A0724
Title: Causal transportability with local interference: A framework for policy effect prediction across regions
Authors: Gary Hettinger - New York University Grossman School of Medicine (United States) [presenting]
Youjin Lee - Brown University (United States)
Nandita Mitra - University of Pennsylvania (United States)
Abstract: Policymakers often need to anticipate the impact of policies in regions where they were not yet implemented, relying on evaluations from regions where they have. For example, sweetened beverage taxes is studied in U.S. cities like Philadelphia, yet many municipalities are still considering adoption. Substantial heterogeneity in observed effects across regions underscores the need for methods that can reliably transport these effects to new settings. Transporting policy effects requires adjusting for demographic and contextual differences. However, existing approaches often overlook key factors like treatment interference from cross-border shopping and exposure heterogeneity from local economic dynamics. Transportability methods are extended by introducing a formal causal framework that incorporates spatial interference and heterogeneous exposure distributions, both of which may differ between observed and target regions. The method is assessed through simulation studies, and it is applied to estimate the impact of Philadelphia's beverage tax in other U.S. regions. Leveraging retail scanner data, zip-code-level Census demographics, and geographic proximity measures, tax effects are estimated in untreated areas, and the estimate's performance in already-taxed regions is assessed. Findings highlight the importance of accounting for regional variation not only in population characteristics, but also in spatial spillovers and economic dynamics, when transporting policy effects.
