Gender differences in risk attitudes are frequently observed, although recent literature has shown that they are context dependent rather than ubiquitous. In this paper we try to rationalize the heterogeneity of results investigating experimentally whether the presence of a safe option among the set of alternatives explains why females are more risk averse than males. We manipulate three widely used risk elicitation methods finding that the availability of a safe option causally affects risk attitudes. The presence of a riskless alternative does not entirely explain the gender gap but it has a significant effect in triggering or magnifying (when already present) such differences. Despite the pronounced instability that usually characterizes the measurement of risk preferences, we show, estimating a structural model, that the effect of a safe option is remarkably stable accross tasks. This paper constitutes the first successful attempt to shed light on the determinants of gender differences in risk attitudes.