Evaluating Stock Index Return Value-at-Risk Estimates in South Africa

Abstract

This article evaluates the performance of a range of alternative volatility models in forecasting volatility and value-at-risk (VaR) in the context of the Basle regulatory framework, using stock index return data from South Africa. We extend the current research in emerging markets by considering a wider selection of GARCH-based models, including a variety of asymmetric and long memory models. Our results suggest that models incorporating both asymmetric and long memory attributes generally outperform all other methods in estimating VaR across the three percentiles we considered. These findings are similar to the volatility forecasting exercise we also conduct. More generally, we find that the standard RiskMetrics model is consistently outperformed by all the GARCH-type models we have analysed in the context of VaR modelling. Finally, our results emphasise the importance of using the stringent probability criteria prescribed by the Basle regulatory framework, and of employing out-of-sample forecast evaluation techniques for the selection of forecasting models that provide accurate VaR estimates.