A common temptation in class action litigation is to fashion procedures based on “rough justice” to avoid overburdening the courts or attempting to redress alleged mass harm. Over the past decade, as storage and computing power have increased exponentially, it has become increasingly tempting to use statistical sampling as a proxy for the actual adjudication of facts in class or mass actions. The idea is that if the facts regarding a statistically significant subset of a class can be evaluated for a particular issue or set of issues, then the results of the evaluation of the sample can be extrapolated across the rest of the class.
One jurisdiction in particular where this approach has gained traction has been California. There, the use of statistical sampling has been recognized for several years as a means of apportioning damages in some cases. See Bell v. Farmers Ins. Exchange (2004) 115 Cal.App.4th 715 [9 Cal.Rptr.3d 544] (Bell III). However, in recent years, plaintiffs have attempted to use statistical sampling as proof of liability, not simply as a means of apportioning damages when liability has been established or (as in Bell III) it is not contested. This approach was harshly criticized in Part III of Justice Scalia’s majority opinion in Wal-Mart v. Dukes, (notably, this was the portion of the Dukes opinion with which all nine justices concurred):
The Court of Appeals believed that it was possible to replace such proceedings with Trial by Formula. A sample set of the class members would be selected, as to whom liability for sex discrimination and the backpay owing as a result would be determined in depositions supervised by a master. The percentage of claims determined to be valid would then be applied to the entire remaining class, and the number of (presumptively) valid claims thus derived would be multiplied by the average backpay award in the sample set to arrive at the entire class recovery— without further individualized proceedings. [internal citation omitted]. We disapprove that novel project.
Earlier this year, in Duran v. U.S. Bank National Association, No. A125557 & A126827 (Cal. App., Feb. 6, 2012), a division of the California Court of Appeal agreed with the above-quoted dicta in Dukes and rejected an attempt to use statistical sampling to prove liability an a wage and hour class action. The plaintiff had presented testimony from statistician Richard Drogin, who had also served as an expert for the plaintiffs in Dukes. Drogin presented a random sampling analysis that purported to estimate the percentage of the defendant’s employees that had been misclassified for purposes of entitlement to overtime pay. The trial court adopted a sampling approach that was modeled on (but not exactly the same as) Drogin’s proposal.
The Court of Appeal held that the trial court’s approach was improper and that it violated defendant’s due process rights for a variety of reasons, including that 1) the use of statistics to estimate the total number of employees who had been misclassified deprived the defendant an opportunity to present relevant evidence and individualized defenses as to individual plaintiffs’ alleged misclassification; 2) the court’s statistical methodology was flawed because it arbitrarily used a sample of 20 employees without any basis for concluding that the sample was statistically significant; 3) even the use of sampling as to damages was improper because the methodology used had an unacceptably high margin of error.
The Duran opinion is worthy of careful study for anyone considering the use of statistics in class certification proceedings, both in the employment context and in other types of class actions. The opinion examines many of the due process problems with allowing proof of liability through statistical sampling, the most significant of which is that it tends to deprive a defendant of presenting evidence in its defense that it would be able to present in an individual case. It also provides an additional illustration of what the Supreme Court considered an improper “trial by formula” in Dukes.