Do we measure overconfidence? A closer look at the interval production task

Abstract

The most common test for overconfidence in the form of miscalibration—the Interval Production task (IP)—is based on the assumption that people internalize requested confidence levels. We demonstrate experimentally that decision makers’ perceived confidence is, however, unaffected by variations in the requested confidence level. In addition, we find large heterogeneity in perceived confidence that the traditional IP measure fails to account for. We show that the alternative measure based on decision makers’ perceived confidence by contrast yields coherent, moderate overconfidence levels. Our evidence suggests that the consistency of the two measures is limited and that they are related to different individual characteristics.