Mark Twain famously said that “there are lies, damn lies, and statistics.” A recent decision by the California Supreme Court provides a good example of why. In Duran v. U.S. Bank, N.A., the Court put the kibosh on a trial court’s decision — in a wage-and-hour class action — to impose a “trial plan” which accepted evidence based only on a small sample of the class members. The Court found that the trial court’s sampling plan created huge statistical biases in plaintiff’s favor, while violating the defendant’s due process rights. We’ve previously discussed Duran. What employers should realize, however, is that Duran’s reasoning should extend far beyond California state law, and far beyond just the trial management stage. With plaintiffs across the country hungry like a wolf to short-circuit employers’ due process rights, the California Supreme Court’s decision provides a model that employers can use to fight back certification efforts for FLSA collective actions and state wage-and-hour class actions.
Duran’s reasoning does not depend upon statutory language or precedent specific to California. Instead, Duran depends upon well-accepted statistical precepts, taught nationwide in our colleges and universities. These statistical precepts, such as “selection bias” and the “margin of error” apply to all contexts in which sampling is used. Similarly, by giving defendants the opportunity to impeach a sampling plan, Duran recognized that any sampling plan is just an imperfect model of the real world; it is not religious dogma that must be accepted on faith. This understanding applies to all sampling models, regardless of whether the context is trial management or something else. Any time a court uses sampling without employing Duran’s safeguards, it engages in bad science.
This is important, because sampling “is a procedure often used in FLSA actions.” LaFleur v. Dollar Tree Stores, Inc. (E.D. Va. Mar. 7, 2014). And, while courts have used sampling to streamline FLSA trials, they have also employed sampling at the certification stage, and even to limit a defendant’s discovery rights to a fraction of FLSA opt-in plaintiffs. See Roussell v. Brinker Int’l, Inc. (5th Cir. 2011) (trial management); Indergit v. Rite Aid Corp. (S.D.N.Y. June 17, 2014) (certification); Smith v. Lowe’s Home Centers, Inc. (S.D. Ohio 2006) (discovery). Even worse, courts have frequently made the same statistical blunders that Duran rebuked. In Roussell and Indergit, for instance, the courts permitted the parties to cherry-pick witnesses to form the representative “sample.” Yet, as Duran showed, such a procedure necessarily creates a biased “sample” of the class, a statistical flaw known as a “selection bias.” Similarly, in Reich v. S. New England Telecommunications Corp. (2d Cir. 1997), the Second Circuit sustained a verdict predicted on testimony from just 2.5% out of 1500 opt-in plaintiffs, finding that a trial needs to contain only “very small sample of representative evidence” from collective members. Yet Reich contained no analysis about the margin of error such a small sample would create (it’s 15.49%, at the standard 95% confidence interval), even assuming the sample was construed in an unbiased, truly representative fashion.
Notably, none of these decisions predicate their reasoning on statutory or regulatory language. The haphazard sampling procedures they endorse are judge-made. And what judges make, judges can easily correct or distinguish. In comes Duran, to explain just why these judge-made rules should be revised. Indeed, in a world where even expert opinions must conform with solid science under Daubert, it is disappointing that judges have too frequently conjured up trial and discovery plans which limit factual testimony to rigged, biased, or sloppily-constructed unscientific samples.
In short, because good sampling science forms the root of Duran, Duran’s reasoning necessarily applies to any kind of sampling plan — whether proposed at the discovery, final certification, liability, or damages phase — in any court, under any wage-and-hour statute. For this reason, employers should study Duran carefully when hit with a wage-and-hour class or collective action, and use Duran’s persuasive logic to make shoddy sampling proposals come undone.