Globalization of equity markets and its determinants

Abstract

We are witnessing a phase of globalization where markets across the globe are getting inter linked and equity markets are no exception. Traditionally, difference in equity markets across the globe was seen as a primary technique for geographic diversification. But this is now fast becoming a thing of past. With markets getting coupled at the pace they are, international diversification is becoming increasingly difficult for the investors. This paper aims to answer two questions Has the correlation across world equity markets undergone a structural change and what are the major determinants of this correlation. The first 2 modules aim to answer the first question while Module 3 answers the second question. In both the cases our hypothesis was that correlations across markets have undergone a change after 2000.Results from Module 1 show that 2000 is indeed an inflection point and the pattern is most distinctive in the case of S and P-Nikkei and S and P-Sensex pairs. This prompted us to do our further analysis on these 2 pairs. The second module focused on establishing this statistically. We have shown that when the returns of the indices are regressed against each other, the regressions undergo a structural change after 2000. Once we have established this, we move on to Module 3 where we attempt to explain the determinants of correlation. We segment the determinants into three categories Global, country specific and contagion factors. The analysis is done for all the three pairs. While some results are as expected, we also get a number of results that are counter-intuitive. We then attempt to interpret each result and provide a plausible explanation for them. The results show that GDP differential and rate differential are key determinants for explaining the correlation of S and P-Sensex pair while for S and P-Nikkei pair, the significant determinants are GDP differential, Inflation differential and returns on world equity Index. Apart from that we run separate regressions for pre and post 2000 data and show that model is not statistically significant in case of pre 200 data points. This further reinforces our findings in Module 1and 2. The paper begins by briefly defining the problem and then we move on to the modules.