grizzard pre-suit period but did during both the pre-Grizzard and post-suit periods, testified unequivocally that D & L was a good place to work and always treated him fairly.40 We also note the lack of evidence regarding the number of employees, black and white, who received temporary upgrades to craft positions in non-maintenance departments. See Wards Cove, 490 U.S. at 651-53, 109 S.Ct. at 2122 (noting that plaintiffs should not unnecessarily segment the employer's work force in an attempt to prove discrimination). We therefore agree with the district court's conclusion that D & L did not discriminate in selecting employees for temporary upgrade assignments in the rack maintenance and general maintenance departments.

The statistical evidence presented by the plaintiffs does not present the gross statistical disparity that, when considered in light of the entire record, would require a reversal of the district court's ultimate findings that no pattern of discrimination existed with regard to D & L's hiring and promotion practices. Nor is the historical and anecdotal evidence introduced by the plaintiffs sufficient to support a finding of disparate treatment. Accordingly, we AFFIRM the district court's judgment on the class claims.

The district judge rejected the individual claims of various plaintiffs challenging the treatment they received pursuant to D & L's hiring, promotion, and termination policies. "In a disparate treatment case involving an individual, the plaintiff initially has the burden of showing a prima facie case of discrimination." Carroll, 708 F.2d at 195. A plaintiff meets this initial burden by demonstrating

(i) that he belongs to a racial minority; (ii) that he applied and was qualified for a job for which the employer was seeking applicants; (iii) that, despite his qualifications, he was rejected; and (iv) that, after his rejection, the position remained open and the employer continued to seek applica[tions] from persons [who possessed the plaintiff's] qualifications.

McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802, 93 S. Ct. 1817, 1824, 36 L. Ed. 2d 668 (1973). "Establishment of the prima facie case in effect creates a presumption that the employer unlawfully discriminated against the employee." Burdine, 450 U.S. at 254, 101 S.Ct. at 1094. If the plaintiff establishes a prima facie case of discrimination, "the burden shifts to the employer to rebut the presumption of discrimination by producing evidence that the plaintiff was rejected, or someone else was preferred, for a legitimate, non-discriminatory reason." Carroll, 708 F.2d at 195. If the employer "carries this burden of production, the presumption raised by the prima facie case is rebutted" and drops from the case. Burdine, 450 U.S. at 255 & n. 10, 101 S. Ct. at 1095 & n. 10. "The plaintiff then has 'the full and fair opportunity to demonstrate,' through presentation of his own case and through cross-examination of the defendant's witnesses, 'that the proffered reason was not the true reason for the employment decision,' and that race was." St. Mary's, --- U.S. at ----, 113 S. Ct. at 2747 (quoting Burdine, 450 U.S. at 256, 101 S.Ct. at 1095). The plaintiff at all times bears the burden of persuading the court that he or she was the victim of intentional discrimination.41 Id. --- U.S. at ----, 113 S. Ct. at 2747-48.

* A number of individual plaintiffs claim that D & L unlawfully discriminated against them on the basis of race by refusing to give them applications during the Grizzard years--October 1982 through April 1986. The district court found that some of the plaintiffs were not credible because, during the time they allegedly could not obtain applications, other blacks were given applications.42 Furthermore, the district court found that D & L properly refused to give many of the plaintiffs applications because the Company either was not hiring when the plaintiffs were seeking applications or was hiring but had already received a sufficient number applications for the positions to be filled. The district court's findings are not clearly erroneous. An extensive review of the record has convinced us that it contains no evidence indicating that the plaintiffs were not given applications because of their race.43 The only evidence supporting the plaintiffs' allegations that D & L refused to give blacks employment applications is the plaintiffs' unsupported assertions concerning their collective belief that D & L discriminated against blacks.44 Under D & L's application policy, see part I.A supra, there were legitimate, nondiscriminatory reasons why D & L did not give applications to the plaintiffs. Because the plaintiffs failed to demonstrate that race was the true reason for D & L's refusal to give them applications, we uphold the district court's finding that D & L did not discriminate against the plaintiffs. See Odom v. Frank, 3 F.3d 839, 849 (5th Cir.1993) (holding that anecdotal evidence and mere speculation will not support a finding that an employer followed an unwritten "policy" of discrimination); Carroll, 708 F.2d at 195-96 (noting that the mere suspicion of discrimination is not sufficient to establish disparate treatment).

