This paper examines whether family and community migration experience affect the probability of high school graduation in Mexico once unobserved heterogeneity is accounted for. Bivariate random effects dynamic probit models for cluster data are estimated to control for the endogeneity of education and migrant network variables. Correlation of unobservables across migration and education decisions as well as within groups of individuals such as the family are explicitly controlled for. Results show that migrant networks reduce the likelihood of high school graduation. Negative migrant selection is detected at the individual level while positive migrant selection is found at the family level.