This paper investigates gender differences in depression among the Chinese elderly population using data from the China Health and Retirement Longitudinal Survey (CHARLS). We apply a threefold Blinder-Oaxaca Decomposition method to distinguish among: (1) effects due to the companions having different social characteristics (the endowments effect); (2) effects due to these characteristics having different influences on mental health situation (the coefficients effect); and (3) the interaction of those two effects (the interaction term). Our results reveal that the bulk of the gender depression differential is attributable to differences in the associations between social characteristics and depressive symptoms, indicating sizable differences in the mental health situation between men and women.