Double machine learning-based programme evaluation under unconfoundedness

Summary This paper reviews, applies, and extends recently proposed methods based on double machine learning (DML) with a focus programme evaluation under unconfoundedness. DML-based leverage flexible prediction models to adjust for confounding variables in the estimation of (a) standard average effects, (b) different forms heterogeneous (c) optimal treatment assignment rules. An multiple programmes Swiss Active Labour Market Policy illustrates how enable comprehensive evaluation. Motivated by extreme individualised effect estimates DR-learner, we propose normalised DR-learner (NDR-learner) address this issue. The NDR-learner acknowledges that can be stabilised an normalisation inverse probability weights.