A0183
Title: Comprehensive causal machine learning
Authors: Michael Lechner - University St Gallen (Switzerland) [presenting]
Jana Mareckova - University of Konstanz (Germany)
Abstract: Uncovering the heterogeneity of causal effects at various levels of granularity provides substantial value to decision-makers. Comprehensive approaches to causal effect estimation allow the use of a single causal machine learning approach for the estimation and inference of causal mean effects for all levels of granularity. Focussing on selection-on-observables, the theoretical asymptotic guarantees for one such approach, the modified causal forest (mcf), are provided. The asymptotic and finite sample properties of the mcf to the generalized random forest (rf) and double machine learning (DML) are also compared. The findings indicate that dml-based methods excel for average treatment effects at the population level (ATE) and group level (GATE) with few groups. However, for finer causal heterogeneity, explicitly outcome-centred forest-based approaches are superior. The mcf has three additional benefits: (i) It is the most robust estimator in cases when dml-based approaches underperform because of substantial selectivity; (ii) it is the best estimator for GATEs when the number of groups gets larger; and (iii), it is the only estimator that is internally consistent, in the sense that low-dimen-sional causal ATEs and GATEs are obtained as aggregates of finer-grained causal parameters.
