学术报告：付连艳 辽宁大学

主讲人：付连艳 辽宁大学

题 目： Covariate-Adjusted Nearest Neighbor Matching with a Diverging Number in ReGressi-on Discontinuity Designs

时间：2025年10月18日 15:50-17:20

地点：VSport体育官网新校园 A509

摘要：This paper proposes nearest-neighbor matching-based estimators for Regression Discontinuity Design. The method allows the number of matches M to diverge with the sample size n, and strictly performs covariate adjustment within the optimal bandwidth. In contrast to conventional RD estimators that employ local linear  regression, this nonparametric covariate adjustment avoids assumptions about functional forms. Therefore, the proposed method yields more robust estimation results when covariates exhibit complex nonlinearity, high dimensionality, or missing data. Meanwhile, we demonstrate that the bias-corrected matching estimators derived within the RD framework possess consistency, and the fuzzy regression discontinuity estimator exhibits double robustness. These estimators are semiparametrically efficient if the density functions are sufficiently smooth and the outcome model is consistently estimated. This framework also admits extensions to double machine learning for accommodating high-dimensional data. Furthermore, the simulation and empirical results demonstrate that the method proposed in this paper performs well.