845 F.2d 710
46 Fair Empl.Prac.Cas. 1215,
47 Empl. Prac. Dec. P 38,195, 56 USLW 2662
Marvin WASHINGTON, Plaintiff-Appellant,
ELECTRICAL JOINT APPRENTICESHIP AND TRAINING COMMITTEE OF
NORTHERN INDIANA, Defendant-Appellee.
United States Court of Appeals,
Argued Jan. 11, 1988.
Decided April 28, 1988.
Frank E. Booker, Notre Dame, Ind., for plaintiff-appellant.
William R. Groth, Fillenwarth, Dennerline, Groth & Baird, Indianapolis, Ind., for defendant-appellee.
Before WOOD, Jr., POSNER and KANNE, Circuit Judges.
POSNER, Circuit Judge.
Marvin Washington brought suit against the Electrical Joint Apprenticeship and Training Committee of Northern Indiana, charging racial discrimination in violation of Title VII of the Civil Rights Act of 1964, 42 U.S.C. Secs. 2000e et seq. The district court entered judgment for the Committee after a bench trial.
The Committee is a joint labor-management committee, see 42 U.S.C. Sec. 2000e-2(d), set up by electrical contractors and members of the electricians' union in northern Indiana to train apprentice electricians. Washington applied to the apprenticeship program and was turned down in 1982, having been ranked last among the 72 applicants for 20 openings. The ranking is determined by the applicant's score on a written test, performance in an oral interview, and academic record; there is no fixed weighting of the three components. The written test has been approved by the Department of Labor as job-related and nondiscriminatory. Washington scored "low" on the test; this is the lowest of three possible scores (the others are "medium" and "high"). Most but not all of the 20 applicants who were selected had higher scores.
Washington does not complain about the written test or his score on it, but he argues that the oral interview, and the weight given academic performance in high school and college, systematically and unjustifiably exclude blacks. The only evidence on which he relied in the district court was that none of the 20 applicants selected in 1982, when Washington was turned down, belonged to minority groups, even though more than 5 percent of the population of northern Indiana belongs to such groups. This evidence is enough, he argues, to show that the Committee's selection system has a racially disparate impact and is therefore lawful only if it is shown to be necessary to the Committee's effective operation. See, e.g., Dothard v. Rawlinson, 433 U.S. 321, 97 S.Ct. 2720, 53 L.Ed.2d 786 (1977). The district court, impressed by evidence presented by the Committee showing that between 1972 and 1982 (inclusive) 6.5 percent of the applicants selected for the apprenticeship program were members of minority groups although only 5.3 percent of the population of northern Indiana belongs to such groups, held that Washington had failed to demonstrate a disparate impact, and entered judgment for the Committee.
Washington's low score on a written test whose lawfulness he does not question, and his overall ranking at the very bottom of the applicants for a limited number of positions, make it unlikely that he would benefit from a revamped system of selection. As in any tort case, statutory or otherwise, a plaintiff cannot win a discrimination case if the harm to him would have been the same whether or not the defendant had discriminated. Mt. Healthy City School District Board of Education v. Doyle, 429 U.S. 274, 285-87, 97 S.Ct. 568, 575-76, 50 L.Ed.2d 471 (1977); Button v. Harden, 814 F.2d 382, 383 (7th Cir.1987). But we cannot be sure that Washington was not harmed by the method of selection; some other applicants who scored "low" on the written test were selected. So we must press on.
It might seem that a disparate-impact plaintiff should always be required to use statistics of applicants rather than of population. The idea behind the disparate-impact theory of a Title VII violation is that selection methods not intended to bear heavily on blacks (or other groups favored by the statute) may nevertheless have that effect, maybe by emphasizing academic qualifications that blacks disproportionately fail to meet. If there are no black applicants, or if as high a fraction of black applicants as of white applicants are hired, it might seem that the method of selection could not have a discriminatory effect, and therefore that the relevant statistics are those of the applicants rather than of the pool from which the applicants are drawn. But this analysis is superficial. See Mister v. Illinois Central Gulf R.R., 832 F.2d 1427, 1435-37 (7th Cir.1987). If the nature of the selection method is known, persons unlikely to pass through the filter that the method creates may decide not to waste their time applying.
