Automatic Debiased Machine Learning for Nonseparable Triangular Models
报告人: 范青亮(香港中文大学)
时间:2026-11-19 13:30-14:50
地点:Room 217, Guanghua Building 2
Abstract:
This paper studies inference for structural parameters in nonseparable triangular regression models identified by instrumental variables (IVs) under the control function approach of Imbens and Newey (2009), accommodating endogeneity and high-dimensional covariates in data-rich settings. Using the control function approach, structural parameters are identified through the conditional cumulative distribution function (CDF) of the treatment given covariates and IVs, which needs to be learned from data and complicates debiased inference. Building on this characterization, we show that the influence function in triangular models has two components: one correcting bias from machine learning estimation of the main regression and the other correcting bias from estimation of the conditional CDF in the first stage. We develop automatic estimators of the Riesz representers for both correction terms, which enables asymptotically unbiased inference. We illustrate the usefulness of our method in an empirical application revisiting Acemoglu et al. (2020).
About the Speaker:
Qingliang Fan is an Associate Professor of Economics at The Chinese University of Hong Kong. His current research focuses on causal inference with machine learning, and portfolio studies. His previous work was published in Econometric Theory, Journal of American Statistical Association, Journal of Business & Economics Statistics, Journal of Econometrics, Journal of the Royal Statistical Society: Series B (Statistical Methodology), Review of Economics and Statistics, and Strategic Management Journal. Some of his ongoing research projects relate to labor economics, financial economics, and management science. He is a member of the executive and postgraduate committee of the Computational Social Science Laboratory at CUHK, a fellow of Morningside College, affiliated members of Center for Population Research at Faculty of Social Science, and CUHK-Zhejiang University Joint Research Center for Digital Economy.

Your participation is warmly welcomed!

欢迎扫码关注北大统计科学中心公众号,了解更多讲座信息!