Plaintiffs Chris Harris and Bobbie Sharp contend that D & L unlawfully discriminated against them on the basis of race by refusing to hire them for production positions. The district court found that D & L did not hire Harris and Sharp for valid, race-neutral reasons. For example, Harris, who applied when D & L was hiring only for the night shift, indicated on her application that she attended afternoon and evening classes at a local college. Haynes testified that she believed she did not hire Harris because of the conflict between Harris's classes and the shift for which she was hiring. The district court noted that although Haynes's explanation was not absolute, no evidence suggested that race affected Haynes's decision not to hire Harris. This finding is not clearly erroneous. See Odom, 3 F.3d at 845 ("Although Strader's testimony ... was less than absolute, it constituted at least some evidence. More significantly, it was uncontradicted. As Odom adduced no evidence favorable to his position ..., we are compelled to accept Strader's legitimate, non-discriminatory explanation...."); cf. St. Mary's, --- U.S. at ---- n. 5, 113 S. Ct. at 2751 n. 5 ("The notion that every reasonable employer keeps 'personnel records' on people who never became personnel, showing why they did not become personnel (i.e., in what respects all other people who were hired were better) seems to us highly fanciful--or for the sake of American business we hope it is."). Likewise, Sharp failed to adduce any evidence indicating that D & L unlawfully considered her race when deciding not to hire her. Moreover, at the same time Harris and Sharp were not hired, D & L hired other blacks as production workers. Thus, the district court's finding that D & L did not discriminate against her is not clearly erroneous. See Carroll, 708 F.2d at 195 (noting that the mere suspicion of discrimination does "not establish disparate treatment").

Plaintiffs Nathaniel Cannon, Michael O'Neal, and Gregory Townsend contend that D & L unlawfully discriminated against them by refusing to rehire them after they had been laid off.45 The district court found that these plaintiffs adduced no evidence indicating that race was a factor in D & L's decision not to rehire them. In fact, the record indicates that D & L rehired twenty-two blacks, and only fourteen whites, during the Grizzard years. We hold that the district court did not err in finding that D & L did not rehire Cannon, O'Neal, and Townsend because of legitimate, race-neutral reasons. See Carroll, 708 F.2d at 195 (noting that the mere suspicion of discrimination does "not establish disparate treatment").

Plaintiffs Moses Coleman, Richard Grant, Earnest Hall, Robert Melvin, Alexander Smith, and Kimroy Williams claim that D & L, in failing to promote them to leaders and foremen, discriminated against them on account of their race. The district court found that the plaintiffs failed to demonstrate that race played any part in D & L's failure to promote them. This finding is supported by the record evidence.

In selecting leaders and foremen, D & L considered the employee's attitude, work record, work experience, leadership abilities, willingness to help other employees, commitment to the Company, and seniority to be very important. The record reflects that each plaintiff was lacking in one or more of these attributes during the Grizzard years. See Davis v. Chevron U.S.A., Inc., 14 F.3d 1082, 1087 (5th Cir.1994) (noting that to establish a prima facie case of discrimination, an applicant must be qualified for the sought position at the time he or she applies for it). For example, Gerald Bardwell, Coleman's supervisor, stated that Coleman, when temporarily filling in for foremen, was unable to handle the job and had employees "goofing off on him." Howard Watson testified that department supervisors never recommended Coleman for promotion. Additionally, both Grant and Hall had both turned down previous offers of promotion to leader positions, indicating a lack of commitment to D & L.46 Charlie West, a black supervisor, testified that he did not consider Grant for one promotion because he did not know Grant. Byron Kyle, a department superintendent, testified that Grant did not have any leadership abilities.47 Richard Williams, a black supervisor, and Howard Watson testified that Melvin and Kimroy Williams did not show any interest in being promoted and did only enough work to get by, thereby demonstrating little desire for advancement and no commitment to the Company.48 Tommy Lucas, a foreman in the paint department and former union steward, testified that Smith had an attitude problem.

The plaintiffs failed to demonstrate that their race played any role in D & L's failure to promote them. See Carroll, 708 F.2d at 196 (rejecting a claim that an employee was discriminatorily denied promotion because his supervisors testified that his attitude and lack of dependability were the actual reasons); Pouncy, 668 F.2d at 795 (rejecting an employee's promotion claim because the employee did not possess the initiative or the ability to communicate with and train his co-workers). At most, the plaintiffs proved that D & L may have overlooked one or more of them when considering potential candidates for promotion. Such a showing is insufficient to prove discrimination. Cf. Odom, 3 F.3d at 849-50 (noting that the perpetuation of a "good old boy" network does not amount to racial discrimination). Accordingly, the plaintiffs failed to prove that D & L did not promote them on account of their race.

Daniel Anderson, Jr. and Tommy Hardy argue that D & L refused to give them temporary upgrade assignments in the general maintenance or rack maintenance departments because of their race. Welding experience was necessary for many of the tasks performed by general maintenance employees, and welding was the primary duty of rack maintenance employees. The district court found that D & L did not give Anderson and Hardy temporary upgrade assignments because they lacked the necessary experience to work in the maintenance departments.

Both Anderson and Hardy testified that they had only a limited amount of welding experience.49 Moreover, they presented no evidence demonstrating that what little welding experience they had qualified them to perform the duties of general and rack maintenance employees. Finally, they introduced no evidence suggesting that any white employee who received an upgrade was unqualified (or less qualified than the plaintiffs). Because the district court's finding that Anderson and Hardy were not qualified to receive temporary upgrades is not clearly erroneous, the district court's conclusion that Anderson and Hardy failed to establish a prima facie case of discrimination is supported by the evidence.