So it was fine in principle for Washington to base his case on population statistics rather than applicant statistics. The weakness of his case lies elsewhere--in his failure to show that the Committee's methods of selection actually had a racial impact. The Committee's multi-year statistics on racial impact, which undermined Washington's single-year statistics, had been gathered pursuant to an affirmative action plan that the Committee had adopted in 1972 under prodding by the Department of Labor. The plan had set a target of 6 percent for minority selections. Washington argues that this meant 6 percent each year, not 6 percent averaged over 11 years. This is hardly a plausible interpretation, given the small number of openings each year. If, as in 1982, only 20 selections were made, a 6 percent minimum would become 10 percent, simply because the selection of one member from a minority group would fall short of 6 percent (one is only 5 percent of 20); there would have to be a minimum of two, which would be 10 percent. Washington's interpretation, which thus would have the effect, in some years anyway, of raising the minimum above 10 percent, cannot be correct--and in any event noncompliance with an affirmative action plan is not racial discrimination. Yatvin v. Madison Metropolitan School District, 840 F.2d 412, 415 (7th Cir.1988), and cases cited there.
Putting the affirmative action plan to one side, therefore, the district court could hardly be faulted for refusing to infer discrimination from the experience of a single year, involving so tiny a sample that the nonappearance of a black in it could easily be a product of chance. See, e.g., Soria v. Ozinga Bros., Inc., 704 F.2d 990, 995 (7th Cir.1983); Sengupta v. Morrison-Knudsen Co., 804 F.2d 1072, 1076 (9th Cir.1986); Connolly & Peterson, Use of Statistics in Equal Employment Opportunity Litigation Sec. 2.01 (1985). It is an unacceptable statistical procedure to turn a large sample into a small one by arbitrarily excluding observations. Suppose that, wanting to estimate the proportions of white and black marbles in a large urn without having to count all the marbles, you pick out 30 at random--and then, without looking at any of them, discard 28, thereby reducing your sample to 2. That would be a senseless way to proceed. Other things being equal, large samples are more reliable than small ones. The picture conveyed by eleven years of experience with the electricians' affirmative action program is apt to be more reliable than the experience of one year, which may well have been a fluke. And apparently the percentage of blacks selected since 1982 has been even higher than in the eleven years covered by the Committee's statistics. Perhaps years near 1982 should be weighted more heavily than much earlier (or later) years, but Washington did not suggest that the district court do this. He rested entirely on the 1982 figures. It is nowhere written that every year at least one minority applicant for training as an electrician's apprentice in northern Indiana shall be among the 20 or so most qualified applicants. Otherwise we might be treated to the spectacle of white applicants' bringing suits for discrimination against them in years when black applicants did better than could have been predicted from the percentage of blacks in the population.
It is true that if the Committee had discriminated against Washington and other blacks in 1982, it would not help the Committee to prove that in even years it discriminates against blacks but in odd years against whites, so that on average blacks and whites are treated equally. The Supreme Court rejected a variant of this "bottom line" approach in Connecticut v. Teal, 457 U.S. 440, 102 S.Ct. 2525, 73 L.Ed.2d 130 (1982). In that case, a test for eligibility for promotion excluded a disproportionate number of blacks, but of the blacks who passed the test a higher percentage were promoted than the percentage of whites, with the consequence that if the results of the test were ignored there was no disparate impact. The Court held that the persons excluded could complain about the test even if their fellow blacks who had not been excluded by it had actually been treated better than the whites. But the case would have come out differently if the basis for the disparate impact claim had been not the results of the test but a comparison between the percentage of blacks who took the test and the percentage of blacks who were promoted; for in that comparison, blacks did as well as whites. In our case the plaintiff was comparing the percentage of blacks selected with the percentage of blacks in the applicant pool (i.e., the black population of northern Indiana); and of course it was open to the defendant to show that, when that comparison was based on the proper sample, there was no disparate impact.
There is much unnecessary talk in the briefs about the prima facie case. When a discrimination suit is tried, as it was here, the question whether the plaintiff made out a prima facie case falls out of the picture, United States Postal Service Board of Governors v. Aikens, 460 U.S. 711, 715, 103 S.Ct. 1478, 1481, 75 L.Ed.2d 403 (1983); Benzies v. Illinois Dept. of Mental Health & Developmental Disabilities, 810 F.2d 146, 148 (7th Cir.1987), and the relevant question is whether the plaintiff proved discrimination. The trial judge thought not, and certainly his decision was not clearly erroneous if the only evidence of discrimination was the Committee's failure to select any black applicants in 1982; for that is far too thin a statistical reed on which to hang a finding of racial discrimination.
