题目：A Dynamic Semiparametric Characteristics-based Model for Portfolio Selection

主讲人：李少然，北京大学

时间：2024年11月8日（星期五）上午10:00-11:30

地点：暨南大学（石牌校区）中惠楼106会议室

主讲人简介：

李少然，北京大学经济学院金融系助理教授，2021年于剑桥大学取得博士学位，研究领域为金融计量，实证资产定价和机器学习。研究成果发表在Journal of Econometrics, Journal of Business & Economic Statistics 和 Journal of the Royal Statistical Society: Series A (Statistics in Society)上。

Abstract：

This paper introduces a two-step methodology for portfolio weight selection linked to a characteristics-based factor model with time-varying factor risk premia. This portfolio management strategy extends the expected utility maximization framework to accommodate a large number of assets and to integrate the insight from an asset pricing model. The first step, termed factor tilting, finds flexible factor-mimicking sub-portfolios through linear combinations of characteristics- based factor loadings. The second step, factor timing, dynamically combines these factor-mimicking sub-portfolios based on the predictability of factor premia by a time-varying signal – a single index function of state variables. We develop a two-stage semiparametric estimator and perform hypothesis tests on the significance of state variables. Applying our methodology and theories to data from the Center for Research in Security Prices (CRSP) and the Federal Reserve Economic Data (FRED), we identify significant roles for some dynamic predictors, which yield excellent in-sample and out-of-sample performance consistent with varying levels of investor risk aversion.