CATE meets ML

Abstract For treatment effects—one of the core issues in modern econometric analysis—prediction and estimation are two sides same coin. As it turns out, machine learning methods tool for generalized prediction models. Combined with theory, they allow us to estimate not only average but a personalized effect—the conditional effect (CATE). In this tutorial, we give an overview novel methods, explain them detail, apply via Quantlets real data applications. We study that microcredit availability has on amount money borrowed if 401(k) pension plan eligibility impact net financial assets, as empirical examples. The presented toolbox contains meta-learners, like doubly-robust, R-, T- X-learner, specially designed CATE causal BART random forest. both, example, find positive all observations conflicting evidence heterogeneity. An additional simulation study, where true is known, allows compare different observe patterns similarities.