Exploiting Discovered Regression Discontinuities to Debias Conditioned-on-observable Estimators
Benjamin Jakubowski, Sriram Somanchi, Edward McFowland III, Daniel B. Neill; 24(133):1−57, 2023.
Abstract
This study introduces a novel method for addressing the limitations of standard approaches to regression discontinuity (RD) analysis. RD designs are commonly used to estimate causal effects in situations where randomized experiments are not feasible. However, existing approaches require prior knowledge of treatment discontinuities and only provide local treatment effect estimates near the discontinuity. To overcome these limitations, the proposed method automatically detects RDs at scale, integrates information from multiple discovered discontinuities with an observational estimator, and extrapolates causal effect estimates away from the discovered local RDs. The performance of the method is demonstrated on two synthetic datasets, showing improved results compared to other methods. Additionally, the method is applied to estimate spatially heterogeneous treatment effects in the context of a recent economic development problem.
[abs]