A1756
Title: Multidimensional asset networks and momentum spillover effects
Authors: Chao Yang - Shanghai University of Finance and Economics (China) [presenting]
Abstract: Owing to misspecification bias, it is challenging for traditional asset pricing models to capture the multidimensional network correlations of stock returns in the A-share market. Meanwhile, existing research on momentum spillover effects predominantly focuses on a single dimension of association, resulting in incomplete analysis. This study examines Chinese Shanghai and Shenzhen A-share listed companies from 2006 to 2024, introducing model averaging methods into momentum spillover effect analysis. Five types of spatial weight matrices are constructed: analyst co-coverage, geographic proximity, industry classification, technological similarity, and statistical similarity. Optimal weights are determined by minimizing the Mallows criterion and integrating multi-matrix information to form a comprehensive asset network structure. The return performance and pricing effectiveness of long-short portfolios under single-matrix versus model-averaged matrix specifications are then compared. Findings reveal significant heterogeneity in how different spatial association dimensions explain stock return differentials. Single-dimension portfolios generally exhibit weak alpha significance and time-series stability. In contrast, portfolios constructed using the model averaging method demonstrate clear quantile monotonicity characteristics, with outstanding performance in raw excess returns and cumulative excess returns.
