Co-authored by Andrew McNaught and Andrew Paley
On February 6, 2012 the California Court of Appeal, First District, issued its opinion in Duran, et al. v. U.S. Bank. Readers may recall our posting following the oral argument in this matter. In a matter of first impression, the Court of Appeal considered whether class action plaintiffs may use statistical sampling and representative evidence to establish liability on a class-wide basis. In an extremely thorough and detailed opinion, the court answered that question “no” in this case, and soundly rejected the trial plan implemented by the trial court. In holding that it was unconstitutional because it deprived U.S. Bank of its due process rights, the court stated that “a trial in which one side is almost completely prevented from making its case does not comport with standards of due process.” The court also decertified the class based on the trial court’s erroneous assumption that a finding of liability could be determined by extrapolating findings based on the trial sample of approximately 20 plaintiffs to the entire class of 260.
Plaintiffs filed the case in Alameda County Superior Court , alleging that U.S. Bank’s Business Banking Officers (“BBOs”) were misclassified as exempt employees. After class certification, Judge Robert Freedman granted Plaintiffs’ motion for summary adjudication on the Bank’s defenses of administrative exemption and commission sales exemption. The case then went to a bench trial on the Bank’s remaining defense under the outside sales exemption.
Over the repeated objections of the Bank, the trial court conducted the liability phase of the trial based on a purportedly random sample of 20 class members out of 260 total. Inexplicably, the trial court allowed the individuals selected as part of trial sample to opt out of the sample. Four individuals elected not to appear at trial and alternates were selected. U.S. Bank attempted to proffer evidence from 70 BBOs who signed declarations confirming that they spent more than 50% of their time outside of the office (and therefore were properly classified as exempt employees). The court, however, prohibited the employer from presenting any evidence from class members (or BBOs who opted out) other than those selected to be in the trial sample. The court determined that the Bank had misclassified 19 of the 20 class members in the sample, and then extrapolated from that result that all 260 class members had been misclassified. The court then conducted a damages phase in which it adopted plaintiffs’ expert’s view that based on statistical extrapolation at a 95% confidence level, the average class members worked 11.87 hours of overtime per week, with a margin of error of 43% — in other words, the actual average overtime worked based upon the sample could fall anywhere in the range of 6.7 hours to almost 17 hours per week. The trial court entered judgment against the Bank in the amount of approximately $15 million.
The Court of Appeal reversed the judgment because it found that the trial plan implemented by Judge Freedman did not reflect a statistically significant and reliable methodology. Although it stopped just short of issuing a bright line rule, the Court came tantalizingly close to holding that statistical extrapolation cannot be used to determine collective liability in a wage/hour class action. The decision also recites a litany of reversible errors committed by the trial court that deprived the Bank of fundamental due process. It noted that the trial court conceived of the trial plan entirely on its own, without relying on the advice of expert witnesses, and that “the trial court’s use of this sampling procedure to determine both liability and monetary damages appears to be entirely unprecedented.” The court pointed out that the trial sample was not, in fact, random because a comparatively high number of trial sample plaintiffs opted out of the class before trial, and also because Judge Freedman allowed evidence from the two named plaintiffs (not randomly chosen) to be extrapolated to the entire class.
The Panel further faulted the plan because the trial court refused to permit U.S. Bank to put on any evidence outside of the trial sample, including that related to many class members who testified both in declarations and depositions that they were properly classified. This evidence, has it been allowed and found persuasive, would have established that at least one-third of the class was properly classified. The court was very troubled by the fact that the judgment awarded an average of over $50,000 to each of the 239 absent class members, while the Bank was precluded from putting on evidence that may have prevented at least one-third of them from any recovery, simply because Judge Freedman felt that such evidence was “irrelevant” because it did not comport with the trial plan. As the Court explained, “fundamentally, the issue here is not just that USB was prevented from defending each individual claim but also that USB was unfairly restricted in presenting its defense to class-wide liability.” In reaching this conclusion, the Court cited the U.S. Supreme Court’s unanimous guidance in Wal-Mart Stores v. Dukes, disapproving “trial by formula.” The Court noted that “the same type of ‘trial by formula’ that the U.S. Supreme Court disapproved of in Wal-Mart is essentially what occurred in this case.” In a passage that is sure to warm the hearts of defendants, the Court explained that “we have never advocated that the expediency afforded by class action litigation should take precedence over a defendant’s right to substantive and procedural due process.”
In the Phase II damages trial, the trial court compounded its statistical errors in constructing the trial sample utilized in Phase I. It failed to follow acceptable statistical principles in Phase I, and then utilized those improper liability findings as the basis for restitution calculations in Phase II – in other words, garbage in, garbage out. The improper sampling methodology in Phase II resulted in a 43.3% margin of error in determining the Bank’s more than $14 million aggregate liability on restitution, which the court held was a separate due process violation. While Bell v. Farmers Insurance Exchange held that statistical sampling can be utilized to prove damages in certain instances, the Panel here clearly distinguished Bell. It stated that the case was “manifestly inapposite,” and noted that the only similarity to Bell could be found in the portion of that decision striking down the double-time award because it was based on flawed statistical modeling. The Court of Appeal explicitly distinguished Bell on the additional grounds that the sampling methodology used in that case was agreed to by the parties, and that the sampling was used only to determine collective damages, not liability.
The Duran decision is unquestionably welcome news for employers defending class actions in California. No California appellate court has ever held that statistical sampling may be used to prove liability. While this decision stops just short of firmly establishing that proposition, it certainly casts serious doubt on its viability. Any trial court attempting to craft a class action trial plan will have to be very careful to avoid the statistically improper methodologies employed by this trial court, both as to liability and damages. Another key lesson from Duran is that employers should object clearly and at every possible opportunity to a trial judge imposing schemes that are devoid of statistical support and which may result in due process violations, or attempting to persuade the parties to agree to such schemes. There is also excellent language and analysis in the portion of the decision decertifying the class which employers will want to utilize on decertification motions.