I concur in all of the opinion except as to whether Foss established a genuine issue of fact as to disparate impact. On that issue, I respectfully dissent.
The Portland Area Office of the Indian Health Service had a reduction in force. Mr. Foss, a twenty year employee, sought reassignment into a less senior person’s position. He was not allowed the reassignment because he was not a nurse. He argues that there is nothing about the position that needs a nurse’s training, and nurses are so overwhelmingly female that this unnecessary and inappropriate job requirement has the effect of discriminating against males. The job Foss wanted was “managed care coordinator.” A managed care coordinator does not perform nursing services for patients. Foss’s evidence shows that: (1) eight out of ten Indian Health Service area offices have managed care coordinators who are not nurses; (2) he has served as acting managed care coordinator; and (3) after the person Foss tried to displace left for other reasons, the Portland office put two people in the posi*1136tion, neither of whom had nursing degrees. Foss says that the nursing degree requirement came about simply because a person holding the managed care coordinator position when the job classification was written happened to have one.
The test for disparate impact discrimination is laid out by statute.1 The statute provides that a disparate impact claim is established when:
A complaining party demonstrates that [an employer] uses a particular employment practice that causes a disparate impact on the basis of race, color, religion, sex, or national origin and the [employer] fails to demonstrate that the challenged practice is job related for the position in question and consistent with business necessity. [2]
So to establish a prima facie case of disparate impact, a plaintiff must: (1) identify the practice being challenged; (2) show disparate impact; and (3) prove causation.3 The burden then shifts to the defendant to either: (1) discredit plaintiffs clam of disparate impact, as for example by demonstrating that plaintiffs own statistics do not show it; or (2) produce evidence that the practice is job related and consistent with business necessity.4
The majority does not quote the statute, but its discussion of statistics can only fit into the first step of the statutory analysis because the opinion holds that “Foss failed to make a prima facie case.” That step requires the complaining party to demonstrate that the employer “uses a particular employment practice that causes a disparate impact on the basis of ... sex.” The employment practice at issue is the requirement of a nursing degree. The majority applies the rule that the disparate impact must be shown with respect to the qualified population in the relevant labor market, and says that Foss’s statistics do not show that the sex disparity in the relevant labor market is significant.5 This is the issue on which we diverge.
Foss correctly argues that because only one position was at issue, and only two people, himself and the incumbent, were the labor market for filling it, our exception for an “extremely small universe” applies.6 We do not require statistics in such a case because they would be meaningless. The statute itself does not speak in statistical terms. The statutory prima facie case requires a “disparate impact on the basis of-sex,” which would ordinarily be statistical, but not necessarily. Foss cites the EEOC guideline that says that if a requirement causes the selection rate for one sex to be less than % of the rate for the other, the EEOC will regard the under 80% rate as “evidence of adverse impact.”7 Foss presented statistics showing that 88% of the nurses employed by the Indian Health Service are female, that 91% of the nurses who would be eligible to “bump” into the position were female, and that the labor market for “bumping” in a “RIF” was limited to people who worked for the Indian Health Service. Thus the job requirement of a nursing degree limits those eligible for the position to a pool that is 91% female.
If an arbitrary requirement limits the job to a pool that is 91% one sex, while *1137non-arbitrary job-related requirements would not, that would satisfy the statutory requirement for a prima facie case. The only thing Foss leaves out of his statistical case is the sex composition of IHS employees who could “bump” the incumbent were the nursing degree requirement absent. But he does present a proxy for this unknown number, by showing that a slight majority of the managed care coordinators are male in the Indian Health Service offices (all but two) that do not require a nursing degree.' That supports the inference that with the nursing degree requirement, the position is largely limited to females, but without it, there would be a fairly even sex distribution among the relevant pool of Indian Health Service employees eligible for the job.
The majority opinion says that the inference that the nursing degree limits the eligibles to a pool substantially slanted to one sex is too “conjectural” to establish a genuine issue of fact.8 With the requirement, the position could be expected to be filled by about 90% females and is in fact filled by 100% females. What is too conjectural is the counter inference that the majority opinion tacitly requires. If the nursing degree requirement were absent, then for the eligible pool to still be 90% female, the genuinely “job related ... and consistent with business necessity” requirements would have to leave the qualified pool at about 90% female. Why should the pool of persons qualified to be a managed care coordinator be 90% female even without the challenged nursing degree requirement? That possibility seems unlikely on its face, and is belied by the evidence that in the offices without the nursing degree requirement, the position is filled by a mixed sex group, slightly more male than female.
The majority has invited me into a statistical discussion that makes the case seem more abstruse than it really is. The employer has a requirement for the managed care coordinator position (in this office but not most of its other offices) that effectively and practically limits the job to persons of one sex. With that requirement, nine out of ten people who can fill the job are limited to one sex. And 100% of the people who got the position in the Portland Area Office were of one sex. Without it, the sex distribution of those filling the job is pretty evenly balanced. That is enough to allow a trier of fact to infer that the requirement “causes a disparate impact on the basis of ... sex.” As for whether the requirement of a nursing degree is arbitrary, like requiring high school football experience to be a nurse, or job related, like requiring a law degree to be a public defender, it is the employer’s burden to show that, and the majority errs by not allowing the case to reach that issue.9
. See 42 U.S.C. § 2000e-2(k)(i).
. Id.
. See Rose v. Wells Fargo & Co., 902 F.2d 1417, 1424 (9th Cir.1990).
. See id:
. Maj. Op. at 1134-35.
. See Morita v. Southern California Permanente Medical Group, 541 F.2d 217, 220 (9th Cir.1976).
.See 29 C.F.R. § 1607.4D which provides in pertinent part:
Adverse impact and the "four-fifths rule.” A selection rate for any race, sex, on ethnic group which is less than four-fifths ('%) (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact....
. Maj. Op. at 1135.
. Cf. Washington v. Garrett, 10 F.3d 1421, 1433 (9th Cir.1993) ("[TJhere will always be a question for the factfinder once a plaintiff establishes a prima facie case and raises a genuine issue as to whether the employer's explanation for its action is true.”).