Rationally Misplaced Confidence

Abstract

I show that persistent underconfidence and overconfidence arise from rational Bayesian learning when effort and ability are complementary and payoffs are affected by a general class of shocks. Which arises depends on the decision-making environment. If greater effort either reduces or does not affect the variance of outcomes, then agents become underconfident on average, but if greater effort increases that variance by enough, then agents become overconfident on average. The mechanism is that agents learn away overconfidence and underconfidence at asymmetric rates because (i) Bayesian updating requires that their sensitivity to new information depend on their effort choices and (ii) their optimized effort choices depend on beliefs about their own ability. As one implication, management can induce overconfidence in employees by providing feedback that is conditionally vague.

Keywords: learning, effort, confidence, ability, talent, feedback, appraisals

JEL Classification: D83, G41, M54