Are worst students really more overconfident? A preliminary test of different measures

Aim of this work is to contribute to the discussion about the link between metacognitive abilities and overconfidence. In particular we want to empirically investigate the existence and the sign of the above relation. In the work we therefore elicit forecast evaluation: (1) using data from two sets of midterm exams and question-by-question evaluations of confidence levels, (2) introducing as an incentive a negative evaluation for incorrect forecasts and (3) proposing four different measures of overconfidence and predictive ability. Our results show that (1) there is significant evidence of a good ability of self-evaluation on the side of the best students; but (2) worse metacognition does not seem to explain overconfidence. This suggests that different methods of investigating overconfidence might lead to results which are at least partially different from the ones discussed in the existing literature.