A dynamic network model to measure exposure concentration in the Austrian interbank market

Abstract

Motivated by an original financial network dataset, we develop a statistical methodology to study non-negatively weighted temporal networks. We focus on the characterization of how nodes (i.e. financial institutions) concentrate or diversify the weights of their connections (i.e. exposures) among neighbors. The approach takes into account temporal trends and nodes’ random effects. We consider a family of nested models on which we define and validate a model-selection procedure that can identify those models that are relevant for the analysis. We apply the methodology to an original dataset describing the mutual claims and exposures of Austrian financial institutions between 2008 and 2011. This period allows us to study the results in the context of the financial crisis in 2008 as well as the European sovereign debt crisis in 2011. Our results highlight that the network is very heterogeneous with regard to how nodes send, and in particular receive edges. Also, our results show that this heterogeneity does not follow a significant temporal trend, and so it remains approximately stable over the time span considered.