Network Approach to Capture Comovements of Global Stock Returns

Abstract

This research aims at developing a network-based index to investigate the change in the level and evolutionary pattern of comovement in the global stock markets. A correlation-based stock market network has been constructed for major stock markets in North America, Europe, and Asia. The behavior of key metrics of the networks such as edge density, clustering coefficient, power-law exponent, length of the Minimum Spanning Tree (MST), and the largest eigenvalues of the correlation matrices have been analyzed. We propose network-based comovement indices derived from standard metrics in the literature for regional and global stock markets to capture the level of comovements of stock returns in the market. Further we analyze these comovement indices using non-parametric statistical methods to test for any significant variation in these indices during the crisis phase compared to those during a phase of stability. The collapse of Lehman Brothers in the United States in September 2008 that triggered a global stock market crisis has been used as an example to demonstrate the potential use of the proposed indices in capturing the dynamic behavior of the stock markets. The study reveals the presence of regional influences on the network dynamics. It also unearthed the emergence of a global pattern in the dynamics of comovement indices around the onset of the crisis confirming interdependence of the global stock markets. The findings provide empirical evidences of statistically significant drift in the mean of the comovement index and presence of non-randomness in the evolutionary behavior of the comovement indexes of the stock market networks about the period of crisis. The findings of this research would be useful in identifying early warning signals of a financial crisis.