Games played through agents in the laboratory : a test of Prat & Rustichini's model

From the regulation of sports to lawmaking in parliament, in many situations one group of people ("agents") makes decisions that affect payoffs of others ("principals") who are inactive. As the principals have a stake in the agents' decisions they face an incentive to offer payments in order to sway their decisions. Prat and Rustichini (2003) characterize pure-strategy equilibria of such Games Played Through Agents, in which principals commit to action-contingent transfers to agents. Specifically, they predict the equilibrium outcome in pure strategies to be efficient under some conditions. With field data hard to come by, we test the theory in a series of experimental treatments with human principals and computerized agents. The theory explains the data remarkably well. Subjects predominantly offer payments that implement efficient outcomes. In some treatments offers fall short of equilibrium predictions though. These minor deviations from equilibrium behavior are explored in a quantal response equilibrium framework.