These appeals call upon us to review the judgment of the district court that appellee Jeffboat, Incorporated ("Jeffboat"), did not discriminate against the appellant employees and the class they represent. The case was tried to the court below for almost four months and after the trial the district court ruled that the plaintiffs had failed to establish by a preponderance of the evidence that they were discriminated against. Regretably, we find the findings of fact made after trial by the district court insufficient to permit meaningful appellate review. Therefore, we vacate the judgment of the district court and remand the case for a new trial.

Jeffboat is engaged in the manufacture and construction of barges, towboats and other marine-related equipment. Its facilities are located in Jeffersonville, Indiana, which is close to Louisville, Kentucky. From 1968 to 1980, the number of hourly employees at Jeffboat varied between 700 and 2000 with an average of 1500 such employees. During the same period, the number of salaried employees increased from 140 to 280.

At all relevant times a fairly large percentage of the hourly employees at Jeffboat were black. From 1967 to 1971, the hourly employees were represented by the Industrial Union of Marine and Shipbuilding Workers of America Local 65. In 1971, Local 65 was decertified and replaced by General Drivers, Warehousemen and Helpers, Local Union No. 89.

In 1977, the appellant employees filed a class action complaint charging Jeffboat with racial discrimination against its black employees in promotion, compensation, discipline, seniority, training and other practices. The complaint alleged that Jeffboat's practices violated both Title VII of the Civil Rights Act of 1964, 42 U.S.C. Secs. 2000e et seq. ("Title VII"), and 42 U.S.C. Sec. 1981.1 The classes were made up of all present black Jeffboat bargaining unit (hourly) employees and all such employees who are no longer employed by Jeffboat but who were employed within the relevant limitations period.2

The claim of discrimination in discipline and discharge requires examination of union collective bargaining agreements which regulated the imposition of discipline by Jeffboat. Violations of disciplinary rules were classified as minor, major or intolerable. Further, the collective bargaining agreements specified the maximum allowable penalty for each type of offense. For example, the first minor offense could draw, at most, a written reprimand. A second minor offense was punishable by a written reprimand and a three day suspension.3 Some minor changes in the severity of the penalties were made during the relevant years but, with one exception, the structure of the discipline system remained generally the same.

The one exceptional change noted above was that, after 1974, discipline for tardiness and absenteeism was deleted from the list of minor offenses and was provided for separately from other offenses. At that time also, the maximum penalties for attendance infractions were increased. Apparently, Jeffboat intended to "get tough" on matters of absenteeism and tardiness by imposing maximum penalties for such infractions. However, plaintiffs have alleged, and Jeffboat has not denied, that Jeffboat officials retained substantial discretion in these matters. Evidence was presented at trial that white employees often were not given the maximum penalties for attendance violations. See Brief for Appellants at 16 n. 20.

The plaintiffs conducted a statistical analysis designed to measure the disparity in the discharge rates of black and white hourly employees. The data for the analysis were drawn from information supplied by Jeffboat during discovery. The analysis revealed that black hourly employees were discharged at a significantly higher rate in relation to their numbers than were their white counterparts.4 The results were the same when the analysis was conducted for different time periods and when the study was confined to terminations for specific rule infractions.

With respect to allegations of discrimination in promotions, we must examine the seniority provisions of the collective bargaining agreement. The contract provided for sixteen separate seniority units and also employed the concept of plant wide seniority. Further, the agreement provided that promotions were to be made from within a seniority unit when possible and, if two equally qualified employees applied for a job, seniority would govern.

One of the class claims was that Jeffboat discriminated in the selection of "leadmen." A leadman is an hourly employee who directs other hourly employees apparently as a surrogate for, or an assistant to, a foreman. The collective bargaining agreements provided that leadmen could be selected without regard to seniority and that such selections were not subject to the grievance procedure. Experience as a leadman apparently was a proving ground for ultimate promotion to foreman. Jeffboat's vice president of production estimated at trial that only 25 percent of foremen had not been leadmen. The plaintiffs' statistical analysis of promotions to leadman showed a relatively low statistical disparity between the rates of black and of white promotion. Plaintiffs in their brief allege, however, that no blacks were among the 96 persons promoted to leadman status between 1971 and 1978. Jeffboat, in its brief, has not disputed this allegation.

The complaint also alleges that Jeffboat discriminated in its selection of foremen. Between 1971 and 1978, 99 people were promoted to foreman, of whom only three were black. The district court declined to consider this allegation on the ground that the position of foreman, to which promotion was made, was outside the class. However, as is elaborated below, discrimination against class members in promotion to positions outside the class is a proper class claim--and thus on retrial the discrimination claims relating to promotions to foreman must be considered.

