Overconfident for real? Proper scoring for confidence intervals

Abstract

Studies show that people tend to provide overly narrow confidence intervals for unknown values. Such a form of overconfidence would have an important impact on financial markets, among other domains, leading i.a. to excessive trading. The present study is one of the very few that try to incentivize reporting correct confidence intervals. To this end, a reward scheme is proposed, based on a combination of asymmetric loss functions minimized by appropriate quantiles of a probability distribution. In the experiment I find that incentivized subjects provide wider confidence intervals, obtaining a higher hit rate than the control group. The effect is stronger than that of feedback and explicit warning. These findings suggest that the overly narrow confidence intervals reported elsewhere are partly due to an insufficient mental effort that subjects exert and that they can be induced to do so by the proposed incentive scheme.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 3 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.