Litigation

Rosenberg & McCloud on Choice-of-Law and Class Actions

By Ted Frank

Via Chamblee Burch and Solum, David Rosenberg and Luke McCloud have written a paper proposing that the "problem" of nationwide federal diversity class actions where there are differences in state law that would normally preclude class certification be resolved by a judicially-created "average law," whereby the rights of class members subject to plaintiff-friendly state laws be abrogated to benefit class members who are subject to state laws that are not as plaintiff-friendly. The paper claims to "refute objections to using the average law approach" without once mentioning the Rules Enabling Act, 28 U.S.C. § 2072(b), and its provision that judicially-created procedural rules "shall not abridge, enlarge or modify any substantive right."

Now, if Rosenberg wants to argue that Erie was wrongly decided and there are sound reasons for application of federal common law in federal diversity cases, I'd be sympathetic to that claim, though Erie is so ingrained in our constitutional DNA that I've found even some of the most ambitious conservative constitutional scholars are loathe to question it. And if Erie is to continue to be applied to individual diversity cases, I fail to see how it does not preclude the Rosenberg proposal.

And Rosenberg's "average law" isn't even "federal common law." Rosenberg proposes "statistical sampling" to get to the right result, but that presupposes that individual state laws are one-dimensional algorithms where one plugs in a set of class members and a defendant's actions and the result spits out "$100" or "$200" or some other damages number—never mind that even in the single-state single-law scenario, there are often substantial factual disputes. The paper doesn't explain the practical realities of how, for example, one averages differing states' conceptions of reliance in consumer fraud cases or different statutes of limitation or different unjust enrichment laws or different scienter requirements. These are frequently binary variables not conducive to analog averaging. How can the "average law" be resolved on a summary judgment motion, much less at a manageable jury trial? How does one construct a jury instruction for 72% reliance?

If the answer is "No, no, we're proposing to hold several dozen individual trials, and then aggregate the results across the class," that holds its own problems. In a class of millions, one would need hundreds of trials to avoid the small-sample problem; even if one is satisfied with a "sample of 50 claims" (p. 23), why is statistical-sampling-of-dozens-of-claims-in-a-single-class superior to refusing to certify a class unless it is broken up into multiple subclasses?

Full post as published by Point of Law on August 16, 2010 (boomark / email).