Selection of Movers on Observable Characteristics and the Effect of Place on Health and Health-Care Spending

Understanding the effect of place on health and health-care spending is a long-standing question with important implications for improving health and the efficiency of health-care spending. To answer this question, recent studies have exploited variation in place from elderly persons moving. The key assumption of these studies is that moving is exogenous conditional on observed characteristics (e.g., age and sex) and place-of-origin fixed effects (or person fixed effects). In this article, we document the extent of selection among elderly movers on a set of observable characteristics and estimate differences in spending and hospitalizations associated with such selection. Specifically, we measured the amount of selection between movers and non-movers, and among movers by the type of move made, as characterized by differences in Medicare spending and hospitalization between origin and destination locations. Our analysis shows that there is a substantial amount of selection among movers on observable characteristics not used in previous studies and that such selection is associated with large and economically important differences in Medicare spending and hospitalizations. We also show that the inclusion of person fixed effects does not eliminate the problem from unmeasured confounding due to time-varying effects of the observed characteristics on individual outcomes.