Abstract for: Financial Contagion as a Self-Exciting Point Process

Financial crises rarely appear as isolated shocks. Instead, extreme losses often cluster over time and propagate across markets, suggesting that systemic stress reflects endogenous amplification as well as exogenous news. Standard volatility-based models capture co-movement and persistence, but they do not directly model the timing and cascade structure of extreme stress events. We model extreme negative market movements as events in continuous time and study their arrival process using point process models. Homogeneous and inhomogeneous Poisson processes serve as benchmarks, while univariate and marked multivariate Hawkes processes capture self-excitation and cross-excitation. The framework is applied to daily stress events constructed U.S. Equity Sectors ETFs over two time ranges: the Global Financial Crisis (2008) and the COVID-19 episode (2020). The results show that the GFC displays stronger self-excitation and more persistent sectoral amplification, especially within the financial sector. The COVID-19 episode is characterized by a sharper but more transient intensity spike and a lower degree of endogenous persistence. Furthermore, in both windows, the multivariate Hawkes model tracks the VIX index closely and performs competitively relative to a DCC-GARCH benchmark. Overall, the results support a macro-financial interpretation of crises as self-exciting systems and illustrate how Hawkes models can provide operational monitoring tools for financial stability, including measures of endogeneity, contagion, and persistence. Structuring the paper layout with Gemini. Code for plots it has been generated with the help of Copilot.