Recent evidence on the gender wage gap shows that it has remained stagnant for those with a university degree and is the largest at the top of the earnings distribution. Many studies have explored institutional factors that contribute to the gender wage gap, but there is little evidence on the role of non-cognitive traits, including overconfidence. This is surprising given its prominence in academic and popular literature. We use a measure of overconfidence captured in adolescence to explain the gender wage gap at age 42. Our results show that overconfidence explains approximately 5.5% of the unconditional gender wage gap. This is driven by women being more underconfident, not men being more overconfident. Furthermore, we find negative wage returns on being underconfident for both men and women. Most of this penalty works via occupational sorting, having lower pre-university educational outcomes, and being less likely to study high-return subjects at university. This has implications for the limitations of workplace-based interventions aimed at boosting women's confidence.