A Graph Neural Network for Comparing the Propagation of Stock Market Shocks

Understanding the dynamics of stock market shocks is critical for financial stability and strategic risk management. This paper introduces a novel approach to analyze the propagation mechanisms of stock market shocks within the S&P 500 Index, utilizing a Temporal Graph Convolutional Network (TGCN) model. Our study aims to identify the main factors in the shock propagation, the most affected companies, the order of impact, and the sectorial influence. Utilizing a detailed dataset of closing prices for 498 entities, including all 497 companies in the S&P 500 Index and the index itself, and incorporating data from identified shock periods, the TGCN model effectively captures dynamic inter-company relationships. The analysis compares network structures before and after model training and evaluates company characteristics such as linkage degree and centrality measures. Key findings from our research reveal varying degrees of impact across different sectors and companies, highlighting patterns of interconnectivity and centrality changes during shock periods. By analyzing three distinct periods of stock market shocks, we provide a comparative perspective on how various crises affect financial markets differently. Our results indicate significant implications for investors, risk managers, and researchers in understanding and navigating the economic landscape during turbulent times. The study emphasizes the importance of resilience, adaptability, and strategic risk management in mitigating the effects of such shocks. Overall, this paper offers a comprehensive view of stock market shock propagation, contributing valuable insights to the field of financial analysis and risk management.