The majority of existing research on mobility indicates that students do worse in the year of a school move. This research, however, has been unsuccessful in isolating the causal effects of mobility and often fails to distinguish the heterogeneous impacts of moves, conflating structural moves (mandated by a school's terminal grade) and non-structural moves (induced by residential mobility or by access to a better school) for example. Moreover, there is little evidence on the effects beyond the first year of a move. In this paper, we obtain credibly causal estimates of the impact of mobility on performance in both the short and long run, addressing heterogeneity in the impacts of mobility and the endogeneity of moving. We do so using richly detailed longitudinal data for five cohorts of New York City public school students making standard academic progress from grades 1-8. We estimate the impact of moving to a new school in a model with student fixed effects and two alternative sets of instrumental variables--the grade span of a student's first grade school and foreclosure/building sale--to isolate the causal effect of mobility that is likely planned and mobility that is likely due to unanticipated shocks, respectively. We find negative short-term as well as long-term effects of the structural moves built into the school system. Non-structural moves, however, have a positive effect on academic performance if they are made to join a new school at the beginning of that school's grade span and, thus, more likely made for strategic reasons. Robustness checks indicate results are not sensitive to inclusion of school quality measures, pre-move trends in mobility, or alternative samples. In the conclusions, we discuss the importance of findings on the heterogeneous impact of school moves to the literature and to policy makers.