Novel Approaches to Estimating the Causal Effect of Policy Interventions in the Presence of Spillovers

本研究项目将提供新的因果推理方法来估计公共政策干预的溢出效应。政策干预可能会溢出到那些没有直接受到政策影响，但却生活在直接受影响地区（如城市或县）附近的部分人口。不考虑溢出效应可能会对公共政策的评估产生严重影响，可能会低估或高估政策的总体影响。例如，一个城市对含糖饮料征税可能会导致饮酒者前往附近的城市购买饮料。这可能会破坏在实施该税的城市评估该税对含糖饮料饮用的影响的努力。研究人员将在邻近地区不同模式和程度的政策暴露下调查政策干预的因果效应。这些有待开发的方法将有助于研究人员和政策制定者更好地理解在存在溢出效应的情况下，政策干预对相关结果的影响。将举办短期课程和讲习班，向更广泛的社区传播新方法。此外，还将对一名研究生进行指导，并开发和提供用户友好的软件。本研究项目将开发一种新的因果估计，在更宽松的因果假设下，而不是在差异中差异方法中常用的因果估计。公共政策干预通常使用差异中的差异方法进行评估。然而，这种方法并没有直接考虑到对邻近地区的溢出效应，比如附近的城市或州。利用新的识别假设，研究人员将开发基于灵活建模和机器学习的双重鲁棒估计器。该项目还将引入一个新的因果估计，可用于评估政策干预在各种社区治疗背景下的效果。研究人员将研究识别条件，以确保干预效果具有普遍性和可转移性，适用于不同组成和社区环境的目标人群。新方法将用于评估费城饮料税对费城及其周边未实施该税的县的销量的影响。这项研究也将为其他考虑征收类似消费税的城市提供指导。这项研究的成果，包括统计软件和实施指南，可以被政策制定者用来评估在特定地理区域实施的任何公共政策，这些政策可能会影响到周边地区。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This research project will produce new causal inference methods to estimate spillover effects of public policy interventions. Policy interventions can spill over to portions of the population who are not directly exposed to the policy, but nonetheless live close to regions, such as cities or counties, that are directly affected. Failure to account for spillover effects can have serious implications on the evaluation of public policies, possibly underestimating or overestimating the overall effects of the policy. For instance, a tax on sugar-sweetened beverages in one city may result in beverage drinkers traveling to a nearby city to purchase beverages. This could undermine efforts to assess the effect of the tax on the drinking of sugar-sweetened beverages in the city that implemented the tax. The researchers will investigate the causal effects of policy interventions under varying patterns and degrees of policy exposure in neighboring regions. The methods to be developed will help researchers and policymakers better understand the effect of policy interventions on outcomes of interest in the presence of spillovers. Short courses and workshops will be developed to disseminate the new methods to the broader community. In addition, a graduate student will be mentored, and user-friendly software will be developed and made available.This research project will develop a novel causal estimator under more relaxed causal assumptions than those commonly used in difference-in-differences approaches. Public policy interventions are commonly evaluated using the difference-in-differences approach. However, this approach does not directly account for spillover effects to neighboring regions, such as nearby cities or states. Using the new identification assumptions, the investigators will develop doubly robust estimators based on flexible modeling and machine learning. The project also will introduce a new causal estimand that can be used to evaluate the effect of a policy intervention under various neighborhood treatment contexts. The researchers will investigate identification conditions that ensure that intervention effects are generalizable and transportable to target populations with different compositions and neighborhood environments. The new methods will be used to assess the impact of the Philadelphia beverage tax on volume sales in Philadelphia and its surrounding counties that did not implement the tax. This research also will provide guidance to other cities considering a similar excise tax. The products of this research, including the statistical software and implementation guidelines, can be used by policy makers to assess any public policy that is implemented in a specific geographic region and has the potential to affect its neighborhoods.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.