Friday, June 6, 2014

Lies, Damn Lies, and Statistics

Duran v. U.S. Bank Nat’l Assoc., No. S200923 (Cal. May 29, 2014)

In this significant case that the wage-and-hour class action bar has been eagerly awaiting, the California Supreme Court reverses a wage-and-hour class action that actually went to trial. That makes it, as the court explains, “an exceedingly rare beast.” But it isn’t just the fact that this case was tried that is unusual. The way it was tried is also outside of the ordinary. The trial court selected twenty-one plaintiffs out of a total class of 260. It then tried those claims and extrapolated their results class-wide. It refused to admit any evidence outside of the sample plaintiffs. This all resulted in a finding that the entire class was erroneously misclassified as exempt, and a $15 million judgment ($57,000 per class member) against the defendant. While the court does not go so far to say that trial by sample can never be appropriate, what happened in this case can’t withstand scrutiny.

Sorry for the long post, but for this to make sense, some context about the law and facts at issue is required. Frankly, the court’s unanimous opinion—by Justice Corrigan—contains lots of interesting discussion of wage and hour law, class action management, and the use of statistical sampling evidence. This post isn’t going to do it the justice it deserves, so if you’re interested in these issues, you should probably read the whole opinion. That said, since it’s sixty-five pages, if you aren’t so inclined, I’ll do my best to hit the highlights.

Plaintiffs contend that their employer misclassified them as exempt from the overtime protections in California’s wage and hour laws. The employer contends, however, that they fall under the outside salesperson exemption in Labor Code § 1171. That exception excludes from overtime protection employees who customarily and regularly spend more than half of their time away from the employer’s office engaged in sales activity. Unlike applicable federal standard, the inquiry is largely quantitative—it turns on how much time the employee actually spends out of the office. And as interpreted by the California Supreme Court in two prior cases—Ramirez v. Yosemite Water Co., 20 Cal. 4th 785 (1999) and Sav-On Drug Stores, Inc. v. Superior Court, 34 Cal. 4th 319 (2004)—the employer bears the burden on the issue, so practically it works like an affirmative defense.

The problem presented, then, is how (or, as defense lawyers would prefer, if) a court can litigate this issue on a classwide basis. On one hand, the employees’ job classifications and duties could be so uniform that they can be resolved collectively. That was the case in Sav-On, upon which the trial court in this case relied in certifying the class. But what about when, like here, there is a great deal of variety as to how much time each employee spent in the office, which is the most relevant consideration on exemption? (A worthwhile digression at this point is to note that the named plaintiffs switched several times during the course of the trial court proceedings because the dropped plaintiffs testified to working 60 percent or more of their time out of the office.)

After several years of discovery (the case was filed in 2001) and dueling proposals by the parties, the trial court came up with its own solution. It would “randomly” pick twenty members of the class, who, along with the two named plaintiffs, would be the “representative witness group”—somewhat annoyingly acronym-ized in the opinion as the “RSG.” When one of these representative witnesses didn’t respond to a subpoena, the trial court treated him as a normal absent class member and proceeded with the twenty-one representative witnesses. It then permitted only the selected witnesses to testify at the liability phase of the trial. Indeed, it repeatedly rebuffed defendant’s efforts to introduce testimony from other class members that would have undermined a finding of liability. Having heard this testimony, the court ruled that the entire class was misclassified, because defendant didn’t carry its burden. 

The trial court then, relying on plaintiff’s expert, computed damages by extrapolating from testimony that plaintiffs worked a weighted average of 11.87 hours of overtime per week. It reached that number with a margin of error of 5.14 hours (43.3 percent), based on a confidence level of 95 percent. Tallying up the damages, the court awarded $8.9 million, which, upon applying significant prejudgment interest, came out to a judgment of just shy of $15 million.

The court of appeal reversed and the supreme court agrees. It is one thing to use statistical extrapolation to compute damages in a class action, but it is another thing to rely on statistics to litigate liability. If, like here, there is an affirmative defense that is insufficiently common to litigate on a classwide basis, the defendant has the right to individually litigate it plaintiff-by-plaintiff. While plaintiff’s proof could rely, in part, on a sufficiently reliable statistical proof—for instance, in establishing that an employer’s policy is different in practice than it purports to be in the employee handbook—liability cannot depend on statistics alone.

