在本文中，我们采用广义倾向评分匹配（GPSM）方法，这与连续可变处理交易，以收益从2010年到2017年估计的结果对中国教育与经典明瑟方程的OLS比较，以及as estimates from two instrumental variable methods (i.e., 2SLS and Lewbel). We use the Chinese General Social Survey data, including a subset newly released in 2020. We find that returns to education in China experienced a slight decrease in 2010-2015, but reverted back in 2017. With the more flexible GPSM method, we also find that returns to university education remain higher than returns to secondary or compulsory education. The GPSM estimates are also closer to OLS estimates, compared to both instrumental variable methods.