Anderson further contends that D & L discriminated against him by not permanently promoting him to the rack maintenance department. Anderson and James Bruce, a white employee, submitted bids for a permanent rack maintenance position. Both Anderson and Bruce were interviewed. The notes of the interviewer indicated that he considered Anderson's limited welding experience, Bruce's more extensive welding experience, Anderson's "poor attendance," and Bruce's "very good" attendance in determining who to recommend for the promotion.50 Moreover, the notes also indicate that a black employee with more seniority than Bruce turned down the job. Thus Anderson did not establish a prima facie case of discrimination because he was not qualified for the promotion which he sought.51 Moreover, even if Anderson did establish a prima facie case, D & L demonstrated by a preponderance of the evidence that it rejected Anderson for legitimate reasons. Accordingly, the district court properly found that Anderson failed to demonstrate that he was the victim of discrimination.

Four plaintiffs--Belinda Cox, Robert Haywood III, Synovia Jackson and Jesse Overstreet--contend that the district court erred in finding that D & L did not discharge black employees under its garnishment policy, see part I.D supra, in retaliation for filing EEOC charges.52 To establish a prima facie case of retaliatory discharge under Sec. 2000e-3(a) of Title 42,53 a plaintiff must demonstrate (1) that he engaged in activity protected by Title VII, (2) that an adverse employment action occurred, and (3) that a causal link between participation in the protected activity and the adverse employment decision exists. Shirley v. Chrysler First, Inc., 970 F.2d 39, 42 (5th Cir.1992). Once a plaintiff establishes a prima facie case, the employer bears the burden of articulating some legitimate, nondiscriminatory reason for its action. Id. If the defendant does so, the employee bears the ultimate burden of showing that the reasons given by the employer were a pretext for retaliation. Id.

The parties appear to agree that the plaintiffs passed the first two prongs of the retaliation test, as they filed EEOC charges and subsequently were terminated from their positions.54 They disagree, however, about the causal link between their terminations and the filing of the EEOC charges. The plaintiffs argue that but for the filing of the EEOC complaints, they would not have been fired. D & L counters by arguing that there is no causal connection between the plaintiffs' terminations and their protected activities. D & L further argues that even if the plaintiffs did establish a prima facie case of retaliation, it articulated a legitimate, nondiscriminatory reason for their discharges--violations of the garnishment policy.

The district court found that the plaintiffs failed to show a causal connection between the filing of the charges and their terminations. The district court further found that the evidence demonstrated legitimate, nondiscriminatory reasons for the plaintiffs' terminations--violations of the garnishment policy. These findings are not clearly erroneous. It is undisputed that the plaintiffs did in fact violate the garnishment policy.55 Because the penalty for such a violation was termination, D & L properly terminated the plaintiffs. See Lewis, 750 F.2d at 1279 (holding that when some evidence substantiates a claim of retaliation while other evidence demonstrates that the employer acted for innocent reasons, the district court did not clearly err in rejecting the retaliation claim).

The plaintiffs argue their terminations are causally related to their protected activity because D & L did not discharge other employees who violated the garnishment policy if those employees or their relatives had not filed EEOC charges. The plaintiffs point to seven individuals whom they argue should have been discharged under the policy but were not. Six of the individuals named by the plaintiffs, however, did not violate the garnishment policy.56 Moreover, the mere fact that one individual violated the policy without being fired, does not by itself amount to race-based discrimination. See Lewis, 750 F.2d at 1278 n. 19 (finding that irregular exceptions to a valid policy did not support a retaliation claim); cf. Odom v. Frank, 3 F.3d 839, 850 (5th Cir.1993) (noting that "misfeasance, malfeasance, or nonfeasance--without nexus to ... race--is not actionable"). Furthermore, contrary to the plaintiffs' contention, the evidence indicates that D & L discharged under the garnishment policy other black employees who had not filed EEOC claims.57 Additionally, the plaintiffs presented no evidence suggesting that D & L did not discharge white employees who violated the policy. Consequently, the district court did not err in finding that no causal link existed between the filing of the EEOC charges and the plaintiffs' discharges. We therefore uphold the district court's finding that D & L did not discharge the plaintiffs in retaliation for their participation in protected activity.

The plaintiffs failed to produce sufficient evidence demonstrating that D & L engaged in any race-based discrimination. Accordingly, we AFFIRM the judgment of the district court in all respects.

The facts here clearly show that there have been scores of individuals who have endured racial discrimination at the behest of Douglas and Lomason ("D & L"). Many of these individuals have been pressured by the president of the company to drop this cause of action "for the good of the company." Instead of bowing to the company president's threats of closing the plant and laying off hundreds of people because of this lawsuit, the plaintiffs have stood up for what they believed to be their rights to equal employment. Despite a plethora of evidence which seems to overwhelmingly support the Plaintiff's contentions, the district court found that no racial discrimination occurred. The majority affirms. After carefully reviewing the record, this writer is constrained to dissent.

