Revisiting the Dunning-Kruger effect: Composite measures and heterogeneity by gender

• We test for the Dunning-Kruger effect using composite measures and birth cohort data. • We use tests from ages 5/10/16 and self-estimates of ability at ages 10/16. • We show that the Dunning-Kruger effect exists in our secondary data. • We find that the Dunning-Kruger effect is not gender specific. • Women are underconfident and men are overconfident. The Dunning-Kruger effect (DKE) states that people with lower levels of the ability tend to self-assess their ability less accurately than people with relatively higher levels of the ability. Thus, the correlation between one's objective cognitive abilities and self-assessed abilities is higher at higher levels of objective cognitive abilities. There has been much debate as to whether this effect actually exists or is a statistical artefact. This paper replicates and extends Gignac and Zajenkowski (2020) and Dunkel, Nedelec, and van der Linden (2023) to test whether the DKE exists using several measures of ability and nationally representative data from a British birth cohort study. To do this, we construct a measure of objective cognitive abilities using 18 tests conducted at ages 5, 10, and 16, and a measure of subjective self-assessed abilities using estimates of school performance and being clever at ages 10 and 16. We replicate their models and show that the DKE exists in our secondary data. Importantly, we are the first to look at whether this relationship is heterogeneous by gender and find that while the self-assessment bias is gender specific, the DKE is not. The DKE comes from men relatively overestimating and women relatively underestimating their abilities.