Contagion, Migration, and Misallocation in a Pandemic

This deposit is the replication package for "Contagion, Migration, and Misallocation in a Pandemic" (Journal of Health Economics). It contains all code and data needed to reproduce every quantitative result in the paper and its appendices. RESEARCH QUESTION AND FINDINGS. The paper studies how individuals' migration decisions during a pandemic interact with the spatial allocation of hospital resources, characterising both the laissez-faire equilibrium and the planner's allocation in closed form. Susceptible individuals leave high-infection cities while infected individuals move toward better-resourced ones, but these private decisions diverge from the planner's along two dimensions. Migration of infected agents toward susceptible-rich destinations raises transmission risk, a contagion externality the planner internalises by restricting such flows; migration toward better-resourced cities relieves hospital congestion at the epicentre, a congestion externality the planner internalises by encouraging them. The two externalities work in opposite directions. The framework explains a cross-country pattern that standard misallocation theory cannot: during COVID-19, countries with greater regional dispersion in mortality tended to have lower aggregate mortality. CONTENTS. (1) "MATLAB codes" holds eight driver scripts and eight helper functions that solve the SIR-migration model numerically and produce Figures 2-6, Table 6, and Figures C.1 and D.1. This component requires no external data: every input is a calibrated parameter reported in Table 5 of the paper. (2) "Empirical data and Stata codes" holds smr_analysis.do, the input dataset covid_smr_panel.xlsx, and boot_betas_TableA1.csv (bootstrap replicates retained for verification). A single run of the do-file reproduces Tables 1, 2 and A.1-A.5 and Figure A.1. DATA. The empirical dataset covers nine countries and 264 sub-national regions as a snapshot in late August 2020: United States (51 regions), Japan (47), India (35), China (31), Brazil (27), Italy (21), Spain (19), South Korea (17), Germany (16). All sources are public; no confidential or licensed data are used. Regional standardised mortality ratios are obtained by indirect age standardisation, applying O'Driscoll et al. (2021) age-specific infection-fatality rates to each region's population by five-year age bin, scaled so that the population-weighted within-country mean equals one. The workbook also documents regional age structure and the fatality rates used, and includes a variable dictionary giving the definition, units and source of all 28 panel variables. SOFTWARE. MATLAB R2016b or later (no additional toolboxes) and Stata 14 or later with the user-written module asdoc. HOW TO USE. README.pdf, at the top level of the package, gives step-by-step instructions and a table listing each figure and table in the paper together with the single script that reproduces it. Results are deterministic; the only randomisation is a seeded bootstrap.