Based on Haruvy, Roth, and Unver. 2004. The Dynamics of Law Clerk Matching: An Experimental and Computational Investigation of Proposals for Reform of the Market.

Each year, Federal appellate judges hire law students as clerks for their courts. There are only a limited number of high-quality law clerks available each year. Judges who move quickly can hire the good students as clerks before their colleagues have a chance to act. However, the earlier in a student's law school career that one hires the student, the less information is available about the student's abilities. Whereas in the past, judges might have hired students when they were in their third year of law school, in the 1999-2000 academic year, many judges hired clerks in the fall of the second year of law school, almost two years before employment would begin and before hardly any information about candidates other than first year grades are available. Hiring dates moved still earlier in 2000 and 2001.

Suppose, you are a Federal appellate judge and you must make a decision about when to hire your law clerk. Suppose that if all judges hire students at the end of their third years everyone has a 50-50 chance of getting a good law clerk, and that the odds of getting a good law clerk are the same if everyone hires students at the end of their second years. However, if you hire your law clerk at the end of the students' second year, you are sure to get a good law clerk. But if hire at the end of the third year and all the other judges are hiring at the end of the second year, you have no chance of hiring a good clerk. The same possibilities apply to the other judges. Hiring at the end of the third year has no cost to you. Hiring at the end of the second year is more costly because you have fewer signals (grades, internships, etc.) to make sure that the clerk you hire really is good and thus must do some extra investigative work. You would be willing to pay $40 to avoid having to do this investigative work. A 50-50 chance of hiring a good law clerk is worth $50 to you and a 100 percent chance of hiring a good law clerk is worth $100. Other judges value these outcomes just as you do.

Note that this game matrix is similar to the one you saw in Experiment 2. In that experiment, however, the players were just two judges, whereas here we show how one of the players can be modeled as a group of players who all have the same strategies available and the same payoffs. We don't include the payoffs of Others because they face the same payoffs and we want to focus on the decision that each individual judge will make as they consider what the other judges might do.

Every judge has a dominant strategy to hire early, and so this is the equilibrium outcome of the game. But it would be better if waited until the students were in their third year. Professional associations in law have tried to create rules to remedy the Prisoner's Dilemma that judges find themselves trapped in, but so far they have been unsuccessful at changing the incentives (economists have argued that it is the ease with which binding contracts are made in this market that causes the problem - once a contract is signed, neither party can easily renegotiate or renege when new information becomes available).