Joachim Krueger

Who they are

This page rests on one paper. The vault has verified its bibliographic record at Crossref, which gives the title as "Unskilled, unaware, or both?" with a subtitle on the better-than-average heuristic and statistical regression, the journal, volume 82, issue 2, pages 180-188, and the year 2002. Nothing about his institution or wider career was checked, so none is stated.

The paper starts from the observation that people who score low on a performance test overestimate their performance relative to others, while high scorers slightly underestimate theirs, and that Kruger and Dunning (1999) attributed these asymmetric errors to differences in metacognitive skill. Their abstract reports a replication of the 1999 study in which none of several candidate variables showed evidence of mediation, the test of whether metacognitive skill carried the asymmetry. Asymmetric errors were instead expected from statistical regression and the general better-than-average heuristic, and, consistent with that account, errors were no longer asymmetric when either regression or the better-than-average effect was statistically removed. They add that high performers, rather than low performers, were the more error prone, in that they were more likely to neglect their own estimates of other people's performance when predicting how they themselves stood relative to the group. The vault read the abstract at Europe PMC on 2026-10-25. Mazor and Fleming (2021) later described the paper's argument in the same terms.

Key papers

Krueger, J., & Mueller, R. A. (2002), "Unskilled, unaware, or both? The better-than-average heuristic and statistical regression predict errors in estimates of own performance," Journal of Personality and Social Psychology, 82(2), 180-188.

Key ideas

The Dunning-Kruger pattern can be produced by regression to the mean combined with the tendency of people to rate themselves above average.

On this account, a deficit in metacognition is not needed to explain why the lowest scorers seem to overshoot most.

In their own replication, removing either regression or the better-than-average effect statistically made the errors symmetric, and the more error-prone group was the high performers, who neglected their own estimates of others.

The 2020 test by Gignac and Zajenkowski, with 929 adults, took the same statistical line and tested it with methods that avoid grouping people into quarters.

---