In this paper, we identify female long-term wage returns to college education using the educational expansion between 1960-1990 in West Germany as exogenous variation for college enrollment. We estimate marginal treatment effects to learn about the underlying behavioral structure of women who decide for or against going to college (e.g., whether there is selection into gains). We propose a simple partial identification technique using an adjusted version of the Lee bounds to account for women who select into employment due to having a college education, which we call college-induced selection into employment (CISE). We find that women are, on average, more than 17 percentage points more likely to be employed due to having a college education than without. Taking this CISE into account, we find wage returns of 6-12 percent per year of education completed (average treatment effects on the treated).