Plaintiffs also allege that Jeffboat discriminated against black employees by punishing them in retaliation for their participation in protests against Jeffboat's treatment of black workers. The protests--called "Black Days"--were held on work days, requiring that participants be absent from work. The plaintiffs claim that the discipline imposed for these absences violated Title VII's prohibition against retaliation for protesting unlawful discrimination.

There were several other allegations, all of which were rejected by the district court, including charges that the defendants generally practiced racial discrimination in the selection of employees for higher paying jobs. In addition, plaintiffs allege that at least one of the black plaintiffs, who was an alcoholic, was treated more harshly than white alcoholic employees. There is also evidence that Jeffboat failed to maintain an affirmative action plan as required under a consent decree it had entered into with the Indiana Civil Rights Commission and also as mandated because Jeffboat was a government contractor. In general, the plaintiffs attempted to paint a portrait of Jeffboat as a company afflicted with widespread racism.

In this Title VII case, our ultimate task is "to determine whether the [plaintiffs have] proven that the defendant discriminated against [them]." Epstein v. Secretary, 739 F.2d 274, 278 (7th Cir.1984).5 In Epstein we recognized that in this circuit, our review of subsidiary factual determinations is governed by the clearly erroneous test while our review of the ultimate fact of discrimination is not so constrained. Id., at 278-79. In the case before us, the plaintiffs argue that many of the district court's factual determinations are erroneous. The plaintiffs also contend that the district court ignored, or refused to consider, whole categories of evidence. As is explained below, we agree that the district court erred in its treatment of important portions of the plaintiffs' evidence. In the same vein, we find that the district court's findings of fact are inadequate and fail to comply with FED.R.CIV.P. 52(a). In these circumstances, we think we have little choice but to order a new trial.

The first and perhaps most striking error committed by the district court is its failure to address the significance of what may be called plaintiffs' "comparative evidence." This evidence consists of voluminous anecdotal evidence which compares the treatment of black employees with the treatment of allegedly similarly situated white employees. Comparative evidence was presented to support many of the individual and class claims including the allegations of discrimination in discipline, discharge, promotions, transfers, leaves of absence and attendance policies. See, e.g., Brief for Appellants, notes 19, 20, 26, 70, 82, 93, 103, 104, 105, 156 and accompanying text. This evidence was offered to establish that there was a pattern of treating black employees more severely than white employees. For example, the plaintiffs used comparative evidence to try to prove that black employees were given the maximum allowable penalty under the union contract for attendance violations while white employees were given less severe penalties.

The district court held, in fact, that, as long as the maximum penalty provided by the union contract was not exceeded, there was no violation of the law. See District Court's Findings of Fact, Conclusions of Law and Judgment at 60-61 (hereinafter "Findings"). This was error. As a hypothetical example, suppose the maximum penalty for one unverified absence were a one day suspension. Suppose further that ten white employees and ten black employees were absent for one day and that all ten black employees received one day suspensions but that the ten white employees received only written reprimands, a less severe penalty. From this evidence one could readily infer discrimination against blacks, and the mere fact that all of the black employees received penalties authorized by the collective bargaining agreement would not immunize the employer from liability for racial discrimination.

The district court's failure even to address this comparative evidence requires us to remand the case for retrial. This court, as a court of review, is not equipped to determine, as an original matter, the appropriate weight to be accorded the plaintiffs' evidence. Nor are we able to weigh this evidence against the testimony and documentary evidence presented by the defendant. In many respects, the case before us is thus similar to Rucker v. Higher Educational Aids Board, 669 F.2d 1179, 1183 (7th Cir.1982), in which we ordered a remand because of inadequate subsidiary findings.6 While we recognize that the district court cannot be expected to explain the significance of every bit of evidence in the record, the court's failure in this case to address the sort of evidence we have noted, and other evidence discussed infra, violates the command of Rule 52(a) to "find the facts specially" and precludes effective appellate review. See Kelley v. Everglades Drainage Dist., 319 U.S. 415, 422, 63 S. Ct. 1141, 1145, 87 L. Ed. 1485 (1943); Schneiderman v. United States, 320 U.S. 118, 129-30, 63 S. Ct. 1333, 1338-39, 87 L. Ed. 1796 (1943); Rucker, 669 F.2d at 1183; Denofre v. Transportation Insurance Rating Bureau, 532 F.2d 43, 45 (7th Cir.1976). The district court thus made the necessary ultimate finding that there was no discrimination, but it failed to make the subsidiary findings necessary for us to follow its chain of reasoning. The Supreme Court's observation in determining findings of fact to be inadequate in Schneiderman is thus particularly apposite here: "The pertinent findings of fact ... are but the most general conclusions of ultimate fact. It is impossible to tell from them upon what underlying facts the court relied and whether proper statutory standards were observed." Schneiderman, 320 U.S. at 129-30, 63 S.Ct. at 1338-39 (footnote omitted). See also Dalehite v. United States, 346 U.S. 15, 24 n. 8, 73 S. Ct. 956, 962 n. 8, 97 L. Ed. 1427 (1953); 9 C. WRIGHT & A. MILLER, FEDERAL PRACTICE AND PROCEDURE Sec. 2577 (1971) (appellate court may, due to inadequate findings, "order a new trial.")7