The most appropriate place to begin an analysis of disparate impact claims is the law. The seminal disparate impact case, Griggs v. Duke Power Co., explained that Congress' objective in enacting Title VII was to "achieve equality of employment opportunities and remove barriers that have operated in the past to favor an identifiable group of white employees over other employees." 401 U.S. 424, 429-30 (1971). The Court went on to determine that "practices, procedures, or tests neutral on their face, and even neutral in terms of intent, cannot be maintained if they operate to 'freeze' the status quo of prior discriminatory employment practices." 401 U.S. at 429-30, 91 S. Ct. 849, 853, 28 L. Ed. 2d 158. According to the Court, "good intent or absence of discriminatory intent does not redeem employment procedures or testing mechanisms that operate as 'built-in headwinds' for minority groups and are unrelated to measuring job capability." Id. at 432, 91 S.Ct. at 854.

The Court elucidated the proof requirements for disparate impact cases in Wards Cove Packing Co., Inc. v. Atonio, 490 U.S. 642, 109 S. Ct. 2115, 104 L. Ed. 2d 733 (1989). It ruled that for a Title VII plaintiff to successfully present a disparate impact claim, he or she must point to a particular employment practice that has created such disparate impact. Id. at 657, 109 S.Ct. at 2124-25. The plaintiffs in Wards Cove alleged that the practices in question were, inter alia, nepotism, the creation of separate hiring channels, the practice of preferential treatment in rehiring, and the use of subjective decision making. Id. The Supreme Court concluded that the plaintiffs had properly presented a disparate impact claim by pointing to such practices which, if proved, would support a finding of racial discrimination. Id.

2. Alleged Discriminatory Employment Practices

In the instant case, the district court concluded that Plaintiffs had failed to identify which of D&L's employment practices adversely impacted African Americans. The majority asserts that the Plaintiffs did, in fact, identify an employment practice--namely, D&L's policy which requires applicants to complete applications at the plant. However, the majority asserts that such a practice could not justify resort to the disparate impact theory. Maj. op. at 1284.

The record belies the conclusions of both the majority and the district court. The pretrial order, signed by the district court, agreed to by counsel for each party, and filed just over one month prior to the trial, states plainly, "Plaintiffs allege that the defendant company has discriminated against blacks in hiring, both by restricting access to application forms and by failing and refusing to hire those blacks who are permitted to fill out application forms on the same basis as whites who fill out application forms." R. at 2590. Restricting access to application forms and using different hiring standards for blacks, as opposed to whites, are both employment "practices" which justify the application of the disparate impact theory to the facts of this case.

The majority "found" that the "record contains no evidence indicating that D&L's application procedures affected potential black applicants any differently than potential white applicants." Maj. op. at 1285-86. Reading the record reveals otherwise. The record quite readily shows that D&L's application procedures almost exclusively affected blacks. D&L's expert, Dr. Joan Haworth, testified that the hiring practices initiated by Patty Haynes--specifically the restriction of application forms--caused "wild fluctuations" in the percent of black applicants who were hired.1 The defense "hung its hat" on this explanation. In its argument to the district court during its Rule 41 motion, D&L's attorney contended that Patty Haynes' change in the hiring procedures alone was enough to cause the change in the proportion of black hires from the pre-Grizzard era.

D&L's--and Plaintiffs'--contention that Ms. Haynes practice of restricting access to application forms caused the changes in black hires at the Cleveland plant is completely substantiated by the record.2 D&L and Plaintiffs introduced into evidence dozens of depositions of black individuals who had attempted to obtain, but who were refused, applications from D&L when the company was hiring. Many of the deponents identified black friends and/or family members who were also denied the opportunity to file applications.3 In all, Plaintiffs identified more than fifty blacks to whom D&L refused applications. The vast majority of these people attempted to obtain applications numerous times. Many of them asserted that they became discouraged and stopped trying to obtain applications. Further, there is no dispute that a great number of them sought to apply for jobs at D&L when the company was hiring. Reviewing this evidence, the majority properly claims that "the black individuals turned away at the gate by D&L guards never saw D&L give any applications to whites seeking employment." Maj. op. at 1285-86 (emphasis added). However, that finding probably provides little comfort to the blacks who testified that they not only saw whites obtaining applications, but also saw whites being hired at the same time that blacks were being turned away in the personnel office--as opposed to the gate.4

One would think--and the law certainly requires--that the restriction of applications would impact on whites equally. However, such is far from the case here. As compared to the two and one-half score blacks who were refused applications, D&L presented the resounding number of three whites who had been refused applications. Although dozens of blacks sought to obtain applications while the company was hiring, not one of the white potential applicants sought an application during such periods.

a. The Law

The disparate impact model usually focuses on statistical analysis. Watson v. Fort Worth Bank and Trust, 487 U.S. 977, 987, 108 S. Ct. 2777, 2784-85, 101 L. Ed. 2d 827 (1988). Statistical data is probative, however, only if it properly compares relevant elements. See Hazelwood School District v. United States, 433 U.S. 299, 97 S. Ct. 2736, 53 L. Ed. 2d 768 (1977). The "proper comparison [is] between the racial composition of [the at-issue jobs] and the racial composition of the qualified ... population in the relevant labor market." Wards Cove, 490 U.S. at 651, 109 S.Ct. at 2121 (quoting Hazelwood, 433 U.S. at 308, 97 S.Ct. at 2741-42). However, where labor market statistics are impossible to ascertain, the Supreme Court has recognized that "certain other statistics--such as measures indicating the racial composition of 'otherwise-qualified applicants' for at-issue jobs--are equally probative for this purpose." Id.

b. The Facts

All agree that the best comparison in this case would be the racial composition of the applicant flow into D&L for general factory jobs as compared with the racial composition of the general factory hires. The district court found that such a comparison is possible. The majority found no clear error in that finding. Both are wrong.

