主讲人Speaker: 蔡宗武   教授   美国堪萨斯大学

时间Date & Time: 2025年6月30日（周一），下午14:00--15:30

地点Venue:粤海校区汇星楼565会议室

内容简介/ Abstract:

First, we study the investor sentiment based on online information and propose constructing two investor sentiment indices based on Internet Search Queries, using linear model approaches such as partial least squares (PLS) and LASSO methods, respectively. By examining the relationship between investor sentiment and stock risk premium on overall market level, we find that these sentiment indices have predictive power for both in and out of samples. However, online investor sentiment proxies are generally high-dimensional and nonlinear. Conventionally, we need to run forecasting in two steps: build a low-dimensional index and do a prediction. With a help of machine learning methods such as random forest and XGBoost, prediction can be done in one step to overcome the above difficulties. The results in our empirical study help us have a better understanding of the nonlinearity between online investor sentiment and future stock return. Finally, we propose using a nonparametric generalized method of moment (NPGMM) as in Cai and Li (2008) to estimate the portfolio choices using an investment sentiment index for choosing and updating optimal asset allocations. The empirical findings show that the market timing depending on investor sentiment is nonlinear and varies across assets.

主讲人介绍/Biography of the speaker：

蔡宗武，美国堪萨斯大学经济系计量经济学Charles Oswald Distinguished Professor（杰出或者资深教授）。1982 年取得中国地质大学（武汉）数学学士学位，1988年获得杭州大学（现为浙江大学）统计学硕士学位，1995 年获得美国加州大学戴维斯校区统计学博士学位。主要研究领域包含理论和应用计量经济学、宏观计量经济学、微观计量经济学、经济分析和政策评估、金融计量学、金融与经济大数据、风险管理、非线性和非平稳时间序列建模和检验、非参数函数估计和检验，以及大数据分析与建模等多个领域。蔡宗武教授曾任“中国留美经济学会”会长（2018 年 9 月-2019 年 8 月）和理事长（2020 年 1 月-2020 年 12 月）。蔡宗武教授还是多家国际一流经济学、数据科学、统计学、金融学期刊的副主编，同时也是美国统计协会 Fellow 和《Journal of Econometrics》的 Fellow。他在国际计量经济学、统计学以及数据科学领域有很高的影响力。在计量经济学和统计学领域，取得许多富有创新性的研究成果，具有国际领先地位的学术造诣。在国际顶尖级的经济学与统计学以及金融学等期刊上发表了论文 140 多篇。