Detection and quantification of causal dependencies in multivariate time series: a novel information theoretic approach to understanding systemic risk.

The recent financial crisis has lead to a need for regulators and policy makers to understand and track systemic linkages. We provide a new approach to understanding systemic risk tomography in finance and insurance sectors. The analysis is achieved by using a recently proposed method on quantifying causal coupling strength, which identifies the existence of causal dependencies between two components of a multivariate time series and assesses the strength of their association by defining a meaningful coupling strength using the momentary information transfer (MIT). The measure of association is general, causal and lag-specific, reflecting a well interpretable notion of coupling strength and is practically computable. A comprehensive analysis of the feasibility of this approach is provided via simulated data and then applied to the monthly returns of hedge funds, banks, broker/dealers, and insurance companies.