Monumentally flawed statistics in landmark overtime class action

In Duran v. U.S. Bank National Association, a class of 260 bank officers sued their employer for unpaid overtime resulting from their alleged misclassification as exempt employees. The trial court ignored the advise of both sides’ statistical experts and (poorly) executed their own plan to utilize statistics in the case.

Background

The plaintiffs were loan officers hired as outside salespersons; they argued they were misclassified as exempt from overtime pay and minimum wage requirements. The case hinged on if the plaintiffs spent more than half their time performing sales tasks outside of a U.S. Bank branch. If they did, then the exemption applied, and no overtime damages were due.

During the class certification proceedings, the employer provided testimony from 79 loan officers confirming they spent more than 50% of their day engaged in outside sales. The plaintiffs presented declarations from 34 loan officers confirming the opposite. The court chose to certify the class.

The use of statistics in the trial was deeply flawed

Shaun J. Voigt of Fisher & Phillips LLP wrote that “the trial court largely ignored recommendations from the statistical experts employed by both parties, and instead devised its own trial plan…”

The court’s two phase plan began first by assessing liability, for which the court conducted liability proof testing using the testimony of 20 randomly selected plaintiffs. In the second phase of the trial, damages were calculated. The plaintiffs’ statistical experts reported that class members worked, on average, 11.87 hours of weekly overtime, with a preposterously high margin of error of 43.3%.

The court extrapolated from the small sample the amount of each class member’s uncompensated overtime, and dismissed the defendant’s statistical experts testimony critiquing the flawed opposing expert report. The trial court awarded $14,959,565 in damages to the plaintiffs.

The defendant was barred from presenting any evidence involving class members who were not part of the sample set, despite that evidence being able to illustrate the the variance between class members.

The appeal court decertified the class and reversed the trial court’s decision. A petition for review was granted, and the California Supreme Court found the trial court’s handling of the case to be ‘profoundly flawed’.

What this case made clear

Predominance of common issues is not sufficient evidence to certify a class, and statistical methodologies must be sound to yield reliable results. In this case, the sample size was too small, it was not random, and the margin of error was unreasonably high.

Trial courts must determine a proposed class has predominate common issues, and consider if it is feasible to try the case as a class. A class must be decertified if the need to litigate individual issues makes the class unmanageable.

This case makes it less likely that a class will be certified without evidence that the employer has a uniform policy or practice that violates wage and hour laws.

Particularly related to liability, statistical sampling and surveys cannot replace common proof. Courts must ensure the validity of statistical methodology by relying on the analyses of statistical experts, and class member sampling must be randomly selected and sufficiently large.

This case also highlighted defendants’ guarantee to due process, including the right to offer evidence supporting its defense, even including individual questions.

How will this case impact future cases?

Lower courts in California now have clearer parameters for determining a class. Will Stern, partner at Morrison & Foerster, shared with InsideCounsel why this decision is so important for employers,

“A damages model cannot include unharmed persons in the multiplier. Rather, people without claims have to be subtracted out from class wide damages. This applies in all California class action cases, whether employment or not, and whether statistics are used to prove liability or even simply to prove damages.”

Stern also pointed out that ‘trial by formula’ is acceptable methodology in cases where the variable is quantifiable and objective. For example, one could count the number of devices containing a defect. Testing overtime hours, or the percentage of tasks performed involving professional skills is more subjective. He comments, “…the class action is being used to get money (or assess a defendant) on behalf of people who, outside of a class action, would not have had a claim.”

The California Supreme Court left some important questions unanswered. One thing that is left unresolved is whether statistical evidence and representative testimony can be used to establish liability in a class trial.

 

Relevant Reading

‘The Use of Statistical Sampling Post-Duran‘, Sarah Butler, Law360 and NERA

‘Efficient conversion of hardcopy paper time records and employee files‘ by Dr. Dwight Steward of EmployStats

‘Using statistical evidence to prove liability in class actions‘ by Joshua Fruchter, Esq.