Fewer than two years after the filing of this lawsuit, Patty Haynes testified during a deposition that she started to retain all of the applications after she learned that some employees filed EEOC charges in October of 1984. She explained that she threw away a number of applications just prior to that time, and she was not sure whether she also threw away some of the applications which were filed during the five months preceding the filing of the EEOC charges.

Statistics, which D&L did not dispute, confirm that Ms. Haynes failed to retain all of the applications. Ms. Haynes testified, and the district court found, that she, as a general rule, would accept twelve to fifteen applications per seven hires. This was a fairly constant practice, and once she adopted this practice, she did not deviate therefrom throughout her tenure as personnel manager. The application data which exists after the EEOC charges were filed (October 17, 1984) are consistent with Ms. Haynes' explanation of her practice: Of the applications on hand, almost forty percent came from unhired applicants. Hence, for every six hires, Ms. Haynes was able to consider ten applicants. The application information for the period prior to the filing of the EEOC charges presents a vastly different picture. Of the applications on hand which were filed prior to October 17, 1984, only twenty-two percent were filed by unhired applicants. Contrary to Ms. Haynes' explication of her practice of accepting twelve to fifteen applications per seven hires, and contrary to the district court's acceptance of that explanation, Ms. Haynes would have been required to hire four people out of every five applicants throughout that period.

The difference in the applications is not only substantial, but it is also statistically significant. Dr. Charles Mann, a statistician, testified--and D&L did not dispute--that all things being equal--and according to Ms. Haynes, all things were equal--there were only four chances in one thousand that all of the applications existed.5 Dr. Mann's calculation reveals that Ms. Haynes' uncertainty about her retention of all of the applications was well-founded.6

Even if all of the applications were available--and they clearly are not--reliance on the applications would be inappropriate in light of the fact that the company disproportionately restricted black potential applicants' access to the forms. Barriers and/or practices which deter qualified minorities from applying for jobs impermissibly taints any analysis which employs the use of actual applicant-flow data. Dothard v. Rawlinson, 433 U.S. 321, 330, 97 S. Ct. 2720, 2727, 53 L. Ed. 2d 786 (1977) ("The application process might itself not adequately reflect the actual potential applicant pool, since otherwise qualified people might be discouraged from applying" because of the alleged discriminatory practices); see Wards Cove, 490 U.S. at 653, 109 S.Ct. at 2122-23 ("As long as there are no barriers or practices deterring qualified nonwhites from applying ... if the percentage of selected applicants who are nonwhite is not significantly less than the percentage of qualified applicants who are nonwhite, the employer's selection mechanism probably does not operate with a disparate impact on minorities." (emphasis added)). In light of D&L's restrictions on blacks' attempts to obtain applications, the true applicant-flow information cannot form the basis of the statistical analysis in this case.7 Hence, the district court should have resorted to appropriate proxy information, which the Supreme Court has declared to be "equally probative." Wards Cove, 490 U.S. at 651, 109 S.Ct. at 2121-22.

Plaintiffs proffered seven different standards of availability which served as a proxy for the actual applicant-flow information. Four of those standards were particularly persuasive: 1) the Mississippi State Employment System ("MSES") referrals to D&L for 1985, 2) the fully registered persons with MSES,8 3) the number of general factory hires during the four and three-quarter years preceding Mr. Grizzard's arrival, and 4) the number of people in Bolivar County who are were employed as operators, fabricators, and laborers in factories other than D&L, as shown by the 1980 census.9

The first standard contains all of the application forms of all interested D&L applicants during the time period in issue which were on file with MSES. This standard is untainted by unfair restrictions on blacks' ability to obtain such applications.10 The district court rejected this standard because it was "an extremely small sample." Interestingly enough, however, the court credited Dr. Haworth's statistical analysis of her "pools" of applications, and each such pool was smaller than the pool in this first standard.11 The district court's rejection of this standard is directly antithetical to its acceptance of Dr. Haworth's pools. In this writer's view, it is clearly erroneous. Furthermore, this standard provides a reliable barometer of the general D&L applicant flow during the periods in question. Indeed, Dr. Haworth, herself, testified that this standard, though not perfect, was valid.12 The proportion of black applicants in this standard was 65.6%. When compared with the actual number of blacks hired, the standard showed statistical significance at 8.8 standard deviations.13

The district court rejected the second standard, claiming that this standard is "heavily weighted" in favor of blacks because of their over-utilization of MSES. Dr. Haworth testified that a national study indicated that blacks in large urban areas generally are more likely to use public employment services than whites. She tempered her testimony with the caveat that she did not know whether this study was applicable to Bolivar County, which is small and rural. Dr. Haworth never claimed that the difference in the usage of public employment services between blacks and whites was statistically significant or even substantial. She did, on the other hand, testify that there might be only a trivial difference, or even no difference, in the usage of MSES by blacks and whites in Bolivar County.

