Networks in Risk Spillovers : A Multivariate GARCH Perspective

We propose a spatial approach for modeling risk spillovers using financial time-varying proximity matrices based on observable networks. We show how these methods could be useful in (i) isolating risk channels, risk spreaders and risk receivers, (ii) investigating the role of portfolio composition in risk transfer, and (iii) computing target exposure structures able to reduce the forecasted system variance and thus the risk of the system. Our empirical analysis builds on banks' foreign exposures provided by the Bank of International Settlements (BIS) as a proxy for Euro area cross-country holdings. We find, in the European sovereign bond markets, that Germany, Italy and, to a lesser extent, Greece are playing a central role in spreading risk, and Ireland and Spain are the most susceptible receivers of spillover effects that can be traced back to a physical claim channel: banks' foreign exposures. We additionally show that acting on these physical channels before the sovereign crisis, it would have been possible to have a clear risk mitigation outcome