The Holdout Game: An Experimental Study of an Infinitely Repeated Game with Two-Sided Incomplete Information

Abstract

We investigate experimentally a two-person infinitely repeated game of incomplete information. In the stage game, each player chooses to give in or hold out. Players have privately known costs of giving in and each player receives a fixed benefit whenever at least one player gives in. High cost players have a dominant strategy in the stage game to hold out, and the low cost players ' best response depends on what the opponent does. Equilibrium play to the infinitely repeated game conveys information about the players' type. We investigate two questions: whether there is any evidence that subject behavior approximates belief stationary equilibria, and whether there is evidence that subjects will converge to an equilibrium of the correct state. We conclude that subjects do not adopt symmetric belief stationary strategies for the holdout game. However, we cannot reject the hypotheses that subjects converge towards eventually playing an equilibrium of the correct state (even though they do not always learn the correct state). Behavior of experienced subjects is closer to the predictions of symmetric belief-stationary equilibrium.

Files

Additional details