concurring in part and dissenting in part:
I agree with the majority’s disposition of the class promotion and retaliation claims. The majority properly remanded these class claims for more comprehensive findings of fact. The majority also properly concludes that the district court has jurisdiction to hear Ms. Kobylinski’s claim. I believe, however, that the district court properly found that the plaintiffs had failed to establish a prima facie case of discrimination because the plaintiff’s statistical evidence does not support an inference of discrimination. Therefore, I cannot agree with the majority’s conclusion that the district court improperly discounted the statistical evidence when determining whether a prima facie case had been established.
The majority concludes that the plaintiffs’ statistical proof1 is sufficient to support an inference of discrimination and, thus, that the district court erred in assigning little weight to the evidence. It reasons that the relevant labor market, as defined by the plaintiffs, adequately reflects the qualifications necessary for employment by the Agency.2 Although it concedes that many Agency positions require skill in a foreign language the majority concludes that the plaintiffs’ failure to adjust their definition of the relevant labor market to account for this factor does not create a serious flaw in the statistical comparisons. In support of this conclusion, the majority asserts that we may logically assume that foreign language skills are equally distributed between men and women and that statistical data regarding foreign language skill may not be available. Finally, the majority indicates that the defendant should carry the burden of rebutting the plaintiffs’ statistical evidence by showing that female applicants are not so likely as male applicants to possess the requisite language skills.
I believe that the district court correctly concluded that the plaintiffs failed to establish a prima facie case. The question of whether the plaintiffs have established a prima facie case depends upon the weight assigned the statistical proof presented by the parties. The plaintiff, in order to establish a prima facie case, must produce sufficient evidence to raise an inference of discrimination on the part of the defendant. International Brotherhood of Teamsters v. United States, 431 U.S. 324, 358, 97 S.Ct. 1843, 1866, 52 L.Ed.2d 396 (1977); Valentino v. United States Postal Service, 674 F.2d 56, 67 (D.C.Cir.1982). When the plaintiffs seeks to raise the inference by means of statistical proof, they have the “burden of comparing appropriate groups in terms of minimum objective qualifications ...” Id. at 68. See Id. at 68 n.17, 71 n.24; Wilkins v. University of Houston, 654 F.2d 388, 408 (5th Cir. 1981). Thus, the plaintiffs must either produce statistical proof which compares the Agency’s workforce with that segment of the labor market which possesses the minimum skills necessary for employment by the Agency (including language skills) or demonstrate the validity of their comparisons by showing that foreign language skill is not a minimum qualification for the job categories being compared. The plaintiffs have failed to make either showing; consequently, the statistical evidence is insufficient to raise an inference of discrimination and, thus, is insufficient to establish a prima facie case of discrimination.
The failure of the plaintiffs to include foreign language skill as a factor in defining the relevant labor market seriously flaws the statistical comparison made by *390the plaintiffs. Generally, the relevant labor market is the group of workers from which the employer hires its employees. United States v. Ironworkers Local 86, 443 F.2d 544 (9th Cir.), cert. denied, 404 U.S. 984, 92 S.Ct. 447, 30 L.Ed.2d 367 (1971). When properly defined, the relevant labor market should contain only persons with the minimum qualifications necessary for the pertinent positions. See Hazelwood School District v. United States, 433 U.S. 299, 308 n.13, 97 S.Ct. 2736, 2742 n.13, 53 L.Ed.2d 768 (1977); Ste. Marie v. Eastern Railway Ass’n, 650 F.2d 395 (2nd Cir. 1981). If a comparison is made to an improperly defined labor market, the conclusions drawn may be misleading, because the data are likely to distort the population of a particular minority available to fill a particular position. See Mayor v. Educational Equality League, 415 U.S. 605, 620-21, 94 S.Ct. 1323, 1333, 39 L.Ed.2d 630 (1974); Davis v. Califano, 613 F.2d 957, 964 (D.C.Cir.1979); Wilkins v. University of Houston, 654 F.2d 388, 398 n.13 (5th Cir. 1981). This circuit has concluded, therefore, that the relevant labor market includes only those persons who possess the minimum objective qualifications necessary for a person to be eligible for a particular position. Davis v. Califano, 613 F.2d at 964.
In light of these principles, the labor market used by the plaintiffs for purposes of comparison is overly broad because the foreign language skills were not considered in defining the market.3 For example, the evidence indicated that a substantial disparity exists between the number of women employed by the Agency as writers and the number of available female writers. The relevant market, however, is narrower than the plaintiffs grouping. The district court found that many writers for the Agency must be fluent in a second language. The parties agree that this qualification is necessary, see Appellants’ Brief, p. 50; a writer for the Agency must prepare materials for dissemination in other countries. The plaintiffs’ statistical comparison fails to account for the additional foreign language qualifications; therefore, the data are too general to establish the existence or cause of any disparity. Because of this imprecision, the district court properly refused to infer discriminations from the plaintiffs’ statistical comparisons.
The plaintiffs failed to establish that their evidence was entitled to more weight, by showing that foreign language skill was not a “minimum objective qualification.” Davis v. Califano, 613 F.2d at 964. Similarly, the plaintiffs produced no evidence indicating that language skills are evenly distributed among men and women.4 In short, the plaintiffs have failed to produce evidence which would support an inference of discrimination. Because the plaintiffs have failed to produce sufficient evidence, the district court’s conclusion that the plaintiffs *391did not establish a prima facie case should be affirmed.
. The plaintiffs statistical proof indicated that substantial disparities existed in four job categories: electronic technicians, radio broadcast technicians, writers/editors, and foreign information specialists. The plaintiffs’ expert concluded that these disparities were statistically significant, with less than .05 probability that the disparities resulted from chance.
. The experts of both parties compared the Agency job categories with the most similar job categories used by the Census Bureau in compiling its statistics. Upon completing the “cross-mapping,” the expert then compared, in each job category, the percentage of the external labor force which is female with the percentage of the Agency’s labor force which is female.
. The plaintiffs’ statistics cannot be used to raise an inference of discrimination, as the majority urges, by means of an assumption that foreign language skills are evenly distributed among men and women. Such an assumption cannot be properly made by this court. First, the assumption is based on the premise that language skills are randomly distributed among the population. The distribution may not be random, however, because, as the majority notes, at note 8 supra, many factors not contained in the record may affect the distribution of language skills, such as the relative rate of immigration. Moreover, the majority’s assumption unfairly excuses the plaintiffs’ failure to produce evidence on the matter. Here, for example, the plaintiffs might have produced evidence regarding the distribution of language skills among the male and female applicants for positions with the Agency. This applicant flow data would have been sufficient to indicate the distribution of language skills. The plaintiffs bear the burden of producing such evidence, Valentino v. United States Postal Service, 674 F.2d at 68 n.17, 67 n.24 and this court should not relieve them from meeting this obligation by assuming the condition.
. The plaintiffs have not produced any evidence indicating which jobs require foreign language skills. Thus, we do not know which positions require foreign language skills. The consequences of this failure must be borne by the plaintiffs, because they bear the burden of raising an inference of discrimination, Valentino, 674 F.2d at 68 n.17, and because they bear the burden of demonstrating the relevance of their statistical evidence. See Ste. Marie v. Eastern Railway Ass’n., 650 F.2d 395, 401 n.6 (2nd Cir. 1981). See generally Valentino, 674 F.2d at 71 n.24.