报告题目：Consistent Testing for Structural Changes in Factor Models

报告时间：2025年7月12日下午14；00

报告地点：北湖东校区澳门博彩在线 新楼216室

主办单位：澳门博彩在线 /科研处

报告人：洪永淼

报告人简介：洪永淼，中国科澳门博彩在线 数学与系统科学研究院冠名首席研究员、中国科澳门博彩在线 大学经济与管理澳门博彩在线 院长、发展中国家科澳门博彩在线 院士、计量经济学会会士，并担任中国教育部高等学校经济学类专业教学指导委员会副主任委员。曾任康奈尔大学ErnestS.Liu经济学与国际研究讲席教授，北美华人经济学家学会会长。研究领域涵盖计量经济学理论、时间序列计量经济学、金融计量经济学及统计学。其学术成果发表于Annals of Statistics、Biometrika、Econometric Theory、Econometrica、 International Economic Review、Journal of American Statistical Association、Journal of Businessand Economic Statistics、 Journal of Econometrics、Journal of Political Economy、Journal of Royal Statistical Society (Series B)等国际知名经济、金融与统计期刊。其最新英文专著《现代计量经济学基础:统一框架》广受学界关注。2014至2024年，连续11年入选爱思唯尔“中国高被引学者(经济学/统计学)”榜单，并于2022年荣获国家级教学成果奖(高等教育本科)一等奖。

报告摘要：We introduce a novel consistent test for detecting structural changes in high-dimensional factor models, leveraging a discrete Fourier transform (DFT) approach. In scenarios where structural changes take place, the ability of conventional principal component analysis to accurately estimate common factors and factor loadings is compromised, leading to estimated residuals that embed signals of these changes. Therefore, we are equipped to assess the DFTs of estimated residuals against the null zero spectrum, which assumes no structural changes. The proposed test is consistent against a wide range of both smooth structural changes and abrupt structural breaks with a possibly unknown number of breaks and unknown break dates in time-varying factor loadings. Moreover, the test has an asymptotic N(0, 1) distribution under the null hypothesis, leading to a simple and convenient inference procedure.

 Monte Carlo simulations confirm the test's reliability in terms of size and its exceptional power against a variety of structural changes in factor loadings. When applied to China's macroeconomic data, the proposed test reveals substantial and consistent time-varying factor loadings in the periods following a series of significant historical events, providing insights that may have been missed in prior analyses.