How to Counteract Biased Self-Assessments? An Experimental Investigation of Reactions to Social Information

Abstract

In a lab experiment, we investigate whether social information can improve the accuracy of self-assessments of relative performance. In particular, we compare the effectiveness of di erent types of social information: subjects either learn their close peers' (i) average absolute performance, (ii) average self-assessment or (iii) average bias of self-assessments. Additionally, we explore the demand for the different types of social information. Our results suggest that social information can help debiasing subjects' self-assessments, but not all types of information are equally effective. Only learning about the average bias of peers improves own self-assessments. Subjects are, in general, willing to pay for social information but mostly prefer information about their peers' absolute performance, which is the least helpful type of social information. Consequently, endogenous choice of social information does not further improve self-assessments.

Keywords: social information, feedback, overcondence, experiment, debiasing strategies