Duran v. U.S. Bank National Assn.
A class-action trial plan may use statistical or representative evidence, but it must still preserve the defendant’s due-process right to litigate its affirmative defenses; a plan that extrapolates liability from a small, unrepresentative sample while barring defense proof as to absent class members cannot stand.
Duran v. U.S. Bank National Assn. (2014) 59 Cal.4th 1
Class Actions
Wage & Hour
Statistical Sampling
Due Process
Affirmed
Published
In brief. After certifying a class of 260 business banking officers who claimed they were misclassified as exempt outside salespersons, the trial court tried liability and damages on the work habits of a “representative witness group” of about 20 class members and refused to let the bank introduce evidence about the rest. The extrapolated overtime estimate carried a relative margin of error of plus or minus 43.3 percent. The Supreme Court held the plan deprived the defendant of its due-process right to litigate its affirmative defense as to absent class members, and that the sampling “here was profoundly flawed.” It affirmed the Court of Appeal’s reversal of the multimillion-dollar judgment. (Duran v. U.S. Bank National Assn. (2014) 59 Cal.4th 1, 8, 13.)
Facts
U.S. Bank employed business banking officers (BBOs) to sell bank products such as loans and lines of credit, and classified them as exempt under the outside-salesperson exemption (Lab. Code, § 1171; Wage Order), which applies to employees who spend more than half their working time away from the employer’s place of business engaged in selling. (59 Cal.4th at pp. 10–14.) The plaintiffs alleged that the BBOs in fact spent most of their time in the office and were therefore misclassified and owed overtime. (Ibid.) Because the exemption is an affirmative defense on which the employer bears the burden, and because it turns on how each employee actually allocated his or her time, the case posed in acute form the question of how individualized exemption issues can be tried on a classwide basis. (Id. at pp. 12–14, 24–27.)
Procedural history
The trial court certified a class of 260 BBOs and adopted an ambitious trial plan. (59 Cal.4th at pp. 14–16.) In a first phase, the court would determine liability and the average amount of overtime by taking testimony from a “representative witness group” (RWG) — a sample of roughly 20 class members plus the named plaintiffs — and then extrapolate the result to the entire class. (Id. at pp. 15–16.) Critically, the court refused to permit the bank to introduce evidence about class members outside the RWG bearing on the exemption. (Id. at pp. 16–18.) The plan produced a judgment for the class with an average recovery exceeding $57,000 per member; the extrapolated overtime estimate carried a relative margin of error of plus or minus 43.3 percent. (Id. at pp. 16–20 & fn. 13.) The Court of Appeal reversed, and the Supreme Court affirmed that reversal. (Id. at pp. 18, 50.)
Issue
May a court adjudicate a wage-and-hour class action by extrapolating liability and damages from a small sample of class members while precluding the defendant from introducing evidence — bearing on its affirmative defense — about class members outside the sample? More generally, what conditions must a class-action trial plan satisfy when it relies on statistical or representative proof? (59 Cal.4th at pp. 8–10.)
Holding
No. Although statistical and representative methods may be used in appropriate class actions, “[a]ny class action trial plan, including those involving statistical methods of proof, must allow the defendant to litigate its affirmative defenses.” (59 Cal.4th at p. 8.) The trial plan here — which barred the bank from presenting evidence about absent class members and rested on a sample that was neither randomly selected nor shown to be representative, with an indefensibly large margin of error — violated due process. “[T]he statistical methods used in this case were profoundly flawed,” and the Court of Appeal’s reversal is affirmed. (Id. at pp. 8, 13.)
Reasoning
1. Manageability cannot come at the expense of a defense. The Court reaffirmed that individual issues may be managed in class litigation — including, in proper circumstances, through statistical or representative evidence — but held that managing those issues cannot be accomplished by extinguishing the defendant’s right to contest liability as to individual class members. (59 Cal.4th at pp. 4, 28–33.) The “litigation of relevant affirmative defenses, even [in] a class” proceeding, is a matter of due process, and a trial plan that forecloses it is invalid. (Id. at p. 6.)
2. The trial plan was profoundly flawed. The plan determined the bank’s liability to all 260 class members from the experience of about 20, then forbade the bank from offering evidence that other class members were properly classified. (59 Cal.4th at pp. 14–18, 24–31.) The sample was not randomly selected, no statistical rationale supported a sample size of 20, and the resulting extrapolation carried a relative margin of error of plus or minus 43.3 percent — far too large to support a classwide judgment. (Id. at pp. 16–20, 31–41.) On these facts, “the statistical methods used in this case were profoundly flawed.” (Id. at p. 13.)
