Many people believe that men are generally more willing to take risks than women. Yet, systematic efforts to test this belief—through multi-context studies, literature reviews, and meta-analyses—lead to mixed conclusions. Some studies find clear sex differences, while others highlight inconsistencies and publication bias. These approaches also have important limitations. In addition to being affected by publication bias, literature reviews and meta-analyses suffer from narrow study designs: when most studies use similar samples and measures, summarizing the literature fails to determine whether men are generally more risk-seeking than women. Furthermore, all multi-context studies rely on a single risk measure, which may obscure the fact that sex differences in risk attitudes are context-specific. My study avoids these limitations. My starting point is that context may be important. I therefore describe the dimensions of heterogeneity that could matter using a “design space.” I then assemble a rich dataset (76 different risk measures across 82 countries) to populate this space with 1,835 estimates of sex differences in risk attitudes. To avoid publication bias, I only rely on data unfiltered by the publication process. My results show that context matters: in the gain domain, men are generally more risk-seeking. In the loss domain, sex differences are less systematic, and, on balance, men appear to be less risk-seeking than women.