Using the China Family Panel Studies, we identify the subjects studied by vocational college and university graduates, with the latter group further divided into ordinary and key universities. While the returns are around 8-10% to attending colleges and ordinary universities, there are higher returns of 12-16% per annum to attending the more prestigious key universities. The recent massive expansion of the higher education sector resulted in reduced returns to all HE types, except for graduates who studied subjects other than LEM (law, economics, and management) or STEM (sciences, technology, engineering and math/medicine) at key universities. We further account for selection on observables into subjects and tiers using the doubly robust Inverse Probability Weighted Regression Adjustment method (IPWRA) approach. While these results are tentative, they suggest that pooled OLS and random-effect models substantially underestimate the effect of attending universities that are more prestigious for graduates of both genders in LEM.