Persistent Overconfidence and Biased Memory: Evidence from Managers

17 May 2018
Versione stampabile

2 PM, Seminar Room, Department of Economics and Management, via Inama 5



This paper provides evidence of persistent overconfidence among managers in a firm, who face a high-powered competitive incentive scheme, and have very detailed feedback as well as substantial experience. We measure overconfidence relative to the benchmark of a Bayesian model of belief formation, which makes predictions based on efficient use of managers’ histories of past outcomes. We can reject that manager predictions are consistent with the model with a high level of confidence. We turn next to investigating a potential underlying mechanism, posited by models of motivated beliefs: biased memory (Benabou and Tirole, 2003). We elicit manager recall of recent past performance, and find that managers who performed poorly tend to be inaccurate, and have errors that are asymmetric, in the direction of recalling better than actual performance. We then show empirically that these biased memories of the past help explain overconfident predictions for the future. This provides some of the first field evidence of the role of biased memory in sustaining persistent overconfidence.

The paper is co - authored with David Huffman and Julia Shvets.