The leverage game.

Abstract

This study uses individual bank’s balance sheet data from Argentina, Brazil, Mexico, South Africa, and Taiwan, to obtain cross-correlations over the time-series leverage ratio. A correlation-based filtering procedure is then implemented to obtain leveragebased networks. Results conclude that topological properties from graph theory describe a surprising similar pattern inside the network between countries. These leverage networks are complex and are characterized by heavy-tailed vertex degree distribution, which reveals the emergence of hubs and suggests the potential scale-free nature of the system. Finally, surrogated Monte Carlo simulations confirm that the complex networks, obtained through the filtering procedure, cannot be explained by chance alone.