http://www.uh.edu/econpapers/RePEc/hou/wpaper/201309855.pdf

This paper performs a meta-analysis to investigate how changes over time, model specifications,
differences in data sets, and variable definitions could contribute to the differences in estimates of returns
to education in China. The results show that approximately 10 percent of the variation can be explained
by changes in labor market over time, while the other 45 percent can be explained by differences in
samples used and empirical methods. Return to education has increased approximately 0.2 percentage
points a year since the economic reform, and increases more quickly as the reform progresses; however,
this accelerating trend has reached a stop in the last few years when the global recession hit China. We
also find that returns to education for rural-to-urban migrant workers are 2.3 percentage points lower than
that of urban workers. We conclude that the increasing reward for human capital accumulation over time
signals that China is moving toward a well functioning labor market.