Accounting for spillover when using the augmented synthetic control method: estimating the effect of localized COVID-19 lockdowns in Chile Volume: XX

Taylor Krajewski ∗ Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A, Michael Hudgens ∗ Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A · 2025

Read Accounting for spillover when using the augmented synthetic control method: estimating the effect of localized COVID-19 lockdowns in Chile Volume: XX by Taylor Krajewski ∗ Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A, Michael Hudgens ∗ Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-16244.txt.

Introduction

A lockdown in one town leaks into the next: workers cross lines, and so does the virus. Taylor Krajewski and Michael Hudgens build a method in 2025 that prices both the direct hit and the leak. Their test case is Chile's local COVID-19 lockdowns, town by town.

The base method forges a fake twin of each treated town from untouched ones. The twist here sorts the donor towns by how much lockdown touched them too. Donors fall into clean and brushed groups, so direct and spill effects split apart. Fake-data tests show the split cuts bias. Policy reads that ignore the leak will misprice local rules, the work warns. The frame suits any local act with cross-line reach.

Catalog record

CREATORTaylor Krajewski ∗ Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A, Michael Hudgens ∗ Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-16244

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Accounting for spillover when using the augmented synthetic control method: estimating the effect of localized COVID-19 lockdowns in Chile Volume: XX Taylor Krajewski ∗ Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A, Michael Hudgens ∗ Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A Abstract The implementation of public health policies, particularly in a single or small set of units (e.g., regions), can create complex dynamics with effects extending beyond the directly treated areas. This paper examines the direct effect of COVID-19 lockdowns in Chile on the comunas where they were enacted and the spillover effect from neighboring comunas. To draw inference about these effects, the Augmented Synthetic Control Method (ASCM) is extended to account for interference between neighboring units by introducing a stratified control framework. Specifically, Ridge ASCM with stratified controls (ASCM-SC) is proposed to partition control units based on treatment exposure. By leveraging control units that are untreated, or treated but outside the treated unit’s neighborhood, this method estimates both the direct and spillover effects of intervention on treated and neighboring units. Simulations demonstrate improved bias reduction under various data-generating processes. ASCM-SC is applied to estimate the direct and total (direct + indirect) effects of COVID-19 lockdowns in Chile at the start of the COVID-19 pandemic. This method provides a more flexible approach for estimating the effects of public health interventions in settings with interference. \pagerange Accounting for spillover when using the augmented synthetic control method: estimating the effect of localized COVID-19 lockdowns in Chile – References \artmonth XXX XXX/XXX.x keywords Causal Inference, Interference, Synthetic Controls, COVID-19 1 Introduction Synthetic control methods (SCMs) have gained much popularity in panel data settings with a goal of estimating a treatment (or intervention) effect on a single or limited number of treated units. These methods construct a synthetic control, a weighted average of untreated units’ outcomes, with weights determined to minimize the difference in pre-treatment outcomes between the treated unit and control units. The synthetic control serves as a proxy for the counterfactual outcome had the treated unit not experienced the intervention, and in turn, the effect of the treatment in the post-treatment period is estimated via the difference between the treated unit’s outcome and that of its synthetic control. The original SCM ( Abadie and Gardeazabal 2003 ; Abadie et al. 2010 ; Abadie 2021 ) was developed for estimating the treatment effect on a single treated unit. Since its inception, SCM has been extended and adapted in a myriad of ways. Abadie 2021 provides a detailed overview of many recent SCM-related methodological developments. Many SCM modifications continue to focus on a single treated unit while altering assumptions for greater applicability. For example, regression-based methods that allow for extrapolation ( Arkhangelsky et al. 2021 ; Chernozhukov et al. 2019 ; Doudchenko and Imbens 2016 ), methods that incorporate matrix completion/estimation ( Amjad et al. 2019 ; Athey et al. 2020 ), and bias-correction methods ( Abadie and L’Hour 2021 ; Arkhangelsky et al. 2021 ; Ben-Michael et al. 2021 ; Chernozhukov et al. 2019 ) have been proposed. Other modifications have extended SCM to account for multiple treated units ( Ben-Michael et al. 2022 ; Abadie and L’Hour 2021 ; Robbins et al. 2017 ; Kreif et al. 2016 ; Hainmueller 2012 ). The majority of advancements in synthetic control methodology assume no interference (or spillover) between units, i.e., the treatment of one unit does not affect the outcomes of any other units. However, this assumption may be questionable in settings where units are in close proximity, and units close to the treated unit(s) may experience spillover effects of the treatment. Increasing attention has been paid recently to estimating causal effects in the presence of interference with a goal of disentangling the direct and spillover (or indirect) effects of treatment in both randomized and observational studies ( Forastiere et al. 2021 ; Ogburn and VanderWeele 2014 ; Tchetgen Tchetgen and VanderWeele 2012 ) . However, only a handful of SCM methods allow for spillover effects. Some methods have extended SCM to incorporate spillover effects under the partial interference assumption, which limits interference to predefined groups or neighborhoods. For example, Menchetti and Bojinov 2022 use a multivariate Bayesian structural time series model to estimate direct and spillover effects, while Grossi et al. 2024 apply penalized SCM to estimate synthetic controls for a single treated unit and its neighbors, relying on untreated units outside the neighborhood as controls. Other approaches relax the partial interference assumption but still assume a priori knowledge of units experiencing spillover. Melnychuk 2024 developed Iterative SCM to estimate the direct effect on a single treated unit, iteratively removing spillover bias by creating synthetic controls for spillover-affected units. Cao and Dowd 2019 assume a known linear spillover structure and estimate synthetic controls for all units (treated and controls) but allow all units, including those potentially affected by spillover, to serve as controls in estimating direct and spillover effects. Di Stefano and Mellace 2024 introduce the inclusive SCM (iSCM) method, which does not require prior knowledge of the spillover structure but does require prior identification of potentially affected units (units that could be indirectly affected by the treatment) and at least one unaffected control for accurate estimation. Their approach allows potentially affected units and other treated units to serve as controls, solving a system of equations to estimate direct and spillover effects. The work in this manuscript is motivated by a setting in which several units receive treatment, and the units receiving treatment are spatially clustered. While the methods described above focus on estimating the direct effect of treatment on a single treated unit and the spillover effects onto (neighboring) untreated units, our approach estimates the direct effect of treatment on a treated unit as well as the additional spillover effect of a neighboring unit receiving treatment (i.e., spillover between treated units). Specifically, we aim to evaluate the causal effect of localized lockdowns in the initial months of the COVID-19 pandemic on the regions that initially went under lockdown and their neighboring regions that also went under lockdown. Methodologically, our approach extends the Augmented Synthetic Control Method (ASCM) ( Ben-Michael et al. 2021 ) by incorporating the Stable Neighbor Treatment Value Assumption (SNTVA), which accounts for interference between neighboring units ( Agarwal et al. 2023 ) . We develop Ridge ASCM with stratified controls (ASCM-SC) to estimate direct, total, and spillover effects by stratifying control units based on their exposure to treatment. This approach can improve the estimation of the causal impact of interventions in settings where multiple neighboring units are treated, and spillover effects are likely to occur between treated units. The paper is organized as follows: Section 2 provides details of the motivating dataset. Section 3 presents the methodology, providing an overview of SCM and ASCM and presenting a modification that accounts for potential spillover effects. In Section 4 results from a simulation study are presented, and in Section 5 the method is applied to estimate the effects of COVID-19 comuna-level lockdowns in Chile. Section 6 concludes with a discussion. 2 Motivating Example During the first year of the COVID-19 pandemic, the country of Chile

Read the complete text