MODELLING DEPENDENCE IN STI COMPONENT STOCKS: AN ANALYSIS WITH R-VINE COPULA MODELS

Abstract

In finance, modelling high-dimensional datasets is often made difficult by the underlying complex dependence. In particular, in view of the 2007-08 global financial crisis and the increasing volatility at financial markets globally, it is therefore critical to study these dependence structures appropriately. While standard multivariate copulas may be more restrictive to explain for the underlying complicated dependencies, vine copulas may provide a convenient alternative as it allows flexible modelling of high- dimensional dependencies through a rich class of bivariate copula families. In this regard, this thesis uses the regular vine copula class to model the dependence structure of a 28-dimensional STI component dataset. We first apply a heuristic method to specify the R-vine, then we evaluate its goodness-of-fit using a selection of likelihood-based tests and Value-at-risk backtesting. Through the test results, we conclude that using a regular vine is indeed appropriate to model high-dimensional dependencies.