Testing the Selection Effect: A New Theoretical Framework with Empirical Tests

Recent law and economics scholarship has produced much theoretical and empirical work on how and why legal disputes are settled and litigated. One of the most significant developments in this literature, attributable to the work of William Baxter and the combined efforts of George Priest and Benjamin Klein, has been the formation of a theory about both the selection of disputes for trial and the rates of success that plaintiffs enjoy for those cases that are resolved at trial. The basic theory contains two components. The selection effect refers to the proposition that the selection of tried cases is not a random sample of the mass of underlying cases. Rather, those cases that tend to be clear for either the plaintiff or the defendant under the applicable legal rules settle relatively quickly, leaving only the difficult cases for trial. Since these tried cases are not representative of the larger class of disputed cases, it is risky to draw any inference from the outcome of tried disputes to the soundness of the legal rules by which these disputes are decided. The sample of tried cases may contain many victories for the plaintiff and many for the defendant. One cannot infer from that fact alone, however, the fairness or desirability of the underlying legal rules. If the rules are heavily weighted for the plaintiff, then the closeness of the tried cases is consistent with there being many cases in which plaintiffs recover handsomely without litigation. Similarly, if the rules are weighted heavily for the defendant, then the total mass of cases brought will, by the time of trial, be reduced, as defendants will be able to settle many cases on favorable terms. The observation of hotly contested trials is therefore wholly consistent with the underlying rules that are skewed toward the plaintiff, toward the defendant, or toward neither side. The power of the selection effect is generally recognized in the academic literature. It has been the basis of complex litigation models and has been subjected to empirical testing and debate. Closely akin to but clearly distinguishable from, the selection effect, is the so-called 50 percent hypothesis. This hypothesis is a more specific prediction than the selection effect. The 50 percent hypothesis posits that the set of tried cases culled from the mass of underlying disputes will result in 50 percent victories for plaintiffs and 50 percent victories for defendants. The 50 percent hypothesis can conveniently be regarded as the limiting case of the selection effect theory. It follows, therefore, that any empirical corroboration of it generates powerful support for the more general selection hypothesis. This article suggests the incompleteness of existing methods of statistically testing the 50 percent hypothesis and reformulates the criteria for accepting or rejecting the hypothesis