Academic journal article Economic Inquiry

A Simple Principal-Agent Experiment for the Classroom

Academic journal article Economic Inquiry

A Simple Principal-Agent Experiment for the Classroom

Article excerpt

Interest in the use of classroom experiments has increased dramatically in the past decade.(1) As we were surveying available demonstrations for a booklet, Ortmann and Colander [1995], we found that there were a number of classroom experiments readily available to demonstrate pricing institutions such as auctions, and symmetric game problems of the prisoner's dilemma variety which are relevant to public goods/externalities/oligopoly/cartel situations. However, there is a relative dearth of classroom experiments illustrating asymmetric game problems.

The most well-known of these asymmetric game problems are moral hazard, principal-agent games - games in which one of the players (the agent) is informed about a key aspect of the game while the other (the principal) is not. We found the lack of classroom experiments involving such games surprising and unsatisfying because they have become a prominent staple in many textbooks as in Stiglitz [1993]; Colander [1995]; Carlton and Perloff [1995]; and Mishkin [1995].

This note addresses the imbalance. We describe a simple, flexible and instructive moral hazard experiment for use in a variety of classes, including introductory courses. The experiment can be used to illustrate the importance of information, and the power of reputational enforcement in principal-agent interactions. Such issues are the key for an understanding of modern theories of the firm, such as Holmstrom and Tirole [1989], Kreps [1990a; 1990b], Stiglitz [1993], Carlton and Perloff [1995], and Mishkin [1995], and the role of market forces in assuring contractual performance as in Klein and Leffler [1981]. We begin by discussing prisoner's dilemma and principal-agent games and their relevance to economics. Then we describe the experiment and its likely results.


Prisoner's dilemma games are symmetric game problems. They are of interest because they capture situations in which the collectively optimal decision will not (necessarily) be achieved through individual optimization as in Axelrod [1984]. Players' sets of conflicting choices are interchangeable; individuals face identical dilemmas. Specifically, both prisoners are presented with the option to confess, or not to confess. Prisoners face the dilemma that their self-interest suggests that they ought to confess; yet, if they both follow their self-interested individual optimization they will be worse off than if they both cooperated. Cartel and free-rider problems fall within this type of symmetric game problem.

Moral hazard, principal-agent interactions are asymmetric game problems. They can be conceptualized as asymmetric prisoner's dilemma games as in Rasmusen [1989]. The asymmetry results from the different set of choices that the principal and the agent face. For example, consider a buyer (a principal) who takes his stereo for repairs. The repair person (the agent) diagnoses the source of the problem and promises to get a high quality part to fix it. The buyer has to decide if he wants to trust the repair person, both as regards her diagnosis and the promise to use high quality parts. If he decides not to trust her, he can get both a second opinion beforehand and have the parts checked afterwards. Both activities would require additional time and cost and may therefore not be desirable. The dilemma is obvious: the agent faces the temptation to renege on a promise that would make agent and principal together better off. (This is the moral hazard aspect of the principal-agent game presently discussed.) The principal, knowing this, is confronted with the dilemma of trusting or not trusting the agent. What makes this asymmetric "dilemma of trust" interesting and relevant to many decisions is that, similar to the symmetric prisoner's dilemma game, the agents' individual and collective rankings of outcomes differ.

In the simplest version of this asymmetric game the two players can choose between two actions each so that there are four possible outcomes. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.