Causal Inference with Random Forests

Many scientific and engineering challenges—ranging from personalized medicine to customized marketing recommendations—require an understanding of treatment heterogeneity. We develop a non-parametric causal forest for estimating heterogeneous treatment effects that is closely inspired by Breiman’s widely used random forest algorithm. Given a potential outcomes framework with unconfoundedness, we show that causal forests are pointwise consistent for the true treatment effect, and have an asymptotically Gaussian and centered sampling distribution. We also propose a practical estimator for the asymptotic variance of causal forests. In both simulations and an empirical application, we find causal forests to be substantially more powerful than classical methods based on nearest-neighbor matching, especially as the number of covariates increases. Our theoretical results rely on a generic asymptotic normality theory for a large family of random forest algorithms. To our knowledge, this is the first set of results that allows random forests, including classification and regression forests, to be used for valid statistical inference. This talk is based on joint work with Susan Athey, Bradley Efron, and Trevor Hastie.