CSRDA Discussion Paper Series

No. 84 Double/debiased Machine Learning for Causal Inference on Survival Function
Daijiro Kabata, Mototsugu Shintani
Daijiro KabataOsaka Metropolitan University
Mototsugu ShintaniThe University of Tokyo
doubly robust estimatorsurvival analysispseudo-observations
Goal 8: Decent Work and Economic Growth
Japanese Panel Study of Employment Dynamics 2020

This paper discusses the use of double/debiased machine learning (DML) for estimating the average treatment effect (ATE) on a survival function using pseudo-observations. Through simulations, we demonstrate the double robustness property of our method and its improved performance, compared to existing estimators in the presence of many covariates. In our empirical example, the method is applied in evaluating the effect of the e-learning program participation on the job-finding rate among individuals who are seeking employment.