Implementation in the K-Level Thinking Environment

Abstract

Standard mechanism design theory mostly relies on Nash equilibrium concepts. However studies of experimental games suggest that Nash Equilibria are rarely played and provide evidence that subjects may be thinking finite number of iterations. The purpose of this paper is to find out whether the standard expected externality mechanism (Arrow, D'Aspermont, Gerard-Varet) retains its properties under iterational thinking. The optimal strategies of finitely-rational players generally deviate from Bayesian Nash Equilibria, though the latter are often good approximation of the outcomes of iterative thinking.