ANALYZING SPATIALLY VARYING DETERMINANTS OF INTERNAL MIGRATION SHIFTS DURING THE COVID-19 PANDEMIC IN THE UNITED STATES

Abstract

This study examines county-level changes in internal migration during the early COVID-19 pandemic, leveraging large-scale mobile device location data (MDLD) to quantify net migration flows with unprecedented temporal granularity. The primary objective is to identify and explain the socioeconomic, demographic, housing, political, and labor-market factors associated with pandemic-era migration shifts, while accounting for spatial heterogeneity in these relationships. A multi-stage spatial modeling framework is implemented, combining Ordinary Least Squares (OLS), Lasso-based variable selection, and Multiscale Geographically Weighted Regression (MGWR). The MGWR framework improves model fit and distinguishes spatially stable from spatially varying effects. Population density, renter share, age structure, and political orientation exhibit consistent nationwide associations, whereas public transit use, group-quarters residence, rurality, and sectoral employment display pronounced regional variation. Diagnostic results confirm minimal residual spatial dependence, underscoring the robustness of the proposed MDLD-based spatial approach.