Investigation of dependence among global stock exchanges via copula

This thesis examines the dependence among international stock exchanges by using copulas that is relatively new but rapidly evolving approach, which allows us to examine both the structure of dependence and the degree of dependence. The importance of this approach is modeling and capturing dependencies among random variables, regardless of the form of the marginal distribution. Correlation measures reflect the degree of dependence but not the structure and shape of dependence. Another advantage of the copulas is that conventional multivariate distributions are used for random variables with the same marginal distribution. However, since many variables may have different marginal distributions, this causes a problem for applications. This problem can be solved through the use of copulas. Copula models allow for multivariate distributions based on marjinal distributions that have not same distribution family and is heterogeneous. Thus, the researcher get rid of taking into account only the present multivariate distributions. Nowadays, copulas are a basic tool for constructing multivariate distributions. In general, financial time series do not fit the normal distribution, which is known to have mostly leptokurtic distribution and fat tail. In this case, the assumption of normality is not provided for most parametric tests and it is not appropriate to use linear correlation for dependence. Therefore, we used the copulas, which are not affected by the forms of the marginal distribution and allow us to measure non-linear dependence. The aim of this study is to investigate the dependence among the global stock exchange indices such as FTSE-100, S&P-500, Nasdaq, and Nikkei-225, which have the largest trading volume in the world, and to model the dependence among the indices with the appropriate copula, using the advantages of copulas