Overconfidence is a Social Signal, not a Judgment Bias

Abstract

Accurate assessments of one's abilities are important in many domains. These assessments form the basis for the choice of education, of career, of whether or not to start a business, or of the wage asked in negotiations. Evidence from psychology and economics indicates that many individuals overestimate their ability, both absolute (their real ability compared to the estimated ability) and relative (their real ranking compared to estimated ranking). These results are typically taken as evidence that individuals are overconfident in their abilities. Recent modeling efforts in economics have shown that this need not be irrational, but can be the outcome of rational Bayesian updating, thus providing a potential explanation for overconfidence. In this paper, we provide three results: First, we show that Bayesian updating imposes restrictions on relative ability judgments that can be tested empirically. Using data on 1015 individuals' confidence judgments about two cognitive tests, we test these restrictions and reject the hypothesis that our data are generated by Bayesian updating and truthful statements. Second, we test whether self-image concerns contribute to overconfidence. Our results suggest that individuals do not care about inferences created by signals of their ability, but rather consume the signals themselves - they like to hear that they are good. Third, we provide evidence that personality characteristics strongly affect confidence judgments: for example, subjects scoring high in the trait social dominance have more confident judgments about their ability, but they do not have in reality a better performance. Our evidence suggests that overconfidence in statements is bias induced by social concerns.