Seminars

Abstract: The Regression Discontinuity (RD) design is often employed to study treatment assigned by policy cutoffs. However, the policy relevance of the treatment effect around the cutoff is limited since these cutoffs are often created by policymakers given the treatment effect heterogeneity along the running variable. In this paper, we shift the focus to how the treatment effect heterogeneity around the cutoff can help improve the selectiveness of the policy. We show that variables that capture the most variation of the treatment effect would provide the largest overall benefit by selecting people who benefit the most. This result is then applied to the health benefits of hyperlipidemia diagnosis assigned by the cholesterol reference range. Using administrative data from 6 million health checks in Taiwan, we find that the diagnosis reduces the short-term risk of complications like strokes and heart failures by 0.156 percentage points (14.5%). Treatment effect heterogeneity is systematically examined with causal forest estimators. Age stands out among a rich set of variables: for the oldest 20%, the health benefit is 3.900 times stronger. As age captures the largest share of the variance in health benefits, including age in a new reference range alongside cholesterol level would bring the most health benefit among participants.