AN EXPERIMENTAL-STUDY OF CONSTANT-SUM CENTIPEDE GAMES

In this paper, we report the results of a series of experiments on a v ersion of the centipede game in which the total payoff to the two play ers is constant. Standard backward induction arguments lead to a uniqu e Nash equilibrium outcome prediction, which is the same as the predic tion made by theories of ''fair'' or ''focal'' outcomes. We find that subjects frequently fail to select the unique Nash outcome prediction. While this behavior was also observed in McKelvey and Palfrey (1992) in the ''growing pie'' version of the game they studied, the Nash outc ome was not ''fair'', and there was the possibility of Pareto improvem ent by deviating from Nash play. Their findings could therefore be exp lained by small amounts of altruistic behavior. There are no Pareto im provements available in the constant-sum games we examine. Hence, expl anations based on altruism cannot account for these new data. We exami ne and compare two classes of models to explain these data. The first class consists of non-equilibrium modifications of the standard ''Alwa ys Take'' model. The other class we investigate, the Quantal Response Equilibrium model, describes an equilibrium in which subjects make mis takes in implementing their best replies and assume other players do s o as well. One specification of this model fits the experimental data best, among the models we test, and is able to account for all the mai n features we observe in the data.