We study the role of second-order beliefs in explaining the gender competition gap. To do so, we extend an existing meta analysis to gather performance and choice data on various tasks used in the literature. Building on this, we run an online experiment to collect belief data on these tasks. First, we note that the extent of the competition gap dependents on the task used. Second, we document important gender differences in beliefs over performance. While men severely overestimate the male performance advantage for all tasks, women report more nuanced beliefs. Further, women’s beliefs about the performance differences display much stronger correlation with the competition gap than the actual performance differences. To test our hypothesis that second-order beliefs causally affect competitiveness, we plan a lab experiment wherein we vary participants’ second-order beliefs.