Abstract: The paper studies the dynamic conditional correlations (DCC Multivariate GARCH models) of risks for 31 major financial markets, using Expected Shortfall as proxies for these markets' risk. We selected several specifications of the Generalized Orthogonal GARCH model and of the Copula Asymmetric Generalized DCC model. Most of the GARCH-Copula models outperform standard DCC-GARCH and GO-GARCH models. We further study the nature of the processes driving these correlation series, finding the correlations non-stationary (but not ‘explosive’) and exhibiting multifractal properties. Moreover, some DCCs may act as triggers (at least in a ‘nonlinear’ Granger sense) for others. Finally, we show evidence of cross-market risk spread during the 2007-2010 turmoil, pandemic and Ukrainian war crises.

Keywords: GO-GARCH DCC; Copula-GARCH DCC; multivariate affine Normal-Inverse Gaussian distribution; Expected Shortfall; financial crises

JEL codes: C10, C13, G15, G32

DOI: ...