Abstract:

This paper investigates volatility spillovers, contagion dynamics, and systemic risk across global financial markets using a hybrid framework that combines econometric modeling, machine learning, and network analysis. Using daily data on equities, bonds, foreign exchange, commodities, and cryptocurrencies from 2005 to 2025, we analyze how systemic linkages evolve across asset classes and crisis periods. Results from DCC-GARCH models and the Diebold–Yilmaz spillover index show that volatility transmission is strongly regime-dependent, with spillovers intensifying during the Global Financial Crisis, the European Sovereign Debt Crisis, and the COVID-19 pandemic. Machine learning models, particularly Long Short-Term Memory (LSTM) networks, outperform traditional econometric approaches in forecasting volatility during periods of severe market stress, while Random Forest models identify monetary policy shocks, oil market volatility, and cryptocurrency crashes as key systemic drivers. Network analysis reveals a structural transformation in global systemic risk: while U.S. and European equity markets remain central contagion hubs, cryptocurrencies and DeFi tokens have emerged as net transmitters of volatility in the post-pandemic era. Overall, these findings challenge traditional diversification strategies, highlight the growing systemic relevance of digital assets, and demonstrate the value of hybrid econometric AI network approaches for financial stability analysis.

Keywords:

Volatility spillovers, Systemic risk, Cryptocurrencies, Machine learning, Financial stability.

JEL classifications:

C32, C58, G01.

How to Cite:

Oumaima Abouzaid and Faouzi Boussedra. Volatility Spillovers, Systemic Risk, and the Rise of Digital Assets: A Hybrid Approach Using Econometrics, Machine Learning, and Network Analysis. [ref]: vol.24.2026. available at: https://refpress.org/ref-vol24-a1