This is a past event
The seminar will be held via Teams on November 6th from 3pm - 4:15pm (contact: bs-research@abdn.ac.uk for link).
Join Dr Murat Mazibas, Senior Lecturer in Banking and Finance at the University of Dundee.
Abstract:
This study explores the evolution of style analysis in fund management, advancing from Sharpe's foundational 1992 framework to a more nuanced approach that accommodates the complexities of modern investment strategies. While Sharpe's method was instrumental in deciphering mutual fund styles, its application was limited to long-only portfolios. Recognising the limitations in addressing strategies that incorporate leverage, short positions, and cash reserves, Agarwal & Naik (2002) expanded this framework to include a broader range of investment strategies. Despite these advancements, existing models still fail to fully capture the intricate dynamics of style factors, particularly their significant correlations and nonlinear relationships with portfolio returns.
Addressing these gaps, our research introduces a sophisticated dynamic framework that integrates Machine Learning (ML) algorithms to unravel the complex interplay between style factors and portfolio performance. This approach transcends traditional linear regression, employing advanced ML algorithms to explore potential nonlinear interactions and multicollinearity among style factors. Furthermore, we innovate by applying Shapley values to quantify each factor's contribution to portfolio returns, enhancing the fairness and accuracy of performance attribution. Additionally, we leverage the Local Interpretable Model-agnostic Explanations (LIME) method to improve the interpretability of our ML-based style analysis, offering clear insights into the driving forces behind fund performance.
Our paper makes significant contributions by employing ML in style analysis to reveal the true investment style of funds with complex strategies, marking the first integration of Shapley values for equitable return attribution among style factors. This novel combination of ML algorithms, Shapley values, and LIME methodology marks a substantial advancement in style analysis, providing a deeper, more accurate, and comprehensible evaluation of investment styles amid the diversity of complex investment strategies. Through this research, we offer a rigorous and equitable framework that aids in strategy assessment, risk management, and informed decision-making, enhancing our understanding of the intricate relationships between style factors and fund performance.
- Speaker
- Dr Murat Mazibas
- Hosted by
- University of Aberdeen Business School