北航经管学院“融实论坛”系列讲座
（2024年第7期，总第64期）
讲座题目：Growing the Efficient Frontier on Panel Trees
讲座时间：2024.7.8（周一）14:30-16:00
讲座嘉宾：冯冠豪 副教授
讲座形式：线下新主楼 A1148
主持人：部慧 副教授
讲座嘉宾 简介
冯冠豪博士目前是香港城市大学商学院统计副教授，他于2017年从芝加哥大学布斯商学院获得博士学位。冯冠豪博士专注于开发新的实证方法，包括机器学习、贝叶斯统计和金融计量，以解决实证资产定价中的大数据问题。他的研究成果已发表在Journal of Finance, Journal of Financial and Quantitative Analysis, Journal of Econometrics, International Economic Review等期刊上。冯教授同时主持多项研究基金，包括香港研究资助局的ECS和GRF基金，以及国家自然科学青年科学家基金。他的研究得到了金融业界的认可，包括INQUIRE Europe和香港货币金融研究所的研究奖项，还有AQR Insight Award。
讲座概要
Estimating the efficient frontier using individual stock returns is a long-standing empirical challenge. We develop a new class of tree-based models (P-Trees) for analyzing (unbalanced) panel data, which grows top-down to split the cross section of asset returns for constructing characteristics-managed test assets and recovering the stochastic discount factor under the global criteria of mean-variance efficiency. When applied to U.S. equities, boosted (multi-factor) P-Trees significantly advance the efficient frontier relative to those constructed with commonly used factors and test assets. Second, P-Tree test assets are diversified and exhibit significant unexplained alphas against benchmark models. Third, the unified P-Tree factors outperform popular observable and latent factor models in pricing cross-sectional returns and generating substantial investment gains. Finally, an over-parameterized Random P-Forest model delivers an exceptional out-of-sample Sharpe ratio, revealing the complexity of panel stock returns. Beyond asset pricing, our framework offers a sparse, interpretable, and computationally efficient alternative to recent machine learning and AI models for analyzing panel data through goal-oriented clustering.
