We devise an experiment to explore principal-agent behavior. A principal can propose one of three contract menus to two agents of unknown type (high or low). All of these contract menus favor the principal, but to varying degrees, and are designed to separate the agents into types. The agents independently choose to either accept one of the two contracts offered in the proposed menu or to reject the contract menu altogether. If either agent rejects the menu offered, all individuals receive a reservation monetary payoff lower than that available by accepting a contract. We vary this reservation payoff in two treatments. The game is played for 15 periods, with random and anonymous re-matching; after each period, the participants learn the choices of all members of their triad. Standard theory predicts that each agent will accept a contract (which one depends on her type) from even the most lopsided menu. However, we observe numerous rejections of the more inequitable menus and approach an equilibrium where one of the less lopsided contract menus is proposed (which depends on the reservation payoffs) and agents accept the menu, selecting actions according to their types. This behavior is consistent with some parameterizations of all recent models of social preferences, suggesting that there is value in including nonpecuniary utility in mechanism design and contracts.

Co-author Antonio Cabrales