This paper investigates the case of interference, when a unit's treatment also affects other units' outcome. When interference is at work, policy evaluation mostly relies on the use of randomized experiments under cluster interference and binary treatment. Instead, we consider a non-experimental setting under continuous treatment and network interference. In particular, we define spillover effects by specifying the exposure to network treatment as a weighted average of the treatment received by units connected through physical, social or economic interactions. We provide a generalized propensity score-based estimator to estimate both direct and spillover effects of a continuous treatment. Our estimator also allows to consider asymmetric network connections characterized by heterogeneous intensities. To showcase this methodology, we investigate whether and how spillover effects shape the optimal level of policy interventions in agricultural markets. Our results show that, in this context, neglecting interference may underestimate the degree of policy effectiveness.

翻译：本文研究了干扰情况，即一个单位的处理也会影响其他单位的结果。当干扰存在时，政策评估主要依赖于聚类随机实验和二元处理。相反，我们考虑了连续处理和网络干扰下的非实验环境。具体而言，我们通过将网络处理暴露度定义为通过物理、社会或经济交互连接的单位的处理加权平均值，来界定溢出效应。我们提出了一种基于广义倾向性得分的估计量，以估计连续处理的直接效应和溢出效应。该估计量还允许考虑以异质强度为特征的非对称网络连接。为展示这一方法，我们研究了溢出效应是否以及如何影响农业市场政策干预的最优水平。结果表明，在此背景下，忽略干扰可能会低估政策有效性程度。