Equal Employment Opportunity Commission v. Kaplan Higher Education Corp.
2014 U.S. App. LEXIS 6490
6th Cir.2014Background
- The EEOC sued Kaplan alleging its use of credit checks for certain positions has a disparate impact on African-American applicants under Title VII.
- Kaplan runs credit checks for positions with access to student financial information after past internal theft and fraud; checks flag listings like bankruptcies or judgments for review.
- The EEOC relied solely on statistical proof from expert Kevin Murphy, who used GIS vendor data (4,670 applicants) but developed a litigation-specific method to infer race from DMV driver-license photos.
- Murphy had five untrained "race raters" classify photos into five racial categories; 11.7% lacked consensus and Murphy provided raters applicants’ names (which they were told to ignore).
- Murphy produced a nonrepresentative analytic sample (1,090 applicants) with a higher overall fail rate than the full GIS pool, and submitted multiple untimely reports.
- The district court excluded Murphy’s testimony under Rule 702/Daubert for unreliable methodology and unrepresentative sampling, then granted summary judgment to Kaplan; the Sixth Circuit affirmed.
Issues
| Issue | Plaintiff's Argument | Defendant's Argument | Held |
|---|---|---|---|
| Admissibility of expert testimony under Fed. R. Evid. 702/Daubert | Murphy’s race-rating and statistical analysis are reliable and admissible to show disparate impact | Murphy’s methodology was untested, not peer-reviewed, uncontrolled, and administered by unqualified raters | Excluded Murphy’s testimony as unreliable; district court did not abuse discretion |
| Reliability of race-identification method (photo "race rating") | Anecdotal corroboration and rater consensus support reliability | Method was novel, untested, lacked standards, and raters had no relevant expertise | Method failed Daubert factors (testing, error rate, peer review, standards); unreliable |
| Representativeness of sample used for statistical inference | Sample of 1,090 applicants accurately reflects the applicant pool to show disparate impact | Sample overrepresented credit "fails" and was not shown to be representative of the full GIS pool | Sample held nonrepresentative; unreliable basis for disparate-impact proof |
| Burden of proof for admissibility | EEOC argued court should scrutinize exclusion as error | Kaplan argued EEOC bears burden to prove admissibility by preponderance | Court reiterated proponent (EEOC) bears burden to establish admissibility; EEOC failed to meet it |
Key Cases Cited
- Gen. Elec. Co. v. Joiner, 522 U.S. 136 (1997) (district-court evidentiary rulings on expert testimony reviewed for abuse of discretion)
- Daubert v. Merrell Dow Pharm., 509 U.S. 579 (1993) (trial-court gatekeeping factors for expert reliability)
- Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999) (Daubert gatekeeping applies to non-scientific expert testimony and gives courts latitude on factors)
- Nelson v. Tenn. Gas Pipeline Co., 243 F.3d 244 (6th Cir. 2001) (proponent bears burden to prove expert admissibility)