This paper examines the importance of differences in reservation wages for the gender wage gap. Based on two waves of rich survey data for a sample of newly unemployed individuals in Germany, we perform a decomposition analysis including measures for reservation wages, detailed information on education, socio-demographics, labor market history, as well as personality traits. In order to address the potential endogeneity of reservation wages we exploit a generated instrumental variable strategy that relies on heteroscedasticity of the error terms. Our findings indicate that the gender wage gap becomes small and statistically insignificant once we control for reservation wages. Moreover, we perform a subgroup analysis that provides valuable insights about the importance of potentially unobserved characteristics that affect reservation wages and realized wages simultaneously. Reasons for differences in reservation wages could arise from productivity differences, the fact that women anticipate discrimination and different unobserved traits or preferences.