Statistical Issues Arising in Class-Action Equal Employment Cases

• Activity Number: 510
• Type: Invited
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Scientific and Public Affairs Advisory Committee
• Author(s): Weiwen Miao* and Joseph L. Gastwirth
• Companies: Haverford College and The George Washington University
• Keywords: Breslow-Day test ; Class action ; Discrimination ; Homogeneity ; Odds ratio ; Power

Abstract

The statistical analysis presented to the Dukes v. Wal-Mart II case is reviewed and an alternative analysis of the data is presented. Serious statistical issues occurred in both the plaintiff's and defendant's analyses. Assuming that the plaintiffs correctly defined the pool of employees eligible for promotion to the Manager Trainee position, it will be seen that a proper combination analysis shows that plaintiffs had a stronger argument than was presented to the court. In particular, the Breslow-Day test for binomial data shows that the odds ratios of women promotion rates to that of men are the same for all districts in the three California regions and in the state as a whole. The Gastwirth-Greenhouse analog of the Mantel-Haenszel estimated common odds ratio is 0.475. Thus, the data were consistent with a general policy that women have about one-half the odds of promotion to that of men. The analysis also demonstrates that the court's requirement that statistical significance be shown for the majority of districts within each region is far too stringent.