A network-based method for visual identification of systemic risks

ABSTRACT Financial markets provide vast numbers of signals about the performance of companies,;banks, assets and economies. These signals can be used by risk managers;and regulators to better understand economic dependencies, correlations and phase;transitions. In this paper, we present a methodology for mapping multiple dimensions;of time series data into two-dimensional visual layouts by applying methods;from statistics and network theory. The methodology involves identifying important;correlations between the time series as well as monitoring individual series to determine;which ones have extreme return values compared with their past performance.;Analysis is presented visually to give quick insight into a complex system moving;in time; for example, systemically important assets are easily recognizable as those;that are central in the minimum spanning tree structure of the correlation matrix, and;systemic events are visible as large numbers of assets having extreme values. We;present historical scenarios to illustrate the methodology.