ESTIMATES OF OWN LETHAL RISKS AND ANCHORING EFFECTS.

The paper presents an experiment showing that some of the results in Benjamin, Dougan and Buschena (2001) (BDB hereafter) should be attributed to an anchoring effect. More precisely, it appears that, when asked to evaluate successively the number of deaths per lethal risks for two different populations, respondents anchor their answers in the second survey, on the answers they gave in the first survey. The experimental outcomes also indicate that, contrary to BDB's conjecture, agents' estimates of their own lethal risks exhibit the traditional biases (i.e. agents overestimate rare risks, and underestimate common risks). However, if the quality of an estimate is measured not only by its mean, but also by its variance, then the present experiment cannot dismiss unambiguously BDB's hypothesis that agents have better information about their own risks.