Study Of The Stock Market Cross-correlation With Cross Sample Entropy

In this thesis we employed the Cross-sample entropy method to investigate the development of cross correlations among international equity returns at the market and industry level over the period2004-2011. The Global Financial Crisis of2007/2009is being used as a case study. We explore daily data from eight market indices from China, Hong Kong, Germany, Japan, Taiwan, the UK, and the USA for a period from2004to2011. We also computed cross correlations between each two of the eight selected stocks from the eight sectors (Basic Materials, Financials, Health Care, Technology, Consumer Goods, Industrial Goods, Services, and Conglomerates) of the Dow Jones Industrial Average. The calculation method of confidence interval of SampEn (sample entropy) was extended and applied to cross-SampEn. The cross-SampEn and its confidence interval for every two of stock markets time series in periods2004-2007(before the global financial crisis) and2008-2011(after the global financial crisis) are calculated. The same calculation method was applied to the selected stocks indices.The results show that the cross-SampEn of the market indices became lower after the global financial crisis indicating a lower asynchrony (meaning got correlated after the crisis). The results also showed that every two stocks indices returns series based on the stocks arrangement in (i.e. stocks selected based on closest relationships) become higher after the global financial crisis, indicating a higher asynchrony(meaning less correlated after the crisis) while on the other hand, for the stocks indices based on least close relationships become lower the global financial crisis, indicating a higher asynchrony(meaning got correlated after the crisis). Comparing the correlation coefficient and the cross-SampEn, we further found that the cross-SampEn is superior in describing the correlation between time series and thereby providing an alternative and promising method for the demonstration of correlations among time series. The results further showed that there are still weakly correlated markets and the influence of the Crisis differs from country to country.