Evaluating (weighted) dynamic treatment effects by double machine learning
Authors: Hugo Bodory, Martin Huber, and Lukas Laers
Abstract: We consider evaluating the causal eects of dynamic treatments, i.e. of multiple treatment sequences in various periods, based on double machine learning to control for observed, time-varying covariates in a data-driven way under a selection-on-observables assumption. To this end, we make use of so-called Neyman-orthogonal score functions, which imply the robustness of treatment eect estimation to moderate (local) misspecications of the dynamic outcome and treatment models. This robustness property permits approximating outcome and treatment models by double machine learning even under high dimensional covariates and is combined with data splitting to prevent overtting. In addition to eect estimation for the total population, we consider weighted estimation that permits assessing dynamic treatment eects in specic subgroups, e.g. among those treated in the rst treatment period. We demonstrate that the estimators are asymptotically normal and √n-consistent under specic regularity conditions and investigate their nite sample properties in a simulation study. Finally, we apply the methods to the Job Corps study in order to assess dierent sequences of training programs under a large set of covariates.