نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشجوی دکتری گروه اقتصاد، دانشکده مدیریت و اقتصاد، واحد علوم و تحقیقات، دانشگاه آزاد اسلامی
2 استاد گروه اقتصاد، دانشکده اقتصاد دانشگاه شهید بهشتی
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
This study proposes a novel framework for identifying and exploiting lead–lag relationships among stocks in the Iranian capital market. Leveraging the theory of path signature and Lévy area computation, dynamic lead–lag matrices were constructed over rolling time windows. Subsequently, Hermitian clustering was employed to uncover the evolving leader–follower structure within the market. Based on this structure, three portfolio strategies were developed: a global leader–follower portfolio (GP), a cluster-based portfolio (CP), and an aggregated cluster portfolio (GCP). The strategies were tested on daily price data of selected stocks from major industries over a multi-year horizon. Empirical results reveal that the GCP strategy achieved an annual return of 6% with a Sharpe ratio of 0.33, outperforming traditional methods such as Granger causality and CP. Notably, the lagged cross-correlation (CCF) approach, despite its simplicity, yielded a strong annual return of 11.5% and a Sharpe ratio of 0.61, surpassing all classical models. Furthermore, an industry-weighted portfolio based on aggregated inter-industry lead–lag flows outperformed all strategies, with an impressive annual return of 31.3% and a Sharpe ratio of 1.51. These findings indicate that temporal and cross-symbol dependencies hold valuable predictive information for portfolio design and market timing. Moreover, the integration of intra-industry and inter-industry analyses provides additional insight into the structural flow of information within the market. The proposed framework offers a robust foundation for dynamic portfolio optimization and highlights the value of modern mathematical tools in financial modeling.
کلیدواژهها [English]
)