3. Statistical proof is a tool, not a shortcut. The Court was careful not to condemn statistical methods categorically. It “disavow[ed] any sweeping conclusion as to whether or when sampling should be available as a tool for proving liability in a class action” (59 Cal.4th at p. 40), and acknowledged that “representative testimony and sampling may sometimes be appropriate tools for managing individual issues in a class action.” But “these statistical methods cannot so completely undermine a defendant’s right to present relevant evidence.” (Id. at p. 35.) If such methods are used, the Court held, the trial plan must be developed with expert input, the sample must be randomly selected and adequate in size, the margin of error must be kept within acceptable limits, and the defendant must retain a mechanism to present proof of its affirmative defenses. (Id. at pp. 7–10, 33–42.)
4. Liability is not the same as the extent of liability. The Court drew a line the trial plan had erased: decisions about “the fact of liability” may not be “reframed as questions about the extent of liability.” (59 Cal.4th at p. 37.) Whether a given class member was misclassified at all is a liability question the defendant is entitled to contest; statistical extrapolation may, in a proper case, help quantify damages, but it cannot be used to manufacture liability that the defendant was never permitted to dispute. The Court nonetheless added that “[t]his is not to say that an employer’s liability for misclassification may never be decided on a classwide basis.” (Ibid.) Because the plan here did none of what a valid plan requires, the judgment could not stand. (Id. at pp. 42–50.)
5. The concurrence. Justice Liu concurred, agreeing that reversal was required because the RWG “was not selected at random but in a manner biased in plaintiffs’ favor,” the court “used no known statistical rationale in picking a sample size of 20,” and it “tolerated a margin of error at the damages phase that was undoubtedly too large.” (59 Cal.4th at p. 50 (conc. opn. of Liu, J.).) He wrote separately to emphasize that the majority took “an appropriately cautious approach” that did not disfavor statistical methods, which remain legitimate when properly applied, and that trial courts should continue to be “procedurally innovative” in managing class actions. (Ibid.)
Significance
Duran is the Supreme Court’s definitive statement on the limits of “trial by formula” in California class actions. Decided two years before the United States Supreme Court’s related caution in Tyson Foods, Inc. v. Bouaphakeo (2016) 577 U.S. 442, it establishes that statistical sampling is a tool, not a substitute for due process. It is the natural counterweight to Sav-On Drug Stores, Inc. v. Superior Court (2004) 34 Cal.4th 319 and Brinker Restaurant Corp. v. Superior Court (2012) 53 Cal.4th 1004: certification remains generous, but the plan for trying common issues must protect the defendant’s right to defend. Its reasoning now reverberates beyond the class-action setting, framing the manageability and trial-plan debates in large representative PAGA actions left open after Estrada v. Royalty Carpet Mills, Inc. (2024) 15 Cal.5th 263. See the Review’s practice guide on wage-and-hour class actions and PAGA.
Key quotes
“Any class action trial plan, including those involving statistical methods of proof, must allow the defendant to litigate its affirmative defenses.” (Duran, supra, 59 Cal.4th at p. 8.)
“While representative testimony and sampling may sometimes be appropriate tools for managing individual issues in a class action, these statistical methods cannot so completely undermine a defendant’s right to present relevant evidence.” (Id. at p. 35.)
Practice pointer
If your certification or trial plan relies on representative or statistical proof, build it to survive Duran: retain a qualified statistician early, design a genuinely random and adequately sized sample, quantify and cap the margin of error (43.3 percent is the cautionary benchmark for what is too high), and preserve a concrete mechanism for the defendant to raise its affirmative defenses. Never structure a plan that simply forecloses the defense from contesting individual class members — that is the precise flaw that sank the Duran judgment. Keep the line between the fact of liability and the extent of liability sharp in your trial plan. On the defense side, Duran is the first authority to cite against any plaintiff’s plan that would extrapolate liability from an unvalidated sample, and increasingly the first authority to cite on manageability in oversized PAGA actions.
Open questions
Duran held that sampling must be done right but deliberately declined to prescribe how. What sample sizes, randomization protocols, and margins of error suffice — and when individual variation is so great that no representative plan can fairly resolve a class’s claims — remain case-specific questions that Duran leaves to trial courts and their experts, subject to the due-process floor it establishes. The decision’s reach into representative PAGA litigation, where there is no certification gatekeeper but the same trial-management pressures, is the most actively contested frontier after Estrada.