Dr. Mark Bendick, Plaintiff's labor economist, testified that a study on the usage of public employment agencies in southern, rural areas--including Sunflower County, Mississippi, a rural county which is adjacent to Bolivar County--revealed that the public employment service usage in southern, rural areas is quite different from such usage in large urban areas. While the study did not compare blacks' and whites' usage of employment services in rural areas, Dr. Bendick testified that no study shows that rural blacks use unemployment services at a disproportionately higher rate than rural whites. The district court had no evidence contrary to Dr. Bendick's testimony on this subject, for the company did not even attempt to dispute the testimony. The district court's conclusion that the second standard is "heavily weighted" in favor of blacks due to their "over-utilization" of MSES is therefore highly speculative and is not supported by the record. In light of the Supreme Court's disapproval of a court's supply of explanations of statistical information which is not supported by the record,14 this Court should not countenance the judicial activism by the district court. No sufficient reason exists in the record--or otherwise, for that matter--for rejecting the overall MSES registrant information. This, the second standard of availability, reveals that 65% of the MSES registrants who were over eighteen and had none of the earlier-described characteristics were black--just six-tenths of one percent smaller than the proportion of blacks in the first standard. This standard demonstrates statistical significance at 8.5 standard deviations.

The district court ruled that in the third relevant standard of availability, Plaintiffs failed "to take into account the distinct possibility that the hiring rate prior to the period beginning October, 1982, may indeed have been discriminatory in favor of black applicants." Op. at 18 (emphasis added). Again, the district court improperly supplied reasons for ignoring Plaintiffs' statistical analysis which no party espoused at any time throughout the trial or during depositions. See supra note 13.

The court further explained that the differences in Patty Haynes' employment procedures perhaps accounted for the change in employment rates of blacks. The district court's statement of the obvious is actually a recitation of the disparate impact theory. Its attempt to explain away Plaintiffs' reliance upon standard three in fact embraces Plaintiffs' Title VII claim. There is no valid reason for disregarding this standard. Indeed, in light of the district court's back-door recognition that Ms. Haynes' practices may have caused the hiring disparities, there is every reason to consider this standard as a valid reference for a statistical analysis of the disparate impact claim. This third standard, which shows that 64.6% of the hires were black, is just four-tenths of one percent lower than the second standard and equates to 8.3 standard deviations. The consistency of the first three standards is amazing and, in this writer's view, lends greatly to their credibility and probative value.

The final standard, though rejected by the district court, is much more precise than standards accepted by the Supreme Court in Dothard v. Rawlinson and International Brotherhood of Teamsters v. United States. Those cases compared the make-up of defendants' employees to the minority presence in the general population. In Dothard, the Supreme Court reviewed Alabama's penitentiary system for sex discrimination in hiring. The Court upheld the district court's reliance upon generalized, national statistics which demonstrated that Alabama's hiring system discriminated against women. Dothard, 433 U.S. at 329-331, 97 S.Ct. at 2726-28. Similarly, the Teamsters Court compared the proportion of minority employees with the proportion of minorities in the general, area-wide population. Teamsters, 431 U.S. 324, 337 n. 17, 97 S. Ct. at 1855 n. 17. The Court in Hazelwood School District v. United States explained that the Teamsters Court's use of general population figures was "highly probative" because the jobs in question were low-skill jobs, which almost everyone in the population could acquire. 433 U.S. at 308 n. 13, 97 S. Ct. at 2741-42 n. 13. The Wards Cove Court further explicated that "where figures for the general population might ... accurately reflect the pool of qualified job applicants, ... we have ... permitted plaintiffs to rest their prima facie cases on such statistics as well." 490 U.S. at 651 n. 6, 109 S. Ct. at 2121 n. 6.

The final statistical standard in this case is much more probative than general population statistics,15 for it hones that number down to the actual number of blacks who are not only in the general population, who are not only in the civilian labor market, but who also are qualified and have shown interest in the general factory-type work at issue in this case.16 The district court's basis for rejecting this standard finds no support in the record. The court first spurned this standard because it claimed that a comparison between hired people and unhired people was inherently unreliable. While such a comparison certainly is not perfect, the facts of this case clearly prove that such a comparison greatly favors D&L, not the Plaintiffs, for the unemployment rate among blacks in Bolivar County at all relevant times has far exceeded the unemployment rates of whites in the county.17 Hence, the proportion of black-unemployed job-seekers would obviously exceed the proportion of black general factory workers.

