A payoff uncertainty explanation of results in experimental centipede games

This paper investigates how well a simple two-sided incomplete information game with a full support prior can be used to explain non-Nash equilibrium outcomes observed in the centipede game experiments. I estimate the variance of the uncertainty about preferences in several Versions of the model, select two models, and compare these models to the estimation results of the altruism model and the quantal response models. It is shown that the two selected models have a better fit than the one-parameter altruism and quantal response models and that the estimated variance can explain all qualitative features of these experimental results. Classification Numbers: C72, C19, C44, C91. (C) 1999 Academic Press.