Wednesday, April 23, 2014
The Sixth Circuit just handed a resounding defeat to the EEOC in its disparate impact challenge to Kaplan Higher Education Corp.'s use of credit history to select employees, including those who will have access to student financial information. The opening paragraph signals the result by noting that the EEOC itself "runs credit checks on applicants for 64 of the agency's 97 positions." Apparently, hypocrisy isn't an asset for a federal enforcement agency.
But that point, however interesting, isn't the most provocative aspect. That honor goes to the court's treatment of the expert report upon which the EEOC based its claim that use of credit checks screen out African American applicants at a greater rate than whites. Upholding the district court's findings that the expert report was inadmissible under Daubert, the Sixth Circuit affirmed the lower court's dismissal of the EEOC complaint: without the report, the EEOC couldn't prove the requisite disparity of impact.
As the case was presented, the Commission seemed to have data on all applicants screened by one of Kaplan's vendors, so it was a simple matter to identify the exclusionary effect of the credit checks by name of applicant. The catch was, of course, that the data did not identify race, which meant that racial impact remained to be proved.
The EEOC subpoenaed state driving license records. Eleven states identified the race of license holders, but 36 states responded by providing color copies of photographs of license holders. This is where expert Murphy (or maybe I should now call him non-expert Murphy) came in. He set up a procedure for "race rating" by five raters who were experienced in "multicultural, multiracial, treatment outcome research." If 4 of the 5 agreed on a particular race for each applicant, he or she was so classified. The raters were unable to reach such a 4/5's consensus on 11.7% of the photographs, but the remainder were rated.
Now there were some technical problems with the expert report. For example, Murphy provided applicant names to the raters, which could obviously have influenced their ratings. And the racially-identified individuals were drawn from only a sample of those subjected to credit checks by one vendor, which meant that other vendors' checks were simply excluded. Add to that, there was some objective reason to doubt that the sample was in fact representative -- although Murphy did do limited "anecdotal" cross-checking of his raters' identifications with racial data provided by a state DMV, finding a 95.7% agreement.
The Sixth Circuit, however, focused not only on these kinds of problems by finding neither peer review of the technique or theory Murphy employed nor standards controlling the technique's operation. The latter referred to Murphy's failure to provide the raters with any particular standard: "they just eyeballed the DMV photos." Since the Daubert test was failed, Murphy's report was excluded, and with him went the EEOC's disparate impact case.
While it's possible to dismiss the case as another EEOC litigation failure, that's too easy a reaction. Practically speaking, the disparate impact theory obviously depends on proof of impact on different races, which means that there has to be some way to identify the races of those affected. Putting aside genetic testing (which has its own conceptual problems not to mention being clearly infeasible), the only obvious ways are self-identification or appearance.
The 11 states that provided racial data were almost certainly providing self-identified race. I suppose the EEOC could have written to all the applicants in the other states and asked them what race they were. Not likely to get much of a response -- other than predictable howls of outrage.
In any event, if those states having racial identity data didn't generate enough cases to reach statistical significance, the EEOC's "race rating" technique seems not implausible, the problems with Murphy's study aside.
But this takes us to deeper questions as to who counts as black (or white or any other race). There's been some interesting scholarship on "regarded as" discrimination, and I concur with most of the writers (albeit not all courts) that, for purposes of Title VII disparate treatment liability, what matters is the perception of the putative discriminator, not the "real" race of the victim.
But we're talking about disparate impact, and regarded as analysis seems inapposite: no one is regarding anyone as of any race is a pure impact case. The only question is whether the challenged practice falls more heavily on one racial group than another. It would seem, then, that we'd need to figure out who belongs in which group, and, for that purpose, appearance is the obvious marker. It remains to be seen whether Kaplan will be an outlier or a precursor for what has rarely been a problem in the past: who counts as what for purposes of impact analysis.