This research project developed new causal inference methods to estimate spillover effects of public policy interventions. While public policies are often evaluated using the difference-in-differences approach, this method typically assumes that units not directly exposed to the policy remain unaffected. In reality, policies can influence neighboring regions -- such as nearby cities or states -- through spillover effects (Figure 1). Failing to account for these effects can lead to biased or misleading conclusions about a policy's impact. This research is motivated by the Philadelphia beverage tax, which was associated with a substantial decrease in the volume of taxed beverage sales within Philadelphia, but also with an increase in beverage sales in bordering counties (due to cross-bordering shopping)  that were not subject to the excise tax.

To address the limitations of existing methods in accounting for spillover effects, we extended difference-in-differences approaches to identify causal effects under various spillover scenarios. First, we introduced novel causal estimands for evaluating policy effects in the presence of spillovers. We then established identification conditions and developed corresponding estimators based on these assumptions. Our proposed estimators are both robust and efficient: they rely on fewer modeling assumptions while leveraging the available data as fully as possible. We also developed causal inference methods to examine the heterogeneous impacts of the policy across different demographic and geographic subgroups (Figure 2). Lastly, we developed a new causal framework to evaluate policy effects in counterfactual treatment scenarios, providing causal quantities for designing effective policies for populations subject to various neighborhood statuses (Figure 3).

Our proposed causal inference methods can help policy makers better understand the effect of the Philadelphia beverage tax on outcomes such as beverage sales in the context of cross-border shopping. Importantly, these methods also enable policymakers to generalize the results to other cities considering a beverage excise tax. More broadly, our toolkit can be applied to evaluate a wide range of public policies (e.g., gasoline excise taxes, COVID-19 masking requirements, community policing, marijuana legalization) that are implemented in specific geographic regions but may affect -- and be affected by -- surrounding areas that have not implemented the policy. 

Through this project, we advised a graduate student, incorporated the developed methods into causal inference courses, published findings in statistics and health policy journals, and presented results at national and international conferences. We also provided public health and policy researchers with well-documented software to implement the proposed methods, along with guidance on how to interpret the results for policy applications. 

This Project Outcomes Report for the General Public is displayed verbatim as submitted by the Principal Investigator (PI) for this award. Any opinions, findings, and conclusions or recommendations expressed in this Report are those of the PI and do not necessarily reflect the views of the National Science Foundation; NSF has not approved or endorsed its content.

Last Modified: 07/24/2025
Modified by: Youjin Lee