« Debating the Infield Fly Rule in Penn Law Review | Main | Up with Cuba »

Thursday, March 24, 2016

Mixed Feelings about Tyson Foods

On Tuesday, the Court decided Tyson Foods v. Bouaphakeo, which asked the Court to decide whether and when plaintiffs can use statistical samples (what Scalia disparagingly called “Trial by Formula”) to prove class liability.  The case was the culmination of a long fight against statistical sampling by class defendants, whose dream has always been to take this sort of evidence offline.

I joined Allan Erbsen’s excellent amicus brief on neither side—which was a pitch at the then-conservative majority to (1) abjure broader rulings governing the use of statistical proof  by (2) vacating the judgment on narrow, case-specific grounds.

That seemed like a pretty good strategy before Justice Scalia’s demise, if you assumed the conservative majority was primed to reverse. But it’s a different Court now, and it affirmed the judgment in a five-justice majority opinion written by Justice Kennedy, with a concurrence by the Chief Justice (joined in part by Justice Alito, who also joined Justice Thomas’s dissent.).

The result, at least on an initial reading, is relief for those of us who worried the Court might drastically curtail opportunities to use statistical sampling to prove class claims. But on the general law of class certification, the majority opinion strikes me as a departure from past precedent, and not necessarily in a good way.

More below the line.

The case involved a class action/FLSA collective action on behalf of 3,300+ employees of a Tyson Foods processing plant, who alleged Tyson Foods failed to pay them for time spent putting on and taking off (“donning and doffing”) job gear, leading to loss of overtime compensation due under the Fair Labor Standards Act and a state statute. The class members needed to prove they had worked over 40 hours a week—a precondition for recovering overtime pay. To prove this, plaintiffs turned to an expert time study by Dr. Kenneth Mericle, who calculated average donning/doffing times based on randomish samples; a second expert then used these averages to estimate that all but 212 members of the class had worked over 40 hours/week inclusive of donning and doffing.

Yet the sampling blurred some significant variations in the time spent on donning and doffing across different job types… variations that if adequately accounted for could reveal that a significant number of class members (according to some estimates, up to half the class) had never met the 40 hour a week threshold.

One of the late, great Richard Nagareda’s many accomplishments was to “demarcate the proper line between class certification and summary judgment.” Class certification, he argued, asks whether class claims pose liability questions that are resolvable as a unit—e.g. susceptible to common answers. And he embraced the rigorous scrutiny approach to certification ultimately adopted in Wal-Mart: When plaintiffs seek to demonstrate that issues can be resolved based on common evidence, courts should rigorously assess whether the evidence proves what it claims (by a preponderance of the evidence) if insufficiency of that proof would leave the defendant’s liability to class members dependent on dissimilarities across the class.

By contrast, if an issue raised by the class can only be proven by some form of class-wide evidence, because it  turns on some common set of underlying events (e.g., “the concern is not that the class exhibits some fatal dissimilarity, but, rather, a fatal similarity”), then this rigorous threshold inquiry is inappropriate. In that case, the question ought to be the usual one applicable to any evidentiary submission: whether the evidence is admissible and might persuade a reasonable jury.

Finally and crucially, Nagareda also cautioned that courts must ensure that aggregate proof is consistent with the substantive liability standard.   The question at the certification stage thus is not simply whether claims are supported by some common proof—its whether that proof reflects a proper substantive theory of liability.

The respondents argued within this framework. I won’t detail their substantive analysis. You can read it here, starting on page 35 (where they analogize the Mt. Clemens liability scheme to the burden-shifting scheme in securities fraud), through page 45 (where they argue certification is appropriate based, in part, on an analogy to the Court’s recent 10b-5 certification cases).

Justice Thomas, though, rightly complains that Kennedy doesn’t grapple with the substantive liability standard in the serious way the respondents do. To me, some of Kennedy’s core reasoning seems to be that:

  • use of statistical sampling is consistent with the substantive law when it is the only “practicable” means of proving the claim (see slip op. 10; this is not, though, the only situation in which the Court thinks plaintiffs can employ sampling);
  • proving that the defendant caused the injury plaintiffs complain of (loss of overtime) required some kind of a statistical study (see slip op. 12: “there were no alternative means for the employees to establish their hours worked,” thanks to the defendant's bad recordkeeping); as a result, relying on statistical evidence in this case was consistent with the substantive law;
  • because statistical evidence is the only “practicable” means of proving the claims here, this issue also falls into Nagareda’s second category, where proof of the issue is inherently “common” to the class (e.g. where the concern about the class is “not that it exhibits some fatal dissimilarity but, rather, a fatal similarity”). And so, the court shouldn’t engage in a full bore rigorous analysis of the merits of plaintiff’s expert submissions—it should assess the expert evidence under Daubert (waived here) and summary judgment standards. See slip op. 14-15.

Maybe this interpretation is uncharitable.   If I’m reading the opinion right, though, this is a pretty striking example of upholding class certification upon the “mere raising of common questions by way of expert submissions” (Nagareda) without first asking if those submissions demonstrate the plaintiffs are seeking to resolve the class claims consistently with the substantive law. After all, its not enough to conclude statistical evidence in the abstract would comport with substantive law—the district court has to find that this statistical study did so, because its methodology took account of variables the substantive law cares about. (And for the record, Dr. Mericle noted that the time study “could [be] repeat[ed] using a different methodology,” see Civ Pro Scholars Br. at 19).

Justice Thomas, by contrast, rightly starts with the substantive liability standard, which he argues requires proof the defendant caused each class member an individual injury. That meant that plaintiffs’ time study had to account for significant donning and doffing differences across distinct job types, something the study’s sampling method didn’t do. In our brief, we argued the Court could simply vacate the judgment and remand, allowing plaintiffs to take another crack at substantively copacetic class trial plan.

It’s probably unwise to make a lot of this opinion—it may be a blip during a transition on the Court. And a future Court may build some of the missing substantive analysis into, say, a heightened Daubert “fit” test in the class cert. context (the Court has yet to clarify Daubert’s application to expert submissions at the class cert. stage). But it’s also possible this may be the beginning of a long, slow drift away from the substanc-ized approach to certification that increasingly characterized the Court’s opinions during the late Scalia era.

Posted by Mark Moller on March 24, 2016 at 05:18 PM | Permalink


Post a comment