题目： Sequential Inference in Multiple Bubble Regimes

报告人：涂云东

报告时间：2024年12月16日  14：00

地点： 综合楼615会议室

报告人简介：

涂云东，北京大学光华管理学院和北京大学统计科学中心联席教授。入选“日出东方”北大光华青年人才，北京大学优秀博士学位论文指导教师(2017,2021,2024)，北京大学优秀研究生导师(2024)，教育部“长江学者奖励计划”青年长江学者，国家杰出青年科学基金获得者。2004年和2006年先后获武汉大学理学学士学位和经济学硕士学位，2012年获美国加州大学河滨分校经济学博士学位。亚太青年计量经济学者会议发起人和主要组织者。40余篇学术论文发表在多个国际国内知名专业杂志。著作教材《时间序列分析》由人民邮电出版社于2022年9月出版。研究领域涵盖时间序列分析、非参数计量方法、大数据分析、金融计量和预测等。

报告摘要：

This paper considers sequential inference in a nonstationary autoregressive process with cyclical structural breaks, which describes financial bubble periods consisting of the normal market period, economic exuberance and market crash. Each time series period begins with a unit root process, then transits to a mildly explosive process and eventually becomes a mildly integrated process. A two-step sequential least squares procedure is proposed to estimate the break dates.  The first step identifies all the bubble collapse dates sequentially by minimizing the squared error loss, while the second step estimates the remaining breaks in samples split by the bubble collapse date estimates. The convergence rates and asymptotic distributions of break date estimators are derived. Finite sample simulations demonstrate that our method enjoys superior detection rates for bubble periods, improved computational efficiency and estimation accuracy for break dates. The empirical analysis of the S&P 500 stock price index successfully identifies the black Monday in 1987, the dot-com bubble and the subprime mortgage crisis.