Causal Machine Learning for Heterogeneous Treatment Effect Estimation in the 2008 Oregon Health Insurance Experiment

Public health insurance programs impact individuals differently based on their health status, demographics, and socioeconomic background, yet most studies focus on average treatment effects (ATE) rather than heterogeneous treatment effects (HTE). This research expands upon the 2008 Oregon Health Insurance Experiment (OHIE) by estimating HTE to analyze how Medicaid’s benefits vary across different subpopulations. Using data from the OHIE, we apply causal machine learning techniques, including meta-learners (T-Learner, S-Learner, X-Learner), Double Machine Learning (DML), and Causal Forests, to identify subgroup-specific effects of Medicaid coverage. The study begins by replicating the original ATE findings as a baseline, followed by the application of modern causal inference methods to uncover disparities in Medicaid’s impact across different groups. Our analysis considers preexisting health conditions, income, and healthcare access, revealing which populations benefit most (or least) from Medicaid expansion. By identifying these variations, we provide insights that can inform more equitable public health policies and improve resource allocation strategies. This research also contributes to the broader field of causal inference in health economics, demonstrating how advanced machine learning techniques can refine treatment effect estimation in randomized controlled trials (RCTs).

Recommended citation: Your Name, You. (2010). "Paper Title Number 2." Journal 1. 1(2).
Download Paper | Download Slides