Analyzing Crisis in Global Financial Indices

Abstract

We apply the Random Matrix Theory and complex network techniques to 20 global financial indices and study the correlation and network properties before and during the financial crisis of 2008 respectively. We find that the largest eigenvalue deviate significantly from the upper bound which shows a strong correlation between financial indices. By using a sliding window of 25 days we find that largest eigenvalue represent the collective information about the correlation between global financial indices and its trend indicate the market conditions. It is confirmed that eigenvectors corresponding to second largest eigenvalue gives useful information about the sector formation in the global financial indices. We find that these clusters are formed on the basis of the geographical location. The correlation network is constructed using threshold method for different values of threshold θ in the range 0 to 0.9, at θ=0.2 the network is fully connected. At θ=0.6, the Americas, Europe and Asia/Pacific form different clusters before the crisis but during the crisis Americas and Europe are strongly linked. If we further increase the threshold to 0.9 we find that European countries France, Germany and UK consistently constitute the most tightly linked markets before and during the crisis. We find that the structure of Minimum Spanning Tree before the crisis is more star like whereas during the crisis it changes to be more chain like. Using the multifractal analysis, we find that Hurst exponents of financial indices increases during the period of crisis as compared to the period before the crisis. The empirical results verify the validity of measures, and this has led to a better understanding of complex financial markets.