A Pure Test of Backward Induction

This dissertation proposes a simple computerized game to serve as a pure test of backward induction and then tests the game in the laboratory. One of the fundamental assumptions of neoclassical economic theory is that human beings function as fully rational agents who maximize their utility over multidimensional alternatives under economic constraints. However, numerous studies have shown systematic deviation from rational decision making in a laboratory setting. While no single explanation is obvious for this suboptimal behavior, the literature suggests other motivations (besides maximizing utility) may be at play, including reciprocity, trust, reputation, and welfare. The "Race to 21" game we test renders these other-regarding preferences irrelevant; therefore we call it a "pure" test of backward induction.

Chapter one introduces the game, as well as tests the effect of adding incentive payments in several places along the path of play. Chapter two continues by analyzing how each different intermediate incentive affects the speed of learning in the game. Chapter three concludes with a look at whether individual differences among laboratory subjects explain some of our experimental results. Common to all chapters is the result that incentive payments offered on the subgame perfect equilibrium path near the midpoint of the game particularly enhance the use of backward induction among subjects.