Multiscale Network Analysis for Financial Contagion

Contagion in financial markets has been one the most active areas of research, especially during the last decade and due to the major incidents during the Global Financial Crisis and the European Financial Crisis. However, two of the most important questions that remain after a financial crisis are what are the determinants of the crisis and how can we forecast an incident based on suitable indicators. The purpose of this study is twofold. First, to develop a measure of contagion based on the multiscale nature of the financial contagion. Second, to examine how financial contagion is spread in the US economy in different frequencies based on the proposed measure. We assert that important information on an upcoming crisis, not observed in the original data, may be revealed by performing a time-frequency analysis of the time-series and the cross-section of stock returns. We use wavelet analysis to decompose the returns and network analysis to compute various network characteristics related to contagion. Our proposed methodology allow us to: understand the short-, mid- and long-term connections of the network, bring out structures/relations that are not visible initially and mask the true connections between companies, study how the networks measures change over scale, and finally, examine the distribution of contagion at different time-horizons and scales.