Duran boasts a long and tortured history. Filed in 2001 and involving events dating back to 1997, it featured wage-and-hour claims for overtime pay brought against U.S. Bank by a putative class of business banking officers who alleged that the bank misclassified them as outside salespeople.
The original trial court judge certified a class of 260 bank employees and awarded $15 million. An appellate court reversed the award, and the California Supreme Court affirmed the reversal in a lengthy opinion. See Duran v. U.S. Bank Nat’l Ass’n, 59 Cal. 4th 1 (Cal. 2014). The California high court addressed the “hotly contested” question of “whether statistical sampling can legitimately be used to prove a defendant’s liability to absent class members.” Id. at 39. The court acknowledged the use of statistical sampling in support of both liability and damages but invalidated the plaintiffs’ survey of bank employees as being neither random nor scientific.
The U.S. Supreme Court’s Bouaphakeo Opinion
Duran was remanded to a new trial judge. While on remand, the U.S. Supreme Court issued the Tyson Foods v. Bouaphakeo opinion definitively holding that representative proof from a sample could be used to show the predominance of common questions. 136 S. Ct. at 1040. The Court’s opinion was not, however, a blanket approval of sampling in the class context. Rather, the Court held, the use of such a sample “will depend on the purpose for which it is being introduced and on the underlying cause of action.” Id.
On remand, and perhaps encouraged by these developments, the Duran plaintiffs submitted a newer and bigger statistical sample. The plaintiffs randomly selected 160 class members for a survey that drew 87 responses—a 54 percent response rate. Of those responding, more than 95 percent reported spending less than half their time performing outside sales for the bank. The survey respondents reported an average of 23 hours of weekly overtime. The plaintiffs’ expert reported a margin of error of less than 5 percent.
But the trial court rejected the plaintiffs’ new sample. Citing the bank’s expert’s opinion, the court found that far larger sample sizes were needed to reach such a low error margin. The court also cited selection bias—as constructed, the sample invited a self-selected sample of respondents. Also, the 23 hours of overtime reported in the new sample dwarfed the original sample’s 14 hours of overtime. The court cited this discrepancy as evidence of the new sample’s unreliability and found that individual testimony would be required to prove a predominance of common issues.
The California appellate court affirmed the trial court’s order denying certification. In particular, the appellate court noted the discrepancies between overtime hours reported between the two samples. “In individual actions, the effects of the passage of time upon each witness’s memory can be explored by the fact finder. In a class action, it is impossible to do so. . . .” Duran, 2018 WL 456769, at *12.
Statistical Sampling of RMBSs: An Emphasis on Context
The most recent Duran opinion underlines the fact that although statistical sampling can be used to support class certification, it cannot be based on a flawed model. Other recent sampling cases highlight the point made by the U.S. Supreme Court in Bouaphakeo—that the use of sampling depends on context. Sampling must serve a purpose and it must relate to and support an underlying cause of action.
This emphasis on context is highlighted in a string of RMBS cases. In those cases in which plaintiffs alleged general fraud, statistical sampling of thousands of packaged loans was used successfully to show broader deception. See, e.g., Nat’l Credit Union Admin. Bd. v. Goldman Sachs & Co., No. 11-cv-06521-GW-JEM (C.D. Cal. Feb. 10, 2014); Fed. Hous. Fin. Agency v. JPMorgan Chase & Co., 2012 WL 6000885, at *31 (S.D.N.Y. Dec. 3, 2012).
But in more recent cases involving slightly different RMBS-related claims—those involving loan repurchase agreements—courts have rejected statistical sampling. See, e.g., Homeward Residential, Inc. v. Sand Canyon Corp., 2017 WL 5256760, at *7 (S.D.N.Y. Nov. 13, 2017); BlackRock Allocation Target Shares v. Wells Fargo Bank N.A., 2017 WL 953550, at *1–3 (S.D.N.Y. Mar. 10, 2017). Those courts have held that in the loan repurchase context, where the overarching pooling and servicing agreements reference the specific characteristics of individual loans, statistical sampling does not substitute the need for a loan-by-loan analysis.
The question of whether statistical sampling can be used to support class certification has been decided affirmatively. But as the recent Duran case and RMBS cases show, when presented with statistical sampling, the courts must now address questions of whether the sampling is reliable and whether sampling fits within the context of a particular case.