Enhancing index-tracking performance: Leveraging characteristic-based factor models for reduced estimation errors
提出一个结合特征因子模型的混合整数优化框架，通过减少估计误差来降低指数追踪的跟踪误差，实证表明该方法优于传统线性与二次规划方法。
This paper addresses the challenge of minimizing tracking error in passive portfolio management by reducing estimation errors commonly encountered in traditional optimization methods. We introduce an innovative cardinality-constrained mixed-integer optimization framework that incorporates characteristic-based factor models to enhance index-tracking performance. By leveraging these models, our approach aims to minimize errors stemming from estimation uncertainty. In an empirical analysis, we benchmark the tracking errors of our approach against traditional methods, examining both linear and quadratic programs. We further evaluate robustness across various stock market indices, time periods, solvers, and transaction costs. The results indicate that our method consistently reduces estimation errors, achieving superior tracking performance relative to conventional techniques. These findings provide crucial guidance for efficiently optimizing index-tracking portfolios while accommodating practical constraints. • New optimization framework reduces tracking error by minimizing estimation uncertainty. • Characteristics-based factor models outperform traditional methods in tracking portfolios. • Linear formulations are preferred for simplicity and stability in index-tracking applications. • Consistent performance across indices, time periods, and transaction cost scenarios. • Optimal factor model specifications minimize complexity while maximizing tracking accuracy.