Primary Submission Category: Difference in Differences, Synthetic Control, Methods for Panel and Longitudinal Data
Causal Supply–Demand Decomposition with Movers
Authors: Zhu Shen, Jose Zubizarreta,
Presenting Author: Zhu Shen*
Mover designs are widely used to analyze regional differences by leveraging individual relocations to separate place-specific effects from population composition. Although commonly used to address causal questions, these designs are rarely formulated within an explicit causal framework, making the resulting estimates difficult to interpret for policy-relevant decomposition. We develop a causal framework for movers that defines origin–destination–time–specific causal effects and makes the associated identification assumptions explicit. These cohort-specific mover effects serve as building blocks for a general supply–demand decomposition of regional differences. We show that, in multi-period and multi-region settings, observed cross-sectional outcome differences can be decomposed into causal supply effects, defined by holding cohort composition fixed, and demand effects arising from population heterogeneity. To estimate these effects, we propose a weighting-based estimator that recovers cohort-specific mover effects by approximately balancing pre-move covariates between movers and appropriate stayer comparison groups. Applying the framework to Medicare data on patient moves across Hospital Referral Regions, we uncover substantial heterogeneity in mover effects by origin, destination, and timing, and show how place-specific factors and population composition jointly contribute to geographic variation in health care utilization.