Abstract. This paper uses Chinese micro data and new semi-parametric methods to estimate the current return to college education allowing for heterogeneous returns and for self-selection into schooling based on them. OLS and IV methods do not properly account for this sorting. Our estimates suggest that, for a randomly selected young person from an urban area, college attendance leads to a 43% increase in lifetime earnings (nearly 11% annually) in 2000, compared with just 36% (nearly 9% annually) for those who do not attend. Our evidence suggests that the return to education has increased substantially in China since the early 1990s.