A0699
Title: Spatial multivariate GARCH models and financial spillovers
Authors: Rosella Giacometti - University of Bergamo (Italy) [presenting]
Gabriele Torri - University of Bergamo (Italy)
Kamonkay Rujirarangsan - University of Bergamo (Italy)
Michela Cameletti - Universita degli Studi di Bergamo (Italy)
Abstract: The risk spillover is estimated among European banks from equity log return data via conditional value at risk (CoVaR). The joint dynamic of returns is modelled with a spatial DCC-GARCH, which allows the conditional variance of log returns of each bank to depend on past volatility shocks to other banks and their past squared returns in a parsimonious way. The backtesting of the resulting risk measures provides evidence that (i) the multivariate GARCH model with Students t distribution is more accurate than both the standard multivariate Gaussian model and the filtered historical simulation (FHS), (ii) the introduction of a spatial component improves the assessment of risk profiles and the market risk spillovers.
