Matching-type estimators using the propensity score are the major workhorse in active labour market policy evaluation. This work investigates if machine learning algorithms for estimating the propensity score lead to more credible estimation of average treatment effects on the treated using a...