SIER Working Paper Series

We propose an econometric procedure to test for the presence of overconfidence using data collected by ranking experiments. Our approach applies the techniques from the moment inequality literature. Although a ranking experiment is a typical way to collect data for the analysis of overconfidence, Benoit and Dubra (2011) show that a ranking experiment may generate data that indicate overconfidence even if participants are purely rational Bayesian updaters. Instead, the authors provide a set of inequalities that are consistent with purely rational Bayesian updaters. We propose the application of the tests of moment inequalities developed by Romano et al. (2014) to test such a set of inequalities. Then, we examine the data from Svenson (1981) on driving safety. Our results indicate the presence of overconfidence with respect to safety among US subjects tested by Svenson. However, other cases tested do not show evidence of overconfidence. We also apply our method to re-examine and confirm the results of Benoit et al. (2015).

Keywords: overconfidence; ranking experiments; moment inequality; driving safety

JEL classification: C12, D03, D81, R41