And in any event, the statistical method used by the court was far too biased and unreliabile to withstand scrutiny. No proof was offered that a sample of twenty would be sufficiently representative of the whole class. And by including the named plaintiffs—who were specifically chosen because they worked more than 50 percent of their hours in the office—the court introduced a selection bias into the sample. This was compounded when the court excluded one of the representative witnesses because his job duties differed from what the court deemed to be the norm. And then the randomness of the sample was further diminished by nonresponse bias when one of the representative witnesses failed to show and the court simply returned to treating him as an absent member of the class. Notably, the rate of opting out was significantly higher for the representative witnesses than for the class as a whole. 

This was all made even worse when the court categorically refused to hear testimony from class members outside of the sample. That included seventy-five members of the class who submitted declarations attesting that they spent the majority of their time out of the office, as well as the four previously named plaintiffs, who similarly testified at their depositions. Under the circumstances, there was no reasonable way for the court to interpret the testimony of twenty-one witnesses as warranting a finding that the defendant was liable to every member of the class. Even plaintiff’s expert admitted that, given the margin of error, as many as 13 percent of the class may have qualified for the exemption. Indeed, the 43 percent margin of error on the total hours calculation was too large to permit its use even for the calculation of damages.

In coming to its decision, the court also provides some interesting explanation on the manageability of individual issues in class actions. The court notes that even if class issues predominate, class treatment is inappropriate if individual issues of proof aren’t manageable. In cases where individual issues are presented, the trial court should make a reasoned, informed decision about manageability at the certification stage. If the plaintiff proposes to manage individual issues in part through statistical proof, the court should consider the rigor of a trial plan employing statistical proof as part of certifying the class. Moreover, the court should remain cognizant of the individual issues even after a class is certified. If the case becomes unmanageable, it should decertify the class. Here, on remand, the trial court should expressly consider these factors in determining whether the case should proceed to retrial as a class action.


Justice Liu—who is developing a reputation for this sort of thing—fully joins the opinion but adds a thoughtful fifteen page concurrence. He (gently) suggests that the opinion is a little too categorical in reducing the outside sales exemption to the question of did the employee actually work more than 50 percent of the time out of the office. As Justice Liu reads Ramirez and Sav-on, what really matters are the actual requirements of the job—which are based on the employer’s reasonable expectations. How an employee spends her time is certainly relevant to that determination, but it is not necessarily dispositive of the exemption. If, for instance, an employer were to actually require its employees to spend most their time performing various tasks that can only be done inside the office, the employees would be non-exempt on a classwide basis, even if some of them actually spent less than 50 percent of their day there.

Given this standard, Justice Liu notes that representative sampling can be appropriate for management of individual issues. But in order to do so reliably, consideration of individual issues and the variability of the class must inform the design of any sampling plan. A way to address the individual issues must be “baked into” the plan’s design. Further, the defendant must be permitted to use evidence of individual issues to attack the reliability and representativeness of the plan’s results. 

As Justice Liu explains:

The important point is that neither an aggregate method of proof (like sampling or representative witness testimony) nor individualized evidence (like a declaration) is necessarily dispositive when the ultimate issue at trial is to determine “the employer‘s realistic expectations” or “the realistic requirements of the job.” The two types of evidence must be considered and weighed alongside each other, and more broadly, they must be considered and weighed together with the full range of evidence bearing on the ultimate issue, including the employer‘s job description, company policies, industry customs, and testimony of supervisors or managers who monitored, evaluated, or otherwise set expectations for employees in the class. 
(citation omitted)

Justice Liu agrees with the opinion of the court that the unreliability of the representative sample procedure used by the trial court and its refusal to consider individualized evidence warrant reversal. But he notes that the “trial court focused on the right question on the merits: What were the realistic requirements of the . . . position?” So if on remand, the trial court were to properly address both the common and individual issues in its trial plan, its conclusion that the whole class was non-exempt could potentially be sustained, were it based on substantial evidence.

No comments:

Post a Comment