A large number of recent experimental studies show that women are less likely to sort into competitive environments. While part of this effect may be explained by gender differences in risk attitudes and overconfidence, previous studies have attributed the majority of the gender gap to gender differences in a separate 'competitiveness' trait. We re-examine this result using a powerful novel experimental technique that allows us to separate competitiveness from alternative explanations by experimental design. In contrast to the literature, the results from our experiment imply that the whole gender gap is driven by risk attitudes and overconfidence. We show that our results are due to our experimental approach, which circumvents concerns raised against the regression-based method used by previous studies. Our results have important implications for policy and future research.