Poor Judicial Performance: When should the Parties Care?

Abstract

The judicial system is imperfect. Were that not so, there would be no need for the burden of proof or the standard of proof. In the language of mechanism design, the imperfection results from the fact that information about the underlying case is not verifiable. In the language of procedural law, there is no admissible to prove an alleged fact to be true. The court decides against the party whose burden it is to prove this fact. In economic theory, verifiability is dichotomous: either a piece of private information held by the agent can be observed by the competent third party, or it cannot. Procedural law is more fine-grained. The standard of proof ranges from preponderance of the evidence up to beyond reasonable doubt. In economic terms this range of standards reflects a normative judgement on the relative weight of type I versus type II errors. In private lawsuits, the legal order in principle is indifferent between the plaintiff and the defendant. In criminal law, every reasonable effort should be exerted to prevent an innocent person from being convicted. The paper by Ben Hermalin can be read as formalising the standard of proof, at least when the underlying case is about contract (Hermalin [2008]). It tells the parties under which conditions they should care about the resulting possibility of materially wrong court decisions, and when this is irrelevant for their interaction. The logic of the argument is easiest to see in an example. It is given in Figure 1. For expository reasons, the example is more restrictive than the model. In a second step, these additional assumptions are dropped. There are three agents: the parties, Nature, and the court. In the initial contract, from an action space Λ, running from ax to αΊ, the parties stipulate the desired action to be taken by the agent, say a3. To every possible action in Λ, Nature attaches a set of possible signals. This set is defined in two dimensions. The total set of signals, A, is a copy of the action space, running from a' to αη. But some actions may only generate some signals. In the example, if the agent behaves as stipulated, only signals in A3 = {αι,α2, <*3,α4} occur. Moreover, each signal has