The major empirical challenge for estimating the driving force of the gender difference is the omitted variable bias because the proposed factor is usually intertwined with other observable and unobservable characteristics. This paper adopts a novel approach to tackle this identification problem and examines how the relative income defines gender roles in the context of mortgage signing order. We first document that men sign first 89% of the time when a mixed-gender couple applies for a housing loan. The signing order reflects purely behavioral preferences, as there is no difference in the legal rights or obligations of either the first or second borrowers. We show that relative income is the major determinant of the observed gender gap in signing order and we exclude other possible explanations such as discrimination, loan officer bias and social norms. To isolate relative income from differences in other dimensions, we compare the signing order of the mixed- and same-gender couples and find that the effects of relative income not only persists but also are fairly comparable between the two groups. Finally, we provide consistent macro-level evidences by both cross-sectional and dynamic analysis and show that gender gap in signing order is higher in states with larger gender wage gap and experiences a declining trend over time.