Financial correlation networks

Abstract

We construct a financial network based on the correlations of the assets. Referred to as a correlation network, its nodes are the assets and its edges are the pair-wise correlations. The network is inspected by both spectral and statistical analyses. We find that these analyses provide complementary information regarding the interactions of securities in the market and their clustering as well as a hierarchy in the market structure. Market portfolio dominating behavior indicates scale free type of interactions where a small number of assets linked to many others accounts for most of the activities. We further introduce a pricing model that uses the interrelations of emerging dominant correlational motifs. Those that are in accord with piecewise stationary behavior are found to successfully define the future price bounds.