The Dunning-Kruger effect is (mostly) a statistical artefact: Valid approaches to testing the hypothesis with individual differences data

Gignac (2020) · Intelligence · Read the paper

Gignac and Zajenkowski argue that the usual way of testing the Dunning-Kruger hypothesis, comparing self-assessed and measured ability across quarters of the sample, is confounded by the better-than-average effect and by regression to the mean. They propose two alternative tests for individual-differences data, the Glejser test of heteroscedasticity and quadratic regression, and show with simulated data that the quartile method can produce the pattern when none is built in. In a community sample, the quartile method showed the familiar picture while both alternative tests found no effect. They conclude that the effect is mostly, though not entirely, a statistical artefact, and that its true size may be much smaller than reported.

Gignac and Zajenkowski argue that the usual way of testing the Dunning-Kruger hypothesis, comparing self-assessed and measured ability across quarters of the sample, is confounded by the better-than-average effect and by regression to the mean. They propose two alternative tests for individual-differences data, the Glejser test of heteroscedasticity and quadratic regression, and show with simulated data that the quartile method can produce the pattern when none is built in. In a community sample, the quartile method showed the familiar picture while both alternative tests found no effect. They conclude that the effect is mostly, though not entirely, a statistical artefact, and that its true size may be much smaller than reported.

Written from the abstract and text of an openly hosted copy of the paper.

What our sources record

Recorded when this source was checked, figures included, with page references and quotation marks removed.

Gignac, G. E., & Zajenkowski, M. (2020). The Dunning-Kruger effect is (mostly) a statistical artefact: Valid approaches to testing the hypothesis with individual differences data. Intelligence, 80, 101449. doi:10.1016/j.intell.2020.101449