Systemic Financial Risk in China: A Complex Network Analysis Based on Returns and Sentiment

自2008年金融危机后，系统性金融风险研究受到广泛关注，遏制风险溢出、构建传播网络及识别关键节点是提升金融市场风险管理能力的关键。本研究构建了基于股票收益和投资者情绪的双层复杂网络，结合皮尔逊、肯德尔相关系数及互信息方法捕捉市场波动和投资者行为，以DCC-GJR-GARCH-CoVAR模型衡量银行、证券、保险、信托和房地产等金融机构的风险溢出效应，并通过信息熵法聚合中心性指标，揭示不同网络中中心性对系统性风险的影响。结果表明，加入情绪因素和互信息构建的网络可提高模型拟合度，更准确刻画风险传递路径，不同网络位置的公司对系统性风险的贡献存在显著差异，突显市场参与者在网络中的重要性，为风险监测和政策制定提供新视角。Since the 2008 financial crisis, systemic financial risk research has received widespread attention. Containing risk spillovers, constructing transmission networks, and identifying key nodes are crucial to enhancing the financial market’s risk management capabilities. This study constructs a dual-layer complex network based on stock returns and investor sentiment, employing Pearson correlation, Kendall correlation, and mutual information methods to capture market co-movement characteristics and investor behavior patterns. The DCC-GJR-GARCH-CoVAR model is utilized to measure the risk spillover effects among financial institutions, including banks, securities firms, insurance companies, trust institutions, and real estate firms. Additionally, centrality indicators are aggregated using the entropy method to reveal the impact of centrality in different networks on systemic risk. The results indicate that incorporating sentiment factors and constructing networks using mutual information enhances model fitting accuracy and provides a more precise depiction of risk transmission pathways. Moreover, firms occupying different network positions exhibit significant variations in their contributions to systemic risk, underscoring the importance of market participants in the network structure. These findings offer new insights for risk monitoring and policy formulation.