What Duran really said about statistical samples

When the state Supreme Court issued its long-awaited decision in Duran v. U.S. Bank, 59 Cal. 4th 1 (2014), earlier this year, for most of the legal community, the takeaway was that the court had significantly limited the use of statistical sampling evidence in class action cases. On a close reading, however, Duran turns out to be not especially path-breaking, neither really curbing nor expanding the use of statistical evidence.

The high profile question in Duran was whether statistical sampling could be used to prove liability, as well as damages. The court essentially declined to rule on this issue. Prior California decisions - notably Bell v. Farmers Insurance Exchange, 115 Cal. App. 4th 715, 720-21 (2004) - had approved the use of sampling to determine aggregate damages after liability had been established. In Brinker Restaurant Corp. v. Superior Court, 53 Cal. 4th 1004 (2012), Justice Kathryn Werdegar, in a concurrence, encouraged the use of "statistical analysis" as a tool "to render manageable determinations of the extent of liability," a phrase seemingly halfway between liability and damages. But no reported California state court case had yet taken that step. In the end, the Duran court did not prohibit or endorse using statistical sampling to prove liability or provide guidance as to when it would be appropriate: "We need not reach a sweeping conclusion as to whether or when statistical sampling should be available as a tool for proving liability in a class action."

Nor did the Duran court have anything particularly new to say about the characteristics of a valid sampling methodology. The guidelines identified by the court consisted of common principles of statistical modeling: (1) the sampling plan should be developed with input from experts (i.e., statisticians); (2) the sample group must be sufficiently large; (3) the sample must be randomly selected; and (4) the estimate derived from sampling must have a low margin of error. These established principles - all of them apparently violated by the trial judge in Duran - were also articulated in Bell.

The express focus of Duran turned out to be the right of a defendant to assert affirmative defenses, notwithstanding the use of statistical sampling evidence. The court held that, "if liability is to be established on a classwide basis, defendants must have an opportunity to present proof of their affirmative defenses within whatever method the court and the parties fashion to try these issues." In broad strokes, this is not controversial or new. The Duran court in fact noted that in Sav-On Drug Stores Inc. v. Superior Court, 34 Cal. 4th 319, 329-30 (2004), it had encouraged the use of statistical evidence in class actions, but also said that defendant is entitled to present evidence of "wide variations" in class members' circumstances. Bell said a "trial management plan would raise due process issues if it served to restrict [defendant's] right to present evidence against the claims." In Brinker, Werdegar acknowledged that individual issues arising from an affirmative defense might preclude class certification, despite implementation of case management methods like statistical sampling. She said that determination should be left to the discretion of the trial judge.

The Duran court also said much the same thing but offered little guidance as to how that discretion should be exercised. To the contrary, the Duran opinion is notable for stating both sides of the question in a way the leaves the issue up in the air. For example, the court said that, "[n]o case, to our knowledge, holds that a defendant has due process right to litigate an affirmative defense as to each individual class member." Shortly thereafter, the court stated that "[i]f a defense depends upon questions individual to each class member, the statistical model must be designed to accommodate these case-specific deviations." The court goes on to observe, "[i]f statistical methods are ultimately incompatible with the nature of plaintiffs' claims or the defendant's defenses, resort to statistical proof may not be appropriate." As one commentator complained, "[e]very declaration of guiding legal principle is accompanied by a countervailing qualification," and the opinion provides little guidance on when individualized issues overwhelm common ones. 20 No. 3 Westlaw Journal Bank & Lender Liability 8 (June 30, 2014) (citation omitted).

Interestingly, the court appeared to have before it a prime example of the sort of case where statistical sampling may not appropriate. The class as certified in Duran was small, consisting of 260 members, so there was not as much necessity for sampling or efficiency gained thereby as in a case involving thousands of claims. The defendant filed over 75 declarations from class members stating that they were subject to an affirmative defense. That defense - the outside salesperson exemption from overtime compensation law - implicated proof that was exceedingly individualized, i.e., whether an individual employee subject to minimal supervision worked 50 percent or more of the time outside the office. Circumstances like these call into question the utility of statistical sampling, as well as its accuracy versus individual adjudication.

The Duran court also failed to acknowledge that statistical sampling is not by nature antithetical to affirmative defenses, even those defenses that turn on individual questions. Affirmative defenses will be raised and determined in adjudication of the sample cases. If the sample is large enough, the extrapolation of the sample result will accurately reflect the effect of affirmative defenses on the defendant's overall liability. Judge Jed Rakoff in the Southern District of New York made a similar point in Assured Guaranty Municipal Corp. v. Flagstar Bank, FSB, 920 F. Supp. 2d 475, 512 (S.D.N.Y 2013): "The very purpose of creating a representative sample of sufficient size is so that, despite the unique characteristics of the individual members populating the underlying pool, the sample is nonetheless reflective of the proportion of the individual members in the entire pool exhibiting any given characteristic." As the court in Bell held, a party "may object to statistical sampling on due process grounds only to the extent that the procedure affected its overall liability for damages."

In Duran, the court had the opportunity and was expected to clarify and provide guidance on the use of statistical sampling evidence, a critical tool in cases where the claims and/or claimants are too numerous for individual adjudication. It is fair to say that the Duran court not only failed in this missionm but introduced even more lack of clarity on the subject. As before, it will fall primarily to the trials courts to develop the rules on how and when to use statistical sampling.