Distinguishing the Effect of Overconfidence from Rational Best-Response on Information Aggregation

Abstract

This article studies the causal effect of individuals' overconfidence and bounded rationality on information aggregation by using a new multiperiod game in which agents are rewarded for submitting accurate estimates of an unknown asset's value based on (i) their private information and (ii) others' past estimates. By carrying out laboratory sessions of this game in which subjects' overconfidence is a treatment variable, I find that overconfidence affects the information aggregation process by increasing the dispersion of estimates and decreasing the rate of estimates' convergence. However, even when subjects are not overconfident, qualitatively similar deviations from the fully rational model predictions are observed. I show that this can be explained by subjects' strategic response to errors.