Metalearners for estimating heterogeneous treatment effects using machine learning

Significance Estimating and analyzing heterogeneous treatment effects is timely, yet challenging. We introduce a unifying framework for many conditional average treatment effect estimators, and we propose a metalearner, the X-learner, which can adapt to structural properties, such as the smoothness and sparsity of the underlying treatment effect. We present its favorable properties, using theory and simulations. We apply it, using random forests, to two field experiments in political science, where it is shown to be easy to use and to produce results that are interpretable. There is growing interest in estimating and analyzing heterogeneous treatment effects in experimental and observational studies. We describe a number of metaalgorithms that can take advantage of any supervised learning or regression method in machine learning and statistics to estimate the conditional average treatment effect (CATE) function. Metaalgorithms build on base algorithms—such as random forests (RFs), Bayesian additive regression trees (BARTs), or neural networks—to estimate the CATE, a function that the base algorithms are not designed to estimate directly. We introduce a metaalgorithm, the X-learner, that is provably efficient when the number of units in one treatment group is much larger than in the other and can exploit structural properties of the CATE function. For example, if the CATE function is linear and the response functions in treatment and control are Lipschitz-continuous, the X-learner can still achieve the parametric rate under regularity conditions. We then introduce versions of the X-learner that use RF and BART as base learners. In extensive simulation studies, the X-learner performs favorably, although none of the metalearners is uniformly the best. In two persuasion field experiments from political science, we demonstrate how our X-learner can be used to target treatment regimes and to shed light on underlying mechanisms. A software package is provided that implements our methods.

[1]  R. F.,et al.  Mathematical Statistics , 1944, Nature.

[2]  L. L. Cam,et al.  On the Asymptotic Theory of Estimation and Testing Hypotheses , 1956 .

[3]  Jaroslav Hájek,et al.  On basic concepts of statistics , 1967 .

[4]  D. Rubin.  Estimating causal effects of treatments in randomized and nonrandomized studies. , 1974 .

[5]  P. Bickel,et al.  Mathematical Statistics: Basic Ideas and Selected Topics , 1977 .

[6]  C. J. Stone,et al.  Optimal Global Rates of Convergence for Nonparametric Regression , 1982 .

[7]  Lucien Birgé.  Approximation dans les espaces métriques et théorie de l'estimation , 1983 .

[8]  D. Rubin,et al.  The central role of the propensity score in observational studies for causal effects , 1983 .

[9]  P. Robinson.  ROOT-N-CONSISTENT SEMIPARAMETRIC REGRESSION , 1988 .

[10]  T. Speed,et al.  On the Application of Probability Theory to Agricultural Experiments. Essay on Principles. Section 9 , 1990 .

[11]  R. Pieters,et al.  Working Paper , 1994 .

[12]  J. Heckman,et al.  Making the Most out of Programme Evaluations and Social Experiments: Accounting for Heterogeneity in Programme Impacts , 1997 .

[13]  Adam Krzyzak,et al.  A Distribution-Free Theory of Nonparametric Regression , 2002, Springer series in statistics.

[14]  Carolyn Pillers Dobler.  Mathematical Statistics: Basic Ideas and Selected Topics , 2002 .

[15]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[16]  Christopher W. Larimer,et al.  Social Pressure and Voter Turnout: Evidence from a Large-Scale Field Experiment , 2008, American Political Science Review.

[17]  Jake Bowers,et al.  Attributing Effects to a Cluster-Randomized Get-Out-the-Vote Campaign , 2009 .

[18]  Alexandre B. Tsybakov,et al.  Introduction to Nonparametric Estimation , 2008, Springer series in statistics.

[19]  Dorota Kurowicka,et al.  Generating random correlation matrices based on vines and extended onion method , 2009, J. Multivar. Anal..

[20]  Christopher B. Mann,et al.  Is There Backlash to Social Pressure? A Large-scale Field Experiment on Voter Mobilization , 2010 .

[21]  H. Chipman,et al.  BART: Bayesian Additive Regression Trees , 2008, 0806.3286.

[22]  Jennifer L. Hill,et al.  Bayesian Nonparametric Modeling for Causal Inference , 2011 .

[23]  D. Green,et al.  Modeling Heterogeneous Treatment Effects in Survey Experiments with Bayesian Additive Regression Trees , 2012 .

[24]  B. Efron.  Estimation and Accuracy After Model Selection , 2014, Journal of the American Statistical Association.

[25]  Lu Tian,et al.  A Simple Method for Detecting Interactions between a Treatment and a Large Number of Covariates , 2012, 1212.2995.

[26]  Jean-Philippe Vert,et al.  Consistency of Random Forests , 2014, 1405.2881.

[27]  G. Imbens,et al.  Machine Learning Methods for Estimating Heterogeneous Causal Eects , 2015 .

[28]  Susan Athey,et al.  Recursive partitioning for heterogeneous causal effects , 2015, Proceedings of the National Academy of Sciences.

[29]  Melissa R. Michelson.  The Risk of Over-Reliance on the Institutional Review Board: An Approved Project Is Not Always an Ethical Project , 2016, PS: Political Science & Politics.

[30]  For real this time: Talking to people about gay and transgender issues can change their prejudices , 2016 .

[31]  W. Newey,et al.  Double machine learning for treatment and causal parameters , 2016 .

[32]  David E. Broockman,et al.  Durably reducing transphobia: A field experiment on door-to-door canvassing , 2016, Science.

[33]  P. Richard Hahn,et al.  Bayesian Regression Tree Models for Causal Inference: Regularization, Confounding, and Heterogeneous Effects , 2017, 1706.09523.

[34]  David E. Broockman,et al.  The Minimal Persuasive Effects of Campaign Contact in General Elections: Evidence from 49 Field Experiments , 2017, American Political Science Review.

[35]  The Design of Field Experiments With Survey Outcomes: A Framework for Selecting More Efficient, Robust, and Ethical Designs , 2017, Political Analysis.

[36]  Xinkun Nie,et al.  Learning Objectives for Treatment Effect Estimation , 2017 .

[37]  Trevor Hastie,et al.  Some methods for heterogeneous treatment effect estimation in high dimensions , 2017, Statistics in medicine.

[38]  Stefan Wager,et al.  Estimation and Inference of Heterogeneous Treatment Effects using Random Forests , 2015, Journal of the American Statistical Association.

[39]  Jennifer Hill,et al.  Automated versus Do-It-Yourself Methods for Causal Inference: Lessons Learned from a Data Analysis Competition , 2017, Statistical Science.

[40]  Jasjeet S. Sekhon,et al.  Inference on a New Class of Sample Average Treatment Effects , 2017, Journal of the American Statistical Association.