This report addresses the scope of China’s left-behind phenomenon and its roots in migration and education policies. It reviews evidence about disadvantages associated with left-behind status and discusses recent policy responses to the left-behind phenomenon. Empirical evidence is drawn from a national study of middle school students and a 15-year longitudinal case-study of children from rural Gansu, China. While a number of prior studies have shown mixed findings about the scale of educational disadvantage of left-behind children, compared to other groups, evidence presented here indicates that even after adjusting for school or community and household socioeconomic status, there are multiple domains in which homes of left-behind children are disadvantaged. They tend to live in households characterized by poorer health resources, cultural resources and social resources. By definition, they lose access, at least temporarily, to the “human capital” of their absent parents. Children in the short term thus experience more physiological, psychological, and (in the national comparison) educational disadvantages than their non-left-behind counterparts. In the long-term, our case study from Gansu Province suggests that father absence is associated with reduced educational attainment and possibly greater propensity to migrate, but not employment or long-term family relations. Overall, disadvantages appear to be more consistent and more generalized for mother-absent and dual-parent-absent families than for father-absent families. We discuss policy responses, and possible policy strategies, in the closing segment of the report. Policy reforms that obviate the need for children to be left behind are one evident solution to the problem, and some steps appear to be happening in this direction, but local resistance may be substantial. More immediately, boarding schools and community centers are commonly-proposed policy solutions to address the immediate needs of left-behind children, with promise but some clear pitfalls. Other possible supports are discussed.