The District Court also rejected this last standard because not all D&L applicants have prior production-worker experience.18 The court noted that many applicants have employment backgrounds in the service or manufacturing industries. While the court's observation is entirely true, it is, at the same time, completely irrelevant to the issue at hand--what proportion of applicants at D&L were black, and hence what is a proper standard of availability for this case. Each applicant, regardless of his or her work experience, displayed interest in general factory work, just as those working in production-type jobs displayed such interest. While a comparison of the number of production workers with the number of applicants to service industry or manufacturing jobs would be unpersuasive--since those interested in the latter jobs might have no interest in the former job--the standard here focuses upon the interest which each person, whether worker or applicant, has shown in general factory work, the work at issue in this case. The district court's rationale for rejecting the final standard entirely misses the boat.

The last standard, which to me is completely valid, shows that 70.1% of the employees in general factory-type jobs in Bolivar County, excluding D&L employees, are black. Statistical significance, at 11.3 standard deviations, plainly shows that, like the other standards, the hiring results during the Grizzard years did not happen by chance.19 Under any of these standards, it seems abundantly clear that Plaintiffs have successfully made out a prima facie case of race discrimination in D&L's hiring practices,20 and the district court so found during trial.

D&L, in essence, conceded Plaintiffs' disparate impact21 case in hiring by agreeing that the changes Patty Haynes made in her hiring practices--namely, her restriction of application forms--resulted in the significant change in the proportion of blacks hired. Even without such a concession, all credible evidence demonstrates that D&L's hiring during the Grizzard era was discriminatory with respect to race. This Court should therefore reverse on this issue and remand for the district court to decide damages.B. Promotions

Like the district court's analysis and the majority's review of the hiring data, this writer finds that the analysis and review of the promotion statistics are greatly misleading. The majority, quoting Lewis v. National Labor Relations Board, correctly sets forth the proper statistical framework upon which claims of promotion discrimination claims must be based: "In establishing an inference of discrimination from statistical evidence, the 'required comparison [is] to a qualified pool of employees presumptively eligible for promotion.' " Maj. op. at 1290 (quoting Lewis, 750 F.2d 1266, 1275 (5th Cir.1985) (emphasis in original)).

Unlike this case, in both Lewis, which this writer authored, and Pouncy v. Prudential Insurance Company of America, to which the majority refers, there were minimum objective qualifications for promotions which precluded comparison of the proportion of minorities promoted to the proportion of minorities in the overall work force. The promotion pool was smaller than the entire work force. In Pouncy, only those employed at level eleven could be promoted to level twenty. Pouncy, 499 F. Supp. 427, 454 (S.D.Tex.1980), aff'd, 668 F.2d 795 (5th Cir.1982). Similarly, in Lewis, only employees who had reached the GS-12 field examiner level or the GS-13 field attorney level were "presumptively eligible" for promotion. Lewis, 750 F.2d at 1275. Comparison with the entire work force in those cases was improper, since not all of the employees had reached the required employment level for promotion.

This case is quite different. Here, every production worker is "presumptively eligible" for promotion to the leader positions in question, and every leader is "presumptively eligible" for promotion to the foreman positions in question. There is no educational, skill, or other objective requirement for promotion to those positions at D&L.22 With the exception of the attendance record,23 all of the "qualifications" which the D&L supervisors thought were important were entirely subjective.24 An unsupported claim that blacks possess these subjective characteristics at a substantially lower proportion than whites is, in itself, a discriminatory claim in which the district court apparently finds solace.

D&L's entire work force, which was approximately 70% black, is the proper pool to which the promotions should be compared,25 for that is the pool from which promotions came.26 At no time have promotions approximated 70%. Interestingly enough, however, prior to Mr. Grizzard's arrival at D&L, 63.8% of the promotions went to blacks. That number plummeted to 22.7%27 during his tenure as plant manager and returned to 63.7% after Plaintiffs filed this lawsuit. When asked why the number of black promotions had declined so drastically, neither Mr. Grizzard nor any supervisor so questioned could offer an explanation.

Contrary to the majority's reliance on the fact that a number of blacks had turned down promotions, Byron Kyle, supervisor over one half of the plant, Howard Watson, production manager for the Cleveland plant, and Charlie Lofton, supervisor over the press area in the plant, all testified that prior to Mr. Grizzard's arrival, blacks and whites turned down promotions. They asserted that the number of people who turned down such promotions did not change during Mr. Grizzard's tenure or after his departure. In fact, a substantial number of blacks who turned down promotions during the Grizzard years had also turned down promotions during the pre-Grizzard years. These employees were repeatedly offered promotions.28 Based upon such testimony, blacks' rejection of promotions could not be the reason for the stark decline in black promotions. Indeed, no supervisor ever even intimated that such was the case.

2. Comparison With Supervisors in Bolivar County

Dr. Haworth and the district court agree that a comparison of the promotions at D&L with the percentage of blacks in supervisory positions in Bolivar County is more relevant than a comparison of the promotions to the applicable labor pool. Such a comparison is highly questionable.

