Networks of spillover effects to spot systemic risk in the banking industry: testing Granger causality in a high dimensional VAR model

Sammanfattning: In this essay the evolution of global systemic risk is investigated by creating networks (and communities) of contagion among 88 banks that are deemed systemically important on a global level (particularly western economy). The contagion is defined by a Granger Causality relation between two individuals in the network, this is tested using the Post double selection method described in Margartitella et al. 2021. The method allows to get rid of the High Dimensionality issue arising in testing a VAR(p) model and estimate models over small numbers of time observations. It is found that the method is useful in shaping the evolution of systemic risk in the relevant years and could be used in further application for example creating an index of global systemic risk.