On Shock Propagation in Financial Networks

Sammanfattning: This thesis develops a simplified financial network model for an interbank lending system which is then analyzed in terms of contagion when exposed to external liquidity shocks. The aim is to understand how individual institutions and the network structure affect the shock propagation and finding factors that increase respectively decrease the systemic risk of the network. The network structures analyzed are mainly the ring graph, the complete graph, and the directed tree graph, given an ex-post and an ex-ante perspective. The first result indicates that traditional centrality measures are not capable of identifying systemically important institutions. The second result concerns the interconnections in the network structure, where it is concluded that if one institution or all institutions are subject to a certain shock, a complete structure always performs better than or equally as well as the denser structure of a ring graph, in terms of number of defaulting institutions, whereas if multiple institutions, but less than all of them, are exposed, the complete graph may perform worse. The last result shows that in acyclic tree graphs, a higher number of offspring in the k-regular tree graph and an offspring distribution with less variance in the random tree graph, can restrict the contagion respectively reduce the probability of shock propagation further down the tree.