Based on CHNS 2011, the paper used quantile regression model to explore the distribution characteristics of rates of return to education in different quantiles of income distribution. The results show that the rates of return to education become lower as the quantiles of income distribution get higher. In low quantiles, the corresponding rates of return are higher, while in high quantiles, the rates of return are relatively low, which indicates that low income groups benefit more from education expansion to increase their income. In terms of education levels, junior and senior high school education has no obvious effects on improving the income of low income groups, while they can narrow the income gap by accepting vocational education, university education or above. In addition, the labor market characteristics and employment features have a significant effect on the rate of return to education in each quantile of income. The conclusions of this paper are as follows: education expansion policy itself does not lead to the income gap among residents, while the changing rates of return to different education levels due to labor market characteristics and employment features have greater impact on income distribution than the changed education level itself. In order to reflect the differences of productivity through the rates of return to education, it is needed to reduce the wage differences among different departments and correct the distortion of the labor market.