While we believe that the district court's failure to make findings with respect to the plaintiffs' comparative evidence is alone sufficient to merit retrial, there are other alleged errors which we find it appropriate to resolve so as to provide guidance on retrial.

For example, the district court appears to have refused to consider plaintiffs' claim that promotions to foreman and other supervisory positions were awarded in a discriminatory fashion. The district court wrote:

The positions of foreman, general foreman, superintendent, and all other salaried promotions, are outside the terms of the collective bargaining agreements [sic] and Local Union No. 89. The class, being limited to hourly employees within the bargaining unit of Local Union No. 89, does not include the positions of foreman, general foreman and superintendent. Thus limited, the plaintiffs' claims that defendant used subjective criteria for the selection of these salaried positions is not a proper class claim.

Findings at 57. Contrary to the district court's conclusion, we do not think that the fact that a promotion would place the complainant outside of the plaintiff class can immunize the promotion from Title VII scrutiny. The Supreme Court's recent decision in Hishon v. King & Spalding, --- U.S. ----, 104 S. Ct. 2229, 2235, 81 L. Ed. 2d 59 (1984), illustrates the error of the district court's reasoning. In Hishon, the Supreme Court decided that although the status of partner falls outside the domain of Title VII, advancement to partnership in a law firm was a "term, condition or privilege of employment for purposes of Title VII." 104 S. Ct. at 2235. Likewise, in this case, consideration for promotion to supervisory positions appears to be a privilege of employment and, as such, promotion decisions may not be made in a discriminatory manner. Therefore, on remand the district court should consider the evidence relative to the alleged promotion practices. See United States Postal Service Board of Governors v. Aikens, 460 U.S. 711, 103 S. Ct. 1478, 75 L. Ed. 2d 403 (1983) (applying traditional Title VII prima facie case criteria to allegation of discrimination in promotion).

Another deficiency which makes it extremely difficult to review the findings of the district court is the inconclusive discussion of the subjective factors involved in many of the employment decisions challenged here. We recognize that frequently subjective and other intangible factors may influence employment decisions and that even subjective misjudgments may not necessarily be the basis for Title VII liability. See Nellis v. Brown County, 722 F.2d 853, 860-61 (7th Cir.1983). However, merely stating that there is a subjective factor in a decision cannot alone serve to rebut a prima facie showing of discrimination. The district court, and, for that matter, the defendant on appeal, did not specify what subjective factors they relied on. And, in some instances where the defendant has offered the subjective nature of a decision as a purportedly nondiscriminatory explanation for an employment action, the plaintiffs have presented evidence tending to rebut these explanations relying on subjective factors. The district court, however, did not make any findings regarding this evidence offered by the plaintiffs. The mere ritual invocation of "subjective factors" does not serve to rebut all evidence tending to establish a Title VII or a Sec. 1981 violation. Therefore, the district court's finding that subjective factors entered into many of the decisions does not relieve the court of its obligation to make specific subsidiary findings that would facilitate appellate review.

In addition, to a significant extent, the allegedly sui generis nature of many of the defendant's decisions convinced the district court that plaintiffs' statistical evidence was useless. For example, the district court held that "[b]ecause each discharge and/or disciplinary action is conducted on a case-by-case basis and involves a wide variety of conduct and rule violations, such is not readily susceptible to statistical analysis." Findings at 62. A similar conclusion was reached as to the statistics on promotions. See Findings at 58. In this connection, we note here that the statistical disparities shown in this case were, in some instances, extremely large. Of course, determining the technical validity of statistical evidence is often highly problematic. See Soria v. Ozinga Bros., Inc., 704 F.2d 990 (7th Cir.1983). However, the plaintiffs argue that in this case their "statistics were more finely tuned" than those accepted by the Supreme Court in Teamsters v. United States, 431 U.S. 324, 337, 97 S. Ct. 1843, 1855, 52 L. Ed. 2d 396 (1977) or Hazelwood School District v. United States, 433 U.S. 299, 97 S. Ct. 2736, 53 L. Ed. 2d 768 (1977). Brief for Appellants at 60. See also Castaneda v. Partida, 430 U.S. 482, 97 S. Ct. 1272, 51 L. Ed. 2d 498 (1977).

