By using the survey data of peasant workers in Jilin Province and sample survey data in rural areas of Jilin Province, the paper applies quantile regression and Heckman TwoStep sample selection bias methods in the study of the effect of human capital on the income peasant workers. The study shows that in different points of the quantile regression, return to education is different. In 25% quantile regression return to education is the largest (1.89 percent), while the median regression return to education is the smallest (1.40 percent). The income gap of longtime migrant workers group is greater than a short period of time working group. The income gap of the receiving Other organizations training group is greater than the group of receiving business training. As a result of amendments to sample selection bias, return to education rises to 4.55 percent, so the study in the income of migration of rural labor does not lose sight of the sample selection bias.