题　目：Harness Machine Learning with Carry

嘉　宾：施展 副教授 清华大学

主持人：张澄宇 助理教授 上海交通大学安泰经济与管理学院

时　间：2025年6月20日（周五）13:30-15:00

地　点：上海交通大学 徐汇校区安泰浩然楼306室

内容简介:

This paper proposes a hybrid machine learning (ML) framework for fixed-income securities, by integrating ML-based forecasts of price returns with bond carry to form composite expected return estimates. Using a comprehensive dataset of corporate bonds over the 1992–2019 period, we illustrate that the proposed ML estimators deliver superior forecasting accuracy and dominate conventional ML models applied directly to total returns. This dichotomy highlights bond markets’ efficiency in pricing carry-related risks and ML’s incremental value in capturing deviations from yield-implied expectations. High-yield bonds exhibit the greatest return predictability, driven by nonlinear credit risk and liquidity shocks, whereas carry strategies remain superior for bonds issued by unlisted firms. Portfolio tests reveal that long-short strategies based on pure ML approaches generate gross monthly alphas of 1.74% per month, but transaction costs erode 60–80% of gains due to high turnover. In contrast, carry-driven ML strategies sustain net returns of 1.15% with lower turnover. Our results highlight the necessity of integrating asset-class-specific structure into ML frameworks and caution against direct adoption of equity-style strategies in fixed income.

演讲人简介：

 施展博士现任清华大学五道口金融学院长聘副教授，中国保险与养老金研究中心副主任。施展于2014年毕业于美国宾夕法尼亚州立大学Smeal商学院，获得金融学博士学位。施展的研究领域是固定收益，市场微观结构，动态公司金融、金融科技。他的论文多次发表在 Journal of Finance、Journal of Financial Economics、Management Science、Review of Finance和《经济研究》、《金融研究》等国内外顶级金融期刊上，他的研究成果曾获得多个国内外学术奖项，包括Western Finance Association 2014年年会的最佳博士生论文奖、中国金融学术研究网年会最佳论文奖和第九届高等学校科学研究青年成果奖（人文社会科学）。

欢迎广大师生参加!