The district court found the plaintiffs' discipline statistics to be of "no value" because "each disciplinary action and/or discharge utilizes a separate factual situation, considering the employee's past and present conduct, and a separate rule violation." Findings at 62. This observation may contain a grain of truth; but it does not necessarily signify that the statistics in question are worthless. We note that statistics are an effective means of measuring the cumulative effects of employment actions and that statistically significant disparities in the treatment of comparably situated persons are often quite probative of unlawful discrimination. Of course, the difficulty of showing comparability of situations is a significant limitation. See Segar v. Smith, 738 F.2d 1249, 1277-78 (D.C.Cir.1984); Lilly v. Harris-Teeter Supermarket, 720 F.2d 326, 336 & n. 20 (4th Cir.1983), cert. denied, --- U.S. ----, 104 S. Ct. 2154, 80 L. Ed. 2d 539 (1984). See also Castaneda v. Partida, 430 U.S. 482, 497 n. 17, 97 S. Ct. 1272, 1281 n. 17, 51 L. Ed. 2d 498 (1977).

Plaintiffs argue that the district court's reliance on the subjective nature of the imposition of discipline would effectively shield all but the most unarguably objectively comparable practices from statistical analysis. We recognize that the probative value of statistical evidence may be significantly undermined by information which casts into question the comparability of data. As noted, the district court generally failed to identify the subjective and other factors that led it to disregard the statistical evidence. This failure makes it impossible for us to determine on review what weight the statistical evidence should have been accorded.

Further, in a case such as this where the statistical pool consists of persons already employed by the defendant, courts have sometimes assumed that desirable employee traits are evenly distributed among the relevant pool. See Segar v. Smith, 738 F.2d at 1276-77. In fact, in Segar the court noted that it is appropriate to exclude subjective factors from statistical analysis because "subjective criteria may well serve as a veil of seeming legitimacy behind which illegal discrimination is operating." Id. at 1276. See also Harrell v. Northern Electric Co., 672 F.2d 444, modified on other grounds, 679 F.2d 31 (5th Cir.), cert. denied, 459 U.S. 1037, 103 S. Ct. 449, 74 L. Ed. 2d 603 (1982). We recognize, however, that the probative value of statistical evidence may be diminished where an employer genuinely relies on subjective factors in making decisions and where data are incomparable. For present purposes, we urge the district court to reexamine the statistical evidence and to make specific findings about the factors affecting its reliability and about the weight, if any, to be assigned it.

The district court also refused to consider the plaintiffs' promotion statistics.8 The district court found that the plaintiffs' statistical analysis of promotions failed to narrow the applicable labor pool to employees with the requisite minimum qualifications. Findings at 58. The district court did not specify what the required qualifications were or what the proper labor pool would have been. To the extent that these minimum qualifications are subjective, the district court's decision here suffers from the same infirmities as its decision with respect to disciplinary statistics. If the minimum qualifications are significantly objective, the district court should have specified what these objective qualifications were. Without these findings, we are unable to determine whether, and to what extent, the promotion statistics should have been taken into account. Cf. Davis v. Califano, 613 F.2d 957, 964 (D.C.Cir.1979) ("only the minimum objective qualifications necessary for one to be eligible for promotion must be considered in the statistical data") (emphasis in the original).

In particular, we think that the significance of all of the statistics may be related to the probative impact of the comparative evidence. The stark statistical disparities presented here, if accompanied by representative anecdotal evidence that black and white employees were treated differently in similar situations, may present highly relevant evidence which the district court should consider on remand. See Valentino v. United States Postal Service, 674 F.2d 56, 68 (D.C.Cir.1982) (quoting Teamsters v. United States, 431 U.S. 324, 338-39, 97 S. Ct. 1843, 1855-56, 52 L. Ed. 2d 396 (1977)) (anecdotal evidence may " '[bring] the cold numbers convincingly to life' "). However, we cannot usurp the role of the district court in evaluating and weighing the evidence. And the plaintiffs' arguments with respect to the statistical evidence must be considered by the district court in the first instance.

