I present a model of parental decision-making with endogenous migration to study the human capital of children in migrant families, focusing on the case of rural China. I derive two main results related to (i) the effect of parental migration on children’s human capital, and (ii) the effects of government policies that increase services for children. First, I show that if government spending is a substitute for parental spending and parent time with the child is weakly complementary to spending on the child, then children’s human capital decreases as they become left-behind by migrant parents. This sheds light on a puzzle in the empirical literature about the ambiguous effects of parental migration on the human capital of left-behind children. Second, in some cases, government policies that increase services for rural children have the unintended consequence of separating children from parents due to endogenous migration effects, thereby reducing child human capital. More broadly, the analysis demonstrates how focusing on marginal treatment effects can facilitate signing comparative statics that are otherwise difficult to sign.