Double Robust, Flexible Adjustment Methods for Causal Inference: An Overview and an Evaluation

Double robust methods for flexible covariate adjustment in causal inference have proliferated in recent years. Despite their apparent advantages, these methods are rarely used by social scientists. It is also unclear whether these methods actually outperform more traditional methods in finite samples. This paper has two aims: It is a guide to some of the latest methods in double robust, flexible covariate adjustment using machine learning, and it compares these methods to more traditional statistical methods and flexible "single robust" methods. It does this by using both simulated data where the treatment effect estimate is known, and then by replicating some prominent articles in sociology that use simpler methods. Double robust methods covered include Augmented Inverse Probability Weighting (AIPW), Targeted Maximum Likelihood Estimation (TMLE), and Double/Debiased Machine Learning (DML). Results suggest that some of these methods do outperform traditional methods in a wide range of simulations, but only slightly. In particular, the top performers are TMLE and AIPW in conjunction with flexible machine learning estimators. But G-computation with the same flexible estimators obtains almost identical results, and simple regression methods have only slightly higher bias and are much more computationally efficient. In the article replications, the application of double robust methods substantively changes some, but not all, of the results, highlighting the importance of comparing performance from multiple estimators.