Identification of Treatment Effect Heterogeneity as a Variable Selection Problem

 

  Abstract

Identification of treatment effect heterogeneity plays an essential role in a number of situations frequently encountered by applied researchers. They include (1) selecting the most effective treatment from a large number of available treatments, (2) ascertaining subpopulations for which a treatment is effective or harmful, (3) designing individualized optimal treatment regimes, (4) testing for the existence or lack of heterogeneous treatment effects, and (5) generalizing causal effect estimates obtained from an experimental sample to a target population. In this paper, we formulate the identification of heterogeneous treatment effects as a variable selection problem. We then propose a method that adapts the Support Vector Machine classifier by placing separate sparsity constraints over the pre-treatment parameters and causal heterogeneity parameters of interest. To fit the proposed model with multiple regularization parameters, we develop a computationally efficient algorithm based on a generalized cross-validation statistic. Our simulation studies show that the proposed method tends to yield a lower false discovery rate than some commonly used alternatives. For empirical illustrations, we apply the proposed method to well-known randomized field experiments from the social sciences. (Last revised, December 2011).
You may also be interested in Imai, Kosuke and Aaron Strauss. (2011). ``Estimation of Heterogeneous Treatment Effects from Randomized Experiments, with Application to the Optimal Planning of the Get-out-the-vote Campaign..'' Political Analysis, Vol. 19, No. 1 (Winter), pp. 1-19.

© Kosuke Imai
  Last modified: Fri Aug 13 17:18:17 EDT 2010