【1149期】 5月31日计量经济学学术讲座：Estimation and Inference of Counterfactual CDF in a High-dimension Framework: With An Application to OB Decomposition（蔡俊，讲师，华中科技大学）

【主题】Estimation and Inference of Counterfactual CDF in a High-dimension Framework: With An Application to OB Decomposition

【报告人】蔡俊（讲师，华中科技大学管理学院）

【时间】2023年5月31日周三10:00-11:30

【地点】经济学院401会议室

【语言】中文

【摘要】Counterfactual CDF estimation and inference is the foundation for distribution effect analysis, ATE and QTE. High dimensional covariates help justify the unconfoundedness assumption for causal inference and alleviate the concerns of endogeneity resulted from omitted variables. This paper considers estimation and inference of counterfactual CDF in a high-dimension framework with an application to Oaxaca-Blinder decomposition. We propose two semi-parameric estimators: double-machine learning estimator and propensity score double debias estimator on the counterfactual CDFs. Asymptotics are derived for our proposed estimators which are prove to be equivalent in the large samples. We demonstrate the good finite sample properties of proposed estimators by Monte Carlo simulations and compared them with existing estimators. We also apply the proposed method to CHIP2018 data on discriminations of hukou system and gender in the Chinese labor market, which yields new insights with consideration of high dimensional covariates.