Nearly four years ago, the California Supreme Court issued its decision in the case of Duran v. U.S. Bank National Ass’n, 59 Cal. 4th 1 (2014), in which it virtually catalogued the many problems inherent in the plaintiffs’ statistical case that purported to demonstrate that a class of 260 outside salespeople were misclassified as exempt. We blogged that decision here. In a nutshell, the plaintiffs, with the trial court’s approval, had attempted to prove their case with an alleged statistical study that was plagued with problems involving sample size, questions about the statistical sample (e.g., the plaintiffs skewed the pool with class members having stronger claims), poor controls and faulty statistical methodology. This tactic worked at the trial court level, where the plaintiffs recovered approximately $15 million before a jury. The court of appeals, however, reversed, and the California Supreme Court similarly found that the verdict had to be set aside. The court in particular criticized the trial court’s trial plan and remanded the case to essentially start over from scratch, including an admonition to revisit the issue of class certification and a question as to whether any trial plan could adequately address both class and individual issues. The case was a very strong defense win in a jurisdiction known for being plaintiff friendly.

Predictably, on remand both sides moved aggressively. The defendant, for its part, moved the court affirmatively to deny certification. The plaintiff both sought additional discovery (that motion was denied) and hired a survey expert in an effort to poll the class members. Following still more procedural wrangling over whether and how such a survey could be conducted, or how the results might or might not support certification, the court ultimately granted the defendant’s motion and decertified the class. In its order, the trial court specifically noted that irrespective of the survey, the proposed class was still mired in individual issues (such as exactly how much time each class member spent in actual outside sales) that presented “severe manageability problems.”

The court of appeals affirmed. It noted continuing problems with the plaintiffs’ statistical approach and evidence. Among other problems, it noted that class members gave far larger numbers in response to the survey sent in 2015 on remand than in response to the one from 2008 used at the original trial. Although the two surveys asked the same question differently (one was total hours and the other just overtime hours), the results ranged from 13.76 hours of overtime to 23.18 hours. As both courts observed, “[e]ven a non-mathematician can readily see this is a difference of nearly ten hours per week.” The disparity in numbers simply underscored the problem that the survey did not account for bias by the respondents, who obviously skewed their answers to achieve a higher figure. Moreover, the court found that the method proposed by the plaintiffs focused on their own case and not sufficiently on the employer’s affirmative defense that, irrespective of the number of hours, at least half of their time was spent on outside sales, a largely individual inquiry and one that could also involve testimony from those outside the sample. It concluded that the trial court properly decertified the case.

The use of junk statistics in wage and hour cases is on the rise. Emboldened by the United States Supreme Court’s holding in Tyson Foods, Inc. v. Bouaphakeo (see our March 25, 2016 blog post), plaintiffs are increasingly seeking to use representative evidence to support wage and hour claims. As this most recent Duran decision demonstrates, however, a statistical case must meet basic standards and in many cases will fail on close scrutiny. Employers facing such cases should decide whether to hire their own statistical experts to point out holes in the plaintiffs’ approach.

The bottom line: Statistical arguments can be effective in wage and hour cases, but they can also be subject to challenges based on sample size, bias and statistical methodology.