经济统计论坛
(2026年第4期，总第41期）
报告题目：Identifying structural breaks with high-frequency data
时间：2026年6月26日，10：00-11：00
地点：A1128
报告人：孙宇澄教授 首都经济贸易大学
邀请人：崔文昊副教授
报告摘要：
We consider a continuous-time factor model to characterize prices for a large number of assets, and propose a procedure to identify common breaks in loadings of the continuous component of the observable factors. Our identification procedure utilizes the local principal component analysis of the residuals obtained from regressions of assets' intraday returns on observable factors, as a structural break can yield a pseudo latent factor in the residuals. Consequently, structural breaks can be identified within shrinking time intervals. We reveal the consistency property of the proposed identification method under some regularities. Given the identification result, we further develop a precise estimator for the locations of the breaks, and estimators for factor loadings and integrated volatility matrices, as well as demonstrating that the proposed method also enables us to determine whether there are omitted factors. We evaluate the finite sample performance of our method by Monte Carlo simulations. In real data analysis, we detect several breaks in the US stock market for the period 2015-2020, while some of the breaks may be associated with significant events. We further take advantage of the break identification result for portfolio allocation, and evaluate the performance of the resultant portfolios.
个人简介：
孙宇澄，西班牙庞培法布拉大学金融学博士，现任首都经济贸易大学国际经济管理学院教授，博士生导师。主要研究领域为计量经济学和高频金融计量，研究成果发表于Journal of Applied Econometrics, Econometric Reviews, Journal of Business & Economic Statistics, Journal of Econometrics, Journal of the American Statistical Association 以及Econometric Theory 等权威学术期刊，并主持完成国家自然科学基金青年项目。
