The United States is a highly mobile society. Between 2010 and 2011, 12 percent of Americans (roughly 35 million people) moved residences, approximately two-thirds of whom moved within the same county. In the schooling context, mobility is often viewed as a disruptive phenomenon that interrupts student learning and has negative consequences for performance.  A wide, largely descriptive, literature on the impacts of residential mobility on student performance tends to support this view—finding negative associations between mobility and achievement. Such negative associations are also found in NYC, where students who move residences at least once between third and eighth grade score approximately 0.1 SDs lower in both eighth grade reading and math than their peers who never move.

There are several weaknesses of the prior literature on this topic, however. First, it often ignores the fundamental differences between movers and non-movers. Second, it tends to conflate residential and school mobility, making it impossible to determine whether negative relationships reflect the impacts of residential mobility itself or of school mobility, which often accompanies residential moves. Finally, it imposes homogeneity of impacts, despite theories of mobility that would suggest differential impacts based on the type of move being made (i.e. near versus far, to a better or worse neighborhood, etc.).

Using a rich, longitudinal dataset that allows us to track student residential moves during elementary and middle school, we aim to improve the current literature by producing causal estimates of the impacts of residential mobility on student performance and providing evidence of heterogeneity in these effects. In this paper we answer the following questions:

1. How does residential mobility affect academic outcomes?
2. Does residential mobility have any effect above and beyond school mobility?
   
3. How does the effect of mobility depend on the type of move being made (i.e. near versus far, into different housing, in response to foreclosure)?

Using a simple conceptual model of residential mobility, we predict that the effects of residential mobility on student performance are likely heterogeneous, with the direction of the impact depending on the events precipitating the move as well as household tastes. We propose two strategies for identifying the causal effect of residential mobility:  a student fixed effects model and an instrumental variables model with student fixed effects. In our student fixed effects models, we find that any residential mobility is associated with a decrease in both ELA and math achievement.  Furthermore, while residential moves made in isolation are negatively related to performance, those made in conjunction with school moves are positively related to both ELA and math achievement. In our IV models, we explore multiple sets of instruments and find that our results are sensitive to the choice of instruments, demonstrating significant heterogeneity of impacts. When using distance and future housing type as instruments, we find positive impacts of mobility on both ELA and math ranging in size from 0.1 to 0.5SDs.  Conversely, when using moves into public housing and foreclosure as instruments we find no or negative impacts of mobility.