The role of the SGT Density with Conditional Volatility, Skewness and Kurtosis in the Estimation of VaR: A Case of the Stock Exchange of Thailand

Abstract

One of primary tools used to assess the financial risk is Value-at-Risk (VaR). It turns to be a standard measure of downward risk among financial intermediaries and regulators recently as it summarized the risk into just a single and easy-to-understand number. Despite the simplicity of VaR's concept, an accurate calculation of VaR is still challenging. This paper aims to propose an alternative approach which is believed to provide more accurate VaR rather than the traditional ones. Instead of the conventional Gaussian distribution, the more flexible skewed generalized t (SGT) density function is assumed for return series. Its volatility is characterized by eight types of GARCH process. Meanwhile, conditional skewness and kurtosis is modeled to exhibit time-varying feature by their past information set and autoregressive term. Daily returns on the SET index will be used to explore the performance of estimated VaR. The finding shows that this new approach can provide more accurate and robust estimates of the actual VaR threshold, especially with TS-GARCH model, than any other approaches that have been applied earlier.