Is the Selection of Cases for Trial Biased?

Abstract

ARE there systematic differences between those cases that end in a trial and those cases that result in a negotiated settlement?' In a recent paper Priest and Klein claim that trials are a biased sample from the population of cases, all of which end in either a trial or settlement.2 In a cleverly constructed model, they argue that those cases that tend to go to trial are those in which the probability of a defendant verdict is 50 percent. Thus trials are not a random sample from the full range of possible probabilities but rather a selected sample from those cases in which each litigant has a fifty-fifty chance of winning. Consequently they assert that the likelihood of success in a trial is invariant to the standard of the decision employed by the judge or jury. This result creates serious problems for those who want to do empirical work on trials. For example, a researcher might not be able to observe biases of particular judges in favor of the plaintiffs or the effect of differing standards of proof on the likelihood of a defendant's verdict because, whatever the bias or standards, the success rate of plaintiffs will hold constant at 50 percent. Furthermore, if trials are a biased sample of those cases that are either settled or tried, then studying jury trials will give a distorted view of negotiated settlements. This is an especially serious problem because relatively few cases go to trial, while information on the outcome of a negotiated settlement typically is not available to the public.