A co-jump network approach to systemic risk measurement: Evidence from the U.S. financial market

This study proposes a co-jump network framework for measuring systemic risk by more accurately capturing latent interconnectedness among financial institutions. To address key limitations of traditional network connectedness measures, namely reliance on nonpublic data, a focus on one-sided tail dependence, and insensitivity to discontinuous shocks, we estimate a Tobit-LASSO model to identify significant co-jump dependence after controlling for macro common factors and institution-level characteristics. We then establish the co-jump network and define a realized systemic jump beta as the risk metric. Empirical results based on U.S. financial data show that the proposed method outperforms benchmark models, including the CoVaR network approach and MES (DCC-GARCH based), in capturing the structural shift in systemic risk during the post-COVID period, when the persistent effects of the pandemic shock were not fully offset by the short-term market rebound. The analysis also highlights liquidity providers and highly leveraged intermediaries as primary contributors to systemic risk, while insurers are comparatively underrepresented. Furthermore, a mean–variance strategy augmented with our systemic risk measure is shown to help avoid extreme downside losses. Overall, the co-jump network offers a theoretically grounded and empirically practical tool for risk management that benefits both regulators and investors.