Another issue that has not been adequately dealt with by the district court is the question whether Jeffboat retaliated against the plaintiffs for participating in protests against unlawful discrimination. In this connection, section 704(a) of the Civil Rights Act of 1964, 42 U.S.C. Sec. 2000e-3(a) prohibits retaliation by an employer against its employees for opposing unlawful employment practices.9 The plaintiffs alleged that Jeffboat retaliated unlawfully for their participation in "Black Days," which were a series of protests against Jeffboat's allegedly discriminatory practices. The district court rejected this claim on the following ground:

The record discloses and the Court finds that the defendant treated absenteeism and tardiness on these "Black Days" in the same manner as it treated absenteeism and tardiness of Caucasian employees for the Louisville, Kentucky, anti-school busing protest days. In both instances, the absenteeism and tardiness was [sic] processed under the terms of the applicable collective bargaining agreements. No special treatment, either in the form of retaliation or favoritism, was shown to either the Negro or Caucasian employees who were absent on these days.

Findings at 65. The district court thus concluded that Jeffboat did not retaliate against black employees for their participation in Black Days.

But the district court's discussion, as quoted above, fails to mention the elements of a claim for unlawful retaliation. To prevail on such a claim, the plaintiffs would have had to show, as a threshold matter, that they reasonably believed that Jeffboat practiced unlawful discrimination and that their opposition to the discrimination caused the employer to take the adverse action alleged. See McCluney v. Jos. Schlitz Brewing Co., 728 F.2d 924, 928 (7th Cir.1984). Whether the retaliatory action taken by the employer was authorized by the collective bargaining agreement would not appear to be dispositive, or even highly relevant. The treatment of white employees who protested school busing in Louisville, Kentucky seems similarly inapposite. The latter protests by whites seem to be wholly unrelated to practices at Jeffboat and are not dispositive.

Of course, employees may not engage in conduct which is "excessively disloyal or hostile or disruptive and damaging to the employer's business" and then claim the shield of Sec. 704(a) as protection against adverse action. Croushorn v. Board of Trustees, 518 F. Supp. 9, 25 (M.D.Tenn.1980). See Hochstadt v. Worcester Foundation for Experimental Biology, Inc., 545 F.2d 222, 230-31 (1st Cir.1976). Further, only lawful activity is protected by Sec. 704(a). Croushorn, 518 F. Supp. at 25. There is a paucity of authority directly addressing the question whether or when concerted absence from work is protected.10 However, the cases establish that under Sec. 704(a), "a court must balance the setting in which the activity arises and the interests and motivations of both employer and employee." Hochstadt, 545 F.2d at 232. On remand, the district court should consider the retaliation claim in light of the burdens of proof and established law on the subject as developed in cases such as Rucker, supra; McCluney, supra; EEOC v. St. Anne's Hospital, 664 F.2d 128 (7th Cir.1981) and Hochstadt, supra. Specifically, the district court should determine whether the elements of a claim under McCluney have been established. If the McCluney elements are present, the district court should then proceed to balance any disruption caused by plaintiffs' absences against Black Days' advancement of Title VII's policy of eliminating discrimination.

Other issues have been raised in this case that need not be addressed in detail because they will be more susceptible of resolution after a new trial and new findings of fact. In particular, the issue whether Jeffboat lived up to its affirmative action obligation under the consent decree with the Indiana Civil Rights Commission will need to be addressed at a new trial. Further, we are not certain what effect the seniority provisions of the union contracts have on plaintiffs' claims with respect to promotion in rank and to higher paying jobs. We recognize that Title VII protects the operation of the seniority agreement unless the agreement was adopted with the intent to discriminate. 42 U.S.C. Sec. 2000e-2(h). See Teamsters v. United States, 431 U.S. 324, 97 S. Ct. 1843, 52 L. Ed. 2d 396 (1977). The district court found that the seniority agreement in this case was not adopted for discriminatory purposes and that often the most senior "qualified" employee was chosen for a new job. However, we are unable to discern from the court's findings precisely what effect the seniority agreement had on the promotion decisions or whether there might have been discrimination in the identification of qualified applicants for a job. On retrial, the district court must make written findings as to the crucial issues of fact and not be content with mental notes. See Rucker, 669 F.2d at 1183.

We emphasize that we have not specifically addressed other issues raised on appeal. On remand, the district court, in exercising its judgment in matters normally within its province, should not be constrained by our failure to address an issue here. Further, the district court should, in general, not feel bound by the law of the case doctrine or some like constraint to adhere to the modes of analysis or procedures pursued at the first trial unless it is persuaded of their correctness.

The judgment is vacated and the case is remanded for a new trial on all of plaintiffs' claims.