обобщенный метод соответствия,
функция отклика на непрерывное воздействие,
метод разности разностей,
оценка политики,
несмещенный алгоритм машинного обучения,
условный средний эффект воздействия
Abstract
Continuous treatment effect evaluation is complicated by confounders. The dose is commonly related to observed characteristics of the treated unit. This paper provides an overview of approaches for evaluating continuous treatment effects. They are 1) the dose-response function; 2) application to panel data based on difference-in-differences; and 3) conditional treatment effect evaluation with debiased machine learning. Empirical examples of the teleworkability effect on unemployment indicators during the COVID‑19 period illustrate the approaches.
