Modeling dependence across stock markets using copulas.

Abstract

An important issue in multivariate statistical modeling is the choice of the appropriate dependence measure. Correlation has many pitfalls as it is associated with the elliptical distributions assumption of normality which fails in the presence of extreme endpoints either in marginals or in higher dimensions. Copulas offer an alternative measure of dependence which overcomes the limitations of correlation, and they also determine the type of dependence whether it is linear, upper tail or lower tail. This research serves to explore the appropriateness of copulas in modeling bivariate dependence amongst five SADC stock markets with an objective of assessing the effectiveness of regional integration. Archimedean copulas, due to their desirable properties, were examined using both parametric and non-parametric techniques.

Non-parametric estimation gave profound results signifying the appropriateness of the Gumbel copula in dependence modeling which indicated that investors had chances of portfolio diversification across the region as the markets were prone to booming together.