A1960
Title: Mapping risk contagion through graphical models and early warning signals in Taiwan's stock market
Authors: Huei-Wen Teng - National Yang Ming Chiao Tung University (Taiwan) [presenting]
Nok Hang Chan - National Yang Ming Chiao Tung University (Hong Kong)
Po-Yao Hsu - National Yang Ming Chiao Tung University (Taiwan)
Abstract: Graphical models are leveraged to investigate financial risk in Taiwan's stock market, focusing on the dynamic interactions among the 30 largest market-capitalization firms from January 1985 to April 2025. Granger causality networks are constructed to capture directional predictability among stock returns and track how the resulting network structure evolves over four decades of market history. Analysis reveals a consistent pattern in the dynamics of network centrality over time, aligning with the widely held view that technology firms dominate the Taiwanese market. By measuring network connectivity through degree of causality, number of connections, closeness, and eigenvector centrality, an early warning mechanism is proposed for identifying periods of heightened systemic risk. Empirical results show that sharp increases in aggregate connectivity tend to precede major market downturns, providing a forward-looking signal of contagion risk. The findings offer practical implications for institutional risk management and individual investment strategies, and demonstrate the value of graphical models as a tool for financial risk monitoring in an emerging-market setting.
