This study examines the determinants of college graduates' work migration decision and explores its impact on college graduates' starting salaries in China, using most recent nationally representative CSLM 2011 survey data. This study is the first one that incorporates student characteristics, institution attributes and regional characteristics (both economic and non-economic factors) in the regression analysis on determinants of work migration in the Chinese context. When investigating work migration's effects on graduates' starting salaries, in addition to the OLS model, the study also employs alternative identification strategies including instrumental variable method and propensity score matching method to account for the potential endogeneity of work migration and reduce selection bias. In addition, this study addresses sample selection issue with Heckman correction technique. The results reveal that the following variables have a significant positive impact on college graduates' work migration: study migration, science or engineering major, student leader in high school, passed CET4, passed CET6, engineering-concentrated institution, from 985 institutions, and from 211 institutions.

College graduates who possess the above individual characteristics or are from the above institutions are more likely to migrate to work. Meanwhile, this study also finds that female, provincial GDP per capita, provincial population, provincial area size, and provincial ECI score are significant and negative determinants of work migration. In terms of work migration's impact on college graduates' starting salaries, the weighted and unweighted OLS regression analyses reveal that new graduates who decide to migrate for work enjoy a 9.9% and 8.6% starting monthly salary premium over those graduates who do not do so, respectively. Three different PSM schemes are used to conduct PS-adjusted regressions and the results also show that work migration has a statistically significant positive impact on graduates' starting salaries. Consistent with the PSM findings, the magnitude of the coefficient estimate for work migration in IV regression is larger than the OLS estimate. Specifically, for the regression using two instrumental variables, the results find that college graduates who decide to migrate for work have a 15.8% higher starting salary compared to those who would not.

These findings of this study have important implications for educational policy makers and higher education institutions in China.