According to Dr. Haworth, most of the supervisory positions to which she compared D&L's promotions were precision craft positions. The "bulk" of those positions were held by craftsmen and craftswomen. Dr. Haworth testified that blacks held only 33.4% of the precision craft jobs in Bolivar County. On the contrary, 70.1% of the general factory jobs belonged to blacks. To therefore compare the leader and foreman positions to precision craft, as opposed to the general factory pool from which the promotions in question came is like comparing apples to watermelons. The comparison is simply irrelevant.29

A proper analysis of this issue reveals that D&L improperly considered race in its promotion practices. D&L did not provide a race neutral reason why, when 70% of the promotion pool is made up of presumptively qualified blacks, only 22.7% of the leader/foreman promotions went to blacks during Mr. Grizzard's tenure--a number which had dropped from 64.8%. It appears that the company tried to make the promotion disparity disappear. Because there is no legitimate explanation for the decrease in minority promotions, Plaintiffs must prevail. This Court should reverse and remand for a trial on the damage issue here.

C. Temporary Upgrades to Rack/General Maintenance30

The majority's review of the temporary upgrade evidence is also questionable. The majority's contention that welding experience was needed for both general maintenance and rack maintenance positions is completely refuted by Wallace Bailey, the assistant supervisor over the maintenance department. Bailey testified at trial that one who sought a temporary upgrade needed to have a mechanical background, have a familiarity with electrical equipment, have the ability to understand written and verbal instructions, and for rack, as opposed to general, maintenance positions, possess welding experience. Bailey averred that one would not be disqualified if he did not possess all of these qualities. In fact, he asserted that he would not even exclude a person simply because that person could not weld. Nevertheless, the majority contends that no blacks were qualified for the general maintenance or rack maintenance positions. The evidence appears to be to the contrary.

Tommy Hardy: Contrary to the majority's recitation of the facts, Tommy Hardy did not have "limited" welding experience. To the contrary, Hardy testified that he ran a body shop in which he welded auto parts. He started the body shop in 1975 or 1976, and it was still in operation at the time of trial. Hardy's welding experience was therefore very extensive. Hardy averred that the type of welding done at D&L was the same type of welding which he performed in his body shop. His work in his body shop also revealed that he had mechanical ability. Hardy testified that he informed Lamar Hays and Terry Lamb, supervisors in the maintenance shop, of his interest in temporary upgrades numerous times.

Daniel Anderson: Anderson testified that he had limited welding experience. However, he asserted that he had substantial mechanical ability. He had taken mechanics classes throughout his four years in high school and had weekly repaired his or other employees' machines when D&L's maintenance personnel were unavailable. Wallace Bailey testified that he would be interested in a production worker who could take apart machines, as Anderson could.

Alfred Kemp, Sr.: Wallace Bailey testified during trial that Alfred Kemp, Sr., had an excellent background in maintenance. In fact, prior to James Grizzard's arrival at the plant, Kemp received both rack and general maintenance upgrades. The district court found that Howard Watson blocked every temporary upgrade which Kemp could have received during the years in question. That finding is unsupported by the record. While Watson testified that he would prohibit Kemp from receiving temporary upgrades if Kemp were needed for his general factory work, Watson conceded that he did not know whether he had blocked any such upgrades during the years in question. Wallace Bailey asserted that he knew of no reason why Kemp was not given a single upgrade from August 1982 to August 1985.

Arthur Perry: According to Wallace Bailey, Arthur Perry was qualified for general maintenance work. However, Bailey could not state a reason why Perry had not been given upgrades prior to the filing of this lawsuit.

Absent some other explanation--which the company did not give--D&L's temporary upgrades reveal grave racial disparities. The majority correctly sets forth the temporary upgrade information in footnote 36. Were the temporary upgrade hours insubstantial, the percentages set forth by the majority in that footnote might not seem consequential. However, the number of hours involved in both the general and rack maintenance upgrades is considerable. During the Grizzard years, white employees were given 1199.4 hours of upgrades in general maintenance. Blacks received just ninety-seven.

In the rack maintenance area, Edward Otto Wolfe, a white employee, received 745 hours of upgrades during the Grizzard pre-suit period. No blacks were awarded such an upgrade during this period. The majority places some import--import which appears to be unsupported by the record--on the fact that Mr. Wolfe was the only person who received rack maintenance upgrades during the Grizzard era. James Grizzard testified in deposition that the maintenance department should not have solely awarded the upgrades to Mr. Wolfe. Likewise, Wallace Bailey asserted that he did not know why Mr. Wolfe was awarded all of the rack maintenance upgrades. The majority's supply of a reason, albeit an irrelevant and unfounded reason, would appear to be improper. See supra note 13.

Why whites received 1199.4 hours of general maintenance upgrades and 745 hours of rack maintenance upgrades when blacks received just ninety-seven hours of general maintenance and no hours of rack maintenance upgrades will likely never be known by this or any other court. In one of his depositions, Mr. Grizzard unequivocally refused to explicate why the company failed to provide everyone the opportunity to compete for temporary upgrades. He refused to provide an explanation even when instructed to give one by D&L's attorney.

The numbers--and the named Plaintiffs' protests--speak loudly enough for themselves in the temporary upgrade claim. Only silence comes from D&L. Therefore, this Court should reverse and remand for trial on damages.

For the reasons heretofore stated, this writer would reverse and remand.

I dissent.