It hardly matters whether we say that Washington's sole statistic is not enough evidence or that it is rebutted by the defendant's statistics, which are based on a much larger sample. The cases are not clear on whether a defendant can use statistics to knock out the plaintiff's prima facie case of disparate impact--that is, to prevent it even from arising--or just to rebut it once it has arisen. See Shidaker v. Tisch, 833 F.2d 627, 632 (7th Cir.1987); Babrocky v. Jewel Food Co. & Retail Meatcutters Union, Local 320, 773 F.2d 857, 867 (7th Cir.1985); Connolly & Peterson, supra, Sec. 2.02. This is surely a nonissue. The term "prima facie case," as usually understood in law, can mean either a strong enough case to defeat a motion for a directed verdict, provided the defendant has put in no evidence of his own, or a strong enough case to shift the burden of production to the defendant, so that if the defendant produces no evidence of his own the plaintiff is entitled to a directed verdict. See, e.g., Texas Dept. of Community Affairs v. Burdine, 450 U.S. 248, 254 n. 7, 101 S.Ct. 1089, 1094 n. 7, 67 L.Ed.2d 207 (1981). The latter is the usual meaning in discrimination cases. See, e.g., id.; Ridenour v. Lawson Co., 791 F.2d 52, 55-56 (6th Cir.1986). Either way, once the defendant puts in his own evidence, the prima facie case is of no significance; so it can make no difference whether in this case the defendant's statistical evidence prevented the prima facie case from coming into existence or refuted it. This should be clear enough when a case has been tried, but it is true of every case in which the defendant puts in evidence. This case could easily have been disposed of on summary judgment, since the statistical picture as rounded out by the defendant showed that there was no genuine issue of material fact.
Washington's principal argument on appeal is that the minority population is not the right universe for determining whether the Committee's selection system is excluding a disproportionate number of blacks or other members of minority groups. The proper universe, he argues, is the set of applicants for the Committee's apprenticeship program. For lack of other opportunities, proportionately more blacks than whites in northern Indiana might want to be electricians. Suppose that half the applicants for the apprenticeship program were black; if only 6.5 percent of the applicants selected were black, this might suggest that the selection system was unfairly stacked against blacks. The figures in the record are not so dramatic as those in our example but they do show that in 1981 and 1982 there were 329 white applicants for the apprenticeship training program and 38 nonwhites, that 44 whites and only 1 nonwhite were selected, and therefore that whites were accepted at a rate more than four times as great as blacks (13.3 percent versus 2.6 percent).
As explained earlier, there is no magic in applicant statistics as opposed to statistics of population. The fact that so many blacks applied to the Committee's apprenticeship program is some evidence that the selection system is not perceived as racially biased, since if blacks thought they would be turned down on account of their race, why would they bother to apply in such numbers? See Dothard v. Rawlinson, supra, 433 U.S. at 330, 97 S.Ct. at 2727. But this reasoning is not conclusive. Applicants may be poorly informed about the effects of the selection system, may be desperate, or may simply consider the cost (principally, their time) of applying for the program acceptable even in relation to the small chance of their being accepted. Hence a case for disparate impact could be mounted from the above statistics, see e.g., Regner v. City of Chicago, 789 F.2d 534, 538 (7th Cir.1986), and it might well be strong enough to establish disparate impact and thereby shift to the defendant the burden of proving that its selection system was necessary to the effective operation of the apprenticeship program.
The only problem with this approach--but it is disabling--is that, as Washington's counsel candidly acknowledged at argument, the applicant-statistics approach to proving disparate impact was not presented to the district judge. The data underlying the statistics were in the record, being part of the back-up for the statistics presented by the Committee to rebut Washington's statistical evidence, but Washington did not direct the district judge's attention to them, argue that the disparate acceptance rate of white versus black applicants showed discrimination, or even make the computations recited above--they appeared for the first time in Washington's brief on appeal. In the district court Washington tried to prove disparate impact only by a comparison between the percentage of blacks selected in 1982 (zero) and the percentage of minority-group members in the relevant population.
A ground not presented in the trial court is forfeited, Erff v. MarkHon Industries, Inc., 781 F.2d 613, 618-19 (7th Cir.1986), with exceptions inapplicable here, such as when the ground demonstrates an absence of subject matter jurisdiction. It is unfair both to the district judge and to the defendant, and an unreasonable burden on this court, for the plaintiff to be allowed to present a new theory of liability in the court of appeals and to argue, as he must do to gain a reversal, that the judge erred in dismissing his case. Maybe had Washington presented this theory in the district court the defendants would have countered it effectively, see System Operations, Inc. v. Scientific Games Development Corp., 555 F.2d 1131, 1144-45 (3d Cir.1977); and if not the court might have ruled for him, thus obviating this appeal. While we appreciate the stringencies under which Washington's counsel, having been requested by the district court to represent Washington (who is indigent), labored, we cannot allow orderly procedure to be circumvented in this manner.