A large empirical literature dating back to the 1970s documents systematic biases in individuals’ confidence assessments of own performance. Much of the evidence points to either overconfidence or a “hard-easy effect” (overconfidence in easy tasks and under-confidence in difficult tasks). The bulk of this literature, however, rests on data generated from non-incentivised self-reports of confidence. More recently, the robustness of conclusions from this line of research has been challenged by studies from experimental economists which use incentivised tasks to elicit confidence judgements and find that overconfidence is considerably reduced. Indeed, in these recent studies, underconfidence is the typical finding.
In this Nottingham School of Economics working paper, Murad, Sefton and Starmer examine judgements of confidence in own performance using an experiment that combines two key design features. First, the authors compare confidence judgements across incentivised and non-incentivised confidence elicitation procedures. The non-incentivised procedure, designed to be very similar to those used in much of the background psychology literature, reproduces the hard-easy effect, while, in line with the recent experimental economics literature, the incentivised procedure produces a general tendency towards under-confidence. The second key feature is that the CeDEx researchers independently measure the risk attitudes of our participants. This allows the authors to filter out the effect of risk attitudes on measures of confidence elicited via the incentivised procedure, and also allows to investigate the relationship between risk attitudes and non-incentivised reported confidence. They find that filtering out the effect of risk attitudes on measured confidence reduces, but does not eliminate, the under-confidence found using our incentivised procedure. The CeDEx researchers also find a significant relationship between self-reported confidence and risk attitudes: individuals who are more risk averse tend to express lower confidence. These findings suggest that overconfidence is not as general a phenomenon as is sometimes claimed, and is, to our knowledge, the first study identifying a systematic relation between risk attitudes and confidence.
Download the revised version of this paper 2015-26 in PDF format
CeDEx Discussion Paper 2014-18, How do risk attitudes affect measured confidence? by Zahra Murad, Martin Sefton and Chris Starmer, November 2014
Zahra Murad, Martin Sefton and Chris Starmer
View all CeDEx discussion papers | View all School of Economics featured discussion papers
Posted on Saturday 1st November 2014